Applied Evidence

Had Sir Isaac Newton attempted family medicine, he likely would have been uncomfortable with its nonlinear aspect typified by unpredictable disease courses and treatment responses.

Linearity forms the basis of our knowledge… Life in a Newtonian world is ordered and predictable, where causes are directly linked to effects and behavior is linear or cyclic (periodic). In this world, stability and predictability define a healthy system. Furthermore, by understanding the parts of a system, we understand the system. As physicians, we are trained to expect this linear, predictable, reductionistic view of health.

…but it does not reflect the human system. However, humans are complex adaptive systems, characterized by multiple interconnected and interdependent parts at levels from the microscopic to the community. Interactions change over time, producing synergistic nonlinear behavior as components periodically self-organize into functional groups.

TABLE 1 compares the Newtonian world view with that of complexity science. Although all of the characteristics of complexity science are relevant to family physicians, this article will focus on the nonlinear behavior of patients as the visible, unpredictable, and often frustrating manifestation of the complexity characteristics. TABLE 2 defines specific characteristics of nonlinearity.

In understanding nonlinearity—as depicted in 4 patient cases presented here—family physicians can learn to

• expect the unexpected
• reduce unpredictability by learning about patients and their context
• attack patient resistance by seeking epiphanies or using positive attractors
• recognize the sensitivity of our patients’ trajectories and use or anticipate it
• promote the healthy benefits of nonlinearity.

TABLE 1
Basic tenets of Newtonian and complexity world views

TABLE 2
Nonlinear characteristics relevant to family practice

Nonlinearity as a truer model of health

Although our basic medical knowledge is built on a reductionistic approach that assumes linear dynamics, our models rarely account for more than 30% of whatever outcome we are investigating. Clinical providers are often faced with the unexpected.

Although linearity suggests that illness should respond in predictable ways regardless of the environment, family physicians know that context is critical. In addition, the human condition is often nonlinear; nonlinear dynamics (chaotic or random dynamics) have been documented in physiology,¹ psychology,^2,3 sociology,⁴ business,^5-7 and economics.⁸

In fact, nonlinear dynamics are often a sign of health. For example, mood may vary in linear patterns among patients with affective disorders; therapy for mood disorders may work by changing the pathological linear dynamics in mood into more healthy nonlinear dynamics.⁹ Linear (or periodic) dynamics often indicate a pathological condition.^10,11

As science and medicine begin to embrace the nonlinearity of complexity science, we must anticipate, recognize, and apply nonlinearity to the care of our patients. This is particularly important for family physicians.

Applying nonlinearity to patient cases

The following cases demonstrate characteristics of nonlinear dynamics (TABLE 2).

Case 1: Sensitivity to initial conditions

I.C. is a 25-year-old teacher who is 6 weeks postpartum. Recently, while at a local shopping mall, she experienced a sudden onset of chest discomfort, palpitations, dizziness, trembling, and a sense of impending doom. The episode peaked in intensity within 3 minutes and lasted 20 minutes after leaving the mall. Although she has not experienced another attack, she has progressively limited her activities since, until now, she has not been able to bring herself to re-enter the mall for fear of another attack. In fact, she reports intense anxiety in anticipation of possibly visiting the mall and has begun limiting her driving in general.

Agoraphobia is linked to the location and interpretation of the first panic attack.¹² This demonstrates the concept of sensitivity to initial conditions whereby small differences in starting values result in very different behaviors later. In other words, apparently minor differences in a patient’s initial physical and emotional state can translate into drastically different outcomes over time.

This emphasizes the need for physicians to pay attention to detail during stressful events that patients experience. For example, if a patient experiences the first panic attack in a self-perceived “safe” environment or interprets the attack as a normal response, she may avoid the disabling consequence of agoraphobia and remain functional. There are other examples of this sensitivity to small changes, such as siblings of similar genetic make-up and environment who exhibit markedly different health as adults may do so because of “minor” life events each experienced.

Similarly, patients with chronic stable disease who, after a minor event, suddenly change their disease trajectory may be demonstrating sensitivity to initial conditions. This sensitivity has been proposed as an explanation for sudden infant death syndrome¹³ (SIDS) and “brittle” diabetes.¹⁴ Treatment response may depend on sensitivity to initial conditions; cases documenting placebo effects or unpredictable potassium excretion on re-administration of potassium-sparing diuretics are examples.¹⁵

Implications for management. Sensitivity to initial conditions has several implications for patient management.

First, we need to recognize the impact it has on patients. Minor life changes can alter the trajectories of patients, so we need to seek the patient’s perspective on stressors they experience. This inherent instability means that “watchful waiting” is a viable approach in some patients because illness may resolve without intervention. It also means that we need to be watchful for signs of an unhealthy trajectory developing in patients even after minor stressors.

Second, sensitivity to initial conditions implies that nonlinear behaviors can change with minor but well-timed interventions. The importance of chronotherapy (timed dosing based on biological rhythms) is receiving increasing attention. Drug efficacy often varies with the time of day.^16,17 Though focused on matching circadian rhythms, chronotherapy may be valuable in nonlinear systems if, through our in-depth knowledge of the patient and context, we can identify a point of leverage when an otherwise “ineffective” treatment may be effective for the patient at a specific point in time. There may be justification for re-administering a previously ineffective treatment if you believe the responsiveness of the patient may have changed. Sensitivity to initial conditions may also explain the effectiveness of placebos.

• For the patient above, minimizing the impact of sensitivity to initial conditions requires immediate access to the patient during a subsequent sensitive time. If the patient has a family history of panic disorder or has had panic attacks in the past, you could anticipate that the patient may experience a panic attack in the future and prepare her for it by discussing the chemical basis for panic (and its lack of serious physical consequences) and by encouraging her to contact you day-or-night immediately after experiencing one so that you can help her to identify a nonthreatening (even if illogical) cause for it, thus preventing the fears that lead to agoraphobia.
• For the practitioner, sensitivity to initial conditions emphasizes the need to understand the details surrounding a stressor by asking patients about their perceptions of the events, the circumstances, and how they are being affected by the stressor. Using this sensitivity for treatment implies focusing on the timing of interventions, and considering re-administration of treatments or even the use of placebos.

Case 2: Effects of attractors

A.T. is a 47-year-old factory worker with a 30-year history of alcohol consumption. His daily intake consisted of a case of beer until he quit 3 years ago. He has periodically suffered relapses consisting of 3 or 4 days of binge drinking followed by prolonged abstinence. Although his wife and 2 children are supportive of his efforts at abstinence, his son dramatically increased his alcohol consumption when his father stopped his daily consumption. In addition, his teenage daughter began experimenting with drugs 2 years ago.

Alcoholism serves as an attractor, controlling not only the patient’s behavior, but the behavior of the family.¹⁸ Attractors limit the range of possible behaviors and thereby resist or limit changes in a patient’s course. Attractors may be internalized models or belief systems that lead to recurring patterns of behavior, even though sensitivity to initial conditions prevents one from predicting the specific path the system will follow; patterns are predictable, the path followed is not. The combination of attractors and sensitivity to initial conditions ensures nonlinearity.

Alcoholism is not the only example of a factor that molds behavior and resists change. Our lives are governed by repeating patterns of behavior. Lifestyle routines are deeply ingrained and resist change even for medically important reasons. These lifestyle patterns may be due to attractors and may explain the resistance to change that many of our diabetic patients exhibit. Similarly, dysfunctional families often display counterproductive patterns of behavior that are resistant to even the best counseling.

Implications for management. The presence of attractors suggests several implications for patient management.

First, we should anticipate resistance and not be frustrated when it occurs.

Second, we can attack the attractor itself,¹⁹ by identifying another, more positive attractor in the patient’s life and reinforcing it to diminish the negative attractor’s impact. For example, instead of simply criticizing the inactive lifestyle ingrained in a hypercholesterolemic patient, we reinforce the positive attractor of the patient’s affection for his grandchildren and use that attractor to get the patient to exercise.

For strongly negative attractors (eg, alcohol use), we could simply attack the attractor itself without providing an alternate attractor. Though this approach is more risky because of the unpredictability of what the patient will substitute, if the attractor is bad enough, we may be willing to allow the patient to choose any other attractor, assuming that it must be more positive than the original.

• For the patient above, simply attacking the negative attractor may lead to other negative behaviors (eg, smoking). The best approach may be to focus on positive attractors (ie, wife, children, social relationships, hobbies), perhaps even positive attractors for the entire family, to move him or them away from the negative attractor.
• For the practitioner, it is imperative to identify the attractors producing recurrent detrimental behaviors that need to be changed. Using attractors for management means that potential positive attractors need to be identified through exploration with patients and reinforced while attacking the negative attractors currently producing the unhealthy behavior.

Case 3: Bifurcation effects

B.I. is a 50-year-old plumber who has had type 2 diabetes for more than 10 years. Though he has regularly seen his physician and taken his medications, his diabetes control has been poor (hemoglobin A_1c=10.2). He admits that compliance with his diet and exercise has been “spotty” at best. Six months ago, his older brother began dialysis for end-stage renal disease secondary to diabetes. Within 1 week of his brother’s first dialysis session, B.I. began walking 30 minutes each night and eliminated evening snacks. Consequently, he has lost 22 pounds, and his hemoglobin A_1c has dropped to 8.1.

Sudden dramatic changes (bifurcations) can occur in nonlinear systems as the system reaches a “tipping point.” In this case, chronic noncompliance suddenly changed to compliance after a meaningful event.²⁰ These bifurcations represent a qualitative change in behavior linked to a change in an attractor. Hence, epiphanies may represent behavioral bifurcations.²¹ Such epiphanies are important in premature menopause²² and initial family decisions to hospitalize a mentally ill relative.²³

Bifurcations have been best documented in cardiovascular disease. Pulsus paradoxicus, pulsus alternans in congestive heart failure (CHF), and cardiac movement in tamponade reflect bifurcations in the system as minor changes cause the system to cross a “tipping point” and produce sudden drastic effects.¹⁰ Similarly, bifurcations in heart rhythm are seen in sick sinus syndrome and ST-T alternans in ventricular tachycardia.²⁴ Paradoxical behavior of the PR interval²⁵ and the disastrous effect of the R-on-T phenomenon are other examples. However, bifurcation dynamics are also important in psychosocial behavior. Sudden drastic changes in mood have been documented in patients with generalized anxiety disorder.²⁶

Implications for management.

• For the patient above, we can first look for events that could serve as epiphanies (eg, development of lung cancer in a relative of our tobacco-dependent patient) and use them to alter behavior. Many physicians already look for consequences of diabetes in friends and relatives to motivate their patients.
• For the practitioner, the existence of bifurcations implies that sudden unforeseen behavior should be expected and should not be a source of frustration. From a management perspective, drastic changes in patient behavior can be achieved by exploring patients’ lives and recognizing and reinforcing epiphanies.

Case 4: Self-organized behavior

S.O. is a 40-year-old housewife with a long history of intermittent anxiety, usually in response to a family stressor. She presents with extreme apprehension and insomnia. On examination, she is restless and mildly tachycardic. Upon further questioning, she denies any recent adverse events but, in fact, reports that her husband recently received a promotion including a significant increase in salary. In reviewing her chart, you notice that you have diagnosed her with adjustment disorder with anxious mood on 3 previous occasions after adverse family stressors. However, her reactions have often been out-of-proportion to the level of stress and she has occasionally reported significant stressors (eg, death of a sister) without subsequent anxiety.

Neurologic systems tend to organize themselves in response to external events and internal models. These self-organized systems consist of tenuously linked parts at the edge of stability balanced between periodic and chaotic behavior. They react to stressors in patterned ways, but the magnitude of the reaction can vary from little or no response to a catastrophic reaction. Because such self-organization can be temporary, with groups periodically forming and dissolving, behavior over time is random without recurrent patterns.

With this patient, varying degrees of stress (even positive events) result in varying degrees of dysfunction with little relationship between the magnitude of stress and the magnitude of dysfunction. The periodic collapse in response to cumulative stress is not the only example of self-organized behavior.

Self-organization is believed to be critical in a variety of neuropsychiatric conditions from personality disorders² and conversion reactions to adult consequences of childhood adversity.²⁷ Patterns of detoxification in groups of alcoholics demonstrate self-organized behavior.²⁸ Self-organization is important to understanding self-regulation and behavior in families.^29,30 Even social interaction patterns among groups of patients on psychiatric wards show self-organized behavior as unstable groups form, dissolve, and reform.³¹

Implications for management. If non-linearity indeed reflects health and helps to keep patients in good health, we should be promoting nonlinear behavior. Studies have shown that frequent small interventions can keep a system that is prone to periodic behavior in nonlinearity.^32,33 Similarly, because nonlinear systems can display a spectrum of behaviors from periodic-to-self-organized behavior-to-chaotic dynamics depending upon their resources and interconnectedness, social systems exhibiting periodic behavior may move into nonlinearity in response to increased resources and decreased restraints,⁵ or to increased interconnectedness.³⁴

Perhaps we can train systems to maximize their variability. For example, exercise programs that used variable intensities and durations may promote a cardiovascular system capable of responding to whatever stressor comes along.

• For the patient above, the self-organized behavior is detrimental, producing over-reaction to stressors; a more chaotic mood pattern would minimize the impact of stressors. The best approach to achieve this may be to increase resources and decrease restraints. Thus, providing the patient with several ways of dealing with stress (ie, multiple treatment modalities including relaxation techniques, self-hypnosis, meditation, PRN anxiolytics) while promoting connectedness with others (ie, support groups, internet, church contacts, meditation) may increase chaotic variability.
• For the practitioner, self-organized behavior may explain the apparent random response to stress in patients. Such unstable behavior can be managed by providing multiple interventions simultaneously (ie, behavioral, pharmacological, social) or temporally (eg, frequent reinforcements of desired behavior) to encourage healthy nonlinearity.

Nonlinearity of primary versus specialty care

Do patients in primary care exhibit a different degree of nonlinearity than those seen in specialty care settings? Generally, yes. Mental illness, for instance, tends to be more severe among psychiatric patients than among primary care patients,^35-37 and CHF is more severe among cardiology patients.³⁸

Differences in severity of illness are important because, in some cases, the more severe the illness, the more periodic the dynamics.^9,39,40 Thus, the nonlinearity decreases as the severity increases. Because diseases exhibiting periodic dynamics should have a more predictable response to therapy, we would expect more severe illnesses to respond more predictably.⁴¹ This pattern has indeed been observed. Prognosis and predictability of treatment response is related to severity of illness in CHF, acute myocardial infarction, depression, and agoraphobia.^38,42-47

Thus, for both biomedical and psychosocial problems, predictability of treatment response correlates with the severity of illness. If patients seen in specialty settings have more severe disease, then we should expect that primary care patients exhibit more nonlinear behavior and are thus less predictable in their response.

Learning to see differently

Though trained to approach medical problems looking through “linear lenses,” we see nonlinear behavior all the time in our patients. If nonlinear processes represent health, then when systems are using healthy, nonlinear dynamics, they are resistant to disruptive external stressors. However, when such systems transition into periodicity due to illness, they may become predictable and more amenable to intervention, permitting physicians to treat them and hopefully restore the healthy, nonlinear dynamics.

Sensitivity to minor changes in their environment, resistance to change, sudden dramatic change in behavior, and intermittent collapses characterize behaviors in many patients. If we understand the nonlinear nature of these behaviors, we will be better able to help our patients.

Expect the unexpected, reduce unpredictability by learning about patients and their contexts, attack resistance by seeking epiphanies or using positive attractors, recognize the sensitivity of our patients’ trajectories and use or anticipate it when possible, and promote the healthy benefits of nonlinearity.

CORRESPONDENCE
David A. Katerndahl, MD, MA, Department of Family and Community Medicine, University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Dr. MC 7795, San Antonio, TX 78229-3900. E-mail: [email protected].

Had Sir Isaac Newton attempted family medicine, he likely would have been uncomfortable with its nonlinear aspect typified by unpredictable disease courses and treatment responses.

Linearity forms the basis of our knowledge… Life in a Newtonian world is ordered and predictable, where causes are directly linked to effects and behavior is linear or cyclic (periodic). In this world, stability and predictability define a healthy system. Furthermore, by understanding the parts of a system, we understand the system. As physicians, we are trained to expect this linear, predictable, reductionistic view of health.

…but it does not reflect the human system. However, humans are complex adaptive systems, characterized by multiple interconnected and interdependent parts at levels from the microscopic to the community. Interactions change over time, producing synergistic nonlinear behavior as components periodically self-organize into functional groups.

TABLE 1 compares the Newtonian world view with that of complexity science. Although all of the characteristics of complexity science are relevant to family physicians, this article will focus on the nonlinear behavior of patients as the visible, unpredictable, and often frustrating manifestation of the complexity characteristics. TABLE 2 defines specific characteristics of nonlinearity.

In understanding nonlinearity—as depicted in 4 patient cases presented here—family physicians can learn to

• expect the unexpected
• reduce unpredictability by learning about patients and their context
• attack patient resistance by seeking epiphanies or using positive attractors
• recognize the sensitivity of our patients’ trajectories and use or anticipate it
• promote the healthy benefits of nonlinearity.

TABLE 1
Basic tenets of Newtonian and complexity world views

TABLE 2
Nonlinear characteristics relevant to family practice

Nonlinearity as a truer model of health

Although our basic medical knowledge is built on a reductionistic approach that assumes linear dynamics, our models rarely account for more than 30% of whatever outcome we are investigating. Clinical providers are often faced with the unexpected.

Although linearity suggests that illness should respond in predictable ways regardless of the environment, family physicians know that context is critical. In addition, the human condition is often nonlinear; nonlinear dynamics (chaotic or random dynamics) have been documented in physiology,¹ psychology,^2,3 sociology,⁴ business,^5-7 and economics.⁸

In fact, nonlinear dynamics are often a sign of health. For example, mood may vary in linear patterns among patients with affective disorders; therapy for mood disorders may work by changing the pathological linear dynamics in mood into more healthy nonlinear dynamics.⁹ Linear (or periodic) dynamics often indicate a pathological condition.^10,11

As science and medicine begin to embrace the nonlinearity of complexity science, we must anticipate, recognize, and apply nonlinearity to the care of our patients. This is particularly important for family physicians.

Applying nonlinearity to patient cases

The following cases demonstrate characteristics of nonlinear dynamics (TABLE 2).

Case 1: Sensitivity to initial conditions

I.C. is a 25-year-old teacher who is 6 weeks postpartum. Recently, while at a local shopping mall, she experienced a sudden onset of chest discomfort, palpitations, dizziness, trembling, and a sense of impending doom. The episode peaked in intensity within 3 minutes and lasted 20 minutes after leaving the mall. Although she has not experienced another attack, she has progressively limited her activities since, until now, she has not been able to bring herself to re-enter the mall for fear of another attack. In fact, she reports intense anxiety in anticipation of possibly visiting the mall and has begun limiting her driving in general.

Agoraphobia is linked to the location and interpretation of the first panic attack.¹² This demonstrates the concept of sensitivity to initial conditions whereby small differences in starting values result in very different behaviors later. In other words, apparently minor differences in a patient’s initial physical and emotional state can translate into drastically different outcomes over time.

This emphasizes the need for physicians to pay attention to detail during stressful events that patients experience. For example, if a patient experiences the first panic attack in a self-perceived “safe” environment or interprets the attack as a normal response, she may avoid the disabling consequence of agoraphobia and remain functional. There are other examples of this sensitivity to small changes, such as siblings of similar genetic make-up and environment who exhibit markedly different health as adults may do so because of “minor” life events each experienced.

Similarly, patients with chronic stable disease who, after a minor event, suddenly change their disease trajectory may be demonstrating sensitivity to initial conditions. This sensitivity has been proposed as an explanation for sudden infant death syndrome¹³ (SIDS) and “brittle” diabetes.¹⁴ Treatment response may depend on sensitivity to initial conditions; cases documenting placebo effects or unpredictable potassium excretion on re-administration of potassium-sparing diuretics are examples.¹⁵

Implications for management. Sensitivity to initial conditions has several implications for patient management.

First, we need to recognize the impact it has on patients. Minor life changes can alter the trajectories of patients, so we need to seek the patient’s perspective on stressors they experience. This inherent instability means that “watchful waiting” is a viable approach in some patients because illness may resolve without intervention. It also means that we need to be watchful for signs of an unhealthy trajectory developing in patients even after minor stressors.

Second, sensitivity to initial conditions implies that nonlinear behaviors can change with minor but well-timed interventions. The importance of chronotherapy (timed dosing based on biological rhythms) is receiving increasing attention. Drug efficacy often varies with the time of day.^16,17 Though focused on matching circadian rhythms, chronotherapy may be valuable in nonlinear systems if, through our in-depth knowledge of the patient and context, we can identify a point of leverage when an otherwise “ineffective” treatment may be effective for the patient at a specific point in time. There may be justification for re-administering a previously ineffective treatment if you believe the responsiveness of the patient may have changed. Sensitivity to initial conditions may also explain the effectiveness of placebos.

• For the patient above, minimizing the impact of sensitivity to initial conditions requires immediate access to the patient during a subsequent sensitive time. If the patient has a family history of panic disorder or has had panic attacks in the past, you could anticipate that the patient may experience a panic attack in the future and prepare her for it by discussing the chemical basis for panic (and its lack of serious physical consequences) and by encouraging her to contact you day-or-night immediately after experiencing one so that you can help her to identify a nonthreatening (even if illogical) cause for it, thus preventing the fears that lead to agoraphobia.
• For the practitioner, sensitivity to initial conditions emphasizes the need to understand the details surrounding a stressor by asking patients about their perceptions of the events, the circumstances, and how they are being affected by the stressor. Using this sensitivity for treatment implies focusing on the timing of interventions, and considering re-administration of treatments or even the use of placebos.

Case 2: Effects of attractors

A.T. is a 47-year-old factory worker with a 30-year history of alcohol consumption. His daily intake consisted of a case of beer until he quit 3 years ago. He has periodically suffered relapses consisting of 3 or 4 days of binge drinking followed by prolonged abstinence. Although his wife and 2 children are supportive of his efforts at abstinence, his son dramatically increased his alcohol consumption when his father stopped his daily consumption. In addition, his teenage daughter began experimenting with drugs 2 years ago.

Alcoholism serves as an attractor, controlling not only the patient’s behavior, but the behavior of the family.¹⁸ Attractors limit the range of possible behaviors and thereby resist or limit changes in a patient’s course. Attractors may be internalized models or belief systems that lead to recurring patterns of behavior, even though sensitivity to initial conditions prevents one from predicting the specific path the system will follow; patterns are predictable, the path followed is not. The combination of attractors and sensitivity to initial conditions ensures nonlinearity.

Alcoholism is not the only example of a factor that molds behavior and resists change. Our lives are governed by repeating patterns of behavior. Lifestyle routines are deeply ingrained and resist change even for medically important reasons. These lifestyle patterns may be due to attractors and may explain the resistance to change that many of our diabetic patients exhibit. Similarly, dysfunctional families often display counterproductive patterns of behavior that are resistant to even the best counseling.

Implications for management. The presence of attractors suggests several implications for patient management.

First, we should anticipate resistance and not be frustrated when it occurs.

Second, we can attack the attractor itself,¹⁹ by identifying another, more positive attractor in the patient’s life and reinforcing it to diminish the negative attractor’s impact. For example, instead of simply criticizing the inactive lifestyle ingrained in a hypercholesterolemic patient, we reinforce the positive attractor of the patient’s affection for his grandchildren and use that attractor to get the patient to exercise.

For strongly negative attractors (eg, alcohol use), we could simply attack the attractor itself without providing an alternate attractor. Though this approach is more risky because of the unpredictability of what the patient will substitute, if the attractor is bad enough, we may be willing to allow the patient to choose any other attractor, assuming that it must be more positive than the original.

• For the patient above, simply attacking the negative attractor may lead to other negative behaviors (eg, smoking). The best approach may be to focus on positive attractors (ie, wife, children, social relationships, hobbies), perhaps even positive attractors for the entire family, to move him or them away from the negative attractor.
• For the practitioner, it is imperative to identify the attractors producing recurrent detrimental behaviors that need to be changed. Using attractors for management means that potential positive attractors need to be identified through exploration with patients and reinforced while attacking the negative attractors currently producing the unhealthy behavior.

Case 3: Bifurcation effects

B.I. is a 50-year-old plumber who has had type 2 diabetes for more than 10 years. Though he has regularly seen his physician and taken his medications, his diabetes control has been poor (hemoglobin A_1c=10.2). He admits that compliance with his diet and exercise has been “spotty” at best. Six months ago, his older brother began dialysis for end-stage renal disease secondary to diabetes. Within 1 week of his brother’s first dialysis session, B.I. began walking 30 minutes each night and eliminated evening snacks. Consequently, he has lost 22 pounds, and his hemoglobin A_1c has dropped to 8.1.

Sudden dramatic changes (bifurcations) can occur in nonlinear systems as the system reaches a “tipping point.” In this case, chronic noncompliance suddenly changed to compliance after a meaningful event.²⁰ These bifurcations represent a qualitative change in behavior linked to a change in an attractor. Hence, epiphanies may represent behavioral bifurcations.²¹ Such epiphanies are important in premature menopause²² and initial family decisions to hospitalize a mentally ill relative.²³

Bifurcations have been best documented in cardiovascular disease. Pulsus paradoxicus, pulsus alternans in congestive heart failure (CHF), and cardiac movement in tamponade reflect bifurcations in the system as minor changes cause the system to cross a “tipping point” and produce sudden drastic effects.¹⁰ Similarly, bifurcations in heart rhythm are seen in sick sinus syndrome and ST-T alternans in ventricular tachycardia.²⁴ Paradoxical behavior of the PR interval²⁵ and the disastrous effect of the R-on-T phenomenon are other examples. However, bifurcation dynamics are also important in psychosocial behavior. Sudden drastic changes in mood have been documented in patients with generalized anxiety disorder.²⁶

Implications for management.

• For the patient above, we can first look for events that could serve as epiphanies (eg, development of lung cancer in a relative of our tobacco-dependent patient) and use them to alter behavior. Many physicians already look for consequences of diabetes in friends and relatives to motivate their patients.
• For the practitioner, the existence of bifurcations implies that sudden unforeseen behavior should be expected and should not be a source of frustration. From a management perspective, drastic changes in patient behavior can be achieved by exploring patients’ lives and recognizing and reinforcing epiphanies.

Case 4: Self-organized behavior

S.O. is a 40-year-old housewife with a long history of intermittent anxiety, usually in response to a family stressor. She presents with extreme apprehension and insomnia. On examination, she is restless and mildly tachycardic. Upon further questioning, she denies any recent adverse events but, in fact, reports that her husband recently received a promotion including a significant increase in salary. In reviewing her chart, you notice that you have diagnosed her with adjustment disorder with anxious mood on 3 previous occasions after adverse family stressors. However, her reactions have often been out-of-proportion to the level of stress and she has occasionally reported significant stressors (eg, death of a sister) without subsequent anxiety.

Neurologic systems tend to organize themselves in response to external events and internal models. These self-organized systems consist of tenuously linked parts at the edge of stability balanced between periodic and chaotic behavior. They react to stressors in patterned ways, but the magnitude of the reaction can vary from little or no response to a catastrophic reaction. Because such self-organization can be temporary, with groups periodically forming and dissolving, behavior over time is random without recurrent patterns.

With this patient, varying degrees of stress (even positive events) result in varying degrees of dysfunction with little relationship between the magnitude of stress and the magnitude of dysfunction. The periodic collapse in response to cumulative stress is not the only example of self-organized behavior.

Self-organization is believed to be critical in a variety of neuropsychiatric conditions from personality disorders² and conversion reactions to adult consequences of childhood adversity.²⁷ Patterns of detoxification in groups of alcoholics demonstrate self-organized behavior.²⁸ Self-organization is important to understanding self-regulation and behavior in families.^29,30 Even social interaction patterns among groups of patients on psychiatric wards show self-organized behavior as unstable groups form, dissolve, and reform.³¹

Implications for management. If non-linearity indeed reflects health and helps to keep patients in good health, we should be promoting nonlinear behavior. Studies have shown that frequent small interventions can keep a system that is prone to periodic behavior in nonlinearity.^32,33 Similarly, because nonlinear systems can display a spectrum of behaviors from periodic-to-self-organized behavior-to-chaotic dynamics depending upon their resources and interconnectedness, social systems exhibiting periodic behavior may move into nonlinearity in response to increased resources and decreased restraints,⁵ or to increased interconnectedness.³⁴

Perhaps we can train systems to maximize their variability. For example, exercise programs that used variable intensities and durations may promote a cardiovascular system capable of responding to whatever stressor comes along.

• For the patient above, the self-organized behavior is detrimental, producing over-reaction to stressors; a more chaotic mood pattern would minimize the impact of stressors. The best approach to achieve this may be to increase resources and decrease restraints. Thus, providing the patient with several ways of dealing with stress (ie, multiple treatment modalities including relaxation techniques, self-hypnosis, meditation, PRN anxiolytics) while promoting connectedness with others (ie, support groups, internet, church contacts, meditation) may increase chaotic variability.
• For the practitioner, self-organized behavior may explain the apparent random response to stress in patients. Such unstable behavior can be managed by providing multiple interventions simultaneously (ie, behavioral, pharmacological, social) or temporally (eg, frequent reinforcements of desired behavior) to encourage healthy nonlinearity.

Nonlinearity of primary versus specialty care

Do patients in primary care exhibit a different degree of nonlinearity than those seen in specialty care settings? Generally, yes. Mental illness, for instance, tends to be more severe among psychiatric patients than among primary care patients,^35-37 and CHF is more severe among cardiology patients.³⁸

Differences in severity of illness are important because, in some cases, the more severe the illness, the more periodic the dynamics.^9,39,40 Thus, the nonlinearity decreases as the severity increases. Because diseases exhibiting periodic dynamics should have a more predictable response to therapy, we would expect more severe illnesses to respond more predictably.⁴¹ This pattern has indeed been observed. Prognosis and predictability of treatment response is related to severity of illness in CHF, acute myocardial infarction, depression, and agoraphobia.^38,42-47

Thus, for both biomedical and psychosocial problems, predictability of treatment response correlates with the severity of illness. If patients seen in specialty settings have more severe disease, then we should expect that primary care patients exhibit more nonlinear behavior and are thus less predictable in their response.

Learning to see differently

Though trained to approach medical problems looking through “linear lenses,” we see nonlinear behavior all the time in our patients. If nonlinear processes represent health, then when systems are using healthy, nonlinear dynamics, they are resistant to disruptive external stressors. However, when such systems transition into periodicity due to illness, they may become predictable and more amenable to intervention, permitting physicians to treat them and hopefully restore the healthy, nonlinear dynamics.

Sensitivity to minor changes in their environment, resistance to change, sudden dramatic change in behavior, and intermittent collapses characterize behaviors in many patients. If we understand the nonlinear nature of these behaviors, we will be better able to help our patients.

Expect the unexpected, reduce unpredictability by learning about patients and their contexts, attack resistance by seeking epiphanies or using positive attractors, recognize the sensitivity of our patients’ trajectories and use or anticipate it when possible, and promote the healthy benefits of nonlinearity.

CORRESPONDENCE
David A. Katerndahl, MD, MA, Department of Family and Community Medicine, University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Dr. MC 7795, San Antonio, TX 78229-3900. E-mail: [email protected].

Had Sir Isaac Newton attempted family medicine, he likely would have been uncomfortable with its nonlinear aspect typified by unpredictable disease courses and treatment responses.

Linearity forms the basis of our knowledge… Life in a Newtonian world is ordered and predictable, where causes are directly linked to effects and behavior is linear or cyclic (periodic). In this world, stability and predictability define a healthy system. Furthermore, by understanding the parts of a system, we understand the system. As physicians, we are trained to expect this linear, predictable, reductionistic view of health.

…but it does not reflect the human system. However, humans are complex adaptive systems, characterized by multiple interconnected and interdependent parts at levels from the microscopic to the community. Interactions change over time, producing synergistic nonlinear behavior as components periodically self-organize into functional groups.

TABLE 1 compares the Newtonian world view with that of complexity science. Although all of the characteristics of complexity science are relevant to family physicians, this article will focus on the nonlinear behavior of patients as the visible, unpredictable, and often frustrating manifestation of the complexity characteristics. TABLE 2 defines specific characteristics of nonlinearity.

In understanding nonlinearity—as depicted in 4 patient cases presented here—family physicians can learn to

• expect the unexpected
• reduce unpredictability by learning about patients and their context
• attack patient resistance by seeking epiphanies or using positive attractors
• recognize the sensitivity of our patients’ trajectories and use or anticipate it
• promote the healthy benefits of nonlinearity.

TABLE 1
Basic tenets of Newtonian and complexity world views

TABLE 2
Nonlinear characteristics relevant to family practice

Nonlinearity as a truer model of health

Although our basic medical knowledge is built on a reductionistic approach that assumes linear dynamics, our models rarely account for more than 30% of whatever outcome we are investigating. Clinical providers are often faced with the unexpected.

Although linearity suggests that illness should respond in predictable ways regardless of the environment, family physicians know that context is critical. In addition, the human condition is often nonlinear; nonlinear dynamics (chaotic or random dynamics) have been documented in physiology,¹ psychology,^2,3 sociology,⁴ business,^5-7 and economics.⁸

In fact, nonlinear dynamics are often a sign of health. For example, mood may vary in linear patterns among patients with affective disorders; therapy for mood disorders may work by changing the pathological linear dynamics in mood into more healthy nonlinear dynamics.⁹ Linear (or periodic) dynamics often indicate a pathological condition.^10,11

As science and medicine begin to embrace the nonlinearity of complexity science, we must anticipate, recognize, and apply nonlinearity to the care of our patients. This is particularly important for family physicians.

Applying nonlinearity to patient cases

The following cases demonstrate characteristics of nonlinear dynamics (TABLE 2).

Case 1: Sensitivity to initial conditions

I.C. is a 25-year-old teacher who is 6 weeks postpartum. Recently, while at a local shopping mall, she experienced a sudden onset of chest discomfort, palpitations, dizziness, trembling, and a sense of impending doom. The episode peaked in intensity within 3 minutes and lasted 20 minutes after leaving the mall. Although she has not experienced another attack, she has progressively limited her activities since, until now, she has not been able to bring herself to re-enter the mall for fear of another attack. In fact, she reports intense anxiety in anticipation of possibly visiting the mall and has begun limiting her driving in general.

Agoraphobia is linked to the location and interpretation of the first panic attack.¹² This demonstrates the concept of sensitivity to initial conditions whereby small differences in starting values result in very different behaviors later. In other words, apparently minor differences in a patient’s initial physical and emotional state can translate into drastically different outcomes over time.

This emphasizes the need for physicians to pay attention to detail during stressful events that patients experience. For example, if a patient experiences the first panic attack in a self-perceived “safe” environment or interprets the attack as a normal response, she may avoid the disabling consequence of agoraphobia and remain functional. There are other examples of this sensitivity to small changes, such as siblings of similar genetic make-up and environment who exhibit markedly different health as adults may do so because of “minor” life events each experienced.

Similarly, patients with chronic stable disease who, after a minor event, suddenly change their disease trajectory may be demonstrating sensitivity to initial conditions. This sensitivity has been proposed as an explanation for sudden infant death syndrome¹³ (SIDS) and “brittle” diabetes.¹⁴ Treatment response may depend on sensitivity to initial conditions; cases documenting placebo effects or unpredictable potassium excretion on re-administration of potassium-sparing diuretics are examples.¹⁵

Implications for management. Sensitivity to initial conditions has several implications for patient management.

First, we need to recognize the impact it has on patients. Minor life changes can alter the trajectories of patients, so we need to seek the patient’s perspective on stressors they experience. This inherent instability means that “watchful waiting” is a viable approach in some patients because illness may resolve without intervention. It also means that we need to be watchful for signs of an unhealthy trajectory developing in patients even after minor stressors.

Second, sensitivity to initial conditions implies that nonlinear behaviors can change with minor but well-timed interventions. The importance of chronotherapy (timed dosing based on biological rhythms) is receiving increasing attention. Drug efficacy often varies with the time of day.^16,17 Though focused on matching circadian rhythms, chronotherapy may be valuable in nonlinear systems if, through our in-depth knowledge of the patient and context, we can identify a point of leverage when an otherwise “ineffective” treatment may be effective for the patient at a specific point in time. There may be justification for re-administering a previously ineffective treatment if you believe the responsiveness of the patient may have changed. Sensitivity to initial conditions may also explain the effectiveness of placebos.

• For the patient above, minimizing the impact of sensitivity to initial conditions requires immediate access to the patient during a subsequent sensitive time. If the patient has a family history of panic disorder or has had panic attacks in the past, you could anticipate that the patient may experience a panic attack in the future and prepare her for it by discussing the chemical basis for panic (and its lack of serious physical consequences) and by encouraging her to contact you day-or-night immediately after experiencing one so that you can help her to identify a nonthreatening (even if illogical) cause for it, thus preventing the fears that lead to agoraphobia.
• For the practitioner, sensitivity to initial conditions emphasizes the need to understand the details surrounding a stressor by asking patients about their perceptions of the events, the circumstances, and how they are being affected by the stressor. Using this sensitivity for treatment implies focusing on the timing of interventions, and considering re-administration of treatments or even the use of placebos.

Case 2: Effects of attractors

A.T. is a 47-year-old factory worker with a 30-year history of alcohol consumption. His daily intake consisted of a case of beer until he quit 3 years ago. He has periodically suffered relapses consisting of 3 or 4 days of binge drinking followed by prolonged abstinence. Although his wife and 2 children are supportive of his efforts at abstinence, his son dramatically increased his alcohol consumption when his father stopped his daily consumption. In addition, his teenage daughter began experimenting with drugs 2 years ago.

Alcoholism serves as an attractor, controlling not only the patient’s behavior, but the behavior of the family.¹⁸ Attractors limit the range of possible behaviors and thereby resist or limit changes in a patient’s course. Attractors may be internalized models or belief systems that lead to recurring patterns of behavior, even though sensitivity to initial conditions prevents one from predicting the specific path the system will follow; patterns are predictable, the path followed is not. The combination of attractors and sensitivity to initial conditions ensures nonlinearity.

Alcoholism is not the only example of a factor that molds behavior and resists change. Our lives are governed by repeating patterns of behavior. Lifestyle routines are deeply ingrained and resist change even for medically important reasons. These lifestyle patterns may be due to attractors and may explain the resistance to change that many of our diabetic patients exhibit. Similarly, dysfunctional families often display counterproductive patterns of behavior that are resistant to even the best counseling.

Implications for management. The presence of attractors suggests several implications for patient management.

First, we should anticipate resistance and not be frustrated when it occurs.

Second, we can attack the attractor itself,¹⁹ by identifying another, more positive attractor in the patient’s life and reinforcing it to diminish the negative attractor’s impact. For example, instead of simply criticizing the inactive lifestyle ingrained in a hypercholesterolemic patient, we reinforce the positive attractor of the patient’s affection for his grandchildren and use that attractor to get the patient to exercise.

For strongly negative attractors (eg, alcohol use), we could simply attack the attractor itself without providing an alternate attractor. Though this approach is more risky because of the unpredictability of what the patient will substitute, if the attractor is bad enough, we may be willing to allow the patient to choose any other attractor, assuming that it must be more positive than the original.

• For the patient above, simply attacking the negative attractor may lead to other negative behaviors (eg, smoking). The best approach may be to focus on positive attractors (ie, wife, children, social relationships, hobbies), perhaps even positive attractors for the entire family, to move him or them away from the negative attractor.
• For the practitioner, it is imperative to identify the attractors producing recurrent detrimental behaviors that need to be changed. Using attractors for management means that potential positive attractors need to be identified through exploration with patients and reinforced while attacking the negative attractors currently producing the unhealthy behavior.

Case 3: Bifurcation effects

B.I. is a 50-year-old plumber who has had type 2 diabetes for more than 10 years. Though he has regularly seen his physician and taken his medications, his diabetes control has been poor (hemoglobin A_1c=10.2). He admits that compliance with his diet and exercise has been “spotty” at best. Six months ago, his older brother began dialysis for end-stage renal disease secondary to diabetes. Within 1 week of his brother’s first dialysis session, B.I. began walking 30 minutes each night and eliminated evening snacks. Consequently, he has lost 22 pounds, and his hemoglobin A_1c has dropped to 8.1.

Sudden dramatic changes (bifurcations) can occur in nonlinear systems as the system reaches a “tipping point.” In this case, chronic noncompliance suddenly changed to compliance after a meaningful event.²⁰ These bifurcations represent a qualitative change in behavior linked to a change in an attractor. Hence, epiphanies may represent behavioral bifurcations.²¹ Such epiphanies are important in premature menopause²² and initial family decisions to hospitalize a mentally ill relative.²³

Bifurcations have been best documented in cardiovascular disease. Pulsus paradoxicus, pulsus alternans in congestive heart failure (CHF), and cardiac movement in tamponade reflect bifurcations in the system as minor changes cause the system to cross a “tipping point” and produce sudden drastic effects.¹⁰ Similarly, bifurcations in heart rhythm are seen in sick sinus syndrome and ST-T alternans in ventricular tachycardia.²⁴ Paradoxical behavior of the PR interval²⁵ and the disastrous effect of the R-on-T phenomenon are other examples. However, bifurcation dynamics are also important in psychosocial behavior. Sudden drastic changes in mood have been documented in patients with generalized anxiety disorder.²⁶

Implications for management.

• For the patient above, we can first look for events that could serve as epiphanies (eg, development of lung cancer in a relative of our tobacco-dependent patient) and use them to alter behavior. Many physicians already look for consequences of diabetes in friends and relatives to motivate their patients.
• For the practitioner, the existence of bifurcations implies that sudden unforeseen behavior should be expected and should not be a source of frustration. From a management perspective, drastic changes in patient behavior can be achieved by exploring patients’ lives and recognizing and reinforcing epiphanies.

Case 4: Self-organized behavior

S.O. is a 40-year-old housewife with a long history of intermittent anxiety, usually in response to a family stressor. She presents with extreme apprehension and insomnia. On examination, she is restless and mildly tachycardic. Upon further questioning, she denies any recent adverse events but, in fact, reports that her husband recently received a promotion including a significant increase in salary. In reviewing her chart, you notice that you have diagnosed her with adjustment disorder with anxious mood on 3 previous occasions after adverse family stressors. However, her reactions have often been out-of-proportion to the level of stress and she has occasionally reported significant stressors (eg, death of a sister) without subsequent anxiety.

Neurologic systems tend to organize themselves in response to external events and internal models. These self-organized systems consist of tenuously linked parts at the edge of stability balanced between periodic and chaotic behavior. They react to stressors in patterned ways, but the magnitude of the reaction can vary from little or no response to a catastrophic reaction. Because such self-organization can be temporary, with groups periodically forming and dissolving, behavior over time is random without recurrent patterns.

With this patient, varying degrees of stress (even positive events) result in varying degrees of dysfunction with little relationship between the magnitude of stress and the magnitude of dysfunction. The periodic collapse in response to cumulative stress is not the only example of self-organized behavior.

Self-organization is believed to be critical in a variety of neuropsychiatric conditions from personality disorders² and conversion reactions to adult consequences of childhood adversity.²⁷ Patterns of detoxification in groups of alcoholics demonstrate self-organized behavior.²⁸ Self-organization is important to understanding self-regulation and behavior in families.^29,30 Even social interaction patterns among groups of patients on psychiatric wards show self-organized behavior as unstable groups form, dissolve, and reform.³¹

Implications for management. If non-linearity indeed reflects health and helps to keep patients in good health, we should be promoting nonlinear behavior. Studies have shown that frequent small interventions can keep a system that is prone to periodic behavior in nonlinearity.^32,33 Similarly, because nonlinear systems can display a spectrum of behaviors from periodic-to-self-organized behavior-to-chaotic dynamics depending upon their resources and interconnectedness, social systems exhibiting periodic behavior may move into nonlinearity in response to increased resources and decreased restraints,⁵ or to increased interconnectedness.³⁴

Perhaps we can train systems to maximize their variability. For example, exercise programs that used variable intensities and durations may promote a cardiovascular system capable of responding to whatever stressor comes along.

• For the patient above, the self-organized behavior is detrimental, producing over-reaction to stressors; a more chaotic mood pattern would minimize the impact of stressors. The best approach to achieve this may be to increase resources and decrease restraints. Thus, providing the patient with several ways of dealing with stress (ie, multiple treatment modalities including relaxation techniques, self-hypnosis, meditation, PRN anxiolytics) while promoting connectedness with others (ie, support groups, internet, church contacts, meditation) may increase chaotic variability.
• For the practitioner, self-organized behavior may explain the apparent random response to stress in patients. Such unstable behavior can be managed by providing multiple interventions simultaneously (ie, behavioral, pharmacological, social) or temporally (eg, frequent reinforcements of desired behavior) to encourage healthy nonlinearity.

Nonlinearity of primary versus specialty care

Do patients in primary care exhibit a different degree of nonlinearity than those seen in specialty care settings? Generally, yes. Mental illness, for instance, tends to be more severe among psychiatric patients than among primary care patients,^35-37 and CHF is more severe among cardiology patients.³⁸

Differences in severity of illness are important because, in some cases, the more severe the illness, the more periodic the dynamics.^9,39,40 Thus, the nonlinearity decreases as the severity increases. Because diseases exhibiting periodic dynamics should have a more predictable response to therapy, we would expect more severe illnesses to respond more predictably.⁴¹ This pattern has indeed been observed. Prognosis and predictability of treatment response is related to severity of illness in CHF, acute myocardial infarction, depression, and agoraphobia.^38,42-47

Thus, for both biomedical and psychosocial problems, predictability of treatment response correlates with the severity of illness. If patients seen in specialty settings have more severe disease, then we should expect that primary care patients exhibit more nonlinear behavior and are thus less predictable in their response.

Learning to see differently

Though trained to approach medical problems looking through “linear lenses,” we see nonlinear behavior all the time in our patients. If nonlinear processes represent health, then when systems are using healthy, nonlinear dynamics, they are resistant to disruptive external stressors. However, when such systems transition into periodicity due to illness, they may become predictable and more amenable to intervention, permitting physicians to treat them and hopefully restore the healthy, nonlinear dynamics.

Sensitivity to minor changes in their environment, resistance to change, sudden dramatic change in behavior, and intermittent collapses characterize behaviors in many patients. If we understand the nonlinear nature of these behaviors, we will be better able to help our patients.

Expect the unexpected, reduce unpredictability by learning about patients and their contexts, attack resistance by seeking epiphanies or using positive attractors, recognize the sensitivity of our patients’ trajectories and use or anticipate it when possible, and promote the healthy benefits of nonlinearity.

CORRESPONDENCE
David A. Katerndahl, MD, MA, Department of Family and Community Medicine, University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Dr. MC 7795, San Antonio, TX 78229-3900. E-mail: [email protected].

Shin splints and stress fractures are injuries we are accustomed to seeing among younger athletes. A less wellknown painful injury becoming more prevalent—perhaps especially among young female athletes—is chronic compartment syndrome (CCS).

The patient’s history offers more important clues to CCS than does the physical exam. Direct measurement of intracompartmental pressure—as described in this article—is the best means of confirming the diagnosis. Because it’s an invasive procedure, first rule out shin splints and other more common disorders. In older, even nonathletic persons, who can also suffer from CCS, the list of disorders to rule out grows longer.

Outpatient surgical treatment relieves the pain of CCS and most often prevents recurrences.

Clinical presentation

The classic presentation of CCS is a young athlete who complains of an aching pain or a crampy ache in the lower leg during exercise, in an area of the leg corresponding to 1 of its 4 compartments (FIGURE 1). Symptoms are bilateral in 70% to 80% of athletes who develop compartment syndrome.^1,2,3

Symptom traits vary. Other patients describe a sensation of tightness, and some even notice a swelling or bulge over the affected compartment. Many patients describe a numbness or tingling at the affected site that occasionally radiates to the foot. They also often report a feeling of weakness and, in severe cases, may develop true muscle weakness such as drop foot (slap sign). Less often, patients with CCS report a severe shooting pain.

Symptoms usually resolve within minutes to hours following the cessation of activity. However, they can linger for 1 to 2 days following intense activity, especially when the athlete continues to participate despite symptoms.³

Symptoms follow regular pattern. While the onset of symptoms can vary among CCS sufferers, patients often note that symptoms recur consistently at a specific distance (if running or cycling) or duration of exercise.⁷ Temporary rest from or complete cessation of the associated activity may relieve symptoms, but a return to full participation inevitably leads to more severe symptoms. Those who ignore symptoms and continue their activity may over weeks or months experience a worsening in severity and an earlier onset of symptoms.²

Clinical exam much less helpful than the history. With most patients who have CCS, the exam is completely normal. In fact, the real value of the exam may be to detect signs associated with other disorders in the differential (TABLE).

However, occasionally positive findings support the diagnosis of CCS. The finding most commonly encountered is muscle herniation, seen in 40% to 50% of patients.^7,8 These hernias result from a defect in the overlying fascia and are evidenced by a soft tissue bulge that becomes more prominent with muscle contraction. Hernias can be tender to palpation or asymptomatic.²

Another finding, in the absence of herniation, is tenderness of the affected area when palpated. Less frequently, the examiner may note swelling or tightness over the affected compartment.

A few patients with anterior or lateral compartment syndromes have significant weakness on dorsiflexion.^2,8

Examining the patient immediately after exercise may increase the sensitivity of all tests for abnormal sensations and weakness.

TABLE 1
Differential diagnosis for chronic compartment syndrome

Rule out other disorders early

Since confirmatory testing for CCS is invasive, rule out alternative diagnoses before proceeding with the workup. Two much more common diagnoses in young athletes are shin splints and stress fractures. Diffuse tenderness along the posteromedial aspect of the tibia suggests shin splints. Point tenderness along the anterior or posteromedial tibia suggests a stress fracture.

Another common diagnosis to consider in all age groups is disk herniation. Concomitant low back pain or pain that radiates from above the knee in a dermatomal distribution makes this diagnosis more likely. In older patients, spinal stenosis can cause radicular pain that must be distinguished from CCS. A thorough neurologic examination is imperative to rule out these conditions.

Also consider claudication with older patients, particularly those who have risk factors for peripheral vascular disease.

Peripheral neuropathy and gastrocnemius/soleus strain are other common diagnoses that are part of the differential. Uncommon yet important diagnoses to consider include osteomyelitis and tumor.

Rare disorders

Though much less common than the above diagnoses, popliteal artery entrapment syndrome (PAES) and peroneal nerve entrapments are important to discuss since their presentations can be almost identical to CCS.

PAES can mimic posterior CCS in presentation and should be part of the differential for any young patient complaining of exertional calf pain. PAES is a rare condition in which the popliteal artery becomes compressed from strenuous exercise, resulting in ischemia. The underlying anatomic anomaly is the relationship of the popliteal artery to the adjacent musculotendinous structures.⁹ If you suspect PAES, evaluate for a diminished dorsalis pedis pulse with either passive plantar flexion or active dorsiflexion. Patients with exertional calf pain and a positive test result require imaging. These patients should undergo Doppler imaging at rest, with passive plantar flexion, and with active dorsiflexion. Equivocal or incomplete studies require arteriography.¹⁰

Similarly, peroneal nerve entrapments can masquerade as anterior/lateral compartment syndromes. While the common, deep, and superficial peroneal nerves can all be entrapped, the superficial branch is most frequently involved.¹¹ The superficial branch exits the lateral compartment approximately 10 cm proximal to the lateral malleolus. Tenderness at this site upon active dorsiflexion, a positive tinel sign at this site, or lateral leg pain upon plantar flexion and inversion all suggest superficial peroneal entrapment.¹¹ Electromyography, oftentimes performed after compartment pressures are found to be normal, is necessary to confirm the diagnosis.

Direct pressure measurement best evaluation

Despite attempts to develop noninvasive testing for CCS, the gold standard remains the direct measurement of intracompartmental pressures. There is some controversy regarding the upper limit of normal, but most authors agree that a resting pressure >15 mm Hg is abnormally elevated.^1,2,5 Other cutoffs cited in the literature include a 1-minute post-exercise value greater than 30 mm Hg, and a 5-minute post-exercise reading >25 mm Hg.⁵ Turnipseed and colleagues recommend checking post-exercise values in patients with modestly elevated resting pressures of 16 to 24 mm Hg to confirm the diagnosis. Because of the pathophysiology of the syndrome, it seems appropriate to also check 5- or 10-minute post-exercise pressures in patients with typical clinical presentations and normal resting pressures.

Orthopedic, vascular, and general surgeons receive training in compartmental pressure measurement. There are 4 techniques for measurement: the simple needle, the Whitesides infusion, the wick catheter, and the slit catheter. Simple needle devices (such as the Stryker stic, FIGURE 2) have increased in popularity because they are handheld, portable, contain a scale within the device for direct measurement, and are easily operated in a clinic setting.¹⁴

Diagnostic tools that may yet prove useful

Many other diagnostic modalities have been studied in the last 10 years and may eventually prove helpful. Near infrared spectroscopy has demonstrated a relative deoxygenation during exercise in CCS patients and a delayed reoxygenation in the immediate post-exercise period.¹⁵

MIBI scintigraphy has shown decreased muscle perfusion at peak exercise when compared with next-day imaging.^16,17

Laser Doppler flow studies conducted in the post-exercise period have demonstrated a delayed peak hyperemia in CCS patients, different from the gradual decrease in blood flow seen in controls.¹⁸

Finally, multiple studies have been conducted done with magnetic resonance imaging. Comparisons of pre- and postexercise imaging have shown a greater increase in signal intensity in patients with CCS.¹⁹ Despite these findings, though, no clear clinical indications for these tools exist.

FIGURE 2
Measurement with Stryker stic

Treatment options

Conservative measures unconvincing

A few studies have addressed the efficacy of conservative treatment of CCS, but most physicians have found such therapy unsuccessful. Modalities tried include nonsteroidal anti-inflammatory drugs (NSAIDs), diuretics, compression, stretching, orthoses, massage, prolonged rest, change in training surface, change in footwear, change in sport, heat, electrostimulation, and whirlpool.²⁰

One small study showed that compartment pressures in middle-aged men with CCS decreased following treatment with diuretics.²¹ Another small study showed that athletes receiving compartmental massage increased the duration of time they were able to exercise before developing symptoms.²² A third trial found competitive runners to have significantly higher pressures than competitive cyclists, suggesting runners with CCS may be able to bike as an alternative form of exercise.²³

The few data regarding other modalities suggests that only prolonged rest or cessation of sport is effective.^24,25 These options, however, may be impractical for competitive athletes who often have no desire to modify their sports participation. With little available evidence, the decision to offer conservative treatment and the type of conservative therapy to recommend remain controversial.

Fasciotomy track record good

Fasciotomy is the most widely used surgical treatment for CCS. The surgeon makes 1 to 3 small skin incisions, exposes the underlying fascia of the affected compartment, and incises the fascia (FIGURES 3 AND 4). Most of these operations are performed under local anesthesia and last less than an hour. Complications are rare, but can include seroma, hemorrhage, cellulitis, peripheral nerve injury, popliteal injury, tibial artery injury, and deep vein thrombosis (DVT).^1,2

Patient satisfaction with the operation ranges from 80% to 95% in most published series.^1,2,3,5,26 The typical postoperative course involves crutches for the first few days, followed by nonimpact aerobic training for 2 to 4 weeks. Impact aerobic training is next, with gradual return to sport by 2 to 3 months depending on the intensity of sports participation and type of sport practiced.^1,4

Faciectomy preferred by some. Though published long-term follow-up data are few, Turnipseed et al found an 11% CCS recurrence rate among patients undergoing fasciotomies before 1985, compared with a 6% recurrence among patients who received fasciectomies between 1991 and 2001.¹ The latter procedure involves excision of strips of fascia, which theoretically should decrease the risk of scarring over the compartment and subsequent recurrence of symptoms. This is one reason fasciectomy has become the procedure of choice for some physicians, while others reserved it for patients with recurrent CCS. Until a randomized study is performed, it will be difficult to compare the efficacy of fasciotomy and faciectomy.

FIGURE 3
5 days postsurgery

FIGURE 4
6 months later

Acknowledgments

The author would like to thank Dr. William Turnipseed, MD, Department of Vascular Surgery, University of Wisconsin, for his assistance in the preparation of this review article.

CORRESPONDING AUTHOR
Jon Englund, MD, Memorial Healthplex Sports Medicine Institute, 111 West Jefferson, South Bend, IN 46601. E-mail: [email protected]

Shin splints and stress fractures are injuries we are accustomed to seeing among younger athletes. A less wellknown painful injury becoming more prevalent—perhaps especially among young female athletes—is chronic compartment syndrome (CCS).

The patient’s history offers more important clues to CCS than does the physical exam. Direct measurement of intracompartmental pressure—as described in this article—is the best means of confirming the diagnosis. Because it’s an invasive procedure, first rule out shin splints and other more common disorders. In older, even nonathletic persons, who can also suffer from CCS, the list of disorders to rule out grows longer.

Outpatient surgical treatment relieves the pain of CCS and most often prevents recurrences.

Clinical presentation

The classic presentation of CCS is a young athlete who complains of an aching pain or a crampy ache in the lower leg during exercise, in an area of the leg corresponding to 1 of its 4 compartments (FIGURE 1). Symptoms are bilateral in 70% to 80% of athletes who develop compartment syndrome.^1,2,3

Symptom traits vary. Other patients describe a sensation of tightness, and some even notice a swelling or bulge over the affected compartment. Many patients describe a numbness or tingling at the affected site that occasionally radiates to the foot. They also often report a feeling of weakness and, in severe cases, may develop true muscle weakness such as drop foot (slap sign). Less often, patients with CCS report a severe shooting pain.

Symptoms usually resolve within minutes to hours following the cessation of activity. However, they can linger for 1 to 2 days following intense activity, especially when the athlete continues to participate despite symptoms.³

Symptoms follow regular pattern. While the onset of symptoms can vary among CCS sufferers, patients often note that symptoms recur consistently at a specific distance (if running or cycling) or duration of exercise.⁷ Temporary rest from or complete cessation of the associated activity may relieve symptoms, but a return to full participation inevitably leads to more severe symptoms. Those who ignore symptoms and continue their activity may over weeks or months experience a worsening in severity and an earlier onset of symptoms.²

Clinical exam much less helpful than the history. With most patients who have CCS, the exam is completely normal. In fact, the real value of the exam may be to detect signs associated with other disorders in the differential (TABLE).

However, occasionally positive findings support the diagnosis of CCS. The finding most commonly encountered is muscle herniation, seen in 40% to 50% of patients.^7,8 These hernias result from a defect in the overlying fascia and are evidenced by a soft tissue bulge that becomes more prominent with muscle contraction. Hernias can be tender to palpation or asymptomatic.²

Another finding, in the absence of herniation, is tenderness of the affected area when palpated. Less frequently, the examiner may note swelling or tightness over the affected compartment.

A few patients with anterior or lateral compartment syndromes have significant weakness on dorsiflexion.^2,8

Examining the patient immediately after exercise may increase the sensitivity of all tests for abnormal sensations and weakness.

TABLE 1
Differential diagnosis for chronic compartment syndrome

Rule out other disorders early

Since confirmatory testing for CCS is invasive, rule out alternative diagnoses before proceeding with the workup. Two much more common diagnoses in young athletes are shin splints and stress fractures. Diffuse tenderness along the posteromedial aspect of the tibia suggests shin splints. Point tenderness along the anterior or posteromedial tibia suggests a stress fracture.

Another common diagnosis to consider in all age groups is disk herniation. Concomitant low back pain or pain that radiates from above the knee in a dermatomal distribution makes this diagnosis more likely. In older patients, spinal stenosis can cause radicular pain that must be distinguished from CCS. A thorough neurologic examination is imperative to rule out these conditions.

Also consider claudication with older patients, particularly those who have risk factors for peripheral vascular disease.

Peripheral neuropathy and gastrocnemius/soleus strain are other common diagnoses that are part of the differential. Uncommon yet important diagnoses to consider include osteomyelitis and tumor.

Rare disorders

Though much less common than the above diagnoses, popliteal artery entrapment syndrome (PAES) and peroneal nerve entrapments are important to discuss since their presentations can be almost identical to CCS.

PAES can mimic posterior CCS in presentation and should be part of the differential for any young patient complaining of exertional calf pain. PAES is a rare condition in which the popliteal artery becomes compressed from strenuous exercise, resulting in ischemia. The underlying anatomic anomaly is the relationship of the popliteal artery to the adjacent musculotendinous structures.⁹ If you suspect PAES, evaluate for a diminished dorsalis pedis pulse with either passive plantar flexion or active dorsiflexion. Patients with exertional calf pain and a positive test result require imaging. These patients should undergo Doppler imaging at rest, with passive plantar flexion, and with active dorsiflexion. Equivocal or incomplete studies require arteriography.¹⁰

Similarly, peroneal nerve entrapments can masquerade as anterior/lateral compartment syndromes. While the common, deep, and superficial peroneal nerves can all be entrapped, the superficial branch is most frequently involved.¹¹ The superficial branch exits the lateral compartment approximately 10 cm proximal to the lateral malleolus. Tenderness at this site upon active dorsiflexion, a positive tinel sign at this site, or lateral leg pain upon plantar flexion and inversion all suggest superficial peroneal entrapment.¹¹ Electromyography, oftentimes performed after compartment pressures are found to be normal, is necessary to confirm the diagnosis.

Direct pressure measurement best evaluation

Despite attempts to develop noninvasive testing for CCS, the gold standard remains the direct measurement of intracompartmental pressures. There is some controversy regarding the upper limit of normal, but most authors agree that a resting pressure >15 mm Hg is abnormally elevated.^1,2,5 Other cutoffs cited in the literature include a 1-minute post-exercise value greater than 30 mm Hg, and a 5-minute post-exercise reading >25 mm Hg.⁵ Turnipseed and colleagues recommend checking post-exercise values in patients with modestly elevated resting pressures of 16 to 24 mm Hg to confirm the diagnosis. Because of the pathophysiology of the syndrome, it seems appropriate to also check 5- or 10-minute post-exercise pressures in patients with typical clinical presentations and normal resting pressures.

Orthopedic, vascular, and general surgeons receive training in compartmental pressure measurement. There are 4 techniques for measurement: the simple needle, the Whitesides infusion, the wick catheter, and the slit catheter. Simple needle devices (such as the Stryker stic, FIGURE 2) have increased in popularity because they are handheld, portable, contain a scale within the device for direct measurement, and are easily operated in a clinic setting.¹⁴

Diagnostic tools that may yet prove useful

Many other diagnostic modalities have been studied in the last 10 years and may eventually prove helpful. Near infrared spectroscopy has demonstrated a relative deoxygenation during exercise in CCS patients and a delayed reoxygenation in the immediate post-exercise period.¹⁵

MIBI scintigraphy has shown decreased muscle perfusion at peak exercise when compared with next-day imaging.^16,17

Laser Doppler flow studies conducted in the post-exercise period have demonstrated a delayed peak hyperemia in CCS patients, different from the gradual decrease in blood flow seen in controls.¹⁸

Finally, multiple studies have been conducted done with magnetic resonance imaging. Comparisons of pre- and postexercise imaging have shown a greater increase in signal intensity in patients with CCS.¹⁹ Despite these findings, though, no clear clinical indications for these tools exist.

FIGURE 2
Measurement with Stryker stic

Treatment options

Conservative measures unconvincing

A few studies have addressed the efficacy of conservative treatment of CCS, but most physicians have found such therapy unsuccessful. Modalities tried include nonsteroidal anti-inflammatory drugs (NSAIDs), diuretics, compression, stretching, orthoses, massage, prolonged rest, change in training surface, change in footwear, change in sport, heat, electrostimulation, and whirlpool.²⁰

One small study showed that compartment pressures in middle-aged men with CCS decreased following treatment with diuretics.²¹ Another small study showed that athletes receiving compartmental massage increased the duration of time they were able to exercise before developing symptoms.²² A third trial found competitive runners to have significantly higher pressures than competitive cyclists, suggesting runners with CCS may be able to bike as an alternative form of exercise.²³

The few data regarding other modalities suggests that only prolonged rest or cessation of sport is effective.^24,25 These options, however, may be impractical for competitive athletes who often have no desire to modify their sports participation. With little available evidence, the decision to offer conservative treatment and the type of conservative therapy to recommend remain controversial.

Fasciotomy track record good

Fasciotomy is the most widely used surgical treatment for CCS. The surgeon makes 1 to 3 small skin incisions, exposes the underlying fascia of the affected compartment, and incises the fascia (FIGURES 3 AND 4). Most of these operations are performed under local anesthesia and last less than an hour. Complications are rare, but can include seroma, hemorrhage, cellulitis, peripheral nerve injury, popliteal injury, tibial artery injury, and deep vein thrombosis (DVT).^1,2

Patient satisfaction with the operation ranges from 80% to 95% in most published series.^1,2,3,5,26 The typical postoperative course involves crutches for the first few days, followed by nonimpact aerobic training for 2 to 4 weeks. Impact aerobic training is next, with gradual return to sport by 2 to 3 months depending on the intensity of sports participation and type of sport practiced.^1,4

Faciectomy preferred by some. Though published long-term follow-up data are few, Turnipseed et al found an 11% CCS recurrence rate among patients undergoing fasciotomies before 1985, compared with a 6% recurrence among patients who received fasciectomies between 1991 and 2001.¹ The latter procedure involves excision of strips of fascia, which theoretically should decrease the risk of scarring over the compartment and subsequent recurrence of symptoms. This is one reason fasciectomy has become the procedure of choice for some physicians, while others reserved it for patients with recurrent CCS. Until a randomized study is performed, it will be difficult to compare the efficacy of fasciotomy and faciectomy.

FIGURE 3
5 days postsurgery

FIGURE 4
6 months later

Acknowledgments

The author would like to thank Dr. William Turnipseed, MD, Department of Vascular Surgery, University of Wisconsin, for his assistance in the preparation of this review article.

CORRESPONDING AUTHOR
Jon Englund, MD, Memorial Healthplex Sports Medicine Institute, 111 West Jefferson, South Bend, IN 46601. E-mail: [email protected]

Shin splints and stress fractures are injuries we are accustomed to seeing among younger athletes. A less wellknown painful injury becoming more prevalent—perhaps especially among young female athletes—is chronic compartment syndrome (CCS).

The patient’s history offers more important clues to CCS than does the physical exam. Direct measurement of intracompartmental pressure—as described in this article—is the best means of confirming the diagnosis. Because it’s an invasive procedure, first rule out shin splints and other more common disorders. In older, even nonathletic persons, who can also suffer from CCS, the list of disorders to rule out grows longer.

Outpatient surgical treatment relieves the pain of CCS and most often prevents recurrences.

Clinical presentation

The classic presentation of CCS is a young athlete who complains of an aching pain or a crampy ache in the lower leg during exercise, in an area of the leg corresponding to 1 of its 4 compartments (FIGURE 1). Symptoms are bilateral in 70% to 80% of athletes who develop compartment syndrome.^1,2,3

Symptom traits vary. Other patients describe a sensation of tightness, and some even notice a swelling or bulge over the affected compartment. Many patients describe a numbness or tingling at the affected site that occasionally radiates to the foot. They also often report a feeling of weakness and, in severe cases, may develop true muscle weakness such as drop foot (slap sign). Less often, patients with CCS report a severe shooting pain.

Symptoms usually resolve within minutes to hours following the cessation of activity. However, they can linger for 1 to 2 days following intense activity, especially when the athlete continues to participate despite symptoms.³

Symptoms follow regular pattern. While the onset of symptoms can vary among CCS sufferers, patients often note that symptoms recur consistently at a specific distance (if running or cycling) or duration of exercise.⁷ Temporary rest from or complete cessation of the associated activity may relieve symptoms, but a return to full participation inevitably leads to more severe symptoms. Those who ignore symptoms and continue their activity may over weeks or months experience a worsening in severity and an earlier onset of symptoms.²

Clinical exam much less helpful than the history. With most patients who have CCS, the exam is completely normal. In fact, the real value of the exam may be to detect signs associated with other disorders in the differential (TABLE).

However, occasionally positive findings support the diagnosis of CCS. The finding most commonly encountered is muscle herniation, seen in 40% to 50% of patients.^7,8 These hernias result from a defect in the overlying fascia and are evidenced by a soft tissue bulge that becomes more prominent with muscle contraction. Hernias can be tender to palpation or asymptomatic.²

Another finding, in the absence of herniation, is tenderness of the affected area when palpated. Less frequently, the examiner may note swelling or tightness over the affected compartment.

A few patients with anterior or lateral compartment syndromes have significant weakness on dorsiflexion.^2,8

Examining the patient immediately after exercise may increase the sensitivity of all tests for abnormal sensations and weakness.

TABLE 1
Differential diagnosis for chronic compartment syndrome

Rule out other disorders early

Since confirmatory testing for CCS is invasive, rule out alternative diagnoses before proceeding with the workup. Two much more common diagnoses in young athletes are shin splints and stress fractures. Diffuse tenderness along the posteromedial aspect of the tibia suggests shin splints. Point tenderness along the anterior or posteromedial tibia suggests a stress fracture.

Another common diagnosis to consider in all age groups is disk herniation. Concomitant low back pain or pain that radiates from above the knee in a dermatomal distribution makes this diagnosis more likely. In older patients, spinal stenosis can cause radicular pain that must be distinguished from CCS. A thorough neurologic examination is imperative to rule out these conditions.

Also consider claudication with older patients, particularly those who have risk factors for peripheral vascular disease.

Peripheral neuropathy and gastrocnemius/soleus strain are other common diagnoses that are part of the differential. Uncommon yet important diagnoses to consider include osteomyelitis and tumor.

Rare disorders

Though much less common than the above diagnoses, popliteal artery entrapment syndrome (PAES) and peroneal nerve entrapments are important to discuss since their presentations can be almost identical to CCS.

PAES can mimic posterior CCS in presentation and should be part of the differential for any young patient complaining of exertional calf pain. PAES is a rare condition in which the popliteal artery becomes compressed from strenuous exercise, resulting in ischemia. The underlying anatomic anomaly is the relationship of the popliteal artery to the adjacent musculotendinous structures.⁹ If you suspect PAES, evaluate for a diminished dorsalis pedis pulse with either passive plantar flexion or active dorsiflexion. Patients with exertional calf pain and a positive test result require imaging. These patients should undergo Doppler imaging at rest, with passive plantar flexion, and with active dorsiflexion. Equivocal or incomplete studies require arteriography.¹⁰

Similarly, peroneal nerve entrapments can masquerade as anterior/lateral compartment syndromes. While the common, deep, and superficial peroneal nerves can all be entrapped, the superficial branch is most frequently involved.¹¹ The superficial branch exits the lateral compartment approximately 10 cm proximal to the lateral malleolus. Tenderness at this site upon active dorsiflexion, a positive tinel sign at this site, or lateral leg pain upon plantar flexion and inversion all suggest superficial peroneal entrapment.¹¹ Electromyography, oftentimes performed after compartment pressures are found to be normal, is necessary to confirm the diagnosis.

Direct pressure measurement best evaluation

Despite attempts to develop noninvasive testing for CCS, the gold standard remains the direct measurement of intracompartmental pressures. There is some controversy regarding the upper limit of normal, but most authors agree that a resting pressure >15 mm Hg is abnormally elevated.^1,2,5 Other cutoffs cited in the literature include a 1-minute post-exercise value greater than 30 mm Hg, and a 5-minute post-exercise reading >25 mm Hg.⁵ Turnipseed and colleagues recommend checking post-exercise values in patients with modestly elevated resting pressures of 16 to 24 mm Hg to confirm the diagnosis. Because of the pathophysiology of the syndrome, it seems appropriate to also check 5- or 10-minute post-exercise pressures in patients with typical clinical presentations and normal resting pressures.

Orthopedic, vascular, and general surgeons receive training in compartmental pressure measurement. There are 4 techniques for measurement: the simple needle, the Whitesides infusion, the wick catheter, and the slit catheter. Simple needle devices (such as the Stryker stic, FIGURE 2) have increased in popularity because they are handheld, portable, contain a scale within the device for direct measurement, and are easily operated in a clinic setting.¹⁴

Diagnostic tools that may yet prove useful

Many other diagnostic modalities have been studied in the last 10 years and may eventually prove helpful. Near infrared spectroscopy has demonstrated a relative deoxygenation during exercise in CCS patients and a delayed reoxygenation in the immediate post-exercise period.¹⁵

MIBI scintigraphy has shown decreased muscle perfusion at peak exercise when compared with next-day imaging.^16,17

Laser Doppler flow studies conducted in the post-exercise period have demonstrated a delayed peak hyperemia in CCS patients, different from the gradual decrease in blood flow seen in controls.¹⁸

Finally, multiple studies have been conducted done with magnetic resonance imaging. Comparisons of pre- and postexercise imaging have shown a greater increase in signal intensity in patients with CCS.¹⁹ Despite these findings, though, no clear clinical indications for these tools exist.

FIGURE 2
Measurement with Stryker stic

Treatment options

Conservative measures unconvincing

A few studies have addressed the efficacy of conservative treatment of CCS, but most physicians have found such therapy unsuccessful. Modalities tried include nonsteroidal anti-inflammatory drugs (NSAIDs), diuretics, compression, stretching, orthoses, massage, prolonged rest, change in training surface, change in footwear, change in sport, heat, electrostimulation, and whirlpool.²⁰

One small study showed that compartment pressures in middle-aged men with CCS decreased following treatment with diuretics.²¹ Another small study showed that athletes receiving compartmental massage increased the duration of time they were able to exercise before developing symptoms.²² A third trial found competitive runners to have significantly higher pressures than competitive cyclists, suggesting runners with CCS may be able to bike as an alternative form of exercise.²³

The few data regarding other modalities suggests that only prolonged rest or cessation of sport is effective.^24,25 These options, however, may be impractical for competitive athletes who often have no desire to modify their sports participation. With little available evidence, the decision to offer conservative treatment and the type of conservative therapy to recommend remain controversial.

Fasciotomy track record good

Fasciotomy is the most widely used surgical treatment for CCS. The surgeon makes 1 to 3 small skin incisions, exposes the underlying fascia of the affected compartment, and incises the fascia (FIGURES 3 AND 4). Most of these operations are performed under local anesthesia and last less than an hour. Complications are rare, but can include seroma, hemorrhage, cellulitis, peripheral nerve injury, popliteal injury, tibial artery injury, and deep vein thrombosis (DVT).^1,2

Patient satisfaction with the operation ranges from 80% to 95% in most published series.^1,2,3,5,26 The typical postoperative course involves crutches for the first few days, followed by nonimpact aerobic training for 2 to 4 weeks. Impact aerobic training is next, with gradual return to sport by 2 to 3 months depending on the intensity of sports participation and type of sport practiced.^1,4

Faciectomy preferred by some. Though published long-term follow-up data are few, Turnipseed et al found an 11% CCS recurrence rate among patients undergoing fasciotomies before 1985, compared with a 6% recurrence among patients who received fasciectomies between 1991 and 2001.¹ The latter procedure involves excision of strips of fascia, which theoretically should decrease the risk of scarring over the compartment and subsequent recurrence of symptoms. This is one reason fasciectomy has become the procedure of choice for some physicians, while others reserved it for patients with recurrent CCS. Until a randomized study is performed, it will be difficult to compare the efficacy of fasciotomy and faciectomy.

FIGURE 3
5 days postsurgery

FIGURE 4
6 months later

Acknowledgments

The author would like to thank Dr. William Turnipseed, MD, Department of Vascular Surgery, University of Wisconsin, for his assistance in the preparation of this review article.

CORRESPONDING AUTHOR
Jon Englund, MD, Memorial Healthplex Sports Medicine Institute, 111 West Jefferson, South Bend, IN 46601. E-mail: [email protected]

Given the US population’s increasing interest in herbal and alternative therapies, it is just a matter of time before you will be fielding questions from patients about Ayurveda—a comprehensive system of health care that originated in India roughly 3000 years ago, emphasizing the relationships among body, mind, and spirit. Diabetes has been a particular focus of this medical system, and a few Ayurvedic herbs do indeed appear to exert glucose lowering mechanisms.

But how strong is the evidence? Though Ayurveda involves the interaction of many kinds of therapy, this review examines only the literature pertaining to herbal remedies your patients with diabetes are likely to ask about.

Ayurveda more than herbs

Ayurveda (Sanskrit for “knowledge of life” or “knowledge of longevity”) has now spread beyond India’s borders to include the rest of the Indian subcontinent, Sri Lanka, Malaysia, Mauritius, South Africa, Japan, Russia, Europe, and North America. At its root, Ayurveda seeks to restore an individual’s innate harmony. Primary Ayurvedic treatments include diet, exercise, meditation, herbs, massage, exposure to sunlight, controlled breathing, and detoxification treatments.

Dosha imbalance. Ayurveda considers human beings a microcosm of nature. All living and nonliving things are believed to consist of 5 basic elements: ether (space), air, fire, water, and earth. In humans these elements combine and manifest in the human body as 3 humors, or doshas, known as vata, pitta, and kapha.

The doshas govern all biological, psychological, and pathophysiological functions. They are fundamental to human health, and an imbalance of the doshas brings on illness. Consequently, Ayurveda seeks to bring the doshas back into harmony. Balanced doshas, as well as good-quality tissues (dhatus), proper digestion, and elimination of excretions (malas), are considered essential in Ayurveda for maintaining health.

Dosha body type influences treatment. In addition, humans are endowed at birth with 1 of 7 body types, depending on which dosha or combination of the 3 basic doshas dominates. Body types can be dominated by a single humor (vata, pitta, or kapha), or they can be dominated by combinations: vata-kapha (when vata and kapha are present in almost equal amounts); vata-pitta; pitta-kapha; or vata-pitta-kapha. A person’s dosha body type is expressed physically and emotionally; for example, a person with a vata-dominant body type will have a thin frame and an insecure temperament. The Ayurvedic physician takes body type and imbalances among the doshas into consideration when treating a patient.

Application to diabetes. The Indian word for diabetes is madhumeha, from madhu, meaning “sweet/sweetness,” and meha, meaning excessive urination. Classical Ayurvedic therapy for madhumeha (diabetes) begins with an assessment of the dosha imbalance. In all types of pramehas (urinary disorders), kapha is vitiated, but in madhumeha, vata is often aggravated as well. Therefore, therapies will be directed at both vata and kapha simultaneously to restore the balance of the doshas.^1,2

Herbs are a commonly used treatment in Ayurveda. Ethnobotanical studies of traditional herbal remedies used for diabetes around the world have identified more than 1200 species of plants with hypoglycemic activity. These plants are broadly distributed throughout 725 different genera. The pharmacopoeia of India is especially rich in herbal treatments for diabetes. Eighty-five percent of the 20 antidiabetes plants most widely used around the world are prescribed in India.³ A few of the herbs commonly used by Ayurveda practitioners to treat diabetes are summarized in the TABLE.

In response to the increasing interest in Ayurveda and herbal treatments for diabetes in this country, the National Center for Complimentary and Alternative Medicine (NCCAM) requested we conduct a systematic review of Ayurvedic therapy for diabetes. This paper reports the results of that review, and is a condensation of an evidence report available at www.ahrq.gov/clinic/epcsums/ayurvsum.htm.

TABLE
Ayurvedic characteristics of herbs commonly used to treat diabetes

Methods

Identification of literature

We used the search term “Ayurveda” plus the names of 16 major botanicals characteristically used in Ayurveda for an initial search of the Western literature. The herbal terms were added to the search to increase its sensitivity, making it possible to find studies that used Ayurvedic herbal therapy without necessarily being directly identified as Ayurvedic studies.

We were concerned that a large body of literature existed in India that could not be obtained through the conventional search strategy outlined above. Hence, we decided to conduct a search for Ayurvedic literature from the Indian subcontinent. A physician member of the research team, fluent in English and Hindi and who had trained in India, went to India to identify Ayurvedic literature available there.

We therefore used several sources for our search:

• An initial “waterfront” search, which already had identified 120 articles on diabetes and Ayurveda (which was used by NCCAM to establish the rationale for selecting diabetes as the focus of this review).⁴ This search is detailed in our Evidence Report (available at www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=hstat1.chapter.95372) and included Medline, HealthSTAR, Allied and Complementary Medicine, MANTIS, CAB HEALTH, BIOSIS Previews, and EMBASE.
• The in-person literature search in India—we obtained 16 volumes of abstracts from various Indian Ayurvedic journals, as compiled by the Central Council for Research in Ayurveda and Siddha (CCRAS) library. This yielded 318 titles.
• An additional online search of the previously mentioned databases for common botanicals used in the treatment of diabetes. This search yielded 773 additional articles.
• A search of the online CINAHL database, which yielded 70 articles.
• The reference lists of important articles in the Ayurveda/diabetes literature. We identified review articles of herbal treatment of diabetes and checked to see if they focused on any of the Ayurvedic herbal therapies we had identified for diabetes. Checking the bibliographies of all the articles we identified from any source, we found an additional 30 titles that were potentially relevant to our search.
• An RCT or a CCT or a natural experiment with a comparison arm that did not receive an herb or Ayurvedic therapy. RCTs of any size were included. For CCTs or natural experiments, the study had to have at least 1 treatment arm that contained at least 10 patients.
• Because of a paucity of studies meeting these design criteria, we also assessed cohort or case series data that used a “pre/post” method of analysis, if their sample included at least 10 subjects.
• The study had to test Ayurveda as a method or herbs used as a single agent, a formula acting as a single agent, or a limited combination of products (no more than 3) acting as a single agent. The agent had to be dispensed more than 1 time to the study patients (in other words, no single-dose studies).
• The study had to report on at least 1 of 3 outcome measures at 30 days minimum following the start of the study: glycosylated hemoglobin (Hb A_1c), fasting blood glucose, or postprandial blood glucose at either 2 hours (preferred) or 1 hour (acceptable).

Data extraction

Detailed information from each study was extracted using a specialized form. The physician reviewers, working independently, extracted data in duplicate and resolved disagreements by consensus. A senior physician resolved any remaining disagreements. Information extracted from articles included the diagnostic criteria used to make the diagnosis (Ayurvedic or Western or both); the location in which the study was done; the subject population (age, gender, other demographics); the sample size in each arm of the study; the interventions used (primarily the names of the individual herbs or mineral and the manner of their preparation); the length of the study; and the outcomes. To evaluate the quality of the design and execution of trials, we collected information on the study design, appropriateness of randomization, blinding, description of withdrawals and dropouts, and concealment of allocation.^5,6 A quality score was calculated for each trial using a system developed by Jadad.⁵ Empirical evidence has shown that studies scoring 2 or less on the Jadad scale report exaggerated results compared with studies scoring 3 or more.⁷ While other elements of the design and execution of controlled trials have been proposed as quality measures, empirical evidence supporting their use as generic quality measures is lacking.⁸

Data synthesis

The studies we reviewed were judged clinically to be too heterogeneous to support a quantitative synthesis; therefore, our review is narrative. However, to promote the comparison of results across studies we extracted outcome data in a common format and present it in graphs for visual comparison. For studies that reported on both 1-hour and 2-hour postprandial blood glucose levels, we chose the 2-hour measurement for the analysis. For studies that reported more than 1 follow-up time, we selected the follow-up time that was closest to 3 months.

Twenty-two studies were selected for further analysis based on the criteria described above. These studies represented either RCTs (2) or CCTs/natural experiments (5) with a comparison arm that did not receive an herb, or studies that had pre/post comparison data available (8).

We considered 2 subgroups of eligible studies separately: the RCTs and CCTs or natural experiments that had nonherbal/Ayurvedic comparison arm(s) were considered 1 subgroup, and all other eligible (pre/post data available) studies made up the second subgroup. The RCT/CCT/natural experiments subgroup allowed us to do a comparison between herbal and nonherbal arms within a study, providing a stronger level of evidence than pre/post comparisons. For these studies, we were able to compare the effects of different treatments after the placebo or nonherbal arm results were taken into account by estimating a difference of differences statistic and its 95% confidence interval. This statistic is equal to the difference between the pre- and post-treatment means in the Ayurvedic arm minus the analogous difference in the nonherbal/Ayurvedic arm. The pre/post studies subgroup did not allow a concurrent comparison, but the difference between pre- and post-treatment means could be estimated along with a 95% confidence interval.⁴ Further details are given in our Evidence Report (available online at www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=hstat1.chapter.95372).

For each study subgroup, we assessed the possibility of publication bias graphically by evaluating a funnel plot of the study summary statistics for asymmetry, which would result from the nonpublication of small trials with negative outcomes. Because graphical evaluation can be subjective, we also conducted an adjusted rank-correlation test⁹ and a regression asymmetry test¹⁰ as formal statistical tests for publication bias.

Results

Description of the evidence

As a result of our search, we considered a total of 1311 articles (FIGURE 1). We reviewed in depth 62 studies, of which 22 were analyzed statistically. Ten studies came from the Western literature, and 12 came from the Indian literature. There were 7 randomized controlled trials, controlled clinical trials, or natural experiments with a treatment/no treatment comparison. There were 15 case series or cohort studies using a “pre/post” analysis. TABLE W2 (available online at www.jfponline.com) summarizes the characteristics of studies.

In terms of interventions, there was no study that tested Ayurveda as a system of care; all tested herbs. Of the 22 studies described narratively in this article, the 2 most commonly tested single herbs were Coccinia indica and fenugreek. The Ayurvedic formula Ayush-82 was the most commonly tested herbal formula. Diet was used as an intervention or co-intervention in 15 of the 22 studies.

FIGURE 1
Ayurveda literature review strategy of focused search for human diabetes articles

Quality of the evidence

Only the RCTs, CCTs, and natural experiments were given a Jadad score, which rates studies on a 0 to 5 scale.⁵ Of the 7 such studies, 3 received a Jadad score of zero, 3 studies scored a 1, and only 1 study received a 4 on this scale. This is the only high quality study in our analysis.

Controlled trials/natural experiments with nonherbal comparison arm(s)

There were 7 RCTs, CCTs or natural experiments that met the inclusion criteria. The data for these 7 studies are reported in FIGURE 2.

Coccinia indica. Azad Khan, Akhtar, and Mahtab (1979) performed a randomized controlled trial to assess the hypoglycemic properties of a preparation of the herb C indica on uncontrolled and untreated type 2 diabetes patients.¹¹ Thirty-eight patients with type 2 diabetes who were attending the outpatient department of the Bangladesh Institute of Research and Rehabilitation in Diabetes in Dhaka, Bangladesh, were recruited for the study. Only uncontrolled or newly diagnosed patients were enlisted. The intervention consisted of a freeze-dried powder of crushed leaves of C indica, given in a dose of 900 mg twice a day for 6 weeks. Control patients received placebo twice a day for the same period. Of the 16 patients receiving C indica, 10 patients showed a marked improvement in the glucose tolerance tests to values closer to normal, while none in the placebo arm showed marked improvement. The Jadad score for the trial was 4. This study was identified from the Eastern literature search.

Holy basil. Agrawal, Rai, and Singh (1996) reported the results of a randomized, placebo controlled crossover trial of the effects of a leaf extract of holy basil (Ocimum sanctum) on patients with type 2 diabetes.¹² Forty patients were recruited from community advertisements in Kanpur, India, and through visits to the offices of local dieticians. All subjects had a 5-day run-in period during which they all consumed a tea made from holy basil leaves. This was followed by an 8-week experimental period. The patients were randomly assigned to drink the holy basil leaf tea for 4 weeks followed by placebo leaves for 4 weeks, or the reverse (the placebo leaves first, followed by the holy basil leaf tea). The authors concluded that holy basil leaf powder caused significant reduction in the level of fasting blood sugar and postprandial blood sugar, and a moderate reduction in cholesterol. The Jadad score for this trial is 1. This study was identified from the Western literature search.

Cinnamonum tamala. Chandola, Tripathi, and Udupa (1980) assessed the hypoglycemic effects of C tamala in type 2 diabetes patients.¹³ Thirty-two patients were recruited from the outpatient departments of the Institute of Medical Sciences, Varanasi, India, and received the intervention and 8 patients with type 2 diabetes served as controls. The socioeconomic characteristics of the patients were not noted, and the authors did not report if the 2 arms of patients were statistically equivalent on important prognostic variables at the start of the trial. The intervention consisted of leaves of C tamala (tejpatra) that were dried, pulverized, and filtered through a fine mesh, and given as 2 heaped teaspoons 4 times a day before meals and tea for a total of 1 month. All patients were advised to follow a 1800-calorie diet with restriction of sugar, potato, and rice. The control patients were only instructed to follow the above diet with no other medications. Fasting blood sugars were assessed in both arms of patients before the trial and at the end of 1 month. The authors reported the mean fasting blood sugar improved from 153 mg/dL to 113 mg/dL (P<.001) at the end of 1 month in patients treated with C tamala, while there was no improvement in control patients. The Jadad score for this study was 0. This study was identified from the Eastern literature search.

Coccina indica. Kamble, Jyotishi, Kamalakar, et al (1996) studied the effect of fresh leaves of C indica on blood sugar levels and on hyperlipidemia in noninsulin-dependent diabetes mellitus (NIDDM, type 2) patients at an Ayurvedic clinic in Nagpur, India.14 Patients were classified into 4 arms: a healthy comparison arm (arm I, 15 patients); a NIDDM comparison arm (arm II, 30 patients); NIDDM patients treated with Coccinia (arm III, 25 patients); and a NIDDM arm treated with chlorpropamide (arm IV, 15 patients). Patients were given the Coccinia (decocted from fresh leaves, dried into a paste, and made into a 3-g tablet) twice a day for 12 weeks, although control group data only exist at 6 weeks. For patients in arm III, the study showed a statistically significant reduction in mean fasting blood sugar, from 160 to 118 mg/dL (P<.001) at 12 weeks, and a significant decrease in blood sugar levels after glucose tolerance tests: the 2-hour level dropped from 308 mg/dL to 142 mg/dL (P<.001). These results were roughly equivalent to those found in the chlorpropamide arm at 6 weeks, with any differences being not statistically significant. Additionally, after 12 weeks of treatment, the Coccinia treated patients showed significant within group reductions in levels of cholesterol (P<.001), phospholipid (P<.05), triglyceride (P<.001), and free fatty acid (P<.001) compared with untreated patients with diabetes. The Jadad score for this study was 0. This study was identified from the Eastern literature search.

Jamun beej. Kohli and Singh (1993) studied the effects of Jamun beej (Eugenia jambolana) on fasting blood sugar, glucose tolerance test, and symptomatic relief of symptoms associated with diabetes.¹⁵ Thirty patients identified from the outpatient unit of the Sunderlal Hospital in Varanasi, India, who had NIDDM confirmed by glucose tolerance test, were given E jambolana crushed into a fine powder at a dose of 4 g 3 times a day for 3 months. No dietary restrictions were included in the study. A second arm of 6 patients with confirmed NIDDM was given chlorpropamide (250 mg/d). No demographic information was provided on these patients. The authors reported a statistically significant reduction in mean fasting blood sugar of 51.86 mg/dL at 2 months (P<.001) but not at 3 months. The study also demonstrated a significant reduction in the response to the glucose tolerance test at both 2 months (P<.001) and at 3 months (P<.01), but comparisons with the patients receiving chlorpropamide were not statistically significant. The Jadad score total is 1. This study was identified from the Eastern literature search.

Gymnema sylvestre. Baskaran, Ahamath, Shanmugasundaram, et al (1990) studied the effects of GS4, a specific isolate of G sylvestre, on the blood sugar of 47 type 2 diabetes patients.¹⁶ The intervention group consisted of 22 patients with diabetes on oral hypoglycemic therapy (continued during the trial), plus herbal therapy with Gymnema at a dose of 400 mg/d. The control group consisted of 25 patients with type 2 diabetes on conventional medications alone. Although the arms were reasonably matched for age and Broca index (percent of ideal body weight) based on our analysis of individual patient data in the study, the 2 arms were statistically different at baseline with respect to fasting blood sugar. The authors reported that in the intervention group at 12 months of follow-up, mean fasting blood sugar improved from 174 mg/dL to 124 mg/dL (P<.001) and Hb A_1c improved from 11.9% to 8.48% (P<.001). Changes in the control group over this time were not significant. Virtually all patients in arm I developed hypoglycemic symptoms, and the dose of their normal oral hypoglycemic agent needed to be changed or stopped. The Jadad score is 0. This study was identified from the Western literature search.

GS4. Shanmugasundaram, Rajeswari, Baskaran, et al (1990b) reported the effect of an herbal compound GS4 on blood glucose, insulin requirements, and cholesterol in patients with type 1 and type 2 diabetes.¹⁷ The intervention group consisted of 23 type 1 patients and 4 type 2 patients (ages 44–50 years, 25% female) who continued their insulin in addition to taking the herbal intervention, and the control group consisted of 37 type 1 patients who were on insulin therapy. The clinical equivalence of the 2 arms at baseline was not mentioned. The intervention drug was GS4, a recrystallized precipitate of the alcoholic extract of the acidic fraction of G sylvestre; at dose of 400 mg/d. At 6 to 8 months of follow-up, all patients in the intervention group developed hypoglycemic episodes, and their insulin doses were reduced by 10 units at a time; differences compared to controls were statistically significant. The Jadad score was 1. This study was identified from the Western literature search.

Summary. In summary, comparisons of the studied herbs with either placebo, diet, or as an adjunct to medical hypoglycemic therapy consistently reported statistically significant benefits in glucose control. Studies comparing herbs to the oral hypoglycemic chlorpropamide did not report differences between groups.

FIGURE 2
Glucose-lowering effects of Ayurvedic herbs compared with nonherbal treatments

Case series or cohort studies without a concurrent comparison group

Fifteen studies had treatment arms that reported blood glucose or Hb A_1c values before and after herbal interventions—ie, that made available pre/post data.^13,18-31 Herbs were tested singly (7 studies), in combination (7 studies), or both singly and in combination (1 study). All of these studies are limited by the lack of randomization. Many are also limited by small sample size and short study duration.

The data are presented in FIGURE 3 in the following order: studies or arms of studies using a single herbal preparation first, followed by studies that use a formula or a group of herbs and minerals. Some studies with multiple arms will have listings in more than one place. As FIGURE 3 shows, most pre/post comparison studies using single herbs and formulas favor treatment. Hb A_1c is measured in only 5 of the studies. Two of those studies seem to strongly favor treatment; 2 are almost equivocal; and 1 does not favor treatment.

The small number of RCTs/CCTs/natural experiments measuring postprandial blood glucose and Hb A_1c, and the small number of case series and cohorts measuring Hb A_1c, make the assessment of publication bias for these outcomes problematic. For the remaining outcomes, only slight evidence of publication bias was evident among the case series and cohort studies for the fasting blood glucose outcome (adjusted rank correlation test P=.97; regression asymmetry test P=.06). For this outcome, we have trial data available, a setting in which no evidence of publication bias is seen.

FIGURE 3
Effect of Ayurvedic herbs on blood glucose and Hb A_1c

Discussion

No controlled trials have assessed Ayurveda as a system of care. Furthermore, among studies of herbal preparations, controlled trials are rare and mostly of poor quality. Still, we found evidence suggesting that some Ayurvedic herbs have clinically important hypoglycemic effects.

The best-quality RCT reported beneficial results for C indica. Other studies have generally reported benefits of C indica; however, results have been inconsistent.

A single trial of holy basil reported a benefit in controlling fasting blood glucose only. G sylvestre was used in a number of studies, again with generally beneficial results. Gymnema was also a constituent in 2 of the formulas that reported beneficial results: D-400 and Ayush-82.

Methodological shortcomings. The evidence we found falls short of proof. As noted, the individual studies we assessed had significant methodological shortcomings. Few used a randomized, placebo-controlled design, despite the feasibility of such methods. Studies generally were underpowered to determine even large effect sizes. Many studies involved an extremely small number of subjects (fewer than 10). A number of studies failed to use appropriate statistical methods in reporting the results. Furthermore, the duration of most studies was short, and only intermediate outcomes (measure of blood sugar control) were reported. Longer term studies that assess clinical outcomes are required. Lastly, there is great heterogeneity in the available literature on Ayurvedic treatment for diabetes, despite the fact that the overwhelming majority of studies test herbal therapy. Heterogeneity exists in the herbs and formulas tested and in the method of their preparation.

No true studies of Ayurveda. None of the identified studies assessed Ayurveda as a system of diagnosis and care. The literature reported mainly studies of single herbs; however, in clinical Ayurvedic practice, herbs are usually given in combination. Formulas can include minerals and metals in addition to multiple plants.

Ayurvedic practitioners assert that the sum of a formula is more potent than the constituents taken individually and, further, that the mineral and metal components boost the formula’s potency. Additional Ayurvedic therapies such as diet, yoga, or special purification routines, should be controlled for when recruiting and assessing patients. It is unclear, perhaps doubtful, that taking a single component of Ayurvedic is sufficient to represent the holistic effect of Ayurvedic interventions.

Limitations of our review. Our searches showed that herbal therapies were emphasized in published studies of Ayurveda. However, no conclusions can be drawn about the use of Ayurveda as a whole system for the treatment of diabetes. Although extensive efforts were made to obtain the relevant literature from India, studies not published in English were excluded, as were MD thesis reports. A handful of studies could not be obtained. Therefore we may not have included in this review all clinical trials. By specifically hand searching documents published in India, we identified that an important body of Ayurvedic literature is not identifiable using traditional systematic review methods.

In summary, we found preliminary evidence suggesting that some herbs and herbal formulas used in Ayurvedic therapy can have clinically important glucose-lowering effects. C indica, holy basil, and Gymnema are particularly worthy of additional study.

Acknowledgments

This research was performed by the Southern California Evidence-Based Practice Center based at RAND, Santa Monica, CA under contract with the Agency for Healthcare Research and Quality (Contract No. 290-97-0001) with support from the National Center for Complementary and Alternative Medicine. The authors of this article are responsible for its contents. No statement in this article should be construed as an official position of the Agency for Healthcare Research and Quality, the National Center for Complementary and Alternative Medicine, National Institutes of Health, or the US Department of Health and Human Services.

CORRESPONDING AUTHOR
Paul G. Shekelle, MD, PhD, West Los Angeles VA Medical Center, 111G, 11301 Wilshire Blvd, Los Angeles, CA 90073. E-mail: [email protected]. At the time of this study, Dr. Shekelle was a Senior Research Associate of the Veterans Affairs Health Services Research and Development Service.

Given the US population’s increasing interest in herbal and alternative therapies, it is just a matter of time before you will be fielding questions from patients about Ayurveda—a comprehensive system of health care that originated in India roughly 3000 years ago, emphasizing the relationships among body, mind, and spirit. Diabetes has been a particular focus of this medical system, and a few Ayurvedic herbs do indeed appear to exert glucose lowering mechanisms.

But how strong is the evidence? Though Ayurveda involves the interaction of many kinds of therapy, this review examines only the literature pertaining to herbal remedies your patients with diabetes are likely to ask about.

Ayurveda more than herbs

Ayurveda (Sanskrit for “knowledge of life” or “knowledge of longevity”) has now spread beyond India’s borders to include the rest of the Indian subcontinent, Sri Lanka, Malaysia, Mauritius, South Africa, Japan, Russia, Europe, and North America. At its root, Ayurveda seeks to restore an individual’s innate harmony. Primary Ayurvedic treatments include diet, exercise, meditation, herbs, massage, exposure to sunlight, controlled breathing, and detoxification treatments.

Dosha imbalance. Ayurveda considers human beings a microcosm of nature. All living and nonliving things are believed to consist of 5 basic elements: ether (space), air, fire, water, and earth. In humans these elements combine and manifest in the human body as 3 humors, or doshas, known as vata, pitta, and kapha.

The doshas govern all biological, psychological, and pathophysiological functions. They are fundamental to human health, and an imbalance of the doshas brings on illness. Consequently, Ayurveda seeks to bring the doshas back into harmony. Balanced doshas, as well as good-quality tissues (dhatus), proper digestion, and elimination of excretions (malas), are considered essential in Ayurveda for maintaining health.

Dosha body type influences treatment. In addition, humans are endowed at birth with 1 of 7 body types, depending on which dosha or combination of the 3 basic doshas dominates. Body types can be dominated by a single humor (vata, pitta, or kapha), or they can be dominated by combinations: vata-kapha (when vata and kapha are present in almost equal amounts); vata-pitta; pitta-kapha; or vata-pitta-kapha. A person’s dosha body type is expressed physically and emotionally; for example, a person with a vata-dominant body type will have a thin frame and an insecure temperament. The Ayurvedic physician takes body type and imbalances among the doshas into consideration when treating a patient.

Application to diabetes. The Indian word for diabetes is madhumeha, from madhu, meaning “sweet/sweetness,” and meha, meaning excessive urination. Classical Ayurvedic therapy for madhumeha (diabetes) begins with an assessment of the dosha imbalance. In all types of pramehas (urinary disorders), kapha is vitiated, but in madhumeha, vata is often aggravated as well. Therefore, therapies will be directed at both vata and kapha simultaneously to restore the balance of the doshas.^1,2

Herbs are a commonly used treatment in Ayurveda. Ethnobotanical studies of traditional herbal remedies used for diabetes around the world have identified more than 1200 species of plants with hypoglycemic activity. These plants are broadly distributed throughout 725 different genera. The pharmacopoeia of India is especially rich in herbal treatments for diabetes. Eighty-five percent of the 20 antidiabetes plants most widely used around the world are prescribed in India.³ A few of the herbs commonly used by Ayurveda practitioners to treat diabetes are summarized in the TABLE.

In response to the increasing interest in Ayurveda and herbal treatments for diabetes in this country, the National Center for Complimentary and Alternative Medicine (NCCAM) requested we conduct a systematic review of Ayurvedic therapy for diabetes. This paper reports the results of that review, and is a condensation of an evidence report available at www.ahrq.gov/clinic/epcsums/ayurvsum.htm.

TABLE
Ayurvedic characteristics of herbs commonly used to treat diabetes

Methods

Identification of literature

We used the search term “Ayurveda” plus the names of 16 major botanicals characteristically used in Ayurveda for an initial search of the Western literature. The herbal terms were added to the search to increase its sensitivity, making it possible to find studies that used Ayurvedic herbal therapy without necessarily being directly identified as Ayurvedic studies.

We were concerned that a large body of literature existed in India that could not be obtained through the conventional search strategy outlined above. Hence, we decided to conduct a search for Ayurvedic literature from the Indian subcontinent. A physician member of the research team, fluent in English and Hindi and who had trained in India, went to India to identify Ayurvedic literature available there.

We therefore used several sources for our search:

• An initial “waterfront” search, which already had identified 120 articles on diabetes and Ayurveda (which was used by NCCAM to establish the rationale for selecting diabetes as the focus of this review).⁴ This search is detailed in our Evidence Report (available at www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=hstat1.chapter.95372) and included Medline, HealthSTAR, Allied and Complementary Medicine, MANTIS, CAB HEALTH, BIOSIS Previews, and EMBASE.
• The in-person literature search in India—we obtained 16 volumes of abstracts from various Indian Ayurvedic journals, as compiled by the Central Council for Research in Ayurveda and Siddha (CCRAS) library. This yielded 318 titles.
• An additional online search of the previously mentioned databases for common botanicals used in the treatment of diabetes. This search yielded 773 additional articles.
• A search of the online CINAHL database, which yielded 70 articles.
• The reference lists of important articles in the Ayurveda/diabetes literature. We identified review articles of herbal treatment of diabetes and checked to see if they focused on any of the Ayurvedic herbal therapies we had identified for diabetes. Checking the bibliographies of all the articles we identified from any source, we found an additional 30 titles that were potentially relevant to our search.
• An RCT or a CCT or a natural experiment with a comparison arm that did not receive an herb or Ayurvedic therapy. RCTs of any size were included. For CCTs or natural experiments, the study had to have at least 1 treatment arm that contained at least 10 patients.
• Because of a paucity of studies meeting these design criteria, we also assessed cohort or case series data that used a “pre/post” method of analysis, if their sample included at least 10 subjects.
• The study had to test Ayurveda as a method or herbs used as a single agent, a formula acting as a single agent, or a limited combination of products (no more than 3) acting as a single agent. The agent had to be dispensed more than 1 time to the study patients (in other words, no single-dose studies).
• The study had to report on at least 1 of 3 outcome measures at 30 days minimum following the start of the study: glycosylated hemoglobin (Hb A_1c), fasting blood glucose, or postprandial blood glucose at either 2 hours (preferred) or 1 hour (acceptable).

Data extraction

Detailed information from each study was extracted using a specialized form. The physician reviewers, working independently, extracted data in duplicate and resolved disagreements by consensus. A senior physician resolved any remaining disagreements. Information extracted from articles included the diagnostic criteria used to make the diagnosis (Ayurvedic or Western or both); the location in which the study was done; the subject population (age, gender, other demographics); the sample size in each arm of the study; the interventions used (primarily the names of the individual herbs or mineral and the manner of their preparation); the length of the study; and the outcomes. To evaluate the quality of the design and execution of trials, we collected information on the study design, appropriateness of randomization, blinding, description of withdrawals and dropouts, and concealment of allocation.^5,6 A quality score was calculated for each trial using a system developed by Jadad.⁵ Empirical evidence has shown that studies scoring 2 or less on the Jadad scale report exaggerated results compared with studies scoring 3 or more.⁷ While other elements of the design and execution of controlled trials have been proposed as quality measures, empirical evidence supporting their use as generic quality measures is lacking.⁸

Data synthesis

The studies we reviewed were judged clinically to be too heterogeneous to support a quantitative synthesis; therefore, our review is narrative. However, to promote the comparison of results across studies we extracted outcome data in a common format and present it in graphs for visual comparison. For studies that reported on both 1-hour and 2-hour postprandial blood glucose levels, we chose the 2-hour measurement for the analysis. For studies that reported more than 1 follow-up time, we selected the follow-up time that was closest to 3 months.

Twenty-two studies were selected for further analysis based on the criteria described above. These studies represented either RCTs (2) or CCTs/natural experiments (5) with a comparison arm that did not receive an herb, or studies that had pre/post comparison data available (8).

We considered 2 subgroups of eligible studies separately: the RCTs and CCTs or natural experiments that had nonherbal/Ayurvedic comparison arm(s) were considered 1 subgroup, and all other eligible (pre/post data available) studies made up the second subgroup. The RCT/CCT/natural experiments subgroup allowed us to do a comparison between herbal and nonherbal arms within a study, providing a stronger level of evidence than pre/post comparisons. For these studies, we were able to compare the effects of different treatments after the placebo or nonherbal arm results were taken into account by estimating a difference of differences statistic and its 95% confidence interval. This statistic is equal to the difference between the pre- and post-treatment means in the Ayurvedic arm minus the analogous difference in the nonherbal/Ayurvedic arm. The pre/post studies subgroup did not allow a concurrent comparison, but the difference between pre- and post-treatment means could be estimated along with a 95% confidence interval.⁴ Further details are given in our Evidence Report (available online at www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=hstat1.chapter.95372).

For each study subgroup, we assessed the possibility of publication bias graphically by evaluating a funnel plot of the study summary statistics for asymmetry, which would result from the nonpublication of small trials with negative outcomes. Because graphical evaluation can be subjective, we also conducted an adjusted rank-correlation test⁹ and a regression asymmetry test¹⁰ as formal statistical tests for publication bias.

Results

Description of the evidence

As a result of our search, we considered a total of 1311 articles (FIGURE 1). We reviewed in depth 62 studies, of which 22 were analyzed statistically. Ten studies came from the Western literature, and 12 came from the Indian literature. There were 7 randomized controlled trials, controlled clinical trials, or natural experiments with a treatment/no treatment comparison. There were 15 case series or cohort studies using a “pre/post” analysis. TABLE W2 (available online at www.jfponline.com) summarizes the characteristics of studies.

In terms of interventions, there was no study that tested Ayurveda as a system of care; all tested herbs. Of the 22 studies described narratively in this article, the 2 most commonly tested single herbs were Coccinia indica and fenugreek. The Ayurvedic formula Ayush-82 was the most commonly tested herbal formula. Diet was used as an intervention or co-intervention in 15 of the 22 studies.

FIGURE 1
Ayurveda literature review strategy of focused search for human diabetes articles

Quality of the evidence

Only the RCTs, CCTs, and natural experiments were given a Jadad score, which rates studies on a 0 to 5 scale.⁵ Of the 7 such studies, 3 received a Jadad score of zero, 3 studies scored a 1, and only 1 study received a 4 on this scale. This is the only high quality study in our analysis.

Controlled trials/natural experiments with nonherbal comparison arm(s)

There were 7 RCTs, CCTs or natural experiments that met the inclusion criteria. The data for these 7 studies are reported in FIGURE 2.

Coccinia indica. Azad Khan, Akhtar, and Mahtab (1979) performed a randomized controlled trial to assess the hypoglycemic properties of a preparation of the herb C indica on uncontrolled and untreated type 2 diabetes patients.¹¹ Thirty-eight patients with type 2 diabetes who were attending the outpatient department of the Bangladesh Institute of Research and Rehabilitation in Diabetes in Dhaka, Bangladesh, were recruited for the study. Only uncontrolled or newly diagnosed patients were enlisted. The intervention consisted of a freeze-dried powder of crushed leaves of C indica, given in a dose of 900 mg twice a day for 6 weeks. Control patients received placebo twice a day for the same period. Of the 16 patients receiving C indica, 10 patients showed a marked improvement in the glucose tolerance tests to values closer to normal, while none in the placebo arm showed marked improvement. The Jadad score for the trial was 4. This study was identified from the Eastern literature search.

Holy basil. Agrawal, Rai, and Singh (1996) reported the results of a randomized, placebo controlled crossover trial of the effects of a leaf extract of holy basil (Ocimum sanctum) on patients with type 2 diabetes.¹² Forty patients were recruited from community advertisements in Kanpur, India, and through visits to the offices of local dieticians. All subjects had a 5-day run-in period during which they all consumed a tea made from holy basil leaves. This was followed by an 8-week experimental period. The patients were randomly assigned to drink the holy basil leaf tea for 4 weeks followed by placebo leaves for 4 weeks, or the reverse (the placebo leaves first, followed by the holy basil leaf tea). The authors concluded that holy basil leaf powder caused significant reduction in the level of fasting blood sugar and postprandial blood sugar, and a moderate reduction in cholesterol. The Jadad score for this trial is 1. This study was identified from the Western literature search.

Cinnamonum tamala. Chandola, Tripathi, and Udupa (1980) assessed the hypoglycemic effects of C tamala in type 2 diabetes patients.¹³ Thirty-two patients were recruited from the outpatient departments of the Institute of Medical Sciences, Varanasi, India, and received the intervention and 8 patients with type 2 diabetes served as controls. The socioeconomic characteristics of the patients were not noted, and the authors did not report if the 2 arms of patients were statistically equivalent on important prognostic variables at the start of the trial. The intervention consisted of leaves of C tamala (tejpatra) that were dried, pulverized, and filtered through a fine mesh, and given as 2 heaped teaspoons 4 times a day before meals and tea for a total of 1 month. All patients were advised to follow a 1800-calorie diet with restriction of sugar, potato, and rice. The control patients were only instructed to follow the above diet with no other medications. Fasting blood sugars were assessed in both arms of patients before the trial and at the end of 1 month. The authors reported the mean fasting blood sugar improved from 153 mg/dL to 113 mg/dL (P<.001) at the end of 1 month in patients treated with C tamala, while there was no improvement in control patients. The Jadad score for this study was 0. This study was identified from the Eastern literature search.

Coccina indica. Kamble, Jyotishi, Kamalakar, et al (1996) studied the effect of fresh leaves of C indica on blood sugar levels and on hyperlipidemia in noninsulin-dependent diabetes mellitus (NIDDM, type 2) patients at an Ayurvedic clinic in Nagpur, India.14 Patients were classified into 4 arms: a healthy comparison arm (arm I, 15 patients); a NIDDM comparison arm (arm II, 30 patients); NIDDM patients treated with Coccinia (arm III, 25 patients); and a NIDDM arm treated with chlorpropamide (arm IV, 15 patients). Patients were given the Coccinia (decocted from fresh leaves, dried into a paste, and made into a 3-g tablet) twice a day for 12 weeks, although control group data only exist at 6 weeks. For patients in arm III, the study showed a statistically significant reduction in mean fasting blood sugar, from 160 to 118 mg/dL (P<.001) at 12 weeks, and a significant decrease in blood sugar levels after glucose tolerance tests: the 2-hour level dropped from 308 mg/dL to 142 mg/dL (P<.001). These results were roughly equivalent to those found in the chlorpropamide arm at 6 weeks, with any differences being not statistically significant. Additionally, after 12 weeks of treatment, the Coccinia treated patients showed significant within group reductions in levels of cholesterol (P<.001), phospholipid (P<.05), triglyceride (P<.001), and free fatty acid (P<.001) compared with untreated patients with diabetes. The Jadad score for this study was 0. This study was identified from the Eastern literature search.

Jamun beej. Kohli and Singh (1993) studied the effects of Jamun beej (Eugenia jambolana) on fasting blood sugar, glucose tolerance test, and symptomatic relief of symptoms associated with diabetes.¹⁵ Thirty patients identified from the outpatient unit of the Sunderlal Hospital in Varanasi, India, who had NIDDM confirmed by glucose tolerance test, were given E jambolana crushed into a fine powder at a dose of 4 g 3 times a day for 3 months. No dietary restrictions were included in the study. A second arm of 6 patients with confirmed NIDDM was given chlorpropamide (250 mg/d). No demographic information was provided on these patients. The authors reported a statistically significant reduction in mean fasting blood sugar of 51.86 mg/dL at 2 months (P<.001) but not at 3 months. The study also demonstrated a significant reduction in the response to the glucose tolerance test at both 2 months (P<.001) and at 3 months (P<.01), but comparisons with the patients receiving chlorpropamide were not statistically significant. The Jadad score total is 1. This study was identified from the Eastern literature search.

Gymnema sylvestre. Baskaran, Ahamath, Shanmugasundaram, et al (1990) studied the effects of GS4, a specific isolate of G sylvestre, on the blood sugar of 47 type 2 diabetes patients.¹⁶ The intervention group consisted of 22 patients with diabetes on oral hypoglycemic therapy (continued during the trial), plus herbal therapy with Gymnema at a dose of 400 mg/d. The control group consisted of 25 patients with type 2 diabetes on conventional medications alone. Although the arms were reasonably matched for age and Broca index (percent of ideal body weight) based on our analysis of individual patient data in the study, the 2 arms were statistically different at baseline with respect to fasting blood sugar. The authors reported that in the intervention group at 12 months of follow-up, mean fasting blood sugar improved from 174 mg/dL to 124 mg/dL (P<.001) and Hb A_1c improved from 11.9% to 8.48% (P<.001). Changes in the control group over this time were not significant. Virtually all patients in arm I developed hypoglycemic symptoms, and the dose of their normal oral hypoglycemic agent needed to be changed or stopped. The Jadad score is 0. This study was identified from the Western literature search.

GS4. Shanmugasundaram, Rajeswari, Baskaran, et al (1990b) reported the effect of an herbal compound GS4 on blood glucose, insulin requirements, and cholesterol in patients with type 1 and type 2 diabetes.¹⁷ The intervention group consisted of 23 type 1 patients and 4 type 2 patients (ages 44–50 years, 25% female) who continued their insulin in addition to taking the herbal intervention, and the control group consisted of 37 type 1 patients who were on insulin therapy. The clinical equivalence of the 2 arms at baseline was not mentioned. The intervention drug was GS4, a recrystallized precipitate of the alcoholic extract of the acidic fraction of G sylvestre; at dose of 400 mg/d. At 6 to 8 months of follow-up, all patients in the intervention group developed hypoglycemic episodes, and their insulin doses were reduced by 10 units at a time; differences compared to controls were statistically significant. The Jadad score was 1. This study was identified from the Western literature search.

Summary. In summary, comparisons of the studied herbs with either placebo, diet, or as an adjunct to medical hypoglycemic therapy consistently reported statistically significant benefits in glucose control. Studies comparing herbs to the oral hypoglycemic chlorpropamide did not report differences between groups.

FIGURE 2
Glucose-lowering effects of Ayurvedic herbs compared with nonherbal treatments

Case series or cohort studies without a concurrent comparison group

Fifteen studies had treatment arms that reported blood glucose or Hb A_1c values before and after herbal interventions—ie, that made available pre/post data.^13,18-31 Herbs were tested singly (7 studies), in combination (7 studies), or both singly and in combination (1 study). All of these studies are limited by the lack of randomization. Many are also limited by small sample size and short study duration.

The data are presented in FIGURE 3 in the following order: studies or arms of studies using a single herbal preparation first, followed by studies that use a formula or a group of herbs and minerals. Some studies with multiple arms will have listings in more than one place. As FIGURE 3 shows, most pre/post comparison studies using single herbs and formulas favor treatment. Hb A_1c is measured in only 5 of the studies. Two of those studies seem to strongly favor treatment; 2 are almost equivocal; and 1 does not favor treatment.

The small number of RCTs/CCTs/natural experiments measuring postprandial blood glucose and Hb A_1c, and the small number of case series and cohorts measuring Hb A_1c, make the assessment of publication bias for these outcomes problematic. For the remaining outcomes, only slight evidence of publication bias was evident among the case series and cohort studies for the fasting blood glucose outcome (adjusted rank correlation test P=.97; regression asymmetry test P=.06). For this outcome, we have trial data available, a setting in which no evidence of publication bias is seen.

FIGURE 3
Effect of Ayurvedic herbs on blood glucose and Hb A_1c

Discussion

No controlled trials have assessed Ayurveda as a system of care. Furthermore, among studies of herbal preparations, controlled trials are rare and mostly of poor quality. Still, we found evidence suggesting that some Ayurvedic herbs have clinically important hypoglycemic effects.

The best-quality RCT reported beneficial results for C indica. Other studies have generally reported benefits of C indica; however, results have been inconsistent.

A single trial of holy basil reported a benefit in controlling fasting blood glucose only. G sylvestre was used in a number of studies, again with generally beneficial results. Gymnema was also a constituent in 2 of the formulas that reported beneficial results: D-400 and Ayush-82.

Methodological shortcomings. The evidence we found falls short of proof. As noted, the individual studies we assessed had significant methodological shortcomings. Few used a randomized, placebo-controlled design, despite the feasibility of such methods. Studies generally were underpowered to determine even large effect sizes. Many studies involved an extremely small number of subjects (fewer than 10). A number of studies failed to use appropriate statistical methods in reporting the results. Furthermore, the duration of most studies was short, and only intermediate outcomes (measure of blood sugar control) were reported. Longer term studies that assess clinical outcomes are required. Lastly, there is great heterogeneity in the available literature on Ayurvedic treatment for diabetes, despite the fact that the overwhelming majority of studies test herbal therapy. Heterogeneity exists in the herbs and formulas tested and in the method of their preparation.

No true studies of Ayurveda. None of the identified studies assessed Ayurveda as a system of diagnosis and care. The literature reported mainly studies of single herbs; however, in clinical Ayurvedic practice, herbs are usually given in combination. Formulas can include minerals and metals in addition to multiple plants.

Ayurvedic practitioners assert that the sum of a formula is more potent than the constituents taken individually and, further, that the mineral and metal components boost the formula’s potency. Additional Ayurvedic therapies such as diet, yoga, or special purification routines, should be controlled for when recruiting and assessing patients. It is unclear, perhaps doubtful, that taking a single component of Ayurvedic is sufficient to represent the holistic effect of Ayurvedic interventions.

Limitations of our review. Our searches showed that herbal therapies were emphasized in published studies of Ayurveda. However, no conclusions can be drawn about the use of Ayurveda as a whole system for the treatment of diabetes. Although extensive efforts were made to obtain the relevant literature from India, studies not published in English were excluded, as were MD thesis reports. A handful of studies could not be obtained. Therefore we may not have included in this review all clinical trials. By specifically hand searching documents published in India, we identified that an important body of Ayurvedic literature is not identifiable using traditional systematic review methods.

In summary, we found preliminary evidence suggesting that some herbs and herbal formulas used in Ayurvedic therapy can have clinically important glucose-lowering effects. C indica, holy basil, and Gymnema are particularly worthy of additional study.

Acknowledgments

This research was performed by the Southern California Evidence-Based Practice Center based at RAND, Santa Monica, CA under contract with the Agency for Healthcare Research and Quality (Contract No. 290-97-0001) with support from the National Center for Complementary and Alternative Medicine. The authors of this article are responsible for its contents. No statement in this article should be construed as an official position of the Agency for Healthcare Research and Quality, the National Center for Complementary and Alternative Medicine, National Institutes of Health, or the US Department of Health and Human Services.

CORRESPONDING AUTHOR
Paul G. Shekelle, MD, PhD, West Los Angeles VA Medical Center, 111G, 11301 Wilshire Blvd, Los Angeles, CA 90073. E-mail: [email protected]. At the time of this study, Dr. Shekelle was a Senior Research Associate of the Veterans Affairs Health Services Research and Development Service.

Given the US population’s increasing interest in herbal and alternative therapies, it is just a matter of time before you will be fielding questions from patients about Ayurveda—a comprehensive system of health care that originated in India roughly 3000 years ago, emphasizing the relationships among body, mind, and spirit. Diabetes has been a particular focus of this medical system, and a few Ayurvedic herbs do indeed appear to exert glucose lowering mechanisms.

But how strong is the evidence? Though Ayurveda involves the interaction of many kinds of therapy, this review examines only the literature pertaining to herbal remedies your patients with diabetes are likely to ask about.

Ayurveda more than herbs

Ayurveda (Sanskrit for “knowledge of life” or “knowledge of longevity”) has now spread beyond India’s borders to include the rest of the Indian subcontinent, Sri Lanka, Malaysia, Mauritius, South Africa, Japan, Russia, Europe, and North America. At its root, Ayurveda seeks to restore an individual’s innate harmony. Primary Ayurvedic treatments include diet, exercise, meditation, herbs, massage, exposure to sunlight, controlled breathing, and detoxification treatments.

Dosha imbalance. Ayurveda considers human beings a microcosm of nature. All living and nonliving things are believed to consist of 5 basic elements: ether (space), air, fire, water, and earth. In humans these elements combine and manifest in the human body as 3 humors, or doshas, known as vata, pitta, and kapha.

The doshas govern all biological, psychological, and pathophysiological functions. They are fundamental to human health, and an imbalance of the doshas brings on illness. Consequently, Ayurveda seeks to bring the doshas back into harmony. Balanced doshas, as well as good-quality tissues (dhatus), proper digestion, and elimination of excretions (malas), are considered essential in Ayurveda for maintaining health.

Dosha body type influences treatment. In addition, humans are endowed at birth with 1 of 7 body types, depending on which dosha or combination of the 3 basic doshas dominates. Body types can be dominated by a single humor (vata, pitta, or kapha), or they can be dominated by combinations: vata-kapha (when vata and kapha are present in almost equal amounts); vata-pitta; pitta-kapha; or vata-pitta-kapha. A person’s dosha body type is expressed physically and emotionally; for example, a person with a vata-dominant body type will have a thin frame and an insecure temperament. The Ayurvedic physician takes body type and imbalances among the doshas into consideration when treating a patient.

Application to diabetes. The Indian word for diabetes is madhumeha, from madhu, meaning “sweet/sweetness,” and meha, meaning excessive urination. Classical Ayurvedic therapy for madhumeha (diabetes) begins with an assessment of the dosha imbalance. In all types of pramehas (urinary disorders), kapha is vitiated, but in madhumeha, vata is often aggravated as well. Therefore, therapies will be directed at both vata and kapha simultaneously to restore the balance of the doshas.^1,2

Herbs are a commonly used treatment in Ayurveda. Ethnobotanical studies of traditional herbal remedies used for diabetes around the world have identified more than 1200 species of plants with hypoglycemic activity. These plants are broadly distributed throughout 725 different genera. The pharmacopoeia of India is especially rich in herbal treatments for diabetes. Eighty-five percent of the 20 antidiabetes plants most widely used around the world are prescribed in India.³ A few of the herbs commonly used by Ayurveda practitioners to treat diabetes are summarized in the TABLE.

In response to the increasing interest in Ayurveda and herbal treatments for diabetes in this country, the National Center for Complimentary and Alternative Medicine (NCCAM) requested we conduct a systematic review of Ayurvedic therapy for diabetes. This paper reports the results of that review, and is a condensation of an evidence report available at www.ahrq.gov/clinic/epcsums/ayurvsum.htm.

TABLE
Ayurvedic characteristics of herbs commonly used to treat diabetes

Methods

Identification of literature

We used the search term “Ayurveda” plus the names of 16 major botanicals characteristically used in Ayurveda for an initial search of the Western literature. The herbal terms were added to the search to increase its sensitivity, making it possible to find studies that used Ayurvedic herbal therapy without necessarily being directly identified as Ayurvedic studies.

We were concerned that a large body of literature existed in India that could not be obtained through the conventional search strategy outlined above. Hence, we decided to conduct a search for Ayurvedic literature from the Indian subcontinent. A physician member of the research team, fluent in English and Hindi and who had trained in India, went to India to identify Ayurvedic literature available there.

We therefore used several sources for our search:

• An initial “waterfront” search, which already had identified 120 articles on diabetes and Ayurveda (which was used by NCCAM to establish the rationale for selecting diabetes as the focus of this review).⁴ This search is detailed in our Evidence Report (available at www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=hstat1.chapter.95372) and included Medline, HealthSTAR, Allied and Complementary Medicine, MANTIS, CAB HEALTH, BIOSIS Previews, and EMBASE.
• The in-person literature search in India—we obtained 16 volumes of abstracts from various Indian Ayurvedic journals, as compiled by the Central Council for Research in Ayurveda and Siddha (CCRAS) library. This yielded 318 titles.
• An additional online search of the previously mentioned databases for common botanicals used in the treatment of diabetes. This search yielded 773 additional articles.
• A search of the online CINAHL database, which yielded 70 articles.
• The reference lists of important articles in the Ayurveda/diabetes literature. We identified review articles of herbal treatment of diabetes and checked to see if they focused on any of the Ayurvedic herbal therapies we had identified for diabetes. Checking the bibliographies of all the articles we identified from any source, we found an additional 30 titles that were potentially relevant to our search.
• An RCT or a CCT or a natural experiment with a comparison arm that did not receive an herb or Ayurvedic therapy. RCTs of any size were included. For CCTs or natural experiments, the study had to have at least 1 treatment arm that contained at least 10 patients.
• Because of a paucity of studies meeting these design criteria, we also assessed cohort or case series data that used a “pre/post” method of analysis, if their sample included at least 10 subjects.
• The study had to test Ayurveda as a method or herbs used as a single agent, a formula acting as a single agent, or a limited combination of products (no more than 3) acting as a single agent. The agent had to be dispensed more than 1 time to the study patients (in other words, no single-dose studies).
• The study had to report on at least 1 of 3 outcome measures at 30 days minimum following the start of the study: glycosylated hemoglobin (Hb A_1c), fasting blood glucose, or postprandial blood glucose at either 2 hours (preferred) or 1 hour (acceptable).

Data extraction

Detailed information from each study was extracted using a specialized form. The physician reviewers, working independently, extracted data in duplicate and resolved disagreements by consensus. A senior physician resolved any remaining disagreements. Information extracted from articles included the diagnostic criteria used to make the diagnosis (Ayurvedic or Western or both); the location in which the study was done; the subject population (age, gender, other demographics); the sample size in each arm of the study; the interventions used (primarily the names of the individual herbs or mineral and the manner of their preparation); the length of the study; and the outcomes. To evaluate the quality of the design and execution of trials, we collected information on the study design, appropriateness of randomization, blinding, description of withdrawals and dropouts, and concealment of allocation.^5,6 A quality score was calculated for each trial using a system developed by Jadad.⁵ Empirical evidence has shown that studies scoring 2 or less on the Jadad scale report exaggerated results compared with studies scoring 3 or more.⁷ While other elements of the design and execution of controlled trials have been proposed as quality measures, empirical evidence supporting their use as generic quality measures is lacking.⁸

Data synthesis

The studies we reviewed were judged clinically to be too heterogeneous to support a quantitative synthesis; therefore, our review is narrative. However, to promote the comparison of results across studies we extracted outcome data in a common format and present it in graphs for visual comparison. For studies that reported on both 1-hour and 2-hour postprandial blood glucose levels, we chose the 2-hour measurement for the analysis. For studies that reported more than 1 follow-up time, we selected the follow-up time that was closest to 3 months.

Twenty-two studies were selected for further analysis based on the criteria described above. These studies represented either RCTs (2) or CCTs/natural experiments (5) with a comparison arm that did not receive an herb, or studies that had pre/post comparison data available (8).

We considered 2 subgroups of eligible studies separately: the RCTs and CCTs or natural experiments that had nonherbal/Ayurvedic comparison arm(s) were considered 1 subgroup, and all other eligible (pre/post data available) studies made up the second subgroup. The RCT/CCT/natural experiments subgroup allowed us to do a comparison between herbal and nonherbal arms within a study, providing a stronger level of evidence than pre/post comparisons. For these studies, we were able to compare the effects of different treatments after the placebo or nonherbal arm results were taken into account by estimating a difference of differences statistic and its 95% confidence interval. This statistic is equal to the difference between the pre- and post-treatment means in the Ayurvedic arm minus the analogous difference in the nonherbal/Ayurvedic arm. The pre/post studies subgroup did not allow a concurrent comparison, but the difference between pre- and post-treatment means could be estimated along with a 95% confidence interval.⁴ Further details are given in our Evidence Report (available online at www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=hstat1.chapter.95372).

For each study subgroup, we assessed the possibility of publication bias graphically by evaluating a funnel plot of the study summary statistics for asymmetry, which would result from the nonpublication of small trials with negative outcomes. Because graphical evaluation can be subjective, we also conducted an adjusted rank-correlation test⁹ and a regression asymmetry test¹⁰ as formal statistical tests for publication bias.

Results

Description of the evidence

As a result of our search, we considered a total of 1311 articles (FIGURE 1). We reviewed in depth 62 studies, of which 22 were analyzed statistically. Ten studies came from the Western literature, and 12 came from the Indian literature. There were 7 randomized controlled trials, controlled clinical trials, or natural experiments with a treatment/no treatment comparison. There were 15 case series or cohort studies using a “pre/post” analysis. TABLE W2 (available online at www.jfponline.com) summarizes the characteristics of studies.

In terms of interventions, there was no study that tested Ayurveda as a system of care; all tested herbs. Of the 22 studies described narratively in this article, the 2 most commonly tested single herbs were Coccinia indica and fenugreek. The Ayurvedic formula Ayush-82 was the most commonly tested herbal formula. Diet was used as an intervention or co-intervention in 15 of the 22 studies.

FIGURE 1
Ayurveda literature review strategy of focused search for human diabetes articles

Quality of the evidence

Only the RCTs, CCTs, and natural experiments were given a Jadad score, which rates studies on a 0 to 5 scale.⁵ Of the 7 such studies, 3 received a Jadad score of zero, 3 studies scored a 1, and only 1 study received a 4 on this scale. This is the only high quality study in our analysis.

Controlled trials/natural experiments with nonherbal comparison arm(s)

There were 7 RCTs, CCTs or natural experiments that met the inclusion criteria. The data for these 7 studies are reported in FIGURE 2.

Coccinia indica. Azad Khan, Akhtar, and Mahtab (1979) performed a randomized controlled trial to assess the hypoglycemic properties of a preparation of the herb C indica on uncontrolled and untreated type 2 diabetes patients.¹¹ Thirty-eight patients with type 2 diabetes who were attending the outpatient department of the Bangladesh Institute of Research and Rehabilitation in Diabetes in Dhaka, Bangladesh, were recruited for the study. Only uncontrolled or newly diagnosed patients were enlisted. The intervention consisted of a freeze-dried powder of crushed leaves of C indica, given in a dose of 900 mg twice a day for 6 weeks. Control patients received placebo twice a day for the same period. Of the 16 patients receiving C indica, 10 patients showed a marked improvement in the glucose tolerance tests to values closer to normal, while none in the placebo arm showed marked improvement. The Jadad score for the trial was 4. This study was identified from the Eastern literature search.

Holy basil. Agrawal, Rai, and Singh (1996) reported the results of a randomized, placebo controlled crossover trial of the effects of a leaf extract of holy basil (Ocimum sanctum) on patients with type 2 diabetes.¹² Forty patients were recruited from community advertisements in Kanpur, India, and through visits to the offices of local dieticians. All subjects had a 5-day run-in period during which they all consumed a tea made from holy basil leaves. This was followed by an 8-week experimental period. The patients were randomly assigned to drink the holy basil leaf tea for 4 weeks followed by placebo leaves for 4 weeks, or the reverse (the placebo leaves first, followed by the holy basil leaf tea). The authors concluded that holy basil leaf powder caused significant reduction in the level of fasting blood sugar and postprandial blood sugar, and a moderate reduction in cholesterol. The Jadad score for this trial is 1. This study was identified from the Western literature search.

Cinnamonum tamala. Chandola, Tripathi, and Udupa (1980) assessed the hypoglycemic effects of C tamala in type 2 diabetes patients.¹³ Thirty-two patients were recruited from the outpatient departments of the Institute of Medical Sciences, Varanasi, India, and received the intervention and 8 patients with type 2 diabetes served as controls. The socioeconomic characteristics of the patients were not noted, and the authors did not report if the 2 arms of patients were statistically equivalent on important prognostic variables at the start of the trial. The intervention consisted of leaves of C tamala (tejpatra) that were dried, pulverized, and filtered through a fine mesh, and given as 2 heaped teaspoons 4 times a day before meals and tea for a total of 1 month. All patients were advised to follow a 1800-calorie diet with restriction of sugar, potato, and rice. The control patients were only instructed to follow the above diet with no other medications. Fasting blood sugars were assessed in both arms of patients before the trial and at the end of 1 month. The authors reported the mean fasting blood sugar improved from 153 mg/dL to 113 mg/dL (P<.001) at the end of 1 month in patients treated with C tamala, while there was no improvement in control patients. The Jadad score for this study was 0. This study was identified from the Eastern literature search.

Coccina indica. Kamble, Jyotishi, Kamalakar, et al (1996) studied the effect of fresh leaves of C indica on blood sugar levels and on hyperlipidemia in noninsulin-dependent diabetes mellitus (NIDDM, type 2) patients at an Ayurvedic clinic in Nagpur, India.14 Patients were classified into 4 arms: a healthy comparison arm (arm I, 15 patients); a NIDDM comparison arm (arm II, 30 patients); NIDDM patients treated with Coccinia (arm III, 25 patients); and a NIDDM arm treated with chlorpropamide (arm IV, 15 patients). Patients were given the Coccinia (decocted from fresh leaves, dried into a paste, and made into a 3-g tablet) twice a day for 12 weeks, although control group data only exist at 6 weeks. For patients in arm III, the study showed a statistically significant reduction in mean fasting blood sugar, from 160 to 118 mg/dL (P<.001) at 12 weeks, and a significant decrease in blood sugar levels after glucose tolerance tests: the 2-hour level dropped from 308 mg/dL to 142 mg/dL (P<.001). These results were roughly equivalent to those found in the chlorpropamide arm at 6 weeks, with any differences being not statistically significant. Additionally, after 12 weeks of treatment, the Coccinia treated patients showed significant within group reductions in levels of cholesterol (P<.001), phospholipid (P<.05), triglyceride (P<.001), and free fatty acid (P<.001) compared with untreated patients with diabetes. The Jadad score for this study was 0. This study was identified from the Eastern literature search.

Jamun beej. Kohli and Singh (1993) studied the effects of Jamun beej (Eugenia jambolana) on fasting blood sugar, glucose tolerance test, and symptomatic relief of symptoms associated with diabetes.¹⁵ Thirty patients identified from the outpatient unit of the Sunderlal Hospital in Varanasi, India, who had NIDDM confirmed by glucose tolerance test, were given E jambolana crushed into a fine powder at a dose of 4 g 3 times a day for 3 months. No dietary restrictions were included in the study. A second arm of 6 patients with confirmed NIDDM was given chlorpropamide (250 mg/d). No demographic information was provided on these patients. The authors reported a statistically significant reduction in mean fasting blood sugar of 51.86 mg/dL at 2 months (P<.001) but not at 3 months. The study also demonstrated a significant reduction in the response to the glucose tolerance test at both 2 months (P<.001) and at 3 months (P<.01), but comparisons with the patients receiving chlorpropamide were not statistically significant. The Jadad score total is 1. This study was identified from the Eastern literature search.

Gymnema sylvestre. Baskaran, Ahamath, Shanmugasundaram, et al (1990) studied the effects of GS4, a specific isolate of G sylvestre, on the blood sugar of 47 type 2 diabetes patients.¹⁶ The intervention group consisted of 22 patients with diabetes on oral hypoglycemic therapy (continued during the trial), plus herbal therapy with Gymnema at a dose of 400 mg/d. The control group consisted of 25 patients with type 2 diabetes on conventional medications alone. Although the arms were reasonably matched for age and Broca index (percent of ideal body weight) based on our analysis of individual patient data in the study, the 2 arms were statistically different at baseline with respect to fasting blood sugar. The authors reported that in the intervention group at 12 months of follow-up, mean fasting blood sugar improved from 174 mg/dL to 124 mg/dL (P<.001) and Hb A_1c improved from 11.9% to 8.48% (P<.001). Changes in the control group over this time were not significant. Virtually all patients in arm I developed hypoglycemic symptoms, and the dose of their normal oral hypoglycemic agent needed to be changed or stopped. The Jadad score is 0. This study was identified from the Western literature search.

GS4. Shanmugasundaram, Rajeswari, Baskaran, et al (1990b) reported the effect of an herbal compound GS4 on blood glucose, insulin requirements, and cholesterol in patients with type 1 and type 2 diabetes.¹⁷ The intervention group consisted of 23 type 1 patients and 4 type 2 patients (ages 44–50 years, 25% female) who continued their insulin in addition to taking the herbal intervention, and the control group consisted of 37 type 1 patients who were on insulin therapy. The clinical equivalence of the 2 arms at baseline was not mentioned. The intervention drug was GS4, a recrystallized precipitate of the alcoholic extract of the acidic fraction of G sylvestre; at dose of 400 mg/d. At 6 to 8 months of follow-up, all patients in the intervention group developed hypoglycemic episodes, and their insulin doses were reduced by 10 units at a time; differences compared to controls were statistically significant. The Jadad score was 1. This study was identified from the Western literature search.

Summary. In summary, comparisons of the studied herbs with either placebo, diet, or as an adjunct to medical hypoglycemic therapy consistently reported statistically significant benefits in glucose control. Studies comparing herbs to the oral hypoglycemic chlorpropamide did not report differences between groups.

FIGURE 2
Glucose-lowering effects of Ayurvedic herbs compared with nonherbal treatments

Case series or cohort studies without a concurrent comparison group

Fifteen studies had treatment arms that reported blood glucose or Hb A_1c values before and after herbal interventions—ie, that made available pre/post data.^13,18-31 Herbs were tested singly (7 studies), in combination (7 studies), or both singly and in combination (1 study). All of these studies are limited by the lack of randomization. Many are also limited by small sample size and short study duration.

The data are presented in FIGURE 3 in the following order: studies or arms of studies using a single herbal preparation first, followed by studies that use a formula or a group of herbs and minerals. Some studies with multiple arms will have listings in more than one place. As FIGURE 3 shows, most pre/post comparison studies using single herbs and formulas favor treatment. Hb A_1c is measured in only 5 of the studies. Two of those studies seem to strongly favor treatment; 2 are almost equivocal; and 1 does not favor treatment.

The small number of RCTs/CCTs/natural experiments measuring postprandial blood glucose and Hb A_1c, and the small number of case series and cohorts measuring Hb A_1c, make the assessment of publication bias for these outcomes problematic. For the remaining outcomes, only slight evidence of publication bias was evident among the case series and cohort studies for the fasting blood glucose outcome (adjusted rank correlation test P=.97; regression asymmetry test P=.06). For this outcome, we have trial data available, a setting in which no evidence of publication bias is seen.

FIGURE 3
Effect of Ayurvedic herbs on blood glucose and Hb A_1c

Discussion

No controlled trials have assessed Ayurveda as a system of care. Furthermore, among studies of herbal preparations, controlled trials are rare and mostly of poor quality. Still, we found evidence suggesting that some Ayurvedic herbs have clinically important hypoglycemic effects.

The best-quality RCT reported beneficial results for C indica. Other studies have generally reported benefits of C indica; however, results have been inconsistent.

A single trial of holy basil reported a benefit in controlling fasting blood glucose only. G sylvestre was used in a number of studies, again with generally beneficial results. Gymnema was also a constituent in 2 of the formulas that reported beneficial results: D-400 and Ayush-82.

Methodological shortcomings. The evidence we found falls short of proof. As noted, the individual studies we assessed had significant methodological shortcomings. Few used a randomized, placebo-controlled design, despite the feasibility of such methods. Studies generally were underpowered to determine even large effect sizes. Many studies involved an extremely small number of subjects (fewer than 10). A number of studies failed to use appropriate statistical methods in reporting the results. Furthermore, the duration of most studies was short, and only intermediate outcomes (measure of blood sugar control) were reported. Longer term studies that assess clinical outcomes are required. Lastly, there is great heterogeneity in the available literature on Ayurvedic treatment for diabetes, despite the fact that the overwhelming majority of studies test herbal therapy. Heterogeneity exists in the herbs and formulas tested and in the method of their preparation.

No true studies of Ayurveda. None of the identified studies assessed Ayurveda as a system of diagnosis and care. The literature reported mainly studies of single herbs; however, in clinical Ayurvedic practice, herbs are usually given in combination. Formulas can include minerals and metals in addition to multiple plants.

Ayurvedic practitioners assert that the sum of a formula is more potent than the constituents taken individually and, further, that the mineral and metal components boost the formula’s potency. Additional Ayurvedic therapies such as diet, yoga, or special purification routines, should be controlled for when recruiting and assessing patients. It is unclear, perhaps doubtful, that taking a single component of Ayurvedic is sufficient to represent the holistic effect of Ayurvedic interventions.

Limitations of our review. Our searches showed that herbal therapies were emphasized in published studies of Ayurveda. However, no conclusions can be drawn about the use of Ayurveda as a whole system for the treatment of diabetes. Although extensive efforts were made to obtain the relevant literature from India, studies not published in English were excluded, as were MD thesis reports. A handful of studies could not be obtained. Therefore we may not have included in this review all clinical trials. By specifically hand searching documents published in India, we identified that an important body of Ayurvedic literature is not identifiable using traditional systematic review methods.

In summary, we found preliminary evidence suggesting that some herbs and herbal formulas used in Ayurvedic therapy can have clinically important glucose-lowering effects. C indica, holy basil, and Gymnema are particularly worthy of additional study.

Acknowledgments

This research was performed by the Southern California Evidence-Based Practice Center based at RAND, Santa Monica, CA under contract with the Agency for Healthcare Research and Quality (Contract No. 290-97-0001) with support from the National Center for Complementary and Alternative Medicine. The authors of this article are responsible for its contents. No statement in this article should be construed as an official position of the Agency for Healthcare Research and Quality, the National Center for Complementary and Alternative Medicine, National Institutes of Health, or the US Department of Health and Human Services.

CORRESPONDING AUTHOR
Paul G. Shekelle, MD, PhD, West Los Angeles VA Medical Center, 111G, 11301 Wilshire Blvd, Los Angeles, CA 90073. E-mail: [email protected]. At the time of this study, Dr. Shekelle was a Senior Research Associate of the Veterans Affairs Health Services Research and Development Service.

Signs and symptoms of hypoglycemia vary considerably among patients with type 2 diabetes, making the condition easy to miss. Moreover, the most common symptoms are not necessarily the first symptoms.

If hypoglycemia occurs repeatedly, it can start a vicious cycle of physiologic reactions that mask or diminish the symptoms that warn patients of an impending episode. This may lead to hypoglycemia unawareness and hypoglycemic episodes of increasing severity. Fear of hypoglycemia, particularly of nocturnal events, may discourage patients from more intensive glycemic control, particularly using insulin. Such fear may even lead them to reduce their antidiabetic medication dosage, resulting in poor glycemic control.¹

Breaking this cycle and restoring normal physiologic responses is one focus of this article, as is teaching patients how to monitor their blood glucose levels and how to correlate low blood glucose with the signs and symptoms of hypoglycemia.

Other therapies and strategies that we discuss in this article:

• Newer insulin analogs and the associated risk of hypoglycemia with each
• Appropriate combination of insulin with oral antidiabetic medications
• The long-acting analog insulin glargine used as basal insulin to lower the incidence of hypoglycemia, including nocturnal and severe hypoglycemia
• Rapid-acting insulin analogs (aspart, glulisine, and lispro) used in basalprandial insulin regimens.

First symptoms vary among individuals

Symptoms of hypoglycemia result primarily from a lowered glucose level in the brain and its effects on the central and autonomic nervous systems (FIGURE 1). A decrease in glucose below physiologic levels has acute consequences for brain function because the brain has an immediate requirement for glucose and little capacity for storage.

Two types of symptoms

Neuroglycopenia and the inhibition of neuronal metabolism causes sensations of warmth, weakness, fatigue, difficulty concentrating, confusion, behavioral changes, and in the most severe cases, a loss of consciousness, seizures, brain damage, and even death.^2-4

Neurogenic symptoms are mediated by the hormones and neurotransmitters secreted in response to low brain glucose levels (FIGURE 1). The gluconeogenic actions of the autonomic nervous system produce the classic warning symptoms—tremulousness, pounding heart, anxiety, sweating, hunger, and tingling sensations—that usually precede the symptoms of hypoglycemia.^2-4 This is particularly so in iatrogenic hypoglycemia.

These direct symptoms of neuroglycopenia are the ones patients typically identify with hypoglycemia. The most common symptoms of hypoglycemia are therefore not necessarily the first symptoms of hypoglycemia (TABLE 1).⁵ For example, though most patients experience sweating as a symptom of hypoglycemia, the first symptom might be trembling or anxiety, depending on the individual.⁵

TABLE 1
Signs and symptoms most commonly associated with hypoglycemia are not always the first to appear

Factors influencing frequency and severity of hypoglycemia

Aggressive diabetes management commonly causes mild-to-moderate hypoglycemia, defined as a blood glucose value <60 mg/dL, that can be managed by the patient without assistance.

Severe hypoglycemia—a blood glucose value <50 mg/dL—is relatively uncommon in type 2 diabetes and requires the assistance of another person to manage, since neurological impairment may render patients unable to treat themselves.^2,6 Severe hypoglycemia, whether in patients with type 1 or type 2 diabetes, can have debilitating consequences, including seizures or coma or even death.⁷

Long-standing type 2 disease. Hypoglycemia is more common in patients with type 1 diabetes than in those with type 2, but it can occur in type 2 diabetes patients who require insulin or are treated intensively with combinations of oral agents.⁶ As type 2 diabetes progresses,⁸ the incidence of hypoglycemic events increases, as endogenously produced insulin declines and is replaced by exogenous insulin.^5,9 In fact, the prevalence of severe episodes (eg, requiring assistance of another person to administer glucose or glucagon) in patients with type 2 diabetes was comparable to that exhibited among patients with type 1 diabetes if they had been on insulin therapy for the same length of time.^5,10

Nocturnal hypoglycemia. This event poses a special concern because the warning signs of hypoglycemia may be blunted during sleep. It has been reported that as many as 29% to 56% of all adult patients treated with insulin have an overnight glucose profile that indicates hypoglycemia occurs at night.^11-13 However, it is important to note that the extent of the problem of nocturnal hypoglycemia is difficult to assess since overnight monitoring of glucose levels is required.

Additional insights from the UKPDS

Hypoglycemia in type 2 diabetes has not received rigorous attention in clinical trials. However, the United Kingdom Prospective Diabetes Study (UKPDS) was a large longitudinal trial in type 2 diabetes that included hypoglycemia as an outcome measure and thus provides some helpful information.

Events with insulin>sulfonylureas>diet. The 6-year follow-up revealed that the cumulative proportion of patients reporting 1 or more hypoglycemic events (of any type) was 76% for those using insulin, 45% among those taking sulfonylureas, and 3% for those on diet alone. Expressed as events per patient year, this was 37%, 17%, and 0.9%, respectively. When only “major” events (those requiring third-party assistance or hospital admission) were considered, the proportion of patients per year reporting 1 or more such events, was 2.3% for insulin, 0.7% for sulfonylureas, and 0.03% for diet alone. The cumulative proportion over 6 years was 3.3% of participants using sulfonylureas, 11.2% of those using insulin, and 0.15% of those on diet therapy.⁸

Metformin increases risk. The cumulative proportion of obese patients reporting any hypoglycemic event was 17.6% for those taking metformin vs 2.8% for those on diet. Severe hypoglycemia (as defined earlier) occurred in 2.4% of participants using metformin compared with 0.4% of those on diet therapy.⁸

Findings from other studies

Interestingly, in a recent systematic review of randomized controlled trials comparing insulin monotherapy with insulin plus oral antidiabetic agents, 13 of 14 studies reporting hypoglycemia demonstrated no difference in events.¹⁴

The occurrence of hypoglycemia among patients on metformin monotherapy in the UKPDS study is notable since, theoretically, hypoglycemia should not occur with agents whose mechanisms of action do not increase insulin secretion (biguanides, thiazolidinediones [TZDs], or α-glucosidase inhibitors),¹ Since newer classes such as TZDs, α-glucosidase inhibitors, and meglitinides were not available when UKPDS was initiated, the trial does not provide data on these classes.

In a small comparative study of insulin combined with either metformin or a TZD, it appeared that metformin combination was associated with fewer occurrences of hypoglycemia; however, the small patient sample limits generalizability of the finding.¹⁵

FIGURE 1
How hypoglycemia occurs

Normally, as blood glucose levels (red) rise, insulin secretion increases, circulating insulin levels (blue) rise, and hepatic glucose production is inhibited. As glucose is disposed and circulating levels decrease due to insulin action, insulin levels then drop and hepatic glucose production begins again.

Hypoglycemia in diabetes can result from an excess of endogenous or exogenous insulin (iatrogenic hypoglycemia). In healthy patients, high insulin and falling glucose levels suppress insulin production and stimulate a hormone-mediated burst of glucose production. In patients with diabetes, the loss of physiologic control of insulin secretion coupled with exogenous administration of insulin or insulin secretagogues can interfere with the normal physiologic response to low blood glucose levels, resulting in hypoglycemia.¹

With intense insulin regimens, the incidence of hypoglycemia can be as high as 30%, in contrast to 12% for patients treated with diet alone and 16% for those taking oral agents.⁶

With secretagogues, it has been suggested that the incidence of hypoglycemia is higher with the older, longer acting sulfonylurea agents.^1,6,16-19 Although populationbased data on hypoglycemic rates associated with combination therapy with oral antidiabetic agents are not available, numerous clinical studies have reported rates of 10% to 20% for any hypoglycemic event.^20-23

Heightening patient awareness, and yours

Because the signs of hypoglycemia vary considerably among individuals, they can easily be missed.³ In addition, repeated episodes of hypoglycemia can alter the normal regulatory responses and diminish the most important signs of a drop in glucose levels.^1,2,24,25 The loss of the physiologic warning signs is thought to stem from dampening and eventual loss of the neuroadrenal response to low glucose levels in the brain (FIGURE 1). A vicious cycle is set up, whereby reduction in the neurogenic response attenuates hypoglycemic symptoms, causing more episodes to occur and become more severe as they are repeated. This cycle can be broken, and the normal physiologic response restored, if hypoglycemic events can be avoided for just a few weeks.^2,26

Key points for patients. The main strategy for managing hypoglycemia is educating patients about the early symptoms of hypoglycemia and how to self-treat when they occur. Reinforce the need to time meals consistently and to limit the amount of carbohydrate ingested.

Advise patients to monitor blood glucose levels frequently, and to learn to relate a drop in glucose levels to hypoglycemic symptoms.²

Counsel patients to eat a snack or, preferably, drink fruit juice to counteract hypoglycemia. Patients may also carry glucose tablets, which are convenient and less tempting than candy.²⁷

Glucagon is indicated for severe cases.

Whenever possible, a patient’s family members (particularly in the case of children) should be educated too.

Ask regularly about episodes. Finally, act to identify problems by querying patients and family members at every visit about hypoglycemia episodes, probing for information about awareness, severity and timing of the episodes, and how these events relate to dosing, meals, and activities.²⁷ If hypoglycemia recurs, analyze the dosing regimen and consider flexible insulin dosing.¹

Anti-hypoglycemia strategies for each new phase of therapy

As the course of diabetes therapy moves, typically, from oral medications to insulin to combination regimens, drug selections can be made in part to reduce the risk of hypoglycemia.

Oral agents: Insulin sensitizers, newer agents generally better

As noted earlier, among oral agents, insulin sensitizers are generally thought to have lower rates of hypoglycemia.

Newer sulfonylureas such as glimepiride and the rapid-acting meglitinides may also cause fewer hypoglycemic events.

Given the progressive decline of endogenous insulin secretion, combination therapy with secretagogues or insulin is eventually required for most patients.

Insulin analogs

A number of rapid-, short-, intermediateand long-acting insulin analogs have been introduced, and many of them make it possible to mimic different phases of physiologic insulin secretion (FIGURE 2). One of the newer analogs less likely to cause hypoglycemia is glargine, a long-acting insulin with a steady, relatively consistent action profile over a 24-hour period, closely mimicking normal basal pancreatic insulin secretion.²⁸

FIGURE 2
Plasma insulin levels with newer analogs

TABLE 2
Strategies for avoiding and addressing hypoglycemia

Insulin mixtures helpful when meal times guaranteed

Other insulins include mixtures of regular insulin and long-acting insulin available in split mixed or premixed formulations. These mixtures are intended to cover insulin peaks at mealtimes with twice-daily administration.

Mixed insulin formulations are often perceived as relatively convenient and simple to use, but they require meals to be taken within set time frames, without a great deal of flexibility. Since the ratios of the insulin components are fixed, and designed to work with meals consumed on a fixed schedule, hypoglycemia can occur if patients miss a meal. In addition, the time-activity profile of the insulin may not match the postprandial glucose peak even if the meal is consumed, and will increase the chance of postprandial hypoglycemia.

Basal insulin plus oral regimens

For patients with type 2 diabetes, adding basal insulin to oral regimens can significantly improve glycemic control. Ideally, basal insulin therapy provides a sustained and relatively constant concentration of insulin throughout the day. In the past, neutral protamine Hagedorn (NPH) insulin, a longer-acting insulin, was used for basal insulin therapy, and regular insulin was used to cover prandial insulin needs. Ultralente, also used as a basal insulin, has a relatively unpredictable timeactivity profile.²⁸

Insulin glargine superior to NPH. In a recent clinical trial, patients with type 2 diabetes whose glucose levels were inadequately controlled on oral antidiabetic medications were given bedtime insulin glargine or NPH insulin.²⁹ The insulin doses were titrated using a simple algorithm targeting a fasting plasma glucose (FPG) of ≤100 mg/dL to reach recommended glycosylated hemoglobin (Hb A_1c) levels. Though no significant difference in glycemic control was found between insulin glargine and NPH insulin, significantly fewer hypoglycemic episodes occurred with insulin glargine therapy. The 24-hour distribution per patient-year of hypoglycemia for glargine vs NPH is shown in FIGURE 3.

Specifically, nearly 25% more patients treated with insulin glargine than with NPH insulin reached target Hb A_1c levels of ≤7.0% without nocturnal hypoglycemia. Moreover, the overall incidence of any hypoglycemic event (eg, plasma-referenced glucose ≤72 mg/dL) and severe hypoglycemia (eg, patient required assistance of another person, and had a glucose level <56 mg/dL or prompt recovery after glucose or glucagons) was lower with insulin glargine than with NPH insulin. Results from other studies and a recent metaanalysis have been similar.^30-32

Thus, using insulin glargine as basal insulin allows patients to reach recommended targets with fewer episodes of hypoglycemia, and can help address patients’ fear that can be a barrier to initiating insulin therapy in type 2 diabetes. Two recent studies have reported that dosing of insulin glargine can be flexible—morning or bedtime administration yields comparable low rates of hypoglycemia.^30,31

FIGURE 3
Hourly hypoglycemia rate with glargine much less than with NPH

Basal insulin plus prandial insulin

For patients who cannot otherwise reach Hb A_1c goals, basal insulin therapy may be supplemented with prandial insulin. Newer, rapid-acting analogs used for the prandial component are insulin lispro, insulin glulisine, or insulin aspart. Although this approach is physiologically more rational than regimens using conventional insulins, data are limited for use in type 2 diabetes.

The incidence of nocturnal hypoglycemia was evaluated in a study of patients with type 1 diabetes and impaired hypoglycemic awareness who were treated with 1 of 2 regimens: insulin lispro in a basal-prandial combination with NPH insulin, or twice-daily, premixed NPH/regular insulin.³³ Results showed that the incidence of nocturnal hypoglycemia was lower in patients receiving the insulin lispro regimen.

Another study, comparing insulin aspart and regular insulin as the prandial component in a basal-prandial regimen with NPH, showed that postprandial glucose control and Hb A_1c levels were significantly better after 1 year of treatment in the insulin aspart group than in the group receiving regular insulin, without an increased risk for hypoglycemia.³⁴ These results suggest that treatment with rapidacting insulin analogs could be helpful in avoiding hypoglycemia in patients with type 2 diabetes when a basal-prandial insulin regimen is indicated.

Avoiding hypoglycemia in the elderly

Elderly patients may be at increased risk for iatrogenic hypoglycemia. A populationbased study of patients presenting to an emergency room with severe hypoglycemic symptoms reported that rates of such events among elderly patients with type 2 diabetes and multiple comorbidities approached the rates among patients with type 1 diabetes.³⁵

Creatinine clearance is often decreased in elderly patients, slowing elimination of oral agents and insulin and potentially resulting in sustained pharmacological action and creating a greater risk for hypoglycemia.

Furthermore, there is evidence that the neurogenic symptoms of hypoglycemia are reduced in elderly patients, diminishing awareness of hypoglycemia.³⁶

In the demented elderly, malnutrition, weight loss, and anorexia may exacerbate the risk for hypoglycemia. For elderly patients with tertiary disease (eg, cerebrovascular accident, myocardial infarction, congestive heart failure, blindness, chronic renal failure), the risk for hypoglycemia and subsequent comorbidity may outweigh the benefits of strict glycemic control.^3,37 Elderly patients may have comorbid conditions that increase risk of falls (eg, poor vision, neurologic conditions), and hypoglycemic episodes may further increase the risk of falls and lead to morbidity (eg, fragility fracture in patients with osteopenia or osteoporosis).

Because the elderly are at a greater risk for hypoglycemia, a switch to a less restrictive diet, such as a “no concentrated sweets” diet, is an option, with control of glucose levels through the administration of oral agents or insulin.³⁶ This may also promote a better quality of life, considering that many of these patients already have secondary and tertiary complications of diabetes, prevention of which is not a realistic goal.

CORRESPONDING AUTHOR
William Cefalu, MD, Professor and Chief, Division of Nutrition and Chronic Diseases, Pennington Biomedical Research Center, Louisiana State University, 6400 Perkins Road, Baton Rouge, LA 70808. Email: [email protected]

Signs and symptoms of hypoglycemia vary considerably among patients with type 2 diabetes, making the condition easy to miss. Moreover, the most common symptoms are not necessarily the first symptoms.

If hypoglycemia occurs repeatedly, it can start a vicious cycle of physiologic reactions that mask or diminish the symptoms that warn patients of an impending episode. This may lead to hypoglycemia unawareness and hypoglycemic episodes of increasing severity. Fear of hypoglycemia, particularly of nocturnal events, may discourage patients from more intensive glycemic control, particularly using insulin. Such fear may even lead them to reduce their antidiabetic medication dosage, resulting in poor glycemic control.¹

Breaking this cycle and restoring normal physiologic responses is one focus of this article, as is teaching patients how to monitor their blood glucose levels and how to correlate low blood glucose with the signs and symptoms of hypoglycemia.

Other therapies and strategies that we discuss in this article:

• Newer insulin analogs and the associated risk of hypoglycemia with each
• Appropriate combination of insulin with oral antidiabetic medications
• The long-acting analog insulin glargine used as basal insulin to lower the incidence of hypoglycemia, including nocturnal and severe hypoglycemia
• Rapid-acting insulin analogs (aspart, glulisine, and lispro) used in basalprandial insulin regimens.

First symptoms vary among individuals

Symptoms of hypoglycemia result primarily from a lowered glucose level in the brain and its effects on the central and autonomic nervous systems (FIGURE 1). A decrease in glucose below physiologic levels has acute consequences for brain function because the brain has an immediate requirement for glucose and little capacity for storage.

Two types of symptoms

Neuroglycopenia and the inhibition of neuronal metabolism causes sensations of warmth, weakness, fatigue, difficulty concentrating, confusion, behavioral changes, and in the most severe cases, a loss of consciousness, seizures, brain damage, and even death.^2-4

Neurogenic symptoms are mediated by the hormones and neurotransmitters secreted in response to low brain glucose levels (FIGURE 1). The gluconeogenic actions of the autonomic nervous system produce the classic warning symptoms—tremulousness, pounding heart, anxiety, sweating, hunger, and tingling sensations—that usually precede the symptoms of hypoglycemia.^2-4 This is particularly so in iatrogenic hypoglycemia.

These direct symptoms of neuroglycopenia are the ones patients typically identify with hypoglycemia. The most common symptoms of hypoglycemia are therefore not necessarily the first symptoms of hypoglycemia (TABLE 1).⁵ For example, though most patients experience sweating as a symptom of hypoglycemia, the first symptom might be trembling or anxiety, depending on the individual.⁵

TABLE 1
Signs and symptoms most commonly associated with hypoglycemia are not always the first to appear

Factors influencing frequency and severity of hypoglycemia

Aggressive diabetes management commonly causes mild-to-moderate hypoglycemia, defined as a blood glucose value <60 mg/dL, that can be managed by the patient without assistance.

Severe hypoglycemia—a blood glucose value <50 mg/dL—is relatively uncommon in type 2 diabetes and requires the assistance of another person to manage, since neurological impairment may render patients unable to treat themselves.^2,6 Severe hypoglycemia, whether in patients with type 1 or type 2 diabetes, can have debilitating consequences, including seizures or coma or even death.⁷

Long-standing type 2 disease. Hypoglycemia is more common in patients with type 1 diabetes than in those with type 2, but it can occur in type 2 diabetes patients who require insulin or are treated intensively with combinations of oral agents.⁶ As type 2 diabetes progresses,⁸ the incidence of hypoglycemic events increases, as endogenously produced insulin declines and is replaced by exogenous insulin.^5,9 In fact, the prevalence of severe episodes (eg, requiring assistance of another person to administer glucose or glucagon) in patients with type 2 diabetes was comparable to that exhibited among patients with type 1 diabetes if they had been on insulin therapy for the same length of time.^5,10

Nocturnal hypoglycemia. This event poses a special concern because the warning signs of hypoglycemia may be blunted during sleep. It has been reported that as many as 29% to 56% of all adult patients treated with insulin have an overnight glucose profile that indicates hypoglycemia occurs at night.^11-13 However, it is important to note that the extent of the problem of nocturnal hypoglycemia is difficult to assess since overnight monitoring of glucose levels is required.

Additional insights from the UKPDS

Hypoglycemia in type 2 diabetes has not received rigorous attention in clinical trials. However, the United Kingdom Prospective Diabetes Study (UKPDS) was a large longitudinal trial in type 2 diabetes that included hypoglycemia as an outcome measure and thus provides some helpful information.

Events with insulin>sulfonylureas>diet. The 6-year follow-up revealed that the cumulative proportion of patients reporting 1 or more hypoglycemic events (of any type) was 76% for those using insulin, 45% among those taking sulfonylureas, and 3% for those on diet alone. Expressed as events per patient year, this was 37%, 17%, and 0.9%, respectively. When only “major” events (those requiring third-party assistance or hospital admission) were considered, the proportion of patients per year reporting 1 or more such events, was 2.3% for insulin, 0.7% for sulfonylureas, and 0.03% for diet alone. The cumulative proportion over 6 years was 3.3% of participants using sulfonylureas, 11.2% of those using insulin, and 0.15% of those on diet therapy.⁸

Metformin increases risk. The cumulative proportion of obese patients reporting any hypoglycemic event was 17.6% for those taking metformin vs 2.8% for those on diet. Severe hypoglycemia (as defined earlier) occurred in 2.4% of participants using metformin compared with 0.4% of those on diet therapy.⁸

Findings from other studies

Interestingly, in a recent systematic review of randomized controlled trials comparing insulin monotherapy with insulin plus oral antidiabetic agents, 13 of 14 studies reporting hypoglycemia demonstrated no difference in events.¹⁴

The occurrence of hypoglycemia among patients on metformin monotherapy in the UKPDS study is notable since, theoretically, hypoglycemia should not occur with agents whose mechanisms of action do not increase insulin secretion (biguanides, thiazolidinediones [TZDs], or α-glucosidase inhibitors),¹ Since newer classes such as TZDs, α-glucosidase inhibitors, and meglitinides were not available when UKPDS was initiated, the trial does not provide data on these classes.

In a small comparative study of insulin combined with either metformin or a TZD, it appeared that metformin combination was associated with fewer occurrences of hypoglycemia; however, the small patient sample limits generalizability of the finding.¹⁵

FIGURE 1
How hypoglycemia occurs

Normally, as blood glucose levels (red) rise, insulin secretion increases, circulating insulin levels (blue) rise, and hepatic glucose production is inhibited. As glucose is disposed and circulating levels decrease due to insulin action, insulin levels then drop and hepatic glucose production begins again.

Hypoglycemia in diabetes can result from an excess of endogenous or exogenous insulin (iatrogenic hypoglycemia). In healthy patients, high insulin and falling glucose levels suppress insulin production and stimulate a hormone-mediated burst of glucose production. In patients with diabetes, the loss of physiologic control of insulin secretion coupled with exogenous administration of insulin or insulin secretagogues can interfere with the normal physiologic response to low blood glucose levels, resulting in hypoglycemia.¹

With intense insulin regimens, the incidence of hypoglycemia can be as high as 30%, in contrast to 12% for patients treated with diet alone and 16% for those taking oral agents.⁶

With secretagogues, it has been suggested that the incidence of hypoglycemia is higher with the older, longer acting sulfonylurea agents.^1,6,16-19 Although populationbased data on hypoglycemic rates associated with combination therapy with oral antidiabetic agents are not available, numerous clinical studies have reported rates of 10% to 20% for any hypoglycemic event.^20-23

Heightening patient awareness, and yours

Because the signs of hypoglycemia vary considerably among individuals, they can easily be missed.³ In addition, repeated episodes of hypoglycemia can alter the normal regulatory responses and diminish the most important signs of a drop in glucose levels.^1,2,24,25 The loss of the physiologic warning signs is thought to stem from dampening and eventual loss of the neuroadrenal response to low glucose levels in the brain (FIGURE 1). A vicious cycle is set up, whereby reduction in the neurogenic response attenuates hypoglycemic symptoms, causing more episodes to occur and become more severe as they are repeated. This cycle can be broken, and the normal physiologic response restored, if hypoglycemic events can be avoided for just a few weeks.^2,26

Key points for patients. The main strategy for managing hypoglycemia is educating patients about the early symptoms of hypoglycemia and how to self-treat when they occur. Reinforce the need to time meals consistently and to limit the amount of carbohydrate ingested.

Advise patients to monitor blood glucose levels frequently, and to learn to relate a drop in glucose levels to hypoglycemic symptoms.²

Counsel patients to eat a snack or, preferably, drink fruit juice to counteract hypoglycemia. Patients may also carry glucose tablets, which are convenient and less tempting than candy.²⁷

Glucagon is indicated for severe cases.

Whenever possible, a patient’s family members (particularly in the case of children) should be educated too.

Ask regularly about episodes. Finally, act to identify problems by querying patients and family members at every visit about hypoglycemia episodes, probing for information about awareness, severity and timing of the episodes, and how these events relate to dosing, meals, and activities.²⁷ If hypoglycemia recurs, analyze the dosing regimen and consider flexible insulin dosing.¹

Anti-hypoglycemia strategies for each new phase of therapy

As the course of diabetes therapy moves, typically, from oral medications to insulin to combination regimens, drug selections can be made in part to reduce the risk of hypoglycemia.

Oral agents: Insulin sensitizers, newer agents generally better

As noted earlier, among oral agents, insulin sensitizers are generally thought to have lower rates of hypoglycemia.

Newer sulfonylureas such as glimepiride and the rapid-acting meglitinides may also cause fewer hypoglycemic events.

Given the progressive decline of endogenous insulin secretion, combination therapy with secretagogues or insulin is eventually required for most patients.

Insulin analogs

A number of rapid-, short-, intermediateand long-acting insulin analogs have been introduced, and many of them make it possible to mimic different phases of physiologic insulin secretion (FIGURE 2). One of the newer analogs less likely to cause hypoglycemia is glargine, a long-acting insulin with a steady, relatively consistent action profile over a 24-hour period, closely mimicking normal basal pancreatic insulin secretion.²⁸

FIGURE 2
Plasma insulin levels with newer analogs

TABLE 2
Strategies for avoiding and addressing hypoglycemia

Insulin mixtures helpful when meal times guaranteed

Other insulins include mixtures of regular insulin and long-acting insulin available in split mixed or premixed formulations. These mixtures are intended to cover insulin peaks at mealtimes with twice-daily administration.

Mixed insulin formulations are often perceived as relatively convenient and simple to use, but they require meals to be taken within set time frames, without a great deal of flexibility. Since the ratios of the insulin components are fixed, and designed to work with meals consumed on a fixed schedule, hypoglycemia can occur if patients miss a meal. In addition, the time-activity profile of the insulin may not match the postprandial glucose peak even if the meal is consumed, and will increase the chance of postprandial hypoglycemia.

Basal insulin plus oral regimens

For patients with type 2 diabetes, adding basal insulin to oral regimens can significantly improve glycemic control. Ideally, basal insulin therapy provides a sustained and relatively constant concentration of insulin throughout the day. In the past, neutral protamine Hagedorn (NPH) insulin, a longer-acting insulin, was used for basal insulin therapy, and regular insulin was used to cover prandial insulin needs. Ultralente, also used as a basal insulin, has a relatively unpredictable timeactivity profile.²⁸

Insulin glargine superior to NPH. In a recent clinical trial, patients with type 2 diabetes whose glucose levels were inadequately controlled on oral antidiabetic medications were given bedtime insulin glargine or NPH insulin.²⁹ The insulin doses were titrated using a simple algorithm targeting a fasting plasma glucose (FPG) of ≤100 mg/dL to reach recommended glycosylated hemoglobin (Hb A_1c) levels. Though no significant difference in glycemic control was found between insulin glargine and NPH insulin, significantly fewer hypoglycemic episodes occurred with insulin glargine therapy. The 24-hour distribution per patient-year of hypoglycemia for glargine vs NPH is shown in FIGURE 3.

Specifically, nearly 25% more patients treated with insulin glargine than with NPH insulin reached target Hb A_1c levels of ≤7.0% without nocturnal hypoglycemia. Moreover, the overall incidence of any hypoglycemic event (eg, plasma-referenced glucose ≤72 mg/dL) and severe hypoglycemia (eg, patient required assistance of another person, and had a glucose level <56 mg/dL or prompt recovery after glucose or glucagons) was lower with insulin glargine than with NPH insulin. Results from other studies and a recent metaanalysis have been similar.^30-32

Thus, using insulin glargine as basal insulin allows patients to reach recommended targets with fewer episodes of hypoglycemia, and can help address patients’ fear that can be a barrier to initiating insulin therapy in type 2 diabetes. Two recent studies have reported that dosing of insulin glargine can be flexible—morning or bedtime administration yields comparable low rates of hypoglycemia.^30,31

FIGURE 3
Hourly hypoglycemia rate with glargine much less than with NPH

Basal insulin plus prandial insulin

For patients who cannot otherwise reach Hb A_1c goals, basal insulin therapy may be supplemented with prandial insulin. Newer, rapid-acting analogs used for the prandial component are insulin lispro, insulin glulisine, or insulin aspart. Although this approach is physiologically more rational than regimens using conventional insulins, data are limited for use in type 2 diabetes.

The incidence of nocturnal hypoglycemia was evaluated in a study of patients with type 1 diabetes and impaired hypoglycemic awareness who were treated with 1 of 2 regimens: insulin lispro in a basal-prandial combination with NPH insulin, or twice-daily, premixed NPH/regular insulin.³³ Results showed that the incidence of nocturnal hypoglycemia was lower in patients receiving the insulin lispro regimen.

Another study, comparing insulin aspart and regular insulin as the prandial component in a basal-prandial regimen with NPH, showed that postprandial glucose control and Hb A_1c levels were significantly better after 1 year of treatment in the insulin aspart group than in the group receiving regular insulin, without an increased risk for hypoglycemia.³⁴ These results suggest that treatment with rapidacting insulin analogs could be helpful in avoiding hypoglycemia in patients with type 2 diabetes when a basal-prandial insulin regimen is indicated.

Avoiding hypoglycemia in the elderly

Elderly patients may be at increased risk for iatrogenic hypoglycemia. A populationbased study of patients presenting to an emergency room with severe hypoglycemic symptoms reported that rates of such events among elderly patients with type 2 diabetes and multiple comorbidities approached the rates among patients with type 1 diabetes.³⁵

Creatinine clearance is often decreased in elderly patients, slowing elimination of oral agents and insulin and potentially resulting in sustained pharmacological action and creating a greater risk for hypoglycemia.

Furthermore, there is evidence that the neurogenic symptoms of hypoglycemia are reduced in elderly patients, diminishing awareness of hypoglycemia.³⁶

In the demented elderly, malnutrition, weight loss, and anorexia may exacerbate the risk for hypoglycemia. For elderly patients with tertiary disease (eg, cerebrovascular accident, myocardial infarction, congestive heart failure, blindness, chronic renal failure), the risk for hypoglycemia and subsequent comorbidity may outweigh the benefits of strict glycemic control.^3,37 Elderly patients may have comorbid conditions that increase risk of falls (eg, poor vision, neurologic conditions), and hypoglycemic episodes may further increase the risk of falls and lead to morbidity (eg, fragility fracture in patients with osteopenia or osteoporosis).

Because the elderly are at a greater risk for hypoglycemia, a switch to a less restrictive diet, such as a “no concentrated sweets” diet, is an option, with control of glucose levels through the administration of oral agents or insulin.³⁶ This may also promote a better quality of life, considering that many of these patients already have secondary and tertiary complications of diabetes, prevention of which is not a realistic goal.

CORRESPONDING AUTHOR
William Cefalu, MD, Professor and Chief, Division of Nutrition and Chronic Diseases, Pennington Biomedical Research Center, Louisiana State University, 6400 Perkins Road, Baton Rouge, LA 70808. Email: [email protected]

Signs and symptoms of hypoglycemia vary considerably among patients with type 2 diabetes, making the condition easy to miss. Moreover, the most common symptoms are not necessarily the first symptoms.

If hypoglycemia occurs repeatedly, it can start a vicious cycle of physiologic reactions that mask or diminish the symptoms that warn patients of an impending episode. This may lead to hypoglycemia unawareness and hypoglycemic episodes of increasing severity. Fear of hypoglycemia, particularly of nocturnal events, may discourage patients from more intensive glycemic control, particularly using insulin. Such fear may even lead them to reduce their antidiabetic medication dosage, resulting in poor glycemic control.¹

Breaking this cycle and restoring normal physiologic responses is one focus of this article, as is teaching patients how to monitor their blood glucose levels and how to correlate low blood glucose with the signs and symptoms of hypoglycemia.

Other therapies and strategies that we discuss in this article:

• Newer insulin analogs and the associated risk of hypoglycemia with each
• Appropriate combination of insulin with oral antidiabetic medications
• The long-acting analog insulin glargine used as basal insulin to lower the incidence of hypoglycemia, including nocturnal and severe hypoglycemia
• Rapid-acting insulin analogs (aspart, glulisine, and lispro) used in basalprandial insulin regimens.

First symptoms vary among individuals

Symptoms of hypoglycemia result primarily from a lowered glucose level in the brain and its effects on the central and autonomic nervous systems (FIGURE 1). A decrease in glucose below physiologic levels has acute consequences for brain function because the brain has an immediate requirement for glucose and little capacity for storage.

Two types of symptoms

Neuroglycopenia and the inhibition of neuronal metabolism causes sensations of warmth, weakness, fatigue, difficulty concentrating, confusion, behavioral changes, and in the most severe cases, a loss of consciousness, seizures, brain damage, and even death.^2-4

Neurogenic symptoms are mediated by the hormones and neurotransmitters secreted in response to low brain glucose levels (FIGURE 1). The gluconeogenic actions of the autonomic nervous system produce the classic warning symptoms—tremulousness, pounding heart, anxiety, sweating, hunger, and tingling sensations—that usually precede the symptoms of hypoglycemia.^2-4 This is particularly so in iatrogenic hypoglycemia.

These direct symptoms of neuroglycopenia are the ones patients typically identify with hypoglycemia. The most common symptoms of hypoglycemia are therefore not necessarily the first symptoms of hypoglycemia (TABLE 1).⁵ For example, though most patients experience sweating as a symptom of hypoglycemia, the first symptom might be trembling or anxiety, depending on the individual.⁵

TABLE 1
Signs and symptoms most commonly associated with hypoglycemia are not always the first to appear

Factors influencing frequency and severity of hypoglycemia

Aggressive diabetes management commonly causes mild-to-moderate hypoglycemia, defined as a blood glucose value <60 mg/dL, that can be managed by the patient without assistance.

Severe hypoglycemia—a blood glucose value <50 mg/dL—is relatively uncommon in type 2 diabetes and requires the assistance of another person to manage, since neurological impairment may render patients unable to treat themselves.^2,6 Severe hypoglycemia, whether in patients with type 1 or type 2 diabetes, can have debilitating consequences, including seizures or coma or even death.⁷

Long-standing type 2 disease. Hypoglycemia is more common in patients with type 1 diabetes than in those with type 2, but it can occur in type 2 diabetes patients who require insulin or are treated intensively with combinations of oral agents.⁶ As type 2 diabetes progresses,⁸ the incidence of hypoglycemic events increases, as endogenously produced insulin declines and is replaced by exogenous insulin.^5,9 In fact, the prevalence of severe episodes (eg, requiring assistance of another person to administer glucose or glucagon) in patients with type 2 diabetes was comparable to that exhibited among patients with type 1 diabetes if they had been on insulin therapy for the same length of time.^5,10

Nocturnal hypoglycemia. This event poses a special concern because the warning signs of hypoglycemia may be blunted during sleep. It has been reported that as many as 29% to 56% of all adult patients treated with insulin have an overnight glucose profile that indicates hypoglycemia occurs at night.^11-13 However, it is important to note that the extent of the problem of nocturnal hypoglycemia is difficult to assess since overnight monitoring of glucose levels is required.

Additional insights from the UKPDS

Hypoglycemia in type 2 diabetes has not received rigorous attention in clinical trials. However, the United Kingdom Prospective Diabetes Study (UKPDS) was a large longitudinal trial in type 2 diabetes that included hypoglycemia as an outcome measure and thus provides some helpful information.

Events with insulin>sulfonylureas>diet. The 6-year follow-up revealed that the cumulative proportion of patients reporting 1 or more hypoglycemic events (of any type) was 76% for those using insulin, 45% among those taking sulfonylureas, and 3% for those on diet alone. Expressed as events per patient year, this was 37%, 17%, and 0.9%, respectively. When only “major” events (those requiring third-party assistance or hospital admission) were considered, the proportion of patients per year reporting 1 or more such events, was 2.3% for insulin, 0.7% for sulfonylureas, and 0.03% for diet alone. The cumulative proportion over 6 years was 3.3% of participants using sulfonylureas, 11.2% of those using insulin, and 0.15% of those on diet therapy.⁸

Metformin increases risk. The cumulative proportion of obese patients reporting any hypoglycemic event was 17.6% for those taking metformin vs 2.8% for those on diet. Severe hypoglycemia (as defined earlier) occurred in 2.4% of participants using metformin compared with 0.4% of those on diet therapy.⁸

Findings from other studies

Interestingly, in a recent systematic review of randomized controlled trials comparing insulin monotherapy with insulin plus oral antidiabetic agents, 13 of 14 studies reporting hypoglycemia demonstrated no difference in events.¹⁴

The occurrence of hypoglycemia among patients on metformin monotherapy in the UKPDS study is notable since, theoretically, hypoglycemia should not occur with agents whose mechanisms of action do not increase insulin secretion (biguanides, thiazolidinediones [TZDs], or α-glucosidase inhibitors),¹ Since newer classes such as TZDs, α-glucosidase inhibitors, and meglitinides were not available when UKPDS was initiated, the trial does not provide data on these classes.

In a small comparative study of insulin combined with either metformin or a TZD, it appeared that metformin combination was associated with fewer occurrences of hypoglycemia; however, the small patient sample limits generalizability of the finding.¹⁵

FIGURE 1
How hypoglycemia occurs

Normally, as blood glucose levels (red) rise, insulin secretion increases, circulating insulin levels (blue) rise, and hepatic glucose production is inhibited. As glucose is disposed and circulating levels decrease due to insulin action, insulin levels then drop and hepatic glucose production begins again.

Hypoglycemia in diabetes can result from an excess of endogenous or exogenous insulin (iatrogenic hypoglycemia). In healthy patients, high insulin and falling glucose levels suppress insulin production and stimulate a hormone-mediated burst of glucose production. In patients with diabetes, the loss of physiologic control of insulin secretion coupled with exogenous administration of insulin or insulin secretagogues can interfere with the normal physiologic response to low blood glucose levels, resulting in hypoglycemia.¹

With intense insulin regimens, the incidence of hypoglycemia can be as high as 30%, in contrast to 12% for patients treated with diet alone and 16% for those taking oral agents.⁶

With secretagogues, it has been suggested that the incidence of hypoglycemia is higher with the older, longer acting sulfonylurea agents.^1,6,16-19 Although populationbased data on hypoglycemic rates associated with combination therapy with oral antidiabetic agents are not available, numerous clinical studies have reported rates of 10% to 20% for any hypoglycemic event.^20-23

Heightening patient awareness, and yours

Because the signs of hypoglycemia vary considerably among individuals, they can easily be missed.³ In addition, repeated episodes of hypoglycemia can alter the normal regulatory responses and diminish the most important signs of a drop in glucose levels.^1,2,24,25 The loss of the physiologic warning signs is thought to stem from dampening and eventual loss of the neuroadrenal response to low glucose levels in the brain (FIGURE 1). A vicious cycle is set up, whereby reduction in the neurogenic response attenuates hypoglycemic symptoms, causing more episodes to occur and become more severe as they are repeated. This cycle can be broken, and the normal physiologic response restored, if hypoglycemic events can be avoided for just a few weeks.^2,26

Key points for patients. The main strategy for managing hypoglycemia is educating patients about the early symptoms of hypoglycemia and how to self-treat when they occur. Reinforce the need to time meals consistently and to limit the amount of carbohydrate ingested.

Advise patients to monitor blood glucose levels frequently, and to learn to relate a drop in glucose levels to hypoglycemic symptoms.²

Counsel patients to eat a snack or, preferably, drink fruit juice to counteract hypoglycemia. Patients may also carry glucose tablets, which are convenient and less tempting than candy.²⁷

Glucagon is indicated for severe cases.

Whenever possible, a patient’s family members (particularly in the case of children) should be educated too.

Ask regularly about episodes. Finally, act to identify problems by querying patients and family members at every visit about hypoglycemia episodes, probing for information about awareness, severity and timing of the episodes, and how these events relate to dosing, meals, and activities.²⁷ If hypoglycemia recurs, analyze the dosing regimen and consider flexible insulin dosing.¹

Anti-hypoglycemia strategies for each new phase of therapy

As the course of diabetes therapy moves, typically, from oral medications to insulin to combination regimens, drug selections can be made in part to reduce the risk of hypoglycemia.

Oral agents: Insulin sensitizers, newer agents generally better

As noted earlier, among oral agents, insulin sensitizers are generally thought to have lower rates of hypoglycemia.

Newer sulfonylureas such as glimepiride and the rapid-acting meglitinides may also cause fewer hypoglycemic events.

Given the progressive decline of endogenous insulin secretion, combination therapy with secretagogues or insulin is eventually required for most patients.

Insulin analogs

A number of rapid-, short-, intermediateand long-acting insulin analogs have been introduced, and many of them make it possible to mimic different phases of physiologic insulin secretion (FIGURE 2). One of the newer analogs less likely to cause hypoglycemia is glargine, a long-acting insulin with a steady, relatively consistent action profile over a 24-hour period, closely mimicking normal basal pancreatic insulin secretion.²⁸

FIGURE 2
Plasma insulin levels with newer analogs

TABLE 2
Strategies for avoiding and addressing hypoglycemia

Insulin mixtures helpful when meal times guaranteed

Other insulins include mixtures of regular insulin and long-acting insulin available in split mixed or premixed formulations. These mixtures are intended to cover insulin peaks at mealtimes with twice-daily administration.

Mixed insulin formulations are often perceived as relatively convenient and simple to use, but they require meals to be taken within set time frames, without a great deal of flexibility. Since the ratios of the insulin components are fixed, and designed to work with meals consumed on a fixed schedule, hypoglycemia can occur if patients miss a meal. In addition, the time-activity profile of the insulin may not match the postprandial glucose peak even if the meal is consumed, and will increase the chance of postprandial hypoglycemia.

Basal insulin plus oral regimens

For patients with type 2 diabetes, adding basal insulin to oral regimens can significantly improve glycemic control. Ideally, basal insulin therapy provides a sustained and relatively constant concentration of insulin throughout the day. In the past, neutral protamine Hagedorn (NPH) insulin, a longer-acting insulin, was used for basal insulin therapy, and regular insulin was used to cover prandial insulin needs. Ultralente, also used as a basal insulin, has a relatively unpredictable timeactivity profile.²⁸

Insulin glargine superior to NPH. In a recent clinical trial, patients with type 2 diabetes whose glucose levels were inadequately controlled on oral antidiabetic medications were given bedtime insulin glargine or NPH insulin.²⁹ The insulin doses were titrated using a simple algorithm targeting a fasting plasma glucose (FPG) of ≤100 mg/dL to reach recommended glycosylated hemoglobin (Hb A_1c) levels. Though no significant difference in glycemic control was found between insulin glargine and NPH insulin, significantly fewer hypoglycemic episodes occurred with insulin glargine therapy. The 24-hour distribution per patient-year of hypoglycemia for glargine vs NPH is shown in FIGURE 3.

Specifically, nearly 25% more patients treated with insulin glargine than with NPH insulin reached target Hb A_1c levels of ≤7.0% without nocturnal hypoglycemia. Moreover, the overall incidence of any hypoglycemic event (eg, plasma-referenced glucose ≤72 mg/dL) and severe hypoglycemia (eg, patient required assistance of another person, and had a glucose level <56 mg/dL or prompt recovery after glucose or glucagons) was lower with insulin glargine than with NPH insulin. Results from other studies and a recent metaanalysis have been similar.^30-32

Thus, using insulin glargine as basal insulin allows patients to reach recommended targets with fewer episodes of hypoglycemia, and can help address patients’ fear that can be a barrier to initiating insulin therapy in type 2 diabetes. Two recent studies have reported that dosing of insulin glargine can be flexible—morning or bedtime administration yields comparable low rates of hypoglycemia.^30,31

FIGURE 3
Hourly hypoglycemia rate with glargine much less than with NPH

Basal insulin plus prandial insulin

For patients who cannot otherwise reach Hb A_1c goals, basal insulin therapy may be supplemented with prandial insulin. Newer, rapid-acting analogs used for the prandial component are insulin lispro, insulin glulisine, or insulin aspart. Although this approach is physiologically more rational than regimens using conventional insulins, data are limited for use in type 2 diabetes.

The incidence of nocturnal hypoglycemia was evaluated in a study of patients with type 1 diabetes and impaired hypoglycemic awareness who were treated with 1 of 2 regimens: insulin lispro in a basal-prandial combination with NPH insulin, or twice-daily, premixed NPH/regular insulin.³³ Results showed that the incidence of nocturnal hypoglycemia was lower in patients receiving the insulin lispro regimen.

Another study, comparing insulin aspart and regular insulin as the prandial component in a basal-prandial regimen with NPH, showed that postprandial glucose control and Hb A_1c levels were significantly better after 1 year of treatment in the insulin aspart group than in the group receiving regular insulin, without an increased risk for hypoglycemia.³⁴ These results suggest that treatment with rapidacting insulin analogs could be helpful in avoiding hypoglycemia in patients with type 2 diabetes when a basal-prandial insulin regimen is indicated.

Avoiding hypoglycemia in the elderly

Elderly patients may be at increased risk for iatrogenic hypoglycemia. A populationbased study of patients presenting to an emergency room with severe hypoglycemic symptoms reported that rates of such events among elderly patients with type 2 diabetes and multiple comorbidities approached the rates among patients with type 1 diabetes.³⁵

Creatinine clearance is often decreased in elderly patients, slowing elimination of oral agents and insulin and potentially resulting in sustained pharmacological action and creating a greater risk for hypoglycemia.

Furthermore, there is evidence that the neurogenic symptoms of hypoglycemia are reduced in elderly patients, diminishing awareness of hypoglycemia.³⁶

In the demented elderly, malnutrition, weight loss, and anorexia may exacerbate the risk for hypoglycemia. For elderly patients with tertiary disease (eg, cerebrovascular accident, myocardial infarction, congestive heart failure, blindness, chronic renal failure), the risk for hypoglycemia and subsequent comorbidity may outweigh the benefits of strict glycemic control.^3,37 Elderly patients may have comorbid conditions that increase risk of falls (eg, poor vision, neurologic conditions), and hypoglycemic episodes may further increase the risk of falls and lead to morbidity (eg, fragility fracture in patients with osteopenia or osteoporosis).

Because the elderly are at a greater risk for hypoglycemia, a switch to a less restrictive diet, such as a “no concentrated sweets” diet, is an option, with control of glucose levels through the administration of oral agents or insulin.³⁶ This may also promote a better quality of life, considering that many of these patients already have secondary and tertiary complications of diabetes, prevention of which is not a realistic goal.

CORRESPONDING AUTHOR
William Cefalu, MD, Professor and Chief, Division of Nutrition and Chronic Diseases, Pennington Biomedical Research Center, Louisiana State University, 6400 Perkins Road, Baton Rouge, LA 70808. Email: [email protected]

Hyaluronic acid injections can help relieve pain for carefully selected patients with knee osteoarthritis. But this option should be reserved for those whose pain has not responded to adequate trials of systemic therapeutic agents (acetaminophen, nonsteroidal anti-inflammatory drugs [NSAIDs], cyclooxygenase-2 [COX-2] inhibitors), topical agents, or to lifestyle modifications such as weight reduction and exercise.

Hyaluronic acid injections may also be indicated when knee surgery must be delayed for middle-aged persons.¹

In spite of the Food and Drug Administration’s approval of this therapy, uncertainty about its efficacy exists among the medical community. A recent meta-analysis of the effectiveness of intra-articular hyaluronic acid for knee osteoarthritis that included 22 published and unpublished, English and non-English, single or double-blinded, randomized controlled trials in humans showed that hyaluronic acid has only a small effect on pain relief when compared with placebo.²

We provide here a stringent test of the efficacy of viscosupplementation for relieving knee pain from osteoarthritis with a meta-analysis that includes only data from randomized, double-blinded, controlled trials of hyaluronic acid that measured pain using a visual analogue scale (VAS), the most widely accepted method for pain evaluation.

Long time coming

Balazs first proposed hyaluronic acid as a treatment for patients with arthritic diseases in 1942. In the early 1970s, therapeutic studies were begun to test the efficacy of hyaluronic acid on knee osteoarthritis. The results were encouraging and side effects were few.²⁸ With the FDA’s approval in 1998, intra-articular “viscosupplementation” with hyaluronic acid—also called hyaluronan or hyaluronate, and the hylan derivatives of hyaluronic acid—is a welcome option for many of the 16 million older Americans with osteoarthritis of the knee.¹

Methods

Selection of studies

We identified clinical trials of viscosupplementation with hyaluronic acid in humans published in English from 1965 through August 2004 through a computerized literature search of Medline. The keyword used was “hyaluronic acid,” which was combined with “trial” or “osteoarthritis knee” or “viscosupplementation.” We conducted an additional manual search of the reference lists of included articles and review articles. We also searched the Cochrane Library and websites of the Agency for Healthcare Research and Quality (AHRQ) for information on hyaluronic acid in knee osteoarthritis. We identified 1872 articles with this search process.

Of the 1872 articles, we identified by title and abstract 33 that might be pertinent to this study, including 17 randomized trials. We excluded reviews, meta-analyses, comparison trials, and trials reporting VAS as part of the WOMAC (Western Ontario McMaster Universities Index) scale. We attempted to contact authors of the studies that used a double-blind, randomized controlled design, to obtain any data that may not have been included in the publications. Three authors provided the details requested; the others did not respond or stated that additional data were not available. Of the 17 randomized trials we identified, 8 were excluded because they were open, single-blinded, or did not use the VAS to measure pain outcomes.^3-10 The remaining 9 double-blinded, placebo controlled, randomized clinical trials of viscosupplementation with hyaluronic acid for knee osteoarthritis that did use a VAS to measure pain were included in this meta-analysis (TABLE 1). Because one of the studies (by Henderson) included 2 subgroups of pain severity, these were considered as 2 separate trials. The trial by Petrella had 2 treatment groups, one with only hyaluronic acid and the other with hyaluronic acid and NSAIDs. We considered them separately in the analysis, resulting in a total of 11 clinical trials for the meta-analysis.¹⁹ Henceforth in this report, we will refer to 11 rather than 9 clinical trials.

Extraction of data

Two investigators independently extracted the following data for each study: year of publication, study design, mean age, number of patients enrolled in each treatment group, number of doses of treatment used, and outcomes measured. When disagreements between investigators occurred, the point of disagreement was discussed until a consensus was reached. Since the treatment duration and the time post-treatment when pain was assessed varied among the trials, we grouped outcomes into four time intervals: at 1 week, 5 to 7 weeks, 8 to 12 weeks, and 15 to 22 weeks after the last hyaluronic acid injection.

Statistical analysis

The outcome was knee pain reported by patients on activity or at rest, measured using a VAS of 100 mm. The results of the clinical trials were recorded as the mean differences of change from baseline between the treatment and placebo groups. If not reported in the publication or provided by the authors, standard error was imputed using the method of Follman et al.²⁰ We used the method of Chalmers for measuring the quality of randomized trials, with 2 of the authors rating the studies independently.²¹

This clinical trial grading system takes into account the following aspects of the trial to determine a quality score: evaluation of recruitment of subjects, rejection log, therapeutic regimen definition, randomization, blinding, prior estimates of numbers, testing compliance, statistical inference, use of appropriate statistical analysis, handling of withdrawal and side effects, dates of starting and ending, timing and tabulation of events. Because 4 of the trials had poor quality scores, an additional analysis excluding these 4 was performed. The DerSimonian and Laird random-effects model was used to obtain the summary estimates.^22,23

An important element of meta-analysis is exploration of the heterogeneity of the outcomes and the possible causes of heterogeneity if it exists. Heterogeneity is the degree to which results vary from study to study. If a test for heterogeneity is statistically significant, there is significant variability among the treatment effects observed in the trials. We explored heterogeneity using Galbraith plots (FIGURE 1).²⁴ In the absence of heterogeneity, all points fall within the confidence limits. Because we did find heterogeneity among these trials, we developed random-effect regression models to explore 3 possible sources of heterogeneity in the efficacy of hyaluronic acid; pain (measured at rest or on activity), the form used (hyaluronan or hylan G-F20), and the quality of the study method (good or poor).

Publication bias was assessed by the Egger et al regression asymmetry test.²⁵ Analyses were performed using the meta-analytic software program of STATA, Inc (College Station, Tex; available at www.stata.com).

FIGURE 1
Evidence of heterogeneity for the studies evaluated at outcome measurement time (week 1)

Results

The 11 randomized, placebo-controlled, double-blinded clinical trials that met our inclusion criteria are summarized in TABLE 1. Nine trials used hyaluronic acid, hyaluronan, or hyaluronate (all types will be referred to as hyaluronan in the text), and 2 studies used hylan GF-20. Only 3 hyaluronan trials have published outcome data at 15 to 22 weeks follow-up. Treatment was administered to patients as 3 to 5 weekly injections, with the exception of the Grecomoro study in which treatment was administered twice weekly. The control group in 10 trials received intra-articular saline injections as placebo. In the Puhl study the investigators added 0.25 mg of hyaluronic acid to the saline injections to impart viscosity to the solution. The mean age of the subjects for the 11 trials was 63 years.

Eight trials received support from pharmaceutical companies and 3 (the 2 by Henderson and the Grecomoro study) did not disclose any pharmaceutical support. One study was conducted in the United States, 3 in the UK, 3 in Germany, 1 in Sweden, 1 in Italy, and 2 in Canada. Five of the studies had scores over a cutoff quality score of >0.75,^{12,13,15,16,19} indicating they were good-quality randomized controlled trials; the remaining 4 had scores below 0.75.^11,14,16,17

The outcomes of the 11 trials are summarized in TABLE 2. Patients’ pain ratings in both the active treatment and placebo groups improved in all the trials. Mean difference between improvements in treatment and placebo groups are shown in FIGURES 2A-2D for pain assessed at weeks 1, 5 to 7, 8 to 12 and 15 to 22, respectively. In each figure, we show the summary estimate of effect size with all the trials included and after excluding the 4 trials considered of poor quality, shown as “good quality studies.”

The mean difference in pain scores between treatment and placebo at week 1 was 4.4 (95% CI, +1.1, +7.2) and –1.0 (95% CI, -3.2, +1.2) for analysis restricted to the 7 good quality trials. The mean difference in pain scores at 5 to 7 weeks was 17.6 (95% CI, +7.5, +28.0) and 7.2 (95% CI, +2.4, +12.0) for the analysis restricted to the 2 good quality studies. At weeks 8 to 12 the mean difference in VAS between treatment and control was 18.1 (95% CI, +6.3, +29.9), and 7.1 (95% CI, +3.0, +11.3) in the analysis restricted to good quality trials. At weeks 15 to 22, the mean difference was 4.4 (95% CI, –15.3, +24.1). The Egger test was not statistically significant (2.3; P=.096; 95% CI, –0.5, +5.2) suggesting that there is no publication bias.

High heterogeneity was observed at all time intervals except 1 week (FIGURE 1). Of the 5 trials outside the confidence bounds (positioned 2 units above and below the regression line), 4 were poor-quality studies.

TABLE 3 shows the random-effect regression models we used to test the influence on the outcome of type of pain measured (pain with activity or pain at rest), type of medication (hyaluronan or hylan G-F 20), and study quality (good or poor). No significant association between treatment efficacy and type of pain used as outcome variable was observed. Clinical trials using hylan GF-20 showed statistically significant better results than those using hyaluronan at weeks 5 to 7 and 8 to 12. Poor-quality studies showed a larger treatment effect, but the difference was statistically significant only at week 1.

TABLE 3
Regression models to assess the sources of heterogeneity in the meta-analysis

Discussion

This meta-analysis synthesized data from 9 randomized, double-blinded, placebo-controlled trials that evaluated the efficacy of intra-articular hyaluronic acid. Our findings show significantly decreased pain as measured by VAS at 5 to 7 weeks and at 8 to 12 weeks after the last injection. Intra-articular hyaluronic acid was not more effective than placebo in relieving pain at 1 week or at 15 to 22 weeks after the last injection. Because only 3 of the trials assessed patients after 12 weeks, however, the sample size is too small to definitively rule out a significant therapeutic effect after 12 weeks.

Reasons for the differences in efficacy among trials of hyaluronic acid in the treatment of knee osteoarthritis include dose, type, and frequency of administration, genetic or age differences among the study subjects, severity of osteoarthritis, time of follow-up, and quality of the studies. We confirmed that the treatment effect is time dependent. Although our meta-regression analysis (TABLE 2) suggests that hylan GF-20 is more effective than hyaluronan at 5 to 12 weeks, the number of clinical trials is relatively small and both of the hylan G-F20 studies were of poor quality. Therefore, we cannot say with confidence that one form is better than the other. Data in these trials were insufficient to assess the impact of body mass index, genetics, or severity of osteoarthritis.

We did not evaluate functional improvement in this meta-analysis because functional status was not measured in some trials and the assessment methods were too variable in the trials that did assess functional status. Publication bias, or the possibility that unpublished data would contradict the results of published studies, is always a potential source of bias in meta-analysis. However, the Egger test was not statistically significant (6.5; 95% CI, –0.5, +13.5) suggesting that there is no publication bias.²⁵

Finally, the presence of heterogeneity of results indicates there were important differences among the studies. Exclusion of clinical trials considered of poor quality diminished this heterogeneity substantially. Subanalysis restricted to good-quality studies supports the efficacy of intra-articular hyaluronic acid in the treatment of knee osteoarthritis pain, although the effect size is smaller when one considers only the good quality studies.

There are 2 other potential limitations of this meta-analysis. Five studies allowed pain to be treated with analgesics such as acetaminophen or NSAIDs,^12,13,16-18 and use of acetaminophen or NSAIDs may have altered the response to hyaluronic acid treatment. An intention-to-treat analysis was performed in only 2 studies (by Altman and Huskisson),^16,18 wherein a post-hoc and “last observation carried forward” analysis showed a trend favoring hyaluronic acid. The treatment effects may have been smaller had the other trials used an intention to treat analysis.

This meta-analysis confirms that viscosupplementation with hyaluronic acid is modestly effective in short term relief of pain in knee osteoarthritis. Our meta-analysis included only double-blinded, randomized trials published in English language in humans using VAS as the pain outcome measure, and our conclusions are very similar to those of Lo.²

Indications for use. Hyaluronic acid is helpful in relieving pain for carefully selected patients with knee osteoarthritis who have not responded to adequate use of systemic therapeutic agents, including acetaminophen, NSAIDs and COX-2 inhibitors and topical agents, along with lifestyle modification such as weight reduction and exercise.

Patients should have a trial for at least three months of conservative therapy or be unable to tolerate NSAIDs before a decision to give 3- to 5-injection course with hyaluronic acid is made.

Hyaluronic acid may be an option when there is a need to delay knee surgery in middle-aged persons¹ or for patients who have failed other treatments.

Time to pain relief. To improve adherence to treatment, tell patients receiving intra-articular hyaluronic acid that the benefits in pain reduction may not be noticeable until 5 to 10 weeks after the last injection.

Cost. Although the cost of hyaluronic acid treatment is covered by Medicare and most insurance plans for symptomatic osteoarthritis of knee, documentation in patient medical records should indicate the signs and symptoms supporting the diagnosis and functional impairment. Objective data to support a diagnosis of osteoarthritis such as x-ray, arthroscopy report, computed tomography scan, or magnetic resonance imaging should be available in the event of a review.

The cost of 1 hyaluronic acid (30 mg/mL) injection is approximately $230. Considering a course of 3 to 5 weekly knee injections, and adding other pharmacy, hospital, or clinic charges, the cost per treatment may exceed $1000 per knee.^26,27 The cost-benefit of pain control with viscosupplementation must be carefully compared with other therapeutic agents and regimens currently available for knee osteoarthritis management.

FIGURE 3
Studies evaluated at outcome measurement time (week 1)

Acknowledgments

Jeff Welge, PhD, University of Cincinnati Medical Center, Center for Bio-statistical Services, for comments on final data analysis and presentation, Marie Marley, PhD for editing the manuscript, and Charity Noble for help in preparation of this manuscript.

CORRESPONDING AUTHOR
Arvind Modawal, MD, MPH, MRCGP, University of Cincinnati Department of Family Medicine/Geriatrics, PO Box 670582, Cincinnati, OH 45267-0582. E-mail: [email protected]

Hyaluronic acid injections can help relieve pain for carefully selected patients with knee osteoarthritis. But this option should be reserved for those whose pain has not responded to adequate trials of systemic therapeutic agents (acetaminophen, nonsteroidal anti-inflammatory drugs [NSAIDs], cyclooxygenase-2 [COX-2] inhibitors), topical agents, or to lifestyle modifications such as weight reduction and exercise.

Hyaluronic acid injections may also be indicated when knee surgery must be delayed for middle-aged persons.¹

In spite of the Food and Drug Administration’s approval of this therapy, uncertainty about its efficacy exists among the medical community. A recent meta-analysis of the effectiveness of intra-articular hyaluronic acid for knee osteoarthritis that included 22 published and unpublished, English and non-English, single or double-blinded, randomized controlled trials in humans showed that hyaluronic acid has only a small effect on pain relief when compared with placebo.²

We provide here a stringent test of the efficacy of viscosupplementation for relieving knee pain from osteoarthritis with a meta-analysis that includes only data from randomized, double-blinded, controlled trials of hyaluronic acid that measured pain using a visual analogue scale (VAS), the most widely accepted method for pain evaluation.

Long time coming

Balazs first proposed hyaluronic acid as a treatment for patients with arthritic diseases in 1942. In the early 1970s, therapeutic studies were begun to test the efficacy of hyaluronic acid on knee osteoarthritis. The results were encouraging and side effects were few.²⁸ With the FDA’s approval in 1998, intra-articular “viscosupplementation” with hyaluronic acid—also called hyaluronan or hyaluronate, and the hylan derivatives of hyaluronic acid—is a welcome option for many of the 16 million older Americans with osteoarthritis of the knee.¹

Methods

Selection of studies

We identified clinical trials of viscosupplementation with hyaluronic acid in humans published in English from 1965 through August 2004 through a computerized literature search of Medline. The keyword used was “hyaluronic acid,” which was combined with “trial” or “osteoarthritis knee” or “viscosupplementation.” We conducted an additional manual search of the reference lists of included articles and review articles. We also searched the Cochrane Library and websites of the Agency for Healthcare Research and Quality (AHRQ) for information on hyaluronic acid in knee osteoarthritis. We identified 1872 articles with this search process.

Of the 1872 articles, we identified by title and abstract 33 that might be pertinent to this study, including 17 randomized trials. We excluded reviews, meta-analyses, comparison trials, and trials reporting VAS as part of the WOMAC (Western Ontario McMaster Universities Index) scale. We attempted to contact authors of the studies that used a double-blind, randomized controlled design, to obtain any data that may not have been included in the publications. Three authors provided the details requested; the others did not respond or stated that additional data were not available. Of the 17 randomized trials we identified, 8 were excluded because they were open, single-blinded, or did not use the VAS to measure pain outcomes.^3-10 The remaining 9 double-blinded, placebo controlled, randomized clinical trials of viscosupplementation with hyaluronic acid for knee osteoarthritis that did use a VAS to measure pain were included in this meta-analysis (TABLE 1). Because one of the studies (by Henderson) included 2 subgroups of pain severity, these were considered as 2 separate trials. The trial by Petrella had 2 treatment groups, one with only hyaluronic acid and the other with hyaluronic acid and NSAIDs. We considered them separately in the analysis, resulting in a total of 11 clinical trials for the meta-analysis.¹⁹ Henceforth in this report, we will refer to 11 rather than 9 clinical trials.

Extraction of data

Two investigators independently extracted the following data for each study: year of publication, study design, mean age, number of patients enrolled in each treatment group, number of doses of treatment used, and outcomes measured. When disagreements between investigators occurred, the point of disagreement was discussed until a consensus was reached. Since the treatment duration and the time post-treatment when pain was assessed varied among the trials, we grouped outcomes into four time intervals: at 1 week, 5 to 7 weeks, 8 to 12 weeks, and 15 to 22 weeks after the last hyaluronic acid injection.

Statistical analysis

The outcome was knee pain reported by patients on activity or at rest, measured using a VAS of 100 mm. The results of the clinical trials were recorded as the mean differences of change from baseline between the treatment and placebo groups. If not reported in the publication or provided by the authors, standard error was imputed using the method of Follman et al.²⁰ We used the method of Chalmers for measuring the quality of randomized trials, with 2 of the authors rating the studies independently.²¹

This clinical trial grading system takes into account the following aspects of the trial to determine a quality score: evaluation of recruitment of subjects, rejection log, therapeutic regimen definition, randomization, blinding, prior estimates of numbers, testing compliance, statistical inference, use of appropriate statistical analysis, handling of withdrawal and side effects, dates of starting and ending, timing and tabulation of events. Because 4 of the trials had poor quality scores, an additional analysis excluding these 4 was performed. The DerSimonian and Laird random-effects model was used to obtain the summary estimates.^22,23

An important element of meta-analysis is exploration of the heterogeneity of the outcomes and the possible causes of heterogeneity if it exists. Heterogeneity is the degree to which results vary from study to study. If a test for heterogeneity is statistically significant, there is significant variability among the treatment effects observed in the trials. We explored heterogeneity using Galbraith plots (FIGURE 1).²⁴ In the absence of heterogeneity, all points fall within the confidence limits. Because we did find heterogeneity among these trials, we developed random-effect regression models to explore 3 possible sources of heterogeneity in the efficacy of hyaluronic acid; pain (measured at rest or on activity), the form used (hyaluronan or hylan G-F20), and the quality of the study method (good or poor).

Publication bias was assessed by the Egger et al regression asymmetry test.²⁵ Analyses were performed using the meta-analytic software program of STATA, Inc (College Station, Tex; available at www.stata.com).

FIGURE 1
Evidence of heterogeneity for the studies evaluated at outcome measurement time (week 1)

Results

The 11 randomized, placebo-controlled, double-blinded clinical trials that met our inclusion criteria are summarized in TABLE 1. Nine trials used hyaluronic acid, hyaluronan, or hyaluronate (all types will be referred to as hyaluronan in the text), and 2 studies used hylan GF-20. Only 3 hyaluronan trials have published outcome data at 15 to 22 weeks follow-up. Treatment was administered to patients as 3 to 5 weekly injections, with the exception of the Grecomoro study in which treatment was administered twice weekly. The control group in 10 trials received intra-articular saline injections as placebo. In the Puhl study the investigators added 0.25 mg of hyaluronic acid to the saline injections to impart viscosity to the solution. The mean age of the subjects for the 11 trials was 63 years.

Eight trials received support from pharmaceutical companies and 3 (the 2 by Henderson and the Grecomoro study) did not disclose any pharmaceutical support. One study was conducted in the United States, 3 in the UK, 3 in Germany, 1 in Sweden, 1 in Italy, and 2 in Canada. Five of the studies had scores over a cutoff quality score of >0.75,^{12,13,15,16,19} indicating they were good-quality randomized controlled trials; the remaining 4 had scores below 0.75.^11,14,16,17

The outcomes of the 11 trials are summarized in TABLE 2. Patients’ pain ratings in both the active treatment and placebo groups improved in all the trials. Mean difference between improvements in treatment and placebo groups are shown in FIGURES 2A-2D for pain assessed at weeks 1, 5 to 7, 8 to 12 and 15 to 22, respectively. In each figure, we show the summary estimate of effect size with all the trials included and after excluding the 4 trials considered of poor quality, shown as “good quality studies.”

The mean difference in pain scores between treatment and placebo at week 1 was 4.4 (95% CI, +1.1, +7.2) and –1.0 (95% CI, -3.2, +1.2) for analysis restricted to the 7 good quality trials. The mean difference in pain scores at 5 to 7 weeks was 17.6 (95% CI, +7.5, +28.0) and 7.2 (95% CI, +2.4, +12.0) for the analysis restricted to the 2 good quality studies. At weeks 8 to 12 the mean difference in VAS between treatment and control was 18.1 (95% CI, +6.3, +29.9), and 7.1 (95% CI, +3.0, +11.3) in the analysis restricted to good quality trials. At weeks 15 to 22, the mean difference was 4.4 (95% CI, –15.3, +24.1). The Egger test was not statistically significant (2.3; P=.096; 95% CI, –0.5, +5.2) suggesting that there is no publication bias.

High heterogeneity was observed at all time intervals except 1 week (FIGURE 1). Of the 5 trials outside the confidence bounds (positioned 2 units above and below the regression line), 4 were poor-quality studies.

TABLE 3 shows the random-effect regression models we used to test the influence on the outcome of type of pain measured (pain with activity or pain at rest), type of medication (hyaluronan or hylan G-F 20), and study quality (good or poor). No significant association between treatment efficacy and type of pain used as outcome variable was observed. Clinical trials using hylan GF-20 showed statistically significant better results than those using hyaluronan at weeks 5 to 7 and 8 to 12. Poor-quality studies showed a larger treatment effect, but the difference was statistically significant only at week 1.

TABLE 3
Regression models to assess the sources of heterogeneity in the meta-analysis

Discussion

This meta-analysis synthesized data from 9 randomized, double-blinded, placebo-controlled trials that evaluated the efficacy of intra-articular hyaluronic acid. Our findings show significantly decreased pain as measured by VAS at 5 to 7 weeks and at 8 to 12 weeks after the last injection. Intra-articular hyaluronic acid was not more effective than placebo in relieving pain at 1 week or at 15 to 22 weeks after the last injection. Because only 3 of the trials assessed patients after 12 weeks, however, the sample size is too small to definitively rule out a significant therapeutic effect after 12 weeks.

Reasons for the differences in efficacy among trials of hyaluronic acid in the treatment of knee osteoarthritis include dose, type, and frequency of administration, genetic or age differences among the study subjects, severity of osteoarthritis, time of follow-up, and quality of the studies. We confirmed that the treatment effect is time dependent. Although our meta-regression analysis (TABLE 2) suggests that hylan GF-20 is more effective than hyaluronan at 5 to 12 weeks, the number of clinical trials is relatively small and both of the hylan G-F20 studies were of poor quality. Therefore, we cannot say with confidence that one form is better than the other. Data in these trials were insufficient to assess the impact of body mass index, genetics, or severity of osteoarthritis.

We did not evaluate functional improvement in this meta-analysis because functional status was not measured in some trials and the assessment methods were too variable in the trials that did assess functional status. Publication bias, or the possibility that unpublished data would contradict the results of published studies, is always a potential source of bias in meta-analysis. However, the Egger test was not statistically significant (6.5; 95% CI, –0.5, +13.5) suggesting that there is no publication bias.²⁵

Finally, the presence of heterogeneity of results indicates there were important differences among the studies. Exclusion of clinical trials considered of poor quality diminished this heterogeneity substantially. Subanalysis restricted to good-quality studies supports the efficacy of intra-articular hyaluronic acid in the treatment of knee osteoarthritis pain, although the effect size is smaller when one considers only the good quality studies.

There are 2 other potential limitations of this meta-analysis. Five studies allowed pain to be treated with analgesics such as acetaminophen or NSAIDs,^12,13,16-18 and use of acetaminophen or NSAIDs may have altered the response to hyaluronic acid treatment. An intention-to-treat analysis was performed in only 2 studies (by Altman and Huskisson),^16,18 wherein a post-hoc and “last observation carried forward” analysis showed a trend favoring hyaluronic acid. The treatment effects may have been smaller had the other trials used an intention to treat analysis.

This meta-analysis confirms that viscosupplementation with hyaluronic acid is modestly effective in short term relief of pain in knee osteoarthritis. Our meta-analysis included only double-blinded, randomized trials published in English language in humans using VAS as the pain outcome measure, and our conclusions are very similar to those of Lo.²

Indications for use. Hyaluronic acid is helpful in relieving pain for carefully selected patients with knee osteoarthritis who have not responded to adequate use of systemic therapeutic agents, including acetaminophen, NSAIDs and COX-2 inhibitors and topical agents, along with lifestyle modification such as weight reduction and exercise.

Patients should have a trial for at least three months of conservative therapy or be unable to tolerate NSAIDs before a decision to give 3- to 5-injection course with hyaluronic acid is made.

Hyaluronic acid may be an option when there is a need to delay knee surgery in middle-aged persons¹ or for patients who have failed other treatments.

Time to pain relief. To improve adherence to treatment, tell patients receiving intra-articular hyaluronic acid that the benefits in pain reduction may not be noticeable until 5 to 10 weeks after the last injection.

Cost. Although the cost of hyaluronic acid treatment is covered by Medicare and most insurance plans for symptomatic osteoarthritis of knee, documentation in patient medical records should indicate the signs and symptoms supporting the diagnosis and functional impairment. Objective data to support a diagnosis of osteoarthritis such as x-ray, arthroscopy report, computed tomography scan, or magnetic resonance imaging should be available in the event of a review.

The cost of 1 hyaluronic acid (30 mg/mL) injection is approximately $230. Considering a course of 3 to 5 weekly knee injections, and adding other pharmacy, hospital, or clinic charges, the cost per treatment may exceed $1000 per knee.^26,27 The cost-benefit of pain control with viscosupplementation must be carefully compared with other therapeutic agents and regimens currently available for knee osteoarthritis management.

FIGURE 3
Studies evaluated at outcome measurement time (week 1)

Acknowledgments

Jeff Welge, PhD, University of Cincinnati Medical Center, Center for Bio-statistical Services, for comments on final data analysis and presentation, Marie Marley, PhD for editing the manuscript, and Charity Noble for help in preparation of this manuscript.

CORRESPONDING AUTHOR
Arvind Modawal, MD, MPH, MRCGP, University of Cincinnati Department of Family Medicine/Geriatrics, PO Box 670582, Cincinnati, OH 45267-0582. E-mail: [email protected]

Hyaluronic acid injections can help relieve pain for carefully selected patients with knee osteoarthritis. But this option should be reserved for those whose pain has not responded to adequate trials of systemic therapeutic agents (acetaminophen, nonsteroidal anti-inflammatory drugs [NSAIDs], cyclooxygenase-2 [COX-2] inhibitors), topical agents, or to lifestyle modifications such as weight reduction and exercise.

Hyaluronic acid injections may also be indicated when knee surgery must be delayed for middle-aged persons.¹

In spite of the Food and Drug Administration’s approval of this therapy, uncertainty about its efficacy exists among the medical community. A recent meta-analysis of the effectiveness of intra-articular hyaluronic acid for knee osteoarthritis that included 22 published and unpublished, English and non-English, single or double-blinded, randomized controlled trials in humans showed that hyaluronic acid has only a small effect on pain relief when compared with placebo.²

We provide here a stringent test of the efficacy of viscosupplementation for relieving knee pain from osteoarthritis with a meta-analysis that includes only data from randomized, double-blinded, controlled trials of hyaluronic acid that measured pain using a visual analogue scale (VAS), the most widely accepted method for pain evaluation.

Long time coming

Balazs first proposed hyaluronic acid as a treatment for patients with arthritic diseases in 1942. In the early 1970s, therapeutic studies were begun to test the efficacy of hyaluronic acid on knee osteoarthritis. The results were encouraging and side effects were few.²⁸ With the FDA’s approval in 1998, intra-articular “viscosupplementation” with hyaluronic acid—also called hyaluronan or hyaluronate, and the hylan derivatives of hyaluronic acid—is a welcome option for many of the 16 million older Americans with osteoarthritis of the knee.¹

Methods

Selection of studies

We identified clinical trials of viscosupplementation with hyaluronic acid in humans published in English from 1965 through August 2004 through a computerized literature search of Medline. The keyword used was “hyaluronic acid,” which was combined with “trial” or “osteoarthritis knee” or “viscosupplementation.” We conducted an additional manual search of the reference lists of included articles and review articles. We also searched the Cochrane Library and websites of the Agency for Healthcare Research and Quality (AHRQ) for information on hyaluronic acid in knee osteoarthritis. We identified 1872 articles with this search process.

Of the 1872 articles, we identified by title and abstract 33 that might be pertinent to this study, including 17 randomized trials. We excluded reviews, meta-analyses, comparison trials, and trials reporting VAS as part of the WOMAC (Western Ontario McMaster Universities Index) scale. We attempted to contact authors of the studies that used a double-blind, randomized controlled design, to obtain any data that may not have been included in the publications. Three authors provided the details requested; the others did not respond or stated that additional data were not available. Of the 17 randomized trials we identified, 8 were excluded because they were open, single-blinded, or did not use the VAS to measure pain outcomes.^3-10 The remaining 9 double-blinded, placebo controlled, randomized clinical trials of viscosupplementation with hyaluronic acid for knee osteoarthritis that did use a VAS to measure pain were included in this meta-analysis (TABLE 1). Because one of the studies (by Henderson) included 2 subgroups of pain severity, these were considered as 2 separate trials. The trial by Petrella had 2 treatment groups, one with only hyaluronic acid and the other with hyaluronic acid and NSAIDs. We considered them separately in the analysis, resulting in a total of 11 clinical trials for the meta-analysis.¹⁹ Henceforth in this report, we will refer to 11 rather than 9 clinical trials.

Extraction of data

Two investigators independently extracted the following data for each study: year of publication, study design, mean age, number of patients enrolled in each treatment group, number of doses of treatment used, and outcomes measured. When disagreements between investigators occurred, the point of disagreement was discussed until a consensus was reached. Since the treatment duration and the time post-treatment when pain was assessed varied among the trials, we grouped outcomes into four time intervals: at 1 week, 5 to 7 weeks, 8 to 12 weeks, and 15 to 22 weeks after the last hyaluronic acid injection.

Statistical analysis

The outcome was knee pain reported by patients on activity or at rest, measured using a VAS of 100 mm. The results of the clinical trials were recorded as the mean differences of change from baseline between the treatment and placebo groups. If not reported in the publication or provided by the authors, standard error was imputed using the method of Follman et al.²⁰ We used the method of Chalmers for measuring the quality of randomized trials, with 2 of the authors rating the studies independently.²¹

This clinical trial grading system takes into account the following aspects of the trial to determine a quality score: evaluation of recruitment of subjects, rejection log, therapeutic regimen definition, randomization, blinding, prior estimates of numbers, testing compliance, statistical inference, use of appropriate statistical analysis, handling of withdrawal and side effects, dates of starting and ending, timing and tabulation of events. Because 4 of the trials had poor quality scores, an additional analysis excluding these 4 was performed. The DerSimonian and Laird random-effects model was used to obtain the summary estimates.^22,23

An important element of meta-analysis is exploration of the heterogeneity of the outcomes and the possible causes of heterogeneity if it exists. Heterogeneity is the degree to which results vary from study to study. If a test for heterogeneity is statistically significant, there is significant variability among the treatment effects observed in the trials. We explored heterogeneity using Galbraith plots (FIGURE 1).²⁴ In the absence of heterogeneity, all points fall within the confidence limits. Because we did find heterogeneity among these trials, we developed random-effect regression models to explore 3 possible sources of heterogeneity in the efficacy of hyaluronic acid; pain (measured at rest or on activity), the form used (hyaluronan or hylan G-F20), and the quality of the study method (good or poor).

Publication bias was assessed by the Egger et al regression asymmetry test.²⁵ Analyses were performed using the meta-analytic software program of STATA, Inc (College Station, Tex; available at www.stata.com).

FIGURE 1
Evidence of heterogeneity for the studies evaluated at outcome measurement time (week 1)

Results

The 11 randomized, placebo-controlled, double-blinded clinical trials that met our inclusion criteria are summarized in TABLE 1. Nine trials used hyaluronic acid, hyaluronan, or hyaluronate (all types will be referred to as hyaluronan in the text), and 2 studies used hylan GF-20. Only 3 hyaluronan trials have published outcome data at 15 to 22 weeks follow-up. Treatment was administered to patients as 3 to 5 weekly injections, with the exception of the Grecomoro study in which treatment was administered twice weekly. The control group in 10 trials received intra-articular saline injections as placebo. In the Puhl study the investigators added 0.25 mg of hyaluronic acid to the saline injections to impart viscosity to the solution. The mean age of the subjects for the 11 trials was 63 years.

Eight trials received support from pharmaceutical companies and 3 (the 2 by Henderson and the Grecomoro study) did not disclose any pharmaceutical support. One study was conducted in the United States, 3 in the UK, 3 in Germany, 1 in Sweden, 1 in Italy, and 2 in Canada. Five of the studies had scores over a cutoff quality score of >0.75,^{12,13,15,16,19} indicating they were good-quality randomized controlled trials; the remaining 4 had scores below 0.75.^11,14,16,17

The outcomes of the 11 trials are summarized in TABLE 2. Patients’ pain ratings in both the active treatment and placebo groups improved in all the trials. Mean difference between improvements in treatment and placebo groups are shown in FIGURES 2A-2D for pain assessed at weeks 1, 5 to 7, 8 to 12 and 15 to 22, respectively. In each figure, we show the summary estimate of effect size with all the trials included and after excluding the 4 trials considered of poor quality, shown as “good quality studies.”

The mean difference in pain scores between treatment and placebo at week 1 was 4.4 (95% CI, +1.1, +7.2) and –1.0 (95% CI, -3.2, +1.2) for analysis restricted to the 7 good quality trials. The mean difference in pain scores at 5 to 7 weeks was 17.6 (95% CI, +7.5, +28.0) and 7.2 (95% CI, +2.4, +12.0) for the analysis restricted to the 2 good quality studies. At weeks 8 to 12 the mean difference in VAS between treatment and control was 18.1 (95% CI, +6.3, +29.9), and 7.1 (95% CI, +3.0, +11.3) in the analysis restricted to good quality trials. At weeks 15 to 22, the mean difference was 4.4 (95% CI, –15.3, +24.1). The Egger test was not statistically significant (2.3; P=.096; 95% CI, –0.5, +5.2) suggesting that there is no publication bias.

High heterogeneity was observed at all time intervals except 1 week (FIGURE 1). Of the 5 trials outside the confidence bounds (positioned 2 units above and below the regression line), 4 were poor-quality studies.

TABLE 3 shows the random-effect regression models we used to test the influence on the outcome of type of pain measured (pain with activity or pain at rest), type of medication (hyaluronan or hylan G-F 20), and study quality (good or poor). No significant association between treatment efficacy and type of pain used as outcome variable was observed. Clinical trials using hylan GF-20 showed statistically significant better results than those using hyaluronan at weeks 5 to 7 and 8 to 12. Poor-quality studies showed a larger treatment effect, but the difference was statistically significant only at week 1.

TABLE 3
Regression models to assess the sources of heterogeneity in the meta-analysis