Applied Evidence

Diabetes affects 18.2 million people in the United States.⁵ It is a leading cause of morbidity and mortality.⁵ Diabetes is associated with more than $90 billion in direct medical costs and with an estimated $40 billion in indirect costs (ie, disability, work loss, and premature mortality).⁵ Type 2 diabetes accounts for 90% to 95% of total cases. It disproportionately affects certain minority populations, and it is increasingly being identified in children and adolescents.^5-7

The obesity epidemic has received tremendous media and public attention and has sobering implications for the development of metabolic syndrome, which is often the precursor to diabetes, and cardiovascular disease (CVD).^8,9 In 1 study, the hazard ratio for development of diabetes was 3.85 for individuals who gained 20 kg or more over approximately 10 years, as compared with individuals whose weight remained relatively stable over the same period.¹⁰ Obesity is associated with elevations in free fatty acid levels and in other compounds which contribute to insulin resistance¹¹ —a key underlying defect in metabolic syndrome and type 2 diabetes. Thus, particularly in genetically predisposed individuals, obesity (through elevated fatty acids) may be the factor that drives insulin resistance to clinical diabetes. In addition, elevated free fatty acid levels also may contribute to the other key underlying defect in type 2 diabetes: progressive pancreatic β-cell loss which results in insulin deficiency.¹² Similar to the trend in adults, childhood obesity has now reached epidemic proportions and has been associated with the increased prevalence of glucose intolerance, metabolic syndrome, and biomarkers of increased CVD risk in this young population.^13-15

Type 2 diabetes in adolescents

CASE 1 Suspected new-onset type 2 diabetes

A 16-year-old Hispanic boy presents for treatment of weight loss. He is a defensive tackle on the high school football team and has a strong appetite. His height is 6 ft 1 in and he weighs 250 lb; his body mass index (BMI) is 32 kg/m², and most of the fat is abdominally distributed (waist circumference, 42 in). Both of his parents are obese, and a grandfather and aunt have type 2 diabetes. Further examination reveals a blood pressure (BP) level of 135/87 mm Hg and acanthosis nigricans. A random fingerstick test shows a blood glucose level of 240 mg/dL.

Until recently, type 2 diabetes rarely was observed in children, occurring in only 1% to 2% of children with new-onset diabetes.¹⁶ However, depending on the sampling strategy, race or ethnicity of the population, and the region of the country sampled, it now has been estimated that type 2 diabetes accounts for anywhere from 8% to 45% of all diabetes reported among children and adolescents.^6,16 Prevalence estimates of pediatric type 2 diabetes in population-based studies range from 4.1 per 1,000 in the National Health and Nutrition Survey (NHANES III) to 50.9 per 1,000 among Pima Indians in Arizona.¹⁷ Unlike the trends of increasing incidence and prevalence of type 2 diabetes in the pediatric population, a similar trend in type 1 diabetes has not been observed.⁷

The diagnosis and treatment of type 2 diabetes in pediatric patients can be challenging. Type 2 diabetes is associated with an increased BMI and is more common in adolescents than in younger children.⁶ Puberty is associated with relative insulin resistance and, along with the time-related effects of obesity, resistance may play a role in the subsequent onset of type 2 diabetes in pediatric patients. However, as an increasing number of children become obese at an earlier age, the onset of disease may occur earlier.

Other risk factors for type 2 diabetes in childhood include having a first- or second-degree relative with the disease or non-European ancestry (Americans of African, Hispanic, Asian, Pacific Islander, or Native American descent). In addition, acanthosis nigricans is a skin condition that serves as a specific, though insensitive, marker of insulin resistance. Among adolescent females, polycystic ovary syndrome has been associated with insulin resistance irrespective of excess weight or frank obesity¹⁸ and may be associated with an increased risk of type 2 diabetes.⁶

TABLE 1 provides criteria for screening for type 2 diabetes in children. The adolescent in the case presented has several risk factors for type 2 diabetes, including obesity, a positive family history, Hispanic descent, and acanthosis nigricans. To more definitively diagnose type 2 diabetes, additional laboratory tests are needed, including assessment of blood glucose and insulin production, as well as the presence of autoantibodies.

TABLE 1

Screening recommendations for type 2 diabetes in children

CASE 1 Laboratory assessment

The patient has a fasting plasma glucose (FPG) level of 215 mg/dL. β-Cell or insulin autoantibodies are not detected. He has a normal-to-high fasting C-peptide level, microalbuminuria, and a serum creatinine level of 0.8 mg/dL. His lipid panel results include: total cholesterol, 234 mg/dL; low-density lipoprotein (LDL) cholesterol, 159 mg/dL; high-density lipoprotein (HDL) cholesterol, 45 mg/dL; and triglycerides, 150 mg/dL.

A comparison of the typical presentation and laboratory findings in pediatric type 1 and type 2 diabetes is listed in TABLE 2. This information underscores that there is significant overlap between the 2 disease classifications.

The laboratory results for this patient confirm a diagnosis of type 2 diabetes. C peptide is a marker of endogenous insulin production and is particularly helpful in the classification of diabetes when it is high, which is indicative of type 2 diabetes. However, it should be noted that a finding of low C-peptide levels does not rule out type 2 diabetes since there is a possibility that glucose toxicity is temporarily limiting insulin production. Indeed, in such a case, insulin therapy can be initiated to establish glycemic control and promote endogenous insulin production by reducing glucose toxicity. Similarly, although the absence of pancreatic β-cells and insulin autoantibodies is consistent with a diagnosis of type 2 diabetes, the presence of such antibodies does not completely rule out type 2 diabetes.¹⁹

Like many patients with type 2 diabetes, the current patient exhibits a cluster of CVD risk factors that characterize the metabolic syndrome, including dyslipidemia (elevated total cholesterol, LDL cholesterol, triglyceride levels, and low HDL cholesterol levels), hypertension, glucose intolerance, hyperinsulinemia, and central or abdominal obesity. This is not surprising as an estimated 30% of overweight adolescents, defined as those with BMI at or above the 95th percentile, meet the criteria for metabolic syndrome.¹⁴

TABLE 2

Classification of type 1 and type 2 diabetes in pediatric patients

CASE 1 Initial therapy

The patient is provided with education on lifestyle modifications to reduce weight and obtain better control of glucose, lipids, and BP. Instructions include ways to improve his eating habits and to increase his physical activity when away from the football field, such as limiting time spent watching TV and playing video games and increasing exercise time and effort. The patient also is instructed on self-monitoring for blood glucose (SMBG) and is given a referral to an ophthalmologist to determine whether there is any baseline retinopathy.

At his 3-month follow-up, the patient’s A1C is 8.7%, and his weight has not changed. At this point, lifestyle modifications are reinforced, and pharmacotherapy with extended-release metformin is initiated and titrated to 2,000 mg/d.

Typically, the first approach to managing type 2 diabetes involves dietary changes and instituting an exercise program to reduce weight and to improve insulin sensitivity. This was the approach used in the patient described above. However, as is frequently the case, lifestyle modifications by themselves are insufficient to meet the American Diabetes Association (ADA) or American Association of Clinical Endocrinologists targets for glycemic control (A1C <7% and <6.5%, respectively).^20,21 It is important to note here that the ADA does recommend less stringent glycemic control guidelines in very young children (ie, those aged less than 13 years).¹⁶ As is the case for adults with diabetes, pharmacologic therapies typically are needed in addition to lifestyle changes to lower elevated blood glucose levels. However, unlike the situation with adults, the list of drugs approved by the Food and Drug Administration for use in pediatric patients is limited (ie, currently only metformin and insulin preparations are approved for pediatric use).

Metformin was chosen as the initial therapy in this adolescent for several reasons, including demonstrated efficacy and tolerability in pediatric studies, approval for use in this population, and because it is generally not associated with weight gain. The patient needs a 1.7% reduction in A1C; therefore, metformin was titrated to the maximum daily dosage of 2,000 mg (expected to, on average, reduce A1C by 1% to 2%). In a recent, randomized, double-blind, placebo-controlled trial of pediatric patients (aged 10 to 16 years) with type 2 diabetes,²² metformin 1,000 mg bid significantly reduced baseline-adjusted mean A1C values compared with placebo (7.5% vs 8.6%, P<0.001) without negatively affecting body weight or lipid levels. The adverse events profile of metformin in children was similar to that observed in adults, primarily involving gastrointestinal events.

CASE 1 Monotherapy inadequate

After 6 months on metformin therapy, the patient returns for follow-up. His A1C measure has improved to 7.7%, but his weight remains unchanged. At this point, a discussion is initiated about using combination therapy to bring his diabetes under control.

This case study illustrates an important point about the treatment of type 2 diabetes: pharmacologic therapy with a single antidiabetic agent often is insufficient to reach target goals for glycemic control.² Frequently, this reflects the insidious, progressive nature of the disease, which may be present for years before being recognized. As a general rule, combination therapy involves the use of drugs with different mechanisms of action. Clinical trials evaluating combination therapy generally follow this rationale.

Since there are limited pediatric data for the majority of oral agents, the clinical decision regarding which agents to use in combination often is based on the available data in adults. Information regarding efficacy, safety, and tolerability for the different oral antidiabetic drugs used in adults with type 2 diabetes is provided in TABLE 3.

Secretagogues (ie, sulfonylureas and meglitinides) have been reported to typically reduce A1C values by 1% to 2% at maximal doses. Generally, the shorteracting meglitinides have been considered useful for reducing postprandial hyperglycemia and are taken before meals.²³ Unlike traditional sulfonylureas, glimepiride also has demonstrated efficacy in controlling postprandial hyperglycemia.²⁴ All patients starting therapy with a secretagogue should be counseled on recognition of hypoglycemic symptoms and appropriate self-treatment. While severe hypoglycemia is not common in type 2 diabetes, it can occur with any agent that increases insulin secretion.

The glitazones lower A1C levels by an estimated 1% to 1.5%. Although they primarily reduce insulin resistance, they also may have beneficial effects on blood lipids, BP, and inflammatory markers associated with CVD, suggesting a theoretical benefit for reducing macrovascular complications.²⁵ Glitazones, however, have been associated with weight gain and fluid retention, which in adults may unmask or exacerbate congestive heart failure.²⁵ Additionally, although pioglitazone and rosiglitazone do not appear to carry the same risk of hepatotoxicity as the first member of the class, troglitazone, there have been a few case reports of liver injury or failure in patients treated with pioglitazone^26-28 or rosiglitazone.^29-31 In the absence of pediatric and long-term data, caution is warranted with use of these agents in this population.

α-Glucosidase inhibitors are helpful in controlling mealtime glycemic excursions and may have utility in patients who require smaller reductions in A1C. Use of these relatively safe agents is limited due to gastrointestinal side effects (ie, diarrhea and flatulence are common),³² which can prevent titration to optimal doses. Since the primary mechanism of action is inhibition of carbohydrate digestion, patients who are using these agents in combination with therapies that can cause hypoglycemia should be counseled to use simple sugars (ie, glucose tablets) rather than complex carbohydrates to self-treat hypoglycemia.

If oral agents cannot achieve adequate glycemic control, addition of insulin is a rational, well-tolerated, and effective option.^3,4,33 Insulin therapy is often underutilized and delayed in patients with type 2 diabetes who would benefit from earlier introduction of such therapy.³⁴ The pediatric experience with insulin is extensive, but mostly in type 1 diabetes. The potential need for and benefits of insulin therapy may be discussed with patients as early as at the time of diagnosis to help alleviate some of the anxiety patients have regarding this therapy.³⁵

TABLE 3

Available oral antidiabetic agents

Type 2 diabetes in adults

CASE 2 Woman with chronic diabetes

A 60-year-old woman presents for her annual checkup complaining of weight gain, fatigue, and tingling in her feet. Her recent history includes recurrent urinary tract and yeast infections. When initially diagnosed with type 2 diabetes 5 years ago, she had been counseled to increase activity to reduce weight; however, she had been unable to maintain consistent effort. She is 5 ft 4 in tall and weighs 230 lb (BMI, 39.5 kg/m²) with a waist circumference of 40 in. Her weight has fluctuated, but she has had a net gain of 30 lb over the last 5 years. Education and lifestyle modification efforts were reinitiated 1 year ago. The patient is a smoker and has cut her habit to 5 cigarettes per day. Her A1C and total cholesterol levels 1 year ago were 7.8% and 203 mg/dL, respectively. The patient also had cataract surgery 1 year ago. Her current medications include lisinopril 40 mg/d, furosemide 40 mg bid, and rosiglitazone 4 mg/d. She performs SMBG infrequently.

On examination, her BP is 125/72 mm Hg. A random fingerstick blood glucose test shows 210 mg/dL. She exhibits normal monofilament sensation, vibratory sensation, and ankle jerks. She also shows signs of mild pedal edema with no foot lesions and normal pedal pulses.

There are a number of signs suggesting this patient’s diabetes is inadequately controlled, including her complaints of fatigue and repeated yeast infections. Poorly controlled diabetes also is associated with peripheral neuropathy, which may manifest as a tingling sensation or numbness that begins in the feet and moves upward; however, her physical examination is negative for neuropathy.³⁶ A random blood glucose level is of limited value and should not be relied upon as an indicator of the patient’s glycemic status. An A1C measurement is overdue for this patient and will reflect her overall glycemic control in recent months.

A total cholesterol level is rarely adequate for clinical decision making and thus a follow-up lipid profile is warranted. Her BP is well controlled with her current antihypertension regimen. The ADA guidelines recommend adequate treatment of hypertension (target BP, <130/80 mm Hg) in diabetic patients and suggest use of an angiotensin-converting enzyme (ACE) inhibitor or angiotensin-receptor blocker (ARB) because, in addition to their antihypertensive effects, these agents may provide renal benefits for patients with albuminuria or renal insufficiency.²⁰ More typically, a thiazide rather than a loop diuretic is used in combination with an ACE inhibitor (or another antihypertensive drug class) as treatment for hypertension but, since the patient’s BP is well controlled with her current regimen and she has mild pedal edema, the loop diuretic may be maintained.

Her weight gain and peripheral edema may in part be due to rosiglitazone treatment. Edema that is not caused by congestive heart failure is not a contraindication for glitazone use but requires close monitoring and should lead to a consideration of alternate therapy.²⁵

CASE 2 Laboratory assessment

The patient’s glycemic indices are: FPG, 184 mg/dL; A1C, 9.4%; postprandial glucose (PPG), 311 mg/dL. A random urine sample reveals microalbuminuria (protein/creatinine, 38 μg/mg). Her serum creatinine is 1.6 mg/dL. Her lipid panel reveals: total cholesterol, 271 mg/dL; LDL cholesterol, 165 mg/dL; HDL cholesterol, 35 mg/dL; and triglycerides, 355 mg/dL. Liver function tests are within normal limits. Further evaluation indicates that she does not have congestive heart failure.

The laboratory results indicate very poor control of overall and postmeal glucose levels. In addition, she now has microalbuminuria. Although poor adherence with therapy should be ruled out, the loss of control with a previously effective therapy is not unusual and underscores the progressive nature of diabetes.

Collectively, this patient’s poor glycemic control, dyslipidemia and central obesity, place her at a very high risk for CVD.³⁷

CASE 2 Therapy adjusted

Glimepiride 4 mg/d is added to the rosiglitazone; in addition, atorvastatin 20 mg/d and fenofibrate 160 mg/d are prescribed. Lifestyle modifications (eg, dietary changes, exercise, smoking cessation) are reinforced, and the patient is referred to a certified diabetes educator.

The comprehensive approach taken for this patient is consistent with that advocated by the results in the Steno-2 Study. Gaede et al¹ demonstrated that a targeted, intensified, multifactorial, interventional approach to improving macrovascular and microvascular risk factors in patients with type 2 diabetes reduces the risk of macrovascular and microvascular diabetic complications by about 50% compared with conventional treatment.¹ Specifically, patients receiving intensive therapy had a significantly lower risk of CVD, nephropathy, retinopathy and autonomic neuropathy.¹ Intensive, multifactorial therapy involved dietary interventions; a consistent exercise program; smoking cessation; use of ACE inhibitors (or ARBs for patients intolerant to ACE inhibitors) for renal benefits and combined with diuretics and other agents, if necessary, to treat hypertension; lipid-lowering therapy to treat hyperlipidemia (statins, plus fibrates for isolated cases of hypertriglyceridemia); pharmacotherapy for glucose control; daily vitamin-mineral supplements; and daily aspirin as a secondary measure for the prevention of CVD.

In general, pharmacotherapy of diabetes should be individualized, since not all agents are equally appropriate for all patients. A variety of studies have demonstrated that adding a second antidiabetic agent to a first typically results in additional improvements in glycemic control.^38-47 In the example here, a sulfonylurea was chosen because of its complementary mechanism of action with an insulin sensitizer. Since the patient had an elevated serum creatinine, metformin was not considered to be an appropriate choice. In addition, the combination of a glitazone with sulfonylurea therapy has been reported to achieve reductions in A1C^40,48 at least comparable to those reported in analyses of the combination of a glitazone with metformin.^40,45 TABLE 4 provides reported reductions in A1C that have been observed in clinical trials of various combination regimens in type 2 diabetes. Since these are not head-to-head comparisons of the various regimens, the data simply illustrate the range of A1C reductions that may be achieved with combination therapy. In this patient, glimepiride was chosen because of its weight neutral effect,⁴⁹ potentially lower incidence of hypoglycemia,^50,51 favorable effect on postprandial glucose (that may ameliorate cardiovascular risk),^52-56 and once-daily dosing.

Given the progressive natural history of type 2 diabetes, and the fact that this patient currently requires a >2% A1C reduction, it is reasonable to anticipate that she will eventually need insulin to attain glycemic control. Recently, Riddle et al⁴ demonstrated that addition of a basal insulin (neutral protamine Hagedorn [NPH] or glargine) to existing oral agents reduced the A1C to <7% in the majority of patients with type 2 diabetes. Insulin glargine was associated with significantly less hypoglycemia than NPH insulin. This is an important consideration since hypoglycemia remains a major barrier to insulin therapy in type 2 diabetes.⁵⁷ Sulfonylurea therapy should be maintained when insulin is initiated, as this combination has been demonstrated to be highly effective in improving glycemic control and is associated with a low incidence of hypoglycemia.^3,33 In the current patient, the addition of basal insulin glargine would complement her other antidiabetic therapies. Insulin glargine would primarily normalize her FBG, while glimepiride controls PPG and rosiglitazone improves insulin sensitivity. Thus, this regimen would address the 2 most important defects in type 2 diabetes—insulin deficiency and insulin resistance.

Clinical trials have shown that combination therapy with oral agents and insulin, as well as with multiple oral agents, is effective.^4,38,39 However, more long-term and comparative studies of these multiple-agent combinations are needed. It is important to set expectations with patients that gaining good control of diabetes frequently requires combination therapy with multiple agents with the ultimate goal of avoiding the onset of new complications or of delaying progression of existing complications.

TABLE 4

A1C reductions noted in clinical trials and reports of combination therapy for type 2 diabetes

Summary

The increased prevalence of obesity, metabolic syndrome, and type 2 diabetes in adolescents and adults is an ominous sign of more serious disease in the future. A concerted effort on the part of health care professionals to improve the care of patients with type 2 diabetes is needed to bring this burgeoning problem under better control. Diabetes is a recognized coronary risk equivalent; thus, a comprehensive multifactorial approach that rigorously addresses glycemia, as well as elevated BP and lipids, is recommended.

Most patients with type 2 diabetes will eventually require combination therapy with 2 or more agents to attain and maintain glycemic control.^2,3 In particular, combinations of agents with complementary mechanisms of action (eg, an insulin sensitizer with a secretagogue) demonstrate greater improvements in glycemic control. Based on the progressive nature of diabetes, a principle in the pharmacotherapy of glucose control is that, in absence of untoward effects, if a given agent is secondarily unable to provide adequate glycemic control (ie, there was initial improvement in glucose control and then subsequent deterioration), additional agents—whether oral agents or insulin—should be added rather than substituted.

Disclosures:

Dr. Elasy has done consultation for Aventis Pharmaceuticals. Dr. Levy has received grants/research support from Aventis Pharmaceuticals, Eli Lilly and Co., Merck & Co., Novartis Pharmaceuticals, Novo Nordisk Pharmaceuticals Inc., and Pfizer Inc. He is a consultant and on the speaker’s bureau for Aventis Pharmaceuticals, Eli Lilly and Co., Novartis Pharmaceuticals, Novo Nordisk Pharmaceuticals Inc, Pfizer Inc, Takeda Pharmaceuticals America, Inc., Bristol-Myers Squibb Company, GlaxoSmithKline, and Wyeth-Ayerst Pharmaceuticals. Dr. Davis has received research grants from Aventis Pharmaceuticals, Eli Lilly and Co., and Bayer Pharmaceuticals Corporation.

Diabetes affects 18.2 million people in the United States.⁵ It is a leading cause of morbidity and mortality.⁵ Diabetes is associated with more than $90 billion in direct medical costs and with an estimated $40 billion in indirect costs (ie, disability, work loss, and premature mortality).⁵ Type 2 diabetes accounts for 90% to 95% of total cases. It disproportionately affects certain minority populations, and it is increasingly being identified in children and adolescents.^5-7

The obesity epidemic has received tremendous media and public attention and has sobering implications for the development of metabolic syndrome, which is often the precursor to diabetes, and cardiovascular disease (CVD).^8,9 In 1 study, the hazard ratio for development of diabetes was 3.85 for individuals who gained 20 kg or more over approximately 10 years, as compared with individuals whose weight remained relatively stable over the same period.¹⁰ Obesity is associated with elevations in free fatty acid levels and in other compounds which contribute to insulin resistance¹¹ —a key underlying defect in metabolic syndrome and type 2 diabetes. Thus, particularly in genetically predisposed individuals, obesity (through elevated fatty acids) may be the factor that drives insulin resistance to clinical diabetes. In addition, elevated free fatty acid levels also may contribute to the other key underlying defect in type 2 diabetes: progressive pancreatic β-cell loss which results in insulin deficiency.¹² Similar to the trend in adults, childhood obesity has now reached epidemic proportions and has been associated with the increased prevalence of glucose intolerance, metabolic syndrome, and biomarkers of increased CVD risk in this young population.^13-15

Type 2 diabetes in adolescents

CASE 1 Suspected new-onset type 2 diabetes

A 16-year-old Hispanic boy presents for treatment of weight loss. He is a defensive tackle on the high school football team and has a strong appetite. His height is 6 ft 1 in and he weighs 250 lb; his body mass index (BMI) is 32 kg/m², and most of the fat is abdominally distributed (waist circumference, 42 in). Both of his parents are obese, and a grandfather and aunt have type 2 diabetes. Further examination reveals a blood pressure (BP) level of 135/87 mm Hg and acanthosis nigricans. A random fingerstick test shows a blood glucose level of 240 mg/dL.

Until recently, type 2 diabetes rarely was observed in children, occurring in only 1% to 2% of children with new-onset diabetes.¹⁶ However, depending on the sampling strategy, race or ethnicity of the population, and the region of the country sampled, it now has been estimated that type 2 diabetes accounts for anywhere from 8% to 45% of all diabetes reported among children and adolescents.^6,16 Prevalence estimates of pediatric type 2 diabetes in population-based studies range from 4.1 per 1,000 in the National Health and Nutrition Survey (NHANES III) to 50.9 per 1,000 among Pima Indians in Arizona.¹⁷ Unlike the trends of increasing incidence and prevalence of type 2 diabetes in the pediatric population, a similar trend in type 1 diabetes has not been observed.⁷

The diagnosis and treatment of type 2 diabetes in pediatric patients can be challenging. Type 2 diabetes is associated with an increased BMI and is more common in adolescents than in younger children.⁶ Puberty is associated with relative insulin resistance and, along with the time-related effects of obesity, resistance may play a role in the subsequent onset of type 2 diabetes in pediatric patients. However, as an increasing number of children become obese at an earlier age, the onset of disease may occur earlier.

Other risk factors for type 2 diabetes in childhood include having a first- or second-degree relative with the disease or non-European ancestry (Americans of African, Hispanic, Asian, Pacific Islander, or Native American descent). In addition, acanthosis nigricans is a skin condition that serves as a specific, though insensitive, marker of insulin resistance. Among adolescent females, polycystic ovary syndrome has been associated with insulin resistance irrespective of excess weight or frank obesity¹⁸ and may be associated with an increased risk of type 2 diabetes.⁶

TABLE 1 provides criteria for screening for type 2 diabetes in children. The adolescent in the case presented has several risk factors for type 2 diabetes, including obesity, a positive family history, Hispanic descent, and acanthosis nigricans. To more definitively diagnose type 2 diabetes, additional laboratory tests are needed, including assessment of blood glucose and insulin production, as well as the presence of autoantibodies.

TABLE 1

Screening recommendations for type 2 diabetes in children

CASE 1 Laboratory assessment

The patient has a fasting plasma glucose (FPG) level of 215 mg/dL. β-Cell or insulin autoantibodies are not detected. He has a normal-to-high fasting C-peptide level, microalbuminuria, and a serum creatinine level of 0.8 mg/dL. His lipid panel results include: total cholesterol, 234 mg/dL; low-density lipoprotein (LDL) cholesterol, 159 mg/dL; high-density lipoprotein (HDL) cholesterol, 45 mg/dL; and triglycerides, 150 mg/dL.

A comparison of the typical presentation and laboratory findings in pediatric type 1 and type 2 diabetes is listed in TABLE 2. This information underscores that there is significant overlap between the 2 disease classifications.

The laboratory results for this patient confirm a diagnosis of type 2 diabetes. C peptide is a marker of endogenous insulin production and is particularly helpful in the classification of diabetes when it is high, which is indicative of type 2 diabetes. However, it should be noted that a finding of low C-peptide levels does not rule out type 2 diabetes since there is a possibility that glucose toxicity is temporarily limiting insulin production. Indeed, in such a case, insulin therapy can be initiated to establish glycemic control and promote endogenous insulin production by reducing glucose toxicity. Similarly, although the absence of pancreatic β-cells and insulin autoantibodies is consistent with a diagnosis of type 2 diabetes, the presence of such antibodies does not completely rule out type 2 diabetes.¹⁹

Like many patients with type 2 diabetes, the current patient exhibits a cluster of CVD risk factors that characterize the metabolic syndrome, including dyslipidemia (elevated total cholesterol, LDL cholesterol, triglyceride levels, and low HDL cholesterol levels), hypertension, glucose intolerance, hyperinsulinemia, and central or abdominal obesity. This is not surprising as an estimated 30% of overweight adolescents, defined as those with BMI at or above the 95th percentile, meet the criteria for metabolic syndrome.¹⁴

TABLE 2

Classification of type 1 and type 2 diabetes in pediatric patients

CASE 1 Initial therapy

The patient is provided with education on lifestyle modifications to reduce weight and obtain better control of glucose, lipids, and BP. Instructions include ways to improve his eating habits and to increase his physical activity when away from the football field, such as limiting time spent watching TV and playing video games and increasing exercise time and effort. The patient also is instructed on self-monitoring for blood glucose (SMBG) and is given a referral to an ophthalmologist to determine whether there is any baseline retinopathy.

At his 3-month follow-up, the patient’s A1C is 8.7%, and his weight has not changed. At this point, lifestyle modifications are reinforced, and pharmacotherapy with extended-release metformin is initiated and titrated to 2,000 mg/d.

Typically, the first approach to managing type 2 diabetes involves dietary changes and instituting an exercise program to reduce weight and to improve insulin sensitivity. This was the approach used in the patient described above. However, as is frequently the case, lifestyle modifications by themselves are insufficient to meet the American Diabetes Association (ADA) or American Association of Clinical Endocrinologists targets for glycemic control (A1C <7% and <6.5%, respectively).^20,21 It is important to note here that the ADA does recommend less stringent glycemic control guidelines in very young children (ie, those aged less than 13 years).¹⁶ As is the case for adults with diabetes, pharmacologic therapies typically are needed in addition to lifestyle changes to lower elevated blood glucose levels. However, unlike the situation with adults, the list of drugs approved by the Food and Drug Administration for use in pediatric patients is limited (ie, currently only metformin and insulin preparations are approved for pediatric use).

Metformin was chosen as the initial therapy in this adolescent for several reasons, including demonstrated efficacy and tolerability in pediatric studies, approval for use in this population, and because it is generally not associated with weight gain. The patient needs a 1.7% reduction in A1C; therefore, metformin was titrated to the maximum daily dosage of 2,000 mg (expected to, on average, reduce A1C by 1% to 2%). In a recent, randomized, double-blind, placebo-controlled trial of pediatric patients (aged 10 to 16 years) with type 2 diabetes,²² metformin 1,000 mg bid significantly reduced baseline-adjusted mean A1C values compared with placebo (7.5% vs 8.6%, P<0.001) without negatively affecting body weight or lipid levels. The adverse events profile of metformin in children was similar to that observed in adults, primarily involving gastrointestinal events.

CASE 1 Monotherapy inadequate

After 6 months on metformin therapy, the patient returns for follow-up. His A1C measure has improved to 7.7%, but his weight remains unchanged. At this point, a discussion is initiated about using combination therapy to bring his diabetes under control.

This case study illustrates an important point about the treatment of type 2 diabetes: pharmacologic therapy with a single antidiabetic agent often is insufficient to reach target goals for glycemic control.² Frequently, this reflects the insidious, progressive nature of the disease, which may be present for years before being recognized. As a general rule, combination therapy involves the use of drugs with different mechanisms of action. Clinical trials evaluating combination therapy generally follow this rationale.

Since there are limited pediatric data for the majority of oral agents, the clinical decision regarding which agents to use in combination often is based on the available data in adults. Information regarding efficacy, safety, and tolerability for the different oral antidiabetic drugs used in adults with type 2 diabetes is provided in TABLE 3.

Secretagogues (ie, sulfonylureas and meglitinides) have been reported to typically reduce A1C values by 1% to 2% at maximal doses. Generally, the shorteracting meglitinides have been considered useful for reducing postprandial hyperglycemia and are taken before meals.²³ Unlike traditional sulfonylureas, glimepiride also has demonstrated efficacy in controlling postprandial hyperglycemia.²⁴ All patients starting therapy with a secretagogue should be counseled on recognition of hypoglycemic symptoms and appropriate self-treatment. While severe hypoglycemia is not common in type 2 diabetes, it can occur with any agent that increases insulin secretion.

The glitazones lower A1C levels by an estimated 1% to 1.5%. Although they primarily reduce insulin resistance, they also may have beneficial effects on blood lipids, BP, and inflammatory markers associated with CVD, suggesting a theoretical benefit for reducing macrovascular complications.²⁵ Glitazones, however, have been associated with weight gain and fluid retention, which in adults may unmask or exacerbate congestive heart failure.²⁵ Additionally, although pioglitazone and rosiglitazone do not appear to carry the same risk of hepatotoxicity as the first member of the class, troglitazone, there have been a few case reports of liver injury or failure in patients treated with pioglitazone^26-28 or rosiglitazone.^29-31 In the absence of pediatric and long-term data, caution is warranted with use of these agents in this population.

α-Glucosidase inhibitors are helpful in controlling mealtime glycemic excursions and may have utility in patients who require smaller reductions in A1C. Use of these relatively safe agents is limited due to gastrointestinal side effects (ie, diarrhea and flatulence are common),³² which can prevent titration to optimal doses. Since the primary mechanism of action is inhibition of carbohydrate digestion, patients who are using these agents in combination with therapies that can cause hypoglycemia should be counseled to use simple sugars (ie, glucose tablets) rather than complex carbohydrates to self-treat hypoglycemia.

If oral agents cannot achieve adequate glycemic control, addition of insulin is a rational, well-tolerated, and effective option.^3,4,33 Insulin therapy is often underutilized and delayed in patients with type 2 diabetes who would benefit from earlier introduction of such therapy.³⁴ The pediatric experience with insulin is extensive, but mostly in type 1 diabetes. The potential need for and benefits of insulin therapy may be discussed with patients as early as at the time of diagnosis to help alleviate some of the anxiety patients have regarding this therapy.³⁵

TABLE 3

Available oral antidiabetic agents

Type 2 diabetes in adults

CASE 2 Woman with chronic diabetes

A 60-year-old woman presents for her annual checkup complaining of weight gain, fatigue, and tingling in her feet. Her recent history includes recurrent urinary tract and yeast infections. When initially diagnosed with type 2 diabetes 5 years ago, she had been counseled to increase activity to reduce weight; however, she had been unable to maintain consistent effort. She is 5 ft 4 in tall and weighs 230 lb (BMI, 39.5 kg/m²) with a waist circumference of 40 in. Her weight has fluctuated, but she has had a net gain of 30 lb over the last 5 years. Education and lifestyle modification efforts were reinitiated 1 year ago. The patient is a smoker and has cut her habit to 5 cigarettes per day. Her A1C and total cholesterol levels 1 year ago were 7.8% and 203 mg/dL, respectively. The patient also had cataract surgery 1 year ago. Her current medications include lisinopril 40 mg/d, furosemide 40 mg bid, and rosiglitazone 4 mg/d. She performs SMBG infrequently.

On examination, her BP is 125/72 mm Hg. A random fingerstick blood glucose test shows 210 mg/dL. She exhibits normal monofilament sensation, vibratory sensation, and ankle jerks. She also shows signs of mild pedal edema with no foot lesions and normal pedal pulses.

There are a number of signs suggesting this patient’s diabetes is inadequately controlled, including her complaints of fatigue and repeated yeast infections. Poorly controlled diabetes also is associated with peripheral neuropathy, which may manifest as a tingling sensation or numbness that begins in the feet and moves upward; however, her physical examination is negative for neuropathy.³⁶ A random blood glucose level is of limited value and should not be relied upon as an indicator of the patient’s glycemic status. An A1C measurement is overdue for this patient and will reflect her overall glycemic control in recent months.

A total cholesterol level is rarely adequate for clinical decision making and thus a follow-up lipid profile is warranted. Her BP is well controlled with her current antihypertension regimen. The ADA guidelines recommend adequate treatment of hypertension (target BP, <130/80 mm Hg) in diabetic patients and suggest use of an angiotensin-converting enzyme (ACE) inhibitor or angiotensin-receptor blocker (ARB) because, in addition to their antihypertensive effects, these agents may provide renal benefits for patients with albuminuria or renal insufficiency.²⁰ More typically, a thiazide rather than a loop diuretic is used in combination with an ACE inhibitor (or another antihypertensive drug class) as treatment for hypertension but, since the patient’s BP is well controlled with her current regimen and she has mild pedal edema, the loop diuretic may be maintained.

Her weight gain and peripheral edema may in part be due to rosiglitazone treatment. Edema that is not caused by congestive heart failure is not a contraindication for glitazone use but requires close monitoring and should lead to a consideration of alternate therapy.²⁵

CASE 2 Laboratory assessment

The patient’s glycemic indices are: FPG, 184 mg/dL; A1C, 9.4%; postprandial glucose (PPG), 311 mg/dL. A random urine sample reveals microalbuminuria (protein/creatinine, 38 μg/mg). Her serum creatinine is 1.6 mg/dL. Her lipid panel reveals: total cholesterol, 271 mg/dL; LDL cholesterol, 165 mg/dL; HDL cholesterol, 35 mg/dL; and triglycerides, 355 mg/dL. Liver function tests are within normal limits. Further evaluation indicates that she does not have congestive heart failure.

The laboratory results indicate very poor control of overall and postmeal glucose levels. In addition, she now has microalbuminuria. Although poor adherence with therapy should be ruled out, the loss of control with a previously effective therapy is not unusual and underscores the progressive nature of diabetes.

Collectively, this patient’s poor glycemic control, dyslipidemia and central obesity, place her at a very high risk for CVD.³⁷

CASE 2 Therapy adjusted

Glimepiride 4 mg/d is added to the rosiglitazone; in addition, atorvastatin 20 mg/d and fenofibrate 160 mg/d are prescribed. Lifestyle modifications (eg, dietary changes, exercise, smoking cessation) are reinforced, and the patient is referred to a certified diabetes educator.

The comprehensive approach taken for this patient is consistent with that advocated by the results in the Steno-2 Study. Gaede et al¹ demonstrated that a targeted, intensified, multifactorial, interventional approach to improving macrovascular and microvascular risk factors in patients with type 2 diabetes reduces the risk of macrovascular and microvascular diabetic complications by about 50% compared with conventional treatment.¹ Specifically, patients receiving intensive therapy had a significantly lower risk of CVD, nephropathy, retinopathy and autonomic neuropathy.¹ Intensive, multifactorial therapy involved dietary interventions; a consistent exercise program; smoking cessation; use of ACE inhibitors (or ARBs for patients intolerant to ACE inhibitors) for renal benefits and combined with diuretics and other agents, if necessary, to treat hypertension; lipid-lowering therapy to treat hyperlipidemia (statins, plus fibrates for isolated cases of hypertriglyceridemia); pharmacotherapy for glucose control; daily vitamin-mineral supplements; and daily aspirin as a secondary measure for the prevention of CVD.

In general, pharmacotherapy of diabetes should be individualized, since not all agents are equally appropriate for all patients. A variety of studies have demonstrated that adding a second antidiabetic agent to a first typically results in additional improvements in glycemic control.^38-47 In the example here, a sulfonylurea was chosen because of its complementary mechanism of action with an insulin sensitizer. Since the patient had an elevated serum creatinine, metformin was not considered to be an appropriate choice. In addition, the combination of a glitazone with sulfonylurea therapy has been reported to achieve reductions in A1C^40,48 at least comparable to those reported in analyses of the combination of a glitazone with metformin.^40,45 TABLE 4 provides reported reductions in A1C that have been observed in clinical trials of various combination regimens in type 2 diabetes. Since these are not head-to-head comparisons of the various regimens, the data simply illustrate the range of A1C reductions that may be achieved with combination therapy. In this patient, glimepiride was chosen because of its weight neutral effect,⁴⁹ potentially lower incidence of hypoglycemia,^50,51 favorable effect on postprandial glucose (that may ameliorate cardiovascular risk),^52-56 and once-daily dosing.

Given the progressive natural history of type 2 diabetes, and the fact that this patient currently requires a >2% A1C reduction, it is reasonable to anticipate that she will eventually need insulin to attain glycemic control. Recently, Riddle et al⁴ demonstrated that addition of a basal insulin (neutral protamine Hagedorn [NPH] or glargine) to existing oral agents reduced the A1C to <7% in the majority of patients with type 2 diabetes. Insulin glargine was associated with significantly less hypoglycemia than NPH insulin. This is an important consideration since hypoglycemia remains a major barrier to insulin therapy in type 2 diabetes.⁵⁷ Sulfonylurea therapy should be maintained when insulin is initiated, as this combination has been demonstrated to be highly effective in improving glycemic control and is associated with a low incidence of hypoglycemia.^3,33 In the current patient, the addition of basal insulin glargine would complement her other antidiabetic therapies. Insulin glargine would primarily normalize her FBG, while glimepiride controls PPG and rosiglitazone improves insulin sensitivity. Thus, this regimen would address the 2 most important defects in type 2 diabetes—insulin deficiency and insulin resistance.

Clinical trials have shown that combination therapy with oral agents and insulin, as well as with multiple oral agents, is effective.^4,38,39 However, more long-term and comparative studies of these multiple-agent combinations are needed. It is important to set expectations with patients that gaining good control of diabetes frequently requires combination therapy with multiple agents with the ultimate goal of avoiding the onset of new complications or of delaying progression of existing complications.

TABLE 4

A1C reductions noted in clinical trials and reports of combination therapy for type 2 diabetes

Summary

The increased prevalence of obesity, metabolic syndrome, and type 2 diabetes in adolescents and adults is an ominous sign of more serious disease in the future. A concerted effort on the part of health care professionals to improve the care of patients with type 2 diabetes is needed to bring this burgeoning problem under better control. Diabetes is a recognized coronary risk equivalent; thus, a comprehensive multifactorial approach that rigorously addresses glycemia, as well as elevated BP and lipids, is recommended.

Most patients with type 2 diabetes will eventually require combination therapy with 2 or more agents to attain and maintain glycemic control.^2,3 In particular, combinations of agents with complementary mechanisms of action (eg, an insulin sensitizer with a secretagogue) demonstrate greater improvements in glycemic control. Based on the progressive nature of diabetes, a principle in the pharmacotherapy of glucose control is that, in absence of untoward effects, if a given agent is secondarily unable to provide adequate glycemic control (ie, there was initial improvement in glucose control and then subsequent deterioration), additional agents—whether oral agents or insulin—should be added rather than substituted.

Disclosures:

Dr. Elasy has done consultation for Aventis Pharmaceuticals. Dr. Levy has received grants/research support from Aventis Pharmaceuticals, Eli Lilly and Co., Merck & Co., Novartis Pharmaceuticals, Novo Nordisk Pharmaceuticals Inc., and Pfizer Inc. He is a consultant and on the speaker’s bureau for Aventis Pharmaceuticals, Eli Lilly and Co., Novartis Pharmaceuticals, Novo Nordisk Pharmaceuticals Inc, Pfizer Inc, Takeda Pharmaceuticals America, Inc., Bristol-Myers Squibb Company, GlaxoSmithKline, and Wyeth-Ayerst Pharmaceuticals. Dr. Davis has received research grants from Aventis Pharmaceuticals, Eli Lilly and Co., and Bayer Pharmaceuticals Corporation.

Diabetes affects 18.2 million people in the United States.⁵ It is a leading cause of morbidity and mortality.⁵ Diabetes is associated with more than $90 billion in direct medical costs and with an estimated $40 billion in indirect costs (ie, disability, work loss, and premature mortality).⁵ Type 2 diabetes accounts for 90% to 95% of total cases. It disproportionately affects certain minority populations, and it is increasingly being identified in children and adolescents.^5-7

The obesity epidemic has received tremendous media and public attention and has sobering implications for the development of metabolic syndrome, which is often the precursor to diabetes, and cardiovascular disease (CVD).^8,9 In 1 study, the hazard ratio for development of diabetes was 3.85 for individuals who gained 20 kg or more over approximately 10 years, as compared with individuals whose weight remained relatively stable over the same period.¹⁰ Obesity is associated with elevations in free fatty acid levels and in other compounds which contribute to insulin resistance¹¹ —a key underlying defect in metabolic syndrome and type 2 diabetes. Thus, particularly in genetically predisposed individuals, obesity (through elevated fatty acids) may be the factor that drives insulin resistance to clinical diabetes. In addition, elevated free fatty acid levels also may contribute to the other key underlying defect in type 2 diabetes: progressive pancreatic β-cell loss which results in insulin deficiency.¹² Similar to the trend in adults, childhood obesity has now reached epidemic proportions and has been associated with the increased prevalence of glucose intolerance, metabolic syndrome, and biomarkers of increased CVD risk in this young population.^13-15

Type 2 diabetes in adolescents

CASE 1 Suspected new-onset type 2 diabetes

A 16-year-old Hispanic boy presents for treatment of weight loss. He is a defensive tackle on the high school football team and has a strong appetite. His height is 6 ft 1 in and he weighs 250 lb; his body mass index (BMI) is 32 kg/m², and most of the fat is abdominally distributed (waist circumference, 42 in). Both of his parents are obese, and a grandfather and aunt have type 2 diabetes. Further examination reveals a blood pressure (BP) level of 135/87 mm Hg and acanthosis nigricans. A random fingerstick test shows a blood glucose level of 240 mg/dL.

Until recently, type 2 diabetes rarely was observed in children, occurring in only 1% to 2% of children with new-onset diabetes.¹⁶ However, depending on the sampling strategy, race or ethnicity of the population, and the region of the country sampled, it now has been estimated that type 2 diabetes accounts for anywhere from 8% to 45% of all diabetes reported among children and adolescents.^6,16 Prevalence estimates of pediatric type 2 diabetes in population-based studies range from 4.1 per 1,000 in the National Health and Nutrition Survey (NHANES III) to 50.9 per 1,000 among Pima Indians in Arizona.¹⁷ Unlike the trends of increasing incidence and prevalence of type 2 diabetes in the pediatric population, a similar trend in type 1 diabetes has not been observed.⁷

The diagnosis and treatment of type 2 diabetes in pediatric patients can be challenging. Type 2 diabetes is associated with an increased BMI and is more common in adolescents than in younger children.⁶ Puberty is associated with relative insulin resistance and, along with the time-related effects of obesity, resistance may play a role in the subsequent onset of type 2 diabetes in pediatric patients. However, as an increasing number of children become obese at an earlier age, the onset of disease may occur earlier.

Other risk factors for type 2 diabetes in childhood include having a first- or second-degree relative with the disease or non-European ancestry (Americans of African, Hispanic, Asian, Pacific Islander, or Native American descent). In addition, acanthosis nigricans is a skin condition that serves as a specific, though insensitive, marker of insulin resistance. Among adolescent females, polycystic ovary syndrome has been associated with insulin resistance irrespective of excess weight or frank obesity¹⁸ and may be associated with an increased risk of type 2 diabetes.⁶

TABLE 1 provides criteria for screening for type 2 diabetes in children. The adolescent in the case presented has several risk factors for type 2 diabetes, including obesity, a positive family history, Hispanic descent, and acanthosis nigricans. To more definitively diagnose type 2 diabetes, additional laboratory tests are needed, including assessment of blood glucose and insulin production, as well as the presence of autoantibodies.

TABLE 1

Screening recommendations for type 2 diabetes in children

CASE 1 Laboratory assessment

The patient has a fasting plasma glucose (FPG) level of 215 mg/dL. β-Cell or insulin autoantibodies are not detected. He has a normal-to-high fasting C-peptide level, microalbuminuria, and a serum creatinine level of 0.8 mg/dL. His lipid panel results include: total cholesterol, 234 mg/dL; low-density lipoprotein (LDL) cholesterol, 159 mg/dL; high-density lipoprotein (HDL) cholesterol, 45 mg/dL; and triglycerides, 150 mg/dL.

A comparison of the typical presentation and laboratory findings in pediatric type 1 and type 2 diabetes is listed in TABLE 2. This information underscores that there is significant overlap between the 2 disease classifications.

The laboratory results for this patient confirm a diagnosis of type 2 diabetes. C peptide is a marker of endogenous insulin production and is particularly helpful in the classification of diabetes when it is high, which is indicative of type 2 diabetes. However, it should be noted that a finding of low C-peptide levels does not rule out type 2 diabetes since there is a possibility that glucose toxicity is temporarily limiting insulin production. Indeed, in such a case, insulin therapy can be initiated to establish glycemic control and promote endogenous insulin production by reducing glucose toxicity. Similarly, although the absence of pancreatic β-cells and insulin autoantibodies is consistent with a diagnosis of type 2 diabetes, the presence of such antibodies does not completely rule out type 2 diabetes.¹⁹

Like many patients with type 2 diabetes, the current patient exhibits a cluster of CVD risk factors that characterize the metabolic syndrome, including dyslipidemia (elevated total cholesterol, LDL cholesterol, triglyceride levels, and low HDL cholesterol levels), hypertension, glucose intolerance, hyperinsulinemia, and central or abdominal obesity. This is not surprising as an estimated 30% of overweight adolescents, defined as those with BMI at or above the 95th percentile, meet the criteria for metabolic syndrome.¹⁴

TABLE 2

Classification of type 1 and type 2 diabetes in pediatric patients

CASE 1 Initial therapy

The patient is provided with education on lifestyle modifications to reduce weight and obtain better control of glucose, lipids, and BP. Instructions include ways to improve his eating habits and to increase his physical activity when away from the football field, such as limiting time spent watching TV and playing video games and increasing exercise time and effort. The patient also is instructed on self-monitoring for blood glucose (SMBG) and is given a referral to an ophthalmologist to determine whether there is any baseline retinopathy.

At his 3-month follow-up, the patient’s A1C is 8.7%, and his weight has not changed. At this point, lifestyle modifications are reinforced, and pharmacotherapy with extended-release metformin is initiated and titrated to 2,000 mg/d.

Typically, the first approach to managing type 2 diabetes involves dietary changes and instituting an exercise program to reduce weight and to improve insulin sensitivity. This was the approach used in the patient described above. However, as is frequently the case, lifestyle modifications by themselves are insufficient to meet the American Diabetes Association (ADA) or American Association of Clinical Endocrinologists targets for glycemic control (A1C <7% and <6.5%, respectively).^20,21 It is important to note here that the ADA does recommend less stringent glycemic control guidelines in very young children (ie, those aged less than 13 years).¹⁶ As is the case for adults with diabetes, pharmacologic therapies typically are needed in addition to lifestyle changes to lower elevated blood glucose levels. However, unlike the situation with adults, the list of drugs approved by the Food and Drug Administration for use in pediatric patients is limited (ie, currently only metformin and insulin preparations are approved for pediatric use).

Metformin was chosen as the initial therapy in this adolescent for several reasons, including demonstrated efficacy and tolerability in pediatric studies, approval for use in this population, and because it is generally not associated with weight gain. The patient needs a 1.7% reduction in A1C; therefore, metformin was titrated to the maximum daily dosage of 2,000 mg (expected to, on average, reduce A1C by 1% to 2%). In a recent, randomized, double-blind, placebo-controlled trial of pediatric patients (aged 10 to 16 years) with type 2 diabetes,²² metformin 1,000 mg bid significantly reduced baseline-adjusted mean A1C values compared with placebo (7.5% vs 8.6%, P<0.001) without negatively affecting body weight or lipid levels. The adverse events profile of metformin in children was similar to that observed in adults, primarily involving gastrointestinal events.

CASE 1 Monotherapy inadequate

After 6 months on metformin therapy, the patient returns for follow-up. His A1C measure has improved to 7.7%, but his weight remains unchanged. At this point, a discussion is initiated about using combination therapy to bring his diabetes under control.

This case study illustrates an important point about the treatment of type 2 diabetes: pharmacologic therapy with a single antidiabetic agent often is insufficient to reach target goals for glycemic control.² Frequently, this reflects the insidious, progressive nature of the disease, which may be present for years before being recognized. As a general rule, combination therapy involves the use of drugs with different mechanisms of action. Clinical trials evaluating combination therapy generally follow this rationale.

Since there are limited pediatric data for the majority of oral agents, the clinical decision regarding which agents to use in combination often is based on the available data in adults. Information regarding efficacy, safety, and tolerability for the different oral antidiabetic drugs used in adults with type 2 diabetes is provided in TABLE 3.

Secretagogues (ie, sulfonylureas and meglitinides) have been reported to typically reduce A1C values by 1% to 2% at maximal doses. Generally, the shorteracting meglitinides have been considered useful for reducing postprandial hyperglycemia and are taken before meals.²³ Unlike traditional sulfonylureas, glimepiride also has demonstrated efficacy in controlling postprandial hyperglycemia.²⁴ All patients starting therapy with a secretagogue should be counseled on recognition of hypoglycemic symptoms and appropriate self-treatment. While severe hypoglycemia is not common in type 2 diabetes, it can occur with any agent that increases insulin secretion.

The glitazones lower A1C levels by an estimated 1% to 1.5%. Although they primarily reduce insulin resistance, they also may have beneficial effects on blood lipids, BP, and inflammatory markers associated with CVD, suggesting a theoretical benefit for reducing macrovascular complications.²⁵ Glitazones, however, have been associated with weight gain and fluid retention, which in adults may unmask or exacerbate congestive heart failure.²⁵ Additionally, although pioglitazone and rosiglitazone do not appear to carry the same risk of hepatotoxicity as the first member of the class, troglitazone, there have been a few case reports of liver injury or failure in patients treated with pioglitazone^26-28 or rosiglitazone.^29-31 In the absence of pediatric and long-term data, caution is warranted with use of these agents in this population.

α-Glucosidase inhibitors are helpful in controlling mealtime glycemic excursions and may have utility in patients who require smaller reductions in A1C. Use of these relatively safe agents is limited due to gastrointestinal side effects (ie, diarrhea and flatulence are common),³² which can prevent titration to optimal doses. Since the primary mechanism of action is inhibition of carbohydrate digestion, patients who are using these agents in combination with therapies that can cause hypoglycemia should be counseled to use simple sugars (ie, glucose tablets) rather than complex carbohydrates to self-treat hypoglycemia.

If oral agents cannot achieve adequate glycemic control, addition of insulin is a rational, well-tolerated, and effective option.^3,4,33 Insulin therapy is often underutilized and delayed in patients with type 2 diabetes who would benefit from earlier introduction of such therapy.³⁴ The pediatric experience with insulin is extensive, but mostly in type 1 diabetes. The potential need for and benefits of insulin therapy may be discussed with patients as early as at the time of diagnosis to help alleviate some of the anxiety patients have regarding this therapy.³⁵

TABLE 3

Available oral antidiabetic agents

Type 2 diabetes in adults

CASE 2 Woman with chronic diabetes

A 60-year-old woman presents for her annual checkup complaining of weight gain, fatigue, and tingling in her feet. Her recent history includes recurrent urinary tract and yeast infections. When initially diagnosed with type 2 diabetes 5 years ago, she had been counseled to increase activity to reduce weight; however, she had been unable to maintain consistent effort. She is 5 ft 4 in tall and weighs 230 lb (BMI, 39.5 kg/m²) with a waist circumference of 40 in. Her weight has fluctuated, but she has had a net gain of 30 lb over the last 5 years. Education and lifestyle modification efforts were reinitiated 1 year ago. The patient is a smoker and has cut her habit to 5 cigarettes per day. Her A1C and total cholesterol levels 1 year ago were 7.8% and 203 mg/dL, respectively. The patient also had cataract surgery 1 year ago. Her current medications include lisinopril 40 mg/d, furosemide 40 mg bid, and rosiglitazone 4 mg/d. She performs SMBG infrequently.

On examination, her BP is 125/72 mm Hg. A random fingerstick blood glucose test shows 210 mg/dL. She exhibits normal monofilament sensation, vibratory sensation, and ankle jerks. She also shows signs of mild pedal edema with no foot lesions and normal pedal pulses.

There are a number of signs suggesting this patient’s diabetes is inadequately controlled, including her complaints of fatigue and repeated yeast infections. Poorly controlled diabetes also is associated with peripheral neuropathy, which may manifest as a tingling sensation or numbness that begins in the feet and moves upward; however, her physical examination is negative for neuropathy.³⁶ A random blood glucose level is of limited value and should not be relied upon as an indicator of the patient’s glycemic status. An A1C measurement is overdue for this patient and will reflect her overall glycemic control in recent months.

A total cholesterol level is rarely adequate for clinical decision making and thus a follow-up lipid profile is warranted. Her BP is well controlled with her current antihypertension regimen. The ADA guidelines recommend adequate treatment of hypertension (target BP, <130/80 mm Hg) in diabetic patients and suggest use of an angiotensin-converting enzyme (ACE) inhibitor or angiotensin-receptor blocker (ARB) because, in addition to their antihypertensive effects, these agents may provide renal benefits for patients with albuminuria or renal insufficiency.²⁰ More typically, a thiazide rather than a loop diuretic is used in combination with an ACE inhibitor (or another antihypertensive drug class) as treatment for hypertension but, since the patient’s BP is well controlled with her current regimen and she has mild pedal edema, the loop diuretic may be maintained.

Her weight gain and peripheral edema may in part be due to rosiglitazone treatment. Edema that is not caused by congestive heart failure is not a contraindication for glitazone use but requires close monitoring and should lead to a consideration of alternate therapy.²⁵

CASE 2 Laboratory assessment

The patient’s glycemic indices are: FPG, 184 mg/dL; A1C, 9.4%; postprandial glucose (PPG), 311 mg/dL. A random urine sample reveals microalbuminuria (protein/creatinine, 38 μg/mg). Her serum creatinine is 1.6 mg/dL. Her lipid panel reveals: total cholesterol, 271 mg/dL; LDL cholesterol, 165 mg/dL; HDL cholesterol, 35 mg/dL; and triglycerides, 355 mg/dL. Liver function tests are within normal limits. Further evaluation indicates that she does not have congestive heart failure.

The laboratory results indicate very poor control of overall and postmeal glucose levels. In addition, she now has microalbuminuria. Although poor adherence with therapy should be ruled out, the loss of control with a previously effective therapy is not unusual and underscores the progressive nature of diabetes.

Collectively, this patient’s poor glycemic control, dyslipidemia and central obesity, place her at a very high risk for CVD.³⁷

CASE 2 Therapy adjusted

Glimepiride 4 mg/d is added to the rosiglitazone; in addition, atorvastatin 20 mg/d and fenofibrate 160 mg/d are prescribed. Lifestyle modifications (eg, dietary changes, exercise, smoking cessation) are reinforced, and the patient is referred to a certified diabetes educator.

The comprehensive approach taken for this patient is consistent with that advocated by the results in the Steno-2 Study. Gaede et al¹ demonstrated that a targeted, intensified, multifactorial, interventional approach to improving macrovascular and microvascular risk factors in patients with type 2 diabetes reduces the risk of macrovascular and microvascular diabetic complications by about 50% compared with conventional treatment.¹ Specifically, patients receiving intensive therapy had a significantly lower risk of CVD, nephropathy, retinopathy and autonomic neuropathy.¹ Intensive, multifactorial therapy involved dietary interventions; a consistent exercise program; smoking cessation; use of ACE inhibitors (or ARBs for patients intolerant to ACE inhibitors) for renal benefits and combined with diuretics and other agents, if necessary, to treat hypertension; lipid-lowering therapy to treat hyperlipidemia (statins, plus fibrates for isolated cases of hypertriglyceridemia); pharmacotherapy for glucose control; daily vitamin-mineral supplements; and daily aspirin as a secondary measure for the prevention of CVD.

In general, pharmacotherapy of diabetes should be individualized, since not all agents are equally appropriate for all patients. A variety of studies have demonstrated that adding a second antidiabetic agent to a first typically results in additional improvements in glycemic control.^38-47 In the example here, a sulfonylurea was chosen because of its complementary mechanism of action with an insulin sensitizer. Since the patient had an elevated serum creatinine, metformin was not considered to be an appropriate choice. In addition, the combination of a glitazone with sulfonylurea therapy has been reported to achieve reductions in A1C^40,48 at least comparable to those reported in analyses of the combination of a glitazone with metformin.^40,45 TABLE 4 provides reported reductions in A1C that have been observed in clinical trials of various combination regimens in type 2 diabetes. Since these are not head-to-head comparisons of the various regimens, the data simply illustrate the range of A1C reductions that may be achieved with combination therapy. In this patient, glimepiride was chosen because of its weight neutral effect,⁴⁹ potentially lower incidence of hypoglycemia,^50,51 favorable effect on postprandial glucose (that may ameliorate cardiovascular risk),^52-56 and once-daily dosing.

Given the progressive natural history of type 2 diabetes, and the fact that this patient currently requires a >2% A1C reduction, it is reasonable to anticipate that she will eventually need insulin to attain glycemic control. Recently, Riddle et al⁴ demonstrated that addition of a basal insulin (neutral protamine Hagedorn [NPH] or glargine) to existing oral agents reduced the A1C to <7% in the majority of patients with type 2 diabetes. Insulin glargine was associated with significantly less hypoglycemia than NPH insulin. This is an important consideration since hypoglycemia remains a major barrier to insulin therapy in type 2 diabetes.⁵⁷ Sulfonylurea therapy should be maintained when insulin is initiated, as this combination has been demonstrated to be highly effective in improving glycemic control and is associated with a low incidence of hypoglycemia.^3,33 In the current patient, the addition of basal insulin glargine would complement her other antidiabetic therapies. Insulin glargine would primarily normalize her FBG, while glimepiride controls PPG and rosiglitazone improves insulin sensitivity. Thus, this regimen would address the 2 most important defects in type 2 diabetes—insulin deficiency and insulin resistance.

Clinical trials have shown that combination therapy with oral agents and insulin, as well as with multiple oral agents, is effective.^4,38,39 However, more long-term and comparative studies of these multiple-agent combinations are needed. It is important to set expectations with patients that gaining good control of diabetes frequently requires combination therapy with multiple agents with the ultimate goal of avoiding the onset of new complications or of delaying progression of existing complications.

TABLE 4

A1C reductions noted in clinical trials and reports of combination therapy for type 2 diabetes

Summary

The increased prevalence of obesity, metabolic syndrome, and type 2 diabetes in adolescents and adults is an ominous sign of more serious disease in the future. A concerted effort on the part of health care professionals to improve the care of patients with type 2 diabetes is needed to bring this burgeoning problem under better control. Diabetes is a recognized coronary risk equivalent; thus, a comprehensive multifactorial approach that rigorously addresses glycemia, as well as elevated BP and lipids, is recommended.

Most patients with type 2 diabetes will eventually require combination therapy with 2 or more agents to attain and maintain glycemic control.^2,3 In particular, combinations of agents with complementary mechanisms of action (eg, an insulin sensitizer with a secretagogue) demonstrate greater improvements in glycemic control. Based on the progressive nature of diabetes, a principle in the pharmacotherapy of glucose control is that, in absence of untoward effects, if a given agent is secondarily unable to provide adequate glycemic control (ie, there was initial improvement in glucose control and then subsequent deterioration), additional agents—whether oral agents or insulin—should be added rather than substituted.

Disclosures:

Dr. Elasy has done consultation for Aventis Pharmaceuticals. Dr. Levy has received grants/research support from Aventis Pharmaceuticals, Eli Lilly and Co., Merck & Co., Novartis Pharmaceuticals, Novo Nordisk Pharmaceuticals Inc., and Pfizer Inc. He is a consultant and on the speaker’s bureau for Aventis Pharmaceuticals, Eli Lilly and Co., Novartis Pharmaceuticals, Novo Nordisk Pharmaceuticals Inc, Pfizer Inc, Takeda Pharmaceuticals America, Inc., Bristol-Myers Squibb Company, GlaxoSmithKline, and Wyeth-Ayerst Pharmaceuticals. Dr. Davis has received research grants from Aventis Pharmaceuticals, Eli Lilly and Co., and Bayer Pharmaceuticals Corporation.

In the United States, most ankle injuries are evaluated radiographically, ^1,2 even though only about 15% are found to involve fractures. ³ An estimated 6 million ankle radiographs are performed annually in the US and Canada, costing approximately $300 million dollars (US). The Ottawa Ankle Rules can significantly decrease the number of unnecessary ankle radiographs.

The rules are not a substitute for sound clinical judgment, but augment findings in the history and physical examination to help the clinician determine the appropriateness of ankle films. If the rules are met and radiographs are avoided, it is unlikely, especially with good communication and follow-up, that a patient will turn out to have a significant fracture. Moreover, discussing physical examination findings and the reasons for doing or not doing radiographs will enhance patient-physician communication. As we report in this article, patients’ satisfaction with care seems not to depend on whether x-ray films are ordered.

Applying the Ottawa Ankle Rules

To address the problem of low positive predictive value of ankle radiographs, Stiell and colleagues⁴ at the University of Ottawa and the Ontario Health Ministry developed a set of clinical parameters in the early 1990s to evaluate the need for ankle and midfoot radiographs. Their criteria were based on a multivariable data analysis involving a large number of clinical variables associated with ankle injuries. The resulting rules (Figure) have been shown to decrease the need for films by about 30%.⁴ If a patient does not exhibit any of the criteria, radiographs of the foot or ankle are not needed after trauma (strength of recommendation [SOR]: A).

FIGURE
Ottawa Ankle Rules for determining the need for radiographs Indications for ankle radiographs

Validation of the rules in different settings

A number of studies have tested the negative predictive value of the Ottawa Ankle Rules. A high negative predictive value implies that if the rules are followed, a fracture will not be missed. Most of these studies were conducted in emergency medicine, sports medicine, and orthopedic settings. In a follow-up study, Stiell and colleagues found a sensitivity of 1.0 (95% confidence interval, 0.95–1.0) for ankle fractures and mid-foot fractures. A positive result is a clinically significant fracture, described as one greater than 3 mm. Such avulsion type injuries are treated clinically like a sprain. In the Stiell study, all films were evaluated by radiologists. The authors estimated a decrease in ankle radiographs by 28% if the rules were followed.⁴

Emergency medicine. The sensitivity in this setting was also found to be 1.0, with a negative predictive value of 1.0 when used by physicians. This means that no fractures would be missed by using the rules. Specificity was only 0.19.³ In a multicenter Canadian trial, the results were similar. Over 12,000 adults were evaluated with the ankle rule at 8 teaching and community hospital emergency departments. This study found a significant decrease in the number of ankle radiographs ordered without an increase in the rate of fractures missed.⁵

Orthopedic surgery. In a study involving 153 military cadets at West Point, orthopedic surgeons also showed the sensitivity of these rules was 1.0, with no false negatives. The investigators estimated they could safely forego 40% of all ankle films and 79% of all midfoot films. In the 4 years of training at the Military Academy, an estimated 33% of all cadets suffer an ankle sprain, which speaks to the prevalence of this condition among young, active persons.⁴

Sports medicine. A prospective study in a university sports medicine clinic validated the use of the ankle rules in a population of 94 athletes. The investigators found a sensitivity of 1.0 for both ankle and midfoot injuries, and a reduction by 34% in the number of films ordered.⁷ In that study the authors comment on the value of these rules in the sports medicine venue, where the rate of ankle and midfoot fractures is low (less than 3%, as opposed to up to 20% in the emergency setting).⁸

Family medicine. Little research in family medicine discusses office use of the Ottawa Ankle Rules, but there is a need for a set of evidence-based protocols in evaluating acute ankle injuries. Before establishment of the ankle rules, family physicians used the clinical findings of (1) absence of tenderness on the dorsum of the foot, (2) lack of impaired weight bearing, (3) recentness of injury (more than 12 hours earlier), and (4) absence of additional injuries. Each of these findings had a negative predictive value of least 94%. While family physician researchers involved in one study did not establish a set of decision “rules,” they estimated that using these criteria could reduce unnecessary films by about 30%.⁹

Meta-analysis. To synthesize results from a large number of studies of the Ottawa Ankle Rules, a meta-analysis involving 27 studies including over 15,000 patients found that evidence supported the rules as an instrument for clinically excluding fractures of the ankle and midfoot. This analysis noted the sensitivity of 1.0 found in virtually every study and estimated that the Ottawa Ankle Rules would decrease the need for films by 30% to 40%.¹⁰

The Ottawa Ankle Rules and children

The original Ottawa Ankle Rules were applied strictly to adults, but they also have been studied considerably in the care of children. Pediatric studies have documented sensitivities ranging from 97% to 100% and specificities from 24% to 47%. All studies have shown significant decreases in unnecessary films and significant cost savings. The conclusion of all authors was that the Ottawa Ankle Rules are a cost-effective, highly sensitive test for evaluating acute ankle injuries in children. ^11-14

However, researchers at the University of Colorado prospectively evaluated the use of Ottawa Ankle Rules in children aged <18 years. The previous studies had used the same criteria as used for adults, but because of the uncertainty of the long-term effects of Salter I (epiphyseal injury) or small (<3 mm) avulsion fractures in the pediatric population, those injuries were included in the fracture category. This brought the sensitivity down to 83%, with a negative predictive value of 93%. The authors suggested that the Ottawa Rules not be employed in the pediatric population.¹⁵ Clearly, more work needs to be done in order to clear them for use in children.

Using the rules reduces costs

Researchers analyzed the cost effectiveness of the rules in the US.¹⁶ Variables used to estimate savings included waiting time and lost productivity as well as the obvious medical and radiographic costs. Previous studies have indicated at least a 28% decrease in unneeded ankle films.⁴ By using this admittedly conservative reduction, it was estimated that the savings would range from $18 to $90 million annually (depending on payer mix involved). Even the smaller amount represents a significant cost savings.

Rule modifications could increase specificity

Because of the low specificity of the Ottawa Rules (a large number of false-positive results are still obtained), sports medicine physicians at the University of Buffalo determined that a modified set of ankle rules could increase specificity significantly. These rules, called the “Buffalo Rules,” kept most of the original rules but changed the original area of fibular tenderness from the posterior and inferior malleolar edges to the midline crest. Using this modification maintained the high sensitivity of the original rules and decreased the need for radiographs from 34% to over 50% (SOR: B).⁷

Patient satisfaction not dependent on films

While most radiographs done for acute ankle injuries are not helpful, many physicians believe patients expect them and will be dissatisfied or upset if films are not taken. A recent Canadian study evaluated the claim that patient preferences influenced physician test ordering and compliance with clinical guidelines. Specifically it looked at emergency department implementation of the Ottawa Ankle Rules and patient satisfaction.¹⁷

This study of almost 1000 patients, split between a population who received films and those who did not for an acute ankle or midfoot injury, indicated that patients who did not get radiographs were just as satisfied with their care as those that did. In the study, 76% of physicians supported the use of clinical guidelines, but 78% admitted that patient expectations influenced their decision making. These rules would likely be of great value where radiographic facilities are inconvenient or costly for patients.

Correspondence
Dr. Paul J. Nugent, 4411 Montgomery Rd, Suite 200, Cincinnati, Ohio 45212. E-mail: [email protected].

In the United States, most ankle injuries are evaluated radiographically, ^1,2 even though only about 15% are found to involve fractures. ³ An estimated 6 million ankle radiographs are performed annually in the US and Canada, costing approximately $300 million dollars (US). The Ottawa Ankle Rules can significantly decrease the number of unnecessary ankle radiographs.

The rules are not a substitute for sound clinical judgment, but augment findings in the history and physical examination to help the clinician determine the appropriateness of ankle films. If the rules are met and radiographs are avoided, it is unlikely, especially with good communication and follow-up, that a patient will turn out to have a significant fracture. Moreover, discussing physical examination findings and the reasons for doing or not doing radiographs will enhance patient-physician communication. As we report in this article, patients’ satisfaction with care seems not to depend on whether x-ray films are ordered.

Applying the Ottawa Ankle Rules

To address the problem of low positive predictive value of ankle radiographs, Stiell and colleagues⁴ at the University of Ottawa and the Ontario Health Ministry developed a set of clinical parameters in the early 1990s to evaluate the need for ankle and midfoot radiographs. Their criteria were based on a multivariable data analysis involving a large number of clinical variables associated with ankle injuries. The resulting rules (Figure) have been shown to decrease the need for films by about 30%.⁴ If a patient does not exhibit any of the criteria, radiographs of the foot or ankle are not needed after trauma (strength of recommendation [SOR]: A).

FIGURE
Ottawa Ankle Rules for determining the need for radiographs Indications for ankle radiographs

Validation of the rules in different settings

A number of studies have tested the negative predictive value of the Ottawa Ankle Rules. A high negative predictive value implies that if the rules are followed, a fracture will not be missed. Most of these studies were conducted in emergency medicine, sports medicine, and orthopedic settings. In a follow-up study, Stiell and colleagues found a sensitivity of 1.0 (95% confidence interval, 0.95–1.0) for ankle fractures and mid-foot fractures. A positive result is a clinically significant fracture, described as one greater than 3 mm. Such avulsion type injuries are treated clinically like a sprain. In the Stiell study, all films were evaluated by radiologists. The authors estimated a decrease in ankle radiographs by 28% if the rules were followed.⁴

Emergency medicine. The sensitivity in this setting was also found to be 1.0, with a negative predictive value of 1.0 when used by physicians. This means that no fractures would be missed by using the rules. Specificity was only 0.19.³ In a multicenter Canadian trial, the results were similar. Over 12,000 adults were evaluated with the ankle rule at 8 teaching and community hospital emergency departments. This study found a significant decrease in the number of ankle radiographs ordered without an increase in the rate of fractures missed.⁵

Orthopedic surgery. In a study involving 153 military cadets at West Point, orthopedic surgeons also showed the sensitivity of these rules was 1.0, with no false negatives. The investigators estimated they could safely forego 40% of all ankle films and 79% of all midfoot films. In the 4 years of training at the Military Academy, an estimated 33% of all cadets suffer an ankle sprain, which speaks to the prevalence of this condition among young, active persons.⁴

Sports medicine. A prospective study in a university sports medicine clinic validated the use of the ankle rules in a population of 94 athletes. The investigators found a sensitivity of 1.0 for both ankle and midfoot injuries, and a reduction by 34% in the number of films ordered.⁷ In that study the authors comment on the value of these rules in the sports medicine venue, where the rate of ankle and midfoot fractures is low (less than 3%, as opposed to up to 20% in the emergency setting).⁸

Family medicine. Little research in family medicine discusses office use of the Ottawa Ankle Rules, but there is a need for a set of evidence-based protocols in evaluating acute ankle injuries. Before establishment of the ankle rules, family physicians used the clinical findings of (1) absence of tenderness on the dorsum of the foot, (2) lack of impaired weight bearing, (3) recentness of injury (more than 12 hours earlier), and (4) absence of additional injuries. Each of these findings had a negative predictive value of least 94%. While family physician researchers involved in one study did not establish a set of decision “rules,” they estimated that using these criteria could reduce unnecessary films by about 30%.⁹

Meta-analysis. To synthesize results from a large number of studies of the Ottawa Ankle Rules, a meta-analysis involving 27 studies including over 15,000 patients found that evidence supported the rules as an instrument for clinically excluding fractures of the ankle and midfoot. This analysis noted the sensitivity of 1.0 found in virtually every study and estimated that the Ottawa Ankle Rules would decrease the need for films by 30% to 40%.¹⁰

The Ottawa Ankle Rules and children

The original Ottawa Ankle Rules were applied strictly to adults, but they also have been studied considerably in the care of children. Pediatric studies have documented sensitivities ranging from 97% to 100% and specificities from 24% to 47%. All studies have shown significant decreases in unnecessary films and significant cost savings. The conclusion of all authors was that the Ottawa Ankle Rules are a cost-effective, highly sensitive test for evaluating acute ankle injuries in children. ^11-14

However, researchers at the University of Colorado prospectively evaluated the use of Ottawa Ankle Rules in children aged <18 years. The previous studies had used the same criteria as used for adults, but because of the uncertainty of the long-term effects of Salter I (epiphyseal injury) or small (<3 mm) avulsion fractures in the pediatric population, those injuries were included in the fracture category. This brought the sensitivity down to 83%, with a negative predictive value of 93%. The authors suggested that the Ottawa Rules not be employed in the pediatric population.¹⁵ Clearly, more work needs to be done in order to clear them for use in children.

Using the rules reduces costs

Researchers analyzed the cost effectiveness of the rules in the US.¹⁶ Variables used to estimate savings included waiting time and lost productivity as well as the obvious medical and radiographic costs. Previous studies have indicated at least a 28% decrease in unneeded ankle films.⁴ By using this admittedly conservative reduction, it was estimated that the savings would range from $18 to $90 million annually (depending on payer mix involved). Even the smaller amount represents a significant cost savings.

Rule modifications could increase specificity

Because of the low specificity of the Ottawa Rules (a large number of false-positive results are still obtained), sports medicine physicians at the University of Buffalo determined that a modified set of ankle rules could increase specificity significantly. These rules, called the “Buffalo Rules,” kept most of the original rules but changed the original area of fibular tenderness from the posterior and inferior malleolar edges to the midline crest. Using this modification maintained the high sensitivity of the original rules and decreased the need for radiographs from 34% to over 50% (SOR: B).⁷

Patient satisfaction not dependent on films

While most radiographs done for acute ankle injuries are not helpful, many physicians believe patients expect them and will be dissatisfied or upset if films are not taken. A recent Canadian study evaluated the claim that patient preferences influenced physician test ordering and compliance with clinical guidelines. Specifically it looked at emergency department implementation of the Ottawa Ankle Rules and patient satisfaction.¹⁷

This study of almost 1000 patients, split between a population who received films and those who did not for an acute ankle or midfoot injury, indicated that patients who did not get radiographs were just as satisfied with their care as those that did. In the study, 76% of physicians supported the use of clinical guidelines, but 78% admitted that patient expectations influenced their decision making. These rules would likely be of great value where radiographic facilities are inconvenient or costly for patients.

Correspondence
Dr. Paul J. Nugent, 4411 Montgomery Rd, Suite 200, Cincinnati, Ohio 45212. E-mail: [email protected].

In the United States, most ankle injuries are evaluated radiographically, ^1,2 even though only about 15% are found to involve fractures. ³ An estimated 6 million ankle radiographs are performed annually in the US and Canada, costing approximately $300 million dollars (US). The Ottawa Ankle Rules can significantly decrease the number of unnecessary ankle radiographs.

The rules are not a substitute for sound clinical judgment, but augment findings in the history and physical examination to help the clinician determine the appropriateness of ankle films. If the rules are met and radiographs are avoided, it is unlikely, especially with good communication and follow-up, that a patient will turn out to have a significant fracture. Moreover, discussing physical examination findings and the reasons for doing or not doing radiographs will enhance patient-physician communication. As we report in this article, patients’ satisfaction with care seems not to depend on whether x-ray films are ordered.

Applying the Ottawa Ankle Rules

To address the problem of low positive predictive value of ankle radiographs, Stiell and colleagues⁴ at the University of Ottawa and the Ontario Health Ministry developed a set of clinical parameters in the early 1990s to evaluate the need for ankle and midfoot radiographs. Their criteria were based on a multivariable data analysis involving a large number of clinical variables associated with ankle injuries. The resulting rules (Figure) have been shown to decrease the need for films by about 30%.⁴ If a patient does not exhibit any of the criteria, radiographs of the foot or ankle are not needed after trauma (strength of recommendation [SOR]: A).

FIGURE
Ottawa Ankle Rules for determining the need for radiographs Indications for ankle radiographs

Validation of the rules in different settings

A number of studies have tested the negative predictive value of the Ottawa Ankle Rules. A high negative predictive value implies that if the rules are followed, a fracture will not be missed. Most of these studies were conducted in emergency medicine, sports medicine, and orthopedic settings. In a follow-up study, Stiell and colleagues found a sensitivity of 1.0 (95% confidence interval, 0.95–1.0) for ankle fractures and mid-foot fractures. A positive result is a clinically significant fracture, described as one greater than 3 mm. Such avulsion type injuries are treated clinically like a sprain. In the Stiell study, all films were evaluated by radiologists. The authors estimated a decrease in ankle radiographs by 28% if the rules were followed.⁴

Emergency medicine. The sensitivity in this setting was also found to be 1.0, with a negative predictive value of 1.0 when used by physicians. This means that no fractures would be missed by using the rules. Specificity was only 0.19.³ In a multicenter Canadian trial, the results were similar. Over 12,000 adults were evaluated with the ankle rule at 8 teaching and community hospital emergency departments. This study found a significant decrease in the number of ankle radiographs ordered without an increase in the rate of fractures missed.⁵

Orthopedic surgery. In a study involving 153 military cadets at West Point, orthopedic surgeons also showed the sensitivity of these rules was 1.0, with no false negatives. The investigators estimated they could safely forego 40% of all ankle films and 79% of all midfoot films. In the 4 years of training at the Military Academy, an estimated 33% of all cadets suffer an ankle sprain, which speaks to the prevalence of this condition among young, active persons.⁴

Sports medicine. A prospective study in a university sports medicine clinic validated the use of the ankle rules in a population of 94 athletes. The investigators found a sensitivity of 1.0 for both ankle and midfoot injuries, and a reduction by 34% in the number of films ordered.⁷ In that study the authors comment on the value of these rules in the sports medicine venue, where the rate of ankle and midfoot fractures is low (less than 3%, as opposed to up to 20% in the emergency setting).⁸

Family medicine. Little research in family medicine discusses office use of the Ottawa Ankle Rules, but there is a need for a set of evidence-based protocols in evaluating acute ankle injuries. Before establishment of the ankle rules, family physicians used the clinical findings of (1) absence of tenderness on the dorsum of the foot, (2) lack of impaired weight bearing, (3) recentness of injury (more than 12 hours earlier), and (4) absence of additional injuries. Each of these findings had a negative predictive value of least 94%. While family physician researchers involved in one study did not establish a set of decision “rules,” they estimated that using these criteria could reduce unnecessary films by about 30%.⁹

Meta-analysis. To synthesize results from a large number of studies of the Ottawa Ankle Rules, a meta-analysis involving 27 studies including over 15,000 patients found that evidence supported the rules as an instrument for clinically excluding fractures of the ankle and midfoot. This analysis noted the sensitivity of 1.0 found in virtually every study and estimated that the Ottawa Ankle Rules would decrease the need for films by 30% to 40%.¹⁰

The Ottawa Ankle Rules and children

The original Ottawa Ankle Rules were applied strictly to adults, but they also have been studied considerably in the care of children. Pediatric studies have documented sensitivities ranging from 97% to 100% and specificities from 24% to 47%. All studies have shown significant decreases in unnecessary films and significant cost savings. The conclusion of all authors was that the Ottawa Ankle Rules are a cost-effective, highly sensitive test for evaluating acute ankle injuries in children. ^11-14

However, researchers at the University of Colorado prospectively evaluated the use of Ottawa Ankle Rules in children aged <18 years. The previous studies had used the same criteria as used for adults, but because of the uncertainty of the long-term effects of Salter I (epiphyseal injury) or small (<3 mm) avulsion fractures in the pediatric population, those injuries were included in the fracture category. This brought the sensitivity down to 83%, with a negative predictive value of 93%. The authors suggested that the Ottawa Rules not be employed in the pediatric population.¹⁵ Clearly, more work needs to be done in order to clear them for use in children.

Using the rules reduces costs

Researchers analyzed the cost effectiveness of the rules in the US.¹⁶ Variables used to estimate savings included waiting time and lost productivity as well as the obvious medical and radiographic costs. Previous studies have indicated at least a 28% decrease in unneeded ankle films.⁴ By using this admittedly conservative reduction, it was estimated that the savings would range from $18 to $90 million annually (depending on payer mix involved). Even the smaller amount represents a significant cost savings.

Rule modifications could increase specificity

Because of the low specificity of the Ottawa Rules (a large number of false-positive results are still obtained), sports medicine physicians at the University of Buffalo determined that a modified set of ankle rules could increase specificity significantly. These rules, called the “Buffalo Rules,” kept most of the original rules but changed the original area of fibular tenderness from the posterior and inferior malleolar edges to the midline crest. Using this modification maintained the high sensitivity of the original rules and decreased the need for radiographs from 34% to over 50% (SOR: B).⁷

Patient satisfaction not dependent on films

While most radiographs done for acute ankle injuries are not helpful, many physicians believe patients expect them and will be dissatisfied or upset if films are not taken. A recent Canadian study evaluated the claim that patient preferences influenced physician test ordering and compliance with clinical guidelines. Specifically it looked at emergency department implementation of the Ottawa Ankle Rules and patient satisfaction.¹⁷

This study of almost 1000 patients, split between a population who received films and those who did not for an acute ankle or midfoot injury, indicated that patients who did not get radiographs were just as satisfied with their care as those that did. In the study, 76% of physicians supported the use of clinical guidelines, but 78% admitted that patient expectations influenced their decision making. These rules would likely be of great value where radiographic facilities are inconvenient or costly for patients.

Correspondence
Dr. Paul J. Nugent, 4411 Montgomery Rd, Suite 200, Cincinnati, Ohio 45212. E-mail: [email protected].

In primary care practice, abnormal vaginal discharge is a common complaint. Signs and symptoms of vaginitis—the most common gynecologic diagnosis in primary care¹ —are not specific for any single underlying cause.² Officebased diagnostic testing, which is underused,³ must be employed to ensure accurate diagnosis and effective treatment. (An article on treatment by the same authors will appear in next month’s issue of The journal of family practice.)

In a primary-care study,⁴ vulvovaginal symptoms including vaginal discharge were due to vulvovaginal candidiasis (VVC) in 27% of patients, bacterial vaginosis (BV) in 21%, trichomoniasis in 8%, Chlamydia trachomatis in 2%, Neisseria gonorrhea (GC) in 1%, and no infection in 34%. Several pathogens may coexist.² VVC, BV, and trichomoniasis account for at least 90% of infectious vaginitis.⁵ This review will therefore focus heavily on these causes of vaginal discharge among women of reproductive age, including pregnant women.

Cervicitis and physiologic cervical discharge

Some women may interpret a physiologic increase in cervical mucous production as abnormal. It occurs cyclically prior to ovulation, is typically transparent and colorless, and may be more pronounced in women with an everted cervix.

Chlamydial infection

In the clinical examination of the cervix, 3 characteristics have been associated with chlamydial infection: yellow endocervical discharge, easily induced cervical bleeding, and opaque cervical discharge.⁶ All 3 findings are statistically significant and independently associated with chlamydial infection (odds ratios 2.8, 2.3, and 2.9, respectively). In the primary care study cited above, purulent cervical discharge was found in 6% of women, most commonly testing positive for Chlamydia, less often for GC.⁴

Trichomonas vaginalis may cause cervicitis as well as vaginitis. Mycoplasma genitalium has been proposed as an additional possible pathogen. It was identified in 7% of more than 700 women with mucopurulent cervical discharge seen in a STD clinic with otherwise negative cultures.⁷ With cervical discharge that appears to be purulent, testing is warranted as a minimum for Chlamydia and GC (SOR: B). Screening of asymptomatic women less than 26 years of age for Chlamydia is recommended by the US Preventive Services Task Force (SOR: A).

Bacterial Vaginosis

Bacterial vaginosis (BV) is neither an inflammatory condition nor an STD, but is a shift in vaginal flora from the normal condition in which lactobacilli predominate, to a polymicrobial flora in which gram-positive anaerobes predominate. In addition to annoying vaginal symptoms, BV is associated with increased risks of more serious conditions such as pelvic inflammatory disease (PID), postoperative infections, and pregnancyrelated complications including prematurity. It also increases the likelihood of acquiring HIV in women exposed to the virus.^8,9

Two principal factors put women at risk for acquiring BV: douching and exposure to a new sexual partner, both of which are thought to disrupt the vaginal ecosystem.¹⁰

Relative benefits of diagnostic tests

A gold standard test has not been established for BV. In about 50% of asymptomatic women, culture results are positive for flora such as Gardnerella vaginalis.⁵ While Amsel’s criteria are often used as a reference and generally suffice for the evaluation of symptomatic women, the best candidate for a gold standard test is probably Gram stain assessment using Nugent’s criteria (described in this section).¹¹ Lack of leukocytes in the vaginal fluid supports a diagnosis of BV. A finding of white blood cells in excess of the number of vaginal epithelial cells suggests an inflammatory process (SOR: C).¹²

Amsel’s criteria with wet mount. The diagnostic approach most commonly used in the office is Amsel’s criteria—homogenous discharge, positive whiff-amine test, pH >4.5, and clue cells found on wet-mount microscopy (see How to perform a wet mount ).¹³ Three of 4 criteria deemed positive is considered diagnostic. If Gram stain is used as the reference standard, then Amsel’s criteria have 70% sensitivity and 94% specificity for diagnosing BV.¹⁴ An analysis of the individual criteria follows. The positive and negative predictive values of each compared with the whole group as reference standard is displayed in Table 1 .

Homogenous discharge. A thin, homogenous, grayish discharge is traditionally associated with BV. However, it is not specific to BV, being found commonly also in women with culture results positive for VVC or no diagnosis of vaginitis.^2,15 It is the criterion least likely to be consistent with the whole group, seen in about half of women BVpositive and over one third of women BV-negative using Amsel’s criteria as the reference standard. ¹⁵

Whiff test. The whiff test is performed by adding drops of 10% potassium hydroxide solution to the vaginal fluid. A positive result is a “fishy” amine odor. In a study¹⁶ of 100 women complaining of malodorous discharge, a positive whiff test was predictive of positive culture results for anaerobic flora such as Bacteroides sp. with sensitivity 67%, specificity 94%, and a positive predictive value of 95%. The whiff test was not positive in any of the 5 cases with positive culture results for G vaginalis in the absence of anaerobes. There were also 12 cases positive for anaerobes without G vaginalis.

pH >4.5. Since the abnormal flora of BV is consistently associated with a vaginal pH >4.5, a normal pH excludes a diagnosis of BV.^17,18 The determination of pH in the narrow range around 4.5 is not accurate using standard nitrazine paper. Narrower-range test paper is available and more accurate. Examples include pH paper for 4.5 to 5.5 (Micro Essential Laboratory), FemExam pH and Amines Test Card (Litmus Concepts), pHem-ALERT: pH paper on a stick (Imagyn Gynecology). Cervical mucous, semen, and blood are alkaline and can interfere with pH testing. Estrogen production is also necessary to maintain an acidic environment. A pH of 3.8 to 4.5 is consistent with normal vaginal flora in premenopausal women with normal estrogen production.¹⁷

Clue cells. Clue cells are vaginal epithelial cells coated with coccobacilli giving an appearance as if coated with ground black pepper. Clue cells on wet mount preparation is considered the most accurate of Amsel’s diagnostic criteria for BV.¹⁹ On the other hand, office evaluation of the wet mount is considered by some authors to be unreliable due to dependence on the clinician’s microscopy skills and lack of a durable record of the patient sample.

Gram stain a more objective test. A Gram stain evaluation using Nugent’s criteria has been adopted as the gold standard test for research purposes, including studies of prematurity. The Gram-stained vaginal specimen is scored from 0 to 10 based on semi-quantitative assessment of 3 classes of morphotypes ( Table 2 ): large gram-positive rods (Lactobacilli), small gram-negative rods (Gardnerella and Bacteroides spp.), and small curved gram-variable rods (Mobiluncus spp.).¹¹

Diagnosis of BV is typically made when the Nugent score is 7 or more, which appears qualitatively as dominant morphotypes other than Lactobacilli. Gram staining is more objective and reproducible compared with wet-mount examination, with a sensitivity of 93% and specificity of 70% if Amsel’s criteria are used as the gold standard.¹⁴ It is useful for the evaluation of asymptomatic women. It also provides a durable record of the patient specimen. Compared with Gram stain, Amsel’s criteria tend to underdiagnose cases. We can expect that if screening for BV in pregnancy becomes a recommendation, Gram staining in a clinical laboratory will be the recommended method of diagnosis.

Other diagnostic tests for BV. DNA testing for Gardnerella is accurate for detection, but it is not synonymous with a diagnosis of BV, as described.²⁰ DNA testing is further described under “Differential Diagnosis.” Gram staining is more reliable than gas-liquid chromatography²¹ and an assay for proline aminopeptidase (a metabolic product of some of the bacteria associated with BV).²² Latex agglutination testing for vaginal lactoferrin is a nonspecific marker for leukocytes, and thus inflammation. It is of little clinical utility in the diagnosis of vaginal discharge.²³

TABLE 1
Predictive values of Amsel’s criteria (using 3 of 4 positive as diagnostic reference standard)

TABLE 2
How to use Nugent’s Gram stain criteria to diagnose bacterial vaginosis

Vulvovaginal Candidiasis

Candidiasis is the second most commonly diagnosed vaginitis in the United States. Some experts estimate that 75% of women will have a yeast infection at some point in life and 5% will have recurrent infections.²⁴ However, 10% to 30% of asymptomatic women with normal flora have positive culture results for Candida.^25-29 The proportion of symptomatic women with positive culture results is 20% to 40%.^4,30,31 Complications of VVC are rare,³² though vulvar vestibulitis³³ and chorioamnionitis in pregnancy³² have been reported.

Risk factors. Symptomatic yeast vaginitis has been associated with condom and diaphragm use, recent antibiotic use, receptive oral sex, oral contraceptive use, spermicide use, diabetes, and immunosuppression including AIDS.^31,34-37 The associations with antibiotic use and oral contraceptives are not consistent.^30,38 Although pregnancy has been postulated as a risk factor for symptomatic VVC, prevalence of yeast on culture in pregnant women is similar to that of nonpregnant women.³⁰

Suggestive symptoms. Among women with a culture result positive for Candida, the most common symptom is pruritus or burning.²⁸ Abnormal discharge is a complaint for most symptomatic women with VVC confirmed by culture.² In addition, women may complain of a thick, odorless, cottage cheese–like discharge.³⁹ A thick, curdled-appearing discharge points to a diagnosis of Candida because it is rarely present with BV or trichomoniasis. In one study,²⁸ a thick curdled discharge had a positive predictive value of 84% for diagnosis of VVC by culture (SOR: B). However, a thin discharge does not rule out VVC; in another study, clinicians described discharge as thin in about half of women ultimately diagnosed with VVC by culture in another study (SOR: B).² On exam, vulvar and vaginal erythema are often present but are not specific findings. The accuracy of the clinical exam for VVC is poor compared with culture (SOR: A).^2,30

Pathogens. Candida albicans is present in 80% to 90% of patients with VVC.^5,40 remainder have non-albicans species, including C glabrata and others.²⁸ An increase to almost 20% of non-Candida species in a vaginitis clinic by the mid-1990’s may be related to increased use of imidazoles available over-the-counter.^40,41 Wet mount results are typically negative in the presence of non-Candida VVC.²⁸

Diagnosis of VVC

The gold standard test for diagnosis of VVC is culture. The potassium hydroxide (KOH) wet mount is only 40% to 75% sensitive.^28,29,42,43 False-positive results are also observed with variable frequency.⁴⁴ The pH of the discharge is usually not more than 5.0 with Candida albicans, but may be higher with non-albicans species such as C glabrata.⁴⁵ Culture is recommended for patients with recurrent or persistent symptoms and a negative wet mount result (SOR: B).^5,28,46 Rapid slide latex agglutination testing is not better than microscopy (SOR: B).⁴²

Trichomoniasis

Trichomonas, a motile protozoan with 4 flagella, causes the third most common form of vaginitis in the United States and is more common in some developing countries. Trichomoniasis accounts for no more than 10% of all cases of vaginitis, and it appears to be decreasing since the introduction of metronidazole.^47,48 It is classified as an STD, although transmission is possible by other means if the organism is protected from desiccation—for example, in dirty washcloths or towels and contaminated water. Nonsexual transmission is thought to be uncommon.

Trichomoniasis is associated with GC and Chlamydia infections, and, like them, has been associated with seroconversion to HIV-positive status.⁴⁹ Trichomonads are identified in 30% to 80% of male sexual partners of infected women. In men, trichomoniasis most often is an asymptomatic carrier state.⁵⁰ However, it is the cause of about 10% of cases of nongonococcal urethritis in men.⁵¹

Clinical presentations. Women with trichomoniasis have variable presentations ranging from an asymptomatic carrier state to a malodorous, purulent discharge with vulvovaginal erythema. Punctate hemorrhagic cervical lesions are considered pathognomonic of trichomoniasis, but are seen in only about 2% of cases (SOR: B).⁵²

Diagnosis. Culture for trichomoniasis is the gold standard. Several culture media have been used, most commonly the Diamond medium. Recently introduced is a transport and culture medium for detection of Trichomonas (InPouch TV), which performs as well as Diamond medium (SOR: A).^53-55 A DNA probe is also available and accurate (SOR: A).

Motile trichomonads are seen on wet preparation in only 50% to 80% of culture-positive cases (SOR: B).^50,54,56 Polymorphonuclear leukocytes can be dominant on wet mount, making visualization of trichomonads more difficult. The pH of the vaginal fluid is usually basic.

Trichomonas reported with cervical cytology

Trichomonas may also be reported on Pap smears. A meta-analysis⁵⁷ comparing the pooled sensitivities and specificities of wet mounts and cytology demonstrated low sensitivities of 68% and 58%, respectively, and high specificities, 99.9% and 97%, respectively (SOR: A).

However, since cytology carries a 3% false-positive rate, its results are not diagnostic of trichomoniasis in low-risk, asymptomatic women.^50,57 Treatment may be prescribed empirically based on positive cytology results. However, if an asymptomatic woman were concerned about whether she really has an STD, a positive wet prep would confirm the diagnosis. A negative wet prep should be followed up with culture to reliably rule out disease (SOR: B).

Trichomoniasis in pregnancy

Screening for asymptomatic trichomoniasis in pregnancy has not been recommended. In fact, some evidence suggests that treatment of trichomoniasis in pregnancy is associated with poorer pregnancy outcomes including lower birth weight and more prematurity (SOR: B).^58,59

Aerobic vaginitis

Aerobic vaginitis is a term proposed to describe purulent vaginal discharge with predominance of abnormal aerobic flora.⁶⁰ Aerobic vaginitis, which may be severe, has been reported as the cause of 5% of cases in a series from a specialty vaginitis clinic.⁶¹ The usual predominant microorganisms are group B streptococci, Escherichia coli, and Staphylococcus aureus. It is likely that less severe cases of aerobic vaginitis are not recognized in the primary care setting and are treated as BV or resolve spontaneously (SOR: C). The case series referred to above also reported good therapeutic response to 2% topical clindamycin (SOR: C).⁶¹

Noninfectious Vaginitis

Noninfectious causes of vaginal discharge include physiologic, irritant and allergic, cytolytic vaginitis, desquamative inflammatory vaginitis, collagen vascular disease, and idiopathic vaginitis.

Irritant and allergic vaginitis may result from sensitivities to topical medications, the active or base ingredients of spermicidal products, douching solutions, and the latex of condoms or diaphragms. If a woman with persistent symptoms has been using such intravaginal products, she should stop (SOR: C).

Cytolytic vaginitis is characterized by overgrowth of lactobacilli and cytolysis of squamous cells, including presence of cytoplasmic fragments and intact cells with naked nuclei.⁶² The cause is uncertain but may include a reaction to intravaginal medications or other products such as tampons. It can be found in up to 5% of women with symptoms and signs of vaginitis.^62,63 Symptoms often mimic VVC and may include a white, cheesy discharge. Vaginal pH ranges from 3.5 to 5.5. Recurrences during luteal phase of the menstrual cycle have been described.⁶⁴ Intravaginal antifungals should be discontinued. Baking soda sitz baths or douches are often used, but clinical trial data to support this practice are lacking (SOR: C).

Noninfectious desquamative inflammatory vaginitis (DIV) has also been described.⁶⁵ DIV is an uncommon vaginitis characterized by profuse purulent discharge with epithelial cell exfoliation. It may occur at any time during the reproductive years or after menopause. There is probably a heterogeneous group of causes of DIV. Some cases may correspond to a disorder within the spectrum of lichen planus.⁶⁶ Treatment is usually difficult, though there may be some response to local or systemic corticosteroid therapy (SOR: C).⁶⁵

Differential diagnosis

A comparison of physical examination findings an diagnostic test results for various etiologies of vaginitis is summarized in Table 3 . An algorithmic approach to the differential diagnosis of abnormal vaginal discharge is presented in the Figure . Diagnosis is complicated in that signs and symptoms do little to help differentiate among BV, VVC, and trichomoniasis. A study² of 22 genitourinary symptoms and signs showed that none differentiated among the 3 infections. This lack of clear-cut differences in symptoms also makes self-diagnosis and telephone triage inaccurate.^67,68

A DNA probe testing system (Affirm VP III Microbial ID Test) for differential diagnosis is available but expensive. It identifies Gardnerella, Trichomonas, and Candida albicans with a sensitivity of 90% to 95%.^54,66 The analyzer costs approximately $10,000 and would typically be purchased by a laboratory. Individual test kits cost about $27.

TABLE 3
Comparative findings among causes of vaginitis

FIGURE
Sequence of office tests to evaluate abnormal vaginal discharge

Corresponding author
Linda French, MD, Associate Professor, Department of Family Practice, College of Human Medicine, Michigan State University, B101 Clinical Center, East Lansing, MI 48824. E-mail: [email protected].

In primary care practice, abnormal vaginal discharge is a common complaint. Signs and symptoms of vaginitis—the most common gynecologic diagnosis in primary care¹ —are not specific for any single underlying cause.² Officebased diagnostic testing, which is underused,³ must be employed to ensure accurate diagnosis and effective treatment. (An article on treatment by the same authors will appear in next month’s issue of The journal of family practice.)

In a primary-care study,⁴ vulvovaginal symptoms including vaginal discharge were due to vulvovaginal candidiasis (VVC) in 27% of patients, bacterial vaginosis (BV) in 21%, trichomoniasis in 8%, Chlamydia trachomatis in 2%, Neisseria gonorrhea (GC) in 1%, and no infection in 34%. Several pathogens may coexist.² VVC, BV, and trichomoniasis account for at least 90% of infectious vaginitis.⁵ This review will therefore focus heavily on these causes of vaginal discharge among women of reproductive age, including pregnant women.

Cervicitis and physiologic cervical discharge

Some women may interpret a physiologic increase in cervical mucous production as abnormal. It occurs cyclically prior to ovulation, is typically transparent and colorless, and may be more pronounced in women with an everted cervix.

Chlamydial infection

In the clinical examination of the cervix, 3 characteristics have been associated with chlamydial infection: yellow endocervical discharge, easily induced cervical bleeding, and opaque cervical discharge.⁶ All 3 findings are statistically significant and independently associated with chlamydial infection (odds ratios 2.8, 2.3, and 2.9, respectively). In the primary care study cited above, purulent cervical discharge was found in 6% of women, most commonly testing positive for Chlamydia, less often for GC.⁴

Trichomonas vaginalis may cause cervicitis as well as vaginitis. Mycoplasma genitalium has been proposed as an additional possible pathogen. It was identified in 7% of more than 700 women with mucopurulent cervical discharge seen in a STD clinic with otherwise negative cultures.⁷ With cervical discharge that appears to be purulent, testing is warranted as a minimum for Chlamydia and GC (SOR: B). Screening of asymptomatic women less than 26 years of age for Chlamydia is recommended by the US Preventive Services Task Force (SOR: A).

Bacterial Vaginosis

Bacterial vaginosis (BV) is neither an inflammatory condition nor an STD, but is a shift in vaginal flora from the normal condition in which lactobacilli predominate, to a polymicrobial flora in which gram-positive anaerobes predominate. In addition to annoying vaginal symptoms, BV is associated with increased risks of more serious conditions such as pelvic inflammatory disease (PID), postoperative infections, and pregnancyrelated complications including prematurity. It also increases the likelihood of acquiring HIV in women exposed to the virus.^8,9

Two principal factors put women at risk for acquiring BV: douching and exposure to a new sexual partner, both of which are thought to disrupt the vaginal ecosystem.¹⁰

Relative benefits of diagnostic tests

A gold standard test has not been established for BV. In about 50% of asymptomatic women, culture results are positive for flora such as Gardnerella vaginalis.⁵ While Amsel’s criteria are often used as a reference and generally suffice for the evaluation of symptomatic women, the best candidate for a gold standard test is probably Gram stain assessment using Nugent’s criteria (described in this section).¹¹ Lack of leukocytes in the vaginal fluid supports a diagnosis of BV. A finding of white blood cells in excess of the number of vaginal epithelial cells suggests an inflammatory process (SOR: C).¹²

Amsel’s criteria with wet mount. The diagnostic approach most commonly used in the office is Amsel’s criteria—homogenous discharge, positive whiff-amine test, pH >4.5, and clue cells found on wet-mount microscopy (see How to perform a wet mount ).¹³ Three of 4 criteria deemed positive is considered diagnostic. If Gram stain is used as the reference standard, then Amsel’s criteria have 70% sensitivity and 94% specificity for diagnosing BV.¹⁴ An analysis of the individual criteria follows. The positive and negative predictive values of each compared with the whole group as reference standard is displayed in Table 1 .

Homogenous discharge. A thin, homogenous, grayish discharge is traditionally associated with BV. However, it is not specific to BV, being found commonly also in women with culture results positive for VVC or no diagnosis of vaginitis.^2,15 It is the criterion least likely to be consistent with the whole group, seen in about half of women BVpositive and over one third of women BV-negative using Amsel’s criteria as the reference standard. ¹⁵

Whiff test. The whiff test is performed by adding drops of 10% potassium hydroxide solution to the vaginal fluid. A positive result is a “fishy” amine odor. In a study¹⁶ of 100 women complaining of malodorous discharge, a positive whiff test was predictive of positive culture results for anaerobic flora such as Bacteroides sp. with sensitivity 67%, specificity 94%, and a positive predictive value of 95%. The whiff test was not positive in any of the 5 cases with positive culture results for G vaginalis in the absence of anaerobes. There were also 12 cases positive for anaerobes without G vaginalis.

pH >4.5. Since the abnormal flora of BV is consistently associated with a vaginal pH >4.5, a normal pH excludes a diagnosis of BV.^17,18 The determination of pH in the narrow range around 4.5 is not accurate using standard nitrazine paper. Narrower-range test paper is available and more accurate. Examples include pH paper for 4.5 to 5.5 (Micro Essential Laboratory), FemExam pH and Amines Test Card (Litmus Concepts), pHem-ALERT: pH paper on a stick (Imagyn Gynecology). Cervical mucous, semen, and blood are alkaline and can interfere with pH testing. Estrogen production is also necessary to maintain an acidic environment. A pH of 3.8 to 4.5 is consistent with normal vaginal flora in premenopausal women with normal estrogen production.¹⁷

Clue cells. Clue cells are vaginal epithelial cells coated with coccobacilli giving an appearance as if coated with ground black pepper. Clue cells on wet mount preparation is considered the most accurate of Amsel’s diagnostic criteria for BV.¹⁹ On the other hand, office evaluation of the wet mount is considered by some authors to be unreliable due to dependence on the clinician’s microscopy skills and lack of a durable record of the patient sample.

Gram stain a more objective test. A Gram stain evaluation using Nugent’s criteria has been adopted as the gold standard test for research purposes, including studies of prematurity. The Gram-stained vaginal specimen is scored from 0 to 10 based on semi-quantitative assessment of 3 classes of morphotypes ( Table 2 ): large gram-positive rods (Lactobacilli), small gram-negative rods (Gardnerella and Bacteroides spp.), and small curved gram-variable rods (Mobiluncus spp.).¹¹

Diagnosis of BV is typically made when the Nugent score is 7 or more, which appears qualitatively as dominant morphotypes other than Lactobacilli. Gram staining is more objective and reproducible compared with wet-mount examination, with a sensitivity of 93% and specificity of 70% if Amsel’s criteria are used as the gold standard.¹⁴ It is useful for the evaluation of asymptomatic women. It also provides a durable record of the patient specimen. Compared with Gram stain, Amsel’s criteria tend to underdiagnose cases. We can expect that if screening for BV in pregnancy becomes a recommendation, Gram staining in a clinical laboratory will be the recommended method of diagnosis.

Other diagnostic tests for BV. DNA testing for Gardnerella is accurate for detection, but it is not synonymous with a diagnosis of BV, as described.²⁰ DNA testing is further described under “Differential Diagnosis.” Gram staining is more reliable than gas-liquid chromatography²¹ and an assay for proline aminopeptidase (a metabolic product of some of the bacteria associated with BV).²² Latex agglutination testing for vaginal lactoferrin is a nonspecific marker for leukocytes, and thus inflammation. It is of little clinical utility in the diagnosis of vaginal discharge.²³

TABLE 1
Predictive values of Amsel’s criteria (using 3 of 4 positive as diagnostic reference standard)

TABLE 2
How to use Nugent’s Gram stain criteria to diagnose bacterial vaginosis

Vulvovaginal Candidiasis

Candidiasis is the second most commonly diagnosed vaginitis in the United States. Some experts estimate that 75% of women will have a yeast infection at some point in life and 5% will have recurrent infections.²⁴ However, 10% to 30% of asymptomatic women with normal flora have positive culture results for Candida.^25-29 The proportion of symptomatic women with positive culture results is 20% to 40%.^4,30,31 Complications of VVC are rare,³² though vulvar vestibulitis³³ and chorioamnionitis in pregnancy³² have been reported.

Risk factors. Symptomatic yeast vaginitis has been associated with condom and diaphragm use, recent antibiotic use, receptive oral sex, oral contraceptive use, spermicide use, diabetes, and immunosuppression including AIDS.^31,34-37 The associations with antibiotic use and oral contraceptives are not consistent.^30,38 Although pregnancy has been postulated as a risk factor for symptomatic VVC, prevalence of yeast on culture in pregnant women is similar to that of nonpregnant women.³⁰

Suggestive symptoms. Among women with a culture result positive for Candida, the most common symptom is pruritus or burning.²⁸ Abnormal discharge is a complaint for most symptomatic women with VVC confirmed by culture.² In addition, women may complain of a thick, odorless, cottage cheese–like discharge.³⁹ A thick, curdled-appearing discharge points to a diagnosis of Candida because it is rarely present with BV or trichomoniasis. In one study,²⁸ a thick curdled discharge had a positive predictive value of 84% for diagnosis of VVC by culture (SOR: B). However, a thin discharge does not rule out VVC; in another study, clinicians described discharge as thin in about half of women ultimately diagnosed with VVC by culture in another study (SOR: B).² On exam, vulvar and vaginal erythema are often present but are not specific findings. The accuracy of the clinical exam for VVC is poor compared with culture (SOR: A).^2,30

Pathogens. Candida albicans is present in 80% to 90% of patients with VVC.^5,40 remainder have non-albicans species, including C glabrata and others.²⁸ An increase to almost 20% of non-Candida species in a vaginitis clinic by the mid-1990’s may be related to increased use of imidazoles available over-the-counter.^40,41 Wet mount results are typically negative in the presence of non-Candida VVC.²⁸

Diagnosis of VVC

The gold standard test for diagnosis of VVC is culture. The potassium hydroxide (KOH) wet mount is only 40% to 75% sensitive.^28,29,42,43 False-positive results are also observed with variable frequency.⁴⁴ The pH of the discharge is usually not more than 5.0 with Candida albicans, but may be higher with non-albicans species such as C glabrata.⁴⁵ Culture is recommended for patients with recurrent or persistent symptoms and a negative wet mount result (SOR: B).^5,28,46 Rapid slide latex agglutination testing is not better than microscopy (SOR: B).⁴²

Trichomoniasis

Trichomonas, a motile protozoan with 4 flagella, causes the third most common form of vaginitis in the United States and is more common in some developing countries. Trichomoniasis accounts for no more than 10% of all cases of vaginitis, and it appears to be decreasing since the introduction of metronidazole.^47,48 It is classified as an STD, although transmission is possible by other means if the organism is protected from desiccation—for example, in dirty washcloths or towels and contaminated water. Nonsexual transmission is thought to be uncommon.

Trichomoniasis is associated with GC and Chlamydia infections, and, like them, has been associated with seroconversion to HIV-positive status.⁴⁹ Trichomonads are identified in 30% to 80% of male sexual partners of infected women. In men, trichomoniasis most often is an asymptomatic carrier state.⁵⁰ However, it is the cause of about 10% of cases of nongonococcal urethritis in men.⁵¹

Clinical presentations. Women with trichomoniasis have variable presentations ranging from an asymptomatic carrier state to a malodorous, purulent discharge with vulvovaginal erythema. Punctate hemorrhagic cervical lesions are considered pathognomonic of trichomoniasis, but are seen in only about 2% of cases (SOR: B).⁵²

Diagnosis. Culture for trichomoniasis is the gold standard. Several culture media have been used, most commonly the Diamond medium. Recently introduced is a transport and culture medium for detection of Trichomonas (InPouch TV), which performs as well as Diamond medium (SOR: A).^53-55 A DNA probe is also available and accurate (SOR: A).

Motile trichomonads are seen on wet preparation in only 50% to 80% of culture-positive cases (SOR: B).^50,54,56 Polymorphonuclear leukocytes can be dominant on wet mount, making visualization of trichomonads more difficult. The pH of the vaginal fluid is usually basic.

Trichomonas reported with cervical cytology

Trichomonas may also be reported on Pap smears. A meta-analysis⁵⁷ comparing the pooled sensitivities and specificities of wet mounts and cytology demonstrated low sensitivities of 68% and 58%, respectively, and high specificities, 99.9% and 97%, respectively (SOR: A).

However, since cytology carries a 3% false-positive rate, its results are not diagnostic of trichomoniasis in low-risk, asymptomatic women.^50,57 Treatment may be prescribed empirically based on positive cytology results. However, if an asymptomatic woman were concerned about whether she really has an STD, a positive wet prep would confirm the diagnosis. A negative wet prep should be followed up with culture to reliably rule out disease (SOR: B).

Trichomoniasis in pregnancy

Screening for asymptomatic trichomoniasis in pregnancy has not been recommended. In fact, some evidence suggests that treatment of trichomoniasis in pregnancy is associated with poorer pregnancy outcomes including lower birth weight and more prematurity (SOR: B).^58,59

Aerobic vaginitis

Aerobic vaginitis is a term proposed to describe purulent vaginal discharge with predominance of abnormal aerobic flora.⁶⁰ Aerobic vaginitis, which may be severe, has been reported as the cause of 5% of cases in a series from a specialty vaginitis clinic.⁶¹ The usual predominant microorganisms are group B streptococci, Escherichia coli, and Staphylococcus aureus. It is likely that less severe cases of aerobic vaginitis are not recognized in the primary care setting and are treated as BV or resolve spontaneously (SOR: C). The case series referred to above also reported good therapeutic response to 2% topical clindamycin (SOR: C).⁶¹

Noninfectious Vaginitis

Noninfectious causes of vaginal discharge include physiologic, irritant and allergic, cytolytic vaginitis, desquamative inflammatory vaginitis, collagen vascular disease, and idiopathic vaginitis.

Irritant and allergic vaginitis may result from sensitivities to topical medications, the active or base ingredients of spermicidal products, douching solutions, and the latex of condoms or diaphragms. If a woman with persistent symptoms has been using such intravaginal products, she should stop (SOR: C).

Cytolytic vaginitis is characterized by overgrowth of lactobacilli and cytolysis of squamous cells, including presence of cytoplasmic fragments and intact cells with naked nuclei.⁶² The cause is uncertain but may include a reaction to intravaginal medications or other products such as tampons. It can be found in up to 5% of women with symptoms and signs of vaginitis.^62,63 Symptoms often mimic VVC and may include a white, cheesy discharge. Vaginal pH ranges from 3.5 to 5.5. Recurrences during luteal phase of the menstrual cycle have been described.⁶⁴ Intravaginal antifungals should be discontinued. Baking soda sitz baths or douches are often used, but clinical trial data to support this practice are lacking (SOR: C).

Noninfectious desquamative inflammatory vaginitis (DIV) has also been described.⁶⁵ DIV is an uncommon vaginitis characterized by profuse purulent discharge with epithelial cell exfoliation. It may occur at any time during the reproductive years or after menopause. There is probably a heterogeneous group of causes of DIV. Some cases may correspond to a disorder within the spectrum of lichen planus.⁶⁶ Treatment is usually difficult, though there may be some response to local or systemic corticosteroid therapy (SOR: C).⁶⁵

Differential diagnosis

A comparison of physical examination findings an diagnostic test results for various etiologies of vaginitis is summarized in Table 3 . An algorithmic approach to the differential diagnosis of abnormal vaginal discharge is presented in the Figure . Diagnosis is complicated in that signs and symptoms do little to help differentiate among BV, VVC, and trichomoniasis. A study² of 22 genitourinary symptoms and signs showed that none differentiated among the 3 infections. This lack of clear-cut differences in symptoms also makes self-diagnosis and telephone triage inaccurate.^67,68

A DNA probe testing system (Affirm VP III Microbial ID Test) for differential diagnosis is available but expensive. It identifies Gardnerella, Trichomonas, and Candida albicans with a sensitivity of 90% to 95%.^54,66 The analyzer costs approximately $10,000 and would typically be purchased by a laboratory. Individual test kits cost about $27.

TABLE 3
Comparative findings among causes of vaginitis

FIGURE
Sequence of office tests to evaluate abnormal vaginal discharge

Corresponding author
Linda French, MD, Associate Professor, Department of Family Practice, College of Human Medicine, Michigan State University, B101 Clinical Center, East Lansing, MI 48824. E-mail: [email protected].

In primary care practice, abnormal vaginal discharge is a common complaint. Signs and symptoms of vaginitis—the most common gynecologic diagnosis in primary care¹ —are not specific for any single underlying cause.² Officebased diagnostic testing, which is underused,³ must be employed to ensure accurate diagnosis and effective treatment. (An article on treatment by the same authors will appear in next month’s issue of The journal of family practice.)

In a primary-care study,⁴ vulvovaginal symptoms including vaginal discharge were due to vulvovaginal candidiasis (VVC) in 27% of patients, bacterial vaginosis (BV) in 21%, trichomoniasis in 8%, Chlamydia trachomatis in 2%, Neisseria gonorrhea (GC) in 1%, and no infection in 34%. Several pathogens may coexist.² VVC, BV, and trichomoniasis account for at least 90% of infectious vaginitis.⁵ This review will therefore focus heavily on these causes of vaginal discharge among women of reproductive age, including pregnant women.

Cervicitis and physiologic cervical discharge

Some women may interpret a physiologic increase in cervical mucous production as abnormal. It occurs cyclically prior to ovulation, is typically transparent and colorless, and may be more pronounced in women with an everted cervix.

Chlamydial infection

In the clinical examination of the cervix, 3 characteristics have been associated with chlamydial infection: yellow endocervical discharge, easily induced cervical bleeding, and opaque cervical discharge.⁶ All 3 findings are statistically significant and independently associated with chlamydial infection (odds ratios 2.8, 2.3, and 2.9, respectively). In the primary care study cited above, purulent cervical discharge was found in 6% of women, most commonly testing positive for Chlamydia, less often for GC.⁴

Trichomonas vaginalis may cause cervicitis as well as vaginitis. Mycoplasma genitalium has been proposed as an additional possible pathogen. It was identified in 7% of more than 700 women with mucopurulent cervical discharge seen in a STD clinic with otherwise negative cultures.⁷ With cervical discharge that appears to be purulent, testing is warranted as a minimum for Chlamydia and GC (SOR: B). Screening of asymptomatic women less than 26 years of age for Chlamydia is recommended by the US Preventive Services Task Force (SOR: A).

Bacterial Vaginosis

Bacterial vaginosis (BV) is neither an inflammatory condition nor an STD, but is a shift in vaginal flora from the normal condition in which lactobacilli predominate, to a polymicrobial flora in which gram-positive anaerobes predominate. In addition to annoying vaginal symptoms, BV is associated with increased risks of more serious conditions such as pelvic inflammatory disease (PID), postoperative infections, and pregnancyrelated complications including prematurity. It also increases the likelihood of acquiring HIV in women exposed to the virus.^8,9

Two principal factors put women at risk for acquiring BV: douching and exposure to a new sexual partner, both of which are thought to disrupt the vaginal ecosystem.¹⁰

Relative benefits of diagnostic tests

A gold standard test has not been established for BV. In about 50% of asymptomatic women, culture results are positive for flora such as Gardnerella vaginalis.⁵ While Amsel’s criteria are often used as a reference and generally suffice for the evaluation of symptomatic women, the best candidate for a gold standard test is probably Gram stain assessment using Nugent’s criteria (described in this section).¹¹ Lack of leukocytes in the vaginal fluid supports a diagnosis of BV. A finding of white blood cells in excess of the number of vaginal epithelial cells suggests an inflammatory process (SOR: C).¹²

Amsel’s criteria with wet mount. The diagnostic approach most commonly used in the office is Amsel’s criteria—homogenous discharge, positive whiff-amine test, pH >4.5, and clue cells found on wet-mount microscopy (see How to perform a wet mount ).¹³ Three of 4 criteria deemed positive is considered diagnostic. If Gram stain is used as the reference standard, then Amsel’s criteria have 70% sensitivity and 94% specificity for diagnosing BV.¹⁴ An analysis of the individual criteria follows. The positive and negative predictive values of each compared with the whole group as reference standard is displayed in Table 1 .

Homogenous discharge. A thin, homogenous, grayish discharge is traditionally associated with BV. However, it is not specific to BV, being found commonly also in women with culture results positive for VVC or no diagnosis of vaginitis.^2,15 It is the criterion least likely to be consistent with the whole group, seen in about half of women BVpositive and over one third of women BV-negative using Amsel’s criteria as the reference standard. ¹⁵

Whiff test. The whiff test is performed by adding drops of 10% potassium hydroxide solution to the vaginal fluid. A positive result is a “fishy” amine odor. In a study¹⁶ of 100 women complaining of malodorous discharge, a positive whiff test was predictive of positive culture results for anaerobic flora such as Bacteroides sp. with sensitivity 67%, specificity 94%, and a positive predictive value of 95%. The whiff test was not positive in any of the 5 cases with positive culture results for G vaginalis in the absence of anaerobes. There were also 12 cases positive for anaerobes without G vaginalis.

pH >4.5. Since the abnormal flora of BV is consistently associated with a vaginal pH >4.5, a normal pH excludes a diagnosis of BV.^17,18 The determination of pH in the narrow range around 4.5 is not accurate using standard nitrazine paper. Narrower-range test paper is available and more accurate. Examples include pH paper for 4.5 to 5.5 (Micro Essential Laboratory), FemExam pH and Amines Test Card (Litmus Concepts), pHem-ALERT: pH paper on a stick (Imagyn Gynecology). Cervical mucous, semen, and blood are alkaline and can interfere with pH testing. Estrogen production is also necessary to maintain an acidic environment. A pH of 3.8 to 4.5 is consistent with normal vaginal flora in premenopausal women with normal estrogen production.¹⁷

Clue cells. Clue cells are vaginal epithelial cells coated with coccobacilli giving an appearance as if coated with ground black pepper. Clue cells on wet mount preparation is considered the most accurate of Amsel’s diagnostic criteria for BV.¹⁹ On the other hand, office evaluation of the wet mount is considered by some authors to be unreliable due to dependence on the clinician’s microscopy skills and lack of a durable record of the patient sample.

Gram stain a more objective test. A Gram stain evaluation using Nugent’s criteria has been adopted as the gold standard test for research purposes, including studies of prematurity. The Gram-stained vaginal specimen is scored from 0 to 10 based on semi-quantitative assessment of 3 classes of morphotypes ( Table 2 ): large gram-positive rods (Lactobacilli), small gram-negative rods (Gardnerella and Bacteroides spp.), and small curved gram-variable rods (Mobiluncus spp.).¹¹

Diagnosis of BV is typically made when the Nugent score is 7 or more, which appears qualitatively as dominant morphotypes other than Lactobacilli. Gram staining is more objective and reproducible compared with wet-mount examination, with a sensitivity of 93% and specificity of 70% if Amsel’s criteria are used as the gold standard.¹⁴ It is useful for the evaluation of asymptomatic women. It also provides a durable record of the patient specimen. Compared with Gram stain, Amsel’s criteria tend to underdiagnose cases. We can expect that if screening for BV in pregnancy becomes a recommendation, Gram staining in a clinical laboratory will be the recommended method of diagnosis.

Other diagnostic tests for BV. DNA testing for Gardnerella is accurate for detection, but it is not synonymous with a diagnosis of BV, as described.²⁰ DNA testing is further described under “Differential Diagnosis.” Gram staining is more reliable than gas-liquid chromatography²¹ and an assay for proline aminopeptidase (a metabolic product of some of the bacteria associated with BV).²² Latex agglutination testing for vaginal lactoferrin is a nonspecific marker for leukocytes, and thus inflammation. It is of little clinical utility in the diagnosis of vaginal discharge.²³

TABLE 1
Predictive values of Amsel’s criteria (using 3 of 4 positive as diagnostic reference standard)

TABLE 2
How to use Nugent’s Gram stain criteria to diagnose bacterial vaginosis

Vulvovaginal Candidiasis

Candidiasis is the second most commonly diagnosed vaginitis in the United States. Some experts estimate that 75% of women will have a yeast infection at some point in life and 5% will have recurrent infections.²⁴ However, 10% to 30% of asymptomatic women with normal flora have positive culture results for Candida.^25-29 The proportion of symptomatic women with positive culture results is 20% to 40%.^4,30,31 Complications of VVC are rare,³² though vulvar vestibulitis³³ and chorioamnionitis in pregnancy³² have been reported.

Risk factors. Symptomatic yeast vaginitis has been associated with condom and diaphragm use, recent antibiotic use, receptive oral sex, oral contraceptive use, spermicide use, diabetes, and immunosuppression including AIDS.^31,34-37 The associations with antibiotic use and oral contraceptives are not consistent.^30,38 Although pregnancy has been postulated as a risk factor for symptomatic VVC, prevalence of yeast on culture in pregnant women is similar to that of nonpregnant women.³⁰

Suggestive symptoms. Among women with a culture result positive for Candida, the most common symptom is pruritus or burning.²⁸ Abnormal discharge is a complaint for most symptomatic women with VVC confirmed by culture.² In addition, women may complain of a thick, odorless, cottage cheese–like discharge.³⁹ A thick, curdled-appearing discharge points to a diagnosis of Candida because it is rarely present with BV or trichomoniasis. In one study,²⁸ a thick curdled discharge had a positive predictive value of 84% for diagnosis of VVC by culture (SOR: B). However, a thin discharge does not rule out VVC; in another study, clinicians described discharge as thin in about half of women ultimately diagnosed with VVC by culture in another study (SOR: B).² On exam, vulvar and vaginal erythema are often present but are not specific findings. The accuracy of the clinical exam for VVC is poor compared with culture (SOR: A).^2,30

Pathogens. Candida albicans is present in 80% to 90% of patients with VVC.^5,40 remainder have non-albicans species, including C glabrata and others.²⁸ An increase to almost 20% of non-Candida species in a vaginitis clinic by the mid-1990’s may be related to increased use of imidazoles available over-the-counter.^40,41 Wet mount results are typically negative in the presence of non-Candida VVC.²⁸

Diagnosis of VVC

The gold standard test for diagnosis of VVC is culture. The potassium hydroxide (KOH) wet mount is only 40% to 75% sensitive.^28,29,42,43 False-positive results are also observed with variable frequency.⁴⁴ The pH of the discharge is usually not more than 5.0 with Candida albicans, but may be higher with non-albicans species such as C glabrata.⁴⁵ Culture is recommended for patients with recurrent or persistent symptoms and a negative wet mount result (SOR: B).^5,28,46 Rapid slide latex agglutination testing is not better than microscopy (SOR: B).⁴²

Trichomoniasis

Trichomonas, a motile protozoan with 4 flagella, causes the third most common form of vaginitis in the United States and is more common in some developing countries. Trichomoniasis accounts for no more than 10% of all cases of vaginitis, and it appears to be decreasing since the introduction of metronidazole.^47,48 It is classified as an STD, although transmission is possible by other means if the organism is protected from desiccation—for example, in dirty washcloths or towels and contaminated water. Nonsexual transmission is thought to be uncommon.

Trichomoniasis is associated with GC and Chlamydia infections, and, like them, has been associated with seroconversion to HIV-positive status.⁴⁹ Trichomonads are identified in 30% to 80% of male sexual partners of infected women. In men, trichomoniasis most often is an asymptomatic carrier state.⁵⁰ However, it is the cause of about 10% of cases of nongonococcal urethritis in men.⁵¹

Clinical presentations. Women with trichomoniasis have variable presentations ranging from an asymptomatic carrier state to a malodorous, purulent discharge with vulvovaginal erythema. Punctate hemorrhagic cervical lesions are considered pathognomonic of trichomoniasis, but are seen in only about 2% of cases (SOR: B).⁵²

Diagnosis. Culture for trichomoniasis is the gold standard. Several culture media have been used, most commonly the Diamond medium. Recently introduced is a transport and culture medium for detection of Trichomonas (InPouch TV), which performs as well as Diamond medium (SOR: A).^53-55 A DNA probe is also available and accurate (SOR: A).

Motile trichomonads are seen on wet preparation in only 50% to 80% of culture-positive cases (SOR: B).^50,54,56 Polymorphonuclear leukocytes can be dominant on wet mount, making visualization of trichomonads more difficult. The pH of the vaginal fluid is usually basic.

Trichomonas reported with cervical cytology

Trichomonas may also be reported on Pap smears. A meta-analysis⁵⁷ comparing the pooled sensitivities and specificities of wet mounts and cytology demonstrated low sensitivities of 68% and 58%, respectively, and high specificities, 99.9% and 97%, respectively (SOR: A).

However, since cytology carries a 3% false-positive rate, its results are not diagnostic of trichomoniasis in low-risk, asymptomatic women.^50,57 Treatment may be prescribed empirically based on positive cytology results. However, if an asymptomatic woman were concerned about whether she really has an STD, a positive wet prep would confirm the diagnosis. A negative wet prep should be followed up with culture to reliably rule out disease (SOR: B).

Trichomoniasis in pregnancy

Screening for asymptomatic trichomoniasis in pregnancy has not been recommended. In fact, some evidence suggests that treatment of trichomoniasis in pregnancy is associated with poorer pregnancy outcomes including lower birth weight and more prematurity (SOR: B).^58,59

Aerobic vaginitis

Aerobic vaginitis is a term proposed to describe purulent vaginal discharge with predominance of abnormal aerobic flora.⁶⁰ Aerobic vaginitis, which may be severe, has been reported as the cause of 5% of cases in a series from a specialty vaginitis clinic.⁶¹ The usual predominant microorganisms are group B streptococci, Escherichia coli, and Staphylococcus aureus. It is likely that less severe cases of aerobic vaginitis are not recognized in the primary care setting and are treated as BV or resolve spontaneously (SOR: C). The case series referred to above also reported good therapeutic response to 2% topical clindamycin (SOR: C).⁶¹

Noninfectious Vaginitis

Noninfectious causes of vaginal discharge include physiologic, irritant and allergic, cytolytic vaginitis, desquamative inflammatory vaginitis, collagen vascular disease, and idiopathic vaginitis.

Irritant and allergic vaginitis may result from sensitivities to topical medications, the active or base ingredients of spermicidal products, douching solutions, and the latex of condoms or diaphragms. If a woman with persistent symptoms has been using such intravaginal products, she should stop (SOR: C).

Cytolytic vaginitis is characterized by overgrowth of lactobacilli and cytolysis of squamous cells, including presence of cytoplasmic fragments and intact cells with naked nuclei.⁶² The cause is uncertain but may include a reaction to intravaginal medications or other products such as tampons. It can be found in up to 5% of women with symptoms and signs of vaginitis.^62,63 Symptoms often mimic VVC and may include a white, cheesy discharge. Vaginal pH ranges from 3.5 to 5.5. Recurrences during luteal phase of the menstrual cycle have been described.⁶⁴ Intravaginal antifungals should be discontinued. Baking soda sitz baths or douches are often used, but clinical trial data to support this practice are lacking (SOR: C).

Noninfectious desquamative inflammatory vaginitis (DIV) has also been described.⁶⁵ DIV is an uncommon vaginitis characterized by profuse purulent discharge with epithelial cell exfoliation. It may occur at any time during the reproductive years or after menopause. There is probably a heterogeneous group of causes of DIV. Some cases may correspond to a disorder within the spectrum of lichen planus.⁶⁶ Treatment is usually difficult, though there may be some response to local or systemic corticosteroid therapy (SOR: C).⁶⁵

Differential diagnosis

A comparison of physical examination findings an diagnostic test results for various etiologies of vaginitis is summarized in Table 3 . An algorithmic approach to the differential diagnosis of abnormal vaginal discharge is presented in the Figure . Diagnosis is complicated in that signs and symptoms do little to help differentiate among BV, VVC, and trichomoniasis. A study² of 22 genitourinary symptoms and signs showed that none differentiated among the 3 infections. This lack of clear-cut differences in symptoms also makes self-diagnosis and telephone triage inaccurate.^67,68

A DNA probe testing system (Affirm VP III Microbial ID Test) for differential diagnosis is available but expensive. It identifies Gardnerella, Trichomonas, and Candida albicans with a sensitivity of 90% to 95%.^54,66 The analyzer costs approximately $10,000 and would typically be purchased by a laboratory. Individual test kits cost about $27.

TABLE 3
Comparative findings among causes of vaginitis

FIGURE
Sequence of office tests to evaluate abnormal vaginal discharge

Corresponding author
Linda French, MD, Associate Professor, Department of Family Practice, College of Human Medicine, Michigan State University, B101 Clinical Center, East Lansing, MI 48824. E-mail: [email protected].

We know that bed rest generally is over-prescribed.¹ But what about resting or immobilizing acutely injured limbs—from simple sprains to gunshot wounds? This specific application of rest is controversial.^2-6

Theoretically, rest should reduce further tissue damage by limiting movement, decreasing pain and swelling, and lessening loss of fracture reduction.^7-15 However, mobilization increases blood flow and so reduces muscle atrophy, disuse osteoporosis, adhesions, and joint stiffness.^7-15

The first empirical trials to determine whether rest or mobilization would offer greater benefit were undertaken in the 1980s, using both human and animal models. Systematic reviews have looked at specific injuries: acute ankle sprains¹⁶; ankle ligament rupture¹⁷; isolated ulnar fracture¹⁸; metacarpal fractures¹⁹; collateral ankle ligament sprains²⁰; and soft-tissue ankle injury.²¹ Each review independently suggests benefits for early mobilization.

However, most clinicians employ some period of enforced rest of injured limbs. For example, a survey of orthopedic surgeons in Denmark found early mobilization for ankle fractures is prescribed for only 6% of patients.²²

We systematically reviewed the literature to find trials that compared rest with early mobilization in acute limb injuries.

Methods

Protocol and search strategy

We searched the literature to identify prospective randomized controlled trials (RCTs) comparing rest with mobilization for acute limb injuries, searching the Cochrane Controlled Trials Register and Systematic Reviews, Medline (1966–2002), EMBASE (Rehabilitation and Physical Therapy), and Web of Science, combined with the following terms:

• immobilis*, immobiliz*, mobilis*, mobiliz*,
  injur*, fracture*, limb*, cast*, brace*, splint*,
  leg*, arm*, wrist*, elbow*, joint*, *carpal*,
  *tarsal*, knee*, ankle*, femur*, tibia*, fibula*,
  colles*, ulna*, radi*, humer*, sprain*, soft tissue,
  plaster slab, dislocat*

We accepted all publication languages and translated several papers. We found 445 potentially relevant papers by examining their titles or abstracts (Figure). Studies that did not fulfill our inclusion criteria were discarded.

We obtained full texts for 76 studies and undertook a prospective cited reference search in Web of Science for each study to identify more recent articles. A total of 187 papers were then evaluated further for duplicate data and inclusion criteria.

Trials that met our inclusion criteria:

1. Were randomized and controlled
2. Studied acute limb injuries
3. Compared groups whose only treatment difference was mobilization (studies that compared surgery and immobilization against nonsurgical early mobilization were not included; studies with internal fixation were included only if internal fixation was applied to both groups)
4. Had loss to follow-up of 20%, except where analysis was clearly intention to treat
5. Studied populations not focussed solely on young children, (studies in which the given age range included a small number of patients as young as 11 or 12 years were accepted—in these studies the median age was much higher)
6. adequately reported data collection and statistical analysis.

By these criteria, we excluded 138 studies (a list of which is available from the authors). Finally, 49 studies were included.

FIGURE
Literature search process, and methods by which randomized controlled trials (RCTs) were selected

Study methods quality and validity

Two authors independently scored the validity of each of the included 49 studies, meeting to discuss and resolve any disagreements. We relied on published data, without writing to authors to clarify or obtain information. Included studies were evaluated against 11 standard criteria.²³ Each criterion was scored 0–2. However, the practical maximum possible score was 18 because it was impossible to blind patients or their caregivers to mobilization. We considered the quality of 16 studies scoring 11 to be high.

Results

Across the 49 eligible studies, the different interventions were well described and differentiated, and data collection methods were clearly explained. Duration of immobilization in a cast was described (if at all) from 10 days to 8 weeks. Mobilization strategies varied and included active exercise, orthoses, crutches, or bandages.

Study quality was poor in many respects: several did not include important demographic data (including age, sex, or the numbers randomized to control and experimental group). Most did not report data about the methods of randomization. Many studies used inadequate randomization methods, with only 5 reporting optimal methods. Loss to follow-up was addressed poorly (not described or intention-to-treat analysis was not used). Outcome assessment was not blinded (a serious shortcoming in view of the subjective assessment of some patient outcomes). Reporting of inclusion criteria was inadequate in a few studies.

Given these deficiencies, our discussion emphasizes conclusions supported by studies of high quality.

Reporting of the studies

We divided the 49 studies into 4 groups: lower-limb fractures, other lower-limb injuries, upper-limb fractures, and other upper-limb injuries (Table W, at www.jfponline.com). Each group was further divided into trials using some form of limb support (such as a brace, splint, or a short period of immobilization) or minimal or no support (bandage, crutches, or tape) in the early mobilization group.

Most studies used some form of dynamic treatment for the injured limb, with highest use in lowerlimb injuries (74%); 2 studies included both supported and unsupported mobilization.^24-26 In studies of upper-limb fractures, support was used in 50%. In high-quality studies, 12 out of 16 studies used some support, but results were consistent with studies that did not. The smaller number of studies using minimal support makes comparison with supported mobilization difficult, but results suggest that supported mobilization may result in an earlier return to work or sport, whereas minimally supported patients achieved better composite scores and muscle strength. There was too much clinical heterogeneity to sensibly combine the results; however, we were interested to see whether mobilization was useful across a heterogeneous group of conditions.

Patient-centered outcomes

No study reported any significant improvement with rest on swelling and pain. To the contrary, 13 studies reported significant improvement with early mobilization compared with rest,^16,17,27-37 and 9 studies reported enhanced patient satisfaction.^{16,30,36,38-43} High-quality studies that reported pain levels found no difference or favored early mobilization.^{25,26,30,31,39,44-51} This was significant in only 2 studies of lower-limb nonfractures.^25,26,30 No study reported adverse patient opinion in an early mobilization regime,^{28,31,38,40,44-46,48,51-56} and patients appreciated their ability to use the affected limb.^34,43,49

Functional outcomes

Ten studies measured global function, using composite scores including subjective and objective criteria—pain, stiffness, swelling, use of supports, and ability to climb stairs, play sport, work, and undertake activities of daily life (Table W). Seven studies reported significant improvements with early mobilization—after 6 months (6 studies),^{28,35,42,46,52,57} or 12 months (1 study).³² Four studies of high quality found significant benefit for early mobilization in composite scores, most of these for lower-limb fractures.^46,47,52,54 Significantly more patients with acute ankle sprain, who were immobilized in a cast, sought medical and physiotherapy advice in the initial 3 weeks, compared with those mobilized early with a functional brace.⁵⁰

Return to work and sport

Thirteen studies reported that early mobilization resulted in a significantly earlier return to work (about 30%–60% earlier),^{16,17,30,33,35,36,38-40,52,57-59} especially in lower-limb nonfracture injuries. In 1 study of ankle fractures, patients randomized to wear a cast reported better performance of daily activities at 6 weeks than those mobilized with crutches.²⁸ Perhaps, this reflects practical difficulties associated with using crutches. Five studies of lower-limb injury reported an earlier return to sport,^{16,30,39,57,60} as did 3 studies of postoperative Achilles tendon repair^30,39,60 and 2 of ankle ligament injury or sprains.^16,57 Five studies reporting a significantly earlier return to work or sport were of high quality, as were all studies of lower limb injuries.^{25,26,30,39,46,52}

Range of movement

Fourteen studies of upper- and lower-limb fractures reported significant improvements in range of movement. Significant differences in range of movement were typically observed within 13 weeks of the injury,^{12,27,29,34,38,42,45,46,49,52,61-64} although a few studies reported differences for longer periods. Similarly for lower-limb nonfractures, significantly better ranges of motion were found within short periods,^{25,33,37,39,43} and also after 12 months.^30,39-41 Eleven studies reporting this outcome were of high quality,^{25,26,30,31,39,45,46,48-51,54} and 4 studies found significant benefit for early mobilization.^{25,26,30,46,49} The other 7 high-quality studies favored early mobilization, and 1 found no difference.⁴⁵

Deformity and other complications

Four studies found early mobilization reduced deformity: for angular displacement⁶¹ and radial shortening³² in Colles’ fractures; displacement of metacarpal fractures⁵⁸ and cosmetic deformity in radial fractures.⁴² Two studies were of high quality.^30,65 Ten trials reported no changes in deformity, no loss of fracture reduction, or any other complications with early mobilization of fractures.^{28,31,45,48,50,59,62-64,66} In contrast, for Colles’ fractures, significantly greater dorsal angulation⁶¹ and significant increases in radial tilt and decreased radioulnar joint space⁴⁹ were reported for mobilized wrists.

Patient preferences

Patients expressed preference for a brace following anterior cruciate ligament surgery because it helped reduce pain and swelling⁶⁷; they expressed preference for a cast following surgery for ankle fracture because it improved ability to perform everyday activities.²⁸

Patients with distal radius fractures preferred early treatment with a Tubigrip bandage and a back-slab instead of removable splints, which increased pain and decreased mobilization.⁵³ Generally, patients with fractures preferred rigid support immediately after the injury. Some also expected to have a traditional cast applied after learning their injured wrist was fractured.⁵³

Cost of treatment

Full economic evaluations were not usually satisfactorily undertaken in these clinical studies. The cost of early mobilization was more expensive than immobilization in some studies^8,23 but cheaper in others.^2,9,10,24,68 It is estimated that direct costs to the patient were lower for early mobilization because of the decreased loss of working time.^10,44,68,69

Adherence and supervision

Ten studies excluded patients who were unable to understand the nature of the treatment; had dementia; existing joint disease; drug abuse; alcohol problems; or difficulty with walking aids.^{28,31,38-41,45,46,49,69} Thus, many researchers have assumed that early mobilization requires greater patient responsibility. Some researchers have recommended applying casts for patients deemed “uncooperative” or “unlikely to be compliant,” despite their positive findings for early mobilization.^3,16

Although 10 trials employed physiotherapy or supervised mobilization sessions,^{25,26,28,29,40,41,46,65,70} 11 studies relied on self-controlled mobilization.^{27,30,31,38,39,42,44,45,48,49,52} In addition, some interventions (eg, braces) intrinsically required more supervision with their removal and reapplication than others (eg, functional casts or semirigid bandages).

Discussion

Early mobilization seems to decrease pain, swelling, and stiffness—at least in the short-term—and patients generally prefer it to immobilization. It results in earlier return—to work and to a greater range of motion, which is most significant within the first 2 months of the injury and can be maintained for up to 12 months for nonfracture injuries. Early mobilization does not increase cosmetic or radiological deformity for stable fractures, and patients experience fewer complications and residual symptoms. However, early mobilization may place greater demands on patients and require higher levels of understanding and responsibility, therefore making it unsuitable for some patients.

This systematic review of all upper- and lower-limb injuries, including fractures, consistently found in favor of early mobilization over rest. We acknowledge that the range of injuries reviewed is very limited. However, the clear benefits of mobilization indicated by this review suggest we need research in a wider range of injuries.

Similarly, it would be naïve to assume mobilization is better than immobilization in all circumstances. Harm must occur at some level of increased activity. In addition we know that pain and discomfort are often experienced by patients who demand immobilization (a “palliative” form of management). Finding the ideal level and type of activity must be undertaken empirically. Therefore more quality clinical trials are required to determine and evaluate the best regimens for early mobilization.

Newer studies will probably find that mobilization can be employed more often (and perhaps more vigorously) than we now advise. The best evidence at hand suggests the medical profession generally errs too conservatively on the side of immobilization.

Acknowledgments

Charlotte Nash did the searches, assembled and analysed the tables and wrote the first draft. Charlotte Nash and Sharon Mickan assessed the quality of all included studies and edited drafts. Paul Glasziou and Chris Del Mar initiated the question and edited drafts. This work was supported in part by a scholarship awarded by the Centre for General Practice, The University of Queensland, and funded by the Australian Commonwealth Department of Health and Ageing through the Primary Health Care Research, Evaluation and Development Strategy.

Corresponding author
Professor Chris B. Del Mar, Centre for General Practice, Faculty Heath Science and Medicine, Bond University, Gold Coast, Queensland 4229, Australia. E-mail: [email protected].

We know that bed rest generally is over-prescribed.¹ But what about resting or immobilizing acutely injured limbs—from simple sprains to gunshot wounds? This specific application of rest is controversial.^2-6

Theoretically, rest should reduce further tissue damage by limiting movement, decreasing pain and swelling, and lessening loss of fracture reduction.^7-15 However, mobilization increases blood flow and so reduces muscle atrophy, disuse osteoporosis, adhesions, and joint stiffness.^7-15

The first empirical trials to determine whether rest or mobilization would offer greater benefit were undertaken in the 1980s, using both human and animal models. Systematic reviews have looked at specific injuries: acute ankle sprains¹⁶; ankle ligament rupture¹⁷; isolated ulnar fracture¹⁸; metacarpal fractures¹⁹; collateral ankle ligament sprains²⁰; and soft-tissue ankle injury.²¹ Each review independently suggests benefits for early mobilization.

However, most clinicians employ some period of enforced rest of injured limbs. For example, a survey of orthopedic surgeons in Denmark found early mobilization for ankle fractures is prescribed for only 6% of patients.²²

We systematically reviewed the literature to find trials that compared rest with early mobilization in acute limb injuries.

Methods

Protocol and search strategy

We searched the literature to identify prospective randomized controlled trials (RCTs) comparing rest with mobilization for acute limb injuries, searching the Cochrane Controlled Trials Register and Systematic Reviews, Medline (1966–2002), EMBASE (Rehabilitation and Physical Therapy), and Web of Science, combined with the following terms:

• immobilis*, immobiliz*, mobilis*, mobiliz*,
  injur*, fracture*, limb*, cast*, brace*, splint*,
  leg*, arm*, wrist*, elbow*, joint*, *carpal*,
  *tarsal*, knee*, ankle*, femur*, tibia*, fibula*,
  colles*, ulna*, radi*, humer*, sprain*, soft tissue,
  plaster slab, dislocat*

We accepted all publication languages and translated several papers. We found 445 potentially relevant papers by examining their titles or abstracts (Figure). Studies that did not fulfill our inclusion criteria were discarded.

We obtained full texts for 76 studies and undertook a prospective cited reference search in Web of Science for each study to identify more recent articles. A total of 187 papers were then evaluated further for duplicate data and inclusion criteria.

Trials that met our inclusion criteria:

1. Were randomized and controlled
2. Studied acute limb injuries
3. Compared groups whose only treatment difference was mobilization (studies that compared surgery and immobilization against nonsurgical early mobilization were not included; studies with internal fixation were included only if internal fixation was applied to both groups)
4. Had loss to follow-up of 20%, except where analysis was clearly intention to treat
5. Studied populations not focussed solely on young children, (studies in which the given age range included a small number of patients as young as 11 or 12 years were accepted—in these studies the median age was much higher)
6. adequately reported data collection and statistical analysis.

By these criteria, we excluded 138 studies (a list of which is available from the authors). Finally, 49 studies were included.

FIGURE
Literature search process, and methods by which randomized controlled trials (RCTs) were selected

Study methods quality and validity

Two authors independently scored the validity of each of the included 49 studies, meeting to discuss and resolve any disagreements. We relied on published data, without writing to authors to clarify or obtain information. Included studies were evaluated against 11 standard criteria.²³ Each criterion was scored 0–2. However, the practical maximum possible score was 18 because it was impossible to blind patients or their caregivers to mobilization. We considered the quality of 16 studies scoring 11 to be high.

Results

Across the 49 eligible studies, the different interventions were well described and differentiated, and data collection methods were clearly explained. Duration of immobilization in a cast was described (if at all) from 10 days to 8 weeks. Mobilization strategies varied and included active exercise, orthoses, crutches, or bandages.

Study quality was poor in many respects: several did not include important demographic data (including age, sex, or the numbers randomized to control and experimental group). Most did not report data about the methods of randomization. Many studies used inadequate randomization methods, with only 5 reporting optimal methods. Loss to follow-up was addressed poorly (not described or intention-to-treat analysis was not used). Outcome assessment was not blinded (a serious shortcoming in view of the subjective assessment of some patient outcomes). Reporting of inclusion criteria was inadequate in a few studies.

Given these deficiencies, our discussion emphasizes conclusions supported by studies of high quality.

Reporting of the studies

We divided the 49 studies into 4 groups: lower-limb fractures, other lower-limb injuries, upper-limb fractures, and other upper-limb injuries (Table W, at www.jfponline.com). Each group was further divided into trials using some form of limb support (such as a brace, splint, or a short period of immobilization) or minimal or no support (bandage, crutches, or tape) in the early mobilization group.

Most studies used some form of dynamic treatment for the injured limb, with highest use in lowerlimb injuries (74%); 2 studies included both supported and unsupported mobilization.^24-26 In studies of upper-limb fractures, support was used in 50%. In high-quality studies, 12 out of 16 studies used some support, but results were consistent with studies that did not. The smaller number of studies using minimal support makes comparison with supported mobilization difficult, but results suggest that supported mobilization may result in an earlier return to work or sport, whereas minimally supported patients achieved better composite scores and muscle strength. There was too much clinical heterogeneity to sensibly combine the results; however, we were interested to see whether mobilization was useful across a heterogeneous group of conditions.

Patient-centered outcomes

No study reported any significant improvement with rest on swelling and pain. To the contrary, 13 studies reported significant improvement with early mobilization compared with rest,^16,17,27-37 and 9 studies reported enhanced patient satisfaction.^{16,30,36,38-43} High-quality studies that reported pain levels found no difference or favored early mobilization.^{25,26,30,31,39,44-51} This was significant in only 2 studies of lower-limb nonfractures.^25,26,30 No study reported adverse patient opinion in an early mobilization regime,^{28,31,38,40,44-46,48,51-56} and patients appreciated their ability to use the affected limb.^34,43,49

Functional outcomes

Ten studies measured global function, using composite scores including subjective and objective criteria—pain, stiffness, swelling, use of supports, and ability to climb stairs, play sport, work, and undertake activities of daily life (Table W). Seven studies reported significant improvements with early mobilization—after 6 months (6 studies),^{28,35,42,46,52,57} or 12 months (1 study).³² Four studies of high quality found significant benefit for early mobilization in composite scores, most of these for lower-limb fractures.^46,47,52,54 Significantly more patients with acute ankle sprain, who were immobilized in a cast, sought medical and physiotherapy advice in the initial 3 weeks, compared with those mobilized early with a functional brace.⁵⁰

Return to work and sport

Thirteen studies reported that early mobilization resulted in a significantly earlier return to work (about 30%–60% earlier),^{16,17,30,33,35,36,38-40,52,57-59} especially in lower-limb nonfracture injuries. In 1 study of ankle fractures, patients randomized to wear a cast reported better performance of daily activities at 6 weeks than those mobilized with crutches.²⁸ Perhaps, this reflects practical difficulties associated with using crutches. Five studies of lower-limb injury reported an earlier return to sport,^{16,30,39,57,60} as did 3 studies of postoperative Achilles tendon repair^30,39,60 and 2 of ankle ligament injury or sprains.^16,57 Five studies reporting a significantly earlier return to work or sport were of high quality, as were all studies of lower limb injuries.^{25,26,30,39,46,52}

Range of movement

Fourteen studies of upper- and lower-limb fractures reported significant improvements in range of movement. Significant differences in range of movement were typically observed within 13 weeks of the injury,^{12,27,29,34,38,42,45,46,49,52,61-64} although a few studies reported differences for longer periods. Similarly for lower-limb nonfractures, significantly better ranges of motion were found within short periods,^{25,33,37,39,43} and also after 12 months.^30,39-41 Eleven studies reporting this outcome were of high quality,^{25,26,30,31,39,45,46,48-51,54} and 4 studies found significant benefit for early mobilization.^{25,26,30,46,49} The other 7 high-quality studies favored early mobilization, and 1 found no difference.⁴⁵

Deformity and other complications

Four studies found early mobilization reduced deformity: for angular displacement⁶¹ and radial shortening³² in Colles’ fractures; displacement of metacarpal fractures⁵⁸ and cosmetic deformity in radial fractures.⁴² Two studies were of high quality.^30,65 Ten trials reported no changes in deformity, no loss of fracture reduction, or any other complications with early mobilization of fractures.^{28,31,45,48,50,59,62-64,66} In contrast, for Colles’ fractures, significantly greater dorsal angulation⁶¹ and significant increases in radial tilt and decreased radioulnar joint space⁴⁹ were reported for mobilized wrists.

Patient preferences

Patients expressed preference for a brace following anterior cruciate ligament surgery because it helped reduce pain and swelling⁶⁷; they expressed preference for a cast following surgery for ankle fracture because it improved ability to perform everyday activities.²⁸

Patients with distal radius fractures preferred early treatment with a Tubigrip bandage and a back-slab instead of removable splints, which increased pain and decreased mobilization.⁵³ Generally, patients with fractures preferred rigid support immediately after the injury. Some also expected to have a traditional cast applied after learning their injured wrist was fractured.⁵³

Cost of treatment

Full economic evaluations were not usually satisfactorily undertaken in these clinical studies. The cost of early mobilization was more expensive than immobilization in some studies^8,23 but cheaper in others.^2,9,10,24,68 It is estimated that direct costs to the patient were lower for early mobilization because of the decreased loss of working time.^10,44,68,69

Adherence and supervision

Ten studies excluded patients who were unable to understand the nature of the treatment; had dementia; existing joint disease; drug abuse; alcohol problems; or difficulty with walking aids.^{28,31,38-41,45,46,49,69} Thus, many researchers have assumed that early mobilization requires greater patient responsibility. Some researchers have recommended applying casts for patients deemed “uncooperative” or “unlikely to be compliant,” despite their positive findings for early mobilization.^3,16

Although 10 trials employed physiotherapy or supervised mobilization sessions,^{25,26,28,29,40,41,46,65,70} 11 studies relied on self-controlled mobilization.^{27,30,31,38,39,42,44,45,48,49,52} In addition, some interventions (eg, braces) intrinsically required more supervision with their removal and reapplication than others (eg, functional casts or semirigid bandages).

Discussion

Early mobilization seems to decrease pain, swelling, and stiffness—at least in the short-term—and patients generally prefer it to immobilization. It results in earlier return—to work and to a greater range of motion, which is most significant within the first 2 months of the injury and can be maintained for up to 12 months for nonfracture injuries. Early mobilization does not increase cosmetic or radiological deformity for stable fractures, and patients experience fewer complications and residual symptoms. However, early mobilization may place greater demands on patients and require higher levels of understanding and responsibility, therefore making it unsuitable for some patients.

This systematic review of all upper- and lower-limb injuries, including fractures, consistently found in favor of early mobilization over rest. We acknowledge that the range of injuries reviewed is very limited. However, the clear benefits of mobilization indicated by this review suggest we need research in a wider range of injuries.

Similarly, it would be naïve to assume mobilization is better than immobilization in all circumstances. Harm must occur at some level of increased activity. In addition we know that pain and discomfort are often experienced by patients who demand immobilization (a “palliative” form of management). Finding the ideal level and type of activity must be undertaken empirically. Therefore more quality clinical trials are required to determine and evaluate the best regimens for early mobilization.

Newer studies will probably find that mobilization can be employed more often (and perhaps more vigorously) than we now advise. The best evidence at hand suggests the medical profession generally errs too conservatively on the side of immobilization.

Acknowledgments

Charlotte Nash did the searches, assembled and analysed the tables and wrote the first draft. Charlotte Nash and Sharon Mickan assessed the quality of all included studies and edited drafts. Paul Glasziou and Chris Del Mar initiated the question and edited drafts. This work was supported in part by a scholarship awarded by the Centre for General Practice, The University of Queensland, and funded by the Australian Commonwealth Department of Health and Ageing through the Primary Health Care Research, Evaluation and Development Strategy.

Corresponding author
Professor Chris B. Del Mar, Centre for General Practice, Faculty Heath Science and Medicine, Bond University, Gold Coast, Queensland 4229, Australia. E-mail: [email protected].

We know that bed rest generally is over-prescribed.¹ But what about resting or immobilizing acutely injured limbs—from simple sprains to gunshot wounds? This specific application of rest is controversial.^2-6

Theoretically, rest should reduce further tissue damage by limiting movement, decreasing pain and swelling, and lessening loss of fracture reduction.^7-15 However, mobilization increases blood flow and so reduces muscle atrophy, disuse osteoporosis, adhesions, and joint stiffness.^7-15

The first empirical trials to determine whether rest or mobilization would offer greater benefit were undertaken in the 1980s, using both human and animal models. Systematic reviews have looked at specific injuries: acute ankle sprains¹⁶; ankle ligament rupture¹⁷; isolated ulnar fracture¹⁸; metacarpal fractures¹⁹; collateral ankle ligament sprains²⁰; and soft-tissue ankle injury.²¹ Each review independently suggests benefits for early mobilization.

However, most clinicians employ some period of enforced rest of injured limbs. For example, a survey of orthopedic surgeons in Denmark found early mobilization for ankle fractures is prescribed for only 6% of patients.²²

We systematically reviewed the literature to find trials that compared rest with early mobilization in acute limb injuries.

Methods

Protocol and search strategy

We searched the literature to identify prospective randomized controlled trials (RCTs) comparing rest with mobilization for acute limb injuries, searching the Cochrane Controlled Trials Register and Systematic Reviews, Medline (1966–2002), EMBASE (Rehabilitation and Physical Therapy), and Web of Science, combined with the following terms:

• immobilis*, immobiliz*, mobilis*, mobiliz*,
  injur*, fracture*, limb*, cast*, brace*, splint*,
  leg*, arm*, wrist*, elbow*, joint*, *carpal*,
  *tarsal*, knee*, ankle*, femur*, tibia*, fibula*,
  colles*, ulna*, radi*, humer*, sprain*, soft tissue,
  plaster slab, dislocat*

We accepted all publication languages and translated several papers. We found 445 potentially relevant papers by examining their titles or abstracts (Figure). Studies that did not fulfill our inclusion criteria were discarded.

We obtained full texts for 76 studies and undertook a prospective cited reference search in Web of Science for each study to identify more recent articles. A total of 187 papers were then evaluated further for duplicate data and inclusion criteria.

Trials that met our inclusion criteria:

1. Were randomized and controlled
2. Studied acute limb injuries
3. Compared groups whose only treatment difference was mobilization (studies that compared surgery and immobilization against nonsurgical early mobilization were not included; studies with internal fixation were included only if internal fixation was applied to both groups)
4. Had loss to follow-up of 20%, except where analysis was clearly intention to treat
5. Studied populations not focussed solely on young children, (studies in which the given age range included a small number of patients as young as 11 or 12 years were accepted—in these studies the median age was much higher)
6. adequately reported data collection and statistical analysis.

By these criteria, we excluded 138 studies (a list of which is available from the authors). Finally, 49 studies were included.

FIGURE
Literature search process, and methods by which randomized controlled trials (RCTs) were selected

Study methods quality and validity

Two authors independently scored the validity of each of the included 49 studies, meeting to discuss and resolve any disagreements. We relied on published data, without writing to authors to clarify or obtain information. Included studies were evaluated against 11 standard criteria.²³ Each criterion was scored 0–2. However, the practical maximum possible score was 18 because it was impossible to blind patients or their caregivers to mobilization. We considered the quality of 16 studies scoring 11 to be high.

Results

Across the 49 eligible studies, the different interventions were well described and differentiated, and data collection methods were clearly explained. Duration of immobilization in a cast was described (if at all) from 10 days to 8 weeks. Mobilization strategies varied and included active exercise, orthoses, crutches, or bandages.

Study quality was poor in many respects: several did not include important demographic data (including age, sex, or the numbers randomized to control and experimental group). Most did not report data about the methods of randomization. Many studies used inadequate randomization methods, with only 5 reporting optimal methods. Loss to follow-up was addressed poorly (not described or intention-to-treat analysis was not used). Outcome assessment was not blinded (a serious shortcoming in view of the subjective assessment of some patient outcomes). Reporting of inclusion criteria was inadequate in a few studies.

Given these deficiencies, our discussion emphasizes conclusions supported by studies of high quality.

Reporting of the studies

We divided the 49 studies into 4 groups: lower-limb fractures, other lower-limb injuries, upper-limb fractures, and other upper-limb injuries (Table W, at www.jfponline.com). Each group was further divided into trials using some form of limb support (such as a brace, splint, or a short period of immobilization) or minimal or no support (bandage, crutches, or tape) in the early mobilization group.

Most studies used some form of dynamic treatment for the injured limb, with highest use in lowerlimb injuries (74%); 2 studies included both supported and unsupported mobilization.^24-26 In studies of upper-limb fractures, support was used in 50%. In high-quality studies, 12 out of 16 studies used some support, but results were consistent with studies that did not. The smaller number of studies using minimal support makes comparison with supported mobilization difficult, but results suggest that supported mobilization may result in an earlier return to work or sport, whereas minimally supported patients achieved better composite scores and muscle strength. There was too much clinical heterogeneity to sensibly combine the results; however, we were interested to see whether mobilization was useful across a heterogeneous group of conditions.

Patient-centered outcomes

No study reported any significant improvement with rest on swelling and pain. To the contrary, 13 studies reported significant improvement with early mobilization compared with rest,^16,17,27-37 and 9 studies reported enhanced patient satisfaction.^{16,30,36,38-43} High-quality studies that reported pain levels found no difference or favored early mobilization.^{25,26,30,31,39,44-51} This was significant in only 2 studies of lower-limb nonfractures.^25,26,30 No study reported adverse patient opinion in an early mobilization regime,^{28,31,38,40,44-46,48,51-56} and patients appreciated their ability to use the affected limb.^34,43,49

Functional outcomes

Ten studies measured global function, using composite scores including subjective and objective criteria—pain, stiffness, swelling, use of supports, and ability to climb stairs, play sport, work, and undertake activities of daily life (Table W). Seven studies reported significant improvements with early mobilization—after 6 months (6 studies),^{28,35,42,46,52,57} or 12 months (1 study).³² Four studies of high quality found significant benefit for early mobilization in composite scores, most of these for lower-limb fractures.^46,47,52,54 Significantly more patients with acute ankle sprain, who were immobilized in a cast, sought medical and physiotherapy advice in the initial 3 weeks, compared with those mobilized early with a functional brace.⁵⁰

Return to work and sport

Thirteen studies reported that early mobilization resulted in a significantly earlier return to work (about 30%–60% earlier),^{16,17,30,33,35,36,38-40,52,57-59} especially in lower-limb nonfracture injuries. In 1 study of ankle fractures, patients randomized to wear a cast reported better performance of daily activities at 6 weeks than those mobilized with crutches.²⁸ Perhaps, this reflects practical difficulties associated with using crutches. Five studies of lower-limb injury reported an earlier return to sport,^{16,30,39,57,60} as did 3 studies of postoperative Achilles tendon repair^30,39,60 and 2 of ankle ligament injury or sprains.^16,57 Five studies reporting a significantly earlier return to work or sport were of high quality, as were all studies of lower limb injuries.^{25,26,30,39,46,52}

Range of movement

Fourteen studies of upper- and lower-limb fractures reported significant improvements in range of movement. Significant differences in range of movement were typically observed within 13 weeks of the injury,^{12,27,29,34,38,42,45,46,49,52,61-64} although a few studies reported differences for longer periods. Similarly for lower-limb nonfractures, significantly better ranges of motion were found within short periods,^{25,33,37,39,43} and also after 12 months.^30,39-41 Eleven studies reporting this outcome were of high quality,^{25,26,30,31,39,45,46,48-51,54} and 4 studies found significant benefit for early mobilization.^{25,26,30,46,49} The other 7 high-quality studies favored early mobilization, and 1 found no difference.⁴⁵

Deformity and other complications

Four studies found early mobilization reduced deformity: for angular displacement⁶¹ and radial shortening³² in Colles’ fractures; displacement of metacarpal fractures⁵⁸ and cosmetic deformity in radial fractures.⁴² Two studies were of high quality.^30,65 Ten trials reported no changes in deformity, no loss of fracture reduction, or any other complications with early mobilization of fractures.^{28,31,45,48,50,59,62-64,66} In contrast, for Colles’ fractures, significantly greater dorsal angulation⁶¹ and significant increases in radial tilt and decreased radioulnar joint space⁴⁹ were reported for mobilized wrists.

Patient preferences

Patients expressed preference for a brace following anterior cruciate ligament surgery because it helped reduce pain and swelling⁶⁷; they expressed preference for a cast following surgery for ankle fracture because it improved ability to perform everyday activities.²⁸

Patients with distal radius fractures preferred early treatment with a Tubigrip bandage and a back-slab instead of removable splints, which increased pain and decreased mobilization.⁵³ Generally, patients with fractures preferred rigid support immediately after the injury. Some also expected to have a traditional cast applied after learning their injured wrist was fractured.⁵³

Cost of treatment

Full economic evaluations were not usually satisfactorily undertaken in these clinical studies. The cost of early mobilization was more expensive than immobilization in some studies^8,23 but cheaper in others.^2,9,10,24,68 It is estimated that direct costs to the patient were lower for early mobilization because of the decreased loss of working time.^10,44,68,69

Adherence and supervision

Ten studies excluded patients who were unable to understand the nature of the treatment; had dementia; existing joint disease; drug abuse; alcohol problems; or difficulty with walking aids.^{28,31,38-41,45,46,49,69} Thus, many researchers have assumed that early mobilization requires greater patient responsibility. Some researchers have recommended applying casts for patients deemed “uncooperative” or “unlikely to be compliant,” despite their positive findings for early mobilization.^3,16

Although 10 trials employed physiotherapy or supervised mobilization sessions,^{25,26,28,29,40,41,46,65,70} 11 studies relied on self-controlled mobilization.^{27,30,31,38,39,42,44,45,48,49,52} In addition, some interventions (eg, braces) intrinsically required more supervision with their removal and reapplication than others (eg, functional casts or semirigid bandages).

Discussion

Early mobilization seems to decrease pain, swelling, and stiffness—at least in the short-term—and patients generally prefer it to immobilization. It results in earlier return—to work and to a greater range of motion, which is most significant within the first 2 months of the injury and can be maintained for up to 12 months for nonfracture injuries. Early mobilization does not increase cosmetic or radiological deformity for stable fractures, and patients experience fewer complications and residual symptoms. However, early mobilization may place greater demands on patients and require higher levels of understanding and responsibility, therefore making it unsuitable for some patients.

This systematic review of all upper- and lower-limb injuries, including fractures, consistently found in favor of early mobilization over rest. We acknowledge that the range of injuries reviewed is very limited. However, the clear benefits of mobilization indicated by this review suggest we need research in a wider range of injuries.

Similarly, it would be naïve to assume mobilization is better than immobilization in all circumstances. Harm must occur at some level of increased activity. In addition we know that pain and discomfort are often experienced by patients who demand immobilization (a “palliative” form of management). Finding the ideal level and type of activity must be undertaken empirically. Therefore more quality clinical trials are required to determine and evaluate the best regimens for early mobilization.

Newer studies will probably find that mobilization can be employed more often (and perhaps more vigorously) than we now advise. The best evidence at hand suggests the medical profession generally errs too conservatively on the side of immobilization.

Acknowledgments

Charlotte Nash did the searches, assembled and analysed the tables and wrote the first draft. Charlotte Nash and Sharon Mickan assessed the quality of all included studies and edited drafts. Paul Glasziou and Chris Del Mar initiated the question and edited drafts. This work was supported in part by a scholarship awarded by the Centre for General Practice, The University of Queensland, and funded by the Australian Commonwealth Department of Health and Ageing through the Primary Health Care Research, Evaluation and Development Strategy.

Corresponding author
Professor Chris B. Del Mar, Centre for General Practice, Faculty Heath Science and Medicine, Bond University, Gold Coast, Queensland 4229, Australia. E-mail: [email protected].