Original Research

In the United States, the landscape of health care is changing. Health care reform and fluctuating political and economic climates have affected and will continue to affect the practice of orthopedic surgery. Demand for musculoskeletal care and the costs of providing this care are exceeding available resources—which has led to an evolution in how we practice as individuals and in the institutions where we provide care. Patient safety, quality, and value have become the outcomes of importance. Orthopedic surgeons, as experts in musculoskeletal care, must be a part of these changes. In this review, we offer perspective on the changing face of orthopedic surgery in the modern US health care system.

1. Meeting the demand

Musculoskeletal conditions represent one of the most common and costly health issues in the United States, affecting individuals medically and economically and compromising their quality of life.^1,2 In 2008, more than 110 million US adults (1 in 2) reported having a musculoskeletal condition for more than 3 months, and almost 7% reported that a chronic musculoskeletal condition made routine activities of daily living significantly difficult.¹ Overall, in the United States, some of the most common chronic conditions are musculoskeletal in origin. These conditions include osteoarthritis and back pain.

Osteoarthritis is the leading cause of chronic pain and disability. Physician-diagnosed arthritis is expected to affect 25% of US adults by 2030,³ and in more than one-third of these patients arthritis limits work or other activity.⁴ Back pain is another of the most common debilitating conditions in the United States.^3,5 St Sauver and colleagues⁶ found that back pain is the third most common condition (23.9%) that prompts patients to seek health care—following skin-related problems (42.7%) and osteoarthritis/joint pain (33.6%).

As life expectancy increases, so do expectations of enjoying higher levels of activity into the later years. Patients expect to be as active in their geriatric years as they were in middle age, and many are able to do so. Amid the growing obesity epidemic and increased incidence of chronic comorbidities, however, the aging population not only is at substantial risk for developing a chronic musculoskeletal disorder but may face new challenges in accessing care.

Although orthopedic surgeons specialize in treating musculoskeletal conditions, up to 90% of common nonsurgical musculoskeletal complaints are thought to be manageable in the primary care setting.⁷ With a disproportionate increase in musculoskeletal demand against a relatively constant number of orthopedic providers,⁸ it is becoming increasingly important for nonorthopedists to adequately manage musculoskeletal conditions. Physiatrists, rheumatologists, internists, family practitioners, and the expanding field of sports medicine specialists provide primary care of musculoskeletal conditions. To meet the growing demand and to ensure that patients receive quality, sustainable, effective, and efficient care, orthopedic surgeons should be actively involved in training these providers. As high as the cost of managing musculoskeletal conditions can be, it is far less than the cost resulting from inadequate or improper management. There is already justification for formal development of a specialization in nonoperative management of musculoskeletal care. Establishing this specialization requires a multidisciplinary approach, with orthopedic surgery taking a lead role.

2. The cost equation

As the prevalence of orthopedic conditions increases, so does the cost of delivering musculoskeletal care. The economic implications of meeting this growing demand are an important area of concern for our health care system. Steadily increasing hospital expenses for personnel and services, rising costs of pharmaceuticals and laboratory tests, constant evolution of costly technology, and insurance/reimbursement rates that do not keep pace with rising costs all contribute to the rapid escalation of the “cost of care.”

Health care expenditures accounted for 17.2% of the US gross domestic product (GDP) in 2012 and are expected to represent 19.3% by 2023.⁹ For musculoskeletal disease, direct costs alone are expected to approach $510 billion, equaling 5% of GDP and representing almost 30% of all health care expenditures. In Medicare patients, osteoarthritis is the most expensive condition to treat overall, and 3 other musculoskeletal problems rank highly as well: femoral neck fractures (3rd), back pain (10th), and fractures of all types (16th).¹⁰ Clearly, musculoskeletal care is one of the most prevalent and expensive health conditions in the United States.

Part of the direct costs of care that consistently increase each year are the steadily increasing costs of technology, which is often considered synonymous with orthopedic care. Promotion of new and more costly implants is common in the absence of evidence supporting their use. However, use of new implants and technology is being scrutinized in an effort to strike the proper cost–benefit balance.

To change the slope of the cost curve, orthopedic surgeons should utilize technological advances that are proven to be clinically significant and economically feasible and should avoid modest improvements with limited clinical benefit and higher price tags. Unfortunately, this approach is not being taken. Minor modifications of implant designs are often marketed as “new and improved” to justify increased costs, and these implants often gain widespread use. A few may prove to be clinically better, but most will be only comparable to older, less expensive designs, and some may end up being clinical failures, discovered at great cost to patients and the health care system.^11,12

Orthopedic surgeons have an important role in this decision-making. We should strive for the best, most cost-effective outcomes for our patients. We should reject new technology that does not clearly improve outcomes. At the least, we should use the technology in a manufacturer-supported clinical trial to determine its superiority. Whether the improvement is in technique, implant design, or workflow efficiency, orthopedic surgeons must be actively involved in researching and developing the latest innovations and must help determine their prospective value by considering not only their potential clinical benefits but also their economic implications.

As the political and economic environment becomes more directed at the cost-containment and sustainability of care, there has been a clear shift in focus to quality and value rather than volume, giving rise to the “value-based care” approach. The “value equation,” in which value equals quality divided by cost, requires a clear measure of outcomes and an equally clear understanding of costs. Delivering high-quality care in a cost-conscious environment is an approach that every orthopedic surgeon should adopt. Widespread adoption of the value-based strategy by hospital systems and insurance companies is resulting in a paradigm shift away from more traditional volume-based metrics and in favor of value-based metrics, including quality measures, patient-reported outcomes, Hospital Consumer Assessment of Healthcare Providers and Systems, and physician-specific outcome measures.

The new paradigm has brought the bundled payment initiative (BPI), a strategy included in the Patient Protection and Affordable Care Act. The philosophy behind the BPI model is for hospital systems and physicians to control costs while maintaining and improving the quality of care. Measured by patient metrics (eg, clinical outcomes, patient satisfaction) and hospital metrics (eg, readmission rates, cost of care), bundled payments reimburse hospitals on the basis of cost of an entire episode of care rather than on the basis of individual procedures and services. This approach provides incentives for both physicians and hospitals to promote value-based care while emphasizing coordination of care among all members of the health care team.

Providing the best possible care for our patients while holding our practice to the highest standards is a central tenet of the practice of orthopedic surgery and should be independent of reimbursement strategies. Thus, to increase the value of care, we must establish practice models and strategies to optimize cost-efficiency while improving outcomes. As explained by Porter and Teisberg,¹³ it is important to be conscientious about cost, but above all we must not allow quality of health care delivery to be compromised when trying to improve the “value” of care. Through evidence-based management and a clear understanding of costs, we must develop cost-efficient practice models that sustainably deliver the highest value of care.

3. Evolving practice models

As the health care landscape continues to change, physician practice models evolve accordingly. Although the private practice model once dominated the physician workforce, this is no longer true, as there has been a significant shift to employer-based practice models. The multiple factors at work relate to changing patterns of reimbursement, increasing government regulations, and a general change in recent residency graduates’ expectations regarding work–life balance. Other catalysts are the shift from volume- to value-based care and the recognition that cost-effective health care is more easily achieved when physicians and their institutions are in alignment. Ultimately, physician–institution alignment is crucial in improving care and outcomes.

Physician–institution alignment requires further discussion. Ideally, it should strike the proper balance between physician autonomy and institutional priorities to ensure the highest quality care. Physicians and their institutions should align their interests in terms of patient safety, quality, and economics to create a work environment conducive to both patient/physician satisfaction and institutional success.¹⁴ As identified by Page and colleagues,¹⁵ the primary drivers of physician–institution alignment, specific to orthopedic surgery, are economic, regulatory, and cultural. In economics, implant selection and ancillary services are the important issues; in the regulatory area, cooperative efforts to address expanding state and federal requirements are needed; last, the primary cultural driver is delivery of care to an expanding, diverse patient population.

Physician–institution alignment brings opportunities for “gainsharing,” which can directly benefit individual physicians, physician groups, and departments. Gainsharing is classically defined as “arrangements in which a hospital gives physicians a percentage share of any reduction in the hospital’s costs for patient care attributable in part to the physicians’ efforts.”¹⁶ Modern gainsharing programs can be used by institutions to align the economic interests of physicians and hospitals, with the ultimate goal being to achieve a sustainable increase in the value and quality of care delivered to patients.¹³ Examples include efforts to reduce the cost of orthopedic implants, which is a major cost driver in orthopedic surgery. Our institution realized significant savings when surgeons were directly involved in the implant contracting process with strategic sourcing personnel. These savings were shared with the department to enhance research and education programs. BPI, a risk-sharing program in which Medicare and hospitals participate, incorporates gainsharing opportunities in which each participating physician can receive up to 50% of his or her previous Medicare billings when specific targets are achieved. BPI included 27 musculoskeletal diagnosis–related groups that could be developed into a bundled payment proposal. Our institution participated in a 90-day episode, for primary hip and knee arthroplasty and non–cervical spine fusion, that had very promising results.

Gainsharing offers physicians incentives to meet institution goals of improved outcomes and increased patient satisfaction while increasing oversight and accountability. When physician-specific outcomes do not meet the established goals in key areas (readmissions, thromboembolic complications, infections), it is only logical that steps will be taken to improve outcomes. Although physicians may not be used to this increased scrutiny, the goal of improving outcomes, even if it necessitates a change in an established approach to care, should be welcomed.

Physicians should be rewarded for good outcomes but not suboptimal outcomes. When outcomes are suboptimal, physicians should take a constructive approach to improve them. On the other hand, not being rewarded for unachieved goals can be perceived as being penalized. Additional monitoring may paradoxically lead physicians to avoid more “complex” cases, such as those of patients at higher risk for complications and poorer outcomes. An example is found in patient selection for surgery, in which issues like obesity, diabetes, and heart disease are known to negatively affect outcomes. In these models, “cherry-picking” is a well-recognized risk^17,18 that can compromise our ethical obligation to provide equal access for all patients. To offset this tendency, we should use a risk-stratification model in which all patients are not considered equal in the risks they present. A risk-adjustment approach benefits both patients and providers by identifying modifiable risk factors that can be addressed to positively affect outcomes. This risk-stratification approach further incentivizes the orthopedist to closely work with other health care providers to address the medical comorbidities that may negatively affect surgical outcomes.

4. Patient and physician expectations

Living in a technology-driven society in the age of information has had a major impact on patients’ attitudes and expectations about their care—and therefore on physicians’ practice methods. It is uncommon to evaluate a patient who has not already consulted the Internet about a problem. Patients now have much more information they can use to make decisions about their treatment, and, though many question the accuracy of Internet information, there is no argument that being more informed is beneficial. In this time of shared decision-making, it is absolutely essential that patients keep themselves informed.

It is crucial to align the expectations of both physicians and patients in order to achieve the best outcomes. Gaining a clear understanding of treatment goals, management, and potential complications consistently leads to improved patient satisfaction, more favorable clinical outcomes, and reduced risk of litigation.^19-22 Addressing patient concerns and expectations is significantly enhanced by a strong patient–physician relationship through clinical models focused on patient-centered care.

Now considered a standard of care, the patient-centered model has changed the way we practice. The foundation of the patient-centered approach is to strengthen the patient–physician relationship by empowering patients to become active decision-makers in the management of their own health. The role of orthopedists in this model is to provide patients with information and insight into their conditions in order to facilitate shared decision-making. Our role should be to guide patients to make educated and informed decisions. Doing so enhances communication, thereby strengthening the patient–physician relationship, and places both patient and physician expectations in perspective. Patient-reported outcomes, satisfaction rates, symptomatic burdens, and costs of care are all positively correlated with strong communication and realistic expectations achieved through a patient-centered approach.^21,23

The evolution of clinical practice has been influenced by factors ranging from external forces (eg, changing political and economic climates) to social trends (use of social media and the Internet). Technology has been a driving force in our rapidly changing clinical environment, significantly altering the way we practice. Although we must be careful in how we use it, new technology can certainly work to our advantage. We have a plethora of medical information at our fingertips, and, with physician-directed guidance, our patients can become more informed than ever before. This is the principle of patient-centered medicine and shared decision-making, and its utility will only increase in importance.

5. The role of advocacy

The central tenet of orthopedic practice has always been a focus on patients. We continually strive to improve patient outcomes, reduce costs, and work efficiently in our practices and facilities. Although we can focus on our individual practices, we cannot ignore the influence and impact of the political system on our performance. Federal and state regulations give physicians and insurance companies an uneven playing field. This imbalance requires that physicians be more active in health care policymaking and advocacy. Although we are more involved than ever before, our influence is far less than what we would like it to be, perhaps partly because of the nature of the political process but perhaps also because of physicians’ resistance to becoming involved.

As experts in the treatment of musculoskeletal conditions, we should be at the forefront of health care policy development—a position we have not been able to attain. Although many factors contribute to our lack of a “seat at the table,” we must recognize our reluctance as a group to support advocacy, either financially or through personal time commitment. The American Association of Orthopaedic Surgeons (AAOS) Orthopaedic Political Action Committee has never been able to obtain donations from more than 30% of AAOS members. Although this committee historically has been successful, we could be much more so if we had financial support from 90% of members. There are many ways to be actively involved in advocacy. One way is to join local and state orthopedic societies and support their advocacy efforts. State orthopedic societies work closely with the AAOS Office of Government Relations to coordinate advocacy and direct efforts and resources to areas of greatest need. Knowing local congressional representatives and communicating with them about issues we face in our practices make our issues “real.” Some of our colleagues have even successfully run for office in Congress, and they certainly deserve our support. Advocacy will absolutely play an increasingly important role as federal and state governments expand their involvement in health care. Our role should be to get involved, at least to some degree. We need to recognize that our strength is in our numbers, as the few cannot accomplish nearly as much as the many.

Summary

Orthopedic surgeons are practicing in the midst of almost constant change—evolving patient care, shifts in employment models, advances in technology, modern patient expectations, and an increasingly complex regulatory environment. Even in this context, however, our goal remains unchanged: to give our patients the highest-quality care possible. Our core values as orthopedic surgeons and physicians are dedication, commitment, and service to patients and to our profession. As US health care continues to evolve, we must evolve as well, with an emphasis on expanding our role in the health care policy debate.

In the United States, the landscape of health care is changing. Health care reform and fluctuating political and economic climates have affected and will continue to affect the practice of orthopedic surgery. Demand for musculoskeletal care and the costs of providing this care are exceeding available resources—which has led to an evolution in how we practice as individuals and in the institutions where we provide care. Patient safety, quality, and value have become the outcomes of importance. Orthopedic surgeons, as experts in musculoskeletal care, must be a part of these changes. In this review, we offer perspective on the changing face of orthopedic surgery in the modern US health care system.

1. Meeting the demand

Musculoskeletal conditions represent one of the most common and costly health issues in the United States, affecting individuals medically and economically and compromising their quality of life.^1,2 In 2008, more than 110 million US adults (1 in 2) reported having a musculoskeletal condition for more than 3 months, and almost 7% reported that a chronic musculoskeletal condition made routine activities of daily living significantly difficult.¹ Overall, in the United States, some of the most common chronic conditions are musculoskeletal in origin. These conditions include osteoarthritis and back pain.

Osteoarthritis is the leading cause of chronic pain and disability. Physician-diagnosed arthritis is expected to affect 25% of US adults by 2030,³ and in more than one-third of these patients arthritis limits work or other activity.⁴ Back pain is another of the most common debilitating conditions in the United States.^3,5 St Sauver and colleagues⁶ found that back pain is the third most common condition (23.9%) that prompts patients to seek health care—following skin-related problems (42.7%) and osteoarthritis/joint pain (33.6%).

As life expectancy increases, so do expectations of enjoying higher levels of activity into the later years. Patients expect to be as active in their geriatric years as they were in middle age, and many are able to do so. Amid the growing obesity epidemic and increased incidence of chronic comorbidities, however, the aging population not only is at substantial risk for developing a chronic musculoskeletal disorder but may face new challenges in accessing care.

Although orthopedic surgeons specialize in treating musculoskeletal conditions, up to 90% of common nonsurgical musculoskeletal complaints are thought to be manageable in the primary care setting.⁷ With a disproportionate increase in musculoskeletal demand against a relatively constant number of orthopedic providers,⁸ it is becoming increasingly important for nonorthopedists to adequately manage musculoskeletal conditions. Physiatrists, rheumatologists, internists, family practitioners, and the expanding field of sports medicine specialists provide primary care of musculoskeletal conditions. To meet the growing demand and to ensure that patients receive quality, sustainable, effective, and efficient care, orthopedic surgeons should be actively involved in training these providers. As high as the cost of managing musculoskeletal conditions can be, it is far less than the cost resulting from inadequate or improper management. There is already justification for formal development of a specialization in nonoperative management of musculoskeletal care. Establishing this specialization requires a multidisciplinary approach, with orthopedic surgery taking a lead role.

2. The cost equation

As the prevalence of orthopedic conditions increases, so does the cost of delivering musculoskeletal care. The economic implications of meeting this growing demand are an important area of concern for our health care system. Steadily increasing hospital expenses for personnel and services, rising costs of pharmaceuticals and laboratory tests, constant evolution of costly technology, and insurance/reimbursement rates that do not keep pace with rising costs all contribute to the rapid escalation of the “cost of care.”

Health care expenditures accounted for 17.2% of the US gross domestic product (GDP) in 2012 and are expected to represent 19.3% by 2023.⁹ For musculoskeletal disease, direct costs alone are expected to approach $510 billion, equaling 5% of GDP and representing almost 30% of all health care expenditures. In Medicare patients, osteoarthritis is the most expensive condition to treat overall, and 3 other musculoskeletal problems rank highly as well: femoral neck fractures (3rd), back pain (10th), and fractures of all types (16th).¹⁰ Clearly, musculoskeletal care is one of the most prevalent and expensive health conditions in the United States.

Part of the direct costs of care that consistently increase each year are the steadily increasing costs of technology, which is often considered synonymous with orthopedic care. Promotion of new and more costly implants is common in the absence of evidence supporting their use. However, use of new implants and technology is being scrutinized in an effort to strike the proper cost–benefit balance.

To change the slope of the cost curve, orthopedic surgeons should utilize technological advances that are proven to be clinically significant and economically feasible and should avoid modest improvements with limited clinical benefit and higher price tags. Unfortunately, this approach is not being taken. Minor modifications of implant designs are often marketed as “new and improved” to justify increased costs, and these implants often gain widespread use. A few may prove to be clinically better, but most will be only comparable to older, less expensive designs, and some may end up being clinical failures, discovered at great cost to patients and the health care system.^11,12

Orthopedic surgeons have an important role in this decision-making. We should strive for the best, most cost-effective outcomes for our patients. We should reject new technology that does not clearly improve outcomes. At the least, we should use the technology in a manufacturer-supported clinical trial to determine its superiority. Whether the improvement is in technique, implant design, or workflow efficiency, orthopedic surgeons must be actively involved in researching and developing the latest innovations and must help determine their prospective value by considering not only their potential clinical benefits but also their economic implications.

As the political and economic environment becomes more directed at the cost-containment and sustainability of care, there has been a clear shift in focus to quality and value rather than volume, giving rise to the “value-based care” approach. The “value equation,” in which value equals quality divided by cost, requires a clear measure of outcomes and an equally clear understanding of costs. Delivering high-quality care in a cost-conscious environment is an approach that every orthopedic surgeon should adopt. Widespread adoption of the value-based strategy by hospital systems and insurance companies is resulting in a paradigm shift away from more traditional volume-based metrics and in favor of value-based metrics, including quality measures, patient-reported outcomes, Hospital Consumer Assessment of Healthcare Providers and Systems, and physician-specific outcome measures.

The new paradigm has brought the bundled payment initiative (BPI), a strategy included in the Patient Protection and Affordable Care Act. The philosophy behind the BPI model is for hospital systems and physicians to control costs while maintaining and improving the quality of care. Measured by patient metrics (eg, clinical outcomes, patient satisfaction) and hospital metrics (eg, readmission rates, cost of care), bundled payments reimburse hospitals on the basis of cost of an entire episode of care rather than on the basis of individual procedures and services. This approach provides incentives for both physicians and hospitals to promote value-based care while emphasizing coordination of care among all members of the health care team.

Providing the best possible care for our patients while holding our practice to the highest standards is a central tenet of the practice of orthopedic surgery and should be independent of reimbursement strategies. Thus, to increase the value of care, we must establish practice models and strategies to optimize cost-efficiency while improving outcomes. As explained by Porter and Teisberg,¹³ it is important to be conscientious about cost, but above all we must not allow quality of health care delivery to be compromised when trying to improve the “value” of care. Through evidence-based management and a clear understanding of costs, we must develop cost-efficient practice models that sustainably deliver the highest value of care.

3. Evolving practice models

As the health care landscape continues to change, physician practice models evolve accordingly. Although the private practice model once dominated the physician workforce, this is no longer true, as there has been a significant shift to employer-based practice models. The multiple factors at work relate to changing patterns of reimbursement, increasing government regulations, and a general change in recent residency graduates’ expectations regarding work–life balance. Other catalysts are the shift from volume- to value-based care and the recognition that cost-effective health care is more easily achieved when physicians and their institutions are in alignment. Ultimately, physician–institution alignment is crucial in improving care and outcomes.

Physician–institution alignment requires further discussion. Ideally, it should strike the proper balance between physician autonomy and institutional priorities to ensure the highest quality care. Physicians and their institutions should align their interests in terms of patient safety, quality, and economics to create a work environment conducive to both patient/physician satisfaction and institutional success.¹⁴ As identified by Page and colleagues,¹⁵ the primary drivers of physician–institution alignment, specific to orthopedic surgery, are economic, regulatory, and cultural. In economics, implant selection and ancillary services are the important issues; in the regulatory area, cooperative efforts to address expanding state and federal requirements are needed; last, the primary cultural driver is delivery of care to an expanding, diverse patient population.

Physician–institution alignment brings opportunities for “gainsharing,” which can directly benefit individual physicians, physician groups, and departments. Gainsharing is classically defined as “arrangements in which a hospital gives physicians a percentage share of any reduction in the hospital’s costs for patient care attributable in part to the physicians’ efforts.”¹⁶ Modern gainsharing programs can be used by institutions to align the economic interests of physicians and hospitals, with the ultimate goal being to achieve a sustainable increase in the value and quality of care delivered to patients.¹³ Examples include efforts to reduce the cost of orthopedic implants, which is a major cost driver in orthopedic surgery. Our institution realized significant savings when surgeons were directly involved in the implant contracting process with strategic sourcing personnel. These savings were shared with the department to enhance research and education programs. BPI, a risk-sharing program in which Medicare and hospitals participate, incorporates gainsharing opportunities in which each participating physician can receive up to 50% of his or her previous Medicare billings when specific targets are achieved. BPI included 27 musculoskeletal diagnosis–related groups that could be developed into a bundled payment proposal. Our institution participated in a 90-day episode, for primary hip and knee arthroplasty and non–cervical spine fusion, that had very promising results.

Gainsharing offers physicians incentives to meet institution goals of improved outcomes and increased patient satisfaction while increasing oversight and accountability. When physician-specific outcomes do not meet the established goals in key areas (readmissions, thromboembolic complications, infections), it is only logical that steps will be taken to improve outcomes. Although physicians may not be used to this increased scrutiny, the goal of improving outcomes, even if it necessitates a change in an established approach to care, should be welcomed.

Physicians should be rewarded for good outcomes but not suboptimal outcomes. When outcomes are suboptimal, physicians should take a constructive approach to improve them. On the other hand, not being rewarded for unachieved goals can be perceived as being penalized. Additional monitoring may paradoxically lead physicians to avoid more “complex” cases, such as those of patients at higher risk for complications and poorer outcomes. An example is found in patient selection for surgery, in which issues like obesity, diabetes, and heart disease are known to negatively affect outcomes. In these models, “cherry-picking” is a well-recognized risk^17,18 that can compromise our ethical obligation to provide equal access for all patients. To offset this tendency, we should use a risk-stratification model in which all patients are not considered equal in the risks they present. A risk-adjustment approach benefits both patients and providers by identifying modifiable risk factors that can be addressed to positively affect outcomes. This risk-stratification approach further incentivizes the orthopedist to closely work with other health care providers to address the medical comorbidities that may negatively affect surgical outcomes.

4. Patient and physician expectations

Living in a technology-driven society in the age of information has had a major impact on patients’ attitudes and expectations about their care—and therefore on physicians’ practice methods. It is uncommon to evaluate a patient who has not already consulted the Internet about a problem. Patients now have much more information they can use to make decisions about their treatment, and, though many question the accuracy of Internet information, there is no argument that being more informed is beneficial. In this time of shared decision-making, it is absolutely essential that patients keep themselves informed.

It is crucial to align the expectations of both physicians and patients in order to achieve the best outcomes. Gaining a clear understanding of treatment goals, management, and potential complications consistently leads to improved patient satisfaction, more favorable clinical outcomes, and reduced risk of litigation.^19-22 Addressing patient concerns and expectations is significantly enhanced by a strong patient–physician relationship through clinical models focused on patient-centered care.

Now considered a standard of care, the patient-centered model has changed the way we practice. The foundation of the patient-centered approach is to strengthen the patient–physician relationship by empowering patients to become active decision-makers in the management of their own health. The role of orthopedists in this model is to provide patients with information and insight into their conditions in order to facilitate shared decision-making. Our role should be to guide patients to make educated and informed decisions. Doing so enhances communication, thereby strengthening the patient–physician relationship, and places both patient and physician expectations in perspective. Patient-reported outcomes, satisfaction rates, symptomatic burdens, and costs of care are all positively correlated with strong communication and realistic expectations achieved through a patient-centered approach.^21,23

The evolution of clinical practice has been influenced by factors ranging from external forces (eg, changing political and economic climates) to social trends (use of social media and the Internet). Technology has been a driving force in our rapidly changing clinical environment, significantly altering the way we practice. Although we must be careful in how we use it, new technology can certainly work to our advantage. We have a plethora of medical information at our fingertips, and, with physician-directed guidance, our patients can become more informed than ever before. This is the principle of patient-centered medicine and shared decision-making, and its utility will only increase in importance.

5. The role of advocacy

The central tenet of orthopedic practice has always been a focus on patients. We continually strive to improve patient outcomes, reduce costs, and work efficiently in our practices and facilities. Although we can focus on our individual practices, we cannot ignore the influence and impact of the political system on our performance. Federal and state regulations give physicians and insurance companies an uneven playing field. This imbalance requires that physicians be more active in health care policymaking and advocacy. Although we are more involved than ever before, our influence is far less than what we would like it to be, perhaps partly because of the nature of the political process but perhaps also because of physicians’ resistance to becoming involved.

As experts in the treatment of musculoskeletal conditions, we should be at the forefront of health care policy development—a position we have not been able to attain. Although many factors contribute to our lack of a “seat at the table,” we must recognize our reluctance as a group to support advocacy, either financially or through personal time commitment. The American Association of Orthopaedic Surgeons (AAOS) Orthopaedic Political Action Committee has never been able to obtain donations from more than 30% of AAOS members. Although this committee historically has been successful, we could be much more so if we had financial support from 90% of members. There are many ways to be actively involved in advocacy. One way is to join local and state orthopedic societies and support their advocacy efforts. State orthopedic societies work closely with the AAOS Office of Government Relations to coordinate advocacy and direct efforts and resources to areas of greatest need. Knowing local congressional representatives and communicating with them about issues we face in our practices make our issues “real.” Some of our colleagues have even successfully run for office in Congress, and they certainly deserve our support. Advocacy will absolutely play an increasingly important role as federal and state governments expand their involvement in health care. Our role should be to get involved, at least to some degree. We need to recognize that our strength is in our numbers, as the few cannot accomplish nearly as much as the many.

Summary

Orthopedic surgeons are practicing in the midst of almost constant change—evolving patient care, shifts in employment models, advances in technology, modern patient expectations, and an increasingly complex regulatory environment. Even in this context, however, our goal remains unchanged: to give our patients the highest-quality care possible. Our core values as orthopedic surgeons and physicians are dedication, commitment, and service to patients and to our profession. As US health care continues to evolve, we must evolve as well, with an emphasis on expanding our role in the health care policy debate.

In the United States, the landscape of health care is changing. Health care reform and fluctuating political and economic climates have affected and will continue to affect the practice of orthopedic surgery. Demand for musculoskeletal care and the costs of providing this care are exceeding available resources—which has led to an evolution in how we practice as individuals and in the institutions where we provide care. Patient safety, quality, and value have become the outcomes of importance. Orthopedic surgeons, as experts in musculoskeletal care, must be a part of these changes. In this review, we offer perspective on the changing face of orthopedic surgery in the modern US health care system.

1. Meeting the demand

Musculoskeletal conditions represent one of the most common and costly health issues in the United States, affecting individuals medically and economically and compromising their quality of life.^1,2 In 2008, more than 110 million US adults (1 in 2) reported having a musculoskeletal condition for more than 3 months, and almost 7% reported that a chronic musculoskeletal condition made routine activities of daily living significantly difficult.¹ Overall, in the United States, some of the most common chronic conditions are musculoskeletal in origin. These conditions include osteoarthritis and back pain.

Osteoarthritis is the leading cause of chronic pain and disability. Physician-diagnosed arthritis is expected to affect 25% of US adults by 2030,³ and in more than one-third of these patients arthritis limits work or other activity.⁴ Back pain is another of the most common debilitating conditions in the United States.^3,5 St Sauver and colleagues⁶ found that back pain is the third most common condition (23.9%) that prompts patients to seek health care—following skin-related problems (42.7%) and osteoarthritis/joint pain (33.6%).

As life expectancy increases, so do expectations of enjoying higher levels of activity into the later years. Patients expect to be as active in their geriatric years as they were in middle age, and many are able to do so. Amid the growing obesity epidemic and increased incidence of chronic comorbidities, however, the aging population not only is at substantial risk for developing a chronic musculoskeletal disorder but may face new challenges in accessing care.

Although orthopedic surgeons specialize in treating musculoskeletal conditions, up to 90% of common nonsurgical musculoskeletal complaints are thought to be manageable in the primary care setting.⁷ With a disproportionate increase in musculoskeletal demand against a relatively constant number of orthopedic providers,⁸ it is becoming increasingly important for nonorthopedists to adequately manage musculoskeletal conditions. Physiatrists, rheumatologists, internists, family practitioners, and the expanding field of sports medicine specialists provide primary care of musculoskeletal conditions. To meet the growing demand and to ensure that patients receive quality, sustainable, effective, and efficient care, orthopedic surgeons should be actively involved in training these providers. As high as the cost of managing musculoskeletal conditions can be, it is far less than the cost resulting from inadequate or improper management. There is already justification for formal development of a specialization in nonoperative management of musculoskeletal care. Establishing this specialization requires a multidisciplinary approach, with orthopedic surgery taking a lead role.

2. The cost equation

As the prevalence of orthopedic conditions increases, so does the cost of delivering musculoskeletal care. The economic implications of meeting this growing demand are an important area of concern for our health care system. Steadily increasing hospital expenses for personnel and services, rising costs of pharmaceuticals and laboratory tests, constant evolution of costly technology, and insurance/reimbursement rates that do not keep pace with rising costs all contribute to the rapid escalation of the “cost of care.”

Health care expenditures accounted for 17.2% of the US gross domestic product (GDP) in 2012 and are expected to represent 19.3% by 2023.⁹ For musculoskeletal disease, direct costs alone are expected to approach $510 billion, equaling 5% of GDP and representing almost 30% of all health care expenditures. In Medicare patients, osteoarthritis is the most expensive condition to treat overall, and 3 other musculoskeletal problems rank highly as well: femoral neck fractures (3rd), back pain (10th), and fractures of all types (16th).¹⁰ Clearly, musculoskeletal care is one of the most prevalent and expensive health conditions in the United States.

Part of the direct costs of care that consistently increase each year are the steadily increasing costs of technology, which is often considered synonymous with orthopedic care. Promotion of new and more costly implants is common in the absence of evidence supporting their use. However, use of new implants and technology is being scrutinized in an effort to strike the proper cost–benefit balance.

To change the slope of the cost curve, orthopedic surgeons should utilize technological advances that are proven to be clinically significant and economically feasible and should avoid modest improvements with limited clinical benefit and higher price tags. Unfortunately, this approach is not being taken. Minor modifications of implant designs are often marketed as “new and improved” to justify increased costs, and these implants often gain widespread use. A few may prove to be clinically better, but most will be only comparable to older, less expensive designs, and some may end up being clinical failures, discovered at great cost to patients and the health care system.^11,12

Orthopedic surgeons have an important role in this decision-making. We should strive for the best, most cost-effective outcomes for our patients. We should reject new technology that does not clearly improve outcomes. At the least, we should use the technology in a manufacturer-supported clinical trial to determine its superiority. Whether the improvement is in technique, implant design, or workflow efficiency, orthopedic surgeons must be actively involved in researching and developing the latest innovations and must help determine their prospective value by considering not only their potential clinical benefits but also their economic implications.

As the political and economic environment becomes more directed at the cost-containment and sustainability of care, there has been a clear shift in focus to quality and value rather than volume, giving rise to the “value-based care” approach. The “value equation,” in which value equals quality divided by cost, requires a clear measure of outcomes and an equally clear understanding of costs. Delivering high-quality care in a cost-conscious environment is an approach that every orthopedic surgeon should adopt. Widespread adoption of the value-based strategy by hospital systems and insurance companies is resulting in a paradigm shift away from more traditional volume-based metrics and in favor of value-based metrics, including quality measures, patient-reported outcomes, Hospital Consumer Assessment of Healthcare Providers and Systems, and physician-specific outcome measures.

The new paradigm has brought the bundled payment initiative (BPI), a strategy included in the Patient Protection and Affordable Care Act. The philosophy behind the BPI model is for hospital systems and physicians to control costs while maintaining and improving the quality of care. Measured by patient metrics (eg, clinical outcomes, patient satisfaction) and hospital metrics (eg, readmission rates, cost of care), bundled payments reimburse hospitals on the basis of cost of an entire episode of care rather than on the basis of individual procedures and services. This approach provides incentives for both physicians and hospitals to promote value-based care while emphasizing coordination of care among all members of the health care team.

Providing the best possible care for our patients while holding our practice to the highest standards is a central tenet of the practice of orthopedic surgery and should be independent of reimbursement strategies. Thus, to increase the value of care, we must establish practice models and strategies to optimize cost-efficiency while improving outcomes. As explained by Porter and Teisberg,¹³ it is important to be conscientious about cost, but above all we must not allow quality of health care delivery to be compromised when trying to improve the “value” of care. Through evidence-based management and a clear understanding of costs, we must develop cost-efficient practice models that sustainably deliver the highest value of care.

3. Evolving practice models

As the health care landscape continues to change, physician practice models evolve accordingly. Although the private practice model once dominated the physician workforce, this is no longer true, as there has been a significant shift to employer-based practice models. The multiple factors at work relate to changing patterns of reimbursement, increasing government regulations, and a general change in recent residency graduates’ expectations regarding work–life balance. Other catalysts are the shift from volume- to value-based care and the recognition that cost-effective health care is more easily achieved when physicians and their institutions are in alignment. Ultimately, physician–institution alignment is crucial in improving care and outcomes.

Physician–institution alignment requires further discussion. Ideally, it should strike the proper balance between physician autonomy and institutional priorities to ensure the highest quality care. Physicians and their institutions should align their interests in terms of patient safety, quality, and economics to create a work environment conducive to both patient/physician satisfaction and institutional success.¹⁴ As identified by Page and colleagues,¹⁵ the primary drivers of physician–institution alignment, specific to orthopedic surgery, are economic, regulatory, and cultural. In economics, implant selection and ancillary services are the important issues; in the regulatory area, cooperative efforts to address expanding state and federal requirements are needed; last, the primary cultural driver is delivery of care to an expanding, diverse patient population.

Physician–institution alignment brings opportunities for “gainsharing,” which can directly benefit individual physicians, physician groups, and departments. Gainsharing is classically defined as “arrangements in which a hospital gives physicians a percentage share of any reduction in the hospital’s costs for patient care attributable in part to the physicians’ efforts.”¹⁶ Modern gainsharing programs can be used by institutions to align the economic interests of physicians and hospitals, with the ultimate goal being to achieve a sustainable increase in the value and quality of care delivered to patients.¹³ Examples include efforts to reduce the cost of orthopedic implants, which is a major cost driver in orthopedic surgery. Our institution realized significant savings when surgeons were directly involved in the implant contracting process with strategic sourcing personnel. These savings were shared with the department to enhance research and education programs. BPI, a risk-sharing program in which Medicare and hospitals participate, incorporates gainsharing opportunities in which each participating physician can receive up to 50% of his or her previous Medicare billings when specific targets are achieved. BPI included 27 musculoskeletal diagnosis–related groups that could be developed into a bundled payment proposal. Our institution participated in a 90-day episode, for primary hip and knee arthroplasty and non–cervical spine fusion, that had very promising results.

Gainsharing offers physicians incentives to meet institution goals of improved outcomes and increased patient satisfaction while increasing oversight and accountability. When physician-specific outcomes do not meet the established goals in key areas (readmissions, thromboembolic complications, infections), it is only logical that steps will be taken to improve outcomes. Although physicians may not be used to this increased scrutiny, the goal of improving outcomes, even if it necessitates a change in an established approach to care, should be welcomed.

Physicians should be rewarded for good outcomes but not suboptimal outcomes. When outcomes are suboptimal, physicians should take a constructive approach to improve them. On the other hand, not being rewarded for unachieved goals can be perceived as being penalized. Additional monitoring may paradoxically lead physicians to avoid more “complex” cases, such as those of patients at higher risk for complications and poorer outcomes. An example is found in patient selection for surgery, in which issues like obesity, diabetes, and heart disease are known to negatively affect outcomes. In these models, “cherry-picking” is a well-recognized risk^17,18 that can compromise our ethical obligation to provide equal access for all patients. To offset this tendency, we should use a risk-stratification model in which all patients are not considered equal in the risks they present. A risk-adjustment approach benefits both patients and providers by identifying modifiable risk factors that can be addressed to positively affect outcomes. This risk-stratification approach further incentivizes the orthopedist to closely work with other health care providers to address the medical comorbidities that may negatively affect surgical outcomes.

4. Patient and physician expectations

Living in a technology-driven society in the age of information has had a major impact on patients’ attitudes and expectations about their care—and therefore on physicians’ practice methods. It is uncommon to evaluate a patient who has not already consulted the Internet about a problem. Patients now have much more information they can use to make decisions about their treatment, and, though many question the accuracy of Internet information, there is no argument that being more informed is beneficial. In this time of shared decision-making, it is absolutely essential that patients keep themselves informed.

It is crucial to align the expectations of both physicians and patients in order to achieve the best outcomes. Gaining a clear understanding of treatment goals, management, and potential complications consistently leads to improved patient satisfaction, more favorable clinical outcomes, and reduced risk of litigation.^19-22 Addressing patient concerns and expectations is significantly enhanced by a strong patient–physician relationship through clinical models focused on patient-centered care.

Now considered a standard of care, the patient-centered model has changed the way we practice. The foundation of the patient-centered approach is to strengthen the patient–physician relationship by empowering patients to become active decision-makers in the management of their own health. The role of orthopedists in this model is to provide patients with information and insight into their conditions in order to facilitate shared decision-making. Our role should be to guide patients to make educated and informed decisions. Doing so enhances communication, thereby strengthening the patient–physician relationship, and places both patient and physician expectations in perspective. Patient-reported outcomes, satisfaction rates, symptomatic burdens, and costs of care are all positively correlated with strong communication and realistic expectations achieved through a patient-centered approach.^21,23

The evolution of clinical practice has been influenced by factors ranging from external forces (eg, changing political and economic climates) to social trends (use of social media and the Internet). Technology has been a driving force in our rapidly changing clinical environment, significantly altering the way we practice. Although we must be careful in how we use it, new technology can certainly work to our advantage. We have a plethora of medical information at our fingertips, and, with physician-directed guidance, our patients can become more informed than ever before. This is the principle of patient-centered medicine and shared decision-making, and its utility will only increase in importance.

5. The role of advocacy

The central tenet of orthopedic practice has always been a focus on patients. We continually strive to improve patient outcomes, reduce costs, and work efficiently in our practices and facilities. Although we can focus on our individual practices, we cannot ignore the influence and impact of the political system on our performance. Federal and state regulations give physicians and insurance companies an uneven playing field. This imbalance requires that physicians be more active in health care policymaking and advocacy. Although we are more involved than ever before, our influence is far less than what we would like it to be, perhaps partly because of the nature of the political process but perhaps also because of physicians’ resistance to becoming involved.

As experts in the treatment of musculoskeletal conditions, we should be at the forefront of health care policy development—a position we have not been able to attain. Although many factors contribute to our lack of a “seat at the table,” we must recognize our reluctance as a group to support advocacy, either financially or through personal time commitment. The American Association of Orthopaedic Surgeons (AAOS) Orthopaedic Political Action Committee has never been able to obtain donations from more than 30% of AAOS members. Although this committee historically has been successful, we could be much more so if we had financial support from 90% of members. There are many ways to be actively involved in advocacy. One way is to join local and state orthopedic societies and support their advocacy efforts. State orthopedic societies work closely with the AAOS Office of Government Relations to coordinate advocacy and direct efforts and resources to areas of greatest need. Knowing local congressional representatives and communicating with them about issues we face in our practices make our issues “real.” Some of our colleagues have even successfully run for office in Congress, and they certainly deserve our support. Advocacy will absolutely play an increasingly important role as federal and state governments expand their involvement in health care. Our role should be to get involved, at least to some degree. We need to recognize that our strength is in our numbers, as the few cannot accomplish nearly as much as the many.

Summary

Orthopedic surgeons are practicing in the midst of almost constant change—evolving patient care, shifts in employment models, advances in technology, modern patient expectations, and an increasingly complex regulatory environment. Even in this context, however, our goal remains unchanged: to give our patients the highest-quality care possible. Our core values as orthopedic surgeons and physicians are dedication, commitment, and service to patients and to our profession. As US health care continues to evolve, we must evolve as well, with an emphasis on expanding our role in the health care policy debate.

Posttraumatic stress disorder (PTSD) remains a significant health concern in veterans and military personnel. Whereas the lifetime incidence of PTSD in the U.S. general population is about 7% to 8%, the estimated prevalence of PTSD in deployed U.S. military personnel is higher than the national average, ranging from 11% to 17%.^1,2 These numbers may be even higher, depending on the branch of service, responsibilities within the military, and specific conflict in which the veteran served. For example, one study found that 31% of Vietnam veterans have PTSD, and another recent study has reported PTSD in 28.7% of veterans returning from military service in Iraq and Afghanistan.^3,4

Posttraumatic stress disorder treatment guidelines from both the American Psychiatric Association and the VA and DoD recommend the use of selective serotonin reuptake inhibitors (SSRIs) or serotonin-norepinephrine reuptake inhibitors (SNRIs) as first-line pharmacotherapy for PTSD.^5,6 However, SSRIs and SNRIs seem to be largely ineffective for the management of nighttime PTSD symptoms, such as insomnia and nightmares.^7,8

Related: PTSD Increases Chance of Heart Failure

Researchers hypothesize that the sympathetic nervous system plays a significant role in the hyperarousal component of nighttime PTSD. The heightened responsiveness and disruption in restorative sleep seen in PTSD have been attributed to increased activity of norepinephrine in the central nervous system.⁹ Mechanistically, therapies that attenuate the increased noradrenergic signaling might be effective in the management of nighttime PTSD symptoms.

The body of evidence for the use of adrenergic agents for nighttime PTSD symptoms is growing. Prazosin, a peripherally acting 
α1-adrenergic receptor antagonist, has recently been demonstrated to be effective for nighttime PTSD symptoms in veterans in a series of small, randomized controlled trials.^10-12 Data to support the use of clonidine, a centrally acting α2-adrenergic 
receptor agonist, are generally limited, with the most compelling data coming from a population of civilian Cambodian refugees.^13,14 A 2007 article by Boehnlein and Kinzie includes a thorough review of the preclinical research, case reports, and early clinical studies that have led to the widespread use of these agents for PTSD despite the lack of FDA approval for this indication.^13A previous retrospective review by Byers and colleagues compared the effectiveness and tolerability of prazosin and quetiapine for nighttime PTSD symptoms in veterans.¹⁵ The results of that review suggest that α1-adrenergic agents may be equally effective and better tolerated than alternative medication options (ie, atypical antipsychotics) for this purpose. The present study was adapted from this design to report concurrently on the real-world use of clonidine and prazosin for the treatment of nighttime PTSD
symptoms.

Study Objectives

The primary objective of this retrospective chart review was to describe the experience of patients prescribed clonidine or prazosin for the treatment of nighttime PTSD symptoms, including initial effectiveness. The primary endpoint of initial drug effectiveness was documented improvement of nighttime PTSD symptoms in the patient’s chart within 6 months of the date of first prescription. Clonidine or prazosin was categorized as initially effective if a statement such as “frequency of nightmares decreased” or “patient’s nighttime PTSD symptoms have improved” was made within 6 months after initial prescription of the drug.

The secondary objectives of this study were to evaluate the long-term effectiveness and tolerability of prazosin. The endpoints used to assess these outcomes were the 2-year continuation rates of clonidine and prazosin (as a surrogate marker for long-term effectiveness) and the documented reasons for discontinuation of clonidine and prazosin for the treatment of nighttime PTSD symptoms (in order to assess tolerability).

Methods

An electronic database search was conducted to identify the VA Portland Health Care System (VAPHCS) patients with a diagnosis of PTSD who received a first prescription for clonidine or prazosin for nighttime PTSD symptoms from a VAPHCS mental health provider or primary care provider (PCP) from January 1, 2009, to December 31, 2011. Patients were excluded if they had any history of prior use of the drug being initiated, were co-initiated on both clonidine and prazosin (defined as starting the drugs within 30 days of each other), or had a concomitant diagnosis of schizophrenia, bipolar 
disorder, psychotic disorder, or cognitive disorder as defined in the 
Diagnostic and Statistical Manual of Mental Disorders, 5th Edition. Patients with traumatic brain injury (TBI) were excluded only if it could be determined that the event had resulted in lasting cognitive impairment.

Study Population

All patients with a diagnosis of PTSD who received a first prescription for clonidine during the period specified were screened for inclusion; patients with PTSD who were first prescribed prazosin during the same period were randomly sampled to equalize patient populations. This was done to maximize the data set while examining groups of roughly equal size for each drug, as prazosin is used much more commonly than clonidine for nighttime PTSD symptoms at VAPHCS. The patients in each resulting group were screened to determine whether they met inclusion and exclusion criteria. All subjects included were followed for 2 years from the date of the initial prescription.

Study Design

Initial effectiveness of each agent was determined by reviewing subjects’ progress notes after the initial prescription of clonidine or prazosin for documentation of improvement in symptoms within 6 months of the prescription start date. A decrease in frequency or intensity of nighttime PTSD symptoms, nightmares, or insomnia, as documented in the patient chart, was interpreted as improvement of symptoms.

Long-term continuation was assessed by reviewing subjects’ prescription records, to determine whether prescription(s) for clonidine or prazosin continued for 2 years after the date of the initial prescription.

Any gap between medication fills that resulted in an anticipated period without medication of ≥ 6 months (eg, 9 months after receiving a 
90-day supply) was considered discontinuation of therapy. Prescription refill history was also reviewed, and medication possession ratio (MPR) was calculated to assess whether patients were adherent to the study drug as prescribed. Adherence was defined as an MPR of ≥ 80%. Patients who left the VAPHCS service area but continued to receive care at another VA were assessed for continuation of therapy, but refill data and/or MPR were not assessed.

Tolerability was assessed by reviewing subjects’ medical records to determine whether therapy with clonidine or prazosin was discontinued due to documented adverse effects (AEs). The occurrence of AEs was determined by reviewing progress notes and other chart documentation surrounding the date of discontinuation. If the drug was discontinued but the reason was not explicitly documented or if the prescription expired without a documented reason for nonrenewing, the reason for discontinuation was coded as “not specified.” Discontinuation due to treatment failure, change in symptoms, nonadherence, or other causes was also recorded. If multiple reasons for discontinuation were cited for a single patient, all were included in the data. This project was approved by the institutional review board at the VAPHCS.

Related:Depression and Substance Abuse Intensify Suicide Risk

Statistical Considerations

Based on clinical experience, it was presumed that many of the patients who were prescribed clonidine would be receiving it as a second-line therapy after failing prazosin. Therefore, statistical analysis of the relative effectiveness and tolerability of clonidine and prazosin could not be performed. Neither power nor sample size needed to demonstrate any difference in effectiveness or tolerability between the groups was calculated. All results are expressed using descriptive statistics.

Results

An initial database search for patients with PTSD who received a first prescription for clonidine between January 1, 2009, and December 31, 2011, from a VAPHCS provider yielded a list of 149 patients. The same search criteria applied for prazosin yielded 1,116 patients, 149 of whom were randomly selected for screening. After screening, 42 patients on clonidine and 60 patients on prazosin were included in this analysis (Figure).

Patient Demographics

The average age of the clonidine patients was 38.5 years (range 21-65 years) (Table 1). The clonidine group was primarily male (90%) and white (83%). Eighteen of the 42 patients in the clonidine group had a
baseline PTSD Checklist-Civilian version (PCL-C) score available within the 90 days before the first prescription of clonidine; the average baseline PCL-C score in this subgroup was 
62 ± 12.0 (median 65.5, range 
31-82). Most of the clonidine patients (71%) had a concomitant diagnosis of a depressive disorder. About one-quarter of the group (24%) had previously tried prazosin per prescription records. In 24 patients (57%), the first prescription for clonidine was written by a psychiatrist or psychiatric nurse practitioner; 18 patients (43%) were started on clonidine by PCPs.

The average age of the prazosin patients was 46.1 years (range 21-74 years). The prazosin group was also primarily male (93%) and white (88%). Twenty of the 60 patients in the prazosin group had a baseline PCL-C score available within the 
90 days before the first prescription of prazosin; the average baseline PCL-C score in this subgroup was 55 ± 16.1 (median 64, range 30-72). Most of the prazosin patients (63%) had a concomitant diagnosis of a depressive disorder. Four patients (7%) had previously tried clonidine per prescription records. In 35 patients (58%), the first prescription for prazosin was written by a psychiatrist or psychiatric nurse practitioner; 
25 patients (42%) were started on prazosin by PCPs.

Data pertaining to initial and long-term effectiveness, tolerability, and MPR for both clonidine and prazosin are presented in Table 2.

Clonidine

Of the 42 clonidine patients assessed, 24 (57%) had a positive response to the medication for nighttime PTSD symptoms documented in the Computerized Patient Record System (CPRS) within 6 months of starting therapy. Six months after starting clonidine, 23 patients (55%) continued to take clonidine. Two years after starting therapy, 8 of the original 
42 patients continued on clonidine for an overall 2-year continuation rate of 19%.

Tolerability

Of the 34 patients who discontinued clonidine within 2 years, 13 patients (38%) cited ineffectiveness of therapy as a reason for discontinuation. Another 13 patients (38%) reported discontinuing therapy due to AEs. Sedation (4 patients, 12%), dizziness/hypotension (3 patients, 9%), and paradoxical worsening of PTSD symptoms (4 patients, 12%) were the most common AEs leading to discontinuation. Other AEs cited as reasons for discontinuation were syncope 
(2 patients), erectile dysfunction 
(1 patient), rash (1 patient), myoclonus (1 patient), increased depression (1 patient), and fatigue (1 patient). One patient reported that he had discontinued clonidine due to symptom resolution/lack of need for treatment. In 8 of the 34 patients, no reason for discontinuation was found in chart documentation.

Medication Possession Ratio

Among the 21 evaluable patients who continued to receive clonidine 6 months after initiation, 10 (48%) were determined to be highly adherent to therapy, with an MPR of ≥ 80%. Six of the 21 patients (29%) had an MPR between 50% and 79%, and 
5 patients (24%) had an MPR < 50%.

Of the 8 patients who continued on clonidine at the 2-year mark, 
3 (38%) were adherent to therapy, with an MPR of ≥ 80%. Three more patients (38%) had a 2-year MPR between 50% and 80%, and 2 patients (25%) had an MPR < 50%.

Prazosin

Of the 60 prazosin patients assessed, 32 (53%) had a positive response to the medication for nighttime PTSD symptoms documented in the CPRS within 6 months of starting therapy. Six months after starting prazosin, 36 patients (60%) continued to take prazosin. Two years after starting therapy, 18 of the original 60 patients continued on prazosin for an overall 2-year continuation rate of 30%.

Tolerability

Of the 42 patients who discontinued prazosin within 2 years, six patients (14%) cited ineffectiveness of therapy as a reason for discontinuation. Thirteen patients (31%) reported discontinuing therapy due to AEs. Sedation (3 patients, 7%), dizziness/hypotension (3 patients, 7%), and paradoxical worsening of PTSD symptoms (6 patients, 14%) were the most common AEs leading to discontinuation. Other AEs cited as reasons for discontinuation were headache 
(2 patients), altered mental status (1 patient), and fatigue (1 patient). Three patients reported that they had discontinued clonidine due 
to symptom resolution/lack of need for treatment. Other reasons for discontinuation not related to 
AEs included flight rules (1 patient), changes to antihypertensive regimen (1 patient), refill issues (1 patient), and cost (1 patient). In 15 of 
the 42 patients, no reason for 
discontinuation was found in chart documentation.

Medication Possession Ratio

Among the 31 evaluable patients who continued to receive prazosin 
6 months after initiation, 20 (65%) were determined to be highly adherent to therapy, with an MPR of ≥ 80%. Five of the 31 patients (16%) had an MPR between 50% and 80%, and 6 patients (19%) had an MPR < 50%.

Of the 15 evaluable patients who continued on prazosin at the 2-year mark, 9 (60%) were adherent to therapy, with an MPR of ≥ 80%. Three patients (20%) had a 2-year MPR 
between 50% and 80%, and 3 patients 
(20%) had an MPR < 50%.

Discussion

Although prazosin has been shown to be effective for nighttime PTSD symptoms in both prospective and retrospective evaluations in veterans, this study provides the first evidence to support the use of clonidine in a veteran population.^10-12,15

Interestingly, 42% of the patients assessed received their first prescription of an α2-adrenergic agent for nighttime PTSD symptoms from a PCP. Even with the recent increased focus on integrating mental health into primary care within the VA, this was a surprising finding. Primary care providers at VAPHCS may have a greater role in the outpatient management of PTSD than previously suspected. The information presented here may prove useful and applicable in both psychiatric and primary care treatment settings.

The study results indicated that a majority of subjects initially reported effectiveness with either clonidine or prazosin (53% and 57%, respectively). The initial effectiveness rate for prazosin is similar to those described in previous studies.^10-13,15 The data also support a viable role for clonidine in the treatment of nighttime PTSD symptoms.

Regardless of initial improvement, the study results also suggest that the therapeutic benefit may not persist in the long term, as evidenced by a significant percentage of discontinuations attributed to ineffectiveness (38% for clonidine and 14% for prazosin) and a very low rate of long-term continuation (19% for clonidine and 30% for prazosin at 2 years). This latter observation contrasts with findings from previous studies; Byers and colleagues reported a 2-year prazosin continuation rate of 48.4% in a similar analysis, and Boehnlein and colleagues reported a sustained benefit of clonidine in responders over a 10-year period.^14,15 The wide variety of reasons for discontinuation reported here may help providers who are considering clonidine or prazosin for their patients to anticipate barriers to long-term success.

Part of the discrepancy between these results and previously reported successes with clonidine and prazosin may be attributable to the classic issue of efficacy vs effectiveness. Many of the studies that have informed us on the efficacy and tolerability of prazosin for nighttime PTSD symptoms described outcomes of prospective clinical research. Furthermore, these prospective trials were limited to < 6 months in duration. To date, neither clonidine nor prazosin has been evaluated for long-term efficacy and effectiveness in well-designed, prospective trials. This retrospective analysis may help provide a realistic estimate of the long-term effectiveness of these therapies, especially within the veteran population.

Limitations

This was a single-center, retrospective study conducted primarily in white male patients. Although likely applicable to the U.S. veteran population at large, these data may be poorly generalizable to patient populations outside the VA health care system.

Aside from external validity, this study has several significant limitations. The primary limitation of this project is that it was not designed to allow for statistical comparison of clonidine and prazosin. Such an analysis would have better defined the role of clonidine in PTSD treatment, either by establishing similar effectiveness of clonidine and prazosin for nighttime symptoms or by providing evidence of the superiority of one over the other. In designing the project, investigators suspected based on experience that the majority of patients prescribed clonidine would receive the drug after having already failed first-line therapy with prazosin. Had this been the case, a direct comparison may have been biased in favor of prazosin. In retrospect, however, only 24% of the clonidine group had previously been prescribed prazosin, and only 7% of the prazosin group had been prescribed clonidine. This suggests that clonidine may be used first line more often than the investigators anticipated and that a future direct comparison would be worthwhile.

Second, the subjective data collected for this project required investigators to read and interpret chart notes, although the review of all records by a single investigator helped limit variability in interpretation. At times, information in the CPRS was incomplete in terms of determining continuation of therapy or cause for discontinuation.

Third, although it is implied that a significant number of veterans have combat-related PTSD, the nature of the traumatic event(s) leading to PTSD was not recorded in this study, and no subgroup analysis was done to compare the effect of α2-adrenergic agents between combat- and noncombat-related PTSD. Owing to their exclusion by design, it is also difficult to apply these results to veterans who have lasting cognitive impairment as a result of TBI, who are presumably among those most likely to have experienced traumas that could provoke PTSD.

The design of this project also did not include a subgroup analysis based on antidepressant type, and it is unclear whether the potential pharmacodynamic interaction 
between noradrenergic antidepressants (ie, SNRIs) and anti–
α2-adrenergic agents had any impact on clinical outcomes. The use of complementary nonpharmacologic treatment modalities (ie, psychotherapy, eye movement desensitization and reprocessing) was also not evaluated.

Related: Female Service Members in the Long War

Finally, the primary outcome of patient-reported improvement in symptoms does not provide information on the magnitude or specific nature of benefits derived. Given the retrospective nature, data used in prospectively designed studies (eg, rating scales pertinent to PTSD), which might have helped to quantify the benefit of treatment, was not consistently available. Even a baseline PCL-C score, collected in order to describe the patient population, was available only in 37% of the patients assessed. Furthermore, nighttime PTSD symptoms vary among individuals, but the primary outcome of this study pools any benefits seen in areas such as nightmares, awakenings, night sweats, or sleep quality into a single outcome of symptom improvement.

Conclusions

This study indicates that both clonidine and prazosin may be effective for the treatment of nighttime PTSD symptoms in the veteran population but that their long-term utility may be limited by waning effectiveness, tolerability, and adherence issues. At this time, it is unclear whether either agent has an advantage over the other in terms of effectiveness or tolerability; further studies are needed to address that question.

Despite its limitations, the authors anticipate that this study will provide information regarding the effectiveness and tolerability of clonidine and prazosin to treat nighttime PTSD symptoms. Findings from this study may help clinicians to anticipate the needs and challenges of patients using β2-adrenergic agents for nighttime symptoms of PTSD.

Acknowledgements
The authors wish to acknowledge Brian Wilcox, PharmD, for his assistance in generating patient data reports, and Ronald Brown, RPh, MS, for his guidance regarding data analysis.

Author disclosures
The authors report no actual or potential conflicts of interest with regard to this article.

Disclaimer
The opinions expressed herein are those of the authors and do not necessarily reflect those of Federal Practitioner, Frontline Medical Communications Inc., the U.S. Government, or any of its agencies. This article may discuss unlabeled or investigational use of certain drugs. Please review the complete prescribing information for specific drugs or drug combinations—including indications, contraindications, warnings, and adverse effects—before administering pharmacologic therapy to patients.

Posttraumatic stress disorder (PTSD) remains a significant health concern in veterans and military personnel. Whereas the lifetime incidence of PTSD in the U.S. general population is about 7% to 8%, the estimated prevalence of PTSD in deployed U.S. military personnel is higher than the national average, ranging from 11% to 17%.^1,2 These numbers may be even higher, depending on the branch of service, responsibilities within the military, and specific conflict in which the veteran served. For example, one study found that 31% of Vietnam veterans have PTSD, and another recent study has reported PTSD in 28.7% of veterans returning from military service in Iraq and Afghanistan.^3,4

Posttraumatic stress disorder treatment guidelines from both the American Psychiatric Association and the VA and DoD recommend the use of selective serotonin reuptake inhibitors (SSRIs) or serotonin-norepinephrine reuptake inhibitors (SNRIs) as first-line pharmacotherapy for PTSD.^5,6 However, SSRIs and SNRIs seem to be largely ineffective for the management of nighttime PTSD symptoms, such as insomnia and nightmares.^7,8

Related: PTSD Increases Chance of Heart Failure

Researchers hypothesize that the sympathetic nervous system plays a significant role in the hyperarousal component of nighttime PTSD. The heightened responsiveness and disruption in restorative sleep seen in PTSD have been attributed to increased activity of norepinephrine in the central nervous system.⁹ Mechanistically, therapies that attenuate the increased noradrenergic signaling might be effective in the management of nighttime PTSD symptoms.

The body of evidence for the use of adrenergic agents for nighttime PTSD symptoms is growing. Prazosin, a peripherally acting 
α1-adrenergic receptor antagonist, has recently been demonstrated to be effective for nighttime PTSD symptoms in veterans in a series of small, randomized controlled trials.^10-12 Data to support the use of clonidine, a centrally acting α2-adrenergic 
receptor agonist, are generally limited, with the most compelling data coming from a population of civilian Cambodian refugees.^13,14 A 2007 article by Boehnlein and Kinzie includes a thorough review of the preclinical research, case reports, and early clinical studies that have led to the widespread use of these agents for PTSD despite the lack of FDA approval for this indication.^13A previous retrospective review by Byers and colleagues compared the effectiveness and tolerability of prazosin and quetiapine for nighttime PTSD symptoms in veterans.¹⁵ The results of that review suggest that α1-adrenergic agents may be equally effective and better tolerated than alternative medication options (ie, atypical antipsychotics) for this purpose. The present study was adapted from this design to report concurrently on the real-world use of clonidine and prazosin for the treatment of nighttime PTSD
symptoms.

Study Objectives

The primary objective of this retrospective chart review was to describe the experience of patients prescribed clonidine or prazosin for the treatment of nighttime PTSD symptoms, including initial effectiveness. The primary endpoint of initial drug effectiveness was documented improvement of nighttime PTSD symptoms in the patient’s chart within 6 months of the date of first prescription. Clonidine or prazosin was categorized as initially effective if a statement such as “frequency of nightmares decreased” or “patient’s nighttime PTSD symptoms have improved” was made within 6 months after initial prescription of the drug.

The secondary objectives of this study were to evaluate the long-term effectiveness and tolerability of prazosin. The endpoints used to assess these outcomes were the 2-year continuation rates of clonidine and prazosin (as a surrogate marker for long-term effectiveness) and the documented reasons for discontinuation of clonidine and prazosin for the treatment of nighttime PTSD symptoms (in order to assess tolerability).

Methods

An electronic database search was conducted to identify the VA Portland Health Care System (VAPHCS) patients with a diagnosis of PTSD who received a first prescription for clonidine or prazosin for nighttime PTSD symptoms from a VAPHCS mental health provider or primary care provider (PCP) from January 1, 2009, to December 31, 2011. Patients were excluded if they had any history of prior use of the drug being initiated, were co-initiated on both clonidine and prazosin (defined as starting the drugs within 30 days of each other), or had a concomitant diagnosis of schizophrenia, bipolar 
disorder, psychotic disorder, or cognitive disorder as defined in the 
Diagnostic and Statistical Manual of Mental Disorders, 5th Edition. Patients with traumatic brain injury (TBI) were excluded only if it could be determined that the event had resulted in lasting cognitive impairment.

Study Population

All patients with a diagnosis of PTSD who received a first prescription for clonidine during the period specified were screened for inclusion; patients with PTSD who were first prescribed prazosin during the same period were randomly sampled to equalize patient populations. This was done to maximize the data set while examining groups of roughly equal size for each drug, as prazosin is used much more commonly than clonidine for nighttime PTSD symptoms at VAPHCS. The patients in each resulting group were screened to determine whether they met inclusion and exclusion criteria. All subjects included were followed for 2 years from the date of the initial prescription.

Study Design

Initial effectiveness of each agent was determined by reviewing subjects’ progress notes after the initial prescription of clonidine or prazosin for documentation of improvement in symptoms within 6 months of the prescription start date. A decrease in frequency or intensity of nighttime PTSD symptoms, nightmares, or insomnia, as documented in the patient chart, was interpreted as improvement of symptoms.

Long-term continuation was assessed by reviewing subjects’ prescription records, to determine whether prescription(s) for clonidine or prazosin continued for 2 years after the date of the initial prescription.

Any gap between medication fills that resulted in an anticipated period without medication of ≥ 6 months (eg, 9 months after receiving a 
90-day supply) was considered discontinuation of therapy. Prescription refill history was also reviewed, and medication possession ratio (MPR) was calculated to assess whether patients were adherent to the study drug as prescribed. Adherence was defined as an MPR of ≥ 80%. Patients who left the VAPHCS service area but continued to receive care at another VA were assessed for continuation of therapy, but refill data and/or MPR were not assessed.

Tolerability was assessed by reviewing subjects’ medical records to determine whether therapy with clonidine or prazosin was discontinued due to documented adverse effects (AEs). The occurrence of AEs was determined by reviewing progress notes and other chart documentation surrounding the date of discontinuation. If the drug was discontinued but the reason was not explicitly documented or if the prescription expired without a documented reason for nonrenewing, the reason for discontinuation was coded as “not specified.” Discontinuation due to treatment failure, change in symptoms, nonadherence, or other causes was also recorded. If multiple reasons for discontinuation were cited for a single patient, all were included in the data. This project was approved by the institutional review board at the VAPHCS.

Related:Depression and Substance Abuse Intensify Suicide Risk

Statistical Considerations

Based on clinical experience, it was presumed that many of the patients who were prescribed clonidine would be receiving it as a second-line therapy after failing prazosin. Therefore, statistical analysis of the relative effectiveness and tolerability of clonidine and prazosin could not be performed. Neither power nor sample size needed to demonstrate any difference in effectiveness or tolerability between the groups was calculated. All results are expressed using descriptive statistics.

Results

An initial database search for patients with PTSD who received a first prescription for clonidine between January 1, 2009, and December 31, 2011, from a VAPHCS provider yielded a list of 149 patients. The same search criteria applied for prazosin yielded 1,116 patients, 149 of whom were randomly selected for screening. After screening, 42 patients on clonidine and 60 patients on prazosin were included in this analysis (Figure).

Patient Demographics

The average age of the clonidine patients was 38.5 years (range 21-65 years) (Table 1). The clonidine group was primarily male (90%) and white (83%). Eighteen of the 42 patients in the clonidine group had a
baseline PTSD Checklist-Civilian version (PCL-C) score available within the 90 days before the first prescription of clonidine; the average baseline PCL-C score in this subgroup was 
62 ± 12.0 (median 65.5, range 
31-82). Most of the clonidine patients (71%) had a concomitant diagnosis of a depressive disorder. About one-quarter of the group (24%) had previously tried prazosin per prescription records. In 24 patients (57%), the first prescription for clonidine was written by a psychiatrist or psychiatric nurse practitioner; 18 patients (43%) were started on clonidine by PCPs.

The average age of the prazosin patients was 46.1 years (range 21-74 years). The prazosin group was also primarily male (93%) and white (88%). Twenty of the 60 patients in the prazosin group had a baseline PCL-C score available within the 
90 days before the first prescription of prazosin; the average baseline PCL-C score in this subgroup was 55 ± 16.1 (median 64, range 30-72). Most of the prazosin patients (63%) had a concomitant diagnosis of a depressive disorder. Four patients (7%) had previously tried clonidine per prescription records. In 35 patients (58%), the first prescription for prazosin was written by a psychiatrist or psychiatric nurse practitioner; 
25 patients (42%) were started on prazosin by PCPs.

Data pertaining to initial and long-term effectiveness, tolerability, and MPR for both clonidine and prazosin are presented in Table 2.

Clonidine

Of the 42 clonidine patients assessed, 24 (57%) had a positive response to the medication for nighttime PTSD symptoms documented in the Computerized Patient Record System (CPRS) within 6 months of starting therapy. Six months after starting clonidine, 23 patients (55%) continued to take clonidine. Two years after starting therapy, 8 of the original 
42 patients continued on clonidine for an overall 2-year continuation rate of 19%.

Tolerability

Of the 34 patients who discontinued clonidine within 2 years, 13 patients (38%) cited ineffectiveness of therapy as a reason for discontinuation. Another 13 patients (38%) reported discontinuing therapy due to AEs. Sedation (4 patients, 12%), dizziness/hypotension (3 patients, 9%), and paradoxical worsening of PTSD symptoms (4 patients, 12%) were the most common AEs leading to discontinuation. Other AEs cited as reasons for discontinuation were syncope 
(2 patients), erectile dysfunction 
(1 patient), rash (1 patient), myoclonus (1 patient), increased depression (1 patient), and fatigue (1 patient). One patient reported that he had discontinued clonidine due to symptom resolution/lack of need for treatment. In 8 of the 34 patients, no reason for discontinuation was found in chart documentation.

Medication Possession Ratio

Among the 21 evaluable patients who continued to receive clonidine 6 months after initiation, 10 (48%) were determined to be highly adherent to therapy, with an MPR of ≥ 80%. Six of the 21 patients (29%) had an MPR between 50% and 79%, and 
5 patients (24%) had an MPR < 50%.

Of the 8 patients who continued on clonidine at the 2-year mark, 
3 (38%) were adherent to therapy, with an MPR of ≥ 80%. Three more patients (38%) had a 2-year MPR between 50% and 80%, and 2 patients (25%) had an MPR < 50%.

Prazosin

Of the 60 prazosin patients assessed, 32 (53%) had a positive response to the medication for nighttime PTSD symptoms documented in the CPRS within 6 months of starting therapy. Six months after starting prazosin, 36 patients (60%) continued to take prazosin. Two years after starting therapy, 18 of the original 60 patients continued on prazosin for an overall 2-year continuation rate of 30%.

Tolerability

Of the 42 patients who discontinued prazosin within 2 years, six patients (14%) cited ineffectiveness of therapy as a reason for discontinuation. Thirteen patients (31%) reported discontinuing therapy due to AEs. Sedation (3 patients, 7%), dizziness/hypotension (3 patients, 7%), and paradoxical worsening of PTSD symptoms (6 patients, 14%) were the most common AEs leading to discontinuation. Other AEs cited as reasons for discontinuation were headache 
(2 patients), altered mental status (1 patient), and fatigue (1 patient). Three patients reported that they had discontinued clonidine due 
to symptom resolution/lack of need for treatment. Other reasons for discontinuation not related to 
AEs included flight rules (1 patient), changes to antihypertensive regimen (1 patient), refill issues (1 patient), and cost (1 patient). In 15 of 
the 42 patients, no reason for 
discontinuation was found in chart documentation.

Medication Possession Ratio

Among the 31 evaluable patients who continued to receive prazosin 
6 months after initiation, 20 (65%) were determined to be highly adherent to therapy, with an MPR of ≥ 80%. Five of the 31 patients (16%) had an MPR between 50% and 80%, and 6 patients (19%) had an MPR < 50%.

Of the 15 evaluable patients who continued on prazosin at the 2-year mark, 9 (60%) were adherent to therapy, with an MPR of ≥ 80%. Three patients (20%) had a 2-year MPR 
between 50% and 80%, and 3 patients 
(20%) had an MPR < 50%.

Discussion

Although prazosin has been shown to be effective for nighttime PTSD symptoms in both prospective and retrospective evaluations in veterans, this study provides the first evidence to support the use of clonidine in a veteran population.^10-12,15

Interestingly, 42% of the patients assessed received their first prescription of an α2-adrenergic agent for nighttime PTSD symptoms from a PCP. Even with the recent increased focus on integrating mental health into primary care within the VA, this was a surprising finding. Primary care providers at VAPHCS may have a greater role in the outpatient management of PTSD than previously suspected. The information presented here may prove useful and applicable in both psychiatric and primary care treatment settings.

The study results indicated that a majority of subjects initially reported effectiveness with either clonidine or prazosin (53% and 57%, respectively). The initial effectiveness rate for prazosin is similar to those described in previous studies.^10-13,15 The data also support a viable role for clonidine in the treatment of nighttime PTSD symptoms.

Regardless of initial improvement, the study results also suggest that the therapeutic benefit may not persist in the long term, as evidenced by a significant percentage of discontinuations attributed to ineffectiveness (38% for clonidine and 14% for prazosin) and a very low rate of long-term continuation (19% for clonidine and 30% for prazosin at 2 years). This latter observation contrasts with findings from previous studies; Byers and colleagues reported a 2-year prazosin continuation rate of 48.4% in a similar analysis, and Boehnlein and colleagues reported a sustained benefit of clonidine in responders over a 10-year period.^14,15 The wide variety of reasons for discontinuation reported here may help providers who are considering clonidine or prazosin for their patients to anticipate barriers to long-term success.

Part of the discrepancy between these results and previously reported successes with clonidine and prazosin may be attributable to the classic issue of efficacy vs effectiveness. Many of the studies that have informed us on the efficacy and tolerability of prazosin for nighttime PTSD symptoms described outcomes of prospective clinical research. Furthermore, these prospective trials were limited to < 6 months in duration. To date, neither clonidine nor prazosin has been evaluated for long-term efficacy and effectiveness in well-designed, prospective trials. This retrospective analysis may help provide a realistic estimate of the long-term effectiveness of these therapies, especially within the veteran population.

Limitations

This was a single-center, retrospective study conducted primarily in white male patients. Although likely applicable to the U.S. veteran population at large, these data may be poorly generalizable to patient populations outside the VA health care system.

Aside from external validity, this study has several significant limitations. The primary limitation of this project is that it was not designed to allow for statistical comparison of clonidine and prazosin. Such an analysis would have better defined the role of clonidine in PTSD treatment, either by establishing similar effectiveness of clonidine and prazosin for nighttime symptoms or by providing evidence of the superiority of one over the other. In designing the project, investigators suspected based on experience that the majority of patients prescribed clonidine would receive the drug after having already failed first-line therapy with prazosin. Had this been the case, a direct comparison may have been biased in favor of prazosin. In retrospect, however, only 24% of the clonidine group had previously been prescribed prazosin, and only 7% of the prazosin group had been prescribed clonidine. This suggests that clonidine may be used first line more often than the investigators anticipated and that a future direct comparison would be worthwhile.

Second, the subjective data collected for this project required investigators to read and interpret chart notes, although the review of all records by a single investigator helped limit variability in interpretation. At times, information in the CPRS was incomplete in terms of determining continuation of therapy or cause for discontinuation.

Third, although it is implied that a significant number of veterans have combat-related PTSD, the nature of the traumatic event(s) leading to PTSD was not recorded in this study, and no subgroup analysis was done to compare the effect of α2-adrenergic agents between combat- and noncombat-related PTSD. Owing to their exclusion by design, it is also difficult to apply these results to veterans who have lasting cognitive impairment as a result of TBI, who are presumably among those most likely to have experienced traumas that could provoke PTSD.

The design of this project also did not include a subgroup analysis based on antidepressant type, and it is unclear whether the potential pharmacodynamic interaction 
between noradrenergic antidepressants (ie, SNRIs) and anti–
α2-adrenergic agents had any impact on clinical outcomes. The use of complementary nonpharmacologic treatment modalities (ie, psychotherapy, eye movement desensitization and reprocessing) was also not evaluated.

Related: Female Service Members in the Long War

Finally, the primary outcome of patient-reported improvement in symptoms does not provide information on the magnitude or specific nature of benefits derived. Given the retrospective nature, data used in prospectively designed studies (eg, rating scales pertinent to PTSD), which might have helped to quantify the benefit of treatment, was not consistently available. Even a baseline PCL-C score, collected in order to describe the patient population, was available only in 37% of the patients assessed. Furthermore, nighttime PTSD symptoms vary among individuals, but the primary outcome of this study pools any benefits seen in areas such as nightmares, awakenings, night sweats, or sleep quality into a single outcome of symptom improvement.

Conclusions

This study indicates that both clonidine and prazosin may be effective for the treatment of nighttime PTSD symptoms in the veteran population but that their long-term utility may be limited by waning effectiveness, tolerability, and adherence issues. At this time, it is unclear whether either agent has an advantage over the other in terms of effectiveness or tolerability; further studies are needed to address that question.

Despite its limitations, the authors anticipate that this study will provide information regarding the effectiveness and tolerability of clonidine and prazosin to treat nighttime PTSD symptoms. Findings from this study may help clinicians to anticipate the needs and challenges of patients using β2-adrenergic agents for nighttime symptoms of PTSD.

Acknowledgements
The authors wish to acknowledge Brian Wilcox, PharmD, for his assistance in generating patient data reports, and Ronald Brown, RPh, MS, for his guidance regarding data analysis.

Author disclosures
The authors report no actual or potential conflicts of interest with regard to this article.

Disclaimer
The opinions expressed herein are those of the authors and do not necessarily reflect those of Federal Practitioner, Frontline Medical Communications Inc., the U.S. Government, or any of its agencies. This article may discuss unlabeled or investigational use of certain drugs. Please review the complete prescribing information for specific drugs or drug combinations—including indications, contraindications, warnings, and adverse effects—before administering pharmacologic therapy to patients.

Posttraumatic stress disorder (PTSD) remains a significant health concern in veterans and military personnel. Whereas the lifetime incidence of PTSD in the U.S. general population is about 7% to 8%, the estimated prevalence of PTSD in deployed U.S. military personnel is higher than the national average, ranging from 11% to 17%.^1,2 These numbers may be even higher, depending on the branch of service, responsibilities within the military, and specific conflict in which the veteran served. For example, one study found that 31% of Vietnam veterans have PTSD, and another recent study has reported PTSD in 28.7% of veterans returning from military service in Iraq and Afghanistan.^3,4

Posttraumatic stress disorder treatment guidelines from both the American Psychiatric Association and the VA and DoD recommend the use of selective serotonin reuptake inhibitors (SSRIs) or serotonin-norepinephrine reuptake inhibitors (SNRIs) as first-line pharmacotherapy for PTSD.^5,6 However, SSRIs and SNRIs seem to be largely ineffective for the management of nighttime PTSD symptoms, such as insomnia and nightmares.^7,8

Related: PTSD Increases Chance of Heart Failure

Researchers hypothesize that the sympathetic nervous system plays a significant role in the hyperarousal component of nighttime PTSD. The heightened responsiveness and disruption in restorative sleep seen in PTSD have been attributed to increased activity of norepinephrine in the central nervous system.⁹ Mechanistically, therapies that attenuate the increased noradrenergic signaling might be effective in the management of nighttime PTSD symptoms.

The body of evidence for the use of adrenergic agents for nighttime PTSD symptoms is growing. Prazosin, a peripherally acting 
α1-adrenergic receptor antagonist, has recently been demonstrated to be effective for nighttime PTSD symptoms in veterans in a series of small, randomized controlled trials.^10-12 Data to support the use of clonidine, a centrally acting α2-adrenergic 
receptor agonist, are generally limited, with the most compelling data coming from a population of civilian Cambodian refugees.^13,14 A 2007 article by Boehnlein and Kinzie includes a thorough review of the preclinical research, case reports, and early clinical studies that have led to the widespread use of these agents for PTSD despite the lack of FDA approval for this indication.^13A previous retrospective review by Byers and colleagues compared the effectiveness and tolerability of prazosin and quetiapine for nighttime PTSD symptoms in veterans.¹⁵ The results of that review suggest that α1-adrenergic agents may be equally effective and better tolerated than alternative medication options (ie, atypical antipsychotics) for this purpose. The present study was adapted from this design to report concurrently on the real-world use of clonidine and prazosin for the treatment of nighttime PTSD
symptoms.

Study Objectives

The primary objective of this retrospective chart review was to describe the experience of patients prescribed clonidine or prazosin for the treatment of nighttime PTSD symptoms, including initial effectiveness. The primary endpoint of initial drug effectiveness was documented improvement of nighttime PTSD symptoms in the patient’s chart within 6 months of the date of first prescription. Clonidine or prazosin was categorized as initially effective if a statement such as “frequency of nightmares decreased” or “patient’s nighttime PTSD symptoms have improved” was made within 6 months after initial prescription of the drug.

The secondary objectives of this study were to evaluate the long-term effectiveness and tolerability of prazosin. The endpoints used to assess these outcomes were the 2-year continuation rates of clonidine and prazosin (as a surrogate marker for long-term effectiveness) and the documented reasons for discontinuation of clonidine and prazosin for the treatment of nighttime PTSD symptoms (in order to assess tolerability).

Methods

An electronic database search was conducted to identify the VA Portland Health Care System (VAPHCS) patients with a diagnosis of PTSD who received a first prescription for clonidine or prazosin for nighttime PTSD symptoms from a VAPHCS mental health provider or primary care provider (PCP) from January 1, 2009, to December 31, 2011. Patients were excluded if they had any history of prior use of the drug being initiated, were co-initiated on both clonidine and prazosin (defined as starting the drugs within 30 days of each other), or had a concomitant diagnosis of schizophrenia, bipolar 
disorder, psychotic disorder, or cognitive disorder as defined in the 
Diagnostic and Statistical Manual of Mental Disorders, 5th Edition. Patients with traumatic brain injury (TBI) were excluded only if it could be determined that the event had resulted in lasting cognitive impairment.

Study Population

All patients with a diagnosis of PTSD who received a first prescription for clonidine during the period specified were screened for inclusion; patients with PTSD who were first prescribed prazosin during the same period were randomly sampled to equalize patient populations. This was done to maximize the data set while examining groups of roughly equal size for each drug, as prazosin is used much more commonly than clonidine for nighttime PTSD symptoms at VAPHCS. The patients in each resulting group were screened to determine whether they met inclusion and exclusion criteria. All subjects included were followed for 2 years from the date of the initial prescription.

Study Design

Initial effectiveness of each agent was determined by reviewing subjects’ progress notes after the initial prescription of clonidine or prazosin for documentation of improvement in symptoms within 6 months of the prescription start date. A decrease in frequency or intensity of nighttime PTSD symptoms, nightmares, or insomnia, as documented in the patient chart, was interpreted as improvement of symptoms.

Long-term continuation was assessed by reviewing subjects’ prescription records, to determine whether prescription(s) for clonidine or prazosin continued for 2 years after the date of the initial prescription.

Any gap between medication fills that resulted in an anticipated period without medication of ≥ 6 months (eg, 9 months after receiving a 
90-day supply) was considered discontinuation of therapy. Prescription refill history was also reviewed, and medication possession ratio (MPR) was calculated to assess whether patients were adherent to the study drug as prescribed. Adherence was defined as an MPR of ≥ 80%. Patients who left the VAPHCS service area but continued to receive care at another VA were assessed for continuation of therapy, but refill data and/or MPR were not assessed.

Tolerability was assessed by reviewing subjects’ medical records to determine whether therapy with clonidine or prazosin was discontinued due to documented adverse effects (AEs). The occurrence of AEs was determined by reviewing progress notes and other chart documentation surrounding the date of discontinuation. If the drug was discontinued but the reason was not explicitly documented or if the prescription expired without a documented reason for nonrenewing, the reason for discontinuation was coded as “not specified.” Discontinuation due to treatment failure, change in symptoms, nonadherence, or other causes was also recorded. If multiple reasons for discontinuation were cited for a single patient, all were included in the data. This project was approved by the institutional review board at the VAPHCS.

Related:Depression and Substance Abuse Intensify Suicide Risk

Statistical Considerations

Based on clinical experience, it was presumed that many of the patients who were prescribed clonidine would be receiving it as a second-line therapy after failing prazosin. Therefore, statistical analysis of the relative effectiveness and tolerability of clonidine and prazosin could not be performed. Neither power nor sample size needed to demonstrate any difference in effectiveness or tolerability between the groups was calculated. All results are expressed using descriptive statistics.

Results

An initial database search for patients with PTSD who received a first prescription for clonidine between January 1, 2009, and December 31, 2011, from a VAPHCS provider yielded a list of 149 patients. The same search criteria applied for prazosin yielded 1,116 patients, 149 of whom were randomly selected for screening. After screening, 42 patients on clonidine and 60 patients on prazosin were included in this analysis (Figure).

Patient Demographics

The average age of the clonidine patients was 38.5 years (range 21-65 years) (Table 1). The clonidine group was primarily male (90%) and white (83%). Eighteen of the 42 patients in the clonidine group had a
baseline PTSD Checklist-Civilian version (PCL-C) score available within the 90 days before the first prescription of clonidine; the average baseline PCL-C score in this subgroup was 
62 ± 12.0 (median 65.5, range 
31-82). Most of the clonidine patients (71%) had a concomitant diagnosis of a depressive disorder. About one-quarter of the group (24%) had previously tried prazosin per prescription records. In 24 patients (57%), the first prescription for clonidine was written by a psychiatrist or psychiatric nurse practitioner; 18 patients (43%) were started on clonidine by PCPs.

The average age of the prazosin patients was 46.1 years (range 21-74 years). The prazosin group was also primarily male (93%) and white (88%). Twenty of the 60 patients in the prazosin group had a baseline PCL-C score available within the 
90 days before the first prescription of prazosin; the average baseline PCL-C score in this subgroup was 55 ± 16.1 (median 64, range 30-72). Most of the prazosin patients (63%) had a concomitant diagnosis of a depressive disorder. Four patients (7%) had previously tried clonidine per prescription records. In 35 patients (58%), the first prescription for prazosin was written by a psychiatrist or psychiatric nurse practitioner; 
25 patients (42%) were started on prazosin by PCPs.

Data pertaining to initial and long-term effectiveness, tolerability, and MPR for both clonidine and prazosin are presented in Table 2.

Clonidine

Of the 42 clonidine patients assessed, 24 (57%) had a positive response to the medication for nighttime PTSD symptoms documented in the Computerized Patient Record System (CPRS) within 6 months of starting therapy. Six months after starting clonidine, 23 patients (55%) continued to take clonidine. Two years after starting therapy, 8 of the original 
42 patients continued on clonidine for an overall 2-year continuation rate of 19%.

Tolerability

Of the 34 patients who discontinued clonidine within 2 years, 13 patients (38%) cited ineffectiveness of therapy as a reason for discontinuation. Another 13 patients (38%) reported discontinuing therapy due to AEs. Sedation (4 patients, 12%), dizziness/hypotension (3 patients, 9%), and paradoxical worsening of PTSD symptoms (4 patients, 12%) were the most common AEs leading to discontinuation. Other AEs cited as reasons for discontinuation were syncope 
(2 patients), erectile dysfunction 
(1 patient), rash (1 patient), myoclonus (1 patient), increased depression (1 patient), and fatigue (1 patient). One patient reported that he had discontinued clonidine due to symptom resolution/lack of need for treatment. In 8 of the 34 patients, no reason for discontinuation was found in chart documentation.

Medication Possession Ratio

Among the 21 evaluable patients who continued to receive clonidine 6 months after initiation, 10 (48%) were determined to be highly adherent to therapy, with an MPR of ≥ 80%. Six of the 21 patients (29%) had an MPR between 50% and 79%, and 
5 patients (24%) had an MPR < 50%.

Of the 8 patients who continued on clonidine at the 2-year mark, 
3 (38%) were adherent to therapy, with an MPR of ≥ 80%. Three more patients (38%) had a 2-year MPR between 50% and 80%, and 2 patients (25%) had an MPR < 50%.

Prazosin

Of the 60 prazosin patients assessed, 32 (53%) had a positive response to the medication for nighttime PTSD symptoms documented in the CPRS within 6 months of starting therapy. Six months after starting prazosin, 36 patients (60%) continued to take prazosin. Two years after starting therapy, 18 of the original 60 patients continued on prazosin for an overall 2-year continuation rate of 30%.

Tolerability

Of the 42 patients who discontinued prazosin within 2 years, six patients (14%) cited ineffectiveness of therapy as a reason for discontinuation. Thirteen patients (31%) reported discontinuing therapy due to AEs. Sedation (3 patients, 7%), dizziness/hypotension (3 patients, 7%), and paradoxical worsening of PTSD symptoms (6 patients, 14%) were the most common AEs leading to discontinuation. Other AEs cited as reasons for discontinuation were headache 
(2 patients), altered mental status (1 patient), and fatigue (1 patient). Three patients reported that they had discontinued clonidine due 
to symptom resolution/lack of need for treatment. Other reasons for discontinuation not related to 
AEs included flight rules (1 patient), changes to antihypertensive regimen (1 patient), refill issues (1 patient), and cost (1 patient). In 15 of 
the 42 patients, no reason for 
discontinuation was found in chart documentation.

Medication Possession Ratio

Among the 31 evaluable patients who continued to receive prazosin 
6 months after initiation, 20 (65%) were determined to be highly adherent to therapy, with an MPR of ≥ 80%. Five of the 31 patients (16%) had an MPR between 50% and 80%, and 6 patients (19%) had an MPR < 50%.

Of the 15 evaluable patients who continued on prazosin at the 2-year mark, 9 (60%) were adherent to therapy, with an MPR of ≥ 80%. Three patients (20%) had a 2-year MPR 
between 50% and 80%, and 3 patients 
(20%) had an MPR < 50%.

Discussion

Although prazosin has been shown to be effective for nighttime PTSD symptoms in both prospective and retrospective evaluations in veterans, this study provides the first evidence to support the use of clonidine in a veteran population.^10-12,15

Interestingly, 42% of the patients assessed received their first prescription of an α2-adrenergic agent for nighttime PTSD symptoms from a PCP. Even with the recent increased focus on integrating mental health into primary care within the VA, this was a surprising finding. Primary care providers at VAPHCS may have a greater role in the outpatient management of PTSD than previously suspected. The information presented here may prove useful and applicable in both psychiatric and primary care treatment settings.

The study results indicated that a majority of subjects initially reported effectiveness with either clonidine or prazosin (53% and 57%, respectively). The initial effectiveness rate for prazosin is similar to those described in previous studies.^10-13,15 The data also support a viable role for clonidine in the treatment of nighttime PTSD symptoms.

Regardless of initial improvement, the study results also suggest that the therapeutic benefit may not persist in the long term, as evidenced by a significant percentage of discontinuations attributed to ineffectiveness (38% for clonidine and 14% for prazosin) and a very low rate of long-term continuation (19% for clonidine and 30% for prazosin at 2 years). This latter observation contrasts with findings from previous studies; Byers and colleagues reported a 2-year prazosin continuation rate of 48.4% in a similar analysis, and Boehnlein and colleagues reported a sustained benefit of clonidine in responders over a 10-year period.^14,15 The wide variety of reasons for discontinuation reported here may help providers who are considering clonidine or prazosin for their patients to anticipate barriers to long-term success.

Part of the discrepancy between these results and previously reported successes with clonidine and prazosin may be attributable to the classic issue of efficacy vs effectiveness. Many of the studies that have informed us on the efficacy and tolerability of prazosin for nighttime PTSD symptoms described outcomes of prospective clinical research. Furthermore, these prospective trials were limited to < 6 months in duration. To date, neither clonidine nor prazosin has been evaluated for long-term efficacy and effectiveness in well-designed, prospective trials. This retrospective analysis may help provide a realistic estimate of the long-term effectiveness of these therapies, especially within the veteran population.

Limitations

This was a single-center, retrospective study conducted primarily in white male patients. Although likely applicable to the U.S. veteran population at large, these data may be poorly generalizable to patient populations outside the VA health care system.

Aside from external validity, this study has several significant limitations. The primary limitation of this project is that it was not designed to allow for statistical comparison of clonidine and prazosin. Such an analysis would have better defined the role of clonidine in PTSD treatment, either by establishing similar effectiveness of clonidine and prazosin for nighttime symptoms or by providing evidence of the superiority of one over the other. In designing the project, investigators suspected based on experience that the majority of patients prescribed clonidine would receive the drug after having already failed first-line therapy with prazosin. Had this been the case, a direct comparison may have been biased in favor of prazosin. In retrospect, however, only 24% of the clonidine group had previously been prescribed prazosin, and only 7% of the prazosin group had been prescribed clonidine. This suggests that clonidine may be used first line more often than the investigators anticipated and that a future direct comparison would be worthwhile.

Second, the subjective data collected for this project required investigators to read and interpret chart notes, although the review of all records by a single investigator helped limit variability in interpretation. At times, information in the CPRS was incomplete in terms of determining continuation of therapy or cause for discontinuation.

Third, although it is implied that a significant number of veterans have combat-related PTSD, the nature of the traumatic event(s) leading to PTSD was not recorded in this study, and no subgroup analysis was done to compare the effect of α2-adrenergic agents between combat- and noncombat-related PTSD. Owing to their exclusion by design, it is also difficult to apply these results to veterans who have lasting cognitive impairment as a result of TBI, who are presumably among those most likely to have experienced traumas that could provoke PTSD.

The design of this project also did not include a subgroup analysis based on antidepressant type, and it is unclear whether the potential pharmacodynamic interaction 
between noradrenergic antidepressants (ie, SNRIs) and anti–
α2-adrenergic agents had any impact on clinical outcomes. The use of complementary nonpharmacologic treatment modalities (ie, psychotherapy, eye movement desensitization and reprocessing) was also not evaluated.

Related: Female Service Members in the Long War

Finally, the primary outcome of patient-reported improvement in symptoms does not provide information on the magnitude or specific nature of benefits derived. Given the retrospective nature, data used in prospectively designed studies (eg, rating scales pertinent to PTSD), which might have helped to quantify the benefit of treatment, was not consistently available. Even a baseline PCL-C score, collected in order to describe the patient population, was available only in 37% of the patients assessed. Furthermore, nighttime PTSD symptoms vary among individuals, but the primary outcome of this study pools any benefits seen in areas such as nightmares, awakenings, night sweats, or sleep quality into a single outcome of symptom improvement.

Conclusions

This study indicates that both clonidine and prazosin may be effective for the treatment of nighttime PTSD symptoms in the veteran population but that their long-term utility may be limited by waning effectiveness, tolerability, and adherence issues. At this time, it is unclear whether either agent has an advantage over the other in terms of effectiveness or tolerability; further studies are needed to address that question.

Despite its limitations, the authors anticipate that this study will provide information regarding the effectiveness and tolerability of clonidine and prazosin to treat nighttime PTSD symptoms. Findings from this study may help clinicians to anticipate the needs and challenges of patients using β2-adrenergic agents for nighttime symptoms of PTSD.

Acknowledgements
The authors wish to acknowledge Brian Wilcox, PharmD, for his assistance in generating patient data reports, and Ronald Brown, RPh, MS, for his guidance regarding data analysis.

Author disclosures
The authors report no actual or potential conflicts of interest with regard to this article.

Disclaimer
The opinions expressed herein are those of the authors and do not necessarily reflect those of Federal Practitioner, Frontline Medical Communications Inc., the U.S. Government, or any of its agencies. This article may discuss unlabeled or investigational use of certain drugs. Please review the complete prescribing information for specific drugs or drug combinations—including indications, contraindications, warnings, and adverse effects—before administering pharmacologic therapy to patients.

Clostridium difficile infection (CDI) is the result of a Gram-positive bacterium, whose exotoxins are commonly associated with infectious, watery diarrhea.¹Clostridium difficile infection is associated with a significant cost burden, and over the past several years, the incidence and severity of CDI have been on the rise.^2,3

There are several known risk factors for CDI. The most well-elucidated risk factor is the use of antibiotics, especially fluoroquinolones, clindamycin, broad-spectrum penicillins, and broad-spectrum cephalosporins.^4,5 Other risk factors include advancing age, immunosuppression, a high burden of comorbidities, hospitalization, and antineoplastic agent use.^6-8 Over the past decade, gastric acid 
suppression has come under increased scrutiny as a possible risk factor for CDI; specifically, exposure to proton pump inhibitors (PPIs) and histamine 2 receptor antagonists (H2RAs).^8-14 With the reported overuse of PPIs, the importance of understanding safety risks associated with these agents is becoming increasingly necessary.¹⁵

In 2012, the FDA issued a public safety announcement reporting a possible association between CDI and patients undergoing treatment with PPIs.¹⁶ A large meta-analysis by Janarthanan and colleagues in 2012 evaluated 23 studies with nearly 300,000 patients, showing a 1.6-fold increase in CDI in patients exposed to a PPI.⁸ Another large meta-analysis noted that 39 studies showed a statistically significant association between PPI use and the risk of developing CDI (odds ratio [OR] 1.74) compared with nonusers.¹⁷ A recent study by McDonald and colleagues demonstrated patients with continuous PPI use had an elevated risk of CDI recurrence compared with patients not on continuous PPI therapy.¹⁸ These large studies did not focus analysis on elderly, hospitalized patients with significant comorbidities. There are several proposed mechanisms for the reported association between PPI use and CDI. The most widely accepted mechanism is that gastric acid suppression disrupts normal gastrointestinal flora and allows for bacterial overgrowth.^19-21There are few studies that have evaluated the association between PPI use and CDI in elderly, hospitalized patients. Studies conducted in a similar patient population have demonstrated no association between PPI use and CDI.^22,23 Shah and colleagues reported that treatment with gastric acid antisecretory agents does not increase the risk of developing CDI among elderly, hospitalized patients who also had severe disability.²³ Lowe and colleagues demonstrated no association between PPI therapy and hospitalization for elderly outpatients with CDI.²² A study was needed to determine the association between PPI use and CDI in hospitalized, elderly patients with a high burden of comorbidities.

Related: Cleaning Up? Microfiber May Be Better

Objectives

The primary objective of this study was to determine whether there is an association between PPI exposure and CDI in elderly, hospitalized patients. The secondary objective was to determine the risk factors for the development of CDI in elderly, hospitalized patients.

Methods

Approval for this study was obtained from the Emory University Institutional Review Board and the VA Research and Development Committee. The study was a single-center, retrospective, medical record review of patients with a CDI polymerase chain reaction (PCR) assay, conducted at the Atlanta VAMC between August 20, 2011, and August 20, 2013.

Two reports for the study period were generated from TheraDoc (Premier Inc., Salt Lake City, UT) medical record software: all patients with a positive CDI PCR assay and all patients with a negative CDI PCR assay. All adult inpatients aged ≥ 18 years with a positive CDI PCR assay and diarrhea were included. Patients with CDI were randomly matched 1:1, based on age, with a control patient from a large sample of eligible CDI negative assays. Any duplicate positive CDI PCR assays were deleted, and only the first positive test was analyzed. Confirmation that PCR assay with liquid stool was being performed per manufacturer recommendations was obtained from microbiology laboratory staff.

Patient-specific data were collected from the VA Computerized Patient Record System (CPRS). Potential covariates for analyses were selected based on previous literature regarding possible associations between PPI and CDI. Data were collected on patient age, gender, PPI exposure, PPI agent, PPI dose, concomitant medications, high-risk antibiotic use, comorbidities (including diabetes, chronic renal failure, liver disease, anemia, coagulopathy, myocardial infarction, chronic heart failure, peripheral vascular disease, chronic obstructive pulmonary disease, hypertension, hypothyroidism, and any alcohol or drug abuse), length of hospital stay, bed location, and first vs recurrent CDI. Proton pump inhibitor exposure was defined as use of any PPI during hospitalization or within 2 months prior to hospitalization. High-risk antibiotics were defined as fluoroquinolones, broad-spectrum penicillins, broad-spectrum cephalosporins, and clindamycin.

Statistical Analysis

Two-sided Wilcoxon rank sum and chi-square tests were used to compare the selected variables between CDI cases and non-CDI controls. A multivariate logistic regression model was fitted to the data using CDI as the outcome and PPI use as the main exposure of interest. The large number of covariates of interest relative to the sample size suggests conditional maximum likelihood methods of estimation.²⁴

Separate models were run using each case-control pair as a separate stratum in the model (125 pairs) as well as pooling similar-age strata to reduce the 125 pairs to 46 pooled sets. However, when comparing the Akaike information criterion (AIC; an objective measure to determine relative quality of multivariate models where a lower AIC value is preferred) between these individual and pooled strata models, the model that controlled for 125 individual case-control strata was overwhelmingly suggested as the better model (AIC, 175 vs 255, respectively).²⁵ Analyses were conducted with SAS 9.2 (SAS Institute Inc., Cary, NC).

Results

A total of 128 patients were positive for CDI during the 2-year study period. Three of these patients were excluded from the study due to outpatient status. The remaining 
125 patients were matched 1:1 with patients negative for CDI to yield a total study population of 250 patients.

Baseline demographics are shown in Table 1. The majority of patients included were males with a median age of 66 years. Nearly half of all patients in both groups had chronic renal failure, diabetes, or anemia. Comorbidities were numerous but were not significantly different between the positive and negative CDI groups. No significant difference in immunosuppression or PPI use was detected between the 2 groups. However, there were significantly more patients taking a high-risk antibiotic or an antineoplastic agent in the positive CDI group compared with the negative CDI group. The average length of hospital stay averaged 10 to 12 days and did not statistically differ between the 2 groups.

Crude ORs (cORs) and adjusted ORs (aORs) were calculated for the primary and secondary outcome measures (Table 2). There was not a statistically significant association between PPI use and CDI (cOR 1.10, 95% confidence interval [CI] 0.67-1.82; aOR 1.19, 95% CI 0.66-2.15). Other known risk factors were also evaluated for association. A statistically significant association did not exist between CDI and immunosuppression, antidepressant use, statin use, diabetes, chronic renal failure, liver disease, or anemia. However, the statistical analysis did suggest an association between CDI and high-risk antibiotic use (aOR 2.20, 95% CI 1.22-3.99) and antineoplastic agent use (aOR 5.52, 95% CI 1.77-17.26).

A sensitivity analysis was conducted to determine whether there were differing associations with CDI by PPI dose or specific agent. In both sensitivity analyses, there were no statistically significant differences in CDI between patients who took once-daily vs twice-daily PPI dosing or those who took pantoprazole vs omeprazole.

Discussion

The objective of this study was to evaluate the association between PPI use and CDI in an aging, hospitalized population. When adjusted for known risk factors, there was no association between CDI and patients exposed to PPI therapy.

Previous studies evaluating PPI use and CDI have shown conflicting results. Large meta-analyses have shown an increase in CDI in patients exposed to a PPI, whereas other studies have shown no association. In the studies that did not link PPI use and CDI, patients were elderly, hospitalized, and had other CDI risk factors. The patients in this study were hospitalized, with a median age of 
66 years. They were significantly immunosuppressed and had a very high burden of comorbidities. A possible explanation for the lack of association between PPI use and CDI is that, in patients with several existing risk factors for CDI, adding a PPI confers no additional effect on CDI risk.

Well-known risk factors, including high-risk antibiotic use and antineoplastic chemotherapy use, were confirmed by this study. Other known risk factors, including immunosuppression and diabetes, were not observed to have an association with CDI in this study. This is perhaps for the same reason that PPI exposure did not show a significant association. In a study published in 2010, Howell and colleagues showed that the risk of CDI increased as acid suppression increased in a dose-dependent fashion.⁹ There was no association between PPI dose and PPI agent on the primary outcome measure.

About half of all patients in the current study were exposed to PPI therapy, which was a surprisingly high number. Although this study did not evaluate appropriate use of PPI therapy, it exposes the high rate of PPI use at the study site. It is known that PPI use has associated risks, and it is important that physicians continue to be vigilant in their prescribing habits.

Related: The Importance of an Antimicrobial Stewardship Program

Limitations and Future Directions

Several limitations of this study should be noted. A relatively narrow patient population was examined, which limits the generalizability of these findings. However, health care providers treating older, hospitalized patients with a high burden of comorbidities may find the results meaningful. This study was retrospective and included a relatively small sample size, which may limit the ability to detect a statistically significant difference.

Data were not collected on the duration of PPI therapy. A longer duration of therapy has been shown in previous studies to be significantly associated with CDI.²⁶ It is unclear in this patient population whether there would have been an association between PPI duration of treatment and CDI.

Outpatient PPI exposure was determined using CPRS refill history. Patients were considered to have PPI exposure if they filled ≥ 1 prescription for a PPI within 2 months of hospitalization. Using this methodology to determine PPI exposure may not have identified patients who took over-the-counter PPIs or did not report filling a prescription for a PPI from an outside pharmacy, which would have resulted in an underestimation of PPI use in this sample. Furthermore, it is difficult to determine adherence to a prescribed regimen for outpatients.

Pantoprazole and omeprazole are the formulary PPIs at the study site. Conducting research at an institution with a formulary prevents evaluation of other PPIs, including esomeprazole, rabeprazole, dexlansoprazole, and lansoprazole. This is not seen as a significant limitation, as there have not been significant differences in the PPI agent and CDI widely reported in the literature.

Data on H2RA exposure were not collected. Any possible effect of H2RA exposure and CDI cannot be accounted for in this study. It is not likely that H2RA exposure would be associated with an increased risk of CDI in this patient population, as several studies have shown less of an association between CDI and H2RA compared with CDI and PPI use.

Further investigation to evaluate the association between CDI and PPI exposure in an elderly, hospitalized population is needed. Larger studies 
in these patients that evaluate duration of PPI therapy would be beneficial.

Related: Antidepressants Plus NSAIDs and the Risk of Intracranial Hemorrhage

Conclusion

In an elderly, hospitalized patient population with a high comorbidity burden, this study did not detect a statistically significant association between PPI exposure and CDI. Despite this, providers should continue to consider discontinuation of unnecessary PPI therapy.

Acknowledgements
The authors wish to thank Mehran Salles, PhD, PharmD, for her assistance. Study findings were presented at the 2014 Southeastern Residency Conference in Athens, Georgia, on May 1, 2014.

Author disclosures
The authors report no actual or potential conflicts of interest with regard to this article.

Disclaimer
The opinions expressed herein are those of the authors and do not necessarily reflect those of Federal Practitioner Frontline Medical Communications Inc., the U.S. Government, or any of its agencies. This article may discuss unlabeled or investigational use of certain drugs. Please review complete prescribing information for specific drugs or drug combinations—including indications, contraindications, warnings, and adverse effects—before administering pharmacologic therapy to patients.

Clostridium difficile infection (CDI) is the result of a Gram-positive bacterium, whose exotoxins are commonly associated with infectious, watery diarrhea.¹Clostridium difficile infection is associated with a significant cost burden, and over the past several years, the incidence and severity of CDI have been on the rise.^2,3

There are several known risk factors for CDI. The most well-elucidated risk factor is the use of antibiotics, especially fluoroquinolones, clindamycin, broad-spectrum penicillins, and broad-spectrum cephalosporins.^4,5 Other risk factors include advancing age, immunosuppression, a high burden of comorbidities, hospitalization, and antineoplastic agent use.^6-8 Over the past decade, gastric acid 
suppression has come under increased scrutiny as a possible risk factor for CDI; specifically, exposure to proton pump inhibitors (PPIs) and histamine 2 receptor antagonists (H2RAs).^8-14 With the reported overuse of PPIs, the importance of understanding safety risks associated with these agents is becoming increasingly necessary.¹⁵

In 2012, the FDA issued a public safety announcement reporting a possible association between CDI and patients undergoing treatment with PPIs.¹⁶ A large meta-analysis by Janarthanan and colleagues in 2012 evaluated 23 studies with nearly 300,000 patients, showing a 1.6-fold increase in CDI in patients exposed to a PPI.⁸ Another large meta-analysis noted that 39 studies showed a statistically significant association between PPI use and the risk of developing CDI (odds ratio [OR] 1.74) compared with nonusers.¹⁷ A recent study by McDonald and colleagues demonstrated patients with continuous PPI use had an elevated risk of CDI recurrence compared with patients not on continuous PPI therapy.¹⁸ These large studies did not focus analysis on elderly, hospitalized patients with significant comorbidities. There are several proposed mechanisms for the reported association between PPI use and CDI. The most widely accepted mechanism is that gastric acid suppression disrupts normal gastrointestinal flora and allows for bacterial overgrowth.^19-21There are few studies that have evaluated the association between PPI use and CDI in elderly, hospitalized patients. Studies conducted in a similar patient population have demonstrated no association between PPI use and CDI.^22,23 Shah and colleagues reported that treatment with gastric acid antisecretory agents does not increase the risk of developing CDI among elderly, hospitalized patients who also had severe disability.²³ Lowe and colleagues demonstrated no association between PPI therapy and hospitalization for elderly outpatients with CDI.²² A study was needed to determine the association between PPI use and CDI in hospitalized, elderly patients with a high burden of comorbidities.

Related: Cleaning Up? Microfiber May Be Better

Objectives

The primary objective of this study was to determine whether there is an association between PPI exposure and CDI in elderly, hospitalized patients. The secondary objective was to determine the risk factors for the development of CDI in elderly, hospitalized patients.

Methods

Approval for this study was obtained from the Emory University Institutional Review Board and the VA Research and Development Committee. The study was a single-center, retrospective, medical record review of patients with a CDI polymerase chain reaction (PCR) assay, conducted at the Atlanta VAMC between August 20, 2011, and August 20, 2013.

Two reports for the study period were generated from TheraDoc (Premier Inc., Salt Lake City, UT) medical record software: all patients with a positive CDI PCR assay and all patients with a negative CDI PCR assay. All adult inpatients aged ≥ 18 years with a positive CDI PCR assay and diarrhea were included. Patients with CDI were randomly matched 1:1, based on age, with a control patient from a large sample of eligible CDI negative assays. Any duplicate positive CDI PCR assays were deleted, and only the first positive test was analyzed. Confirmation that PCR assay with liquid stool was being performed per manufacturer recommendations was obtained from microbiology laboratory staff.

Patient-specific data were collected from the VA Computerized Patient Record System (CPRS). Potential covariates for analyses were selected based on previous literature regarding possible associations between PPI and CDI. Data were collected on patient age, gender, PPI exposure, PPI agent, PPI dose, concomitant medications, high-risk antibiotic use, comorbidities (including diabetes, chronic renal failure, liver disease, anemia, coagulopathy, myocardial infarction, chronic heart failure, peripheral vascular disease, chronic obstructive pulmonary disease, hypertension, hypothyroidism, and any alcohol or drug abuse), length of hospital stay, bed location, and first vs recurrent CDI. Proton pump inhibitor exposure was defined as use of any PPI during hospitalization or within 2 months prior to hospitalization. High-risk antibiotics were defined as fluoroquinolones, broad-spectrum penicillins, broad-spectrum cephalosporins, and clindamycin.

Statistical Analysis

Two-sided Wilcoxon rank sum and chi-square tests were used to compare the selected variables between CDI cases and non-CDI controls. A multivariate logistic regression model was fitted to the data using CDI as the outcome and PPI use as the main exposure of interest. The large number of covariates of interest relative to the sample size suggests conditional maximum likelihood methods of estimation.²⁴

Separate models were run using each case-control pair as a separate stratum in the model (125 pairs) as well as pooling similar-age strata to reduce the 125 pairs to 46 pooled sets. However, when comparing the Akaike information criterion (AIC; an objective measure to determine relative quality of multivariate models where a lower AIC value is preferred) between these individual and pooled strata models, the model that controlled for 125 individual case-control strata was overwhelmingly suggested as the better model (AIC, 175 vs 255, respectively).²⁵ Analyses were conducted with SAS 9.2 (SAS Institute Inc., Cary, NC).

Results

A total of 128 patients were positive for CDI during the 2-year study period. Three of these patients were excluded from the study due to outpatient status. The remaining 
125 patients were matched 1:1 with patients negative for CDI to yield a total study population of 250 patients.

Baseline demographics are shown in Table 1. The majority of patients included were males with a median age of 66 years. Nearly half of all patients in both groups had chronic renal failure, diabetes, or anemia. Comorbidities were numerous but were not significantly different between the positive and negative CDI groups. No significant difference in immunosuppression or PPI use was detected between the 2 groups. However, there were significantly more patients taking a high-risk antibiotic or an antineoplastic agent in the positive CDI group compared with the negative CDI group. The average length of hospital stay averaged 10 to 12 days and did not statistically differ between the 2 groups.

Crude ORs (cORs) and adjusted ORs (aORs) were calculated for the primary and secondary outcome measures (Table 2). There was not a statistically significant association between PPI use and CDI (cOR 1.10, 95% confidence interval [CI] 0.67-1.82; aOR 1.19, 95% CI 0.66-2.15). Other known risk factors were also evaluated for association. A statistically significant association did not exist between CDI and immunosuppression, antidepressant use, statin use, diabetes, chronic renal failure, liver disease, or anemia. However, the statistical analysis did suggest an association between CDI and high-risk antibiotic use (aOR 2.20, 95% CI 1.22-3.99) and antineoplastic agent use (aOR 5.52, 95% CI 1.77-17.26).

A sensitivity analysis was conducted to determine whether there were differing associations with CDI by PPI dose or specific agent. In both sensitivity analyses, there were no statistically significant differences in CDI between patients who took once-daily vs twice-daily PPI dosing or those who took pantoprazole vs omeprazole.

Discussion

The objective of this study was to evaluate the association between PPI use and CDI in an aging, hospitalized population. When adjusted for known risk factors, there was no association between CDI and patients exposed to PPI therapy.

Previous studies evaluating PPI use and CDI have shown conflicting results. Large meta-analyses have shown an increase in CDI in patients exposed to a PPI, whereas other studies have shown no association. In the studies that did not link PPI use and CDI, patients were elderly, hospitalized, and had other CDI risk factors. The patients in this study were hospitalized, with a median age of 
66 years. They were significantly immunosuppressed and had a very high burden of comorbidities. A possible explanation for the lack of association between PPI use and CDI is that, in patients with several existing risk factors for CDI, adding a PPI confers no additional effect on CDI risk.

Well-known risk factors, including high-risk antibiotic use and antineoplastic chemotherapy use, were confirmed by this study. Other known risk factors, including immunosuppression and diabetes, were not observed to have an association with CDI in this study. This is perhaps for the same reason that PPI exposure did not show a significant association. In a study published in 2010, Howell and colleagues showed that the risk of CDI increased as acid suppression increased in a dose-dependent fashion.⁹ There was no association between PPI dose and PPI agent on the primary outcome measure.

About half of all patients in the current study were exposed to PPI therapy, which was a surprisingly high number. Although this study did not evaluate appropriate use of PPI therapy, it exposes the high rate of PPI use at the study site. It is known that PPI use has associated risks, and it is important that physicians continue to be vigilant in their prescribing habits.

Related: The Importance of an Antimicrobial Stewardship Program

Limitations and Future Directions

Several limitations of this study should be noted. A relatively narrow patient population was examined, which limits the generalizability of these findings. However, health care providers treating older, hospitalized patients with a high burden of comorbidities may find the results meaningful. This study was retrospective and included a relatively small sample size, which may limit the ability to detect a statistically significant difference.

Data were not collected on the duration of PPI therapy. A longer duration of therapy has been shown in previous studies to be significantly associated with CDI.²⁶ It is unclear in this patient population whether there would have been an association between PPI duration of treatment and CDI.

Outpatient PPI exposure was determined using CPRS refill history. Patients were considered to have PPI exposure if they filled ≥ 1 prescription for a PPI within 2 months of hospitalization. Using this methodology to determine PPI exposure may not have identified patients who took over-the-counter PPIs or did not report filling a prescription for a PPI from an outside pharmacy, which would have resulted in an underestimation of PPI use in this sample. Furthermore, it is difficult to determine adherence to a prescribed regimen for outpatients.

Pantoprazole and omeprazole are the formulary PPIs at the study site. Conducting research at an institution with a formulary prevents evaluation of other PPIs, including esomeprazole, rabeprazole, dexlansoprazole, and lansoprazole. This is not seen as a significant limitation, as there have not been significant differences in the PPI agent and CDI widely reported in the literature.

Data on H2RA exposure were not collected. Any possible effect of H2RA exposure and CDI cannot be accounted for in this study. It is not likely that H2RA exposure would be associated with an increased risk of CDI in this patient population, as several studies have shown less of an association between CDI and H2RA compared with CDI and PPI use.

Further investigation to evaluate the association between CDI and PPI exposure in an elderly, hospitalized population is needed. Larger studies 
in these patients that evaluate duration of PPI therapy would be beneficial.

Related: Antidepressants Plus NSAIDs and the Risk of Intracranial Hemorrhage

Conclusion

In an elderly, hospitalized patient population with a high comorbidity burden, this study did not detect a statistically significant association between PPI exposure and CDI. Despite this, providers should continue to consider discontinuation of unnecessary PPI therapy.

Acknowledgements
The authors wish to thank Mehran Salles, PhD, PharmD, for her assistance. Study findings were presented at the 2014 Southeastern Residency Conference in Athens, Georgia, on May 1, 2014.

Author disclosures
The authors report no actual or potential conflicts of interest with regard to this article.

Disclaimer
The opinions expressed herein are those of the authors and do not necessarily reflect those of Federal Practitioner Frontline Medical Communications Inc., the U.S. Government, or any of its agencies. This article may discuss unlabeled or investigational use of certain drugs. Please review complete prescribing information for specific drugs or drug combinations—including indications, contraindications, warnings, and adverse effects—before administering pharmacologic therapy to patients.

Clostridium difficile infection (CDI) is the result of a Gram-positive bacterium, whose exotoxins are commonly associated with infectious, watery diarrhea.¹Clostridium difficile infection is associated with a significant cost burden, and over the past several years, the incidence and severity of CDI have been on the rise.^2,3

There are several known risk factors for CDI. The most well-elucidated risk factor is the use of antibiotics, especially fluoroquinolones, clindamycin, broad-spectrum penicillins, and broad-spectrum cephalosporins.^4,5 Other risk factors include advancing age, immunosuppression, a high burden of comorbidities, hospitalization, and antineoplastic agent use.^6-8 Over the past decade, gastric acid 
suppression has come under increased scrutiny as a possible risk factor for CDI; specifically, exposure to proton pump inhibitors (PPIs) and histamine 2 receptor antagonists (H2RAs).^8-14 With the reported overuse of PPIs, the importance of understanding safety risks associated with these agents is becoming increasingly necessary.¹⁵

In 2012, the FDA issued a public safety announcement reporting a possible association between CDI and patients undergoing treatment with PPIs.¹⁶ A large meta-analysis by Janarthanan and colleagues in 2012 evaluated 23 studies with nearly 300,000 patients, showing a 1.6-fold increase in CDI in patients exposed to a PPI.⁸ Another large meta-analysis noted that 39 studies showed a statistically significant association between PPI use and the risk of developing CDI (odds ratio [OR] 1.74) compared with nonusers.¹⁷ A recent study by McDonald and colleagues demonstrated patients with continuous PPI use had an elevated risk of CDI recurrence compared with patients not on continuous PPI therapy.¹⁸ These large studies did not focus analysis on elderly, hospitalized patients with significant comorbidities. There are several proposed mechanisms for the reported association between PPI use and CDI. The most widely accepted mechanism is that gastric acid suppression disrupts normal gastrointestinal flora and allows for bacterial overgrowth.^19-21There are few studies that have evaluated the association between PPI use and CDI in elderly, hospitalized patients. Studies conducted in a similar patient population have demonstrated no association between PPI use and CDI.^22,23 Shah and colleagues reported that treatment with gastric acid antisecretory agents does not increase the risk of developing CDI among elderly, hospitalized patients who also had severe disability.²³ Lowe and colleagues demonstrated no association between PPI therapy and hospitalization for elderly outpatients with CDI.²² A study was needed to determine the association between PPI use and CDI in hospitalized, elderly patients with a high burden of comorbidities.

Related: Cleaning Up? Microfiber May Be Better

Objectives

The primary objective of this study was to determine whether there is an association between PPI exposure and CDI in elderly, hospitalized patients. The secondary objective was to determine the risk factors for the development of CDI in elderly, hospitalized patients.

Methods

Approval for this study was obtained from the Emory University Institutional Review Board and the VA Research and Development Committee. The study was a single-center, retrospective, medical record review of patients with a CDI polymerase chain reaction (PCR) assay, conducted at the Atlanta VAMC between August 20, 2011, and August 20, 2013.

Two reports for the study period were generated from TheraDoc (Premier Inc., Salt Lake City, UT) medical record software: all patients with a positive CDI PCR assay and all patients with a negative CDI PCR assay. All adult inpatients aged ≥ 18 years with a positive CDI PCR assay and diarrhea were included. Patients with CDI were randomly matched 1:1, based on age, with a control patient from a large sample of eligible CDI negative assays. Any duplicate positive CDI PCR assays were deleted, and only the first positive test was analyzed. Confirmation that PCR assay with liquid stool was being performed per manufacturer recommendations was obtained from microbiology laboratory staff.

Patient-specific data were collected from the VA Computerized Patient Record System (CPRS). Potential covariates for analyses were selected based on previous literature regarding possible associations between PPI and CDI. Data were collected on patient age, gender, PPI exposure, PPI agent, PPI dose, concomitant medications, high-risk antibiotic use, comorbidities (including diabetes, chronic renal failure, liver disease, anemia, coagulopathy, myocardial infarction, chronic heart failure, peripheral vascular disease, chronic obstructive pulmonary disease, hypertension, hypothyroidism, and any alcohol or drug abuse), length of hospital stay, bed location, and first vs recurrent CDI. Proton pump inhibitor exposure was defined as use of any PPI during hospitalization or within 2 months prior to hospitalization. High-risk antibiotics were defined as fluoroquinolones, broad-spectrum penicillins, broad-spectrum cephalosporins, and clindamycin.

Statistical Analysis

Two-sided Wilcoxon rank sum and chi-square tests were used to compare the selected variables between CDI cases and non-CDI controls. A multivariate logistic regression model was fitted to the data using CDI as the outcome and PPI use as the main exposure of interest. The large number of covariates of interest relative to the sample size suggests conditional maximum likelihood methods of estimation.²⁴

Separate models were run using each case-control pair as a separate stratum in the model (125 pairs) as well as pooling similar-age strata to reduce the 125 pairs to 46 pooled sets. However, when comparing the Akaike information criterion (AIC; an objective measure to determine relative quality of multivariate models where a lower AIC value is preferred) between these individual and pooled strata models, the model that controlled for 125 individual case-control strata was overwhelmingly suggested as the better model (AIC, 175 vs 255, respectively).²⁵ Analyses were conducted with SAS 9.2 (SAS Institute Inc., Cary, NC).

Results

A total of 128 patients were positive for CDI during the 2-year study period. Three of these patients were excluded from the study due to outpatient status. The remaining 
125 patients were matched 1:1 with patients negative for CDI to yield a total study population of 250 patients.

Baseline demographics are shown in Table 1. The majority of patients included were males with a median age of 66 years. Nearly half of all patients in both groups had chronic renal failure, diabetes, or anemia. Comorbidities were numerous but were not significantly different between the positive and negative CDI groups. No significant difference in immunosuppression or PPI use was detected between the 2 groups. However, there were significantly more patients taking a high-risk antibiotic or an antineoplastic agent in the positive CDI group compared with the negative CDI group. The average length of hospital stay averaged 10 to 12 days and did not statistically differ between the 2 groups.

Crude ORs (cORs) and adjusted ORs (aORs) were calculated for the primary and secondary outcome measures (Table 2). There was not a statistically significant association between PPI use and CDI (cOR 1.10, 95% confidence interval [CI] 0.67-1.82; aOR 1.19, 95% CI 0.66-2.15). Other known risk factors were also evaluated for association. A statistically significant association did not exist between CDI and immunosuppression, antidepressant use, statin use, diabetes, chronic renal failure, liver disease, or anemia. However, the statistical analysis did suggest an association between CDI and high-risk antibiotic use (aOR 2.20, 95% CI 1.22-3.99) and antineoplastic agent use (aOR 5.52, 95% CI 1.77-17.26).

A sensitivity analysis was conducted to determine whether there were differing associations with CDI by PPI dose or specific agent. In both sensitivity analyses, there were no statistically significant differences in CDI between patients who took once-daily vs twice-daily PPI dosing or those who took pantoprazole vs omeprazole.

Discussion

The objective of this study was to evaluate the association between PPI use and CDI in an aging, hospitalized population. When adjusted for known risk factors, there was no association between CDI and patients exposed to PPI therapy.

Previous studies evaluating PPI use and CDI have shown conflicting results. Large meta-analyses have shown an increase in CDI in patients exposed to a PPI, whereas other studies have shown no association. In the studies that did not link PPI use and CDI, patients were elderly, hospitalized, and had other CDI risk factors. The patients in this study were hospitalized, with a median age of 
66 years. They were significantly immunosuppressed and had a very high burden of comorbidities. A possible explanation for the lack of association between PPI use and CDI is that, in patients with several existing risk factors for CDI, adding a PPI confers no additional effect on CDI risk.

Well-known risk factors, including high-risk antibiotic use and antineoplastic chemotherapy use, were confirmed by this study. Other known risk factors, including immunosuppression and diabetes, were not observed to have an association with CDI in this study. This is perhaps for the same reason that PPI exposure did not show a significant association. In a study published in 2010, Howell and colleagues showed that the risk of CDI increased as acid suppression increased in a dose-dependent fashion.⁹ There was no association between PPI dose and PPI agent on the primary outcome measure.

About half of all patients in the current study were exposed to PPI therapy, which was a surprisingly high number. Although this study did not evaluate appropriate use of PPI therapy, it exposes the high rate of PPI use at the study site. It is known that PPI use has associated risks, and it is important that physicians continue to be vigilant in their prescribing habits.

Related: The Importance of an Antimicrobial Stewardship Program

Limitations and Future Directions

Several limitations of this study should be noted. A relatively narrow patient population was examined, which limits the generalizability of these findings. However, health care providers treating older, hospitalized patients with a high burden of comorbidities may find the results meaningful. This study was retrospective and included a relatively small sample size, which may limit the ability to detect a statistically significant difference.

Data were not collected on the duration of PPI therapy. A longer duration of therapy has been shown in previous studies to be significantly associated with CDI.²⁶ It is unclear in this patient population whether there would have been an association between PPI duration of treatment and CDI.

Outpatient PPI exposure was determined using CPRS refill history. Patients were considered to have PPI exposure if they filled ≥ 1 prescription for a PPI within 2 months of hospitalization. Using this methodology to determine PPI exposure may not have identified patients who took over-the-counter PPIs or did not report filling a prescription for a PPI from an outside pharmacy, which would have resulted in an underestimation of PPI use in this sample. Furthermore, it is difficult to determine adherence to a prescribed regimen for outpatients.

Pantoprazole and omeprazole are the formulary PPIs at the study site. Conducting research at an institution with a formulary prevents evaluation of other PPIs, including esomeprazole, rabeprazole, dexlansoprazole, and lansoprazole. This is not seen as a significant limitation, as there have not been significant differences in the PPI agent and CDI widely reported in the literature.

Data on H2RA exposure were not collected. Any possible effect of H2RA exposure and CDI cannot be accounted for in this study. It is not likely that H2RA exposure would be associated with an increased risk of CDI in this patient population, as several studies have shown less of an association between CDI and H2RA compared with CDI and PPI use.

Further investigation to evaluate the association between CDI and PPI exposure in an elderly, hospitalized population is needed. Larger studies 
in these patients that evaluate duration of PPI therapy would be beneficial.

Related: Antidepressants Plus NSAIDs and the Risk of Intracranial Hemorrhage

Conclusion

In an elderly, hospitalized patient population with a high comorbidity burden, this study did not detect a statistically significant association between PPI exposure and CDI. Despite this, providers should continue to consider discontinuation of unnecessary PPI therapy.

Acknowledgements
The authors wish to thank Mehran Salles, PhD, PharmD, for her assistance. Study findings were presented at the 2014 Southeastern Residency Conference in Athens, Georgia, on May 1, 2014.

Author disclosures
The authors report no actual or potential conflicts of interest with regard to this article.

Disclaimer
The opinions expressed herein are those of the authors and do not necessarily reflect those of Federal Practitioner Frontline Medical Communications Inc., the U.S. Government, or any of its agencies. This article may discuss unlabeled or investigational use of certain drugs. Please review complete prescribing information for specific drugs or drug combinations—including indications, contraindications, warnings, and adverse effects—before administering pharmacologic therapy to patients.

In 2015, more than 35,660 new cases of liver cancer and 24,550 liver cancer-related deaths are expected to occur in the U.S. About 80% of these cases will consist of hepatocellular carcinoma, (HCC).¹ The incidence of HCC varies throughout the world: Incidence is as low as 5 in 100,000 individuals in North America and ranges up to > 20 in 100,000 individuals in sub-Saharan Africa and Eastern Asia.² Nearly half of all cases of HCC are associated with hepatitis B virus (HBV), and another 25% are associated with hepatitis C virus (HCV). Other risk factors for developing HCC include alcoholic liver disease, nonalcoholic steatohepatitis, flatoxin-contaminated food, diabetes, and obesity.³

Therapeutic options for advanced HCC are limited. The FDA approved sorafenib in 2008 for the treatment of unresectable HCC.⁴ According to the American Association for the Study of Liver Diseases (AASLD) and the Barcelona Clinic Liver Cancer (BCLC) staging system, patients with Stage C liver cancer may undergo a trial of sorafenib.⁴ National Comprehensive Cancer Network (NCCN) clinical guidelines for hepatobiliary cancers reserve sorafenib for patients with inoperable tumors, metastatic disease, or extensive liver tumor burden.⁵ Sorafenib is shown to inhibit multiple intracellular and cell surface kinases. Several of these kinases are thought to be involved in tumor cell signaling, angiogenesis, and apoptosis.⁴ In the Sorafenib HCC Assessment Randomized Protocol (SHARP) trial, median overall survival (OS) was 10.7 months in the sorafenib group and 7.9 months in the placebo group.⁶ The predicted survival rates at 1 year were 44% in the sorafenib group and 33% in the placebo group.⁶The economic impact of oral chemotherapy on health care cannot be discounted. At about $50,000 to $100,000 per quality- 
adjusted life-year, the incremental cost-effectiveness ratio (ICER) of sorafenib over placebo was $62,473 per quality-adjusted life-year in 2007.⁷The purpose of this retrospective chart review was to evaluate sorafenib for efficacy and safety in a veteran population. Veterans have poorer health and more medical conditions compared with nonveterans.⁸ Furthermore, in the VHA, about 170,000 veterans have HCV.⁹ The rate of progression from HCV to HCC is about 3% to 5% annually. More than half of those diagnosed with HCC are late stage, and unfortunately, the 5-year OS rate for patients with liver cancer is 9% and 4% for those patients who are diagnosed at regional and distant stages of the disease.¹ As the practice of oncology grows, it is necessary for pharmacists to be involved in the selection of chemotherapeutic agents in order to provide optimal pharmaceutical care.¹⁰

Related: VIDEO: NAFLD increasingly causing U.S. hepatocellular carcinoma

Methods

A retrospective chart review was conducted to identify patients who were prescribed sorafenib from November 1, 2007, to September 30, 2011, at the VA Greater Los Angeles Healthcare System (VAGLAHS). Inclusion criteria included patients who had a diagnosis of advanced HCC, who were initiated and managed by a VAGLAHS provider and who were eligible for a 1-year safety evaluation period. The study was approved by the VAGLAHS institutional review board.

Baseline demographic, clinical, laboratory, and medication data were collected. Demographic, clinical, laboratory, and medication data were obtained from CPRS (Computerized Patient Record System) and VistA (Veterans Health Information Systems and Technology Architecture). Data were collected on secured servers and saved on encrypted files. The master list was destroyed once the records control schedule was finalized. No identifiers were collected on the data collection sheet.

Standard practice at VAGLAHS is to monitor European Cooperative Oncology Group Performance Status (ECOG-PS), Child-Pugh class, and alpha-fetoprotein (AFP) at initiation and every 3 months and to obtain laboratory data at initiation and every month before each medication refill. Patients were seen in the Oncology Clinic periodically at the provider’s discretion. The time of drug discontinuation and the reason for drug discontinuation were recorded. Time of death at any point was collected to measure OS.

It was determined that a total sample size of 42 patients would be insufficient to achieve 80% power to demonstrate any hypothesized effects. However, the Fisher exact test was used to calculate P values for simple comparison. Patient demographics and clinical characteristics were reported as total numbers and frequencies when applicable. Survival rate was measured from the time of sorafenib initiation to 1 year after therapy initiation. Overall survival was measured from the time of sorafenib initiation to time of death. Duration of therapy was measured from the time of sorafenib initiation to time of discontinuation, either by 
provider or by patient.

Results

There were 83 patients who were prescribed sorafenib between November 1, 2007, and September 30, 2011. Of the 83 patients, 27 patients were ineligible for a 1-year follow-up period, 9 patients were diagnosed with non-HCC, 3 were initiated or managed by providers outside the institution, and 2 were not started on therapy. In all, 42 patients met inclusion criteria and had received at least 1 dose of sorafenib. The primary etiologies for HCC were history of alcohol abuse, HCV, and HBV. The primary risk factors were obesity, smoking, and diabetes. Many patients presented with multiple etiologies and risk factors. Ten patients (23.8%) had moderate-to-severe hepatic impairment (Child-Pugh class B or C). Baseline characteristics of these patients are listed in Table 1.

Efficacy

The median OS was 5.9 months and ranged from 21 days to 60 months. There were 17 patients who survived at the 1-year follow-up, including 1 patient who survived 363 days after treatment initiation, yielding an OS rate of 40.5%. Table 2 presents 1-year survival rates with respect to select baseline data. Baseline factors found to be negligible were age, smoking, alcohol abuse, obesity, presence of HCV, medication possession ratio (MPR), prior treatment, macrovascular invasion, and AFP. Neither initial dose regimen, final dose regimen achieved, or average dose correlated with the survival rate at the 1-year follow-up.

Factors possibly associated with a higher probability of survival were baseline ECOG-PS score and baseline Child-Pugh class (Table 2). Patients with an ECOG-PS score of 0 or 1 had a higher survival rate at 1 year than did patients with an ECOG-PS score of ≥ 2 (50% vs 0%, respectively; P = .113). Patients with Child-Pugh class B or C had a lower survival rate at 1 year than did patients with Child-Pugh class A (51% vs 10%, respectively; 
P =  .028). Other indicators were size of largest hepatic lesion ≤ 5 cm, total bilirubin ≤ 2 mg/dL, concurrent treatment, almost exclusively embolization, and treatment after sorafenib discontinuation, such as another oral chemotherapeutic agent or embolization.

The 17 patients who survived at 1 year were reviewed to see if they shared characteristics that indicated a higher probability of survival. The figure shows the baseline ECOG-PS score and the Child-Pugh class the patients who did and did not survive at the 1-year follow-up. In the first group, all patient possessed an ECOG-PS score of 0 or 1, and only 1 patient presented with Child-Pugh class B or C. In contrast, in the group who did not survive at the 1-year follow-up, there were 4 patients with ECOG-PS scores of > 1 and 9 patients who presented with Child-Pugh class B or C. The mean AFP level of this group was < 200 µg/mL, and only 4 patients were followed by Palliative Care Services. The average normalized MPR of this group was 71.9% compared with 85.3% for those who did not survive at the 1-year follow-up.

In patients who experienced at least 1 adverse event (AE), 16 survived, whereas only 1 who did not experience an AE survived (45.7% vs 14.3%, respectively; P = .210). Thirteen patients who experienced ≥ 3 AEs survived at 1 year; and only 3 patients who reported < 3 AEs survived at 
1 year (61.9% vs 14.0%, respectively; 
P = .011). However, when the number of AEs was normalized to duration of treatment per patient, the median frequency of AEs for all patients was 0.61 AEs per month treated. The difference in survival rates grew smaller and less significant between patients who had a frequency of AEs lower than the median compared with those with a higher ratio (52.4% vs 28.6%, respectively; P = .208). Patients affected by AEs in the first 30 days and 90 days of treatment had a survival rate at the 1-year follow-up of 42.4% and 30.2%, respectively. Patients who experienced dermatologic AEs had a higher survival rate than those who did not have dermatologic AEs (60.0% vs 29.6%, respectively; P = .099). This correlation was not found with 2 other classes of AEs, gastrointestinal (50.0% vs 27.8%; 
P = .208) or neurologic (64.0% vs 41.2%; P = .209).

The median overall time to discontinuation was 3.4 months. The main reasons cited for discontinuing sorafenib at 1 year included symptomatic progression (52.4%), radiographic progression (23.8%), severe AEs (16.7%), and mild-to-moderate AEs (11.9%). There was overlap, as 15 patients discontinued treatment for multiple reasons. For the 22 patients who discontinued medication due to symptomatic progression at 1 year, the median time to discontinuation was 3.8 months. For the 10 patients who discontinued medication due to radiographic progression at 1 year, median time to discontinuation was 5.6 months. Seven patients (16.7%) were still on therapy at 1 year.The study considered the impact of potential dose adjustments on survival rate and safety. The authors compared patients’ prescribed dose with the recommended dose based on the package insert and monthly laboratory values if recorded. The prescribed dose was recorded as appropriate dose, below dose, above dose, or indeterminate due to the lack of current laboratory values. Patients who survived at the 1-year follow-up had a composition of 26%, 21%, 10%, and 43%, respectively. These results were similar to those of patients who did not survive at the 1-year follow-up, 29%, 12%, 30%, and 29%, respectively.

Based on medication refill history and VA acquisition cost, the total prescription drug cost of treating 42 patients with sorafenib was $388,370.40. The total number of days survived for these patients was 16,607 days, which equates to $8,535.87 per year lived.

Safety

Of the 42 patients, 35 patients experienced ≥ 1 AE for a total of 122 AEs reported. The median number of AEs per patient was 2.5. The median time to the first AE was 21 days and ranged from 3 to 244 days. In the first 30 days of treatment, 23 patients (54.7%) reported 47 AEs (39.5%). In the first 90 days of treatment, 33 patients (78.6%) reported 88 AEs (73.9%). Common AEs in both instances were diarrhea, fatigue, erythematous plantar-palmar rash, and nausea (Table 3).

The predominant classes of AEs were GI (39.3%), dermatologic (18.9%), and neurologic (15.6%). Erythematous palmar-plantar rash, also known as hand-foot syndrome, has been noted as a potential dose-limiting sorafenib AE if the rash is recurrent or severe. One patient experienced recurrent grade-2 rashes, and sorafenib was immediately discontinued after an attempt to lower the dose. There were 8 patients who reported serious AEs, and 5 were hospitalized. One patient continued therapy despite GI hemorrhage. The other 4 patients discontinued therapy on hospitalization and were seen for intracranial hemorrhage, GI perforation, acute renal failure, and acute liver failure. In the first 3 cases, sorafenib could not be ruled out as the primary cause of death. None of these patients presented with comorbidities, such as hypertension, which predisposed them to AEs.

Overall, 38 patients ended therapy at the recommended regimen of 400 mg twice daily, and the average total daily dose was 619 mg, just below 80% of the recommended daily dose. Reasons for not achieving 400 mg twice daily included slow titration, AEs, and dose adjustments for compromised renal and hepatic function such as dialysis. Patients who had an ECOG-PS score of 
0 or 1 or Child-Pugh class A reported ≥ 3 AEs, but when normalized to duration of treatment, no difference was observed. No correlations were found for average dose, creatinine clearance, aspartate aminotransferase, platelets, total bilirubin, or weight and number or frequency of AEs.

In regard to potential dose adjustments, the doses (400 mg twice daily, 600 mg daily [400 mg + 200 mg in 2 doses], 200 mg twice daily, and 200 mg daily) did not correlate well with AEs. Patients who had < 3 AEs presented with the breakdown 23%, 16%, 22%, and 38%, similar to patients who had ≥ 3 AEs—30%, 19%, 14%, and 37%. Likewise, patients who had a frequency of AEs lower than the median presented with the breakdown 22%, 22%, 15%, and 40% compared with patients who had more AEs than the median—37%, 9%, 23%, and 31%.

Related: Hepatocellular Carcinoma: To Biopsy or Not?

Discussion

Sorafenib is the only oral oncology medication approved by the FDA for treatment of unresectable HCC.³ Prior to sorafenib, the AASLD recommendation was supportive care for patients presenting with BCLC-Stage C liver cancer. However, guidelines changed when SHARP showed that sorafenib provided a survival benefit with a tolerable AE profile. The survival benefit of sorafenib has been replicated in a few large, multicenter trials. In Asia, Cheng and colleagues saw improved median OS of 6.5 months for sorafenib compared with 4.2 months with placebo, and in Italy, Iavarone and colleagues showed a median OS of 10.5 months without a placebo comparator.^11,12

In the veteran population for this study, the OS rate of 40.5% was similar to the rate reported in the SHARP study, although the patients’ median OS fell short of the time described in SHARP and other trials. The medical complexities involved in treating veterans may explain this difference. The veteran population is heterogeneous with diverse ethnic backgrounds, several comorbidities, and varying degrees of organ dysfunction. The authors compared survival rates of different subgroups to test the hypothesis that the probability of survival while on therapy should not depend on demographics or medical history. However, in this study, patients with minimal impact from HCC, such as mild hepatic 
impairment and high-functional status, demonstrated higher survival rates at 1-year follow-up than did those without significant compromise.Although the high prevalence of HCV and alcohol abuse in the veteran population has resulted in a high incidence of hepatic dysfunction, this study suggests that these factors are independent of survival if liver function or integrity has not been compromised.⁹

Some researchers have hypothesized that clinical toxicities from tyrosine kinase inhibitors may correlate with survival.¹³ The authors noticed that the presentation of dermatologic AEs may reflect improved survival. In this study, patients who experienced ≥ 1 AE and ≥ 3 AEs had survival rates at the 1-year follow-up of 45.7% and 61.9%, respectively. Moreover, patients affected by AEs in the first 90 days of treatment had a survival rate at the 1-year follow-up of 42.4%.

Caution is advised when drawing conclusions from the number of AEs or when they appear, because this may falsely favor correlation. Patients who survive longer have additional time to report an AE. Therefore, the authors also looked at the ratio of AEs over time per patient to consider the number of AEs per duration of treatment and saw that there was little difference in survival rate in this regard. When considering patients affected by AEs only in the first 30 days of treatment, the survival rate at the 1-year follow-up fell to 30.2%.

A more likely factor for the survival of the 17 patients who were alive at the 1-year follow-up was their overall health relative to the rest of the study group. Overall health may indicate survival independent of sorafenib. The group of 17 who survived at the 1-year follow-up reflected a population that was different from the rest of the study population. The subset was generally healthier with better ECOG-PS scores and Child-Pugh classes, was not followed by Palliative Care Services, and had a mean AFP level under the threshold for diagnosis of HCC in patients who present with hepatic lesions and elevated AFP.¹⁴ This subset’s MPR, a surrogate marker for adherence, was less than the accepted threshold in clinical practice for oral medications.¹⁵Evaluating the patient’s dose regimen was expected to reveal a relationship between dosing and clinical outcomes, such as low survival rates with low doses or more AEs with high doses. However, the authors were not able to establish this link. In fact, the median time to discontinuation of 3.⁴ months for the study group, or duration of treatment, was much shorter than the median OS of 5.9 months.

These findings were consistent with Cabibbo and colleagues, who conducted a meta-analysis of survival rates for untreated patients and found that impaired performance status and Child-Pugh class B or C were independently associated with shorter survival.¹⁶ The SHARP study and Cheng and colleagues also attempted to exclude patients who were not Child-Pugh class A in their studies, which suggests a negligible correlation between sorafenib and survival time and a close relationship between baseline clinical status and survival.

The authors determined that prior treatment, including locoregional therapy, was not a factor in predicting survival. This observation is confirmed by the results of a phase 3 study that looked at sorafenib as adjuvant treatment for patients who had no detectable disease after surgical resection or local ablation.¹⁷ The trial did not meet its primary endpoint of improved recurrence-free survival. However, the authors observed in this study that 4 patients who underwent resection of the liver before sorafenib had a mean OS of 2.9 years. One patient, who was alive at the time of the study conclusion, received only 
22 days of sorafenib treatment and survived for 4.9 years after sorafenib discontinuation. Patients who received concurrent or postsorafenib treatment had higher survival rates.

The cost of treatment in this study was found to be $8,535.87 per year lived. Although formal quality of life assessments were not captured, medication was discontinued at the first sign of disease progression or AE as determined by the provider or patient. When the cost of treatment was adjusted to account for median OS time and VA drug acquisition costs, estimated at average wholesale price minus 40%, the cost of treatment was within the threshold of $50,000-$100,000 per quality-adjusted life-year.^7,18Of the 42 patients in this study, 28.6% discontinued therapy due to AEs, compared with 32% observed in the SHARP study. Common GI, dermatologic, and CNS AEs were comparable between the 2 studies. Serious AEs included intracranial hemorrhage, GI hemorrhage, GI perforation, acute liver failure, and acute renal failure; 3 of these events led to death. About 12% of patients experienced bleeding, regardless of severity, compared with the 18% seen in SHARP, despite no prior history of hemorrhage or GI perforation.⁵ The authors did not find any clinical factors at baseline that predisposed patients to AEs. It was also difficult to distinguish between drug-related AEs and general disease progression.

Although the authors did not find a relationship between dose or dose adjustments and the number or frequency of AEs, there were serious adverse outcomes in this study that were also rare complications observed in SHARP. The decision to start sorafenib should not be taken lightly.

Related: Diagnostic Dilemma of Hepatocellular Carcinoma Presenting as Hepatic Angiomyolipoma

Limitations

This retrospective review had several limitations. In SHARP and other large, multicenter trials, patients were continued on therapy until they experienced both symptomatic and radiographic progression. In this study, patients were discontinued at the first sign of progression, either symptomatic or radiographic or both. Had all patients remained on therapy until symptomatic and radiographic signs of progression were observed, there could have been a better correlation between duration of treatment and OS, symptomatic progression, or radiographic progression. The authors acknowledge, however, that there is diminishing benefit of administering chemotherapy when there are known and potentially serious AEs.

The data for this study were limited due to a small sample size, and it was not powered to evaluate for statistically significant characteristics between the patients who survived at the 1-year follow-up and the patients who did not survive at the 1-year follow-up. This information would be useful to identify potential prognostic factors and guide providers in sorafenib management. Finally, a long-term safety profile could not be established, as patients were evaluated for a 1-year period.

Ultimately, HCC is a multifactorial disease, and it is difficult to account for all potential confounding factors. Additional research, including studies comparing sunitinib or a control group to sorafenib, may provide further insight.

Conclusions

In light of these results, the authors believe that sorafenib may be considered for veterans with unresectable HCC and who are contraindicated for alternative treatments. One-year survival rates were similar to those seen in previous studies. However, there was no clear association between the duration of treatment and OS, and although the medication was well tolerated, there were also serious AEs. It is prudent to continually assess the need for therapy throughout the treatment period.

Pharmacists have a critical role in care for oncology patients, from the integration of certified clinical pharmacist practitioners into hematology-oncology clinics to patient monitoring through oral oncology pharmacy programs.^19,20 These 
programs have been shown to improve patient outcomes and decrease overall health care use and may benefit the veteran population.

In this study, a veteran population achieved a survival rate at the 1-year follow-up similar to that found in SHARP: 40.5% vs 44%. However, OS was markedly shorter: 5.9 months vs 10.7 months. Patients with minimal impact from HCC, such as mild hepatic impairment and high functional status, demonstrated higher survival rates at the 1-year follow-up than did those with significant compromise. Thirty-five patients experienced ≥1 AE, most observed within the first 90 days of treatment, and for 3 patients, sorafenib could not be ruled out as the cause of death.

Sorafenib remains a viable therapeutic option for veterans with advanced HCC. However, it is uncertain how much benefit sorafenib affords to the veteran population, especially with the associated risks.

Author disclosures
The authors report no actual or potential conflicts of interest with regard to this article.

Disclaimer
The opinions expressed herein are those of the authors and do not necessarily reflect those of Federal Practitioner, Frontline Medical Communications Inc., the U.S. Government, or any of its agencies. This article may discuss unlabeled or investigational use of certain drugs. Please review the complete prescribing information for specific drugs or drug combinations—including indications, contraindications, warnings, and adverse effects—before administering pharmacologic therapy to patients.

In 2015, more than 35,660 new cases of liver cancer and 24,550 liver cancer-related deaths are expected to occur in the U.S. About 80% of these cases will consist of hepatocellular carcinoma, (HCC).¹ The incidence of HCC varies throughout the world: Incidence is as low as 5 in 100,000 individuals in North America and ranges up to > 20 in 100,000 individuals in sub-Saharan Africa and Eastern Asia.² Nearly half of all cases of HCC are associated with hepatitis B virus (HBV), and another 25% are associated with hepatitis C virus (HCV). Other risk factors for developing HCC include alcoholic liver disease, nonalcoholic steatohepatitis, flatoxin-contaminated food, diabetes, and obesity.³

Therapeutic options for advanced HCC are limited. The FDA approved sorafenib in 2008 for the treatment of unresectable HCC.⁴ According to the American Association for the Study of Liver Diseases (AASLD) and the Barcelona Clinic Liver Cancer (BCLC) staging system, patients with Stage C liver cancer may undergo a trial of sorafenib.⁴ National Comprehensive Cancer Network (NCCN) clinical guidelines for hepatobiliary cancers reserve sorafenib for patients with inoperable tumors, metastatic disease, or extensive liver tumor burden.⁵ Sorafenib is shown to inhibit multiple intracellular and cell surface kinases. Several of these kinases are thought to be involved in tumor cell signaling, angiogenesis, and apoptosis.⁴ In the Sorafenib HCC Assessment Randomized Protocol (SHARP) trial, median overall survival (OS) was 10.7 months in the sorafenib group and 7.9 months in the placebo group.⁶ The predicted survival rates at 1 year were 44% in the sorafenib group and 33% in the placebo group.⁶The economic impact of oral chemotherapy on health care cannot be discounted. At about $50,000 to $100,000 per quality- 
adjusted life-year, the incremental cost-effectiveness ratio (ICER) of sorafenib over placebo was $62,473 per quality-adjusted life-year in 2007.⁷The purpose of this retrospective chart review was to evaluate sorafenib for efficacy and safety in a veteran population. Veterans have poorer health and more medical conditions compared with nonveterans.⁸ Furthermore, in the VHA, about 170,000 veterans have HCV.⁹ The rate of progression from HCV to HCC is about 3% to 5% annually. More than half of those diagnosed with HCC are late stage, and unfortunately, the 5-year OS rate for patients with liver cancer is 9% and 4% for those patients who are diagnosed at regional and distant stages of the disease.¹ As the practice of oncology grows, it is necessary for pharmacists to be involved in the selection of chemotherapeutic agents in order to provide optimal pharmaceutical care.¹⁰

Related: VIDEO: NAFLD increasingly causing U.S. hepatocellular carcinoma

Methods

A retrospective chart review was conducted to identify patients who were prescribed sorafenib from November 1, 2007, to September 30, 2011, at the VA Greater Los Angeles Healthcare System (VAGLAHS). Inclusion criteria included patients who had a diagnosis of advanced HCC, who were initiated and managed by a VAGLAHS provider and who were eligible for a 1-year safety evaluation period. The study was approved by the VAGLAHS institutional review board.

Baseline demographic, clinical, laboratory, and medication data were collected. Demographic, clinical, laboratory, and medication data were obtained from CPRS (Computerized Patient Record System) and VistA (Veterans Health Information Systems and Technology Architecture). Data were collected on secured servers and saved on encrypted files. The master list was destroyed once the records control schedule was finalized. No identifiers were collected on the data collection sheet.

Standard practice at VAGLAHS is to monitor European Cooperative Oncology Group Performance Status (ECOG-PS), Child-Pugh class, and alpha-fetoprotein (AFP) at initiation and every 3 months and to obtain laboratory data at initiation and every month before each medication refill. Patients were seen in the Oncology Clinic periodically at the provider’s discretion. The time of drug discontinuation and the reason for drug discontinuation were recorded. Time of death at any point was collected to measure OS.

It was determined that a total sample size of 42 patients would be insufficient to achieve 80% power to demonstrate any hypothesized effects. However, the Fisher exact test was used to calculate P values for simple comparison. Patient demographics and clinical characteristics were reported as total numbers and frequencies when applicable. Survival rate was measured from the time of sorafenib initiation to 1 year after therapy initiation. Overall survival was measured from the time of sorafenib initiation to time of death. Duration of therapy was measured from the time of sorafenib initiation to time of discontinuation, either by 
provider or by patient.

Results

There were 83 patients who were prescribed sorafenib between November 1, 2007, and September 30, 2011. Of the 83 patients, 27 patients were ineligible for a 1-year follow-up period, 9 patients were diagnosed with non-HCC, 3 were initiated or managed by providers outside the institution, and 2 were not started on therapy. In all, 42 patients met inclusion criteria and had received at least 1 dose of sorafenib. The primary etiologies for HCC were history of alcohol abuse, HCV, and HBV. The primary risk factors were obesity, smoking, and diabetes. Many patients presented with multiple etiologies and risk factors. Ten patients (23.8%) had moderate-to-severe hepatic impairment (Child-Pugh class B or C). Baseline characteristics of these patients are listed in Table 1.

Efficacy

The median OS was 5.9 months and ranged from 21 days to 60 months. There were 17 patients who survived at the 1-year follow-up, including 1 patient who survived 363 days after treatment initiation, yielding an OS rate of 40.5%. Table 2 presents 1-year survival rates with respect to select baseline data. Baseline factors found to be negligible were age, smoking, alcohol abuse, obesity, presence of HCV, medication possession ratio (MPR), prior treatment, macrovascular invasion, and AFP. Neither initial dose regimen, final dose regimen achieved, or average dose correlated with the survival rate at the 1-year follow-up.

Factors possibly associated with a higher probability of survival were baseline ECOG-PS score and baseline Child-Pugh class (Table 2). Patients with an ECOG-PS score of 0 or 1 had a higher survival rate at 1 year than did patients with an ECOG-PS score of ≥ 2 (50% vs 0%, respectively; P = .113). Patients with Child-Pugh class B or C had a lower survival rate at 1 year than did patients with Child-Pugh class A (51% vs 10%, respectively; 
P =  .028). Other indicators were size of largest hepatic lesion ≤ 5 cm, total bilirubin ≤ 2 mg/dL, concurrent treatment, almost exclusively embolization, and treatment after sorafenib discontinuation, such as another oral chemotherapeutic agent or embolization.

The 17 patients who survived at 1 year were reviewed to see if they shared characteristics that indicated a higher probability of survival. The figure shows the baseline ECOG-PS score and the Child-Pugh class the patients who did and did not survive at the 1-year follow-up. In the first group, all patient possessed an ECOG-PS score of 0 or 1, and only 1 patient presented with Child-Pugh class B or C. In contrast, in the group who did not survive at the 1-year follow-up, there were 4 patients with ECOG-PS scores of > 1 and 9 patients who presented with Child-Pugh class B or C. The mean AFP level of this group was < 200 µg/mL, and only 4 patients were followed by Palliative Care Services. The average normalized MPR of this group was 71.9% compared with 85.3% for those who did not survive at the 1-year follow-up.

In patients who experienced at least 1 adverse event (AE), 16 survived, whereas only 1 who did not experience an AE survived (45.7% vs 14.3%, respectively; P = .210). Thirteen patients who experienced ≥ 3 AEs survived at 1 year; and only 3 patients who reported < 3 AEs survived at 
1 year (61.9% vs 14.0%, respectively; 
P = .011). However, when the number of AEs was normalized to duration of treatment per patient, the median frequency of AEs for all patients was 0.61 AEs per month treated. The difference in survival rates grew smaller and less significant between patients who had a frequency of AEs lower than the median compared with those with a higher ratio (52.4% vs 28.6%, respectively; P = .208). Patients affected by AEs in the first 30 days and 90 days of treatment had a survival rate at the 1-year follow-up of 42.4% and 30.2%, respectively. Patients who experienced dermatologic AEs had a higher survival rate than those who did not have dermatologic AEs (60.0% vs 29.6%, respectively; P = .099). This correlation was not found with 2 other classes of AEs, gastrointestinal (50.0% vs 27.8%; 
P = .208) or neurologic (64.0% vs 41.2%; P = .209).

The median overall time to discontinuation was 3.4 months. The main reasons cited for discontinuing sorafenib at 1 year included symptomatic progression (52.4%), radiographic progression (23.8%), severe AEs (16.7%), and mild-to-moderate AEs (11.9%). There was overlap, as 15 patients discontinued treatment for multiple reasons. For the 22 patients who discontinued medication due to symptomatic progression at 1 year, the median time to discontinuation was 3.8 months. For the 10 patients who discontinued medication due to radiographic progression at 1 year, median time to discontinuation was 5.6 months. Seven patients (16.7%) were still on therapy at 1 year.The study considered the impact of potential dose adjustments on survival rate and safety. The authors compared patients’ prescribed dose with the recommended dose based on the package insert and monthly laboratory values if recorded. The prescribed dose was recorded as appropriate dose, below dose, above dose, or indeterminate due to the lack of current laboratory values. Patients who survived at the 1-year follow-up had a composition of 26%, 21%, 10%, and 43%, respectively. These results were similar to those of patients who did not survive at the 1-year follow-up, 29%, 12%, 30%, and 29%, respectively.

Based on medication refill history and VA acquisition cost, the total prescription drug cost of treating 42 patients with sorafenib was $388,370.40. The total number of days survived for these patients was 16,607 days, which equates to $8,535.87 per year lived.

Safety

Of the 42 patients, 35 patients experienced ≥ 1 AE for a total of 122 AEs reported. The median number of AEs per patient was 2.5. The median time to the first AE was 21 days and ranged from 3 to 244 days. In the first 30 days of treatment, 23 patients (54.7%) reported 47 AEs (39.5%). In the first 90 days of treatment, 33 patients (78.6%) reported 88 AEs (73.9%). Common AEs in both instances were diarrhea, fatigue, erythematous plantar-palmar rash, and nausea (Table 3).

The predominant classes of AEs were GI (39.3%), dermatologic (18.9%), and neurologic (15.6%). Erythematous palmar-plantar rash, also known as hand-foot syndrome, has been noted as a potential dose-limiting sorafenib AE if the rash is recurrent or severe. One patient experienced recurrent grade-2 rashes, and sorafenib was immediately discontinued after an attempt to lower the dose. There were 8 patients who reported serious AEs, and 5 were hospitalized. One patient continued therapy despite GI hemorrhage. The other 4 patients discontinued therapy on hospitalization and were seen for intracranial hemorrhage, GI perforation, acute renal failure, and acute liver failure. In the first 3 cases, sorafenib could not be ruled out as the primary cause of death. None of these patients presented with comorbidities, such as hypertension, which predisposed them to AEs.

Overall, 38 patients ended therapy at the recommended regimen of 400 mg twice daily, and the average total daily dose was 619 mg, just below 80% of the recommended daily dose. Reasons for not achieving 400 mg twice daily included slow titration, AEs, and dose adjustments for compromised renal and hepatic function such as dialysis. Patients who had an ECOG-PS score of 
0 or 1 or Child-Pugh class A reported ≥ 3 AEs, but when normalized to duration of treatment, no difference was observed. No correlations were found for average dose, creatinine clearance, aspartate aminotransferase, platelets, total bilirubin, or weight and number or frequency of AEs.

In regard to potential dose adjustments, the doses (400 mg twice daily, 600 mg daily [400 mg + 200 mg in 2 doses], 200 mg twice daily, and 200 mg daily) did not correlate well with AEs. Patients who had < 3 AEs presented with the breakdown 23%, 16%, 22%, and 38%, similar to patients who had ≥ 3 AEs—30%, 19%, 14%, and 37%. Likewise, patients who had a frequency of AEs lower than the median presented with the breakdown 22%, 22%, 15%, and 40% compared with patients who had more AEs than the median—37%, 9%, 23%, and 31%.

Related: Hepatocellular Carcinoma: To Biopsy or Not?

Discussion

Sorafenib is the only oral oncology medication approved by the FDA for treatment of unresectable HCC.³ Prior to sorafenib, the AASLD recommendation was supportive care for patients presenting with BCLC-Stage C liver cancer. However, guidelines changed when SHARP showed that sorafenib provided a survival benefit with a tolerable AE profile. The survival benefit of sorafenib has been replicated in a few large, multicenter trials. In Asia, Cheng and colleagues saw improved median OS of 6.5 months for sorafenib compared with 4.2 months with placebo, and in Italy, Iavarone and colleagues showed a median OS of 10.5 months without a placebo comparator.^11,12

In the veteran population for this study, the OS rate of 40.5% was similar to the rate reported in the SHARP study, although the patients’ median OS fell short of the time described in SHARP and other trials. The medical complexities involved in treating veterans may explain this difference. The veteran population is heterogeneous with diverse ethnic backgrounds, several comorbidities, and varying degrees of organ dysfunction. The authors compared survival rates of different subgroups to test the hypothesis that the probability of survival while on therapy should not depend on demographics or medical history. However, in this study, patients with minimal impact from HCC, such as mild hepatic 
impairment and high-functional status, demonstrated higher survival rates at 1-year follow-up than did those without significant compromise.Although the high prevalence of HCV and alcohol abuse in the veteran population has resulted in a high incidence of hepatic dysfunction, this study suggests that these factors are independent of survival if liver function or integrity has not been compromised.⁹

Some researchers have hypothesized that clinical toxicities from tyrosine kinase inhibitors may correlate with survival.¹³ The authors noticed that the presentation of dermatologic AEs may reflect improved survival. In this study, patients who experienced ≥ 1 AE and ≥ 3 AEs had survival rates at the 1-year follow-up of 45.7% and 61.9%, respectively. Moreover, patients affected by AEs in the first 90 days of treatment had a survival rate at the 1-year follow-up of 42.4%.

Caution is advised when drawing conclusions from the number of AEs or when they appear, because this may falsely favor correlation. Patients who survive longer have additional time to report an AE. Therefore, the authors also looked at the ratio of AEs over time per patient to consider the number of AEs per duration of treatment and saw that there was little difference in survival rate in this regard. When considering patients affected by AEs only in the first 30 days of treatment, the survival rate at the 1-year follow-up fell to 30.2%.

A more likely factor for the survival of the 17 patients who were alive at the 1-year follow-up was their overall health relative to the rest of the study group. Overall health may indicate survival independent of sorafenib. The group of 17 who survived at the 1-year follow-up reflected a population that was different from the rest of the study population. The subset was generally healthier with better ECOG-PS scores and Child-Pugh classes, was not followed by Palliative Care Services, and had a mean AFP level under the threshold for diagnosis of HCC in patients who present with hepatic lesions and elevated AFP.¹⁴ This subset’s MPR, a surrogate marker for adherence, was less than the accepted threshold in clinical practice for oral medications.¹⁵Evaluating the patient’s dose regimen was expected to reveal a relationship between dosing and clinical outcomes, such as low survival rates with low doses or more AEs with high doses. However, the authors were not able to establish this link. In fact, the median time to discontinuation of 3.⁴ months for the study group, or duration of treatment, was much shorter than the median OS of 5.9 months.

These findings were consistent with Cabibbo and colleagues, who conducted a meta-analysis of survival rates for untreated patients and found that impaired performance status and Child-Pugh class B or C were independently associated with shorter survival.¹⁶ The SHARP study and Cheng and colleagues also attempted to exclude patients who were not Child-Pugh class A in their studies, which suggests a negligible correlation between sorafenib and survival time and a close relationship between baseline clinical status and survival.

The authors determined that prior treatment, including locoregional therapy, was not a factor in predicting survival. This observation is confirmed by the results of a phase 3 study that looked at sorafenib as adjuvant treatment for patients who had no detectable disease after surgical resection or local ablation.¹⁷ The trial did not meet its primary endpoint of improved recurrence-free survival. However, the authors observed in this study that 4 patients who underwent resection of the liver before sorafenib had a mean OS of 2.9 years. One patient, who was alive at the time of the study conclusion, received only 
22 days of sorafenib treatment and survived for 4.9 years after sorafenib discontinuation. Patients who received concurrent or postsorafenib treatment had higher survival rates.

The cost of treatment in this study was found to be $8,535.87 per year lived. Although formal quality of life assessments were not captured, medication was discontinued at the first sign of disease progression or AE as determined by the provider or patient. When the cost of treatment was adjusted to account for median OS time and VA drug acquisition costs, estimated at average wholesale price minus 40%, the cost of treatment was within the threshold of $50,000-$100,000 per quality-adjusted life-year.^7,18Of the 42 patients in this study, 28.6% discontinued therapy due to AEs, compared with 32% observed in the SHARP study. Common GI, dermatologic, and CNS AEs were comparable between the 2 studies. Serious AEs included intracranial hemorrhage, GI hemorrhage, GI perforation, acute liver failure, and acute renal failure; 3 of these events led to death. About 12% of patients experienced bleeding, regardless of severity, compared with the 18% seen in SHARP, despite no prior history of hemorrhage or GI perforation.⁵ The authors did not find any clinical factors at baseline that predisposed patients to AEs. It was also difficult to distinguish between drug-related AEs and general disease progression.

Although the authors did not find a relationship between dose or dose adjustments and the number or frequency of AEs, there were serious adverse outcomes in this study that were also rare complications observed in SHARP. The decision to start sorafenib should not be taken lightly.

Related: Diagnostic Dilemma of Hepatocellular Carcinoma Presenting as Hepatic Angiomyolipoma

Limitations

This retrospective review had several limitations. In SHARP and other large, multicenter trials, patients were continued on therapy until they experienced both symptomatic and radiographic progression. In this study, patients were discontinued at the first sign of progression, either symptomatic or radiographic or both. Had all patients remained on therapy until symptomatic and radiographic signs of progression were observed, there could have been a better correlation between duration of treatment and OS, symptomatic progression, or radiographic progression. The authors acknowledge, however, that there is diminishing benefit of administering chemotherapy when there are known and potentially serious AEs.

The data for this study were limited due to a small sample size, and it was not powered to evaluate for statistically significant characteristics between the patients who survived at the 1-year follow-up and the patients who did not survive at the 1-year follow-up. This information would be useful to identify potential prognostic factors and guide providers in sorafenib management. Finally, a long-term safety profile could not be established, as patients were evaluated for a 1-year period.

Ultimately, HCC is a multifactorial disease, and it is difficult to account for all potential confounding factors. Additional research, including studies comparing sunitinib or a control group to sorafenib, may provide further insight.

Conclusions

In light of these results, the authors believe that sorafenib may be considered for veterans with unresectable HCC and who are contraindicated for alternative treatments. One-year survival rates were similar to those seen in previous studies. However, there was no clear association between the duration of treatment and OS, and although the medication was well tolerated, there were also serious AEs. It is prudent to continually assess the need for therapy throughout the treatment period.

Pharmacists have a critical role in care for oncology patients, from the integration of certified clinical pharmacist practitioners into hematology-oncology clinics to patient monitoring through oral oncology pharmacy programs.^19,20 These 
programs have been shown to improve patient outcomes and decrease overall health care use and may benefit the veteran population.

In this study, a veteran population achieved a survival rate at the 1-year follow-up similar to that found in SHARP: 40.5% vs 44%. However, OS was markedly shorter: 5.9 months vs 10.7 months. Patients with minimal impact from HCC, such as mild hepatic impairment and high functional status, demonstrated higher survival rates at the 1-year follow-up than did those with significant compromise. Thirty-five patients experienced ≥1 AE, most observed within the first 90 days of treatment, and for 3 patients, sorafenib could not be ruled out as the cause of death.

Sorafenib remains a viable therapeutic option for veterans with advanced HCC. However, it is uncertain how much benefit sorafenib affords to the veteran population, especially with the associated risks.

Author disclosures
The authors report no actual or potential conflicts of interest with regard to this article.

Disclaimer
The opinions expressed herein are those of the authors and do not necessarily reflect those of Federal Practitioner, Frontline Medical Communications Inc., the U.S. Government, or any of its agencies. This article may discuss unlabeled or investigational use of certain drugs. Please review the complete prescribing information for specific drugs or drug combinations—including indications, contraindications, warnings, and adverse effects—before administering pharmacologic therapy to patients.

In 2015, more than 35,660 new cases of liver cancer and 24,550 liver cancer-related deaths are expected to occur in the U.S. About 80% of these cases will consist of hepatocellular carcinoma, (HCC).¹ The incidence of HCC varies throughout the world: Incidence is as low as 5 in 100,000 individuals in North America and ranges up to > 20 in 100,000 individuals in sub-Saharan Africa and Eastern Asia.² Nearly half of all cases of HCC are associated with hepatitis B virus (HBV), and another 25% are associated with hepatitis C virus (HCV). Other risk factors for developing HCC include alcoholic liver disease, nonalcoholic steatohepatitis, flatoxin-contaminated food, diabetes, and obesity.³

Therapeutic options for advanced HCC are limited. The FDA approved sorafenib in 2008 for the treatment of unresectable HCC.⁴ According to the American Association for the Study of Liver Diseases (AASLD) and the Barcelona Clinic Liver Cancer (BCLC) staging system, patients with Stage C liver cancer may undergo a trial of sorafenib.⁴ National Comprehensive Cancer Network (NCCN) clinical guidelines for hepatobiliary cancers reserve sorafenib for patients with inoperable tumors, metastatic disease, or extensive liver tumor burden.⁵ Sorafenib is shown to inhibit multiple intracellular and cell surface kinases. Several of these kinases are thought to be involved in tumor cell signaling, angiogenesis, and apoptosis.⁴ In the Sorafenib HCC Assessment Randomized Protocol (SHARP) trial, median overall survival (OS) was 10.7 months in the sorafenib group and 7.9 months in the placebo group.⁶ The predicted survival rates at 1 year were 44% in the sorafenib group and 33% in the placebo group.⁶The economic impact of oral chemotherapy on health care cannot be discounted. At about $50,000 to $100,000 per quality- 
adjusted life-year, the incremental cost-effectiveness ratio (ICER) of sorafenib over placebo was $62,473 per quality-adjusted life-year in 2007.⁷The purpose of this retrospective chart review was to evaluate sorafenib for efficacy and safety in a veteran population. Veterans have poorer health and more medical conditions compared with nonveterans.⁸ Furthermore, in the VHA, about 170,000 veterans have HCV.⁹ The rate of progression from HCV to HCC is about 3% to 5% annually. More than half of those diagnosed with HCC are late stage, and unfortunately, the 5-year OS rate for patients with liver cancer is 9% and 4% for those patients who are diagnosed at regional and distant stages of the disease.¹ As the practice of oncology grows, it is necessary for pharmacists to be involved in the selection of chemotherapeutic agents in order to provide optimal pharmaceutical care.¹⁰

Related: VIDEO: NAFLD increasingly causing U.S. hepatocellular carcinoma

Methods

A retrospective chart review was conducted to identify patients who were prescribed sorafenib from November 1, 2007, to September 30, 2011, at the VA Greater Los Angeles Healthcare System (VAGLAHS). Inclusion criteria included patients who had a diagnosis of advanced HCC, who were initiated and managed by a VAGLAHS provider and who were eligible for a 1-year safety evaluation period. The study was approved by the VAGLAHS institutional review board.

Baseline demographic, clinical, laboratory, and medication data were collected. Demographic, clinical, laboratory, and medication data were obtained from CPRS (Computerized Patient Record System) and VistA (Veterans Health Information Systems and Technology Architecture). Data were collected on secured servers and saved on encrypted files. The master list was destroyed once the records control schedule was finalized. No identifiers were collected on the data collection sheet.

Standard practice at VAGLAHS is to monitor European Cooperative Oncology Group Performance Status (ECOG-PS), Child-Pugh class, and alpha-fetoprotein (AFP) at initiation and every 3 months and to obtain laboratory data at initiation and every month before each medication refill. Patients were seen in the Oncology Clinic periodically at the provider’s discretion. The time of drug discontinuation and the reason for drug discontinuation were recorded. Time of death at any point was collected to measure OS.

It was determined that a total sample size of 42 patients would be insufficient to achieve 80% power to demonstrate any hypothesized effects. However, the Fisher exact test was used to calculate P values for simple comparison. Patient demographics and clinical characteristics were reported as total numbers and frequencies when applicable. Survival rate was measured from the time of sorafenib initiation to 1 year after therapy initiation. Overall survival was measured from the time of sorafenib initiation to time of death. Duration of therapy was measured from the time of sorafenib initiation to time of discontinuation, either by 
provider or by patient.

Results

There were 83 patients who were prescribed sorafenib between November 1, 2007, and September 30, 2011. Of the 83 patients, 27 patients were ineligible for a 1-year follow-up period, 9 patients were diagnosed with non-HCC, 3 were initiated or managed by providers outside the institution, and 2 were not started on therapy. In all, 42 patients met inclusion criteria and had received at least 1 dose of sorafenib. The primary etiologies for HCC were history of alcohol abuse, HCV, and HBV. The primary risk factors were obesity, smoking, and diabetes. Many patients presented with multiple etiologies and risk factors. Ten patients (23.8%) had moderate-to-severe hepatic impairment (Child-Pugh class B or C). Baseline characteristics of these patients are listed in Table 1.

Efficacy

The median OS was 5.9 months and ranged from 21 days to 60 months. There were 17 patients who survived at the 1-year follow-up, including 1 patient who survived 363 days after treatment initiation, yielding an OS rate of 40.5%. Table 2 presents 1-year survival rates with respect to select baseline data. Baseline factors found to be negligible were age, smoking, alcohol abuse, obesity, presence of HCV, medication possession ratio (MPR), prior treatment, macrovascular invasion, and AFP. Neither initial dose regimen, final dose regimen achieved, or average dose correlated with the survival rate at the 1-year follow-up.

Factors possibly associated with a higher probability of survival were baseline ECOG-PS score and baseline Child-Pugh class (Table 2). Patients with an ECOG-PS score of 0 or 1 had a higher survival rate at 1 year than did patients with an ECOG-PS score of ≥ 2 (50% vs 0%, respectively; P = .113). Patients with Child-Pugh class B or C had a lower survival rate at 1 year than did patients with Child-Pugh class A (51% vs 10%, respectively; 
P =  .028). Other indicators were size of largest hepatic lesion ≤ 5 cm, total bilirubin ≤ 2 mg/dL, concurrent treatment, almost exclusively embolization, and treatment after sorafenib discontinuation, such as another oral chemotherapeutic agent or embolization.

The 17 patients who survived at 1 year were reviewed to see if they shared characteristics that indicated a higher probability of survival. The figure shows the baseline ECOG-PS score and the Child-Pugh class the patients who did and did not survive at the 1-year follow-up. In the first group, all patient possessed an ECOG-PS score of 0 or 1, and only 1 patient presented with Child-Pugh class B or C. In contrast, in the group who did not survive at the 1-year follow-up, there were 4 patients with ECOG-PS scores of > 1 and 9 patients who presented with Child-Pugh class B or C. The mean AFP level of this group was < 200 µg/mL, and only 4 patients were followed by Palliative Care Services. The average normalized MPR of this group was 71.9% compared with 85.3% for those who did not survive at the 1-year follow-up.

In patients who experienced at least 1 adverse event (AE), 16 survived, whereas only 1 who did not experience an AE survived (45.7% vs 14.3%, respectively; P = .210). Thirteen patients who experienced ≥ 3 AEs survived at 1 year; and only 3 patients who reported < 3 AEs survived at 
1 year (61.9% vs 14.0%, respectively; 
P = .011). However, when the number of AEs was normalized to duration of treatment per patient, the median frequency of AEs for all patients was 0.61 AEs per month treated. The difference in survival rates grew smaller and less significant between patients who had a frequency of AEs lower than the median compared with those with a higher ratio (52.4% vs 28.6%, respectively; P = .208). Patients affected by AEs in the first 30 days and 90 days of treatment had a survival rate at the 1-year follow-up of 42.4% and 30.2%, respectively. Patients who experienced dermatologic AEs had a higher survival rate than those who did not have dermatologic AEs (60.0% vs 29.6%, respectively; P = .099). This correlation was not found with 2 other classes of AEs, gastrointestinal (50.0% vs 27.8%; 
P = .208) or neurologic (64.0% vs 41.2%; P = .209).

The median overall time to discontinuation was 3.4 months. The main reasons cited for discontinuing sorafenib at 1 year included symptomatic progression (52.4%), radiographic progression (23.8%), severe AEs (16.7%), and mild-to-moderate AEs (11.9%). There was overlap, as 15 patients discontinued treatment for multiple reasons. For the 22 patients who discontinued medication due to symptomatic progression at 1 year, the median time to discontinuation was 3.8 months. For the 10 patients who discontinued medication due to radiographic progression at 1 year, median time to discontinuation was 5.6 months. Seven patients (16.7%) were still on therapy at 1 year.The study considered the impact of potential dose adjustments on survival rate and safety. The authors compared patients’ prescribed dose with the recommended dose based on the package insert and monthly laboratory values if recorded. The prescribed dose was recorded as appropriate dose, below dose, above dose, or indeterminate due to the lack of current laboratory values. Patients who survived at the 1-year follow-up had a composition of 26%, 21%, 10%, and 43%, respectively. These results were similar to those of patients who did not survive at the 1-year follow-up, 29%, 12%, 30%, and 29%, respectively.

Based on medication refill history and VA acquisition cost, the total prescription drug cost of treating 42 patients with sorafenib was $388,370.40. The total number of days survived for these patients was 16,607 days, which equates to $8,535.87 per year lived.

Safety

Of the 42 patients, 35 patients experienced ≥ 1 AE for a total of 122 AEs reported. The median number of AEs per patient was 2.5. The median time to the first AE was 21 days and ranged from 3 to 244 days. In the first 30 days of treatment, 23 patients (54.7%) reported 47 AEs (39.5%). In the first 90 days of treatment, 33 patients (78.6%) reported 88 AEs (73.9%). Common AEs in both instances were diarrhea, fatigue, erythematous plantar-palmar rash, and nausea (Table 3).

The predominant classes of AEs were GI (39.3%), dermatologic (18.9%), and neurologic (15.6%). Erythematous palmar-plantar rash, also known as hand-foot syndrome, has been noted as a potential dose-limiting sorafenib AE if the rash is recurrent or severe. One patient experienced recurrent grade-2 rashes, and sorafenib was immediately discontinued after an attempt to lower the dose. There were 8 patients who reported serious AEs, and 5 were hospitalized. One patient continued therapy despite GI hemorrhage. The other 4 patients discontinued therapy on hospitalization and were seen for intracranial hemorrhage, GI perforation, acute renal failure, and acute liver failure. In the first 3 cases, sorafenib could not be ruled out as the primary cause of death. None of these patients presented with comorbidities, such as hypertension, which predisposed them to AEs.

Overall, 38 patients ended therapy at the recommended regimen of 400 mg twice daily, and the average total daily dose was 619 mg, just below 80% of the recommended daily dose. Reasons for not achieving 400 mg twice daily included slow titration, AEs, and dose adjustments for compromised renal and hepatic function such as dialysis. Patients who had an ECOG-PS score of 
0 or 1 or Child-Pugh class A reported ≥ 3 AEs, but when normalized to duration of treatment, no difference was observed. No correlations were found for average dose, creatinine clearance, aspartate aminotransferase, platelets, total bilirubin, or weight and number or frequency of AEs.

In regard to potential dose adjustments, the doses (400 mg twice daily, 600 mg daily [400 mg + 200 mg in 2 doses], 200 mg twice daily, and 200 mg daily) did not correlate well with AEs. Patients who had < 3 AEs presented with the breakdown 23%, 16%, 22%, and 38%, similar to patients who had ≥ 3 AEs—30%, 19%, 14%, and 37%. Likewise, patients who had a frequency of AEs lower than the median presented with the breakdown 22%, 22%, 15%, and 40% compared with patients who had more AEs than the median—37%, 9%, 23%, and 31%.

Related: Hepatocellular Carcinoma: To Biopsy or Not?

Discussion

Sorafenib is the only oral oncology medication approved by the FDA for treatment of unresectable HCC.³ Prior to sorafenib, the AASLD recommendation was supportive care for patients presenting with BCLC-Stage C liver cancer. However, guidelines changed when SHARP showed that sorafenib provided a survival benefit with a tolerable AE profile. The survival benefit of sorafenib has been replicated in a few large, multicenter trials. In Asia, Cheng and colleagues saw improved median OS of 6.5 months for sorafenib compared with 4.2 months with placebo, and in Italy, Iavarone and colleagues showed a median OS of 10.5 months without a placebo comparator.^11,12

In the veteran population for this study, the OS rate of 40.5% was similar to the rate reported in the SHARP study, although the patients’ median OS fell short of the time described in SHARP and other trials. The medical complexities involved in treating veterans may explain this difference. The veteran population is heterogeneous with diverse ethnic backgrounds, several comorbidities, and varying degrees of organ dysfunction. The authors compared survival rates of different subgroups to test the hypothesis that the probability of survival while on therapy should not depend on demographics or medical history. However, in this study, patients with minimal impact from HCC, such as mild hepatic 
impairment and high-functional status, demonstrated higher survival rates at 1-year follow-up than did those without significant compromise.Although the high prevalence of HCV and alcohol abuse in the veteran population has resulted in a high incidence of hepatic dysfunction, this study suggests that these factors are independent of survival if liver function or integrity has not been compromised.⁹

Some researchers have hypothesized that clinical toxicities from tyrosine kinase inhibitors may correlate with survival.¹³ The authors noticed that the presentation of dermatologic AEs may reflect improved survival. In this study, patients who experienced ≥ 1 AE and ≥ 3 AEs had survival rates at the 1-year follow-up of 45.7% and 61.9%, respectively. Moreover, patients affected by AEs in the first 90 days of treatment had a survival rate at the 1-year follow-up of 42.4%.

Caution is advised when drawing conclusions from the number of AEs or when they appear, because this may falsely favor correlation. Patients who survive longer have additional time to report an AE. Therefore, the authors also looked at the ratio of AEs over time per patient to consider the number of AEs per duration of treatment and saw that there was little difference in survival rate in this regard. When considering patients affected by AEs only in the first 30 days of treatment, the survival rate at the 1-year follow-up fell to 30.2%.

A more likely factor for the survival of the 17 patients who were alive at the 1-year follow-up was their overall health relative to the rest of the study group. Overall health may indicate survival independent of sorafenib. The group of 17 who survived at the 1-year follow-up reflected a population that was different from the rest of the study population. The subset was generally healthier with better ECOG-PS scores and Child-Pugh classes, was not followed by Palliative Care Services, and had a mean AFP level under the threshold for diagnosis of HCC in patients who present with hepatic lesions and elevated AFP.¹⁴ This subset’s MPR, a surrogate marker for adherence, was less than the accepted threshold in clinical practice for oral medications.¹⁵Evaluating the patient’s dose regimen was expected to reveal a relationship between dosing and clinical outcomes, such as low survival rates with low doses or more AEs with high doses. However, the authors were not able to establish this link. In fact, the median time to discontinuation of 3.⁴ months for the study group, or duration of treatment, was much shorter than the median OS of 5.9 months.

These findings were consistent with Cabibbo and colleagues, who conducted a meta-analysis of survival rates for untreated patients and found that impaired performance status and Child-Pugh class B or C were independently associated with shorter survival.¹⁶ The SHARP study and Cheng and colleagues also attempted to exclude patients who were not Child-Pugh class A in their studies, which suggests a negligible correlation between sorafenib and survival time and a close relationship between baseline clinical status and survival.

The authors determined that prior treatment, including locoregional therapy, was not a factor in predicting survival. This observation is confirmed by the results of a phase 3 study that looked at sorafenib as adjuvant treatment for patients who had no detectable disease after surgical resection or local ablation.¹⁷ The trial did not meet its primary endpoint of improved recurrence-free survival. However, the authors observed in this study that 4 patients who underwent resection of the liver before sorafenib had a mean OS of 2.9 years. One patient, who was alive at the time of the study conclusion, received only 
22 days of sorafenib treatment and survived for 4.9 years after sorafenib discontinuation. Patients who received concurrent or postsorafenib treatment had higher survival rates.

The cost of treatment in this study was found to be $8,535.87 per year lived. Although formal quality of life assessments were not captured, medication was discontinued at the first sign of disease progression or AE as determined by the provider or patient. When the cost of treatment was adjusted to account for median OS time and VA drug acquisition costs, estimated at average wholesale price minus 40%, the cost of treatment was within the threshold of $50,000-$100,000 per quality-adjusted life-year.^7,18Of the 42 patients in this study, 28.6% discontinued therapy due to AEs, compared with 32% observed in the SHARP study. Common GI, dermatologic, and CNS AEs were comparable between the 2 studies. Serious AEs included intracranial hemorrhage, GI hemorrhage, GI perforation, acute liver failure, and acute renal failure; 3 of these events led to death. About 12% of patients experienced bleeding, regardless of severity, compared with the 18% seen in SHARP, despite no prior history of hemorrhage or GI perforation.⁵ The authors did not find any clinical factors at baseline that predisposed patients to AEs. It was also difficult to distinguish between drug-related AEs and general disease progression.

Although the authors did not find a relationship between dose or dose adjustments and the number or frequency of AEs, there were serious adverse outcomes in this study that were also rare complications observed in SHARP. The decision to start sorafenib should not be taken lightly.

Related: Diagnostic Dilemma of Hepatocellular Carcinoma Presenting as Hepatic Angiomyolipoma

Limitations

This retrospective review had several limitations. In SHARP and other large, multicenter trials, patients were continued on therapy until they experienced both symptomatic and radiographic progression. In this study, patients were discontinued at the first sign of progression, either symptomatic or radiographic or both. Had all patients remained on therapy until symptomatic and radiographic signs of progression were observed, there could have been a better correlation between duration of treatment and OS, symptomatic progression, or radiographic progression. The authors acknowledge, however, that there is diminishing benefit of administering chemotherapy when there are known and potentially serious AEs.

The data for this study were limited due to a small sample size, and it was not powered to evaluate for statistically significant characteristics between the patients who survived at the 1-year follow-up and the patients who did not survive at the 1-year follow-up. This information would be useful to identify potential prognostic factors and guide providers in sorafenib management. Finally, a long-term safety profile could not be established, as patients were evaluated for a 1-year period.

Ultimately, HCC is a multifactorial disease, and it is difficult to account for all potential confounding factors. Additional research, including studies comparing sunitinib or a control group to sorafenib, may provide further insight.

Conclusions

In light of these results, the authors believe that sorafenib may be considered for veterans with unresectable HCC and who are contraindicated for alternative treatments. One-year survival rates were similar to those seen in previous studies. However, there was no clear association between the duration of treatment and OS, and although the medication was well tolerated, there were also serious AEs. It is prudent to continually assess the need for therapy throughout the treatment period.

Pharmacists have a critical role in care for oncology patients, from the integration of certified clinical pharmacist practitioners into hematology-oncology clinics to patient monitoring through oral oncology pharmacy programs.^19,20 These 
programs have been shown to improve patient outcomes and decrease overall health care use and may benefit the veteran population.

In this study, a veteran population achieved a survival rate at the 1-year follow-up similar to that found in SHARP: 40.5% vs 44%. However, OS was markedly shorter: 5.9 months vs 10.7 months. Patients with minimal impact from HCC, such as mild hepatic impairment and high functional status, demonstrated higher survival rates at the 1-year follow-up than did those with significant compromise. Thirty-five patients experienced ≥1 AE, most observed within the first 90 days of treatment, and for 3 patients, sorafenib could not be ruled out as the cause of death.

Sorafenib remains a viable therapeutic option for veterans with advanced HCC. However, it is uncertain how much benefit sorafenib affords to the veteran population, especially with the associated risks.

Author disclosures
The authors report no actual or potential conflicts of interest with regard to this article.

Disclaimer
The opinions expressed herein are those of the authors and do not necessarily reflect those of Federal Practitioner, Frontline Medical Communications Inc., the U.S. Government, or any of its agencies. This article may discuss unlabeled or investigational use of certain drugs. Please review the complete prescribing information for specific drugs or drug combinations—including indications, contraindications, warnings, and adverse effects—before administering pharmacologic therapy to patients.

A 27-year-old white woman presented to the clinic with headaches and decreased vision through her reading glasses while performing near tasks. Her medical history was significant for herpes simplex, hyperlipidemia, and migraine headaches with aura. Her migraines began following an earlier motor vehicle accident, and her most recent magnetic resonance imaging (MRI) showed no abnormalities. Her current medications included prophylactic acyclovir for herpes and acetaminophen and caffeine tablets as needed for headache. She reported no other trauma or surgery and no known allergies. The patient’s best-corrected Snellen visual acuities in both eyes were 20/20 (distance) and 20/30 (near).

Preliminary testing, including pupils, extraocular motilities, confrontation fields, and color vision, were all within normal limits. Her slit-lamp examination was unremarkable. A dilated fundus examination revealed crowded, elevated discs without vessel obscuration, hemorrhage, hyperemia, or drusen (Figure 1). The fundus examination was otherwise unremarkable. Optical coherence tomography of the optic nerves showed increased nerve fiber layer thickness in both eyes (Figure 2). Her blood pressure (BP) at this visit was 
106/77 mm/Hg.

The diagnosis based on these findings was bilateral optic nerve elevation with long-standing migraine headaches. The plan was for the patient to return to the clinic for repeat visual field testing and B-scan ultrasonography to rule out buried optic nerve head drusen.

Two months later, the patient presented to the clinic 19 weeks pregnant and reported that her headaches had increased in frequency, but she had no diplopia. All preliminary testing, including visual acuities, pupil reaction, color vision, and slit-lamp examination remained normal. Fundus examination showed the patient’s nerves were unchanged in appearance from the initial presentation. Visual fields revealed an enlarged blind spot in the right eye and paracentral defects in the left eye. The B-scan testing was negative for optic nerve drusen. Due to the increased frequency of headaches, pregnancy, and suspicious optic nerves, an urgent consult was placed to neurology.

At the neurology appointment 
1 month later, the patient was diagnosed with migraine headache syndrome and idiopathic intracranial hypertension (IIH). The neurologist believed her headaches might have been resulting from analgesic rebound. He suggested that the patient discontinue or decrease use of oral butalbital, acetaminophen and caffeine tablets, and other forms of caffeine. It was decided that divalproxen sodium and verapamil were not feasible due to pregnancy. The neurologist started her on oral acetazolamide 
500 mg twice daily.

The patient returned to her obstetrician 1 month later for a routine 
follow-up; the headaches had worsened and were now accompanied by nausea and vomiting twice daily on average. Her medications still included acetaminophen and caffeine tablets, although it had been recommended she discontinue them, prochlorperazine, and acetazolamide. Due to the worsening of her symptoms and visual fields (eFigure 1), the obstetrician recommended that the patient deliver by cesarean section at 38 to 39 weeks.

(eFigure 1.Visual Fields at Follow-up 1 and 2)
Right eye