Laura A. Drew

The Adherence evaluation After Ischemic Stroke Longitudinal (AVAIL) Registry : Design, rationale, and baseline patient characteristics

BACKGROUND Approximately one third of the 780,000 people in the United States who have a stroke each year have recurrent events. Although efficacious secondary prevention measures are available, levels of adherence to these strategies in patients who have had stroke are largely unknown. Understanding medication-taking behavior in this population is an important step to optimizing the appropriate use of proven secondary preventive therapies and reducing the risk of recurrent stroke. METHODS The Adherence eValuation After Ischemic Stroke Longitudinal (AVAIL) registry is a prospective study of adherence to stroke prevention medications from hospital discharge to 1 year in patients admitted with stroke or transient ischemic attack. The primary outcomes are medication usage as determined by patient interviews after 3 and 12 months. Potential patient-, provider-, and system-level barriers to persistence of medication use are also collected. Secondary outcomes include the rates of recurrent stroke or transient ischemic attack, vascular events, and rehospitalization and functional status as measured by the modified Rankin score. RESULTS The AVAIL enrolled about 2,900 subjects from 106 hospitals from July 2006 through July 2008. The 12-month follow-up will be completed in August 2009. CONCLUSIONS The AVAIL registry will document the current state of adherence and persistence to stroke prevention medications among a nationwide sample of patients. These data will be used to design interventions to improve the quality of care post acute hospitalization and reduce the risks of future stroke and cardiovascular events.

Quality of life and economic outcomes with surgical ventricular reconstruction in ischemic heart failure: results from the Surgical Treatment for Ischemic Heart Failure trial.

BACKGROUND Surgical ventricular reconstruction (SVR) is used in conjunction with coronary artery bypass graft surgery (CABG) to improve left ventricular function and clinical outcomes in selected patients with ischemic heart failure. The impact of SVR on quality of life (QOL) and medical costs is unknown. METHODS We compared CABG plus SVR with CABG alone in 1,000 patients with ischemic heart failure, an anterior wall scar, and a left ventricular ejection fraction <or=0.35. In 991 (99% of eligible), we collected a battery of QOL instruments. The principal, prespecified QOL measure was the Kansas City Cardiomyopathy Questionnaire, which evaluates the effects of heart failure symptoms on QOL using a scale from 0 to 100 with higher scores indicating better QOL. Structured QOL interviews were conducted at baseline, 4, 12, 24, and 36 months post randomization and were >or=92% complete. Cost data were collected on 196 (98%) of 200 patients enrolled in the United States. RESULTS Heart-failure-related QOL outcomes did not differ between the 2 treatment strategies out to 3 years (median Kansas City Cardiomyopathy Questionnaire scores for CABG alone and CABG plus SVR, respectively: baseline 53 versus 54, P = .53; 3 years 85 versus 84, P = .89). There were no treatment-related differences in other QOL measures. In the US patients, total index hospitalization costs averaged over

Quality-of-Life Outcomes With Coronary Artery Bypass Graft Surgery in Ischemic Left Ventricular Dysfunction: A Randomized Trial

14,500 higher for CABG plus SVR (P = .004) due primarily to 4.2 extra postoperative, high-intensity care days in the hospital. CONCLUSIONS Addition of SVR to CABG in patients with ischemic heart failure did not improve QOL but significantly increased health care costs.

Staphylococcus aureus bacteremia among patients with health care-associated fever.

Context A randomized, controlled trial of patients with high-risk coronary artery disease and heart failure previously reported no significant difference between medical therapy alone and medical therapy plus coronary artery bypass grafting when the outcome was death from any cause. Contribution This study from the same trial reports that medical therapy plus coronary artery bypass grafting is better than medical therapy alone when the outcome is quality of life. Caution The type of therapy was not concealed from patients or investigators. Implication This is the first study to examine how these therapies affect quality of life in patients who have coronary artery disease and heart failure. The Editors Clinical trials performed during the 1970s and 1980s defined several major coronary artery disease (CAD) subgroups for which coronary artery bypass grafting (CABG) provided incremental survival, angina relief, or both relative to medical therapy. These early trials formed the foundation for current practice patterns and guideline recommendations on the use of CABG (13). However, patients with severe left ventricular dysfunction (ejection fraction 0.35) were not represented. Thus, management decisions for these patients have largely relied on clinical judgment to extrapolate from those trials and a small group of observational studies(4, 5). The challenges in using this evidence to select treatment for contemporary patients is further complicated by the substantial improvement in medical therapies for both CAD and heart failure over those used in the earlier trials. The STICH (Surgical Treatment for Ischemic Heart Failure) trial was funded by the National Heart, Lung, and Blood Institute in 2002 to provide a comprehensive evaluation of the incremental therapeutic benefits of routine CABG over contemporary guideline-based medical therapy in patients with severe systolic dysfunction due to CAD (6). A major prespecified secondary end point of the trial was health-related quality of life (QOL), which is an outcome that complements the major clinical end points by assessing the patients experience of, and satisfaction with, the 2 therapeutic strategies compared (7, 8). Methods Patient Population and Primary Clinical Results To test the STICH trials surgical revascularization hypothesis, we randomly assigned 1212 patients with site-defined left ventricular ejection fraction of 0.35 or less and CAD suitable for revascularization to CABG or medical therapy (6). Rationale, trial design, and complete inclusion and exclusion criteria have been described previously (7). Patients were enrolled at 99 clinical sites in 22 countries between July 2002 and May 2007. All patients provided informed consent, and study protocol was approved by each sites institutional review board or ethics committee. Median follow-up was 56 months. The primary intention-to-treat comparison showed that 35.7% of patients assigned to CABG and 40.5% of those assigned to medical therapy died (primary analysis: unadjusted hazard ratio for all-cause mortality, 0.86 [95% CI, 0.72 to 1.04]; P= 0.123; secondary analysis: adjusted hazard ratio for all-cause mortality, 0.82 [CI, 0.68 to 0.99]; P= 0.039). Patients assigned to CABG had lower rates of the 2 major secondary clinical end points: death from cardiovascular causes (hazard ratio, 0.81; P= 0.050) and the composite of all-cause mortality and hospitalization for cardiovascular causes (hazard ratio, 0.74; P< 0.001). Health-Related QOL Data Collection We collected QOL data using structured interviews at baseline and 4, 12, 24, and 36 months after randomization. Site coordinators were specially trained by the Duke Clinical Research Institute Outcomes Research Group to conduct baseline interviews. The original research plan called for all patients to be enrolled in North America and all English- and Spanish-language follow-up QOL interviews to be completed via telephone by the Duke Clinical Research Institute. The few patients expected to require French-language interviews were to be interviewed by site coordinators. When enrollment was expanded outside of North America, the plan was modified to have those site coordinators do all QOL interviews in the patients native language. For nonEnglish-speaking participants, translations of the QOL instruments were obtained from the instrument developers or a translation service was used to create validated translations. The New York Heart Association (NYHA) class and Canadian Cardiovascular Society (CCS) angina class were collected on the clinical case report form at baseline and each follow-up interval. QOL Measures A battery of validated measures was used to provide a comprehensive but efficient assessment of QOL. The principal prespecified QOL measure was the Kansas City Cardiomyopathy Questionnaire (KCCQ), which is a 23-item instrument that measures the effect of heart failure symptoms on QOL (9). We used 3 scales from the Seattle Angina Questionnaire to assess the effect of angina symptoms on QOL outcomes (10), the Short Form-12 Survey to provide a brief overall generic measure of health status (11), and 5 individual scales from the Short Form-36 Health Survey to provide a more detailed assessment of areas of functioning and well-being from a generic perspective (12). Depressive symptoms were assessed using the Center for Epidemiologic Studies Depression Scale (13). The Cardiac Self-Efficacy Questionnaire was used to measure patient confidence in controlling disease symptoms and maintaining physical functioning (14). Finally, we used the EuroQol-5D, which is a generic instrument consisting of a 5-dimension single summary health status index and a self-rated visual analogue scale (0 to 100) of current health-related QOL (15). Statistical Analysis All primary comparisons were performed with the treatment group defined according to the intention-to-treat principle. Descriptive statistics included percentages for discrete variables and medians with 25th to 75th percentiles plus means with SDs for continuous variables. The chi-square test was used for discrete variable comparisons. Treatment comparisons of continuous variables were done by using a linear mixed model to account for repeated measures within a patient. The baseline QOL measure was used as one of the repeated measures. In PROC MIXED in SAS, version 9.2 (SAS Institute), the baseline, 4-, 12-, 24-, and 36-month measurements within a patient were fitted using maximum likelihood methods with unstructured covariance matrix (16). At each time point, estimated treatment differences, 95% CIs, and P values were obtained using the model estimates. Finally, the proportion of patients who achieved a clinically important improvement of 5 points or more in KCCQ overall summary scores were compared by treatment group using the chi-square test (17). All analyses were conducted using SAS, version 8 or higher. All reported P values were 2-sided. No adjustment was made in significance levels for multiple comparisons. Benchmarks for clinically significant changes in QOL scores for individual patients were used informally to assess the magnitude of the mean difference between the 2 groups. However, a QOL difference between groups at or exceeding the benchmark was not used as a formal decision rule to define clinical significance. Subgroup Analysis Regional effects were tested as interactions between the 5 regions used in the clinical analyses (Asia, Australia and New Zealand, Europe, North America, and South America) and the KCCQ overall summary score. Other interactions tested from those prespecified in the clinical analysis plan included age, sex, race, current NYHA class, left ventricular ejection fraction, baseline diabetes, CCS angina class, and myocardium viability. Sensitivity Analysis Some QOL data were missing because a small proportion of living patients did not complete a scheduled QOL assessment and a larger proportion of patients died before 1 or more of the scheduled assessments could be done. Almost all cases of missing data among living patients were attributed to administrative reasons rather than their state of health (Supplement 1), which supports the assumption of missing at random and the use of multiple imputation techniques. Supplement 1. Data Collection in the Surgical Treatment of Ischemic Heart Disease (STICH) Trial Quality of Life Substudy The STICH trial had a high overall mortality rate and a treatment-related mortality difference, as noted previously. These deaths produced a nonrandom group of survivors who provided the follow-up QOL data. No consensus exists in the statistical literature about how best to analyze intermediate end points, such as QOL, when death prevents complete data collection. The primary concern this issue raises in a treatment comparison is that a therapy more effective at preventing death may also preserve more patients with poor QOL than the comparison therapy. Thus, when comparing QOL in such a situation using all survivors, the therapy with the worse survival might have an artifactual improvement in QOL measurements. Rubin has proposed that because QOL data that are censored due to death do not exist even in theory and should be regarded as undefined, the most meaningful analysis is to compare QOL for patients in the 2 treatment groups who would have survived with either therapy (18). To accomplish this, we performed a survival average causal effect (SACE) analysis (1821). The SACE estimates of QOL are calculated from weighted averages of the QOL data multiplied by survival probability estimates (from survival models developed in the parent STICH study cohort) specific to the study group. The 95% CIs for the SACE were calculated using 200 repetitions of a nonparametric bootstrap procedure (22). These sensitivity analyses are helpful as supplements to the primary analysis but have their own difficulties, not the least of which are the important but untestable assumptioClinical trials performed during the 1970s and 1980s defined several major coronary artery disease (CAD) subgroups for which coronary artery bypass grafting (CABG) provided incremental survival, angina relief, or both relative to medical therapy, which formed the foundation for current practice patterns and guideline recommendations on the use of CABG (1-3). However, patients with severe left ventricular dysfunction (ejection fraction ≤0.35) were not represented in these early trials. Thus, management decisions for these patients have largely relied on clinical judgment to extrapolate from those trials and a small group of observational studies (4, 5). The challenges in using this evidence to select treatment for contemporary patients is further complicated in that medical therapies for both CAD and heart failure have improved substantially over those used in the earlier trials. The STICH (Surgical Treatment for Ischemic Heart Failure) trial was funded by the National Heart, Lung, and Blood Institute in 2002 to provide a comprehensive evaluation of the incremental therapeutic benefits of routine CABG over contemporary guideline-based medical therapy in patients with severe systolic dysfunction due to CAD (6). A major prespecified secondary end point of the trial was health-related quality of life (QOL), which is an outcome that complements the major clinical end points by assessing the patient’s experience of, and satisfaction with, the 2 therapeutic strategies compared (7, 8).

Similar secondary stroke prevention and medication persistence rates among rural and urban patients.

BACKGROUND Although Staphylococcus aureus bacteremia is a common, serious infection, accurately identifying febrile patients with this diagnosis at the time of initial evaluation is difficult. The purpose of this investigation was to define clinical characteristics present at the time of the initial recognition of fever that were associated with the presence of any bloodstream infection and, in particular, with S. aureus bacteremia. METHODS All patients > or =18 years of age with a new episode of health care-associated fever (temperature > or =38 degrees C) and at least one blood culture drawn were eligible for enrollment into this prospective multicenter cohort study. Multivariable analyses were conducted and internally validated scoring systems were developed to categorize the risk of bacteremia. RESULTS Of 1015 patients enrolled, 181 patients (17.8%) had clinically significant bacteremia, including 77 patients (7.6%) with S. aureus bacteremia. Clinical characteristics associated with S. aureus bacteremia were the presence of a hemodialysis graft or shunt (odds ratio [OR] 3.22; 95% confidence interval [CI], 1.85-5.61), chills (OR 2.38; 95% CI, 1.43-3.98), and a history of S. aureus infection (OR 2.68; 95% CI, 1.38-5.20). Peripheral vascular catheters were inversely associated with S. aureus bacteremia (OR 0.42; 95% CI, 0.26-0.69). Clinical characteristics associated with any bloodstream infection were central venous access, chills, history of S. aureus infection, and hemodialysis access. CONCLUSIONS Among patients with health care-associated fever, the presence of easily recognizable clinical characteristics at the time of obtaining the initial blood cultures can help to identify patients at increased risk for any bloodstream infection, in particular for S. aureus bacteremia.

An evaluation of stroke education in avail registry hospitals

PURPOSE Rural residents are less likely to obtain optimal care for many serious conditions and have poorer health outcomes than those residing in more urban areas. We determined whether rural vs urban residence affected postdischarge medication persistence and 1 year outcomes after stroke. METHODS The Adherence eValuation After Ischemic Stroke-Longitudinal (AVAIL) study is a multicenter registry of stroke patients enrolled in 101 hospitals nationwide. Medications were recorded at hospital discharge and again after 3 and 12 months. Persistence was defined as continuation of prescribed discharge medications. Participants were categorized as living in rural or urban settings by cross-referencing home ZIP code with metropolitan statistical area (MSA) designation. FINDINGS Rural patients were younger, more likely to be white, married, smokers, and less likely to be college graduates. There was no difference in stroke type or working status compared to urban patients, and there were minor differences in comorbid conditions. There were no differences based on rural vs urban residence in medication persistence at 3 or 12 months postdischarge and no differences in outcomes of recurrent stroke or rehospitalization at 12 months. CONCLUSION Despite differences in patient characteristics, there was no difference in medication persistence or outcomes between rural and urban dwellers after hospitalization for ischemic stroke or transient ischemic attack (TIA).

Quality-of-Life Outcomes With Coronary Artery Bypass Graft Surgery in Ischemic Left Ventricular DysfunctionA Randomized TrialQuality-of-Life Outcomes in Surgical Treatment of Ischemic Heart Failure

ABSTRACT The purpose of this study is to explore factors associated with recall of medication education and satisfaction with healthcare provider communication in patients with acute stroke or transient ischemic attack. This is an analysis of data from the AVAIL (Adherence Evaluation of Acute Ischemic Stroke Longitudinal) study. At 3 months after discharge, 2,219 stroke patients from 99 sites were interviewed and asked about their perceptions of education and communication with their healthcare providers as well as their current medication use and knowledge. Results show that less than 2% of the respondents reported not understanding how to take their medications, 4% did not know how to refill their medications, and 5% did not know the reason they were taking them. A vast majority (92%) of participants reported high levels of satisfaction in their communications with healthcare providers after discharge. Although overall understanding and satisfaction was high, older subjects were less likely to recall receiving medication information at discharge or to understand their medications. Similarly, African Americans and patients discharged from an academic hospital were less likely to report receiving a written medication list. This report highlights the success of education efforts and potential areas for additional improvement.

Quality-of-Life Outcomes in Surgical Treatment of Ischemic Heart Failure Quality-of-Life Outcomes With Coronary Artery Bypass Graft Surgery in Ischemic Left Ventricular Dysfunction: A Randomized Trial

Context A randomized, controlled trial of patients with high-risk coronary artery disease and heart failure previously reported no significant difference between medical therapy alone and medical therapy plus coronary artery bypass grafting when the outcome was death from any cause. Contribution This study from the same trial reports that medical therapy plus coronary artery bypass grafting is better than medical therapy alone when the outcome is quality of life. Caution The type of therapy was not concealed from patients or investigators. Implication This is the first study to examine how these therapies affect quality of life in patients who have coronary artery disease and heart failure. The Editors Clinical trials performed during the 1970s and 1980s defined several major coronary artery disease (CAD) subgroups for which coronary artery bypass grafting (CABG) provided incremental survival, angina relief, or both relative to medical therapy. These early trials formed the foundation for current practice patterns and guideline recommendations on the use of CABG (13). However, patients with severe left ventricular dysfunction (ejection fraction 0.35) were not represented. Thus, management decisions for these patients have largely relied on clinical judgment to extrapolate from those trials and a small group of observational studies(4, 5). The challenges in using this evidence to select treatment for contemporary patients is further complicated by the substantial improvement in medical therapies for both CAD and heart failure over those used in the earlier trials. The STICH (Surgical Treatment for Ischemic Heart Failure) trial was funded by the National Heart, Lung, and Blood Institute in 2002 to provide a comprehensive evaluation of the incremental therapeutic benefits of routine CABG over contemporary guideline-based medical therapy in patients with severe systolic dysfunction due to CAD (6). A major prespecified secondary end point of the trial was health-related quality of life (QOL), which is an outcome that complements the major clinical end points by assessing the patients experience of, and satisfaction with, the 2 therapeutic strategies compared (7, 8). Methods Patient Population and Primary Clinical Results To test the STICH trials surgical revascularization hypothesis, we randomly assigned 1212 patients with site-defined left ventricular ejection fraction of 0.35 or less and CAD suitable for revascularization to CABG or medical therapy (6). Rationale, trial design, and complete inclusion and exclusion criteria have been described previously (7). Patients were enrolled at 99 clinical sites in 22 countries between July 2002 and May 2007. All patients provided informed consent, and study protocol was approved by each sites institutional review board or ethics committee. Median follow-up was 56 months. The primary intention-to-treat comparison showed that 35.7% of patients assigned to CABG and 40.5% of those assigned to medical therapy died (primary analysis: unadjusted hazard ratio for all-cause mortality, 0.86 [95% CI, 0.72 to 1.04]; P= 0.123; secondary analysis: adjusted hazard ratio for all-cause mortality, 0.82 [CI, 0.68 to 0.99]; P= 0.039). Patients assigned to CABG had lower rates of the 2 major secondary clinical end points: death from cardiovascular causes (hazard ratio, 0.81; P= 0.050) and the composite of all-cause mortality and hospitalization for cardiovascular causes (hazard ratio, 0.74; P< 0.001). Health-Related QOL Data Collection We collected QOL data using structured interviews at baseline and 4, 12, 24, and 36 months after randomization. Site coordinators were specially trained by the Duke Clinical Research Institute Outcomes Research Group to conduct baseline interviews. The original research plan called for all patients to be enrolled in North America and all English- and Spanish-language follow-up QOL interviews to be completed via telephone by the Duke Clinical Research Institute. The few patients expected to require French-language interviews were to be interviewed by site coordinators. When enrollment was expanded outside of North America, the plan was modified to have those site coordinators do all QOL interviews in the patients native language. For nonEnglish-speaking participants, translations of the QOL instruments were obtained from the instrument developers or a translation service was used to create validated translations. The New York Heart Association (NYHA) class and Canadian Cardiovascular Society (CCS) angina class were collected on the clinical case report form at baseline and each follow-up interval. QOL Measures A battery of validated measures was used to provide a comprehensive but efficient assessment of QOL. The principal prespecified QOL measure was the Kansas City Cardiomyopathy Questionnaire (KCCQ), which is a 23-item instrument that measures the effect of heart failure symptoms on QOL (9). We used 3 scales from the Seattle Angina Questionnaire to assess the effect of angina symptoms on QOL outcomes (10), the Short Form-12 Survey to provide a brief overall generic measure of health status (11), and 5 individual scales from the Short Form-36 Health Survey to provide a more detailed assessment of areas of functioning and well-being from a generic perspective (12). Depressive symptoms were assessed using the Center for Epidemiologic Studies Depression Scale (13). The Cardiac Self-Efficacy Questionnaire was used to measure patient confidence in controlling disease symptoms and maintaining physical functioning (14). Finally, we used the EuroQol-5D, which is a generic instrument consisting of a 5-dimension single summary health status index and a self-rated visual analogue scale (0 to 100) of current health-related QOL (15). Statistical Analysis All primary comparisons were performed with the treatment group defined according to the intention-to-treat principle. Descriptive statistics included percentages for discrete variables and medians with 25th to 75th percentiles plus means with SDs for continuous variables. The chi-square test was used for discrete variable comparisons. Treatment comparisons of continuous variables were done by using a linear mixed model to account for repeated measures within a patient. The baseline QOL measure was used as one of the repeated measures. In PROC MIXED in SAS, version 9.2 (SAS Institute), the baseline, 4-, 12-, 24-, and 36-month measurements within a patient were fitted using maximum likelihood methods with unstructured covariance matrix (16). At each time point, estimated treatment differences, 95% CIs, and P values were obtained using the model estimates. Finally, the proportion of patients who achieved a clinically important improvement of 5 points or more in KCCQ overall summary scores were compared by treatment group using the chi-square test (17). All analyses were conducted using SAS, version 8 or higher. All reported P values were 2-sided. No adjustment was made in significance levels for multiple comparisons. Benchmarks for clinically significant changes in QOL scores for individual patients were used informally to assess the magnitude of the mean difference between the 2 groups. However, a QOL difference between groups at or exceeding the benchmark was not used as a formal decision rule to define clinical significance. Subgroup Analysis Regional effects were tested as interactions between the 5 regions used in the clinical analyses (Asia, Australia and New Zealand, Europe, North America, and South America) and the KCCQ overall summary score. Other interactions tested from those prespecified in the clinical analysis plan included age, sex, race, current NYHA class, left ventricular ejection fraction, baseline diabetes, CCS angina class, and myocardium viability. Sensitivity Analysis Some QOL data were missing because a small proportion of living patients did not complete a scheduled QOL assessment and a larger proportion of patients died before 1 or more of the scheduled assessments could be done. Almost all cases of missing data among living patients were attributed to administrative reasons rather than their state of health (Supplement 1), which supports the assumption of missing at random and the use of multiple imputation techniques. Supplement 1. Data Collection in the Surgical Treatment of Ischemic Heart Disease (STICH) Trial Quality of Life Substudy The STICH trial had a high overall mortality rate and a treatment-related mortality difference, as noted previously. These deaths produced a nonrandom group of survivors who provided the follow-up QOL data. No consensus exists in the statistical literature about how best to analyze intermediate end points, such as QOL, when death prevents complete data collection. The primary concern this issue raises in a treatment comparison is that a therapy more effective at preventing death may also preserve more patients with poor QOL than the comparison therapy. Thus, when comparing QOL in such a situation using all survivors, the therapy with the worse survival might have an artifactual improvement in QOL measurements. Rubin has proposed that because QOL data that are censored due to death do not exist even in theory and should be regarded as undefined, the most meaningful analysis is to compare QOL for patients in the 2 treatment groups who would have survived with either therapy (18). To accomplish this, we performed a survival average causal effect (SACE) analysis (1821). The SACE estimates of QOL are calculated from weighted averages of the QOL data multiplied by survival probability estimates (from survival models developed in the parent STICH study cohort) specific to the study group. The 95% CIs for the SACE were calculated using 200 repetitions of a nonparametric bootstrap procedure (22). These sensitivity analyses are helpful as supplements to the primary analysis but have their own difficulties, not the least of which are the important but untestable assumptioClinical trials performed during the 1970s and 1980s defined several major coronary artery disease (CAD) subgroups for which coronary artery bypass grafting (CABG) provided incremental survival, angina relief, or both relative to medical therapy, which formed the foundation for current practice patterns and guideline recommendations on the use of CABG (1-3). However, patients with severe left ventricular dysfunction (ejection fraction ≤0.35) were not represented in these early trials. Thus, management decisions for these patients have largely relied on clinical judgment to extrapolate from those trials and a small group of observational studies (4, 5). The challenges in using this evidence to select treatment for contemporary patients is further complicated in that medical therapies for both CAD and heart failure have improved substantially over those used in the earlier trials. The STICH (Surgical Treatment for Ischemic Heart Failure) trial was funded by the National Heart, Lung, and Blood Institute in 2002 to provide a comprehensive evaluation of the incremental therapeutic benefits of routine CABG over contemporary guideline-based medical therapy in patients with severe systolic dysfunction due to CAD (6). A major prespecified secondary end point of the trial was health-related quality of life (QOL), which is an outcome that complements the major clinical end points by assessing the patient’s experience of, and satisfaction with, the 2 therapeutic strategies compared (7, 8).

Surgical Treatment of Ischemic Heart Failure Trial

Context A randomized, controlled trial of patients with high-risk coronary artery disease and heart failure previously reported no significant difference between medical therapy alone and medical therapy plus coronary artery bypass grafting when the outcome was death from any cause. Contribution This study from the same trial reports that medical therapy plus coronary artery bypass grafting is better than medical therapy alone when the outcome is quality of life. Caution The type of therapy was not concealed from patients or investigators. Implication This is the first study to examine how these therapies affect quality of life in patients who have coronary artery disease and heart failure. The Editors Clinical trials performed during the 1970s and 1980s defined several major coronary artery disease (CAD) subgroups for which coronary artery bypass grafting (CABG) provided incremental survival, angina relief, or both relative to medical therapy. These early trials formed the foundation for current practice patterns and guideline recommendations on the use of CABG (13). However, patients with severe left ventricular dysfunction (ejection fraction 0.35) were not represented. Thus, management decisions for these patients have largely relied on clinical judgment to extrapolate from those trials and a small group of observational studies(4, 5). The challenges in using this evidence to select treatment for contemporary patients is further complicated by the substantial improvement in medical therapies for both CAD and heart failure over those used in the earlier trials. The STICH (Surgical Treatment for Ischemic Heart Failure) trial was funded by the National Heart, Lung, and Blood Institute in 2002 to provide a comprehensive evaluation of the incremental therapeutic benefits of routine CABG over contemporary guideline-based medical therapy in patients with severe systolic dysfunction due to CAD (6). A major prespecified secondary end point of the trial was health-related quality of life (QOL), which is an outcome that complements the major clinical end points by assessing the patients experience of, and satisfaction with, the 2 therapeutic strategies compared (7, 8). Methods Patient Population and Primary Clinical Results To test the STICH trials surgical revascularization hypothesis, we randomly assigned 1212 patients with site-defined left ventricular ejection fraction of 0.35 or less and CAD suitable for revascularization to CABG or medical therapy (6). Rationale, trial design, and complete inclusion and exclusion criteria have been described previously (7). Patients were enrolled at 99 clinical sites in 22 countries between July 2002 and May 2007. All patients provided informed consent, and study protocol was approved by each sites institutional review board or ethics committee. Median follow-up was 56 months. The primary intention-to-treat comparison showed that 35.7% of patients assigned to CABG and 40.5% of those assigned to medical therapy died (primary analysis: unadjusted hazard ratio for all-cause mortality, 0.86 [95% CI, 0.72 to 1.04]; P= 0.123; secondary analysis: adjusted hazard ratio for all-cause mortality, 0.82 [CI, 0.68 to 0.99]; P= 0.039). Patients assigned to CABG had lower rates of the 2 major secondary clinical end points: death from cardiovascular causes (hazard ratio, 0.81; P= 0.050) and the composite of all-cause mortality and hospitalization for cardiovascular causes (hazard ratio, 0.74; P< 0.001). Health-Related QOL Data Collection We collected QOL data using structured interviews at baseline and 4, 12, 24, and 36 months after randomization. Site coordinators were specially trained by the Duke Clinical Research Institute Outcomes Research Group to conduct baseline interviews. The original research plan called for all patients to be enrolled in North America and all English- and Spanish-language follow-up QOL interviews to be completed via telephone by the Duke Clinical Research Institute. The few patients expected to require French-language interviews were to be interviewed by site coordinators. When enrollment was expanded outside of North America, the plan was modified to have those site coordinators do all QOL interviews in the patients native language. For nonEnglish-speaking participants, translations of the QOL instruments were obtained from the instrument developers or a translation service was used to create validated translations. The New York Heart Association (NYHA) class and Canadian Cardiovascular Society (CCS) angina class were collected on the clinical case report form at baseline and each follow-up interval. QOL Measures A battery of validated measures was used to provide a comprehensive but efficient assessment of QOL. The principal prespecified QOL measure was the Kansas City Cardiomyopathy Questionnaire (KCCQ), which is a 23-item instrument that measures the effect of heart failure symptoms on QOL (9). We used 3 scales from the Seattle Angina Questionnaire to assess the effect of angina symptoms on QOL outcomes (10), the Short Form-12 Survey to provide a brief overall generic measure of health status (11), and 5 individual scales from the Short Form-36 Health Survey to provide a more detailed assessment of areas of functioning and well-being from a generic perspective (12). Depressive symptoms were assessed using the Center for Epidemiologic Studies Depression Scale (13). The Cardiac Self-Efficacy Questionnaire was used to measure patient confidence in controlling disease symptoms and maintaining physical functioning (14). Finally, we used the EuroQol-5D, which is a generic instrument consisting of a 5-dimension single summary health status index and a self-rated visual analogue scale (0 to 100) of current health-related QOL (15). Statistical Analysis All primary comparisons were performed with the treatment group defined according to the intention-to-treat principle. Descriptive statistics included percentages for discrete variables and medians with 25th to 75th percentiles plus means with SDs for continuous variables. The chi-square test was used for discrete variable comparisons. Treatment comparisons of continuous variables were done by using a linear mixed model to account for repeated measures within a patient. The baseline QOL measure was used as one of the repeated measures. In PROC MIXED in SAS, version 9.2 (SAS Institute), the baseline, 4-, 12-, 24-, and 36-month measurements within a patient were fitted using maximum likelihood methods with unstructured covariance matrix (16). At each time point, estimated treatment differences, 95% CIs, and P values were obtained using the model estimates. Finally, the proportion of patients who achieved a clinically important improvement of 5 points or more in KCCQ overall summary scores were compared by treatment group using the chi-square test (17). All analyses were conducted using SAS, version 8 or higher. All reported P values were 2-sided. No adjustment was made in significance levels for multiple comparisons. Benchmarks for clinically significant changes in QOL scores for individual patients were used informally to assess the magnitude of the mean difference between the 2 groups. However, a QOL difference between groups at or exceeding the benchmark was not used as a formal decision rule to define clinical significance. Subgroup Analysis Regional effects were tested as interactions between the 5 regions used in the clinical analyses (Asia, Australia and New Zealand, Europe, North America, and South America) and the KCCQ overall summary score. Other interactions tested from those prespecified in the clinical analysis plan included age, sex, race, current NYHA class, left ventricular ejection fraction, baseline diabetes, CCS angina class, and myocardium viability. Sensitivity Analysis Some QOL data were missing because a small proportion of living patients did not complete a scheduled QOL assessment and a larger proportion of patients died before 1 or more of the scheduled assessments could be done. Almost all cases of missing data among living patients were attributed to administrative reasons rather than their state of health (Supplement 1), which supports the assumption of missing at random and the use of multiple imputation techniques. Supplement 1. Data Collection in the Surgical Treatment of Ischemic Heart Disease (STICH) Trial Quality of Life Substudy The STICH trial had a high overall mortality rate and a treatment-related mortality difference, as noted previously. These deaths produced a nonrandom group of survivors who provided the follow-up QOL data. No consensus exists in the statistical literature about how best to analyze intermediate end points, such as QOL, when death prevents complete data collection. The primary concern this issue raises in a treatment comparison is that a therapy more effective at preventing death may also preserve more patients with poor QOL than the comparison therapy. Thus, when comparing QOL in such a situation using all survivors, the therapy with the worse survival might have an artifactual improvement in QOL measurements. Rubin has proposed that because QOL data that are censored due to death do not exist even in theory and should be regarded as undefined, the most meaningful analysis is to compare QOL for patients in the 2 treatment groups who would have survived with either therapy (18). To accomplish this, we performed a survival average causal effect (SACE) analysis (1821). The SACE estimates of QOL are calculated from weighted averages of the QOL data multiplied by survival probability estimates (from survival models developed in the parent STICH study cohort) specific to the study group. The 95% CIs for the SACE were calculated using 200 repetitions of a nonparametric bootstrap procedure (22). These sensitivity analyses are helpful as supplements to the primary analysis but have their own difficulties, not the least of which are the important but untestable assumptioClinical trials performed during the 1970s and 1980s defined several major coronary artery disease (CAD) subgroups for which coronary artery bypass grafting (CABG) provided incremental survival, angina relief, or both relative to medical therapy, which formed the foundation for current practice patterns and guideline recommendations on the use of CABG (1-3). However, patients with severe left ventricular dysfunction (ejection fraction ≤0.35) were not represented in these early trials. Thus, management decisions for these patients have largely relied on clinical judgment to extrapolate from those trials and a small group of observational studies (4, 5). The challenges in using this evidence to select treatment for contemporary patients is further complicated in that medical therapies for both CAD and heart failure have improved substantially over those used in the earlier trials. The STICH (Surgical Treatment for Ischemic Heart Failure) trial was funded by the National Heart, Lung, and Blood Institute in 2002 to provide a comprehensive evaluation of the incremental therapeutic benefits of routine CABG over contemporary guideline-based medical therapy in patients with severe systolic dysfunction due to CAD (6). A major prespecified secondary end point of the trial was health-related quality of life (QOL), which is an outcome that complements the major clinical end points by assessing the patient’s experience of, and satisfaction with, the 2 therapeutic strategies compared (7, 8).