Aims and Hypotheses
HART was designed to determine whether a patient SM intervention significantly improved clinically important outcomes, relative to an education control, in patients with mild to moderate systolic and/or diastolic HF. Two clinically important composite outcomes served as the primary endpoints: (1) death or hospitalizations for HF; and (2) death or all-cause hospitalizations.
portrays the pathway by which the SM intervention was hypothesized to translate into benefit on the primary endpoints. The intervention provided training in SM skills and was, as such, hypothesized to improve self-efficacy at SM. Since belief about one’s self-efficacy at SM is a prerequisite for performance of SM behaviors,17
improvement in self-efficacy was hypothesized to result in improved adherence to recommended drug therapy and lifestyle changes. Since negative mood undermines adherence18
and may also have a direct effect on clinical endpoints,19
several SM skills were directed toward the management of psychosocial stress, resulting in the expectation that the intervention would improve psychosocial function. Improvement in adherence and psychosocial function were hypothesized to translate into improvement on the primary endpoints and on the secondary endpoints of progression of HF and quality of life.
HART was a single-site, multi-hospital behavioral randomized efficacy trial based at Rush University Medical Center in Chicago and including 9 other collaborating hospitals from the Chicago metropolitan area. The aim was to randomize approximately 900 patients to one of two treatment arms between October, 2001 and October, 2004. Patients underwent their respective HART treatment for one year and were then followed for an additional 1 to 2 years, depending upon the timing of recruitment, resulting in a total on-trial time of 2–3 years. All data collection ended in October, 2006.
Sample size was based upon the assumption that the SM intervention would produce a 25% reduction in the primary event rate, relative to the education control group, from results of prior SM trials.20,21
The base rate for the primary endpoint of death or hospitalization for HF in the education control group was assumed to be 15% per year, from rates observed in the treatment arms of drug trials with similar patients22
since these drugs would likely be the standard of care when HART ended. Dropouts and losses were estimated to be 15% over the duration of the trial and sample size was adjusted by approximately 3% to allow for interim analyses. To account for both of these factors, an inflation factor of 20% was incorporated into the sample size calculations. Assuming a 2-sided alpha of 0.05 and 80% power, this led to a sample size of 900 participants, evenly distributed between the two treatment arms.
Because HART was a behavioral trial, double-blinding is impossible. However, HART was a partially blinded trial. All staff, except for the senior investigators, and all participants were blinded to trial hypotheses by providing neutral names for the randomized treatment arms. The SM treatment was called “Skills Training” and the attention control was called “Enhanced Education.” All investigators and staff, except for 2 biostatisticians and the recruiting nurses, were blinded to the randomization status of the participant. Treatment teams within each trial arm had no contact with patients in the other arm.
Eligibility for HART was kept broad to maximize the generalizability of results. Because the HART intervention focused on the development of skills, not simply the correction of deficits, eligibility was open to all patients, regardless of their current success with adherence. Eligible patients had HF for not less than the prior 3 months defined as either: (1) left ventricular ejection fraction ≤40% by echocardiography, radiographic ventriculography, or radionuclide ventriculography; or (2) diuretic therapy for at least 3 months and 1 previous hospitalization for HF.
Exclusions were factors that would jeopardize the conduct or rigor of the trial. These included: (1) patients for whom the 12-month prognosis was uncertain (i.e., NYHA Class IV, likelihood of cardiac transplant over the next year, symptomatic or sustained ventricular tachycardia not controlled by therapy within the last 3 months, or other illnesses or disorders that limit 12-month survival); (2) patients classified as NYHA Class I who were unlikely to have a primary endpoint over the course of the trial; (3) patients who were unlikely to undergo or benefit from the behavioral treatment (i.e., presence of cognitive dysfunction or psychological co-morbidity such as substance abuse, psychotic disorder, or active suicidal ideation); (4) patients whose symptoms may be eliminated by surgery (e.g., severe aortic stenosis); (5) logistical issues (e.g., patients were already enrolled in a conflicting protocol, patients were non-English speaking); (6) patients whose physicians refused access; (7) patients who indicated that they were unwilling to make lifestyle changes now or in the near future; and (8) patients who had unstable angina, MI, CABG, or PTCA within the last month (this was a temporary exclusion).
shows the number of screened patients who were excluded for each of these reasons. Of all patients screened, 988 (31%) excluded themselves due to logistical problems, being unwilling to join the trial, or being unwilling to make lifestyle changes.
Patient flow from screening to enrollment
Nine recruiting hospitals were chosen because they had a substantial number of patients, had potential to promote diversity in gender and ethnicity, and were located in one of 4 geographic clusters, thus making patient travel to a group treatment feasible. Each collaborating hospital had a local cardiologist in place who served as the local principal investigator of the trial. In general, 3 recruiting strategies were used: inpatient screening, outpatient screening at ongoing clinics, and referrals from local cardiologists and internists. Each local principal investigator could tailor the recruitment approach to the features of his/her setting. In the early phase of recruitment, direct marketing to community volunteers was conducted using mass media but this strategy was abandoned due to unfavorable cost-effectiveness.
From pilot data, the screening to enrollment ratio was expected to be 4 patients screened for every patient enrolled, necessitating a recruitment period of between 2–3 years to obtain the target of 900 patients. shows that Rush and the 9 recruiting hospitals screened 3,154 patients over 3 years to produce the 902 patients enrolled in HART. Thus, 3.5 patients were screened for every patient enrolled.
presents the final cohort of 902 patients by geographic area, age, gender, ethnicity, and HF type. The average age of the HART participant was 63.6 years. Approximately half were women, 40% were minority and 23% had diastolic dysfunction, reflecting the diversity in HF seen in clinical practice. The achievement of this diversity was due to the choice of recruiting hospitals. For example, patients from the 3 central Chicago sites were younger (average age = 58.0 years) and more often ethnic minorities (80.2%) than patients from the northern suburbs (average age = 69.2 years; 11.0% minority).
Recruitment by geographic area, age, gender, ethnicity, and heart failure type
Once a potential participant was identified, a HART nurse coordinator obtained permission from the physician to examine medical records and screen for the presence of exclusion criteria. If the patient was eligible, he/she was contacted and the trial was described. If the patient was interested in participating, a baseline exam was scheduled at the local recruiting hospital. Upon arrival at the baseline exam, the trial was described again, eligibility was confirmed, informed consent was obtained, and the baseline exam was conducted.
At the conclusion of the baseline exam, the patient was provided with instructions for receiving the results of the randomization, beginning the treatment, and upcoming exams and phone calls. The nurse coordinator called the automated randomization service (Moffitt Cancer Center, University of South Florida) to obtain treatment assignment and then mailed a letter of notification to the patient. Follow-up by staff leading the relevant trial arm commenced approximately 2–5 days after receipt of the letter.
Baseline and all annual exams consisted of: (1) a clinical exam assessing height, weight, pulse, respiratory rate, blood pressure, 6-Minute Walk, and a blood draw; (2) a medical history including current medical conditions, current medications, sociodemographic status, and risk factors such as smoking, alcohol consumption, and physical activity; (3) an interview for the assessment of HF symptoms, health-related quality of life, problems with HF management, and depression; and (4) a questionnaire assessing satisfaction with quality of life, self-efficacy, social support, hopelessness, irritability, purpose in life, and salt intake. At the conclusion of each exam, the patient was asked to put a month’s supply of an ace inhibitor (or beta blocker if the patient was not taking an ace inhibitor) into a MEM’s pill cap container (MEMS V Trackcap, AARDEX, Zug, Switzerland) and was taught how to use it for the ensuing month.
The trial protocol was approved by the institutional committees on human research at the central coordinating center and at each of the 9 collaborating hospitals.
The randomization protocol was developed to insure that 10 patients from the same geographic location would be randomized as quickly as possible to a SM group. A stratified block design was used with strata defined as 4 geographic locations in metropolitan Chicago. Within each stratum, a block size of 20 was used to insure that after 20 randomizations, 10 patients were assigned to each treatment arm.
SM treatment featured group-based counseling to help patients develop mastery in problem-solving and in 5 SM skills deemed to be helpful in adhering to medical advice and alleviating any negative affect that would undermine adherence. Group, rather than individual, treatment was chosen because modeling by coping peers, and the resulting vicarious learning, has been shown to be a powerful agent of change.23
Eighteen 2-hour group meetings of 5–10 patients were spread out over the course of one year. At each group meeting, patients received HF education in the form of 18 one-page Heart Failure Tip Sheets from the American Heart Association which summarized basic elements of HF management including medication adherence, sudden weight gain, salt restriction, moderate physical activity, and stress management.
Implementation of these health education tips was accomplished through the mastery of 5 SM skills, chosen because of their relevance for the lifestyle changes required. These skills were: (1) self-monitoring of such targets as weight and daily sodium intake; (2) environmental restructuring of the home and workplace by, for example, placing pill boxes in visible places to remind patients to take them; (3) social support where patients were encouraged to discuss efforts at making lifestyle changes with family and friends as a way to elicit their support; (4) cognitive restructuring where patients were taught to replace stress inducing thoughts with stress reducing ones; and (5) deep breathing as an immediate response to physical or emotional stress. To foster proactivity in invoking these skills when needed, a problem-solving format was used in which patients identified barriers to implementing the HF tips and were encouraged to use the SM skills to overcome them.
All groups were led by health professionals with advanced degrees, experience in conducting groups, and a willingness to follow a protocol. All prospective group leaders underwent standardized training over the course of 2 days. Only those certified at the end of this training were invited to lead. All treatment sessions were taped and randomly selected tapes were reviewed by the supervisor (KJF) to check adherence to the protocol and provide feedback to the leader, as needed. Data reports focusing both on group leaders and individual patients were reviewed by a SM Subcommittee to identify problems and strategies for their resolution. To prevent drift, group leaders met monthly to share successes and strategies for overcoming challenges. These meetings were conducted using the same format as that used in the SM groups for the purpose of modeling desired behaviors.
The use of an attention control, rather than a usual care control, minimizes potential problems that could occur in a behavioral trial with an unblinded usual care control. Among these problems are: (1) the inability to determine whether it is any attention, rather than the specific treatment, that promotes benefit; (2) the inability to mask patients to trial hypotheses; (3) the potential to produce an “underdog” effect that would promote differential dropout in the controls; (4) the potential to produce a “John Henry” effect in which the controls seek treatment on their own and thus essentially cross-over into treatment; and (5) the minimization of any generalized placebo effect where patients in the treated group get better simply because they perceive they are receiving treatment.
Patients randomized to the education control received a phone-based educational intervention which was believed to be the standard of care that would be in place when HART was concluded. Patients received the same 18 American Heart Association Heart Failure Tip Sheets, on the same schedule as the group meetings in the SM group. However, they were mailed to the home and, to insure receipt and check comprehension, a study coordinator placed a phone call within 2–3 days of receipt. If at the time the first call was placed the patient had not read the Tip Sheet, another call was scheduled. Questions about the Tip Sheets were answered but if questions were asked about SM or non-Tip Sheet concerns, the patient was referred back to his/her provider. No new information beyond the Tip Sheet was provided.
Study coordinators conducting the phone calls were trained on purpose, structure, content, data reporting responsibilities, and quality control procedures. Training included role playing to simulate phone call interactions. Ongoing training, as needed, took place in response to special problems (e.g., difficulty in reaching patients, patients asking for SM support).
summarizes the primary and secondary endpoints in HART. Patients or, in the case of death, their family members were queried every 3 months for the occurrence of death or hospitalizations. All secondary endpoints were assessed annually during the physical exam, interview, and self-reported questionnaire. Mortality was confirmed by medical record, death certificate, emergency medical services record, or queries from the national mortality registry. HF admissions were adjudicated by the presence of either shortness of breath, peripheral edema, or chest x-ray evidence of pulmonary edema without evidence of another disease process accounting for symptoms or signs. HF admissions were confirmed if the patient responded to anti-failure therapy or had a documented decrease in LV function.
Description of primary and secondary endpoints
The safety of the patients was monitored continuously by an In-House Safety Committee. Potential adverse events that could be related to the nature or content of the interventions included suicidal attempts, suicidal ideation, a new or recurrent episode of depression or anxiety, substance abuse or recidivism, a report of unusually high levels of distress, depression, or anxiety not meeting diagnostic criteria, and physical injury. Potential serious adverse events that could be directly related to the nature or content of the interventions included death, life-threatening adverse drug/device/product events, inpatient hospitalization or prolonged hospitalization, and persistent or significant disability/incapacity.
The HART cohort was 40% African American and 36% had incomes <$20,000/year. Many of these patients had low literacy levels and functional limitations. To insure that this trial was culturally sensitive to the special barriers that occur in these subgroups, all aspects of the recruitment protocol and treatment arms were studied to determine patient perceptions of the entire research process.40
This resulted in several important protocol modifications. For example, for underserved African Americans, the development of trust came more slowly than in other subgroups, translating into missed SM group meetings early in treatment. Thus, the SM protocol was modified to permit make-up sessions later in the year after trust had developed. For low income patients, reimbursement for transportation was arranged. For low literacy patients, Literacy Chicago
, a non-profit organization dedicated to improving literacy in the underserved, was employed to improve the readability and comprehensibility of all treatment and outcome measures. Offers were made to read consent forms and questionnaires. For patients with functional limitations, all examination and treatment sites were evaluated for their ability to accommodate wheelchairs, consistent with recommendations from the Americans with Disabilities Act.
Trial results will be analyzed according to the intention-to-treat principle. Data from all randomized patients will be analyzed according to the original treatment assignment, regardless of the actual dose of treatment received or possible cross-over between treatment arms. Baseline characteristics will be compared between the 2 treatment arms to assess covariate balance and any imbalances will be adjusted for in multivariate models. Event rates over time will be summarized using Kaplan-Meier survival curves and differences in these curves by treatment arm will be analyzed using the Mantel-Haenszel (logrank) test. The semi-parametric Cox model will be used for multivariate modeling. In the event that the proportionality assumption is not met, treatment differences will be analyzed using the Wilcoxon test and multivariate modeling will be analyzed using logistic regression.
Differential dependency in the HART data exists because the SM intervention was delivered in groups whereas the education control was delivered individually. To assess the significance of this dependency, we will use hierarchical linear models to analyze endpoint data in the SM participants. If group assignment and/or any relevant interactions are found to be significant covariates, they will be incorporated as covariates in subsequent analyses of the trial endpoints.
Treatment differences in secondary endpoints will be examined using t-tests as the primary analytic tool. Change in secondary outcomes will be analyzed using generalized linear models (GLM) which will take into account the full richness of the data obtained via annual assessments. Analyses of treatment efficacy are planned for prespecified subgroups which include age, gender, income, education, ethnicity, functional capacity, systolic/diastolic dysfunction, and adherence. For these analyses, interaction terms (subgroup × treatment) will be evaluated in multivariate modeling that includes the interaction, its main effects, and sociodemographic and medical covariates.