The prospective multi-centre cohort study aimed at answering the following research question: 18 months after phase II, do patients with an unfavourable risk profile who attended an additional phase III programme (IG, intervention group) have on average fewer risk factors outside the target range compared to patients without phase III (KG, control group)?

Methods: At baseline (T0) and 18 months thereafter (T1) patients and physicians were asked to complete a questionnaire to report on clinical parameters, anxiety/depression, health-related quality of life etc. The minimally detectable group difference (α= 0,05, power 80%, min. of 150 patients per group) was defined as a mean of 0,32 risk factors out of the target range between the groups. The primary hypothesis was tested for significance with a stratified Wilkoxon-Mann-Whitney U test. For stratification we used a propensity score. To compare numerical differences between IG and KG we used adjusted differences. We developed a prognostic model to estimate the bias on the primary effect due to potential differences in loss of follow up between the groups. The study protocol had been approved by the local ethics committees.

Results: 13 Austrian rehabilitation clinics recruited 395 patients (recruitment was stopped prematurely before reaching the planned sample size of 450 patients, because the patient inclusion had stagnated). 164 physicians' and 228 patients' questionnaires were returned at T1. Data of 76 IG and 81 KG patients were available for the analysis of the primary endpoint. Despite counter-measures (repeated mailing, phone contacts) the drop out was higher than the expected ~33%. Pseudo-randomisation (due to the pre-defined recruiting process) had been successful to a large extent: at T0 patients differed significantly in a few parameters only (e.g. slightly more patients with angina pectoris and higher motivation for phase III in IG). Three thirds of patients were male, the mean age was 62 years. The most frequent cardiovascular events prior to rehabilitation were St.p. acute coronary syndrome (IG:29%; KG:24%) and aortocoronary bypass operation (IG:18%; KG:24%). At T0 more than 90% of patients had not reached the 'physical fitness' goal (performance in exercise stress test) or the 'body weight' goal (BMI and/or abdominal girth), about two thirds failed to reach the 'blood lipids' goal (cholesterol and/ or triglycerides) and almost 50% the 'blood sugar' goal (fasting blood sugar and/or HbA1c).
At T1, the adjusted mean difference in the change of risk factors out of the target range was -0,06 [-0.7; 0.6] (non-significant; group differences at T0 adjusted by means of the propensity score). Roughly spoken, about one out of twenty IG patients (6 of 100) showed one risk factor out of the target range less at T1 than a KG patient with the same number of not-reached goals and the same stratification characteristics at T0. The stratified effect size Cliff's delta is 0,04[-0,18; 0.26] for the primary endpoint (the chance of a better result for IG patients than under control is 52%, the chance of a worse is 48%). In secondary endpoints, only the change in the exercise stress test (percental increase, stratified: IG +11.4% (±15.10) vs. KG +5.08% (± 14.87), p.strat.= 0.008; Cliff's Delta 0.30 [0.07;0.53]) and the frequency of inpatient re-rehabilitation showed a statistically significant group difference in favour of the intervention. The exploratory analysis (using the prognostic score) did not indicate that the loss of follow up influenced results to the disadvantage of the intervention.

Conclusions: Despite positive trends in some outcomes, the present study could not find any evidence that phase III cardiac rehabilitation results in an effect as defined as minimally detectable in the study protocol (even after an 80%-power conserving adjustment of the significance level, compensating for a smaller than planned sample size).