Exercise-based cardiac rehabilitation for chronic heart failure: the EXTRAMATCH II individual participant data meta-analysis

Background: Current national and international guidelines on the management of heart failure (HF) recommend exercise-based cardiac rehabilitation (ExCR), but do not differentiate this recommendation according to patient subgroups. Objectives: (1) To obtain definitive estimates of the impact of ExCR interventions compared with no exercise intervention (control) on mortality, hospitalisation, exercise capacity and health-related quality of life (HRQoL) in HF patients; (2) to determine the differential (subgroup) effects of ExCR in HF patients according to their age, sex, left ventricular ejection fraction, HF aetiology, New York Heart Association class and baseline exercise capacity; and (3) to assess whether or not the change in exercise capacity mediates for the impact of the ExCR on final outcomes (mortality, hospitalisation and HRQoL), and determine if this is an acceptable surrogate end point. Design: This was an individual participant data (IPD) meta-analysis. Setting: An international literature review. Participants: HF patients in randomised controlled trials (RCTs) of ExCR. Interventions: ExCR for at least 3 weeks compared with a no-exercise control, with 6 months' follow-up. Main outcome measures: All-cause and HF-specific mortality, all-cause and HF-specific hospitalisation, exercise capacity and HRQoL. Data sources: IPD from eligible RCTs. Review methods: RCTs from the Exercise Training Meta-Analysis of Trials for Chronic Heart Failure (ExTraMATCH/ExTraMATCH II) IPD meta-analysis and a 2014 Cochrane systematic review of ExCR (Taylor RS, Sagar VA, Davies EJ, Briscoe S, Coats AJ, Dalal H, et al. Exercise-based rehabilitation for heart failure. Cochrane Database Syst Rev 2014;4:CD003331). Results: Out of the 23 eligible RCTs (4398 patients), 19 RCTs (3990 patients) contributed data to this IPD meta-analysis. There was a wide variation in exercise programme prescriptions across included studies. Compared with control, there was no statistically significant difference in pooled time-to-event estimates in favour of ExCR, although confidence intervals (CIs) were wide: all-cause mortality had a hazard ratio (HR) of 0.83 (95% CI 0.67 to 1.04); HF-related mortality had a HR of 0.84 (95% CI 0.49 to 1.46); all-cause hospitalisation had a HR of 0.90 (95% CI 0.76 to 1.06); and HF-related hospitalisation had a HR of 0.98 (95% CI 0.72 to 1.35). There was a statistically significant difference in favour of ExCR for exercise capacity and HRQoL. Compared with the control, improvements were seen in the 6-minute walk test (6MWT) (mean 21.0 m, 95% CI 1.57 to 40.4 m) and Minnesota Living with Heart Failure Questionnaire score (mean -5.94, 95% CI -1.0 to -10.9; lower scores indicate improved HRQoL) at 12 months' follow-up. No strong evidence for differential intervention effects across patient characteristics was found for any outcomes. Moderate to good levels of correlation (R-trial(2) > 50% and p > 0.50) between peak oxygen uptake (VO(2)peak) or the 6MWT with mortality and HRQoL were seen. The estimated surrogate threshold effect was an increase of 1.6 to 4.6 ml/kg/minute for VO(2)peak. Limitations: There was a lack of consistency in how included RCTs defined and collected the outcomes: it was not possible to obtain IPD from all includable trials for all outcomes and patient-level data on exercise adherence was not sought. Conclusions: In comparison with the no-exercise control, participation in ExCR improved the exercise and HRQoL in HF patients, but appeared to have no effect on their mortality or hospitalisation. No strong evidence was found of differential intervention effects of ExCR across patient characteristics. VO(2)peak and 6MWT may be suitable surrogate end points for the treatment effect of ExCR on mortality and HRQoL in HF. Future studies should aim to achieve a consensus on the definition of outcomes and promote reporting of a core set of HF data. The research team also seeks to extend current policies to encourage study authors to allow access to RCT data for the purpose of meta-analysis.