A Meta-analysis of the effects of Exercise Training on Left Ventricular Remodeling Following Myocardial Infarction: Start early and go longer for greatest exercise benefits on remodeling

A searched was undertaken of OVID MEDLINE (1950 to 2009), Cochrane Central Register of Controlled Trials (1991 to 2009), AMED (1985 to 2009), EMBASE (1988 to 2009), PUBMED (1966 to 2009), SPORT DISCUS (1975 to 2009), SCOPUS (1950 to 2009) and WEB OF SCIENCE (1950 to 2009) using the following medical subject headings: myocardial infarction, post myocardial infarction, post infarction, heart attack, ventricular remodeling, ventricular volumes, ejection fraction, left ventricular function, exercise, exercise therapy, kinesiotherapy, exercise training. Reference lists of all identified studies were also manually searched for further relevant investigations.

Two investigators (MH and AC) independently reviewed the titles and abstracts of all citations to identify studies reporting the effect of exercise training on EF and/or ventricular volumes in recent (≤ 3 months) post-MI patients. We excluded trials that were not randomized, did not have a usual care control group, non-exercise intervention, exercise and other intervention, and non-human studies. We also excluded non-English articles.

Two authors (MH and BE) extracted relevant outcome data and any disagreement was resolved by consensus in discussion with AC. When necessary, original investigators were contacted to clarify data or provide additional data; authors for 3 studies provided further information. Quality was assessed using the Jadad scale [17].

All data was entered into SPSS files and analyzed with SPSS v15 software utilizing meta-analysis and meta-regression scripts created by Lipsey and Wilson [18]. The review conforms to the requirements of PRISMA reporting standards (Additional file 1). Formulae for calculation of outcomes is provided in Additional File 2.

After initial review of 1,033 citations (Figure 1), 19 papers were reviewed in full; of these, seven were excluded. Reasons for this exclusion were: EF data incomplete or un-extractable (n = 3), non-randomized design (n = 1), time from MI to initiation of exercise training was unreported (n = 1), inclusion of patients experiencing MI >3 months (n = 2). One trial [19] randomly assigned subjects 'a priori' to exercise or control groups based on baseline EF (≤ 30 and >30%). Accordingly, data for these sub-groups were analyzed separately.

The selected studies [5‐7, 9‐14, 19‐21](Additional File 3) contained 647 post-MI patients (mean age: 55 years) with impaired LV systolic function (weighted mean EF = 44%). Trials incorporated aerobic training at 60% to 80% of baseline peak oxygen uptake (or heart rate) for 20 to 180 minutes per session (Table 1). The length of training program was 1 to 6 months in duration (Table 1). All trials reported pre and post measures for outcome variable for both treatment groups and for control groups. No trial described randomization procedures or blinding methods. Consequently, trials tended to be assessed as being of low to moderate quality.

Three outcome variables were chosen for analysis. The primary variable was EF for which 12 trials reported data (n = 647). The secondary outcome variables were ESV and EDV for which data was only provided for a subset of trials: ESV (9 trials, n = 475) and EDV (10 trials, n = 512).

As the studies had high levels of statistical, clinical and methodological heterogeneity - as evident for example in the initial meta-analysis of EF (Cochran's Q = 33.57, df = 12, p < 0.01) and variations in effect size differences for each outcome variable, pooling of the results from the trials in a sub-analysis was inappropriate [22]. Therefore, we performed meta-regression of post-infarct LV remodeling. The effects of elements of the exercise interventions identified a priori for analysis were: time post-MI to initiation of exercise training program and length of training program [23].

Meta-regression identified that change in EF effect size difference decreased as the time between MI and initiation of the exercise program lengthened, and increased as the duration of the program increased (Figure 2). Overall, these changes accounted for a significant proportion of the variations in outcomes across the studies (Q = 25.48, df = 2, p < 0.01, R² = 0.76) (Table 2). That is, when study level differences in average time to initiate and average length of program are considered the trials become comparable.

A sufficient number of trials reported data on ESV (9 trials, n = 475) [5‐7, 9‐14] and EDV (10 trials, n = 512) [5‐7, 9‐14, 20] to allow meta-regression of these outcomes. Similar to EF, effects on LV volumes were not homogeneous across studies (ESV, Q = 30.12, df = 8, p < 0.01; EDV, Q = 33.03, df = 9, p < 0.01). Greater reductions in ESV and EDV (as indicated by effect size decreases) occurred with earlier initiation of exercise training and with longer training durations (Tables 3 and 4, Figures 3 and 4). These differences accounted for a significant proportion of the heterogeneity for ESV (Q = 23.89, df = 2, p < 0.05, R² = 0.79) and EDV (Q = 27.42, df = 2, p < 0.01, R² = 0.83). As with EF effect size, when study level differences in average time to initiate and average length of program were considered, effects on ESV and EDV became comparable.

The findings may be influenced strongly by a small number of trials which evaluated exercise training after around one week. Analyses were therefore repeated excluding the four trials with the shortest latency to initiation of the exercise program [5, 6, 12, 13]. After this removal, the analysis no longer had sufficient power to demonstrate significant effects. However, the regression coefficients remained of similar size and direction (For time to initiate: -0.119 to -0.178 (p < 0.15); for length from 0.098 to 0.069 (p < 0.25)) suggesting that the trends identified in the full analysis remained even when these trials are excluded. Removing the same trials from the analyses of ESV and EDV reduced the number of trials in the analyses more dramatically (5 and 6 trials respectively). In both re-analyses, the coefficients for time to initiate again remained in the same direction but decreased substantially in magnitude (ESV: 0.246 to 0.076 and EDV: 0.189 to 0.059) and were no longer significant. Yet, the coefficients for program length were more stable (ESV: -0.249 to -0.138, p < 0.10, and EDV: -0.148 to -0.13, p < 0.05). Hence, particularly for LV remodeling, the influence of time to initiation of exercise training after MI and subsequent length of training program remained after removal of the trials.

There is no current recommendations [30] or consensus [31] as to when exercise training should commence after MI. Most cardiac rehabilitation and secondary prevention programs commence at least four to six weeks after hospital discharge [32]. To achieve maximal anti-remodeling benefits, clinically stable patients after uncomplicated MI should begin aerobic exercise training earlier after hospital discharge (from one week) and should continue training for up to 6 months. As this conclusion represents a potentially significant change from current practice, it is important to take into account the size and safety of this.

Though understandable, there is no evidence from the trials in this review or other observational studies that early commencement of exercise training is harmful. Trials have indicated that risk of adverse events or complications (including: re-infarction, and revascularization ) and EF are not raised by earlier physical activity when compared to activity after 6 weeks - even when pre-discharge (Bruce protocol) stress tests are performed 1-week post-MI [33]. Consistent with these findings, pre-discharge exercise (Bruce protocol) stress testing is safe and feasible in the majority of post-MI patients 3 days after infarction [34]. In trials in this review, no adverse events occurred during the 6-month exercise training sessions initiated at the earliest juncture (around one week post-MI) in people with mild to moderate LV systolic dysfunction [5, 6]. Moreover, clinical events were significantly lower in the trained versus control group during the 6-month period.

In addition to benefitting mortality and being safe, earlier commencement of exercise training appears to markedly increase participation in secondary prevention services. Emerging evidence from observational studies indicates that commencement of cardiac rehabilitation after one week leads to a 90% increase in participation rates compared to a commencement after four weeks [35] and a faster return to work [33].

By providing more specific guidance on basic yet pivotal program design characteristics, these findings are also likely to render existing research evidence on exercise after MI more usable to health professionals and decision-makers [16]. Despite convincing evidence of the benefits of exercise after myocardial infarction [1, 2, 36, 37], secondary prevention in clinical populations remains poor [38] and health services to promote exercise are under-utilized and poorly funded [39]. Health services, including different forms of cardiac rehabilitation [36, 37, 40] should be used more widely to promote exercise earlier and for longer after MI.