Is there any practical application of meta-analytical results in strength training?

Designing resistance training (RT) programs is a complex task that involves the manipulation of numerous variables that interact with each other, influencing the program outcomes (Tan, 1999; Paoli, 2012). The attempt to clearly define the combination of variables which would bring optimal adaptations for different outcomes is undermined by the large number of studies involving RT, the conflicting findings reported by many of them and the lack of methodological clarity and consistency in previous studies' protocols. As such, meta-analyses emerge as an attractive approach since they allow the combination of multiple studies in an attempt to estimate the effect size of a single variable, surpassing possible inadequacies of statistical power within individual studies. With this aggregation of information, a more robust estimation of the effects is possible.

However, Field (2015) has noted a pertinent philosophical objection to these types of analyses that might apply to RT studies; in essence we have a replication crisis. Researchers often attempt to perform replications of the findings from earlier studies, yet frequently they do not adequately replicate the conditions of the original study. For example, one study may examine the effects of low or high set volume whilst participants train at a frequency of twice a week using repetition ranges of 8–12 and perform sets to momentary failure. Another may examine the effects of low or high set volume whilst participants train at a frequency of five times a week using 10 repetitions per set and not having participants perform sets to momentary failure. Though the two studies might appear to be examining whether low or high set volumes produce greater adaptations, they are in fact examining these within the context of different manipulations of other RT variables. There is likely a reason for this lack of proper replication, as was noted by Richard Feynman1. Indeed, we would argue that the currently heterogeneous body of literature on the effects of the manipulation of different RT variables is evidence of this replication crisis being alive and well in our field.

In this current opinion article we explain specifically why it might be unwise to conduct meta-analyses with such heterogeneous RT studies noting the effects of different confounding RT variables, and also suggest that it might be irresponsible to make general estimates of RT effects and propose recommendations.