Commentaries on “Effect of blood‐flow restricted vs heavy‐load strength training on muscle strength: Systematic review and meta‐analysis”

Abstract

We read with great interest the recent meta-analysis by Grønfeldt et al,1 comparing the effects of low-load resistance training associated with blood-flow restriction (LL-BFR) versus conventional high-load resistance training (HL-RT) on muscle adaptations in healthy individuals. We believe this work reinforces the notion, previously established by our group2 and others,3 that both training protocols are equally effective in promoting hypertrophic adaptations in different cohorts. However, Grønfeldt et al1 observed similar increases in maximum strength between LL-BFR vs HL-RT, in contrast to previous reports of superior gains for the latter.2,3 In their current work, Grønfeldt et al1 credit the discrepancies in the results to methodological inadequacies employed in our previous work. In the following lines, we present a series of issues with the current findings by Grønfeldt et al.1 In their recently published paper, the authors criticize our inclusion criteria (eg, non-randomized trials and unilateral experimental designs), the use of multiple outcomes for maximum muscle strength assessment (eg, specific and non-specific) and inclusion of more than one comparator arm in a single study (eg, double counting). To address these critiques, we re-analyzed our previous data, considering all the points raised by Grønfeldt et al.1 As expected, our conclusion remains the same (ie, higher increases in muscle strength for HL-RT as compared with BFR-RT) (Table 1), suggesting that these points did not compromise our conclusion. A crucial difference between the two studies is the meta-analytic design. While Grønfeldt et al1 used a between-groups design (ie, post-intervention comparison), we2 used a within-subjects design (ie, preto post-intervention changes).4-6 Although not consensual, literature advocates in favor of the latter, as this model can increase statistical power, test precision, and allows testing for subject-by-treatment interaction.5,6 In this respect, estimating possible differences in group variability (ie, delta-change variability) is critical to determine between-group differences. Furthermore, potential between-group differences at baseline, due to sampling error in studies with small sample size, can either underor over-estimate the standardized mean difference in a meta-analysis,5 biasing possible between-group differences. For instance, three studies7-9 included in both meta-analyses present between-group differences at baseline ranging from 16% to 54%, favoring the LL-BFR protocol that, despite not statistically significant, may have influenced the analyses of Grønfeldt et al.1 To further explore this point, we re-analyzed the data from Grønfeldt et al1—removing all studies with a baseline between-group difference >15% (3 studies7-9)—but still using their meta-analytic design (ie, post-intervention data model). The results showed that removing these studies led to a different interpretation of the results, with the standardized mean difference favoring HL-RT protocols. This exploratory analysis should be interpreted with caution, given that removing studies based on arbitrary cutoffs may not be ideal. Thus, we also re-analyzed the data published by Grønfeldt et al1 using a within-subject design and including all studies. Not surprisingly, the results corroborate with our previous conclusion of superior gains in maximum strength for HL-RT protocols as compared with LL-BFR.2 Finally, a closer look into Grønfeldt et al1 data revealed a somewhat biased approach in selecting the comparator arm in case of studies with multiple arms. According to the authors, “in studies where multiple LL-BFR exercise protocols were employed, the interventions most similar to the LL-BFR exercise protocols of other included studies were selected to ensure a homogenous comparison between exercise modes.” Although one may argue that this approach can reduce data heterogeneity within a meta-analysis, it is important to highlight that selecting a specific arm could also ignore relevant comparisons between LL-BFR vs HL-RT, precluding proper comparison between protocols. Also, “in studies where multiple modes of maximum strength measurements were reported, the test modality with the highest test-retest reliability was included, prioritized in the following order: 1. isometric dynamometer; 2. isokinetic dynamometer, 3. handheld dynamometer and 4. RM-test.” Although we may agree that this can increase the precision of the LL-BFR estimate, excluding isoinertial strength assessment when more than one strength measurement was taken could also be importantly misleading,