Pooling of Measures Within the Change-of-Direction Speed Meta-analysis

The commentary raises the same concern regarding the inclusion of left and right limb values for the arrowhead agility test in Hwang and Kim [10] and of the T-test and the Illinois agility test in Nawed et al. [11]. We respectfully disagree in both cases. Change-of-direction (COD) speed is often treated as a single construct reflecting rapid deceleration, reorientation and acceleration. Field tests, such as the T-test, Illinois agility test and arrowhead test are widely used as interchangeable proxies in research and practice [12–14].

In line with this, the Cochrane Collaboration Handbook (Sect. 10.3.2) explicitly supports the aggregation of conceptually similar outcomes using the SMD framework when studies assess the same underlying construct with different measurement tools. Accordingly, the inclusion of multiple COD outcomes (e.g., left/right directions or different COD tests) from the same study is consistent with established meta-analytic practice, provided that they reflect the same performance domains. As noted above, the inclusion of multiple outcomes reflects different operationalization of the same construct rather than distinct domains.

We further note that the issue of dependence among effect sizes was examined directly. While methodological work has highlighted failure to account for within-study correlation as a potential source of bias in meta-analysis, particularly when large numbers of highly correlated outcomes are included [15], the magnitude of this issue is context-dependent. In the present analysis, no more than three dependent outcomes were included per study, and sample sizes were comparable across groups. To evaluate this empirically, we conducted a three-level meta-analysis (see Supplementary Material). These analyses demonstrated that within-study dependence contributed negligibly to the total variance for dynamic balance and vertical jump performance, and although some dependence was observed for COD, it did not affect the magnitude or statistical significance of the pooled estimates. Importantly, the results of the three-level models were consistent with those of the original analyses, reinforcing the robustness of the findings. Supplementary to this, numerous previous systematic reviews and meta-analyses have similarly combined multiple COD-related outcomes within a single analysis when assessing functionally comparable tests [16–18].

Decisively, the colleagues’ own sensitivity analysis conducted with the adjustments they advocate returned an estimate of SMD = − 0.47 (95% CI − 0.78 to − 0.17; p = 0.002; _I_2 = 61.9%), which is statistically significant and fully consistent with the direction and magnitude of our original pooled estimate (SMD = − 0.65; 95% CI − 1.11 to − 0.20; p = 0.005; _I_2 = 84%). The difference in magnitude is modest and does not alter the practical effect interpretation. Both analyses therefore support the same conclusion for practitioners.