Pooling of Measures Within the Dynamic Balance Meta-analysis

The commentary contends that including multiple directions of the Y-balance test (YBT) and left/right limb scores represents a unit-of-analysis error. We do not agree with that framing. Chapter 10 of the Cochrane Handbook (§10.3.2) explicitly permits the pooling of different scales or measures that capture the same underlying construct using the standardized mean difference (SMD), and the construct at stake here, dynamic postural control of the lower limbs, is well established in the sports-science and medicine literature as being expressed, variably, through composite YBT scores, through the anterior, posteromedial and posterolateral reach distances and through left and right limb scores. These measures are conceptually coherent with one another and represent different operationalization of the same construct rather than distinct outcomes.

Importantly, restricting inclusion to studies reporting composite YBT scores only would introduce a further source of bias, as computation of the composite score requires normalisation to limb length. In studies where only absolute reach distances are reported and limb length is not available, a composite score cannot be derived. Consequently, a substantial proportion of otherwise eligible studies would be excluded on purely reporting grounds rather than substantive methodological differences. The approach adopted in the original analysis therefore maximises data utilisation while maintaining conceptual consistency of the construct.

The YBT is a standardized derivative of the star excursion balance test (SEBT), developed to enhance measurement consistency and clinical applicability. It can be regarded as an instrumented and simplified version of the SEBT, reducing the assessment from eight reach directions to three. Empirical work has demonstrated strong associations between YBT and SEBT reach distances, supporting their use as closely related definitions of dynamic postural control [4, 5]. Accordingly, combining outcomes derived from these protocols within a single meta-analytic framework is consistent with their shared construct basis and established practice in the literature [6, 7].

We acknowledge that inclusion of multiple outcomes from the same study can, in theory, introduce statistical dependence. Given the conceptual alignment of these measures, the limited number of outcomes per study, and the consistency of findings in sensitivity analyses applying more restrictive outcome selection, we consider any such dependence unlikely to have influenced the direction or interpretation of the pooled effects across the outcomes examined. Furthermore, the inclusion of asymmetry data from Pardos-Mainer et al. [8] is justified, as their assessment followed the methodology of Gonzalo-Skok et al. [9], where balance and inter-limb asymmetry are evaluated as related components of neuromuscular control.

Additionally, two internal features of the original analysis support this position. First, the low heterogeneity observed (_I_2 = 7%) is consistent with these measures capturing a closely related underlying construct, although we recognise that _I_2 alone cannot establish measurement equivalence. Second, the random-effects model accounts for between-study differences in measurement, so the pooled estimate is a robust summary rather than a naive aggregation. The commentary’s own sensitivity analysis, restricted to studies that reported a YBT composite score, yielded a significant positive effect favouring FIFA 11 + (SMD = 0.70; 95% confidence interval [CI] 0.122–1.28; p = 0.018). This is consistent in direction and broadly comparable in magnitude to our original pooled estimate (SMD = 0.37; 95% CI 0.18–0.56; p < 0.001; _I_2 = 7%). The difference in magnitude is modest and well within the range expected given differences in outcome selection and precision. In other words, the colleagues’ preferred analysis and ours converge on the same substantive conclusion: FIFA 11 + improves dynamic balance in youth and adult soccer players. We therefore find no sufficient methodological basis to conclude that readers would be misled by the original analysis.