*2.4. Data Synthesis and Analyses*

The analysis was conducted by first extracting the relevant information from all study groups at baseline and at the end of the intervention. This included the number of participants (*n*), *p* values, mean, standard deviation (SD) and 95% confidence intervals (if available). To compare the effects of protein supplementation against placebo conditions, pooling was used of the continuous data as standardized mean difference (SMD) represented as Cohen's d effect sizes (ES), standard error (SE) and 95% confidence intervals calculated for each main outcome using the reported mean change differences (delta scores), *n* and corresponding SDs [17,19]. If the mean change difference was not reported this was calculated based on the reported pre-and-post mean and SDs in each study. If a study used multiple protein-supplemented groups, we combined the data into an overall protein-supplemented group for subsequent analyses [17]. When a study used multiple performance outcome measures, the relative . V O2peak and lower body 1RM were

prioritised for muscle strength and aerobic endurance [17]. Effect sizes were classed as small (0.2), medium (0.5), and large effects (0.8) [50]. Effect size was calculated using the following equations:

> E1: Cohen's d = (M2 − M1)/SDpooled

$$\text{E2: } \text{SD}\_{\text{pooled}} = \sqrt{(\text{SD}\_1^2 + \text{SD}\_2^2)/2)}$$

**Figure 1.** Flow chart of study retrieval process.

A random-effects model was applied with heterogeneity across studies tested using I2 test. I2 values of 25%, 50%, and 75% were considered low, moderate and high, respectively [17]. Each study was weighted (%) based on its inverse within study variance and between study variance using the Meta-Essentials spreadsheet 1.4 (Microsoft Excel 2016, Washington, DC, USA). Meta-Essentials was used for the meta-analysis, creation of forest and Egger's funnel plots (including the trim and fill method) and running statistical analysis, with alpha set at *p* ≤ 0.05.
