Commentary: BMI is associated with sperm quality and sex hormones in men: a meta-analysis

We have read with great interest the article by Ji and colleagues, entitled "BMI is associated with 10 sperm quality and sex hormones in men: a meta-analysis" (1) . The authors provide valuable evidence 11 indicating that elevated BMI is significantly associated with impaired semen quality and altered sex 12 hormone levels in men (1). Although interpretation of these findings must consider potential 13 heterogeneity arising from differences in participant sources, the consistent adverse effect of BMI 14 across various populations strongly supports its inclusion as a routine component in the clinical 15 evaluation of male fertility (1). A review of the literature identified possible data entry errors in Ji and 16 colleagues'analysis (1), which may have led to inconsistencies in the reported results. In our re-17 analysis using corrected data, the subgroup comparison showed no statistically significant difference 18 in luteinizing hormone (LH) concentrations between overweight and obese individuals (P = 0.25), in 19 contrast to the original study, which reported a significant difference between these groups (1). 20 A critical assessment of the methodological concerns to subgroup analysis is imperative. 21 Specifically, a clear discrepancy exists between the statistical approach described in Section 2.6 and 22 its actual implementation in the study (1). The authors state that a random-effects model would be 23 applied in cases of high heterogeneity (I² > 50% or P < 0.1) -conditions that were explicitly met in 24 their results (I² = 98% for Figure 4A; I² = 96% for Figure 4D) (1). Despite this, the forest plots in 25Figures 4A and4D indicate the use of a fixed-effects model (1).This model misapplication has 26 produced erroneous statistically significant findings: first, regarding total testosterone (TT) 27 concentration in the normal weight versus obesity comparison (Figure 4A), and second, regarding 28 luteinizing hormone (LH) concentration between overweight and obese individuals (Figure 4D) (1). 29The present commentary seeks to correct this analytical error and present a revised analysis using the 30 appropriate random-effects model. 31We utilized the same dataset and adhered to the same inclusion criteria as those specified by Ji et 33 al. All statistical analyses were performed using RevMan software (version 5.3). For continuous 34 variables, the mean difference (MD) was employed for outcomes measured with uniform units, whereas the standardized mean difference (SMD) along with a 95% confidence interval (CI) was 36 applied for outcomes with differing units. Heterogeneity across studies was evaluated using the I² 37 statistic, with values exceeding 25%, 50%, and 75% indicating low, moderate, and high heterogeneity, 38 respectively. When I² was 50% or higher, sensitivity or subgroup analyses were performed, and a 39 random-effects model was adopted. In cases where I² was below 50%, a fixed-effects model was used 40(2). Thus, the selection of the statistical model was determined based on the observed I² value. A p-41 value of less than 0.05 was regarded as statistically significant. 42Revised meta-analysis results 43