Meta-regression analysis and publication bias:
Meta-regression analysis was used to explore possible sources of heterogeneity and to find characteristics of participants or trials with effective treatment effects. The result of univariate meta-regression analysis didn’t show a significant linear association between soy isoflavones dose and effect size of soy isoflavones effect on CRP levels (Coefficient= -0.0004, 95% CI: -0.03, 0.03; P=0.977) (Table 5& Figure 5 ). Also, after adjustment for design, intervention duration, baseline CRP, health status, sample size, region, age, BMI, quality assessment, and publication year of article, dose of soy isoflavones did not have any linear association with effect size of soy isoflavones effect on CRP levels (Coefficient= 0.009, 95% CI: -0.11, 0.13; P=0.847). We also didn’t find a significant association between other studied variables and effect size in univariate meta-regression analysis (table 5 ). Regarding RCTs which indicated soy isoflavones effect, there was not any evidence of publication bias (Egger test p-value=0.954) (Figure 6 ).
Although, univariate meta-regression analysis didn’t show a significant linear association between soy isoflavones dose of combination of soy isoflavones plus soy protein dose and studied effect size (Coefficient= -0.008, 95% CI: -0.02, 0.001; P=0.082) (Table 6 & Figure 7 ). Also, after adjustment for other variables, there was not observed any significant association between dose of soy isoflavones and studied effect size (Coefficient= -0.005, 95% CI: -0.02, 0.01; P=0.447). Although the funnel plot was not visually symmetric for studies included in meta-analysis of soy isoflavones plus soy protein (Figure 8 ), the results from the Egger test did not show evidence of publication bias (Egger test p-value=p=0.781).