4.3.6. Sources of heterogeneity

The potential sources of heterogeneity will be investigated by subgroup and meta-regression analyses [24]. Subgroup analyses will be performed by type of diabetes, HbA1c levels, sex, race, obesity status, CKD status. If more than 10 studies are included in the quantitative synthesis, then subgroup analyses will be supplemented by random effect meta-regression analyses which will allow the effects of multiple factors (called effect modifiers) to be simultaneously investigated [27]. The potential effect modifiers considered will be the following: sex, race, obesity status, and type of study (observational vs experimental). We will use the model F value and its statistical significance to assess whether there is evidence for an association between any of the covariates and the outcome; all the covariates with p-value <0.2 in bivariable models will be added to the multivariable model, in which a p-value <0.05 will be considered statistically significant. The model fit will be assessed by the adjusted R² which measures the proportion of the between-study variance explained by the covariates [28]. To control for the risk of type I error when performing meta-regression with multiple covariates, we will perform Monte Carlo permutation tests to calculate P values adjusted for type I error and we will check if there is a change in statistical significance [28, 29].