2.7. Assessment of heterogeneity

We will utilize I² statistics to assess the heterogeneity of each pairwise comparisons. When I² < 50% and P > .10, we will adopt the fixed effect model since there is no significant heterogeneity. If not, we will perform a random effect model. Subgroup analysis and sensitivity analysis are needed for heterogeneity caused by factors such as trial design and treatment duration. In addition, supposing that the origin of heterogeneity is unaware, only descriptive analysis will be used.

Considering the heterogeneity, we will perform subgroup analysis in the light of the characteristics of research related to the source of heterogeneity. In addition, with regard to different design schemes, we will adopt subgroup analysis in accordance with age, gender, treatment type and course of disease.

Sensitivity analysis will be assessed by excluding the literature one by one to ascertain whether the literature has an impact on heterogeneity. Once the heterogeneity of the study changes, this article may be the source of heterogeneity, and further analysis will be carried out, such as the difference in sample size and the reference standard of outcome indicators. If there is no obvious change before and after exclusion, it manifests that the results are stable and credible.