Background: It has been suggested that any systematic review including meta-analysis of observational studies must consider the risk of bias in individual primary studies. To date, checklists for the reporting of observational epidemiological studies have been developed, but their usefulness is limited. Moreover, the design features of primary studies rather than a design label is important because the risk of bias is affected by the specific features of a study, not by a broad categorization of the approach taken. As such, listing potential confounding factors in the primary studies and stratification of participants into subgroups with respect to potential confounding factors is straightforward and valuable.

Objectives: To analyse the effect estimates by stratification of the participants according to the potential confounding factors.

Methods: As an illustration, we performed a meta-analysis of observational studies identifying the association between hepatitis C virus infection (HCV) and type 2 diabetes (T2D), following the Cochrane guideline for the non-randomised studies. We stratified the participants according to known risk factors age group, gender, and family history of diabetes, and performed subgroup analysis.

Results: We identified observational studies addressing the association between HCV and T2D (k = 21, n = 294437). A positive and significant association between HCV and T2D was observed (odds ratio 1.71; 95% CI = 1.17-2.49). Patients of age >40years (7.47; 95% CI = 5.93-9.42), obesity (0.32: 95% CI = 5.93-9.42) and female sex (0.8: 95% CI = 0.66-0.97) were also significant factors.

Conclusions: Observational epidemiological studies are generally poorly reported. The evaluation of methodological quality and risk of bias consistently across primary studies is difficult. The stratification by potential confounding factors in the interpretation of the findings of meta-analysis of observational studies could be valuable.