A tool for assessing sex/gender bias in epidemiological studies of occupational health: Pilot testing on studies of sedentary behaviour at the workplace and cardiometabolic health.

BACKGROUND

The perspective of sex/gender bias is often missing in tools used to assess study risk of bias in systematic reviews. The aim was to pilot a checklist using an aetiological occupational health research question regarding the impact of sedentary behaviour at the workplace and cardiometabolic health. The checklist examined whether the consideration of sex/gender was associated with different study characteristics.

METHODS

A sex/gender checklist developed based on a synopsis of existing instruments with input from the Cochrane Sex/Gender Methods Group was adapted for the present study. This checklist comprises four categories: 1. "Background and conceptual considerations" (3 items), 2. "Study design" (2 items), 3. "Study procedures of investigation or intervention and statistical analysis" (2 items), and 4. "Presentation and interpretation of findings" (3 items). Two independent reviewers evaluated all included studies. Bivariate and multivariable logistic regression analyses were used to examine the consideration of sex/gender across study designs, years of publication, and risk of bias levels (based on the Scottish Intercollegiate Guidelines Network [SIGN]).

RESULTS

Of the 49 studies evaluated with the checklist, none provided detailed information, 69% (n = 34) provided basic information and 31% (n = 15) no information for the consideration of sex/gender. No intervention study provided information for the first two categories. In the third category, all intervention studies (n = 17) and case-control studies (n = 5) provided basic information on sex/gender, while two of the 23 cohort studies did not. In the fourth category, detailed information was found for all study designs (n = 8). Bivariate analyses revealed no association between the consideration of sex/gender and the year of publication (OR per year = 0.89; 95% CI: 0.65, 1.23). A low risk of bias level was not associated with consideration of sex/gender (OR = 0.60; 95% CI: 0.14; 2.50). Compared to intervention studies the odds of considering sex/gender was increased by a factor of 3.6 (95% CI: 1.0, 12.8) in observational studies.

CONCLUSION

The adapted checklist was applicable to assess the consideration of sex/gender in all studies. None of the primary studies considered sex/gender perspectives in all of the four categories. Further optimisation of the sex/gender tool seems warranted, based on further research on weighting individual categories or items and application of the checklist for occupational epidemiology in general.