Case-control studies in diabetes. Do they really use a case-control design?

Aims: Studies defined as case-control do not always use this design. We aimed to estimate the frequency of mislabelled case-control studies in published articles in the area of diabetes and to identify the predictors of incorrect labelling.

Methods: We searched Medline and Web of Science for articles with "diabetes" and "case control" in title and filtered for language (English/Romance) and period (January 2010-December 2014). Inclusion criteria were: (1) statement to use a case-control design in title, (2) to be a final full-length publication and (3) to have original data in the area of diabetes. Three independent reviewers went through titles, looked for full texts and reviewed them. Discrepancies were settled with a fourth reviewer. Expert epidemiologist advice was requested in case of doubt.

Outcome Variable: case-control mislabelling; addressed predictors: publication year, journal impact factor and journal subject.

Statistics: proportion of mislabelled CC articles and assessment of predictors by multivariate logistic regression analysis.

Results: We retrieved 362 articles, 251 of them fulfilling inclusion criteria. The proportion of mislabelled CC studies was 43.8% (confidence interval 95% 37.7-50.0%). Most mislabelled studies had a cross-sectional design (82.7%). Predictors of mislabelling were publication year, journal impact factor and journal area.

Conclusions: A relevant subset of studies defined as case-control in the area of diabetes correspond to mislabelled cross-sectional studies. Incorrect labelling misleads readers regarding the interpretation of results and the cause-effect hypothesis. Researchers, reviewers and editors should be aware of and commit to settle this issue.