RESUMO
BACKGROUND: Mendelian randomization (MR) analyses have been increasingly used to seek evidence of causal associations. This systematic review aims at characterizing and evaluating the reporting of MR analyses in oncological studies. METHODS: The PubMed database was searched to identify MR cancer studies until December 31, 2017. Two of the authors independently selected and evaluated reporting quality of the studies. Reporting quality in MR studies before 2016 and in 2016/17 was compared. RESULTS: Cancer studies with MR analyses in 2016 and 2017 accounted for 55.8% of the total number of studies identified. In the 77 eligible articles, 39 (50.6%) did not report subjects' characteristics, 53 (68.8%) did not conduct power estimation, 40 (51.9%) did not state all of the first three MR assumptions (i.e., genetic instrument is associated with exposure, is not associated with confounders, and acts on outcome only through exposure), and 31 (40.3%) did not exclude SNPs that diverged from Hardy-Weinberg equilibrium. More studies estimated power in 2016/2017 than before 2016 (p = 0.028). CONCLUSIONS: Some MR cancer studies did not sufficiently report essential information, posing obstacles for critical appraisal. This study proposes for MR analysis a guideline/checklist for future publications in cancer and other biomedical research.
