Type 2 diabetes mellitus (T2DM) is a major public health problem. Evidence suggests an association between diabetes and cancer, but this may be distorted by confounding. This research aimed to identify and assess the evidence suggesting a causal association between T2DM and cancer.

A systematic review was conducted in Pubmed, Embase, CINAHL, Web of Science, and the Cochrane Library for literature on the association between T2DM and cancer, from inception to 7 May 2021. Case-control and cohort studies published in any language were considered. Based on a targeted literature review, a directed acyclic graph for each type of cancer was developed to test whether the causal effects were adequately controlled for potential confounding.

A total of 131 studies with a low risk of bias were selected that reported 415 effect estimates of the association of T2DM with 57 types of cancer. Breast, colorectal, pancreas, prostate, and lung were the cancer sites with the highest number of studies. Causality was claimed in 57 studies, but only 34 percent of the outcomes were adequately controlled for confounders. Of the studies assessing a causal relationship for prostate and pancreatic cancer, 87 and 70 percent adequately controlled for confounding. In contrast, only 29 percent of lung cancer, 27 percent of colorectal cancer, and 17 percent of breast cancer results considered the minimal sufficient adjustment set. Lifestyle variables were identified as key potential confounders in more than 20 types of cancer, but they were not included in the analyses.

Many studies simply reported an association between diabetes and cancer. The policy implications of such studies are unclear. Of the studies claiming a causal link between diabetes and cancer, a large proportion did not adequately control for confounding. It is critical that studies take a systematic approach to identifying potential confounding factors, such as targeted reviews and the development of directed acyclic graph approaches, to estimate causal effects that may be useful in informing health policy.