Type 2 diabetes (T2D) is associated with increased incidence of several cancers, and T2D is increasingly recognized as a heterogeneous condition with clinically distinct subtypes. This study evaluated whether site-specific cancer incidence and cancer screening uptake differ across validated clinical subtypes of newly diagnosed T2D compared with adults without diabetes.
The primary objective was to estimate adjusted hazard ratios for seven site-specific cancers across T2D subtypes. A secondary objective was to assess differences in cancer screening uptake.
The study used electronic health record data from the Epic Cosmos Research Platform covering 2012–2025. Adults with newly diagnosed T2D were identified and classified into four clinical subtypes using validated algorithms applied to routine clinical data. Subtype distribution among the 575,139 adults with T2D was: severe insulin-deficient (SIDD) 21.6%, mild obesity-related (MOD) 23.5%, mild age-related (MARD) 40.7%, and mixed 14.1%.
A comparator cohort of 689,719 adults without diabetes was matched to the T2D cohort on age, sex, and body mass index (BMI).
The study employed a matched retrospective cohort design. Cause-specific Cox proportional hazards models were used to estimate adjusted hazard ratios (HRs) for seven site-specific cancers while accounting for competing risks. Follow-up time had a median of 3.8 years. Screening uptake for mammography and prostate-specific antigen (PSA) testing was analyzed as a secondary outcome.
Overall, cancer incidence varied by T2D subtype. The MARD subtype exhibited the highest overall cancer incidence at 17.3 per 1,000 person-years.
Compared with matched adults without diabetes, several site-specific cancers showed elevated rates across T2D subtypes. The cancers with increased incidence reported across subtypes included colorectal, pancreatic, liver, endometrial, and ovarian cancers. The largest hazards for these site-specific cancers were reported in the SIDD and the mixed phenotypes; the paper reports a notably elevated hazard in SIDD (HR = 3.87, 95% CI = 3.51 to 4.27). The source does not provide site-by-site HRs for each subtype beyond the summary comparisons noted.
In contrast, prostate cancer incidence was lower across all T2D subtypes relative to adults without diabetes, with the most pronounced reduction observed in MOD (HR = 0.60, 95% CI = 0.55 to 0.64).
Breast cancer incidence differed by subtype: rates were higher among the mixed phenotype (HR = 1.12, 95% CI = 1.05 to 1.19) and lower among MOD (HR = 0.85, 95% CI = 0.80 to 0.90).
The study reports that uptake of screening tests—specifically mammography and PSA screening—was lower across T2D subtypes compared with matched adults without diabetes. The source does not provide exact screening percentages or adjusted estimates in the abstract; details beyond the reported direction of difference were not reported in the available source text.
The findings indicate that site-specific cancer incidence and cancer screening behaviors vary by clinically defined T2D subtype. If validated and replicated, subtype classification derived from routine clinical data could help identify patients with newly diagnosed T2D who are at higher or lower risk for specific cancers and who may benefit from tailored surveillance strategies.
The authors caution that further study is required before implementing subtype-guided cancer surveillance in clinical practice.
The analysis used de-identified, HIPAA-limited secondary data and was exempt from IRB review according to the authors. Funding sources declared include the Winship Cancer Institute, Emory Global Diabetes Research Center, and the Pharmaceutical Research and Manufacturers of America Foundation. Epic Cosmos access is available to participating institutions after certification; a link to the platform is provided in the source.
The source abstract and accompanying material describe a large, retrospective EHR-based study with matching on age, sex, and BMI. Specific limitations, including potential residual confounding, details on covariates adjusted in the Cox models, and site-specific HRs for each subtype beyond the summary statistics in the abstract, were not fully reported in the available text. The authors note the need for further study to validate findings and to evaluate the utility of subtype-based cancer surveillance in clinical settings.
In this large EHR-derived cohort, clinical subtypes of newly diagnosed Type 2 diabetes demonstrated different patterns of site-specific cancer incidence and lower uptake of some cancer screening tests compared with matched adults without diabetes. The results suggest subtype-informed risk stratification could be informative, but additional research is required before clinical adoption.