Mortality after incident cancer in people with and without type 2 diabetes: impact of metformin on survival

Abstract

Objective: Type 2 diabetes is associated with an increased risk of several types of cancer and with reduced survival after cancer diagnosis. We examined the hypotheses that survival after a diagnosis of solid-tumor cancer is reduced in those with diabetes when compared with those without diabetes, and that treatment with metformin influences survival after cancer diagnosis.

Research design and methods: Data were obtained from >350 U.K. primary care practices in a retrospective cohort study. All individuals with or without diabetes who developed a first tumor after January 1990 were identified and records were followed to December 2009. Diabetes was further stratified by treatment regimen. Cox proportional hazards models were used to compare all-cause mortality from all cancers and from specific cancers.

Results: Of 112,408 eligible individuals, 8,392 (7.5%) had type 2 diabetes. Cancer mortality was increased in those with diabetes, compared with those without (hazard ratio 1.09 [95% CI 1.06-1.13]). Mortality was increased in those with breast (1.32 [1.17-1.49]) and prostate cancer (1.19 [1.08-1.31]) but decreased in lung cancer (0.84 [0.77-0.92]). When analyzed by diabetes therapy, mortality was increased relative to nondiabetes in those on monotherapy with sulfonylureas (1.13 [1.05-1.21]) or insulin (1.13 [1.01-1.27]) but reduced in those on metformin monotherapy (0.85 [0.78-0.93]).

Conclusions: This study confirmed that type 2 diabetes was associated with poorer prognosis after incident cancer, but that the association varied according to diabetes therapy and cancer site. Metformin was associated with survival benefit both in comparison with other treatments for diabetes and in comparison with a nondiabetic population.

Figure 1

Adjusted* survival using alternative diabetes-related treatment regimens: monotherapy within 90 days of cancer diagnosis. Note: the patterns of survival in the insulin and sulfonylurea groups are the same. *Cox model specification: age, sex, smoking status, year of cancer diagnosis, and Charlson comorbidity index.