Abstract
BACKGROUND: Diabetes is considered as an established risk factor for cancer development. However, the link between diabetes among cancer survivors remains inconclusive. The hypothesis of this study was to assess the hazard ratio (HR) of incidence of diabetes in cancer survivors compared with the HR in the general population. PATIENTS AND METHODS: A comprehensive literature search was performed in PubMed, Embase, and the Cochrane Library from database inception to 15 December 2020 for population-based cohort studies. Summary effect estimates were combined using random-effects models. We also performed subgroup analyses to test sources of heterogeneity and the stability of the results stratified by various study and participant characteristics. RESULTS: Thirteen population-based cohort studies involving 1 686 595 participants were analyzed. The HR for the development of diabetes in cancer survivors was 1.39 [95% confidence interval (CI) 1.29-1.50; I(2) = 82.3%; P < 0.001] compared with that in noncancer controls, among which survivors of hematological, gynecologic, breast, colorectal and urinary tract cancer (all P < 0.05) showed consistent significant results, whereas no significant increased risk was observed for other cancer types. The effects were more prominent in populations of shorter cancer survival duration (<1 year) (HR 2.09, 95% CI 1.32-3.32; P = 0.009). Moreover, cancer survivors with a longer follow-up period (>10 years) had a relatively higher risk of diabetes (HR 1.54, 95% CI 1.34-1.77) than those with a shorter follow-up period. CONCLUSIONS: In this large pooled analysis of population-based cohorts, evidence supports the hypothesis that the risk of developing diabetes is increased in cancer survivors compared with the general population. We should interpret the results with caution for considerable interstudy heterogeneity. However, health policy makers should take this as a challenge for the early prevention and effective intervention of diabetes.