Date of Award


Document Type

Open Access Dissertation


Epidemiology and Biostatistics


Norman J. Arnold School of Public Health

First Advisor

James Hébert


Background: There is growing evidence of association between diabetes and cancer. No studies have been conducted in India evaluating this association. With the current epidemiologic, nutritional and economic transition in India, it becomes extremely important to examine this association in an Indian population. Additionally, difference in association exists based on different cancer subtypes. Research has shown that diabetes is associated with an increased risk of colorectal cancer. However most of these studies suggest detection bias to be one of the probable reasons for this association. Additionally, the common risk factors shared by both these conditions are considered to one of the reasons in the association. Furthermore, very few studies have assessed the association between duration of diabetes and either CRC risk or disease aggressiveness. Even more rarely have studies confirmed the status of type 2 diabetes mellitus (T2DM) while determining the diabetes-CRC association.

Methods: For our first objective, we used the Mumbai Cohort Study (MCS)- a longitudinal study. Diabetes information was collected at baseline and cancer information was received via follow-up questionnaire and confirmed using cancer registry. We also evaluated the association between diabetes and cancer subtypes after creating matched datasets for each cancer subtype. We used Cox Proportional model for cancer incidence and conditional logistic regression for cancer subtypes. For our second and third question, we used the Prostate Lung Colorectal Ovarian (PLCO) Cancer screening trial. Diabetes information was self-reported and collected at baseline and using one of the follow-up questionnaires-supplemental questionnaires. The cancer information was collected using annual survey questionnaire (ASU) administered every year and confirmed using medical records. For our second aim final analysis we use cox proportional hazards model. To evaluate the notion of detection bias, we conducted stratified analysis. In our final question, the diabetes duration was calculated using information on age at diabetes diagnosis. We fit a Cox proportional hazards model for cancer incidence and conducted logistic regression analysis for cancer grade and stage.

Results: In the MCS, we did not observe any significant associations between diabetes and all cancer incidence and cancer subgroups. However the association was in the expected direction. The hazards of all cancer incidence was 1.06 (95%CI=0.75, 1.62) among persons with diabetes as compared to people without diabetes. Among cancer subtypes, there was an increased risk of ‘lip/oral/pharyngeal cancer’ (OR=1.83; 95%CI=0.86, 3.86) and ‘respiratory tract cancer’ among people with diabetes (OR=1.28; 95%CI=0.53, 3.13) respectively. Inverse direction was observed for ‘digestive organ cancer and ‘breast/prostate/uterine/cervical cancer’ among people with diabetes compared to people without diabetes (OR=0.59; 95%CI=0.27, 1.32) and (OR=0.66; 95%CI=0.24, 1.84) respectively, but none of these associations reached statistical significance. For our second aim, we observed a 33% higher risk of CRC among people with diabetes as compared to people without diabetes. After stratifying the results by screening arm, we still found a higher risk among both the screening arms, (HR=1.41, 95%CI=1.13, 1.76) among the control arm (HR=1.22, 95%CI=0.94, 1.58). After stratifying by BMI, the risk was still high among people with diabetes in all the groups.In our final aim, we observed that participants with >10 years of diabetes had a higher risk (HR=1.37; 95%CI: 1.06, 1.77) of CRC incidence compared people without diabetes. An apparently smaller effect was observed among people with(HR=1.13; 95%CI: 0.89, 1.43); however, it was not significant. We did not find significant results in the association between cancer aggressiveness and diabetes.

Conclusion: In Indian population, our findings appear to show a higher hazards of all cancer incidence, lip/oral/pharyngeal and respiratory tract cancer among people with diabetes compared to people without diabetes. They direction of the association is consistent with previous study results. However the association is not significant. Future studies needed to explore this association in detail. Secondly, in the PLCO data, our findings showed an association between diabetes and increased risk of colorectal cancer. Detection might not be the reason for this association. Further studies should include information on other factors like diabetic medications. For our final aim, the CRC risk was higher among people with longer duration of diabetes, even after accounting for the potential confounders.


© 2017, Shraddha S. Vyas

Included in

Epidemiology Commons