TABLE 2.

Associations between diabetes and risk of different molecular subtypes of colorectal cancer

^a Adjusted for: study, sex, age at CRC diagnosis, energy intake, family history, BMI, red meat, processed meat, vegetables, fruit, alcohol, smoking, exercise and aspirin/NSAID use.

^b Error bars represent 95% confidence intervals.

^c Multinomial logistic regression was used to compare colorectal cancer cases to cancer free controls separately for each molecular pathological subtype (polytomous analysis, P).

^d Multinomial logistic regression was used to compare cases of each molecular pathological subtype to all additional cases not belonging to that subtype (case only analysis, P _difference).