We conducted three nested case-control studies to evaluate the risk of breast, colon, and prostate cancers developing in patients exposed to thiazolidinediones (TZDs) compared with other anti-diabetic agents. Cancer cases were matched to five controls by age, gender, calendar year, and time in the database from a cohort of 1 26 971 diabetic patients taking anti-diabetic medication in the US Integrated Healthcare Information Services database. Five hundred thirteen breast cancer cases were matched with 2557 controls, 408 cases of colon cancer were matched with 2027 controls and 643 cases of prostate cancer were matched with 3176 controls. Exposure to an anti-diabetic agent within 90 days preceding the index date was defined as recent exposure and at any time during the follow-up was defined as ever exposed. The adjusted odds ratios and 95%CI of cancer from ever exposure to TZDs compared to oral monotherapy, oral dual therapy, oral triple therapy, insulin monotherapy, insulin and oral therapy and all non-TZD anti-diabetic agents were, respectively for breast cancer: 0.91 (0.69-1.20), 0.80 (0.56-1.14), 0.87 (0.32-2.35), 1.27 (0.61-2.67), 0.71 (0.36-1.37), 0.89 (0.68-1.15); for colon cancer: 1.06 (0.80-1.40), 1.12 (0.77-1.63), 1.73 (0.39-7.78), 4.46 (1.05-19.00), 1.06 (0.50-2.26) 1.03 (0.80-1.32) and for prostate cancer: 1.08 (0.85-1.37), 0.89 (0.66-1.21); 0.82 (0.33-2.06); 1.80 (0.79-4.07), 1.10 (0.55-2.18), 1.04 (0.83-1.31). Results for exposure within 90 days of the date of the cancer were similar. Our findings suggest that the effect of TZDs on the likelihood of development of the cancers studied (colon, prostate and breast) appears to be neutral and do not support a beneficial or deleterious effect of TZD on the cancers studied. Copyright 2006 John Wiley & Sons, Ltd.