Letters to the Editor Subsite-Specific Colorectal Cancer in Diabetic and Nondiabetic Patients To the Editor: We read with interest the report by Limburg et al. (1), a large prospective study of postmeno- pausal women which noted the association between type II diabetes mellitus and incident colorectal cancer to be subsite specific. Specifically, they reported a statistically increased risk of proximal colon cancer (relative risk, 1.9; 95% confidence interval, 1.3-2.6). An alternate way of exploring the question of diabetes and subsite-specific bowel cancer risk is to start with a population of patients with colorectal cancer, looking for differences in the distribution of tumors in the diabetic versus nondiabetic populations. This also removes some of the potential biases in the report by Limburg et al., as they included only female patients, of a relatively tight age range (55-69 years at registration), and relied on a patient questionnaire (returned by 42% of women). Here we report from our prospective colorectal cancer database of 1,139 patients entered over 14 years (from 1990 to 2004), at two Australian Hospitals. Prospective informa- tion on patient comorbidities has been collected on all patients, including diabetes mellitus. Using the same definitions as Limburg et al. (1), there were 308 proximal colon cancers, 456 distal colon cancers, and 365 rectal cancers. Ten patients with multiple primary tumors were excluded from the analysis. Of these colorectal cancer patients, 188 (16.5%) also had diabetes. The subsite distribution of colorectal carcinoma was not statistically different in diabetic and nondiabetic patients; however, there was a trend for more proximal cancers in the diabetic patients, 35.1% having proximal tumors versus 26.8% in the control group. The odds ratio of having proximal tumors compared with other sites in diabetic patients was 1.50 (95% confidence interval, 1.08-2.09); this was consistently adjusted for age and sex. Further break- down of this distribution according to sex revealed similar findings (42.3% versus 33.3% proximal tumors for females and 30.0% versus 21.1% proximal tumors for males). For completeness, the percentage figures for distal colon (36.7% versus 40.1%) and rectal cancers (28.1% versus 32%) were similar for diabetic and nondiabetic patients. Our findings are consistent with the hypothesis that diabetic patients are more likely to develop proximal colon cancers, as reported by Limburg et al. (1) and previously in the Nurses Health Study, which reported a relative risk of 1.64 for proximal cancers (95% confidence interval, 1.04-2.60; ref. 2). Interestingly, both of these studies included only women, whereas our study included both male and female patients. In the only other study with significant numbers of patients with colorectal cancer (3), the authors also reported a slightly higher risk of proximal colon cancers. Based on all of these reports, we agree with Limburg et al. (1) that screening of diabetic patients with sigmoidoscopy alone may be of less value than in the nondiabetic population. E. Lim I.T. Jones Royal Melbourne Hospital, Victoria, Australia P. Gibbs Royal Melbourne Hospital, Victoria, Australia Ludwig Institute of Cancer Research, Victoria, Australia S. McLaughlan I. Faragher I. Skinner M.W. Chao Western Hospital, Victoria, Australia J. Johns Ludwig Institute of Cancer Research, Victoria, Australia References 1. Limburg PJ, Anderson KE, Johnson TW, et al. Diabetes mellitus and subsite-specific colorectal cancer risks in the Iowa Women’s health study. Cancer Epidemiol Biomarkers Prev 2005;14:133 – 7. 2. Hu FB, Manson JE, Liu S, et al. Propective study of adult onset diabetes mellitus (Type 2) and risk of colorectal cancer in women. J Natl Cancer Inst 1999;91:542 – 7. 3. Weiderpass E, Gridley G, Nyren O, Ekbom A, Persson I, Adami HO. Diabetes mellitus and risks of large bowel cancer. J Natl Cancer Inst 1997; 89:660 – 1. Robustness of Case-Control Studies to Population Stratification To the Editors: Using computer simulations, Khlat et al. (1) quantified type I error increase caused by population stratification. They argued that under ‘‘realistic scenarios’’ (where subpopulations account for V10% of the study population and allelic frequency differences are V0.2), the inflation of type I error is of limited concern. Results from both computer simulations (2) and theoretical analyses (3) suggest a more nuanced and complex view of population stratification. Our results are consistent with some of those by Khlat et al. (1). First, a large inflation in type I error can occur when the two subpopulations are equal sized but have moderate marker allele and disease frequency differences. Second, the confounding risk ratio provides a poor measure of the increase in type I error rate under population stratification, as we showed in ref. (3). However, our computer simulations also show that having a mixture of unequal-sized subpopulations (e.g., 10% versus 90%) does not necessarily lead to a reduction in the inflated Cancer Epidemiology, Biomarkers & Prevention 1579 Cancer Epidemiol Biomarkers Prev 2005;14(6). June 2005 Downloaded from http://aacrjournals.org/cebp/article-pdf/14/6/1579/1744889/1579-1582.pdf by guest on 11 July 2022