The choice of the control group is a key issue in case-control studies, particularly in studies of molecular epidemiology. We discuss the potential bias introduced by different options. To exemplify the consequences of different choices, we have analysed two sets of controls in the context of a case-control study on bladder cancer: 55 were patients with urological conditions (cystitis, prostate hypertrophy), while 49 had a miscellany of medical or surgical conditions. We measured DNA adducts in white blood cells (WBC) by (32)P-postlabelling and a series of metabolic polymorphisms (GSTM1, GSTT1, GSTP1, NAT2, NQO1). While no statistically significant differences were found for metabolic polymorphisms, the two series of controls showed different concentrations of DNA adducts, suggesting that conditions related to bladder cancer or intermediate steps leading to bladder cancer, such as chronic cystitis, may be associated with higher adduct levels. An association between DNA adduct levels and infection has been noted before in experimental animals: both in lung and in the skin, an inflammatory response increased the biologically effective doses of polycyclic aromatic hydrocarbons. An alternative explanation is confounding; in fact, after adjustment for the level of consumption of fruit and vegetables (but not for smoking) the difference between the two control groups was no longer statistically significant. In conclusion, the choice of controls in studies of molecular epidemiology has subtle methodological implications, including confounding of metabolic/molecular measurements by complex exposures such as diet.

作者：C, Sacerdote；M, Peluso；A, Munnia；C, Malaveille；P, Vineis

来源：Biomarkers : biochemical indicators of exposure, response, and susceptibility to chemicals 2000 年 5卷 4期