Effect of confounding in the association between circumcision status and urinary tract infection

Objectives. To estimate the impact of confounding in the association between circumcision status and urinary tract infection from epidemiological factors, sample collection, and health-seeking behaviors in the first year of life. Methods. Beginning with the assumption that true urinary tract infection occurred equally regardless of circumcision status, a Markov model incorporating the differences in the rates of prematurity, of urine collection, of false positive urine specimens, and of health-seeking behaviors in infant boys based on circumcision status was developed. Using this model, the rates of false-positive urine cultures, asymptomatic bacteriuria, and true urinary tract infection detected in the first year of life were estimated and contrasted. Error of the model was estimated using Monte Carlo simulations. Results. Keeping the incidence of true urinary tract infection constant between groups, the factors included in the model could account for urinary tract infection being diagnosed 4.27 times more frequently in non-circumcised mates under a year of age. Conclusions. Previously reported differences in the rate of urinary tract infection by circumcision status could be entirely due to sampling and selection bias. Until clinical studies adequately control for sources of bias, circumcision should not be recommended as a preventive for urinary tract infection. (c) 2004 The British Infection Society. Published by Elsevier Ltd. All rights reserved.