Bladder cancer risk in patients with spinal cord injury

Bladder cancer (BC) is the fourth most common cancer in men. It is associated with several risk factors (RF), of which only a few have been evaluated in previous studies. Models incorporating only one or a severely restricted number of RFs did not predict BC well. We employed 19 a priori RFs and 12 interactions in a multivariate logistic regression analysis for the prediction of BC in a sample of subjects with spinal cord injury (SCI), of whom 149 were outpatients (7 with BC) and 4 were inpatients with BC. We also replicated dichotomous predictions for 10 of the 19 RFs that have been most frequently associated with a higher BC risk in the literature. The overall test for 31 predictors was significant (p = 0.0038). A dichotomized predictor correctly identified 9 of 11 BC cases and all 142 but one of the cases without BC. A more parsimonious subset of 21 predictors satisfied a Scheffé-type multiple comparison criterion. Although duration of SCI satisfied a Bonferroni criterion for statistical significance, it did not satisfy a Scheffé criterion. In the replication studies, only dichotomized duration of indwelling catheterization for at least 10 years significantly replicated the previous findings. Results of this study suggest that using multiple risk factors and interactions in a comprehensive statistical model may provide useful screening of patients with SCI for BC risk. Since early identification of BC substantially improves prognosis, such a model may identify patients at highest risk who are most likely to benefit from bladder biopsy.