Bladder cancer has a 70% recurrence rate within five years and a high associated mortality. It commonly occurs in one or both of two predominant growth/behaviour patterns: either well-differentiated, relatively benign exophytic papillary lesions, or flat, poorly differentiated invasive carcinoma usually arising from carcinoma-in-situ. We have used the F344 rat treated with N-butyl-N-(4-hydroxybutyl)nitrosamine (BBN) as a model for the papillary disease, and the BBN-treated B6D2F1 mouse for flat, invasive bladder carcinoma. In the rat, carcinogenesis is a multistage process and several retinoids will delay or even halt the development of bladder cancer. Inhibition of carcinogenesis is not complete, but there is a consistent reduction in the time-related incidence of papillomas and carcinomas and a concomitant improvement in the overall differentiation of the urothelium. In the BBN/mouse model, retinoids also have anticarcinogenic activity but interpretation of the results is more complicated. Unlike the F344 rat, the B6D2F1 mouse has a non-uniform response to BBN; not all mice develop bladder cancer even after treatment with very high doses of BBN and in those that do, more than one mechanism of carcinogenesis may be involved. Individual retinoids differ markedly in their ability to modulate bladder carcinogenesis in rodents; the behaviour of one analogue cannot be predicted automatically from data obtained with another. Combined data from rodent trials in this and other laboratories have identified N-(4-hydroxyphenyl)retinamide (HPR) as the most anticarcinogenic retinoid tested so far for the rodent bladder. It is also less toxic in rodents and better tolerated in humans than either 13-cis-retinoic acid or etretinate, two retinoids currently used in dermatological practice. A prophylactic chemopreventive trial of HPR in bladder cancer patients starting in 1985 will be centered on the Middlesex Hospital, London.