Calculation of carcinogenic potency from long-term animal carcinogenesis experiments.

An index of carcinogenic potency for chemicals tested in chronic animal experiments is described. By analogy with the well-known 'lethal dose 50' (LD50) of quantal bioassay, a 'tumorigenic dose 50' (TD50) may be defined (in the absence both of tumors in the control group and of intercurrent deaths) as that (daily) dose of chemical which gives 50% of the test animals tumors by some fixed age. Tumors in the control (zero-dose) group are handled exactly as for the LD50, and intercurrent deaths are handled by life-table methods. Nonparametric procedures are developed for estimating the TD50 and for constructing confidence intervals. These are based on likelihoods which assume that the tumor hazard is linear in dose.