Probability Prediction Model of Mouse Skin Tumors Based on Chemical Components of Cigarette Smoke Condensates

摘要

Smoke condensates generated from a large number of cigarette blends have been painted on the backs of mice to try to produce histopathological verified tumors. Corresponding samples of each smoke condensate were also analyzed for the concentrations of selected chemical constituents. Relating the biological data to the chemical analyses sheds light on the undesirable smoke condensate constituents and can be used to design a less hazardous cigarette. This relationship is formulated as a mathematical model that predicts the probability that a mouse will be tumor free as a function of the concentration of the applied smoke condensate and eight of the chemical constituents of the smoke condensate. The prediction model is based on 196 data points from the four series of experiments conducted for the Towards A Less Hazardous Cigarette Program. Using a least squares analysis, a 17-term prediction model was found that predicted probabilities as well as a full 66-term second-order model. The prediction model accounted for 76.6% of the total variation and 70.4% of the observed values are within one standard deviation of the predicted values. (ERA citation 05:001951)