Predicting the response of head and neck cancers to radiation therapy with a multivariate modelling system: an analysis of the RTOG head and neck registry.

Two thousand and sixty-six patients with primary head and neck cancers were entered in the RTOG Head and Neck Cancer Registry between 1977 and 1980. Nine hundred and ninety-seven (997) evaluable patients were treated initially with radiation therapy alone. Tumor site, T-stage, N-stage, histology, degree of infiltration, degree of differentiation, patient age, sex and Karnofsky performance score were all prospectively recorded. T-stage (p less than .001), N-stage (p = .007), primary site (p less than .001), and initial Karnofsky performance score (p less than .001) proved to be significant factors independently predictive of primary tumor clearance. Using these factors, a multivariate response model was constructed to predict primary tumor response. The predictive accuracy of the model proved to be highly reliable, and was tested by comparing the predicted vs. observed complete tumor clearance rates for each independent variable. For primary sites, the predicted number of complete responses vs. observed were: oral cavity, 139.8 predicted vs. 139 observed; nasopharynx, 51.8 predicted vs. 51 observed; oropharynx, 174.8 predicted vs. 176 observed, supraglottic larynx, 314.4 predicted vs. 318 observed; glottic larynx, 314.4 predicted vs. 318 observed; and hypopharynx, 49.4 predicted vs. 46 observed. For the entire group of patients predicted to have a 90% or better complete primary tumor response (including T3 and T4 tumors), 94% remained in initial complete remission at the primary site at one year, and 87% at two years. An accurate multivariate response model, such as the one presented in this paper, should prove to be a useful tool in selecting patients with head and neck cancers suitable for treatment with radiation therapy alone.