The effect of patient-to-patient variability on the accuracy of predictive assays of tumor response to radiotherapy: a theoretical evaluation.

Mathematical modeling was used to investigate the relative accuracy that might be expected from predictive assays of tumor response to radiotherapy based on one of the following four tumor characteristics: intrinsic tumor-cell radiosensitivity, doubling time of the clonogenic tumor cells, number of clonogens in the tumor at the start of treatment, and extent of hypoxia in the tumor. In particular, the influence of inter-tumor heterogeneity on predictive accuracy was investigated. Wide patient-to-patient variability in a tumor characteristic contributing to treatment response adds to the accuracy of a predictive assay based on that characteristic, but variability from patient to patient in factors influencing response, but not measured by the assay, has a confounding effect and reduces predictive accuracy. The results of this theoretical study suggest that predictive assays based on intrinsic tumor-cell radiosensitivity are much more likely to be correlated with clinical outcome than are assays based on clonogen doubling time, hypoxic fraction, or clonogen number, since individual differences in tumor radiosensitivity can seriously confound assays based on the other factors. It is concluded that it may be necessary to correct for individual differences in intrinsic radiosensitivity before predictive assays based on other tumor characteristics, such as potential doubling time, might have any detectable clinical significance.