Predicting normal tissue injury in radiation therapy.

We tested three radiobiologic models, the nominal standard dose (NSD), the biologic index of reaction (BIR), and the linear quadratic (LQ) models to determine which best predicts normal tissue injury in radiation therapy. Clinical data for radiation myelopathy, rib fracture and pericardial effusion were used for all three models to predict injury. We assumed that on the average injuries occurred at higher equivalent doses of radiation than were received by patients who were not injured. We used a t-test to determine whether there were in fact significant differences in the mean values of the equivalent doses among the injured and non-injured. The means were calculated for the four sets of injury by the three models. For the LQ model it was necessary to choose a value for the parameter alpha/beta; the results were not sensitive to the choice over the range of 1/2 to 12 Gy. None of these models showed a significant difference between injured and non-injured patients for all four sets of data. The BIR model showed significant differences in three sets, the LQ model was significant in two and marginally significant in one set, and the NSD was significant in two sets. This analysis illustrates therefore, that the linear quadratic model can be adopted for analysis of clinical data with results that are no worse and possibly better than the NSD model.