Alternative models for estimating the radiotherapy retreatment dose for the spinal cord

To review the available experimental animal and patient data on response of the spinal cord to re-irradiation in order to identify appropriate data sets to investigate the clinical potential of models that would allow evaluation of the increase in the retreatment dose with elapsed time from the initial exposure.

Analysis of published data on irradiated rat and primate spinal cord identified results for the rat cervical spinal cord that could be compared, where the development of myelopathy was caused by selective white matter necrosis. This data, although limited, provide some important insights. Two models, derived from simple differential equations, provide a time- and dose-dependency for recovery and could be fitted to these data. These models predict the remaining tolerance, in a phase space above the line that connects the 100% biological effectiveness (BEDTOL) tolerance dose of the first and second treatment courses when these are plotted together. A third, much simpler, linear model, assumed that recovery was time but not initial dose dependent.

The experimental results showed a non-linear time dependency for the change in biological effectiveness (BED) of the re-irradiation dose. Comparison of the three different models paid particular attention to changes in the re-irradiation dose, when the initial radiation dose was either low or high. For each model, cautious data interpretations were also introduced to reduce the effects of the near completeness of recovery with time derived from the important experiments with primates, which include few data points. Model 1 predicts the least recovery following low initial doses, but with greater recovery following larger initial doses. Model 2 allowed no further irradiation after an initial full tolerance dose, but also greater than expected recovery following the use of smaller priming doses. Model 3 gives unrealistically high doses when used after an initial full tolerance irradiation dose.

These results show that it is possible to model these time-dependent relationships for the spinal cord and that Model 1 is probably the most realistic, especially when it is used conservatively. To give greater confidence as to which of the three presented methods is best, further experiments and/or more analysis of human data are necessary. In the meantime clinicians will need to exert caution and judgement as to the choice of the re-irradiation BED, bearing in mind the other clinical factors that influence radio-tolerance. Further research is necessary to provide the safest recommendations and best clinical outcomes. Some suggestions as to what needs to be done are given.