The first James Kirk memorial lecture. What next in fractionated radiotherapy?

Models for predicting the total dose required to produce tolerable normal-tissue injury are becoming less empirical, more realistic, and more specific for different tissue reactions. The trend can be seen by the progression from the "cube root law", through Strandqvist's slope of 0.22, to NSD, TDF and CRE which have separate time and fraction number exponents, to the even better approximations which are now available. The dose-response formulae that can be used, with statistical legitimacy, to define the effect of fraction size (and number) include (1) the linear quadratic(LQ) model; (2) the two-component (TC) multi-target model; and (3) repair - misrepair models. The LQ model offers considerable convenience and requires only two parameters to be determined. The use of a new model often provides fresh insights. The LQ model has emphasized the difference between late and early normal-tissue dependence on dose per fraction which was first shown by exponents greater than the NSD slope of 0.24. Exponents of overall time, e.g. T0.11, yield the wrong shape of time curve, suggesting that most proliferation occurs early, although it really occurs after a delay depending on the turnover time of the tissue. The principles of better time factors are well known but actual values for human tissues are not well determined. Fortunately the time factors are usually small, especially for late reactions. Improved clinical results are being sought by hyperfractionation, by accelerated fractionation, or by continuous low dose rate irradiation as in interstitial implants. New clinical trials are investigating these approaches, which have been suggested by the accumulation of radiobiological data.