Uncertainties in volume definition: Impact on NTCP/TCP estimations

The recent availability of tools which permit a true 3D reconstruction of the dosimetry data of radiotherapy treatments (such as the DVHs), makes it possible to obtain more accurate correlations between clinical data (local control and side effects, i.e. TCP and NTCP) and dose-volume data. However, the values of the parameters for the model which are now available (for instance the Emami-Burman parameters of the Lyman model) are still mainly based on the clinical experience before the 3D radiotherapy planning era; these values have to be considered as a zero point of biological modelling in radiotherapy. Modelling the behaviour of tumours and healthy tissues in radiotherapy will become more and more accurate with the time through careful correlation studies between clinical outcome and DVHs. These investigations have a number of intrinsic uncertainties; among them we wish to focus our attention on the uncertainty in volume definition. In general, DVHs calculated during the treatment planning phase refer to a 'static' situation without including the effect of organ deformation/movement and of set-up error which unavoidably occur during the execution of the treatment; moreover, treatment planning generally cannot take into account the uncertainty in contouring CTV/PTV and organs at risk, which may be estimated through inter-observer variability studies. Recent investigations have shown that these uncertainties may have a significant impact on the DVHs. Methods to incorporate them and/or to quantify their impact have been developed: in particular random errors can be incorporated through convolution methods. At the same time, methods to minimise their effect are becoming available, such as patient immobilisation, set-up correction procedures, breathing control techniques, rectum/bladder filling control. The effect of all these uncertainties on NTCP and TCP calculations in clinical situations have recently been analysed in a small number of investigations. The results of these studies are critically reviewed here. In the future, these uncertainties will have to be taken into account in refining the models so that NTCP/TCP estimations become more reliable.