Comparative analysis of dose volume histogram reduction algorithms for normal tissue complication probability calculations

A model for estimating radiotherapy treatment outcome through the probabili ty of damage to normal tissue and the probability of tumour control is a us eful tool for treatment plan optimization, dose escalation strategies and o ther currently used procedures in radiation oncology. Normal tissue complic ation estimation (NTCP) is here analysed from the point of view of the reli ability and internal consistency of the most popular model. Five different dose volume histogram (DVH) reduction algorithms, applied to the Lyman mode l for NTCP calculation, were analysed and compared. The study was carried o ut for sets of parameters corresponding to quite different expected dose-re sponse relationships. In particular, we discussed the dependence of the mod els on the parameters and on the dose bin size in the DVH. The sensitivity of the different reduction schemes to dose inhomogeneities was analysed, us ing a set of simple DVHs representing typical situations of radiation thera py routine. Significant differences were substantiated between the various reduction methods regarding the sensitivity to the degree of irradiation ho mogeneity, to the model parameters and to the dose bin size; Structural asp ects of the reduction Formalism allowed an explanation for these difference s. This work shows that DVH reduction for NTCP calculation has still to be considered as a very delicate field and used with extreme care, especially for clinical applications, at least until the actual formulations are tuned against strong clinical data.