What should we know about photon dose calculation algorithms used for radiotherapy? Their impact on dose distribution and medical decisions based on TCP/NTCP

The dose calculation algorithms, integrated in a radiotherapy treatment planning system, use different approximations to swiftly compute the dose distributions. Any biological effect is somehow related to the dose delivered to the tissues. Thus, the optimization of treatment planning in radiation oncology requires, as a basis, the most accurate dose calculation to carry out the best possible prediction of the Normal Tissue Complication Probability (NTCP), as well as Tumor Control Probability (TCP). Presently, a number of bio-mathematical models exist to estimate TCP and NTCP from a physical calculated dose using the differential dose volume histogram (dDVH). The purpose of this review is to highlight the link between any change of algorithms and possible significant changes of DVH metrics, TCP, NTCP and even more of estimated Quality-adjusted life years (QALY) based on predicted NTCP. The former algorithms, such as pencil beam convolution (PBC) algorithm with 1D or 3D density correction methods, overestimated the TCP while underestimating NTCP for lung cancer. The magnitude of error depends on the algorithms, the radiobiological models and their assumed radiobiological parameters setting. The over/under estimation of radiotherapy outcomes can reach up to 50% relatively. Presently, the anisotropic analytical algorithm (AAA), collapsed cone convolution algorithm (CCC), Acuros-XB or Monte Carlo are the most recommended algorithms to consistently estimate the TCP/ NTCP outcomes and QALY score, to rank and compare radiotherapy plans, to make a useful medical decision regarding the best plan. This paper points out also that the values of the NTCP radiobiological parameters should be adjusted to each dose calculation algorithm to provide the most accurate estimates.