Room: Exhibit Hall
Purpose: As dose calculation algorithms (DCAs) incorporate limitations and uncertainties, different dose distributions and dose-volume histograms can be predicted by disparate algorithms for the same treatment. This causes a problem for patient-specific dose distributions when estimating normal tissue complication probability (NTCP). Employing algorithm-specific NTCP model parameters can prevent errors stemmed from differences in DCAs. This study aims at evaluating the impact of radiobiology-varying parameters on NTCP calculation for lung complications, and determining how to alter model parameters for lung complications derived for a pencil beam (PB) dose calculation algorithm, to make them authentic for three other common DCAs.
Methods: Treatment plans for 15 breast cancer patients were retrospectively calculated using Monte Carlo (MC), superposition convolution (SP) and Clarkson (Clkr) algorithms to compare NTCP estimates for radiation pneumonitis with those derived from clinically used pencil beam algorithm (PB). For PB calculations, original model parameters were derived from previously published values for LKB model. For MC, SP and Clkr calculations, parameters were ï¬?tted to give the same NTCP as for the PB calculations.
Results: Changing the algorithm without deriving new model parameters caused alterations in NTCP value up to 10 percentage points for the cases investigated. Furthermore, the error introduced could be of the same magnitude as confidence intervals of calculated NTCP values. New NTCP model parameters were tabulated as the algorithm varied from PB to MC, SP, or Clkr. Moving from PB to MC, SP or Clkr necessitated a change in NTCP model parameters.
Conclusion: This paper demonstrates signiï¬?cant shifts of NTCP parameter values observed for LKB model, being comparable in magnitude to the uncertainties of published parameter values. Therefore, quoting the applied DCA when reporting estimates of NTCP parameters for ensuring the correct use of the model is important.