Influence of different dose calculation algorithms on the estimate of NTCP for lung complications

Emma Hedin,Anna Bäck

doi:10.1120/jacmp.v14i5.4316

Abstract

Due to limitations and uncertainties in dose calculation algorithms, different algorithms can predict different dose distributions and dose‐volume histograms for the same treatment. This can be a problem when estimating the normal tissue complication probability (NTCP) for patient‐specific dose distributions. Published NTCP model parameters are often derived for a different dose calculation algorithm than the one used to calculate the actual dose distribution. The use of algorithm‐specific NTCP model parameters can prevent errors caused by differences in dose calculation algorithms. The objective of this work was to determine how to change the NTCP model parameters for lung complications derived for a simple correction‐based pencil beam dose calculation algorithm, in order to make them valid for three other common dose calculation algorithms. NTCP was calculated with the relative seriality (RS) and Lyman‐Kutcher‐Burman (LKB) models. The four dose calculation algorithms used were the pencil beam (PB) and collapsed cone (CC) algorithms employed by Oncentra, and the pencil beam convolution (PBC) and anisotropic analytical algorithm (AAA) employed by Eclipse. Original model parameters for lung complications were taken from four published studies on different grades of pneumonitis, and new algorithm‐specific NTCP model parameters were determined. The difference between original and new model parameters was presented in relation to the reported model parameter uncertainties. Three different types of treatments were considered in the study: tangential and locoregional breast cancer treatment and lung cancer treatment. Changing the algorithm without the derivation of new model parameters caused changes in the NTCP value of up to 10 percentage points for the cases studied. Furthermore, the error introduced could be of the same magnitude as the confidence intervals of the calculated NTCP values. The new NTCP model parameters were tabulated as the algorithm was varied from PB to PBC, AAA, or CC. Moving from the PB to the PBC algorithm did not require new model parameters; however, moving from PB to AAA or CC did require a change in the NTCP model parameters, with CC requiring the largest change. It was shown that the new model parameters for a given algorithm are different for the different treatment types.PACS numbers: 87.53.‐j, 87.53.Kn, 87.55.‐x, 87.55.dh, 87.55.kd

Highlights

Radiation therapy treatments are designed and optimized by considering both tumor control probabilities and normal tissue complications
The objective of this work is to determine how to change the normal tissue complication probability (NTCP) model parameters for lung complications derived for a simple correction-based pencil beam dose calculation algorithm, PB, in order to make them valid for collapsed cone (CC), analytical anisotropic algorithm (AAA), and pencil beam convolution (PBC), using the method described by Brink et al(7) Model parameters for two NTCP models, relative seriality (RS)(9) and Lyman-Kutcher-Burman (LKB),(10,11) are collected from different published studies on different grades of pneumonitis.(1-3,6) The results for CC is compared to the results from Brink et al(7) and De Jaeger et al(6) Possible differences in the NTCP model parameters between
The results are inconsistent for the tangential breast (Tang) plans in Eclipse, as in this case a PBC-to-AAA change sometimes results in a higher NTCP value

Summary

Introduction

Radiation therapy treatments are designed and optimized by considering both tumor control probabilities and normal tissue complications. A NTCP model uses a mathematical expression to describe the relationship between the delivered physical dose and the biological effect in normal tissue. Model parameters are empirically derived by fitting the NTCP predictions for a specific endpoint to the observed clinical outcome for a population of treated patients. The patient population includes individual patient-specific variations (such as radiation sensitivity) and a large number of patients must be included to achieve good precision in the model parameter values. The accuracy of the NTCP estimates depends on the accuracy of the assessment of the delivered dose, as well as uncertainties related to the clinical data material (e.g., difficulties in diagnosis). A low prevalence of the endpoint studied results in poor statistics

Objectives

Methods

Results

Discussion

Conclusion