Preselection Tool for the Model-Based Selection of Head and Neck Cancer Patients for Proton Therapy

Abstract

In the Netherlands, head and neck cancer (HNC) patients are selected for intensity modulated proton therapy (IMPT) according to the ‘model-based approach’ (MBA) which is done by integrating the results of individual in silico planning comparison studies (ΔDose of protons versus photons) in NTCP-models to assess ΔNTCP (expected difference in toxicity rates). However, performing planning comparisons in all patients is labor intensive. Therefore, tools to identify patients that are most likely to benefit from protons are needed. The aim of this study was to develop a preselection tool (PST) for a more effective selection of patients for a planning comparison procedure. According to Dutch National Indication Protocol, NTCP-models for grade≥2 dysphagia, xerostomia, and tube feeding dependence are used for patient selection for IMPT. The ΔNTCP-thresholds used to select patients for IMPT are ≥10%, ≥5% and/or ≥15% for grade ≥2, grade ≥3, summed risk reduction (ΣΔNTCP) for grade 2 side effects, respectively. At present, a VMAT plan is created for each patient with optimal sparing of organ at risks (OARs) relevant for the ΔNTCP-profile (model-based optimization). Patients are only selected for a plan comparison if the NTCP-value of one of the 3 endpoints is higher than thresholds (OLD-PST). If that is the case, a robust IMPT-plan is optimized using the same criteria to assess ΔDose and to calculate the ΔNTCP-profile. The NEW-PST assumes a mean dose of 0 Gy in the OAR part that is outside the PTV+5mm. For the remaining part of OAR, the IMPT dose is assumed to be similar to that with VMAT to obtain adequate target coverage. For this study, an IMPT plan was created when patients qualified for protons according to the OLD-PST. Then, we investigated how many unnecessary plan comparisons could be avoided using the NEW-PST. The primary endpoint was: the patient qualifies for protons. 120 patients with HNC who would require a plan comparison according to OLD-PST were included in the study. A VMAT plan was made in 143 patients of which 120 (84%) were selected for plan comparison based on OLD-PST. 55 (46%) of these patients qualified for proton therapy based on plan comparison while for 65 (54%) patients unnecessary plan comparisons were performed. With the NEW-PST, unnecessary plan comparisons were avoided in 18 of these 65 patients. This translates into a reduction of 27% of unnecessary plan comparisons with the NEW-PST. No patient was denied a treatment plan comparison with the NEW-PST, who eventually qualified for protons. The sensitivity and specificity of the NEW PST were 100% and 47%, positive and negative predictive values were 54% and 100%, respectively. The NEW-PST predicts best scenario IMPT dose based on VMAT plan and reduced almost one third of unnecessary plan comparisons compared to current PST without wrongfully denying patients a plan comparison. This contributes to an even more cost-effective selection of patients for protons according to the model-based approach.