Session Item

In October 2019, we started to treat NSCLC patients with intensity modulated proton radiotherapy (IMPT). We planned to initiate a dose escalation RCT, but we felt that this should only start beyond the steepest part of the learning curve. Therefore, the objective of this study was to explore our learning curve in terms of treatment planning and delivery.

The model-based approach was used to select patients for IMPT. Model-based selection (IMPT compared to VMAT) was based on 2 NTCP-models (Grade ≥2 acute esophageal toxicity (AET) and Grade ≥2 radiation pneumonitis (RP)) and one prediction model for 2-year mortality (2yM). All patients received (chemo)radiation (25x2.4 Gy) for NSCLC. Robust IMPT plans were created using typically three beams and a 3D robustly optimized planning technique (5 times layered rescanning). Initially, maximum target motion allowed for IMPT was 10 mm, from March 2020 on this was extended to 15 mm. Online position verification consisted of daily (consecutive) body surface scanning, 2D kV imaging and cone-beam CT imaging. Weekly repeated (4D) CTs were acquired; the treatment plan was adapted in case of inadequate target coverage visualized on the repeated CT. The following characteristics were evaluated: ΔDose (mean doses for heart, lungs and esophagus) and ΔNTCP (for AET, RP and 2yM), all based on the nominal treatment plan. Treatment delivery times at the gantry and the number of plan adaptations were evaluated also. To evaluate the learning curve, the cohort was split in two: Cohort A ranged from October 2019 to March 31^st 2020 (n=24), Cohort B ranged from April 2020 to September 2021 (n=90).

For patient and treatment characteristics, see Table 1. After radiotherapy plan comparison (IMPT vs VMAT), the median ΔDose (mean) for heart (3.1 Gy vs 4.2 Gy, p=0.087), lung (2.9 Gy vs 3.4 Gy, p=0.13) and esophagus (2.4 Gy vs 2.8 Gy, p=0.48) was larger in Cohort B compared to Cohort A. Median ∆NTCP increased in Cohort B compared to Cohort A: from 4.9% to 6.0% (2yM, p=0.26), 3.2% to 5.7% (RP, p=0.06) and 3.3% to 4.7% (AET, p=0.49), respectively (figure 1). Treatment delivery times decreased from an average of 34 minutes (Cohort A) to 25 minutes (Cohort B). Eleven (46%) of the treatment plans in Cohort A needed adaptation, compared to 20 (22%) in Cohort B. Distribution of the timing of these plan adaptations were equally balanced over the radiotherapy course and were similar for both cohorts. Comparing the first half of Cohort B with the second half of Cohort B: ΔDose, ∆NTCP and treatment delivery parameters were similar for both groups.

The observed increase in ΔDose and ΔNTCP, and more efficient treatment delivery over time indicates that a learning curve was present after the introduction of IMPT. Awareness of the learning curve is important in further optimizing treatment. This may have implications for the initiation of clinical studies while still on the steep part of the learning curve.