ESTRO 2024 - Abstract Book

S4400

Physics - Intra-fraction motion management and real-time adaptive radiotherapy

ESTRO 2024

The SMART trial enrolled 136 patients across 13 sites in 3 countries. Key eligibility criteria included pathologically confirmed adenocarcinoma, BRPC or LAPC as per institutional definition, and receipt of > 3 months of any systemic therapy without distant progression The required prescribed dose was 50 Gy in 5 fractions (biologically effective dose [BED10] = 100 Gy). The highest priority for treatment planning was to meet protocol-specified organ-at-risk (OAR) constraints and secondarily to optimize target volume coverage. All patients were treated on an integrated 0.35T MRgRT system with on-table adaptive re-planning as needed. OARs within 3 cm of the PTV were re-contoured on the daily MRI. During optimization, a PTV opt structure was created by subtracting the gastrointestinal (GI) luminal planning at-risk volumes (PRVs) from the PTV. The plan was adapted if >1 cc of any GI OAR exceeded 33 Gy and/or if target coverage was insufficient. Centralized quality assurance (QA) review was not required for original or adapted plans. Retrospective analysis was performed for all adapted plans. Plan quality metrics evaluated for dose conformity included in this study but not the original protocol: homogeneity index (HI=PTV opt D 2% / D 98% ), high dose conformity index (CI=100% Rx isodose volume/PTV opt volume), and gradient index (R 50% =50% Rx isodose volume/PTV opt volume). Target coverage metrics including D max , D 99% , D 95% , D 90% and D 80% for PTV opt were evaluated. To quantify poor vs. favorable anatomy, the amount of GI luminal OARs overlapping the PTV during adaptation was assessed using a ratio of PTV opt /PTV. A PTV opt /PTV of 1 indicates no overlap of GI PRVs with PTV, while 0 indicates complete overlap. This adaptive anatomical geometry (i.e., PTV opt /PTV) was correlated with CI and D95%. CI was categorized as poor (CI<0.66), fair (0.660.83) based on the acceptability criteria in RTOG 09156. Coverage was categorized as poor (D 95% <90% Rx), fair (90% Rx95% Rx) based on clinical criteria. Statistically significant differences of poor vs. favorable anatomical geometry based on the median PTV opt /PTV were assessed using the Kolmogorov-Smirnov test for all plan quality metrics. Of the total 680 delivered fractions, 616 fractions (90.5%) were adapted. Mean + SD PTV opt /PTV was 0.89 ± 0.09, indicating that overlap with the GI OARs reduced the optimized ablative target volume by 10% on average. For target coverage metrics, mean ± SD [min, max] across all adaptive plans were 65.0 ± 4.8 [50.3, 79.4] Gy (PTV opt D max ), 41.3 ± 5.4 [17.5, 53.6] Gy (PTV opt D 99% ), 45.8 ± 4.6 Gy [20.7, 56.1] (PTV opt D 95% ), 48.1 ± 3.8 [22.1, 57.3] Gy (PTV opt D 90% ), and 50.7 ± 2.9 [24.8, 59.5] Gy (PTV opt D 80% ), respectively. For conformity metrics, mean ± SD were 0.97 ± 0.21 [0.4, 2.1] (CI), 4.11 ± 1.11 [2.3, 10.2] (R 50% ), and 1.45 ± 0.24 [1.1, 3.3] (HI), respectively. Kolmogorov-Smirnov demonstrated statistical significance (p<0.05) between poor and favorable adaptive geometry plans for all plan quality metrics except Dmax and HI. For conformity, poor/fair/good CI correlated with increasing PTV opt /PTV (i.e., reduced GI PRV overlap to the PTV geometry). Specifically, poor CI (n=25, median=0.601), fair CI (n=131, median=0.770), and good CI (n=460, median=1.017) had median PTV opt /PTV of 0.88, 0.89, 0.92, respectively. Conversely, only fair (n=180, median=46.3 Gy) and good (n=239, median=49.4 Gy) PTV D95% correlated with reduced GI PRV overlap to PTV with respective median PTV opt /PTV of 0.88 and 0.93. While poor PTV D95% (n=197, median=40.9 Gy) did not correlate with greater GI PRV overlap to PTV (i.e., median PTV opt /PTV of 0.91). As such, poor target coverage did not correspond to unfavorable GI luminal anatomy. Results:

Conclusion:

The quality of on-table adapted plans from the phase 2 SMART trial was generally favorable although the quality of some plans was poor. Loss of conformity correlated with increasing level of overlap between GI luminal OAR and target. Similarly, adapted plans with good to fair target coverage correlated with an increasing amount of GI OAR to PTV overlap. However, adapted plans with poor target coverage did not correlate with unfavorable GI anatomy and the reasons for this are uncertain. These findings highlight the need for centralized QA in future clinical trials and we plan to evaluate whether these dosimetric results are significantly associated with clinical outcomes.