Demystifying cardiac dose in RTOG-0617
McWilliam, Alan Abravan, Azadeh Banfill, Kathryn
McWilliam, Alan
Abravan, Azadeh
Banfill, Kathryn
Citations
Altmetric:
Abstract
Purpose/Objective(s): The RTOG-0617 trial presented worse overall
survival for patients with lung cancer treated in the high-dose 74 Gy arm
compared to those treated with 60Gy. In multi-variable models, whole
heart volumetric dose parameters (volume of heart receiving at least 5Gy
and 30Gy), and the high-dose treatment arm was found to be associated
with survival, rescinding dose escalation. In this abstract, we repeat this
survival analysis, but consider cardiac sub-regions to find an explanation
for the counterintuitive survival difference observed between the treatment arms.
Materials/Methods: Voxel-wise analysis was performed on dose distributions spatially normalized onto a template patient anatomy. Permutation
testing assessed significance, identifying anatomical regions where dose
was associated with overall survival. Dose to the identified cardiac sub-region was included in multi-variable modelling. Clinical and dosimetric
variables were bootstrapped 500 times into an elastic-net variable selection
model to identify variables associated with survival. Variable importance
was assessed by (1) magnitude of model coefficients and (2) frequency of
selection. Next, multi-variable Cox regression survival models were created to assess significance of dose to the identified cardiac region compared to v5 and v30 parameters used previously.
Results: A total of 488 patients, with accurate dose mapping, were
included in the voxel-wise analysis. A significance region (P < 0.001) was
identified in the base of the heart, centered over the origin of the left coronary artery and atrioventricular node, with median dose 9.5 Gy (high dose
arm 10.2 Gy, low dose arm 9.1 Gy). Bootstrapping of clinical variables
identified mean lung dose (selected 99%), treatment arm (65%), log tumor
volume (66%) and heart region dose (62%) as most important factors associated with survival (v5 and v30 selected < 10%). Multi-variable models
included mean lung dose, log tumor volume, age, gender, treatment arm,
performance status, smoking history, PET staging and histology and either
heart v5, v30 or the identified heart region dose. Heart v5 and v30 were
not found to be associated with survival, with treatment arm (P < 0.05)
and log tumor volume (P < 0.02) being significant. However, heart region
dose was significant (P = 0.02), along with log tumor volume (P = 0.03).
Other parameters, including treatment arm were not significant.
Conclusion: These results show that dose to cardiac sub-regions is more
strongly association with overall survival than whole heart dose parameters. Variable selection showed the importance to include mean lung dose,
suggesting an interplay between lung and cardiac dose. The treatment arm
effect observed in RTOG-0617 lost its significant once the identified base
of heart region dose was included into the model. Our results suggest that
the worse survival of the high dose arm in RTOG617 could be explained
by the dose delivered to the base of the heart
Description
Date
2021
Publisher
Collections
Keywords
Type
Meetings and Proceedings
Citation
McWilliam A, Abravan A, Banfill K, Faivre-Finn C, van Herk M. Demystifying Cardiac Dose in RTOG-0617. International Journal of Radiation Oncology Biology Physics. 2021;111(3):S125-S.