External validation of survival of lung cancer patients due to setup uncertainties towards the heart

Abstract

Purpose or Objective The impact on survival due to heart toxicity for lung cancer patients treated with radiotherapy (RT) has been difficult to quantify, due to correlations between dose to heart and lung with tumor burden. A possible measure that indicates enhanced dose to heart but does not correlate with the other risk factors is the deviation between actual and planned daily distance between isocenter and heart. The average of this daily distance (DeltaD), which can be obtained from CBCT scans, will for positive values indicate larger separation of target and heart and thus reduced heart dose during delivery. DeltaD has previously been reported to impact survival [Johnson-Hart et al IJROBP 2018]. The aim of the current study is to undertake an external validation of this finding in another institution Material and Methods All patients treated in the validation department from April 2010 to end 2015 with daily CBCT, planned dose of 60-66 Gy in 2 Gy fractions and for which a heart delineation was available were collected for analysis. Standard clinical IGRT procedure utilized two registrations for each CBCT 1) tumor region (mask) and 2) overall anatomy (clipbox). The first was used for patient positioning while the latter was used for overall validation of the anatomy. The latter is representative of the heart position, while the first represents the tumor position hence the treatment isocenter. The difference between the registrations can thus be used to estimate the daily shift of the heart relative to the isocenter. Based on clinical data (performance status, GTV volume, Age, T and N stage, histology, tumor stage and dose) a base Cox model predicting survival was created using Lasso for parameter selection. Impact of continuous variable DeltaD was tested by adding DeltaD to the base Cox model. All analyses were performed in R v. 3.6.1. Results 300 patients were included. DeltaD did not correlate statistically significant with any of the clinical factors. Range of DeltaD was -0.6 cm to 0.7 cm. Figure 1 shows Kaplan-Meier plots of patients with positive versus negative DeltaD. It is seen that a split of the curves occurs around 16 months after the start of RT. The base Cox model included performance status, ln(GTV volume) and Age. DeltaD was included in the base model using time dependent Cox regression for which DeltaD effectively was zero until 16 months after RT. This resulted in a statistically significant regression constant of DeltaD of - 1.63 per cm (HR=0.195) with p-value of 0.017 Conclusion As in the original study this validation study demonstrates significant impact on survival due to estimated setup errors in the direction of the heart, even with daily online IGRT corrections. This indicates that dose to heart impacts survival for these patients. However, in contrast to the original study the effect of DeltaD only starts 16 months after RT. This difference in the two studies might reflect differences in treatment or patient cohorts at the two centers, and calls for additional external validation.