Cardiac substructure radiotherapy dose and changes in physical activity and quality of life after chemoradiotherapy for NSCLC: a secondary analysis of the CLARITY prospective study.

Introduction
There is an increasing emphasis on protecting cardiac substructures during radiotherapy planning for locally advanced non-small cell lung cancer (NSCLC). [1,2]. Prior studies have shown distinct associations between cardiac substructure radiotherapy dose and cardiac events [[3], [4], [5], [6], [7]]. Notably, there is a paucity of data examining associations between cardiac substructure dose and declines in patient-reported outcomes (PROs) post-radiotherapy. Two prior studies found associations between whole heart dose and declines in physical activity and quality of life metrics, but cardiac substructure dose was not analyzed and sample sizes were modest [8,9]. The purpose of this study was to determine the longitudinal changes in PROs after contemporary chemoradiotherapy, and to determine the associations between cardiac substructure radiotherapy dose and changes in physical activity and quality of life (fatigue and dyspnea). Additionally, associations between physical activity and quality of life after chemoradiotherapy were assessed.

Methods and materials

Patient population
The study population was a subcohort of CLARITY (NCT04305613), a multi-institutional longitudinal prospective cohort study of patients with NSCLC treated with definitive radiotherapy +/− chemotherapy and immunotherapy. Patients completed standardized physical activity and quality of life questionnaires at baseline, completion of radiotherapy, 6 months post-radiotherapy, and 12 months post-radiotherapy. Participants included in this analysis completed questionnaires at baseline and at least 1 additional follow-up time point. Only patients who received conventionally fractionated radiotherapy (1.8–2 Gy per fraction) with curative intent for locally advanced or oligometastatic NSCLC were included. The study was approved by the central institutional review board (IRB) at the University of Pennsylvania and local site IRBs. All participants provided written, informed consent.

Physical activity and quality of life scales
At each time point, self-reported physical activity was assessed via the Godin-Shephard Leisure-Time Physical Activity Questionnaire (Godin), which quantifies the amount and type of physical activity reported (strenuous, moderate, mild) [10]. Quality of life metrics of fatigue and dyspnea were assessed via the Functional Assessment of Chronic Illness Therapy Fatigue and Dyspnea Scales (hereafter referred to as FACIT-Fatigue and FACIT-Dyspnea) [11]. Higher Godin scores indicate greater physical activity, higher FACIT-Fatigue scores less fatigue, and higher FACIT-Dyspnea raw scores worse dyspnea.

Radiotherapy delivery and cardiac dose metrics
Participants were simulated using 4-dimensional computed tomography or deep inspiration breath hold and were treated primarily with intensity-modulated radiotherapy or proton beam therapy. Digital Imaging and Communications in Medicine (DICOM) radiotherapy records were centrally compiled in the Varian Eclipse treatment planning system (versions 13.0–16.0, Varian Medical Systems, Palo Alto, CA). TheraPanacea 2.1, a clinically validated software was used to auto-segment the whole heart, left ventricle (LV), right ventricle (RV), left atrium (LA), and right atrium (RA) [12,13]. MIM Maestro (version 6.9.6) software was used to auto-segment the left anterior descending coronary artery (LAD). Each auto-segmented contour was manually reviewed and edited as needed by 2 investigators (IH and NYR). For each cardiac structure, the following dose metrics were extracted: mean dose (Gy), minimum dose to the hottest 0.03 cc (D0.03 cc [Gy]; i.e., maximum dose), volume receiving ≥ 5 Gy (V5Gy [%]), V15Gy, and V30Gy. These substructures were chosen a priori as they each have been associated with cardiac events [[4], [5], [6],14]. A total of 30 cardiac dose metrics were tested (6 structures, 5 dose metrics per structure). These dose metrics were chosen as they represent a broad representation of low, medium and high dose exposure to each structure.

Statistical analysis
Clinical characteristics and radiation dose metrics were summarized using medians with interquartile ranges [median (Q1, Q3)] for continuous variables and counts with percentages [n (%)] for categorical variables. Missing baseline Godin scores for 7 patients were imputed using multiple imputation via predictive mean matching, leveraging available Godin scores from subsequent visits.
Longitudinal changes in PROs were assessed using repeated-measures linear regression via generalized estimating equations (GEE) to account for intra-subject correlations. Joint Wald tests were used to evaluate whether PRO scores changed significantly over time.
Associations between radiation dose metrics and changes in PRO scores were assessed using repeated-measures linear regression via GEE, with results presented as coefficients and 95% confidence intervals [CI]. Given the large number of a priori identified cardiac radiation dose metrics of potential importance (n = 30 cardiac dose metrics), feature selection using least absolute shrinkage and selection operator (LASSO) regression with nested cross-validation was applied to further define variable importance [15]. A 10-fold outer cross-validation combined with a 5-fold inner cross-validation using GridSearchCV was used to optimize the regularization parameter. Radiotherapy dose metrics with non-zero coefficients were identified as significant predictors and subsequently evaluated in multivariable GEE models. Mean heart dose, which was hypothesized to be of importance based on clinical judgement a priori, was also evaluated in multivariable models.
Univariable GEE models were fit to evaluate for potential clinical covariates associated with changes in PROs. Covariates with p-values < 0.20, along with clinical judgement, were used to guide covariate selection. Final covariate adjustments included age, sex, timepoint of PRO assessment (baseline, end of radiotherapy, 6 months post-radiotherapy, and 12 months post-radiotherapy), baseline cardiovascular disease (defined as a history of myocardial infarction, arrhythmia, heart failure, coronary disease, percutaneous coronary intervention, cardiac surgery, pacemaker, defibrillator, or stroke), lung mean dose, gross tumor volume (GTV), prior lung surgery, recruiting center, and baseline PRO score. Covariate adjustment was consistent across all multivariable models. An identity link function with an independent working correlation structure was used and a correction for multiple testing was applied using the Holm-Bonferroni method.
All analyses were performed by R 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided p-value < 0.05 was considered statistically significant.

Results
A total of 122 participants were included. Table 1 shows baseline characteristics. The median age was 67 years and 57% were male. Forty-one percent of patients had baseline cardiovascular disease. Median radiotherapy dose/fractionation was 60 Gy in 30 fractions. Most were treated with intensity modulated radiotherapy (IMRT) (51%), volumetric modulated arc therapy (VMAT) (24%) or proton therapy (25%). All patients received concurrent chemotherapy, and 70% received consolidative immunotherapy following radiotherapy. Median GTV was 92 cc. Median whole heart mean dose was 9 Gy, median whole heart maximum dose 64 Gy, median LAD V15Gy 1%, and median lung mean dose 14.5 Gy (Table 2).
Godin physical activity (p = 0.0499), FACIT-Fatigue (p < 0.001), and FACIT-Dyspnea scores (p = 0.0037) worsened from baseline to end of radiotherapy, then recovered to baseline levels thereafter (Fig. 1, Supplementary Table 1). Godin scores, in particular, were very low. In multivariable analysis, an improvement in Godin physical activity scores over time was associated with an improvement in FACIT-Fatigue scores (p = 0.023, Table 3). There was no association between baseline Godin physical activity scores and changes in FACIT-Fatigue or FACIT-Dyspnea scores.
LASSO regression identified several candidate cardiac substructure dose metrics that were of potential importance in determining the variance in PROs (Fig. 2). In multivariable analysis (Table 4), whole heart maximum dose was initially associated with a worsening in both FACIT-Fatigue score (β −0.07 per 1 Gy increase, 95% CI −0.14 – 0.00, p = 0.0499) and FACIT-Dyspnea score (β 0.04 per 1 Gy increase, 95% CI 0.00 – 0.08, p = 0.048) (Supplementary Fig. 1). However, after adjustment for multiple comparisons, no cardiac dose metric was significantly associated with changes in Godin physical activity, FACIT-Fatigue or FACIT-Dyspnea.

Discussion
In 122 participants with NSCLC treated with definitive, conventionally fractionated chemoradiotherapy, patient-reported physical activity, fatigue, and dyspnea worsened by the end of treatment but recovered by 6 months post-treatment. After adjustment for multiple comparisons, no cardiac substructure dose metric was associated with changes in these PROs.
We hypothesize the following potential explanations for the lack of significant association between cardiac dose metrics and changes in PROs. First, modern radiotherapy techniques were used: 99% were treated with IMRT, VMAT or proton therapy. Consequently, median mean heart dose/LAD V15Gy were only 9 Gy/1% in our cohort compared to 15  Gy/38% in RTOG 0617 and 12.3 Gy/13.8% in Atkins et al. in which many patients were treated with 3-dimensional conformal radiotherapy [5,16]. Second, although potentially informative, PROs are not cardiac specific, and thus may not correlate well with subclinical or cardiac dysfunction. PROs can be affected by numerous factors in a lung cancer population, including smoking history, performance status, GTV, chemotherapy or immunotherapy use, prior lung surgery, and lung radiation dose. Future analyses in the CLARITY cohort will evaluate the associations between cardiac dose and changes in quantitative measures of cardiac function, and likely lend further insight into the clinical relevance of cardiac dose.
Krishnan et al. found that among 50 patients with lung cancer and lymphoma, whole heart V5Gy was associated with decreased physical activity but not with changes in fatigue or dyspnea [8]. Paul et al. found that among 46 patients with locally advanced lung cancer, every 10 Gy increase in mean heart dose was associated with a 3.1% reduction in step count per week during chemoradiotherapy [9]. In contrast to this current study, these two studies had relatively smaller sample sizes which precluded robust covariate adjustment. Moreover, our cohort, reflecting current trends in care, had a lower mean heart dose (median 9 Gy versus 11.4 Gy in Paul et al.) and a median LAD V15Gy of only 1%.
Limitations of the current study include a modest sample size in the context of multiple testing, as well as a limited follow up of 12 months which may be insufficient to capture late PRO decline. Longer follow up for PRO assessment at 24 months is ongoing. However, important strengths are also noted: the prospective design, contemporary cohort, centralized contouring, and robust statistical analysis using LASSO, multiple testing correction, and careful consideration of confounders, including lung dose and GTV.
In conclusion, cardiac substructure radiation dose was not significantly associated with changes in PROs after thoracic chemoradiotherapy. Future analyses of the CLARITY study will assess more specific measures of cardiac injury (e.g., quantitative biomarkers, echocardiogram parameters) to further understand the clinical significance of cardiac radiation dose in this population.

Credit Author Statement
Conceptualization: NYR, KK, WZ, SJF, BK
Data curation: ISH, SH, JK, JW, OF, AMS, BK
Formal analysis: NYR, KK, BK
Funding acquisition: BK
Investigation: NYR, KK, ISH, WZ, BK
Methodology: NYR, KK, BK
Project administration: AMS, BK
Resources: JDM, NO, SKJ, RHM, CR, WPL, RC, SNN, LS, MS, AJK, SJF, BK
Software: KK, BK
Supervision: BK
Validation: All authors
Visualization: KK, BK
Writing – original draft: NYR, KK, BK
Writing – review and editing: All authors

Conflicts of Interest/Disclosures
BK: Grants: NIH, American Heart Association
Other funding: Pfizer, Roche, American College of Cardiology
GW: Honoraria for contributing to non-promotional educational events, conference analysis and an advisory board for AstraZeneca. Grants from CRUK and IASLC.
JM: Consulting/Advisory Boards for Alnylam, AstraZeneca, BridgeBio and Pfizer, unrelated to the contents of the manuscript. Research funding from Abbott Laboratories and Myocardial Solutions, unrelated to the contents of the manuscript.
AK: Consulting for Medtronic
All other authors did not report any disclosures.

Data availability
Research data are stored in an institutional repository and may be shared upon reasonable request to the corresponding author.

CRediT authorship contribution statement
Nikhil Yegya-Raman: Writing – review & editing, Writing – original draft, Validation, Methodology, Investigation, Formal analysis, Conceptualization. Kyunga Ko: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Methodology, Investigation, Formal analysis, Conceptualization. Ivy S. Han: Writing – review & editing, Validation, Investigation, Data curation. Joshua D. Mitchell: Writing – review & editing, Validation, Resources. Wei Zou: Writing – review & editing, Validation, Investigation, Conceptualization. Nitin Ohri: Writing – review & editing, Validation, Resources. Salma K. Jabbour: Writing – review & editing, Validation, Resources. Raymond H. Mak: Writing – review & editing, Validation, Resources. Clifford Robinson: Writing – review & editing, Validation, Resources. William P. Levin: Writing – review & editing, Validation, Resources. Leanne Barrett: Writing – review & editing, Validation. Congying Xia: Writing – review & editing, Validation. Eva Berlin: Writing – review & editing, Validation. Paco Bravo: Writing – review & editing, Validation. Marcelo Di Carli: Writing – review & editing, Validation. Roger Cohen: Writing – review & editing, Validation, Resources. Sandra Hutton: Writing – review & editing, Validation, Data curation. Jonathan Keltz: Writing – review & editing, Validation, Data curation. Jessica Wang: Writing – review & editing, Validation, Data curation. Omotayo Fasan: Writing – review & editing, Validation, Data curation. Suneel N. Nagda: Writing – review & editing, Validation, Resources. Angie Seo: Writing – review & editing, Validation. Amanda M. Smith: Writing – review & editing, Validation, Project administration, Data curation. Lova Sun: Writing – review & editing, Validation, Resources. Michael Soike: Writing – review & editing, Validation, Resources. Adam J. Kole: Writing – review & editing, Validation, Resources. Gerard Walls: Writing – review & editing, Validation. Steven J. Feigenberg: Writing – review & editing, Validation, Conceptualization, Resources. Bonnie Ky: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Project administration, Funding acquisition, Methodology, Investigation, Formal analysis, Conceptualization, Data curation, Resources.

Funding
This research was supported by R01HL148272 (BK) and K24HL167127-01A1 (BK).
GW was supported by a Visiting Fellow Scholarship from Fulbright Ireland (Health Research Board Health Impact Award).

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.