Predicting Adverse Cardiac Events After Radiotherapy for Locally Advanced Non–Small Cell Lung Cancer

Background Radiotherapy may cause grade ≥3 cardiac events, necessitating a better understanding of risk factors. The potential predictive role of imaging biomarkers with radiotherapy doses for cardiac event occurrence has not been studied. Objectives The aim of this study was to establish the associations between cardiac substructure dose and coronary artery calcium (CAC) scores and cardiac event occurrence. Methods A retrospective cohort analysis included patients with locally advanced non–small cell lung cancer treated with radiotherapy (2006-2018). Cardiac substructures, including the left anterior descending coronary artery, left main coronary artery, left circumflex coronary artery, right coronary artery, and TotalLeft (left anterior descending, left main, and left circumflex coronary arteries), were contoured. Doses were measured in 2-Gy equivalent units, and visual CAC scoring was compared with automated scoring. Grade ≥3 adverse cardiac events were recorded. Time-dependent receiver-operating characteristic modeling, the log-rank statistic, and competing-risk models were used to measure prediction performance, threshold modeling, and the cumulative incidence of cardiac events, respectively. Results Of the 233 eligible patients, 61.4% were men, with a median age of 68.1 years (range: 34.9-90.7 years). The median follow-up period was 73.7 months (range: 1.6-153.9 months). Following radiotherapy, 22.3% experienced cardiac events, within a median time of 21.5 months (range: 1.7-118.9 months). Visual CAC scoring showed significant correlation with automated scoring (r = 0.72; P < 0.001). In a competing-risk multivariable model, TotalLeft volume receiving 15 Gy (per 1 cc; HR: 1.38; 95% CI: 1.11-1.72; P = 0.004) and CAC score >5 (HR: 2.51; 95% CI: 1.08-5.86; P = 0.033) were independently associated with cardiac events. A model incorporating age, TotalLeft CAC (score >5), and volume receiving 15 Gy demonstrated a higher incidence of cardiac events for a high-risk group (28.9%) compared with a low-risk group (6.9%) (P < 0.001). Conclusions Adverse cardiac events associated with radiation occur in more than 20% of patients undergoing thoracic radiotherapy within a median time of <2 years. The present findings provide further evidence to support significant associations between TotalLeft radiotherapy dose and cardiac events and define CAC as a predictive risk factor.

T he leading noncancer cause of morbidity and mortality among long-term cancer survivors is cardiovascular (CV) disease.[4] Furthermore, more than 10% of patients with lung cancer may experience such cardiac events within 2 years of receiving RT. 3 Thus, minimizing cardiotoxicity after thoracic RT remains a substantial clinical challenge, underscoring the importance of identifying predictive factors associated with these outcomes.
The mean RT doses received by the heart have been associated with cardiotoxicity, [4][5][6][7][8] with an estimated 4% to 16% increased risk for cardiac events per 1 Gy received.However, it has been suggested that total heart dose metrics may serve as a surrogate for damage to specific cardiac substructures. 9cent studies suggest that RT doses to coronary arteries and cardiac substructures could offer better quantification of risk after RT, 10,11 providing more specific metrics for assessing cardiotoxicity.
There is emerging interest in investigating RT doses to cardiac substructures, such as the coronary arteries, as a means to better quantify the risk for cardiac events after RT.This approach can provide valuable insights for guiding future treatment strategies aimed at reducing CV risk.
Additionally, coronary artery calcium (CAC) score, a direct imaging measure of atherosclerotic burden, has been reported as a highly specific imaging biomarker for CV risk, 12,13 and CAC score may provide additional cardiotoxicity risk stratification after RT. 14,15 CAC scoring has shown a strong predictive capability for future CV risk, and the extent of pre-RT CAC burden has demonstrated an independent association with cardiac events after RT, even when adjusting for mean heart dose (MHD). 16Although comprehensive automated Agatston scoring is available, simple visual assessment of CAC scoring can generate risk assessments for CV-related death and all-cause mortality, showing a strong association with outcomes. 17 sought to enhance our understanding of predictive factors associated with higher risks for future adverse cardiac events following thoracic RT by combining clinical, dosimetric, and CAC risk factors.
Furthermore, we investigated the potential of an integrated model to improve the prediction of adverse cardiac events after thoracic RT.This approach aims to provide management solutions for mitigating longterm CV risks among cancer survivors.

METHODS
CLINICAL DATA.A retrospective review was performed with Institutional Review Board approval provided by the Stanford Research Compliance Office on consecutively treated patients with locally advanced non-small cell lung cancer (NSCLC) who received RT as their primary treatment at our institution from 2006 through 2018.All patients were diagnosed with stage IIB to IIIC NSCLC. 18Baseline cardiac risk, including pretreatment cardiac history and data from the American College of Cardiology atherosclerotic CV disease (ASCVD) risk estimator, 19 was recorded.This risk estimator incorporates age, sex, race, smoking history, diabetes status, medications, blood pressure, and cholesterol levels in those with available data.Blood pressure readings were obtained from the initial radiation oncology consul-   No et al contoured on the primary CT data set (Figure 1A) by a single blinded observer (H.J.N.).A 5-mm brush was used to contour the individual cardiac vessels, guided by a cardiac contouring atlas. 20The following metrics were tabulated: mean doses received by the heart, LV, LAD, LMCA, LCx, RCA, combined 3-vessel arteries (including the LAD, LMCA, and LCx, referred to as TotalLeft), and all combined coronary arteries (referred to as TotalCor), as well as the volume receiving 15 Gy (V15) of the heart, LV, LAD, LMCA, CAC SCORING.CAC scores were tabulated using a modified Chiles method, 17 evaluating the initial treatment simulation CT study for all patients. 21though the initial intention was to use automated Agatston scoring, a significant portion of images were obtained with intravenous contrast, rendering automated Agatston methods suboptimal for scoring.
Thus, CAC scoring was performed through visual analysis of individual coronary vessels (Figure 1B).
For scoring, absent CAC was assigned a score of 0, mild CAC a score of 1 (Figure 1C), moderate CAC a score of 2, and heavy CAC a score of 3 for each epicardial vessel (Figure 1D).STATISTICAL ANALYSIS.Follow-up was calculated using the reverse Kaplan-Meier method, with day 0 being the first day of RT.Model performance for dose and cardiac calcification metric was assessed using receiver-operating characteristic (ROC) curves, using a competing risk ROC method (with the R package timeROC).To assess the prediction performance of the cardiac substructure mean dose for cardiac events, time-dependent ROC modeling at 5 years post-RT was performed.For threshold modeling, we used the log-rank statistic, with 95% CIs measured using bootstrap resampling with 1,000 iterations.
Competing-risk models were constructed using Fine-Gray subdistribution hazard models.Competing-risk models, considering death as the competing risk, were used to measure the cumulative incidence of cardiac events.These models were used in both uni-

DOSE TOXICITY, CAC TOXICITY, AND THRESHOLD
MODELING.Patients who experienced cardiac events received a significantly higher MHD compared with others (12.3 Gy vs 8.5 Gy; P ¼ 0.040) (Figure 2B).
Patients who experienced myocardial, conduction, and constrictive events had significantly greater left-sided coronary artery involvement of TotalLeft V15: 1.9 cc vs 0.4 cc (P < 0.01), 1.6 cc vs 0.4 cc (P < 0.01), and 3.1 cc vs 0.5 cc (P < 0.01), respectively, compared with those who did not experience such events.Only dose to the LAD was significantly associated with myocardial events (0.8 cc vs 0.5 cc; The obtained AUCs for the prediction performance of adverse cardiac events, on the basis of the mean dose to cardiac substructures, were 0.  Subsequently, we identified absolute dose constraints in cubic centimeters to the coronary arteries, because mean doses may be influenced by anatomical variation and user delineation.To identify rational thresholds and make comparisons with mean doses reported by others, 10 we used the log-rank statistic to perform threshold modeling.The following optimal cutpoints were identified for mean doses: 15.5 Gy for the LAD, 13.2 Gy for the LMCA, 1.2 Gy for the LCx, and 1.27 Gy for TotalLeft.We chose to quantify V15 thresholds for the LMCA, LCx, LAD, TotalLeft, and LV.Subsequently, the chosen thresholds were based on previously reported mean threshold values, 10 our mean threshold values with potential alternative cutpoints (Supplemental Figure 3), and the choice of 15 Gy as it represents a readily implementable value in clinical settings.We found the following V15 thresholds: 1.53 cc for the LAD, 0.16 cc for the LMCA, 0.71 cc for the LCx, 2.53 cc for TotalLeft, and 36.28 cc for the LV.We identified a CAC cutpoint score of >5 for TotalLeft.Furthermore, when excluding patients who underwent multiple courses of RT, we obtained nearly identical V15 thresholds: 1.53 cc for the LAD, 0.15 cc for the LMCA, 1.59 cc for the LCx, 3.52 cc for TotalLeft, and 5.48 cc for the LV.However, there was a local maximum threshold at 36 cc (Supplemental Figure 4).
We then evaluated whether including CAC and dose metrics could improve model performance.
Given the minimal difference in AUC values for univariable models predicting cardiac events including either MHD or coronary artery substructures, we also wanted to determine whether a multivariable model including CAC might further provide evidence for improved prediction when including coronary artery dose information compared with MHD alone.We repeated our time-dependent ROC analyses at the 5-year mark post-RT for the left coronary arteries' V15 (milliliters), and the results are shown in Supplemental Table 2.
COMPETING-RISK REGRESSION ANALYSIS.We performed both univariable and multivariable competing-risk regression analyses, adjusting for the competing risk for noncardiac death.In the univariable analysis, age, available in all patients, was the only baseline clinical factor significantly associated with cardiac events (Table 2).Although only a few of the patients had calculated ASCVD scores, this showed an association with adverse cardiac events (HR: 1.14; 95% CI: 1.03-1.25;P ¼ 0.008).Nearly all RT dose and CAC factors showed significant associations  No  3D (P < 0.001).
Additionally, there was no significant difference in cardiac events in those with prior RT vs without prior RT that exceeded TotalLeft dose constraints (P ¼ 0.49) or in cardiac events when TotalLeft dose constraints were not exceeded (P ¼ 0.052) (Supplemental Figure 6).

DISCUSSION
We investigated the risk factors for grade $3 adverse cardiac events after RT for locally advanced NSCLC, observing strong associations between coronary artery dose and pretreatment CAC score with the occurrence of cardiac events.With a median follow-up period of more than 6 years, we observed a 5-year cardiac event incidence of 13.8% following thoracic RT (Central Illustration).These cardiac events spanned a time frame of 1.7 to 118.9 months, with an IQR of 32.4 months (IQR: 7.1-39.5 months).
Although the occurrence of such subacute events might not be expected soon after RT, we opted to keep these early events in our analysis to inform the characterization of potential etiologies of cardiotoxicity.Nearly all of the observed cardiac events (88.5%) were either conduction or myocardial.Although high doses to the sinoatrial node have been associated with atrial fibrillation development, 11 preclinical evidence demonstrates that ischemia and infarctions represent major inciting events for remodeling mechanisms, such as fibrosis, for the development of arrhythmias. 24,25Moreover, ischemia has been associated with the onset of arrhythmias after RT. 26 Therefore, to identify subgroups in our cohort at high risk for cardiotoxicity, we used absolute dose volume metrics to obtain dose constraints for individual coronary arteries.
This approach revealed an approximate 2.5-fold increase in cardiac event incidence associated with elevated doses in the left coronary artery (V15 >2.53 cc).Additionally, we identified that CAC imaging biomarkers were associated with cardiac events in a dose-dependent manner.In a composite model, we identified low-and high-risk subgroups by including factors such as increasing age, Total-Left CAC, and V15.This improved understanding of vessel dosimetry and baseline vessel health provides valuable insights to reduce long-term cardiac damage for individuals undergoing RT.
For decades, elevated MHD was used as a dose constraint during RT planning. 4Illustrating the linear increase in cardiotoxicity risk with higher MHD, Darby et al 4 reported a 7.4% increase in cardiac events per mean dose delivered to the heart.However, recent studies have demonstrated that cardiac substructure doses have improved predictive performance for cardiotoxicity. 10,11,27Although we did find an association between MHD and cardiac events in our study, our analysis showed that dose to the left coronary arteries and CAC score had the strongest predictive performance on the basis of ROC analyses.
Furthermore, dose to the LV also displayed predictive capability for cardiac events; however, when combined with CAC, it performed less well compared with dose to the left coronary arteries.The consistency of our findings with the work by Atkins et al 10 is noteworthy.They showed in a cohort of 701 patients that RT doses received by the left coronary arteries had a higher AUC for cardiac events than MHD and were significantly associated with cardiac events.We externally validated the mean dose constraints reported by Atkins et al 10 and obtained a nearly identical mean dose threshold for the LAD (w15 Gy).Given their large cohort size and overlapping potential dose thresholds seen in our cohort (Results and Supplemental Figure 3), we identified dose constraints for the volume in milliliters of left coronary arteries receiving doses >15 Gy.We opted to provide absolute volumetric dose constraints because userdefined structures and anatomical variation may result in less reliable mean or percentage doses.
Of note, we aimed to provide a comprehensive overview of grade $3 adverse cardiac events, which may not be limited to previous definitions, as highlighted by Atkins et al. 10 For example, in the cohort reported by Atkins et al, 10 only approximately 10% of patients experienced cardiac events, but it is unknown how many subsequently developed other cardiac complications that may be secondary to RT exposure of critical structures.In fact, our own data exhibited a similar 8.2% rate of events when considering the same event categories.However, a No et al Cardiac Event Prediction in NSCLC After RT substantial number of other cardiac events, such as the development of atrial fibrillation, suggests that there is a higher occurrence of these clinically meaningful events that also appear to be related to dose received by cardiac structures.
Although doses to the left coronary arteries emerge as factors significantly associated with cardiotoxicity and that may be optimized during radiation treatment planning, we also studied baseline patient risk factors.This included evaluating CAC scores, which could serve as imaging biomarkers for pre-RT cardiotoxicity risk assessment, thus improving risk stratification.In a median follow-up of 73.7 months, radiotherapy (RT)-associated grade $3 adverse cardiac events occurred in 22.3% of those receiving thoracic RT in a median time of 21.5 months.A 5-year cardiac event incidence of 26.7% vs 10.6% (P ¼ 0.004) was seen for high-vs low-dose groups, where left-sided coronary artery volume receiving 15 Gy (V15) exceeding 2.53 cc was considered high.Additionally, high coronary artery calcium (CAC) (score >5) exhibited 28.4% vs 10.7% cardiac events compared with a low score.An integrated model incorporating dose and CAC showed a higher 5-year adverse cardiac event incidences of 6.9% and 28.9% for low-risk vs high-risk patients, respectively.NSCLC ¼ non-small cell lung cancer; OS ¼ overall survival.
No et al CAC has been further supported in studies by Wang et al 12 and Gal et al. 13 In a multivariable ROC model that included both left coronary artery CAC score and TotalLeft V15, we observed an AUC of 0.69, suggesting that these 2 independent risk factors in combination may deserve further study.
Given the strong association between CAC and cardiotoxicity, it is necessary to establish standardized approaches to compute CAC score.Although computationally intensive voxel-calculated scoring approaches have been used, 13,17,21 visual CAC scoring methods, such as the method of Chiles et al 17 we validated, have generated comparable risk assessments.We observed a high correlation between the visual scoring method and Agatston scoring (Supplemental Figure 2).Thus, our visual scoring assessment represents a clinically implementable tool that may be used for risk stratification, further justifying RT dose reduction for left coronary arteries or early medical risk reduction strategies when dose reduction is not possible. 12,28-30Improved image guidance and techniques such as respiratory and cardiac gating could aid in treatment planning to reduce doses to regions with severe calcifications and could further help mitigate cardiotoxicity, as seen in breast cancer RT, in which doses to calcified regions were associated with increased cardiotoxicity. 31wever, because of the 3-dimensional nature of lung RT involving all 3 left coronary arteries, implementing optimal dosing to severe calcifications would likely present challenges.
Our study provides not only valuable external validation of the strong association between the RT dose received by the left coronary arteries and adverse cardiac events 10 but also new evidence supporting the utility of CAC scoring in the stratification of patient risk prior to treatment.Our data may be leveraged in the design of future prospective clinical trials.In fact, there is growing recognition of the necessity for prospective data to incorporate cardiotoxicity endpoints into trial design. 32The landmark study by Darby et al, 4 which highlighted the association of cardiac dose and cardiotoxicity, was followed by a study of more than 1,800 trials that showed no difference in the inclusion of cardiotoxicity endpoints between the pre-Darby and post- which may be subject to inherent biases.However, our comparison with automated computerized scoring did show comparability.
Our cohort included 57 (24.5%) patients who had undergone multiple courses of thoracic RT; however, all grade $3 cardiac events occurred after the delivery of all RT courses.When excluding these patients from our dose constraint threshold modeling, we obtained nearly identical dose thresholds for the coronary arteries compared with the overall cohort.Additionally, through competing-risk regression (Table 2), we did not observe a higher risk for cardiotoxicity in individuals with prior RT.Finally, the incidence of cardiac events among those with prior RT compared with those without, exceeding TotalLeft dose constraints, did not show a significant difference (Supplemental Figure 6).This suggests that including these patients in our analysis does not appear to introduce bias to our results; in fact, their inclusion may provide valuable additional data on dose constraints when accumulating doses across multiple courses of RT.
Finally, we were surprised to find no association between traditional CV risk factors and prior CV disease and outcomes.However, this was likely due to the limited sample size of our cohort and its homogenous nature.The prevalence of smokers and a large number of other high-risk factors among those evaluated may limit the correlation of traditional CV risk factors.

CONCLUSIONS
Radiation-associated grade $3 adverse cardiac events occur in more than 20% of individuals undergoing main coronary artery LV = left ventricle/ventricular MHD = mean heart dose NSCLC = non-small cell lung cancer RCA = right coronary artery ROC = receiver-operating characteristic RT = radiotherapy V15 = volume receiving 15 Gy From the a Department of Radiation Oncology, Stanford University School of Medicine, Stanford, California, USA; b University of Rochester School of Medicine and Dentistry, Rochester, New York, USA; c Department of Medicine, Division of Cardiology, City of Hope Comprehensive Cancer Center, Duarte, California, USA; and the d Division of Cardiovascular Medicine, Department of Medicine, Stanford University School of Medicine, Stanford, California, USA.*Dr No and Ms Guo contributed equally to this work.

2 0 2 3 : 7 7 5 -7 8 7
Constrictive events referred to pericarditis.Valvular events involved any form of valve-related disease.Conduction events included abnormalities requiring a permanent pacemaker or second-or third-degree heart block, as well as newly occurring atrial arrhythmia or ventricular arrhythmia requiring interventions.DOSIMETRIC ANALYSIS.The RT plans were transferred to MIM Radiation Oncology (MIM Software).All doses were converted to equivalent doses in 2-Gy fractions using an a/b ratio of 3 for organs at risk.In cases in which patients underwent multiple courses of thoracic RT, plans from separate computed tomographic (CT) simulation scans were fused on the initial thoracic treatment CT study.Plans not involving the thorax were excluded from the analysis.Subsequently, the heart, left ventricle (LV), left anterior descending coronary artery (LAD), left main coronary artery (LMCA), left circumflex coronary artery (LCx), and right coronary artery (RCA) were

FIGURE 1
FIGURE 1 Visual Scoring of CAC

FIGURE 3
FIGURE 3 Cardiac Event Cumulative Incidence Curves Adjusted for Noncardiac Death

J
A C C : C A R D I O O N C O L O G Y , V O L . 5 , N O .6 , 2 0 2 3 Cardiac Event Prediction in NSCLC After RT D E C E M B E R 2 0 2 3 : 7 7 5 -7 8 7 PERSPECTIVES COMPETENCY IN MEDICAL KNOWLEDGE: In patients with lung cancer treated with thoracic RT, measured pretreatment imaging of CAC alongside RT dose to specific cardiac substructures are associated with adverse cardiac events, which occurred in >20% of our cohort at a median time of <2 years.TRANSLATIONAL OUTLOOK: Future studies are needed to assess the utility of measuring these markers of cardiac risk in those undergoing definitive thoracic RT to inform personalized treatment strategies.
Values are median (range) or n (%) except as indicated.Chemotherapy regimens listed in Supplemental Table3.a Clinical and laboratory information was sufficient to calculate ASCVD score for 18.9% of the total cohort (n ¼ 44).3D ¼ 3-dimensional; ASCVD ¼ atherosclerotic cardiovascular disease;
Cardiac Event Prediction in NSCLC After RT thoracic RT within a median time frame of <2 years.We observed an elevated cardiac event risk following thoracic RT, particularly for patients with moderate to severe CAC who receive elevated RT doses to the left coronary arteries (TotalLeft V15 >2.53 cc).Our RT dose constraints offer external validation to findings in other large cohorts, and our CAC scoring provides new clinical evidence for assessing the baseline risk for cardiotoxicity prior to treatment.With further study, our findings may serve to inform personalized RT planning and facilitating early medical intervention for patients with locally advanced NSCLC who are undergoing thoracic RT.Niska JR, Hu J, Li J, et al.Using novel statistical techniques to accurately determine the predictive dose range in a study of overall survival after definitive radiotherapy for stage III non-small cell lung cancer in association with heart dose.J Cancer Ther.2021;12:505-529.7. van Nimwegen FA, Schaapveld M, Cutter DJ, et al.Radiation dose-response relationship for risk of coronary heart disease in survivors of Hodgkin lymphoma.J Clin Oncol.2016;34:235-243.8. van den Bogaard VAB, Ta BDP, van der Schaaf A, et al.Validation and modification of a prediction model for acute cardiac events in patients with breast cancer treated with radiotherapy based on three-dimensional dose distributions to cardiac substructures.J Clin Oncol.2017;35:1171-1178.with major adverse cardiac events and mortality in patients with non-small cell lung cancer.JAMA Oncol.2021;7:206-219.11.Kim KH, Oh J, Yang G, et al.Association of sinoatrial node radiation dose with atrial fibrillation and mortality in patients with lung cancer.JAMA Oncol.2022;8(11):1624-1634. 12. Wang K, Malkin HE, Patchett ND, et al.Coronary artery calcifications and cardiac risk after radiation therapy for stage III lung cancer.Int J Radiat Oncol Biol Phys.2022;112:188-196.13.Gal R, van Velzen SGM, Hooning MJ, et al.
the risks of breast cancer radiotherapy: evidence from modern radiation doses to the lungs and heart and from previous randomized trials.J Clin Oncol.2017;35:1641-1649.6. 9. Hoppe BS, Bates JE, Mendenhall NP, et al.The meaningless meaning of mean heart dose in mediastinal lymphoma in the modern radiation therapy era.Pract Radiat Oncol.2020;10:e147-e154.10.Atkins KM, Chaunzwa TL, Lamba N, et al.Association of left anterior descending coronary artery radiation dose fications on radiotherapy planning CT scans in patients with breast cancer.JAMA Oncol.2021;7: 1024-1032.14.Greenland P, Blaha MJ, Budoff MJ, Erbel R, Watson KE.Coronary calcium score and cardiovascular risk.J Am Coll Cardiol.2018;72:434-447.15.Shemesh J, Henschke CI, Shaham D, et al.Ordinal scoring of coronary artery calcifications on low-dose CT scans of the chest is predictive of