Coronary Artery Calcifications and Cardiac Risk after Radiotherapy for Stage III Lung Cancer

Abstract

Purpose:

Heart dose and heart disease increase risk for RT-associated cardiac toxicity. We hypothesized that CT coronary calcifications are associated with cardiac toxicity and may help ascertain baseline heart disease.

Methods and Materials:

We analyzed the cumulative incidence of cardiac events in patients with Stage III non-small-cell lung cancer receiving median 74 Gy on prospective dose-escalation trials. Events were defined as symptomatic effusion, pericarditis, unstable angina, infarction, significant arrhythmia, and/or heart failure. Coronary calcifications were delineated on simulation CT’s using MIM (130 HU threshold). Calcifications were defined “None,” “Low,” and “High,” with median volume dividing Low and High.

Results:

Of 109 patients, 26 had cardiac events at median 26 months (range, 1-84 months) post-RT. Median follow-up in surviving patients was 8.8 years (range, 2.3-17.3). On simulation CT’s, 64 (59%) had coronary calcifications with median volume 0.2 cc (range 0.01-8.3). Only 16 patients (15%) had baseline coronary artery disease. Cardiac events occurred in 7% (3/45), 29% (9/31), and 42% (14/33) of patients with No, Low, and High calcifications, respectively. Calcification burden was associated with cardiac toxicity on univariate (Low vs. None: HR 5.0, p=0.015, High vs. None: HR 8.1, p<0.001) and multivariate analyses (Low vs. None: HR 7.0, p=0.005, High vs. None: HR 10.6, p<0.001, heart mean dose: HR 1.1/Gy, p<0.001). Four-year competing risk-adjusted event rates for No, Low, and High calcifications were 4%, 23%, and 34%, respectively.

Conclusions:

Coronary calcifications can be considered a cardiac risk factor that can identify high-risk patients for medical referral and help guide clinicians prior to potentially cardiotoxic cancer treatments.

Introduction

RT-associated cardiac toxicity has been investigated for thoracic malignancies including breast cancer, lymphoma, lung cancer, and gastroesophageal cancers; but the clinical presentation of cardiac toxicity varies widely based on patient age and performance status.^1,2 Patients with lung cancer often have a smoking history and other cardiopulmonary comorbidities, leading to earlier and varied cardiac events including pericarditis, pericardial effusions, arrhythmias, and acute coronary syndromes.^3,4 Increased risks of toxicity have been linked to both higher heart doses and baseline cardiac risk. Though dose is a relatively straightforward metric, ascertainment of baseline cardiac disease (a more nebulous construct) can be complex. While scoring systems (eg, Framingham, QRISK, WHO/ISH, and others) exist in the setting of primary cardiovascular disease prevention, they are not routinely considered by oncologists.⁵

Accurate and quantitative ascertainment of baseline cardiac disease may help manage competing priorities when balancing cancer treatments with cardiac toxicity. For instance, echocardiography is routinely used to evaluate and monitor ejection fraction in patients receiving anthracyclines, trastuzumab, and other cardiotoxic therapies.^6,7 The coronary artery CT calcium also provides a sensitive, reproducible method to assess cardiovascular risk, and has been recommended by cardiology societies for risk stratification and screening.^8-13 CT scans are vital in oncology: almost all patients with lung (and other) cancers have diagnostic CT or PET/CT scans, and all patients who receive radiation have an additional CT “simulation” scan for RT planning. Thus, could quantification of coronary calcium on pre-treatment CT’s be a simple and economic way to risk stratify patients prior to thoracic radiation?

We herein assess the relationship between coronary calcifications, heart dose, and subsequent cardiac events in a previously-studied cohort of patients with Stage III non-small-cell lung cancer (NSCLC) receiving dose-intensified chemoradiotherapy on 6 prospective trials.³ Our hypothesis was that the extent of coronary calcifications observed on radiation simulation CT is associated with risk of symptomatic cardiac events after thoracic radiotherapy. We also hypothesized that calcifications are more reliable than documented medical history for ascertainment of baseline cardiac disease, given that many patients with NSCLC are smokers and may have undiagnosed cardiac conditions.

Materials and Methods

We performed a pooled post-hoc analysis of patients treated on six IRB-approved prospective trials, for which previous analyses correlating toxicity with cardiac and cardiac substructure dose have been published.^3,14 From 1996 to 2009, 127 patients with Stage III NSCLC with ECOG performance status 0-1 received dose-escalated RT using three-dimensional conformal RT (3D-CRT) to 70-90 Gy with induction and (in 5 of 6 trials) concurrent chemotherapy. Chemotherapy regimens varied but did not include any agents considered to be cardiotoxic (eg, anthracyclines or 5-FU). 112 patients were included in the original analysis after excluding patients who did not complete RT or those with inaccessible radiation plans. Heart and substructures were delineated for dosimetric analysis per guidelines published by Feng et al, with the superior slice just inferior to the left pulmonary artery.¹⁵ The primary endpoint was symptomatic cardiac events, defined as symptomatic pericardial effusion, pericarditis, unstable angina, myocardial infarction, significant arrhythmia, and/or heart failure. Additional details can be found in prior reports.^3,14 For the current analysis of coronary calcifications, 3 patients who did not have a non-contrast simulation CT were excluded, leaving 109 patients for final analysis.

Radiographic Assessment of Calcifications

Radiologic assessment of calcifications was performed using MIM software (Cleveland, OH) primarily using non-contrast simulation CT, and secondarily using contrast-enhanced diagnostic CT, with separate analyses performed for each method (Figure 1). Simulation CT’s were performed without contrast with patients free breathing in the treatment planning position. Simulation CT slice thickness was 3mm in 74 patients (68%), 4mm in 5 patients (5%), 5mm in 29 patients (27%), and 8 mm in one patient (1%). For this primary simulation CT dataset, an automated thresholding tool was chosen to minimize bias and maximize objectivity. Coronary calcifications were defined as any areas within the heart with a minimum of 130 Hounsfield Units (HU), which is a standard definition used in the literature (eg, for calculating Agatston score on dedicated non-contrast coronary CT).¹⁶ Obvious artifacts were excluded from the calcification volume.

Figure 1.

Examples of patients (A-C) with cardiac toxicity who had no known cardiac history, but calcifications detected on radiation planning simulation CT (images on left, calcifications delineated in automated fashion shown in red) or diagnostic CT (images on right, calcifications delineated manually shown in blue)

To verify the findings of the simulation CT dataset, we performed a confirmatory analysis on a secondary dataset of any patients with retrievable diagnostic, breath-hold, contrast-enhanced chest CT’s within 3 months before or after the radiation start date (n=74). Diagnostic CT slice thickness was 5mm in 59 patients (80%) and 8mm in 15 patients (20%). For the secondary diagnostic CT dataset, the automated thresholding tool was not used due to the high HU of contrast material and calcifications were thus manually delineated by the primary investigator (KW).

Assessment and Definition of Cardiac Risk

In prior reports, we attempted to quantify cardiac risk using two methods: presence or absence of baseline coronary artery disease (CAD) documented in history and calculation of the WHO / International Society of Hypertension (WHO/ISH) risk score, an algorithm that estimates 10-year risk of a cardiovascular event (in five strata ranging from <10% to ≥40%) based on age, sex, smoking status, diabetes, and systolic blood pressure.¹⁷

To assess the performance of coronary calcifications as a risk factor in the current analysis, the total 3-dimensional volume of calcifications was calculated in MIM. We did not use the Agatston score because it requires assessment of the 2-dimensional calcification area on each CT slice, which is not a capability of MIM software (that allows calculation of 3-dimensional structure volume) and furthermore is dependent on a standard slice thickness of 3 mm.¹⁶ For both the primary simulation CT and secondary diagnostic CT cohorts, we then divided patients into three roughly equivalent tertiles of calcification burden for analysis: (0: None, 1: Low, 2: High), using the median calcification volume of those with calcifications (in each cohort) as a cutpoint between “Low” and “High” calcifications.

Statistical Analysis

Analyses were performed using SAS (Cary, NC). Cumulative incidence of cardiac events (using death as a censoring event) was initially estimated using the Kaplan-Meier method, then with adjustment for the competing risk of death using the Fine and Gray method.¹⁸ A two-tailed p of < 0.05 was considered significant. Univariate analysis was performed for covariates including CAD (presence vs. absence), WHO/ISH 10-yr risk (<10%, 10-20%, and ≥20%), coronary calcification burden (none, Low, and High), and heart mean dose. Other dosimetric parameters including lung, volumetric, and cardiac substructure doses were previously analyzed, but not presently included given the focus on calcifications.

Given the relatively low number of cardiac events, the multivariable analysis was limited to two covariates per analysis (heart dose and calcifications). Calcifications, CAD, and WHO/ISH were not analyzed in the same multivariate model given collinearity and the limited number of events. Multivariable analyses combining heart dose and these other metrics of baseline risk were, however, performed previously.³ The Akaike information criterion (AIC) statistic was also calculated for each baseline risk covariate; the covariate with the lowest AIC is considered the “best predictor” for the competing-risk model. Finally, intraclass correlation coefficient was calculated to assess the agreement between calcification volume as assessed on the primary simulation CT dataset vs. the secondary diagnostic CT dataset.

Results

The final cohort consisted of 109 patients, of whom 26 patients had one or more cardiac events at a median 26 months (range, 1-84 months). Median follow-up in the 11 patients alive at time of analysis was 8.8 years (range, 2.3 to 17.3 years). Median RT dose was 74 Gy (range, 70-90 Gy). Patient characteristics are shown in Table 1 for the entire cohort, and divided by those with vs. without symptomatic cardiac events. Cardiac events included 7 symptomatic pericardial effusion, 5 myocardial infarction, 3 unstable angina, 2 pericarditis, 12 significant arrhythmia, and 1 heart failure. The median heart mean dose for the cohort was 12.4 Gy

Table 1.

Patient and treatment characteristics

Characteristic	All patients	Patients without symptomatic cardiac events	Patients with symptomatic cardiac events
No.	109	83	26
Median age (range, years)	58 (36-82)	57 (39-82)	62 (36-81)
Gender
Male	60 (59%)	43 (52%)	17 (65%)
Female	49 (41%)	40 (48%)	9 (35%)
Tumor laterality
Right	64 (59%)	47 (57%)	17 (65%)
Left	45 (41%)	36 (43%)	9 (35%)
Stage
IIIA	64 (59%)	47 (57%)	17 (65%)
IIIB	45 (41%)	36 (43%)	9 (35%)
ECOG Performance status
0	70 (64%)	50 (60%)	20 (77%)
1	39 (36%)	33 (40%)	6 (23%)
Active/former smoker
No	15 (14%)	13 (16%)	2 (8%)
Yes	94 (86%)	70 (84%)	24 (92%)
Median pack years (range)	37 (5-120)	35 (13-120)	43 (5-100)
WHO/ISH 10-yr risk
<10 %	65 (60%)	54 (65%)	11 (42%)
10-20 %	34 (31%)	22 (27%)	12 (46%)
≥20 %	10 (9%)	7 (8%)	3 (12%)
CAD
No	93 (85%)	76 (92%)	17 (65%)
Yes	16 (15%)	7 (8%)	9 (35%)
Calcifications
None	45 (41%)	42 (51%)	3 (12%)
Low (<0.2 cc)	31 (28%)	22 (27%)	9 (35%)
High (≥0.2 cc)	33 (30%)	19 (23%)	14 (54%)
Prescribed RT dose
70-74 Gy	84 (77%)	65 (78%)	19 (73%)
78-90 Gy	25 (23%)	18 (22%)	7 (27%)
Gross tumor volume (median)	46.1 cc	47.1 cc	44.5 cc
Heart mean dose (median)	12.4 Gy	10.1 Gy	20.4 Gy
Heart V5Gy (median)	37.6 %	34.1 %	55.9 %
Heart V30Gy (median)	17.4 %	12.1 %	28.8 %
Left ventricle mean dose (median)	4.0 Gy	2.9 Gy	9.5 Gy
Left ventricle V5Gy (median)	18.4 %	14.2 %	38.2 %
Left ventricle V30Gy (median)	2.0 %	0.4 %	10.3 %

Abbreviations: ECOG, Eastern Cooperative Oncology Group; WHO/ISH, World Health Organization / International Society of Hypertension; CAD, coronary artery disease; RT, radiotherapy; V5Gy, volume receiving ≥5 Gy; V30Gy, volume receiving ≥30 Gy.

^*Column percentages displayed.

Ascertainment of Calcifications and Cardiac Risk Factors

Of 109 patients, 64 (59%) had evidence of calcifications on simulation CT’s. Using the median calcification volume of 0.2 cc (range 0.01-8.3 cc) as a cutpoint, 31 patients (28%) had “Low” calcifications (<0.2 cc) vs. 33 patients (30%) with “High” calcifications (≥0.2 cc). Of 74 patients with available diagnostic CT’s, 35 (47%) had calcifications with median volume 0.5 cc (range 0.08–2.6 cc). There was good agreement between calcification volume on diagnostic vs. simulation CT’s (intraclass correlation coefficient 0.8).

Table 2 shows the relationship between calcification status and cardiac risk as ascertained using CAD or WHO/ISH 10-yr risk. As expected, calcification burden was collinear and associated with these other measures of cardiac risk (chi-sq p<0.001 for both CAD and WHO/ISH). However, calcifications were detected in 59% of patients; whereas only 15% had a baseline diagnosis of CAD and only 40% had a WHO/ISH risk score above the lowest stratum. Stated differently, though most patients with CAD had High calcifications (13/16, 81%), only a minority of patients with High calcifications had a clinical diagnosis of CAD (13/33, 39%).

Table 2.

Relationship between calcifications and cardiac risk assessed by CAD or WHO/ISH

	CAD		Baseline WHO/ISH 10-yr risk
	No	Yes	Low (<10%)	Medium (10-20%)	High (≥20%)
Calcifications
None	44 (47%)	1 (6%)	37 (57%)	8 (24%)	0
Low (<0.2 cc)	29 (31%)	2 (13%)	16 (25%)	12 (35%)	3 (30%)
High (≥0.2 cc)	20 (22%)	13 (81%)	12 (19%)	14 (41%)	7 (70%)

Association of Cardiac Risk Factors with Symptomatic Cardiac Events

Table 3 shows analyses examining factors associated with symptomatic cardiac events while accounting for time to event and competing risk of death. In addition to heart dose, CAD, and WHO/ISH risk score, coronary calcifications on simulation CT’s appeared to be significantly associated with cardiac toxicity on univariate analysis (Low calcifications: HR 4.98, p=0.015; High calcifications: HR 8.10, p<0.001). The multivariable analysis paired calcification burden with heart mean dose. Both calcification burden and heart mean dose appeared to be independently associated with toxicity (Low calcifications: HR 7.01, p=0.005; High calcifications: HR 10.6, p<0.001; heart mean dose: HR 1.05/Gy, p<0.001). Findings were similar in the confirmatory diagnostic CT cohort. Multivariable analyses also showed that the other cardiac risk factors (CAD and WHO/ISH risk score) were associated with toxicity when paired with heart dose metrics, as presented in a previous report.³ The AIC values for calcifications, CAD, and WHO/ISH risk score were 225, 230, and 236, respectively for the simulation CT cohort, and 127, 127, and 133, respectively for the diagnostic CT cohort. Thus, calcification burden appeared to be the best predictor of toxicity amongst the baseline risk covariates.

Table 3.

Competing risk-adjusted analyses for symptomatic cardiac events

		Sim Cohort (n=109)		Diagnostic Cohort (n=74)
Characteristic	p	HR	p	HR
Univariable analyses:
Age	0.335	1.02	0.200	1.03
ECOG PS (1 vs. 0)	0.125	0.49	0.526	0.69
Gross tumor volume	0.434	1.00	0.259	1.00
Heart mean dose	0.002	1.03 / Gy	0.010	1.03 / Gy
Left ventricle mean dose	0.032	1.03	0.107	1.03
WHO/ISH 10-yr risk
≥20% (vs. <10%)	0.28	1.98	0.81	1.28
10-20% (vs. <10%)	0.04	2.34	0.04	2.97
CAD	<0.001	3.69	<0.001	5.47
Calcification burden
Any (vs. none)	0.002	6.50	0.007	5.71
High (vs. none)	<0.001	8.10	0.004	7.12
Low (vs. none)	0.015	4.98	0.043	4.35
Multivariable analvses:
Heart mean dose	<0.001	1.05	0.003	1.05
Calcification burden
High (vs. none)	<0.001	10.64	<0.001	9.83
Low (vs. none)	0.005	7.01	0.072	3.99

Abbreviations: ECOG PS, Eastern Cooperative Oncology Group performance status; WHO/ISH, World Health Organization / International Society of Hypertension; CAD, coronary artery disease

Figure 2A shows the competing risk-adjusted cumulative incidence of symptomatic cardiac events for the three calcification tertiles. Competing risk-adjusted event rates for patients with no, Low, and High calcifications were 2%, 10%, and 21%, respectively at 2 years, and 4%, 23%, 34%, respectively at 4 years. The crude cardiac event rates were 7% (3 of 45 patients), 29% (9 of 31 patients), and 42% (14 of 33 patients), for patients with no, Low, and High calcifications, respectively.

Figure 2.

Cumulative incidence of competing risk-adjusted symptomatic cardiac events in patients divided based on calcification burden (A), heart mean dose for all patients (B), and heart mean dose for patients with calcifications (C).

Interaction of Calcifications and Heart Dose

The incidence of cardiac events in patients with different heart doses is shown in Figure 2B (for the entire cohort) and Figure 2C (limited to patients with calcifications). Cardiac risk appears to be higher for each heart dose tertile in patients with calcifications. The 4-year competing-risk adjusted cardiac event rate in those with calcifications vs. the overall cohort was 7% vs. 4%, 22% vs. 14%, and 63% vs. 41%, for heart mean dose <10 Gy, 10-20 Gy, and ≥20 Gy, respectively. Figure 3 depicts the estimated 4-year cardiac event rate for patients grouped both by heart dose and calcification burden. Though patient numbers in each bin are low, the overall pattern supports the dual contribution of dose and calcifications to cardiac events.

Figure 3.

Estimated 4-year competing risk-adjusted symptomatic cardiac event rate in patients grouped into 9 unique bins by both calcification burden (No, Low, High calcs) and mean heart dose (<10 Gy, 10-20 Gy, ≥20 Gy).

Discussion

In this analysis of 109 patients with Stage III NSCLC receiving a median 74 Gy on six prospective trials, clinically significant cardiac events were strongly associated with both pre-RT coronary artery calcification burden and heart dose. Calcifications were observed in approximately half of patients and may reliably ascertain baseline cardiac risk prior to cancer treatment. Since CT scans are routinely used for both cancer diagnosis and radiation planning, assessment of coronary calcification burden could be simple, timely, and cost-effective for personalizing risk-adapted care in this and other similar patient populations.

In recent years, many studies have reported that radiation dose to the heart and its components increases risk for a variety of cardiac events and may decrease survival.^4,19-21 These data have increased emphasis on heart-sparing for Stage III NSCLC, an emphasis that has already existed for decades when treating lymphoma and breast cancer. That it took so much longer for NSCLC is likely a reflection of two substantial competing risks: (1) Poor cancer-related prognosis for locally advanced NSCLC, that may now be improving due to advances including radiotherapy techniques, improved staging, and immunotherapy, and (2) Comorbid cardiopulmonary conditions that often co-exist in smokers, both confounding assessment of toxicity and independently worsening patient outcomes. Indeed, it is possible that cardiopulmonary decline ascribed to pre-existing comorbidities may reflect (at least in part) RT-associated cardiac disease, and vice versa.

A previously published analysis established a link between heart dose and toxicity for Stage III NSCLC. In this re-analysis, our goal was to investigate the use of coronary calcifications as a biomarker of comorbid cardiac disease that could help clinicians risk-stratify patients prior to thoracic radiotherapy. In this high-risk population of predominantly active/former smokers (86% of patients), calcifications were present on over half of simulation CT’s, representing significantly greater ascertainment of potential baseline cardiac disease than either CAD or WHO/ISH score. Indeed, many patients with no known CAD or other cardiac history had evidence of substantial calcification burden (Table 1 and Figure 1). Calcifications were not only prevalent, but also strongly associated with an increased incidence of cardiac events after thoracic RT. Both heart dose and calcifications were predictive of toxicity, consistent with the notion that baseline cardiac comorbidities should be considered and optimized, alongside efforts to minimize heart dose. The synergy between pre-RT calcifications and heart dose is visually presented in Figure 2B-C. Compared to the overall study population, patients with calcifications appeared to have an approximately 50% higher risk of cardiac events for each heart dose tertile.

As further background, cardiovascular risk assessment in primary cardiac disease prevention quantifies individual risk and guides use of preventative therapies (eg, statins, aspirin, etc). Traditional risk assessment methods include calculated risk scores including Framingham, QRISK, Pooled Cohort Equation, and WHO/ISH.⁵ In recent decades, extensive studies have shown that coronary calcium CT’s are sensitive and specific for coronary atherosclerosis, potentially superior to traditional cardiovascular risk scores, and are cost-effective for screening.^8-13 As a result, cardiology societies have increasingly adopted coronary calcification assessment into their guidelines.¹³

In oncology, cardiovascular risk assessment is important to balance major risks and benefits of potentially cardiotoxic cancer treatments. A recent American Society of Clinical Oncology Clinical Practice Guideline reviewed high-risk scenarios (including thoracic RT) and recommended screening for modifiable cardiovascular risk factors and consideration of echocardiography for suspected cardiac dysfunction.⁷ However, the extent to which such screening is practiced and/or practical is unknown. Patients with lung and other smoking-associated cancers may be more likely to live in rural communities or other areas with limited healthcare access.^22-24 Those without primary care physicians or reliable records may have undiagnosed comorbid cardiac conditions (eg, CAD). Assessment of coronary calcifications on CT (that patients already have) may therefore provide valuable data for cardiac risk stratification and medical optimization, without requiring additional testing or delaying treatment. Observation of calcifications could also influence clinicians during radiotherapy planning, where competing priorities of heart, lung, and esophageal sparing (along with tumor coverage) must be balanced. For instance, heart sparing could be prioritized when extensive calcifications are observed. Conversely, it may be reasonable to prioritize lung and esophagus (to minimize risks of pneumonitis and esophagitis, respectively) when calcifications and other cardiac risk factors are absent.

The use of incidentally detected calcifications as a clinical tool has been studied in other clinical contexts, including breast and lung cancer screening. Numerous studies have shown associations between calcifications detected on screening mammography with both coronary artery calcium burden and clinical cardiovascular disease. This has led to active interest in using mammography for the dual purposes of breast cancer screening and primary cardiovascular disease prevention.^25,26 In addition, coronary artery calcifications observed on lung cancer screening CT’s were associated with cardiac and all-cause mortality in the National Lung Screening trial.²⁷ These studies are consistent with our findings and illustrate that patients with cancer have coexisting risks of both cardiac and oncologic morbidity. Our study, in addition, demonstrates the potential utility of applying this “bonus” radiologic data in clinical practice prior to cancer treatment, and adds to the evolving field of cardio-oncology. Calcifications may be an easy, highly-sensitive screening tool that can both influence oncologic decision making and coincidentally identify high-risk patients to refer to cardiology.

There are several important limitations to discuss. First, we did not use the well-known Agatston score (that is standardly calculated with cardiac CT screening), instead presenting absolute volume of calcifications. However, we repeated the analysis in a separate diagnostic CT cohort with similar findings, and currently there are multiple accepted methodologies for coronary calcification assessment.^28-30 Second, it is possible that identification of calcifications is too sensitive (~50% of study patients). Some calcifications are not associated with clinically meaningful CAD, and one could argue that clinically diagnosed CAD (only 15% of study patients) is a more relevant metric. However, greater sensitivity/ascertainment may be preferable in this context of risk stratification prior to potentially-cardiotoxic treatment. Third, the study endpoints included non-ischemic events not clearly related to atherosclerotic disease. Nonetheless, calcifications could be considered either a surrogate for subclinical cardiac disease or susceptibility to RT-induced cardiac injury in general. Fourth, cardiac risk factors (and smoking rates) vary across the country and our patient cohort may not be representative of all patients. However, our findings are likely generalizable to the broader population of higher-risk patients with shared causal factors for both lung cancer and cardiovascular disease.

In conclusion, coronary calcifications are prevalent in high-risk smokers with lung cancer and were associated with cardiac toxicity after thoracic radiotherapy. Assessment of coronary calcifications using CT is straightforward, identifies patients for targeted medical management, and can help clinicians balance cardiac comorbidities with potentially cardiotoxic treatments.