By Michael H. Crawford, MD, Editor

SYNOPSIS: The authors applied an automated algorithm to calculate an Agatston coronary artery calcium score from non-ECG-gated planning CT scans in breast cancer patients undergoing radiation therapy. This provided a graded risk calculation that could encourage preventive measures in patients at highest risk of a cardiovascular disease event.

SOURCE: Gal R, van Velzen SGM, Hooning MJ, et al. Identification of risk of cardiovascular disease by automatic quantification of coronary artery calcification on radiotherapy planning CT scans in patients with breast cancer. JAMA Oncol 2021; May 6. doi: 10.1001/jamaoncol.2021.1144. [Online ahead of print].

In older women, breast cancer competes with cardiovascular (CV) disease as the leading cause of death. Most patients with breast cancer undergo a CT scan for planning purposes before starting a course of radiation therapy. Gal et al developed an automatic algorithm for measuring the coronary artery calcium score on these CT scans. Then, they conducted tests to see if these automatically produced scores would identify breast cancer patients at higher risk for CV disease.

In two centers in the Netherlands, breast cancer patients undergoing radiation therapy were included in this retrospective analysis (unless they had known metastatic disease or if their CT scan was taken more than one year before their cancer diagnosis). Clinical data were obtained from the Netherlands Cancer Registry. The CT scans were performed without ECG gating. Gal et al developed a deep learning-based algorithm to produce an Agatston score of coronary artery calcium (CAC). The authors considered five categories of CAC score: 0, 0-10, 11-100, 101-400, and > 400. The primary outcomes were fatal or non-fatal CV disease (CVD) events and all-cause mortality. Patients were followed until December 2018 unless a CVD event occurred. Also, to avoid competing risks, patients were censored for non-CVD events, breast cancer recurrence, or the development of other cancers. The risk analyses were adjusted for age, left or right breast, and concomitant anthracyclines. On average, the 15,915 patients studied were age 59 years; all were followed for an average of 57 months.

The distribution of CAC scores was: 0 in 70% of the population; 1-10 in 10%, 11-100 in 12%, 101-400 in 5%, and > 400 in 3%. A CVD-related hospitalization occurred in 8.4% of patients (0.7% died of CVD). The risk of these outcomes increased with higher calcium scores: 5% with 0 vs. 28% at > 400. The adjusted HRs for CVD risk were 1.1, 1.8, 2.4, and 3.4, respectively, in the four groups with detectable coronary calcium compared to those with a CAC score of 0. If the patient also received anthracyclines, the HR was 5.8. If they received a radiation boost, the HR was 6.1. The CAC score was particularly strong at predicting the coronary artery disease (CAD) subset of CVD (HR, 7.8). Laterality did not affect the results. The authors concluded automated CAC scoring on radiation planning CT scans may be a low-cost tool to identify breast cancer patients at higher risk of CVD. Abrogating this risk could reduce the occurrence of CVD in breast cancer patients.


Radiologists are paying more attention to the presence of calcium in CT scans of the chest and upper abdomen that are conducted for other reasons — specifically, when the scans show calcium in the coronary arteries. By putting this finding into their reports, radiologists hope primary care clinicians include this information into enhanced CVD prevention interventions. The Gal et al study pushed this concept to a higher level. They developed an automated algorithm by deep learning methods, which can calculate an Agatston CAC score from the non-ECG-gated CT scans used to plan radiation therapy in breast cancer patients. The derived CAC score also was excellent at detecting the subset of CVD patients with CAD. Since CVD is common in breast cancer patients and almost all received radiation therapy, this is a potentially simple, cost-effective way to identify the highest-risk patients for developing CVD.

There were other factors that augmented the CVD risk prediction. An extra radiation treatment (boost) was an obvious one. Radiation to the heart can lead to subsequent CAD, valve damage, and even cardiomyopathy. Administering concomitant anthracycline augments the risk of radiation therapy, which should not be unexpected since this chemotherapeutic agent has been associated with cardiomyopathy. There were too few patients in the Gal et al study who received trastuzumab to analyze its potential effect. Interestingly, laterality did not affect the results. Perhaps the heart receives enough radiation through whichever breast is irradiated to cause a detrimental effect.

There were limitations to the Gal et al study. The follow-up was rather short at 4.75 years. The adverse effects of radiation therapy often are not appreciated for 10 years. There was no control group that did not receive radiation therapy, so the effects of the various insults to the heart could not be parsed. The clinical information came from administrative databases, although in this study a rather complete one (i.e., it was focused on cancer patients). There was no information on traditional risk factors and what roles they may have played. Although there were no data on the dose of radiation and its effect on the CAC score, using a boost dose later did increase the risk of CVD. Finally, we do not know if clinicians used this data to change the treatment of the patients, especially regarding CVD risk reduction. If so, did it made any difference in outcomes?

At this point, we should encourage radiologists to continue to note coronary calcium in their CT scan reports and to explore acquiring this software for calculating a CAC score soon. It is easier to make firm recommendations to patients if they are determined to be at high risk. This information from these routine CT scans would help tailor the aggressiveness of our advice to patients.