In a Dutch study reported in JACC: CardioOncology, Leerink et al found that the addition of ejection fraction measurement at the time of first surveillance echocardiogram improved prediction of 10-year risk of left-ventricular systolic dysfunction vs prediction based on cumulative anthracycline and chest-directed radiotherapy doses alone in long-term childhood cancer survivors.

As noted by the investigators, the current International Late Effects of Childhood Cancer Guideline Harmonization Group (IGHG) guidelines recommend performance of echocardiograms once every 5 years to facilitate early detection and treatment of asymptomatic left-ventricular dysfunction in all childhood cancer survivors who received cardiotoxic cancer therapies.

Study Details

The study involved measurement of ejection fraction at time of first surveillance echocardiogram (initial ejection fraction) at > 5 years after cancer diagnosis in a derivation cohort of 299 childhood cancer survivors from a site in Amsterdam and a validation cohort of 218 childhood cancer survivors from a site in Nijmegen. Survivors in both cohorts were aged > 18 years at the first follow-up echocardiogram, and had been treated with anthracyclines, mitoxantrone, and/or chest-directed radiotherapy.

Survivors with a history of heart failure before the first follow-up echocardiogram ≥ 5 years after cancer diagnosis and asymptomatic survivors with ejection fraction < 40% before or at the first echocardiogram were excluded from both cohorts. In the derivation cohort, Cox regression models were constructed including cardiotoxic cancer treatment exposures with and without ejection fraction to estimate the probability of development of left-ventricular dysfunction with an ejection fraction < 40% (LVD40) at 10-year follow-up.

Outcomes and Predictive Performance in Derivation Cohort

In the derivation cohort, median age at cancer diagnosis was 7.22 years (25th–75th percentile = 4.01–11.71 years), median time from diagnosis to first follow-up echocardiogram was 16.74 years (25th-75th percentile = 11.83–23.15 years), and median age at first follow-up echocardiogram was 24.06 years (25th-75th percentile = 19.60–30.71 years). Median follow-up after first echocardiogram was 10.9 years (25th–75th percentile = 8.2–13.1 years).

The cumulative incidence of LVD40 at 10-year follow-up in the entire cohort was 3.7%. The cumulative LVD40 incidence was 11.0% in childhood cancer survivors with midrange initial ejection fraction (40%–49%; 13% of survivors) vs 2.6% in those with preserved ejection fraction (≥ 50%, P = .012).

In multivariate analysis adjusted for anthracycline and chest-directed radiotherapy, childhood cancer survivors with midrange ejection fraction had a higher risk of LVD40 vs those with preserved ejection fraction (hazard ratio [HR] = 7.8, 95% confidence interval [CI] = 2.1–29.5). Lower ejection fraction was associated with increased risk of LVD40 during follow-up, at a hazard ratio of 9.6 (95% CI = 2.8–32.6) per 10-point ejection fraction decrease.

The addition of ejection fraction to the prediction model consisting of anthracycline and chest-directed radiotherapy dose increased the integrated area under the receiver-operating characteristic curve (integrated AUC) from 0.74 (bias = 0.018, 95% CI = 0.55–0.84) to 0.87 (bias = 0.009, 95% CI = 0.71–0.98). The likelihood ratio test comparing predictive performance of the model with vs without ejection fraction had a P value < .0001.

Assessment of continuous net reclassification improvement (cNRI), indicating the proportion of  childhood cancer survivors with accurate change in predicted risk using the model with vs without ejection fraction, showed no significant improvement in risk reclassification among LVD40 cases (cNRI = 0.15, 95% CI = –0.55–0.84), but significant improvement among childhood cancer survivors who did not develop LVD40 (cNRI = 0.50, 95% CI = 0.40–0.60). Based on the model including ejection fraction, 10-year predicted risk of LVD40 was ≤ 3% for 76.3% of childhood cancer survivors, with the model showing sensitivity of 89.8%, specificity of 76.2%, negative predictive value of 99.5%, and positive predictive value of 12.0%.

Performance in Validation Cohort

In the validation cohort, median age at cancer diagnosis was 7.02 years (25th–75th percentile = 4.00–12.46 years), median time from diagnosis to first follow-up echocardiogram was 16.95 years (25th-75th percentile = 12.99–21.70 years), and median age at first follow-up echocardiogram was 22.63 years (25th–75th percentile = 20.05–28.06 years). Median follow-up after first follow-up echocardiogram was 8.9 years (25th–75th percentile = 5.2–10.9 years). The cumulative incidence of LVD40 at 10-year follow-up was 3.6%.

For the model including ejection fraction vs the model including only anthracycline and radiotherapy dose, the integrated AUC increased from 0.72 (bias = –0.003, 95% CI = 0.70–0.77) to 0.86 (bias = –0.003, 95% CI = 0.83–0.89; likelihood ratio P < .001). Based on the model including ejection fraction, predicted 10-year probability of LVD40 was ≤ 3% in 74.8% of childhood cancer survivors, with the model having sensitivity of 85.1%, specificity of 77.1%, negative predictive value of 99.3%, and positive predictive value of 12.2%.  

The investigators concluded: “In childhood cancer survivors, an initial surveillance ejection fraction, in addition to anthracyclines and chest-directed radiotherapy dose, improves the 10-year prediction for LVD40. Through this strategy, both the identification of low-risk survivors in whom the surveillance frequency may be reduced and a group of survivors at increased risk of LVD40 could be identified.”

The prediction model, including a surveillance ejection fraction, cumulative anthracycline, and chest-directed radiotherapy dose, is available online at risk-of-cardiomyopathy.netlify.com.

Jan M. Leerink, MD, of Amsterdam UMC, University of Amsterdam, Department of Clinical and Experimental Cardiology, is the corresponding author for the JACC: CardioOncology article.

Disclosure: The study was supported by the Dutch Heart Foundation. For full disclosures of the study authors, visit jacc.org.