Improved Cardiomyopathy Risk Prediction Using Global Longitudinal Strain and N-Terminal-Pro-B-Type Natriuretic Peptide in Survivors of Childhood Cancer Exposed to Cardiotoxic Therapy

Matthew J Ehrhardt; Qi Liu; Daniel A Mulrooney; Isaac B Rhea; Stephanie B Dixon; John T Lucas, Jr; Yadav Sapkota; Kyla Shelton; Kirsten K Ness; Deo Kumar Srivastava; Aaron McDonald; Leslie L Robison; Melissa M Hudson; Yutaka Yasui; Gregory T Armstrong

doi:10.1200/JCO.23.01796

. 2024 Jan 11;42(11):1265–1277. doi: 10.1200/JCO.23.01796

Improved Cardiomyopathy Risk Prediction Using Global Longitudinal Strain and N-Terminal-Pro-B-Type Natriuretic Peptide in Survivors of Childhood Cancer Exposed to Cardiotoxic Therapy

Matthew J Ehrhardt ^1,^2,^✉, Qi Liu ³, Daniel A Mulrooney ^1,², Isaac B Rhea ⁴, Stephanie B Dixon ^1,², John T Lucas Jr ⁵, Yadav Sapkota ², Kyla Shelton ², Kirsten K Ness ², Deo Kumar Srivastava ⁶, Aaron McDonald ², Leslie L Robison ², Melissa M Hudson ^1,², Yutaka Yasui ², Gregory T Armstrong ²

PMCID: PMC11095874 PMID: 38207238

Abstract

PURPOSE

To leverage baseline global longitudinal strain (GLS) and N-terminal-pro-B-type natriuretic peptide (NT-proBNP) to identify childhood cancer survivors with a normal left ventricular ejection fraction (LVEF) at highest risk of future treatment-related cardiomyopathy.

METHODS

St Jude Lifetime Cohort participants ≥5 years from diagnosis, at increased risk for cardiomyopathy per the International Guideline Harmonization Group (IGHG), with an LVEF ≥50% on baseline echocardiography (n = 1,483) underwent measurement of GLS (n = 1,483) and NT-proBNP (n = 1,052; 71%). Multivariable Cox regression models estimated hazard ratios (HRs) and 95% CIs for postbaseline cardiomyopathy (modified Common Terminology Criteria for Adverse Events ≥grade 2) incidence in association with echocardiogram-based GLS (≥–18) and/or NT-proBNP (>age-sex-specific 97.5th percentiles). Prediction performance was assessed using AUC in models with and without GLS and NT-proBNP and compared using DeLong's test for IGHG moderate- and high-risk individuals treated with anthracyclines.

RESULTS

Among survivors (median age, 37.6; range, 10.2-70.4 years), 162 (11.1%) developed ≥grade 2 cardiomyopathy 5.1 (0.7-10.0) years from baseline assessment. The 5-year cumulative incidence of cardiomyopathy for survivors with and without abnormal GLS was, respectively, 7.3% (95% CI, 4.7 to 9.9) versus 4.4% (95% CI, 3.0 to 5.7) and abnormal NT-proBNP was 9.9% (95% CI, 5.8 to 14.1) versus 4.7% (95% CI, 3.2 to 6.2). Among survivors with a normal LVEF, abnormal baseline GLS and NT-proBNP identified anthracycline-exposed, IGHG-defined moderate-/high-risk survivors at a four-fold increased hazard of postbaseline cardiomyopathy (HR, 4.39 [95% CI, 2.46 to 7.83]; P < .001), increasing to a HR of 14.16 (95% CI, 6.45 to 31.08; P < .001) among survivors who received ≥250 mg/m² of anthracyclines. Six years after baseline, AUCs for individual risk prediction were 0.70 for models with and 0.63 for models without GLS and NT-proBNP (P = .022).

CONCLUSION

GLS and NT-proBNP should be considered for improved identification of survivors at high risk for future cardiomyopathy.

INTRODUCTION

Despite improvements in overall survival rates for children with cancer, long-term survivors continue to incur increased cardiac morbidity and mortality compared with the general population.^1-7 Cardiac conditions are the leading cause of non–cancer-related death 30 years after cancer diagnosis,⁷ accounting for a four-fold excess risk of cardiovascular mortality in childhood cancer survivors over that observed in the age-matched, general population.⁸ Risk of cardiomyopathy is largely attributable to previous exposure to chest-directed radiation and/or anthracycline chemotherapy, with the highest risk observed in those exposed to both.^3-5,9-13 Approximately two thirds of the estimated 500,000 childhood cancer survivors in the United States received one or both treatment exposures, rendering many potentially vulnerable to long-term cardiotoxicity.^13-16

CONTEXT

Key Objective
In childhood cancer survivors with a normal left ventricular ejection fraction (defined as ≥50%) on screening echocardiography, can abnormal global longitudinal strain (GLS) and N-terminal-pro-B-type natriuretic peptide (NT-proBNP) identify those at increased risk of developing future cardiomyopathy?
Knowledge Generated
Abnormal baseline GLS and NT-proBNP identified anthracycline-exposed survivors at a four-fold increased hazard of developing future cardiomyopathy. This risk increased to 14-fold among those who received ≥250 mg/m² of anthracyclines without radiation, identifying individuals who, despite a normal ejection fraction, are at substantially increased short-interval risk of cardiomyopathy.
Relevance (S. Bhatia)
This study identifies potential biomarkers to identify childhood cancer survivors at risk for developing clinically overt cardiomyopathy, setting the stage for targeted interventions among those at highest risk.*

*Relevance section written by JCO Associate Editor Smita Bhatia, MD, MPH, FASCO.

Symptomatic cardiomyopathy is typically preceded by a period of asymptomatic left ventricular dysfunction (ALVD). Left untreated, ALVD is expected to progress to symptomatic cardiomyopathy, leading to a nine-fold increase in the standardized mortality ratio in childhood cancer survivors relative to peers.⁸ Surveillance guidelines have recommended routine echocardiography for survivors exposed to cardiotoxic cancer therapies to maximize early detection and effective intervention before the development of symptomatic cardiomyopathy.^13,17 To date, surveillance intensity and risk prediction have relied upon cumulative doses of cardiotoxic treatment and baseline patient characteristics (eg, age at exposure) and have not accounted for subsequent risk factor acquisition or changes in subclinical biomarkers (eg, N-terminal pro-B-type natriuretic peptide [NT-proBNP] or global longitudinal strain [GLS]). Clear understanding of the benefit of treatment will take decades to achieve, thus it has become common practice to treat anthracycline- or radiation-associated ALVD similarly to adults with ACC/AHA stage B heart failure. Adopting this approach, such surveillance and intervention approaches have been shown to be clinically effective and cost-effective.^18-20

As observed in the general population,^21,22 earlier ALVD detection in cancer survivors may lead to improved effectiveness of interventions to mitigate progression to symptomatic disease. However, identification of those with normal cardiac function assessed by left ventricular ejection fraction (LVEF), who will go on to develop ALVD and symptomatic cardiomyopathy, is dependent upon modalities with greater sensitivity than LVEF alone. Echocardiography-based GLS, for example, has been shown to have superior sensitivity compared with LVEF for detecting cardiac dysfunction.^23-25 Although abnormal GLS has been shown to be associated with increasing cumulative doses of chest-directed radiation and anthracycline exposure in childhood cancer survivors,²⁵ its association with subsequent symptomatic cardiomyopathy remains unknown. Similarly, serum biomarkers (eg, cardiac troponins and NT-proBNP) have been explored as possible indicators of early cardiac injury in long-term survivors. Most report generally low sensitivity (8%-32%) and high specificity (81%-100%) for NT-proBNP^26-37 and similarly low sensitivity for troponin T and I^{26,27,30,32,33,35-37} regarding detection of asymptomatic cardiomyopathy. Enhanced ability to identify survivors at highest risk of subsequently developing symptomatic cardiomyopathy using such biomarkers, beyond the capabilities permitted by treatment exposures and demographic characteristics alone, would have substantial implications on prioritization of interventions seeking to delay or prevent the onset or progression of heart disease.

METHODS

Study Design

We used the institutional review board–approved St Jude Lifetime Cohort study (SJLIFE) cohort, a longitudinal study with prospective clinical health assessments previously described.³⁸ In short, 5-year survivors of childhood cancer, diagnosed between 1962 and 2012 and treated at St Jude Children's Research Hospital, were eligible for study inclusion. Participants undergo systematic medical record abstraction for treatment exposures and chronic health conditions, and prospective, comprehensive laboratory and clinical assessments at study entry and recurring at approximately 5-year intervals.

Study Subjects

We included all SJLIFE participants who completed a baseline echocardiogram at the time of or after their first study visit/campus assessment (≥5 years from cancer diagnosis) that demonstrated a LVEF ≥50% without a history of or ongoing treatment for cardiomyopathy and at least one subsequent echocardiogram (Fig 1).

FIG 1. — Flow diagram. GLS, global longitudinal strain; SJLIFE, St Jude Lifetime Cohort study.

Treatment Exposures

Chemotherapy and radiation exposures were systematically abstracted from medical records.³⁸ Cumulative anthracycline doses were calculated using published doxorubicin-equivalent anthracycline dose ratios.^13,39,40 Maximum dose to the chest was estimated from prescribed radiation fields with potential exposure to the heart.^17,41 The decision to use chest-region dosimetry rather than heart dosimetry was made to use data readily available in most survivorship clinics for risk stratification to result in a clinical practice–relevant model.

Outcomes

The primary outcome was grade 2-5 cardiomyopathy subsequent to the baseline SJLIFE evaluation, defined according to the modified Common Terminology Criteria for Adverse Events (CTCAE) v4.03 (Appendix Table A1, online only).⁴² Cardiomyopathy was primarily assessed by echocardiography and supplemented by clinical records, including documentation of a confirmed diagnosis by a clinical provider and/or pharmacologic intervention.

Predictors

Demographics and baseline number and severity of cardiovascular risk factors (hypertension, diabetes, obesity, and dyslipidemia), graded according to the modified CTCAE v4.03 described above,⁴² were obtained/assessed during the campus visit and from medical record review.³⁸

Echocardiograms were obtained using a Vivid 7 machine (GE Medical Systems, Milwaukee, WI) and indices measured according to the American Society of Echocardiography standards.^43,44 LVEF was measured using the Simpson's biplane method.⁴³ Echocardiograms were read by trained cardiologists. Left ventricular GLS was assessed using speckle tracking–based strain imaging software (EchoPAC PC version 10.0; GE HealthCare, Chicago, IL).⁴⁴ Abnormal GLS was defined as ≥–18% and obtained using two-dimensional echocardiography. Plasma NT-proBNP was collected at baseline and measured with an electrochemiluminescence immunoassay (Elecsys 2008; Roche Diagnostics, Indianapolis, IN), for which the lower limit of detection was 5 pg/mL. NT-proBNP was defined as abnormal if exceeding the 97.5th percentile, consistent with previously used definitions.^27,45 Because NT-proBNP was initially not part of the routine SJLIFE baseline assessment, it was assigned as unavailable for patients assessed early on during study enrollment. Exploratory variables included left ventricular end-diastolic and -systolic volumes (normal v above normal),⁴⁶ left ventricular diastolic dimension (normal v above normal),⁴⁶ cardiopulmonary fitness (low [<7.9 metabolic equivalents of task, METS] v high [≥7.9 METS]),⁴⁷ exercise intolerance (<85% v ≥85 predicted),^48,49 cardiac remodeling phenotype (normal v abnormal geometry),⁴⁶ and single-nucleotide polymorphisms shown to be associated with cardiomyopathy and replicated in at least one or more independent sample (Appendix Table A2).^50-57

International Guideline Harmonization Group Risk Stratification

We used the International Guideline Harmonization Group (IGHG) risk stratification for subsequent analyses, defined according to previous anthracycline (in doxorubicin-equivalent doses) and/or chest radiation exposure as follows: low (1-99 mg/m² and chest-directed radiotherapy <15 Gy), moderate (100-249 mg/m² or chest-directed radiotherapy 15-29 Gy), and high (≥250 mg/m², chest-directed radiotherapy ≥30 Gy, or a combination of cumulative anthracyclines ≥100 mg/m² and chest-directed radiotherapy ≥15 Gy).

Statistical Analysis

Descriptive statistics were calculated for the total study population and those who developed postbaseline cardiomyopathy. At-risk follow-up for cardiomyopathy incidence began at the baseline SJLIFE visit and ended at first occurrence of cardiomyopathy or the most recent visit as of April 30, 2020, whichever occurred earlier. Cumulative incidence was estimated and plotted by IGHG risk group, GLS, and NT-proBNP status.

Multivariable Cox regression models estimated hazard ratios (HRs) and 95% CIs for postbaseline cardiomyopathy incidence for the abnormal GLS and/or NT-proBNP at baseline in each risk group defined by the combinations of IGHG risk groups and anthracycline exposure (hereafter referred to as IGHGxAnthracycline groups), adjusted for age at baseline SJLIFE visit, age at diagnosis, sex, race/ethnicity, number of cardiovascular risk factors at baseline, and IGHGxAnthracycline groups. We performed receiver operating characteristic (ROC) analyses to compare the predictive ability of models with and without GLS and NT-proBNP at 6 years from baseline SJLIFE assessment. Predictive performance of each model was measured by AUC using the time-dependent AUC method of Heagerty and Zhang.⁵⁸ Statistical significance of the AUC difference with and without the GLS and NT-proBNP was evaluated using the DeLong test.⁵⁹ AUCs were compared between the two models for all exposed survivors and IGHG moderate- and high-risk individuals exposed to anthracyclines. To more clearly describe the impact of GLS and NT-proBNP on prediction power, we calculated the positive predictive value (PPV) and negative predictive value (NPV) at 6 years from baseline. Specifically, the PPV was estimated by the proportion of survivors who developed cardiomyopathy within 6 years among survivors with the 162 highest model-predicted risk values (115 IGHG moderate- or high-risk individuals). The NPV was estimated by the proportion who did not develop cardiomyopathy within 6 years among the survivors who were not among the 162 at highest predicted risk. Proportions were estimated in the presence of competing risk (death due to causes other than cardiomyopathy) and censoring; thus, we used the calculation of cumulative incidence in this estimation.

RESULTS

Among 1,483 survivors exposed to anthracyclines and/or chest radiation, the median (range) age was 8.0 (0-23.5) years at diagnosis, 37.6 (10.2-70.4) years at SJLIFE assessment, 20.8 (7.2-47.7) years from cancer diagnosis to baseline assessment, and 5.6 (0.6-11.4) years from baseline assessment to last follow-up (Table 1). All underwent an echocardiogram with GLS and 1,052 (70.9%) had NT-proBNP assessed at baseline. Most participants were non-Hispanic White (83%), whereas sex was evenly distributed (51% male). Baseline GLS values and NT-proBNP levels are shown in Figure 2, stratified by IGHG risk group and development of postbaseline cardiomyopathy. Baseline frequencies of exploratory variables for the total population and those who developed postbaseline cardiomyopathy are shown in Appendix Table A3.

TABLE 1.

Baseline Characteristics of the Total Study Population and Those Who Developed Postbaseline Cardiomyopathy

Characteristic	Total (N = 1,483)	Cardiomyopathy (n = 162)
Age at diagnosis, years, median (range)	8.0 (0-23.5)	9.5 (0.1-21.2)
Age at assessment, years, median (range)	37.6 (10.2-70.4)	41.7 (15.4-61.5)
Time from cancer diagnosis to baseline echo, years, median (range)	20.8 (7.2-47.7)	22.5 (9.4-45.3)
No. of echocardiograms per patient, median (range)	2 (2-9)	3 (2-9)
Duration between baseline and last follow-up, years, median (range)	5.6 (0.6-11.4)	5.1 (0.7-10.0)
Chest radiation dose, Gy, median (range)	26.0 (1.5-62.0)	26.0 (4.5-45.0)
Anthracycline dose, mg/m², median (range)	149.5 (12.3-589.2)	201.8 (24.6-531.1)
Race/ethnicity, No. (%)
White non-Hispanic	1,227 (83)	129 (80)
Black non-Hispanic	178 (12)	28 (17)
Hispanic	51 (3)	2 (1)
Other	27 (2)	3 (2)
Sex, No. (%)
Male	756 (51)	89 (55)
Female	727 (49)	73 (45)
Age at diagnosis, years, No. (%)
0-4	520 (35)	45 (28)
5-9	340 (23)	39 (24)
10-14	360 (24)	46 (28)
≥15	263 (18)	32 (20)
Year of diagnosis, No. (%)
1960-1969	26 (2)	3 (2)
1970-1979	205 (14)	30 (19)
1980-1989	492 (33)	62 (38)
1990-1999	592 (40)	53 (33)
≥2000	168 (11)	14 (9)
Age at assessment, years, No. (%)
5-14	14 (1)	0 (0)
15-19	24 (2)	3 (2)
20-24	24 (2)	1 (1)
25-29	204 (14)	11 (7)
30-39	609 (41)	56 (35)
40-49	428 (29)	62 (38)
≥50	180 (12)	29 (18)
Age at baseline echocardiogram, years, No. (%)
5-9	3 (0)	0 (0)
10-14	16 (1)	1 (1)
15-19	66 (4)	5 (3)
20-24	336 (23)	23 (14)
25-29	332 (22)	28 (17)
30-34	325 (22)	41 (25)
35-39	201 (14)	32 (20)
40-44	124 (8)	17 (10)
≥45	80 (5)	15 (9)
Cancer diagnosis, No. (%)
ALL	571 (39)	39 (24)
AML	82 (6)	7 (4)
Hodgkin lymphoma	233 (16)	38 (23)
Non-Hodgkin lymphoma	131 (9)	17 (10)
CNS	81 (5)	6 (4)
Bone tumor	118 (8)	22 (14)
Soft tissue sarcoma	61 (4)	15 (9)
Wilms tumor	104 (7)	7 (4)
Neuroblastoma	72 (5)	9 (6)
Retinoblastoma	5 (0)	1 (1)
Germ cell tumor	2 (0)	0 (0)
Other	23 (2)	1 (1)
Cancer treatment, No. (%)
Chest radiation (yes/no)	507 (34)	77 (48)
Anthracyclines (yes/no)	1,274 (86)	134 (83)
Modifiable cardiovascular risk factors (CTCAE grade), No. (%)
Hypertension (≥grade 2)	275 (19)	38 (23)
Abnormal glucose metabolism (≥grade 2)	133 (9)	18 (11)
Obesity (≥grade 3)	464 (31)	56 (35)
Dyslipidemia (≥grade 2)	236 (16)	37 (23)
IGHG risk group,^a No. (%)
Low	469 (32)	23 (14)
Moderate	490 (33)	40 (25)
High	524 (35)	99 (61)
Baseline global longitudinal strain, No. (%)
<–18%	1,041 (70)	93 (57)
≥–18%	442 (30)	69 (43)
N-terminal-pro-B-type natriuretic peptide, No. (%)
Normal	815 (55)	76 (47)
Abnormal	237 (16)	47 (29)
Unavailable	431 (29)	39 (24)

Open in a new tab

Abbreviations: CTCAE, Common Terminology Criteria for Adverse Events; IGHG, International Guideline Harmonization Group.

^{^a}

Low (cumulative doxorubicin-equivalent anthracycline dose 1-99 mg/m² and chest-directed radiotherapy <15 Gy), moderate (cumulative doxorubicin-equivalent anthracycline dose 100-249 mg/m² or chest-directed radiotherapy 15-29 Gy), and high (cumulative doxorubicin-equivalent anthracycline dose ≥250 mg/m², chest-directed radiotherapy ≥30 Gy, or a combination of cumulative doxorubicin-equivalent anthracycline dose ≥100 mg/m² and chest-directed radiotherapy ≥15 Gy).

FIG 2. — Baseline (A) GLS values and (B) NT-proBNP levels by IGHG low- (cumulative doxorubicin-equivalent anthracycline dose 1-99 mg/m² and chest-directed radiotherapy <15 Gy), moderate- (cumulative doxorubicin-equivalent anthracycline dose 100-249 mg/m² or chest-directed radiotherapy 15-29 Gy), and high-risk (cumulative doxorubicin-equivalent anthracycline dose ≥250 mg/m², chest-directed radiotherapy ≥30 Gy, or a combination of cumulative doxorubicin-equivalent anthracycline dose ≥100 mg/m² and chest-directed radiotherapy ≥15 Gy) groups for those who did and did not develop postbaseline cardiomyopathy during study follow-up. The horizontal line within each box represents the median value of GLS and NT-proBNP levels, the box encompasses the interquartile range (from the lower quartile to the upper quartile), the upper horizontal line represents the maximum value within 1.5 times the interquartile range from the upper quartile, and circles above the horizontal lines represent additional values exceeding that value. For 14 patients, NT-proBNP exceeded 500. Respective values for these were 504, 515, 531, 560, 610, 711, 768, 817, 867, 1,048, 1,293, 3,127, 3,421, and 11,351 pg/mL. To improve boxplot visibility, these values were plotted as 500 pg/mL. GLS, global longitudinal strain; IGHG, International Guideline Harmonization Group; NT-proBNP, N-terminal-pro-B-type natriuretic peptide.

One hundred sixty-two (10.9%) developed ≥grade 2 cardiomyopathy at a median follow-up of 5.1 (0.7-10.0) years from baseline assessment. Among these, the median age was 9.5 (0.1-21.2) years at cancer diagnosis and 41.7 (15.4-61.5) years at baseline SJLIFE assessment, while the median time from cancer diagnosis to baseline assessment was 22.5 (9.4-45.3) years. The distribution of race/ethnicity and sex was similar to the overall population (Table 1). The 5-year cumulative incidence of cardiomyopathy for survivors with and without abnormal GLS was, respectively, 7.3% (95% CI, 4.7 to 9.9) versus 4.4% (95% CI, 3.0 to 5.7) and abnormal NT-proBNP was 9.9% (95% CI, 5.8 to 14.1) versus 4.7% (95% CI, 3.2 to 6.2; Fig 3).

FIG 3. — Cumulative incidence (in percent) of ≥grade 2 cardiomyopathy by years from baseline St Jude Lifetime Cohort study echocardiogram by (A) IGHG risk group, (B) baseline GLS status, and (C) baseline NT-proBNP status. GLS, global longitudinal strain; IGHG, International Guideline Harmonization Group; NT-proBNP, N-terminal-pro-B-type natriuretic peptide.

IGHG moderate- and high-risk survivors exposed to anthracyclines with a normal GLS had a two-fold increased hazard rate of postbaseline cardiomyopathy if NT-proBNP was abnormal (HR, 2.01 [95% CI, 1.11 to 3.66]; P = .022) or unavailable (HR, 2.32 [95% CI, 1.20 to 4.48]; P = .012), compared with those with both a normal at baseline, adjusting for the IGHGxAnthracycline groups. Survivors had a 4.39-fold increased hazard rate when both baseline GLS and NT-proBNP were abnormal (HR, 4.39 [95% CI, 2.46 to 7.83]; P < .001; Fig 4) or when GLS was abnormal and NT-proBNP unavailable (HR, 4.14 [95% CI, 2.12 to 8.08]; P < .001), adjusting for the IGHGxAnthracycline groups. Abnormal GLS and/or NT-proBNP were not associated with increased risk of cardiomyopathy in IGHG low-risk survivors or in those defined as moderate- to high-risk due to chest radiation only.

FIG 4. — Multivariable associations between IGHG risk group, baseline GLS, baseline N-terminal-pro-B-type natriuretic peptide, and ≥grade 2 postbaseline cardiomyopathy. Adjusted for race, sex, age, age at cancer diagnosis, number of cardiovascular risk factors (hypertension, diabetes, obesity, and dyslipidemia), and IGHG risk group. Values in bold indicate P < .05; ^acumulative doxorubicin-equivalent anthracycline dose <100 mg/m² and chest radiation <15 Gy; ^b≥15 Gy; ^ccumulative doxorubicin-equivalent anthracycline dose ≥100 mg/m². GLS, global longitudinal strain; IGHG, International Guideline Harmonization Group; NT-proBNP, N-terminal-pro-B-type natriuretic peptide; Ref, reference.

Associations were more pronounced in individuals exposed to IGHG-defined high-risk doses of anthracyclines (ie, ≥250 mg/m²) without chest radiation (Appendix Table A4), wherein survivors had a 14-fold increased risk of postbaseline cardiomyopathy incidence if GLS and NT-proBNP were abnormal (HR, 14.16 [95% CI, 6.45 to 31.08]; P < .001) or if GLS was abnormal and NT-proBNP unavailable (HR, 13.36 [95% CI, 5.56 to 32.10]; P < .001) compared with IGHG-defined low-risk survivors (ie, <100 mg/m² and <15 Gy chest radiation) with normal GLS and NT-proBNP levels at baseline.

The results from exploratory models assessing associations with each predictor variable, adjusted for age at diagnosis, race, sex, and age at baseline assessment, are found in Appendix Table A5. Variant rs2229774 was significantly associated with cardiomyopathy (HR, 0.33 [95% CI, 0.15 to 0.71]; P = .005); however, no other exploratory variables were found to be associated with the development of subsequent cardiomyopathy. We repeated the previous multivariable model (Fig 4), adjusting for rs2229774, and observed similar associations (Appendix Table A6).

Among IGHG moderate- and high-risk individuals, 6-year AUCs were 0.74 and 0.70 for models with and without GLS and NT-proBNP, respectively (P = .066; Fig 5). Among these individuals but limiting to those with anthracycline exposure, AUCs were 0.70 and 0.63 with and without GLS and NT-proBNP, respectively (P = .022). The PPV for the model including IGHG moderate- and high-risk individuals increased from 0.220 to 0.316 with the addition of GLS and NT-proBNP. Similarly, among individuals not at IGHG moderate- or high-risk, the NPV increased from 0.871 to 0.888 by adding GLS and NT-proBNP.

DISCUSSION

We observed that the combination of abnormal baseline GLS and NT-proBNP was associated with a four-fold risk of postbaseline cardiomyopathy in childhood cancer survivors already designated as being IGHG-defined moderate- to high-risk based upon treatment exposures alone (doxorubicin-equivalent cumulative anthracycline dose ≥100 mg/m² with or without chest radiation ≥15 Gy). This association was not present in survivors exposed only to chest radiation, suggesting that tailored surveillance assessments may be warranted in survivors based upon treatment exposures received.

Previous groups have reported associations between early reductions in left ventricular function and subsequent symptomatic cardiomyopathy in survivor cohorts but did not assess GLS in survivors with normal LVEF (ie, the majority of survivors) and were limited by relatively short follow-up from treatment exposure (ie, most cases <10 years) and heterogeneous outcome definitions.^60-63 Recently, midrange (40%-49%) but not preserved LVEFs obtained at entry into long-term follow-up in survivors exposed to anthracyclines and chest-directed radiation were shown to be predictive of subsequent development of an LVEF <40%.⁶⁴ Notably, when taken as a whole, individuals with preserved LVEFs were at low risk for cardiomyopathy 10 years from evaluation, suggesting they may not require the more frequent surveillance recommended for those with higher anthracycline exposures. Conversely, by using baseline echocardiographic and serum biomarkers, we identified a subset of individuals with a preserved LVEF at baseline at a significantly increased risk of short-interval development of cardiomyopathy, suggesting that the combination of these biomarkers yields greater sensitivity to detect subsequent cardiac dysfunction.

Our study was strengthened by a comprehensive analysis of several exploratory variables. Although not significantly associated with subsequent cardiomyopathy, lack of association with these variables is particularly valuable when establishing clinically relevant risk prediction models. Of interest was the lack of association with genetic variants previously shown to be associated with anthracycline-related cardiomyopathy.^50-57 We hypothesize that this is related to the fact that the association of GLS and NT-proBNP with cardiomyopathy we observed was identified among individuals already at substantial risk due to treatment exposures alone. Consequently, the strength of the association between these abnormal biomarkers and subsequent cardiomyopathy likely exceeded that contributed by genetics. A similar paradigm has been observed, wherein an otherwise significant association between cancer predisposition mutations and subsequent malignancies in cancer survivors was mitigated in those who received previous radiotherapy.⁶⁵

Although reduced GLS has been reported at higher rates in childhood cancer survivors compared with control populations,²⁵ to our knowledge, it has not been previously used in longitudinal cohorts to identify survivors at highest risk of subsequent cardiomyopathy. Using the longitudinally assessed SJLIFE cohort, we identified individuals at heightened and previously unrecognized risk for short-interval cardiac decline. Existing studies suggest more frequent surveillance intervals are unlikely to be cost-effective even in highest-risk individuals^18-20; therefore, our findings identify a population that may benefit from tertiary (ie, cardioprotective pharmacologic intervention) rather than secondary prevention (ie, surveillance). Such an approach was used in the SUCCOUR trial, which randomly assigned adults actively receiving anthracyclines for cancer to dual cardioprotective therapy with an ACE-I and beta-blocker on the basis of either a ≥12% relative reduction in GLS or a >10% absolute reduction in LVEF. Although 1-year follow-up suggested individuals in the EF-guided arm developed higher rates of chemotherapy-related cardiovascular disease (13.7% v 5.8%; P = .011) and greater decline in LVEF (9.1% v 2.9%; P = .03) compared with the GLS-guided arm,⁶⁶ 3-year follow-up demonstrated that participants in both arms experienced similarly improved LVEFs.⁶⁷ Although the low overall frequency of events likely precluded the ability to detect potentially significant differences between groups, these findings emphasize the importance of additional validation of the study's findings in an independent cohort or of conducting a prospective trial investigating whether or not prophylactic neurohormonal blockade would benefit childhood cancer survivors at highest risk, such as those identified in our study with a history of high-risk exposures and abnormal GLS and NT-proBNP despite a normal EF. Accordingly, a strategy using cardiotoxic chemotherapy exposures and baseline GLS and NT-proBNP values offers a promising path toward precision survivorship.

Our results are timely, given the recently reported findings from the randomized, double-blind, placebo-controlled phase IIB trial (ClinicalTrials.gov identifier: NCT02717507), which suggested that among survivors exposed to doxorubicin-equivalent anthracycline doses ≥250 mg/m² with a baseline LVEF ≥55% and/or shortening fraction ≥28%, 2-year administration of low-dose carvedilol was significantly associated with better left ventricular end-diastolic diameter and end-systolic wall stress compared with placebo.⁶⁸ Notably, treatment effects varied by time since cancer diagnosis (P = .008), suggesting a greater benefit to cardioprotective therapy with carvedilol in longer-term survivors. In a second trial, randomized atorvastatin administration in adults with lymphoma receiving anthracyclines was shown to be associated with a reduction in the incidence of an absolute decline in LVEF ≥10% from baseline to an LVEF <55% at 1 year (9% atorvastatin v 22% placebo; P = .002).⁶⁹ Although longer follow-up of both cohorts is necessary to understand the sustainability of these effects, the results raise the possibility of prophylactic cardioprotective therapy, for example, with carvedilol for all survivors who received ≥250 mg/m² of anthracyclines. However, an all-inclusive approach would carry significant costs and burden on an already overtasked health care system. Conversely, we identify those among an otherwise carvedilol-eligible cohort (ie, ≥250 mg/m² of anthracyclines and normal LVEF) for whom the short-interval risk of cardiomyopathy is profound (ie, ≥250 mg/m², abnormal baseline GLS and NT-proBNP, and normal LVEF), and in whom a prophylactic treatment approach may be most beneficial.

Our findings should be considered in the context of important limitations. First, the timing of the baseline assessment was variable because of SJLIFE enrollment practices. In many cases, survivors had been discharged from the institution for many years before opening of and enrollment onto SJLIFE, and thus were, on average, age 37 years and approximately 20 years from diagnosis at baseline assessment. As a result, our risk estimates may not apply to individuals whose baseline surveillance echocardiograms occur at younger ages. Additionally, the median interval follow-up between baseline echocardiogram and last follow-up was 5.2 years. It is likely that our results underestimate the degree of risk of cardiac decline with longer follow-up, emphasizing the importance of the results for this group of individuals. In addition, we observed an increased risk of cardiomyopathy associated with unavailable NT-proBNP. We suspect this reflects the recruitment emphasis for leukemia and lymphoma survivors in the early years of SJLIFE, before routine NT-proBNP testing, and that if tested, many of these individuals would likely have an abnormal NT-proBNP. Finally, models incorporating GLS and NT-proBNP to understand the cost-effectiveness of both deintensification and intensification screening and/or pharmacologic intervention for those at lowest risk and highest risk, respectively, have not been performed. However, given the challenges associated with prospective tertiary prevention studies, we contend that the degree of risk demonstrated in our models identifies individuals for whom tertiary/pharmacologic intervention should be strongly considered.

In summary, we observed that the combination of abnormal baseline GLS and NT-proBNP was associated with a four-fold risk of postbaseline cardiomyopathy in childhood cancer survivors designated as IGHG-defined moderate- to high-risk on the basis of cardiotoxic cancer treatment exposures alone. The results suggest that more sensitive methods of detecting early myocardial dysfunction may predict those at greater risk for development of symptomatic cardiomyopathy and potentially identify individuals most likely to benefit from tertiary prevention strategies.

APPENDIX

TABLE A1.

Modified Common Terminology Criteria for Adverse Events v4.03 Grades of Cardiomyopathy

Grade	Definition
Grade 1	Not applicable
Grade 2	Resting EF <50%-40%; 10%-19% absolute drop from baseline
Grade 3	Resting EF 39%-20%; >20% absolute drop from baseline; medication indicated or initiated
Grade 4	Resting EF <20%; refractory or poorly controlled heart failure due to drop in ejection fraction; on medical management; intervention such as ventricular assist device, intravenous vasopressor support, or heart transplant indicated
Grade 5	Death

Open in a new tab

Abbreviation: EF, ejection fraction.

TABLE A2.

Genetic Variants Associated With Anthracycline-Associated Cardiomyopathy (replicated one or more times in an independent sample)

Single-Nucleotide Polymorphism	Risk Allele/Genotype	Other Allele/Genotype
rs17863783 (UGT1A6)⁴⁹	T	G
rs1786814 (CELF4)^50,^a	GA/AA	GG
rs2229774 (RARG)⁵¹	A	G
rs2232228 (HAS3)^52,^b	AA	GG
rs4149178 (SLC22A7)⁵³	G	A
rs4982753 (SLC22A17)⁵³	T	C
rs7627754 (ABCC5)⁵⁴	TT	AT/AA
rs7853758 (SLC28A3)⁴⁹	A	G
rs2815063 (Intergenic)⁵⁵	A	C
rs6689879 (MAGI3/PHTF1)^56,^c	C	T

Open in a new tab

^{^a}

rs1786814 (CELF4) assessed as SNP x anthracycline (>300 mg/m²).

^{^b}

rs2232228 (HAS3) assessed as SNP x anthracycline (>250 mg/m²).

^{^c}

Identified in African American survivors of childhood cancer.

TABLE A3.

Baseline Frequencies of Exploratory Variables for the Total Study Population and Those Who Developed Postbaseline Cardiomyopathy

Characteristic	Total (n = 1,483), No. (%)	Cardiomyopathy (n = 162), No. (%)
Left ventricular end-diastolic volume index, mL
Normal	1,260 (85)	135 (84)
Above normal	219 (15)	26 (16)
Left ventricular end-systolic volume index, mL
Normal	1,175 (79)	126 (78)
Above normal	305 (21)	35 (22)
Left ventricular diastolic dimension, mm
Normal	1,441 (97)	156 (97)
Above normal	38 (3)	5 (3)
Cardiopulmonary fitness
High (max ≥7.9 METS; Ref)	405 (43)	34 (39)
Low (max <7.9 METS)	540 (57)	54 (61)
Exercise intolerance
≥85% predicted (Ref)	371 (39)	32 (36)
<85% predicted	590 (61)	58 (64)
Remodeling phenotype
Normal geometry (Ref)	937 (67)	92 (62)
Abnormal	454 (33)	56 (38)
Single-nucleotide polymorphisms (genes)
rs17863783
G/G	1,281 (94)	137 (96)
G/T	84 (6)	5 (4)
T/T	2 (0)	0 (0)
rs1786814
A/A	38 (3)	7 (5)
G/A	423 (31)	33 (23)
G/G	904 (66)	102 (72)
rs2229774
A/A	0 (0)	0 (0)
G/A	190 (14)	7 (5)
G/G	1,176 (86)	135 (95)
rs2232228
A/A	531 (39)	57 (40)
A/G	620 (45)	67 (47)
G/G	214 (16)	18 (13)
rs4149178
A/A	939 (69)	91 (64)
A/G	372 (27)	39 (27)
G/G	56 (4)	12 (8)
rs4982753
C/C	797 (58)	86 (61)
C/T	489 (36)	48 (34)
T/T	79 (6)	8 (6)
rs7627754
A/A	1,021 (75)	99 (70)
A/T	309 (23)	38 (27)
T/T	37 (3)	5 (4)
rs7853758
A/A	51 (4)	7 (5)
G/A	351 (26)	36 (26)
G/G	964 (71)	98 (70)
rs2815063
A/A	58 (4)	9 (6)
C/A	354 (26)	42 (30)
C/C	955 (70)	91 (64)
rs6689879
C/C	114 (8)	15 (11)
T/C	567 (42)	58 (41)
T/T	685 (50)	69 (49)

Open in a new tab

NOTE. Counts and percentages are based on available data.

Abbreviations: METS, metabolic equivalents of task; Ref, reference.

TABLE A4.

Multivariable Associations Comparing Individuals Within IGHG-Defined Risk Groups Based Upon Cumulative Doses of Anthracyclines and Chest Radiation

Variable	HR (95% CI)	P
Race
White, non-Hispanic	Ref
Non-White	1.65 (1.11 to 2.45)	.014
Sex
Male	Ref
Female	0.86 (0.62 to 1.19)	.37
Age at diagnosis, years
0-4	Ref
5-9	1.30 (0.83 to 2.03)	.25
10-14	1.12 (0.71 to 1.77)	.64
15-19	1.00 (0.59 to 1.68)	.99
No. of cardiovascular risk factors
None	Ref
1	1.07 (0.72 to 1.59)	.73
≥2	1.54 (1.01 to 2.34)	.043
IGHG risk group
Low	Ref
Moderate, without anthracycline	0.90 (0.17 to 4.72)	.90
Moderate, with anthracycline	1.39 (0.64 to 3.02)	.41
High, without anthracycline	4.62 (1.77 to 12.07)	.002
High, with anthracycline	3.23 (1.54 to 6.75)	.002
IGHG low-risk
GLS normal/NT-proBNP normal	Ref
GLS abnormal/NT-proBNP normal	0.98 (0.30 to 3.14)	.97
GLS normal/NT-proBNP abnormal	0.00 (0.00 to 2.22)	.97
GLS abnormal/NT-proBNP abnormal	0.00 (0.00 to 6.36)	.99
GLS normal/NT-proBNP missing	1.50 (0.56 to 4.01)	.42
GLS abnormal/NT-proBNP missing	1.07 (0.23 to 4.96)	.94
IGHG moderate-risk without anthracycline
GLS normal/NT-proBNP normal	0.90 (0.17 to 4.72)	.90
GLS abnormal/NT-proBNP normal	0.77 (0.14 to 4.11)	.76
GLS normal/NT-proBNP abnormal	0.62 (0.09 to 4.12)	.62
GLS abnormal/NT-proBNP abnormal	0.93 (0.17 to 5.18)	.93
GLS normal/NT-proBNP missing	1.63 (0.15 to 18.07)	.69
GLS abnormal/NT-proBNP missing	0.86 (0.07 to 10.78)	.91
IGHG moderate-risk without radiation
GLS normal/NT-proBNP normal	1.39 (0.64 to 3.02)	.41
GLS abnormal/NT-proBNP normal	1.48 (0.63 to 3.50)	.37
GLS normal/NT-proBNP abnormal	2.80 (1.20 to 6.52)	.017
GLS abnormal/NT-proBNP abnormal	6.10 (2.58 to 14.42)	<.001
GLS normal/NT-proBNP missing	3.23 (1.38 to 7.53)	.007
GLS abnormal/NT-proBNP missing	5.76 (2.42 to 13.68)	<.001
IGHG high-risk without anthracycline
GLS normal/NT-proBNP normal	4.62 (1.77 to 12.07)	.002
GLS abnormal/NT-proBNP normal	3.96 (1.29 to 12.17)	.016
GLS normal/NT-proBNP abnormal	3.20 (0.86 to 11.93)	.083
GLS abnormal/NT-proBNP abnormal	4.77 (1.67 to 13.62)	.004
GLS normal/NT-proBNP missing	8.37 (1.03 to 68.17)	.047
GLS abnormal/NT-proBNP missing	4.44 (0.56 to 35.54)	.16
IGHG high-risk without radiation
GLS normal/NT-proBNP normal	3.23 (1.54 to 6.75)	.002
GLS abnormal/NT-proBNP normal	3.43 (1.51 to 7.82)	.003
GLS normal/NT-proBNP abnormal	6.49 (2.91 to 14.50)	<.001
GLS abnormal/NT-proBNP abnormal	14.16 (6.45 to 31.08)	<.001
GLS normal/NT-proBNP missing	7.48 (3.09 to 18.14)	<.001
GLS abnormal/NT-proBNP missing	13.36 (5.56 to 32.10)	<.001

Open in a new tab

NOTE. Adjusted for age at baseline.

Abbreviations: GLS, global longitudinal strain; HR, hazard ratio; IGHG, International Guideline Harmonization Group; NT-proBNP, N-terminal-pro-B-type natriuretic peptide; Ref, reference.

TABLE A5.

Model Results Evaluating Associations Between Exploratory Variables and Subsequent Cardiomyopathy

Variable	HR (95% CI)	P
LVEDVi, mL
Normal	Ref
Above normal	1.16 (0.76 to 1.79)	.49
LVESVi, mL
Normal	Ref
Above normal	1.18 (0.80 to 1.73)	.41
Left ventricular diastolic dimension, mm
Normal	Ref
Above normal	1.02 (0.42 to 2.51)	.96
Cardiopulmonary fitness
High (max ≥7.9 METS)	Ref
Low (max <7.9 METS)	0.88 (0.54 to 1.45)	.62
Exercise intolerance
≥85% predicted	Ref
<85% predicted	1.08 (0.70 to 1.69)	.72
Remodeling phenotype
Normal geometry	Ref
Abnormal	0.89 (0.62 to 1.27)	.53
Single-nucleotide polymorphisms
rs17863783—continuous 0/1/2 with 0 = GG	0.49 (0.20 to 1.20)	.12
rs1786814—GA/AA 1/0	0.86 (0.58 to 1.25)	.42
rs2229774—continuous 0/1/2 with 0 = GG	0.33 (0.15 to 0.71)	.005
rs4149178—continuous 0/1/2 with 0 = AA	1.23 (0.93 to 1.64)	.14
rs4982753—continuous 0/1/2 with 0 = CC	0.95 (0.71 to 1.26)	.73
rs7627754—TT 1/0	1.19 (0.48 to 2.96)	.70
rs7853758—continuous 0/1/2 with 0 = GG	1.06 (0.77 to 1.44)	.73
rs2815063—continuous 0/1/2 with 0 = CC	1.16 (0.86 to 1.56)	.34
rs6689879—continuous 0/1/2 with 0 = TT	1.13 (0.87 to 1.45)	.35
rs2232228—categorical with GG as reference
G/G	Ref
A/G	0.97 (0.67 to 1.40)	.87
A/A	0.77 (0.44 to 1.32)	.34

Open in a new tab

NOTE. Each variable reflects a separate model adjusted for age at baseline, race, sex, and age at diagnosis.

Abbreviations: HR, hazard ratio; LVEDVi, left ventricular end-diastolic volume index; LVESVi, left ventricular end-systolic volume index; METS, metabolic equivalents of task; Ref, reference.

TABLE A6.

Multivariable Associations Comparing Individuals Within IGHG-Defined Risk Groups Based Upon Cumulative Doses of Anthracyclines and Chest Radiation Adjusting for rs2229774

Variable	HR (95% CI)	P
Race
White	Ref
Non-White	1.44 (0.93 to 2.24)	.11
Sex
Male	Ref
Female	0.81 (0.57 to 1.14)	.23
Age at diagnosis, years
0-4	Ref
5-9	1.37 (0.85 to 2.21)	.20
10-14	1.08 (0.66 to 1.76)	.77
15-19	1.07 (0.61 to 1.87)	.81
No. of cardiovascular risk factors
None	Ref
1	0.98 (0.64 to 1.49)	.92
≥2	1.30 (0.82 to 2.04)	.27
IGHG risk group
Low	Ref
Moderate, without anthracycline	0.71 (0.08 to 6.41)	.76
Moderate, with anthracycline	1.70 (0.69 to 4.19)	.25
High, without anthracycline	6.71 (2.35 to 19.16)	<.001
High, with anthracycline	3.95 (1.67 to 9.34)	.002
IGHG low-risk group
GLS normal/NT-proBNP normal	Ref
GLS abnormal/NT-proBNP normal	1.15 (0.29 to 4.48)	.84
GLS normal/NT-proBNP abnormal	0.00 (0.00 to 3.34)	.27
GLS abnormal/NT-proBNP abnormal	0.00 (0.00 to 18.6)	.62
GLS normal/NT-proBNP unavailable	2.07 (0.68 to 6.27)	.20
GLS abnormal/NT-proBNP unavailable	1.78 (0.36 to 8.77)	.48
IGHG moderate-/high-risk group, without anthracycline
GLS normal/NT-proBNP normal	Ref
GLS abnormal/NT-proBNP normal	0.72 (0.21 to 2.45)	.60
GLS normal/NT-proBNP abnormal	0.40 (0.08 to 1.91)	.25
GLS abnormal/NT-proBNP abnormal	0.88 (0.30 to 2.58)	.82
GLS normal/NT-proBNP unavailable	1.83 (0.22 to 15.14)	.57
GLS abnormal/NT-proBNP unavailable	0.00 (0.00 to 1.79)	.99
IGHG moderate-/high-risk group, with anthracycline
GLS normal/NT-proBNP normal	Ref
GLS abnormal/NT-proBNP normal	0.98 (0.50 to 1.92)	.96
GLS normal/NT-proBNP abnormal	2.15 (1.15 to 4.01)	.016
GLS abnormal/NT-proBNP abnormal	4.29 (2.33 to 7.89)	<.001
GLS normal/NT-proBNP unavailable	2.66 (1.34 to 5.29)	.005
GLS abnormal/NT-proBNP unavailable	4.24 (2.09 to 8.60)	<.001

Open in a new tab

NOTE. Adjusted for age at baseline and rs2229774.

Abbreviations: GLS, global longitudinal strain; HR, hazard ratio; IGHG, International Guideline Harmonization Group; NT-proBNP, N-terminal-pro-B-type natriuretic peptide; Ref, reference.

Matthew J. Ehrhardt

Honoraria: Optum

Daniel A. Mulrooney

This author is a member of the Journal of Clinical Oncology Editorial Board. Journal policy recused the author from having any role in the peer review of this manuscript.

Isaac B. Rhea

Research Funding: Pfizer (Inst)

Stephanie B. Dixon

Employment: Pacemate

Travel, Accommodations, Expenses: Pacemate

Open Payments Link: https://openpaymentsdata.cms.gov/physician/2791772

Kirsten K. Ness

Consulting or Advisory Role: City of Hope

Deo Kumar Srivastava

Consulting or Advisory Role: General Dynamics Information Technology

Melissa M. Hudson

Employment: Consolidated Medical Practices of Memphis

Consulting or Advisory Role: Oncology Research Information Exchange Network, Princess Máxima Center, VIVA Foundation Singapore

No other potential conflicts of interest were reported.

PRIOR PRESENTATION

Presented in part at the ASCO Annual Meeting, Chicago, IL, June 3-8, 2022, and at the International Symposium on Late Complications After Childhood Cancer, Utrecht, the Netherlands, July 8-9, 2022.

SUPPORT

Supported by a Cancer Center Support (CORE) grant (P30 CA21765) to St Jude Children's Research Hospital, U01 CA195547 (Multiple Principal Investigators: M.M.H. and K.K.N.), R01 CA216354 (Multiple Principal Investigators: Y.Y. and J.Z.), and the American Lebanese Syrian Associated Charities, Memphis, TN.

AUTHOR CONTRIBUTIONS

Conception and design: Matthew J. Ehrhardt, Deo Kumar Srivastava, Melissa M. Hudson, Yutaka Yasui, Gregory T. Armstrong

Financial support: Melissa M. Hudson, Gregory T. Armstrong

Administrative support: Aaron McDonald, Leslie L. Robison, Melissa M. Hudson, Gregory T. Armstrong

Provision of study materials or patients: Kirsten K. Ness, Leslie L. Robison, Melissa M. Hudson, Gregory T. Armstrong

Collection and assembly of data: Matthew J. Ehrhardt, Daniel A. Mulrooney, Kyla Shelton, Kirsten K. Ness, Leslie L. Robison, Melissa M. Hudson, Gregory T. Armstrong

Data analysis and interpretation: Matthew J. Ehrhardt, Qi Liu, Daniel A. Mulrooney, Isaac B. Rhea, Stephanie B. Dixon, John T. Lucas, Yadav Sapkota, Deo Kumar Srivastava, Aaron McDonald, Melissa M. Hudson, Yutaka Yasui, Gregory T. Armstrong

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST