Cardiac Function After Cardiotoxic Treatments for Childhood Cancer—Left Ventricular Longitudinal Strain in Screening

Background: The majority of childhood cancer survivors (CCSs) have been exposed to cardiotoxic treatments and often present with modifiable cardiovascular risk factors. Our aim was to evaluate the value of left ventricular (LV) longitudinal strain for increasing the sensitivity of cardiac dysfunction detection among CCSs. Methods: We combined two national cohorts: neuroblastoma and other childhood cancer survivors treated with anthracyclines. The final data consisted of 90 long-term CCSs exposed to anthracyclines and/or high-dose chemotherapy with autologous stem cell rescue and followed up for > 5 years and their controls (n = 86). LV longitudinal strain was assessed with speckle tracking (Qlab) and LV ejection fraction (EF) by three-dimensional echocardiography (3DE). Results: Of the CCSs, 11% (10/90) had abnormal LV longitudinal strain (i.e., < -17.5%); of those, 70% (7/10) had normal 3DE LV EF. Multivariable linear model analysis demonstrated that follow-up time (p = 0.027), sex (p = 0.020), and BMI (p = 0.002) were significantly associated with LV longitudinal strain. Conversely, cardiac risk group, hypertension, age, cumulative anthracycline dose or exposure to chest radiation were not. Conclusion: LV longitudinal strain is a more sensitive method than LV EF for the detection of cardiac dysfunction among CCSs. Therefore, LV longitudinal strain should be added to the screening panel, especially for those with modifiable cardiovascular risk factors.


INTRODUCTION
The number of childhood cancer survivors (CCSs) reaching adulthood is increasing rapidly (1). In contrast, the doses of anthracyclines and chest irradiation have decreased in modern treatment protocols due their dose-dependent cardiotoxicity (2,3). Some novel drugs, mainly used in adult oncology, also have cardiovascular side effects (4). Although cardiovascular damage coincides with exposure to the toxic treatment, clinical heart failure may not manifest until decades later. Modifiable cardiovascular risk factors, e.g., obesity, hypertension and dyslipidemia (5), are also common among CCSs and potentiate therapy-related adverse events (6,7). As a result, a frequency of clinical heart failure ranging between 0 and 16% among CCSs has been reported post-treatment (8), and subclinical toxicity is even more prevalent (9). Thus, meticulous, long-term followup is of key importance for identifying patients at risk and for offering adequate treatment (3,10). Consequently, a consensus recommendation on cardiomyopathy surveillance of CCSs has recently been put forth (1). However, the most commonly used echocardiographic methods, including ejection fraction (EF) and fractional shortening for the evaluation of systolic function, have some limitations with regard to sensitivity and accuracy (9,11).
Speckle tracking-based LV longitudinal strain is a sensitive method for detecting decreased systolic function, even when the LV EF is still within normal limits (12). The subendocardial longitudinal fibers are prone to subtle injury because of their location; however, the LV EF comprises of radial, circumferential and longitudinal functions and thus deteriorates later than does longitudinal strain (13,14). Left atrial (LA) strain is a promising tool for the detection of LV diastolic dysfunction (15). Guidelines recommend the use of global longitudinal LV strain together with LV EF for screening among adults during and after cancer treatment (4,16). However, the inclusion of global longitudinal LV strain measurements has not yet been applied to children or adolescent protocols. Indeed, the best echocardiographic methods for the detection of asymptomatic cardiotoxicity remain to be determined.
The purpose of this study was to evaluate the additive value of strain imaging with regard to the sensitivity of cardiac dysfunction detection after childhood cancer.

Participants
For this study, we combined data from two previously published cohorts of CCSs, firstly childhood cancer patients treated with anthracyclines, and secondly, a national cohort of neuroblastoma patients (Figure 1) (17,18). The study group consisted of 95 long-term CCSs treated between 1980 and 2006 at the five university hospitals in Finland. The study patients were exposed to cardiotoxic treatments, e.g., anthracyclines and/or high-dose therapy with autologous stem cell rescue, and followed up for >5 years. None of the patients were given dexrazoxane during the survey. Six patients (6%) had heart failure treatment at the time of evaluation: 4 were treated with enalapril and two with enalapril combined with a beta-blocker. Stratification for long-term cardiac risk (i.e., cardiac risk groups) was performed according to the International Late Effects of Childhood Cancer Guideline Harmonization Group based on cumulative anthracycline and chest radiation doses (1). Healthy, age-and sex-matched controls were included for both study groups. The details of the clinical characteristics are shown in Table 1.
The Research Ethics Committees of Helsinki and Tampere University Hospitals approved the study carried out in accordance with the Declaration of Helsinki. All the study subjects and their legal guardian(s) provided a written, informed consent.

Blood Pressure
Right arm blood pressure was measured at rest according to guidelines. Stage one hypertension was defined as blood pressure ≥95th percentile or ≥130/80 mmHg (whichever was lower) for children under 13 years and ≥130/80 mmHg for those ≥13 years (19).
Age-dependent reference values of the LV transmitral to septal mitral annular early diastolic velocity ratio (E/E ′ ) and peak septal myocardial systolic velocity (S ′ ) were used for subjects up to 18 years of age (22). LV E/E ′ >10.0 and S ′ <6.1 were considered abnormal for those over 19 years of age (23).

Three-Dimensional Echocardiography
Analyses of three-dimensional echocardiographs were performed using commercial software (Qlab v9, Philips Medical Systems, Andover, MA, USA), as previously described (24). Briefly, the three-dimensional echocardiography (3DE) LV EF was calculated, and end-diastolic, end-systolic and stroke volumes were indexed to the body surface area (BSA). The LV end-systolic mass (LVMS) was normalized to height (meters) to a power of 2.7 (LVMSi), and the result expressed as g/m 2.7 .

Qlab Peak Systolic Strain Analysis
LV longitudinal peak systolic strain from the four-chamber view was analyzed with Qlab (Philips Qlab, version 10.5, CMQ; Philips Healthcare, Bothell, WA, USA) and designated LV longitudinal strain. In the analysis, the points were placed at the edge of the mitral valve annulus on the septal and lateral sides and apex at end-diastole. The frame chosen by the program also automatically performed the rest of the tracing, which was manually checked and corrected if needed; the case was excluded if correction was not feasible. All systolic strain measurements were analyzed by one investigator blinded to all clinical and FIGURE 1 | Study recruitment and combination flow chart. The number of study patients and their cancer therapy. The majority of the patients, 94% (85/90), were treated with anthracyclines and/or chest radiotherapy. The rest (5/90, 6%) were not exposed but had received high-dose therapy and stem cell rescue. Altogether, 28% (25/90) of the patients had received high-dose therapy and stem cell rescue, and 80% (20/25) of them were also exposed to anthracyclines and/or chest radiotherapy. outcome data. Intraobserver analysis, performed using the Bland-Altman analysis (n = 10), showed mean of difference of 0.12 with limits of agreement between −1.08 and 1.32.

Analysis of Diastolic Function and Atrial Parameters
Velocity vector imaging (VVI) of the left atrium (LA) was used to analyze LA indices and diastolic cardiac function. VVI analysis from the four-chamber view was performed with the VVI analysis program (Syngo USWP 3.0, Siemens Healthineers, Erlangen, Germany) as described (25). In short, manual tracing of the LA was performed using a single still frame in end-systole. Endocardial tracing began at the edge of the mitral valve annulus, extended to the base of the atrium and returned to the other edge of the annulus. The VVI algorithm calculated the velocity vectors for each frame of the cardiac cycle, displaying them for the complete loop. If the endocardial border was not traceable throughout the whole cardiac cycle, it was corrected manually. The parameters calculated were the LA area, LA fractional area change and VVI peak longitudinal strain for LA. The LA area was normalized to BSA, and the result expressed as cm 2 /m 2 .

Statistical Methods
IBM SPSS Statistics version 24 (IBM Corp., Armonk, NY, USA) was used in this study. Categorical data are presented as frequencies and percentages, normally distributed continuous variables as the mean ± SD, and as median and interquartile ranges (IQR) in cases of non-normality. Categorical variables were compared with the chi-square or Fisher's exact-test. Means between two groups were compared using the independent samples t-test and medians with independent-samples using the Mann-Whitney U-test. Means were compared using one-way analysis of variance (ANOVA), and further pairwise comparisons between groups were performed with the Bonferroni method. When the homogeneity of variance was not met, Welch's ANOVA with Tamhane's-test was used for pairwise comparisons. Univariate associations of continuous variables associated with strain were analyzed with linear regression. A multivariable linear model was used to examine associations of multiple variables with strain. The cut-off value for the follow-up time to detect pathological strain was determined by a receiver operating characteristic (ROC) curve. The optimal cut-off value for followup time was chosen by using the Youden Index. A p-value < 0.05 was considered significant.  SCT, stem cell transplantation; TBI, total body irradiation.

RESULTS
A total of 90 CCSs and 86 controls were included (Figure 1). Four survivors (4%) and 9 controls (9%) were excluded due to non-analyzable four-chamber views. The demographics of the subjects are listed in  (Figure 1). The normal strain value used for the Qlab was >-17.5% corresponding with the lower normal limits (mean -1.96 SD) of our controls. The survivors had higher systolic and diastolic blood pressure than the controls, but BSA and body mass index (BMI) for children according to age (26) did not differ between the two groups ( Table 1).
There was more diastolic dysfunction (an abnormally high LV E/E ′ ) in S1 (60%) than in S2 (20%, p = 0.012) and controls (5%, p < 0.001). The LA area index did not differ between the groups. The LA fractional area change was lower in S1 than in S2 (48.6 ± 7.7 vs. 55.9 ± 9.2%, p = 0.041). Despite a trend for lower LA longitudinal strain in S1 than in S2 and C3 (34.5 ± 12.5 vs. 45.5 ± 14.4 and 44.4 ± 13.7%), the difference was not statistically significant (p = 0.066).   β indicates regression coefficient for one-unit increase in continuous variables. a Doxorubicin isotoxic equivalents. b The means between two groups were compared with the independent samples t-test and between cardiac risk groups with the one-way ANOVA.
Linear regression was used to evaluate the association of continuous variable with strain. c Multivariable linear model.

Variables Associated With LV Longitudinal Strain
Variables associated with LV longitudinal strain were also evaluated ( Table 3). According to the univariate analysis, a long follow-up time (p = 0.029), female sex (p = 0.031), high BMI (p = 0.036) and chest radiotherapy (p = 0.004) were significantly associated with low strain. Furthermore, followup time (p = 0.027), female sex (p = 0.020) and BMI (p = 0.002) remained significant variables in the multivariable linear model, whereas cardiac risk group, high blood pressure, age, cumulative anthracycline dose or exposure to chest radiation did not ( Table 3).
The area under the curve (AUC) for the follow-up time to detect pathological strain (i.e., ≤-17.5%) was 0.76 (95% CI 0.63-0.89). The critical cut-off value for the follow-up time to increase the risk of abnormal strain was ≥ 9.24 years (sensitivity 0.90 and specificity 0.60).

DISCUSSION
Our study demonstrates that among CCSs with abnormal LV longitudinal strain, 70% had normal LV EF, suggesting that decreased LV longitudinal strain may emerge as a more sensitive marker of cardiotoxicity than LV EF following treatment for childhood cancer.
Our results are in line with recent data on adults showing abnormal LV longitudinal strain in 28% of CCSs despite normal LV EF (27). For adults, speckle tracking-based strain imaging has been well-validated for the measurement of LV deformation and recommended for echocardiographic, functional follow-up (1,4,16). However, the role of myocardial strain imaging among children is less well-established. In agreement with our results, several studies have found that the global systolic function and LV longitudinal strain are reduced among CCSs (2,(28)(29)(30)(31), though these factors have yet to be proven to be more sensitive than LV EF in the early follow-up.
In this study, we report a decrease in strain in a subgroup of patients with preserved EF. This finding could not be explained by the patient age as a confounding variable, as it did not show a significant correlation with reduced longitudinal strain, either in univariate or multivariable analysis. In addition, although some level of decrease in the strain values has recently been shown to occur due to aging, in a recent study (32), any significant decrease in LV global longitudinal strain did not occur until the 8th decade of life and thus did not potentially affect the patients of our young study groups. This underlines the fact that the observed reduction in LV longitudinal strain among the CCSs is not based on increased age alone but instead on accumulated follow-up time after the original cancer treatment. Similarly, and importantly, male sex is known to be associated with lower longitudinal, circumferential, and radial strain (33), whereas in our study, the S1 patients with reduced strain with preserved EF were predominantly females. This further reduces the possibility of sex explaining the observed phenomenon of strain reduction in this population. Instead, females appear to be at higher risk of deleterious cardiotoxic effects of cytotoxic drugs.
Anthracycline toxicity and secondary cardiovascular risk factors mainly affect the subendocardial fibers (13,14,32), contributing negatively to longitudinal shortening prior to the reduction in LV EF. For example, patients with heart failure and preserved EF compensate for the reduction in longitudinal shortening by increasing twist to maintain normal EF (34). Thus, strain imaging has been proposed as a more sensitive technique to detect myocardial damage than LV EF. Early detection of cardiac failure development may be valuable for treatment with early interventions being considered more efficient than measures taken after abnormal EF is detected or clinical symptoms of dysfunction manifest (3,16,31,35).
The value of LV longitudinal strain in detecting early myocardial dysfunction in cohort studies depends on how many events are assumed to occur simultaneously in the general population. The proportion of abnormal LV functional findings appears to depend on the length of follow-up (2,6,27,31). In our cohort (median follow-up time 8.1; IQR 6.0-13.3 years), the proportion of survivors with preserved LV EF but with abnormal strain was 9%, lower than that reported by two adult studies [28% among those with normal LV EF after a mean follow-up time of 21.6 ± 7.9 years (27) or a median time from diagnosis of 23 years (range 10-48 years) (2)].
However, our results are comparable, albeit clearer, than those reported by others. For example, Slieker et al. (31) recently found reduced longitudinal strain in 7.7% of their 546 CCSs [median time since last anthracycline treatment, 7.9 (IQR, 5.6-10.6) years]. Similarly, three other studies reported significant decreases in the longitudinal function but did not report the prevalence of abnormal strain (median follow-up time, 5.2-13.2 years) (30,34,36).
The risk of cardiac dysfunction has been shown to increase with time (3), in line with our results. Similarly, the recent differing report (31) on the role of LV longitudinal strain in the follow-up of CCSs seems to be impacted by a shorter follow-up time than in our study demonstrating cardiac dysfunction more likely to be detectable by pathological strain when the follow-up time exceeds 9 years (sensitivity 0.90, specificity 0.60). Our results thus indicate that the longitudinal strain putatively offers a potent tool for risk assessment among the CCSs, especially early on and beyond the first decade of follow-up.
Our data are in line with those of Christiansen et al. (27) showing exposure to chest radiotherapy to be more common among the CCSs with abnormal LV longitudinal strain than others. However, in our study, the anthracycline dose did not correlate with the reduced longitudinal strain, again in line with the data of Slieker et al. (31), most likely indicating the absence of a safe dose of anthracyclines among the CCSs.
Both of our CCS groups (S1 with abnormal strain and S2 with normal strain) had a lower LV EF than controls (C3). A similar trend was also observed for the TDI S ′ , an additional sensitive marker of systolic function. The most common and best-established form of anthracycline cardiomyopathy indeed resembles dilated cardiomyopathy, with a thin-walled, large LV and low EF (3).
Our data further show and support those of others that the CCSs are at increased risk of modifiable cardiovascular risk factors such as hypertension and obesity (7). Hypertension induces LV hypertrophy and thus increases LV EF but possibly decreases stroke volume. Consequently, as an early marker of systolic dysfunction, the CCSs may have decreased LV longitudinal function with preserved EF. In our study, those with abnormal LV longitudinal strain (S1) had a higher LV systolic mass index and more hypertension than controls, illustrating this phenomenon. A large study on adult CCSs has shown that survivors with metabolic syndrome are twice as likely to have abnormal global longitudinal strain. Each individual component of the metabolic syndrome increases the risk, but without a higher risk for abnormal LV EF (2). Therefore, cancer treatment-related cardiomyopathy should no longer be solely defined as dilated but rather as mixed and further associated with an increased burden of treatment-related, modifiable risk factors, including hypertension-related LV hypertrophy with abnormal longitudinal strain, as early markers of LV dysfunction with preserved EF. Indeed, cardiomyopathy risk groups based on the anthracycline and chest radiation doses consist of high-risk survivors at an early stage of follow-up (1), but with further follow-up, the risks become less well-delineated due to the impact of modifiable cardiovascular risk factors. The inclusion of LV longitudinal strain in the screening armamentarium might thus improve detection for survivors in the low and moderate cardiomyopathy risk groups but with increased risk of LV dysfunction.
Diastolic LV dysfunction often precedes the systolic. In our study, LV E/E ′ as a marker of diastolic dysfunction was higher among survivors than controls and peaked in group S1. However, there was no difference in the LA area index between the groups, most likely reflecting the fact that atrial restriction prevents enlargement despite diastolic dysfunction. In addition, we observed a trend toward lower LA longitudinal strain in our group S1 compared with the others, even though the difference was not statistically significant. A lower LA fractional area change in S1 compared with S2 demonstrated the same. These results are in line with the study of Morris et al. (15), showing LA strain to be more sensitive than the volume index for detecting LV diastolic dysfunction among adult patients at risk. Nevertheless, LA strain is mostly a research tool, with analysis of the thin atrial wall sometimes being technically challenging.
An earlier study comparing the different systems showed small but statistically significant intervendor variation in the assessment of LV longitudinal strain (37). Thus, ideally, an ultrasound device from the same vendor should be used during the follow-up whenever possible. We studied LV longitudinal strain using the more readily employable Qlab system. Importantly, the lower normal limit for Qlab LV longitudinal strain for our controls was in line with the −2 SD value (i.e., −17.5%) derived from recently published pediatric reference values (38).

Limitations
The most important limitation of this study was its crosssectional nature, rendering future prospective studies important to confirm an increase in the risk of pathological longitudinal strain during and after the first decade of follow-up. Because we used previously acquired echocardiographs from our earlier studies, LV longitudinal strain from the four-chamber views was employed, as opposed to global longitudinal strain. Nonetheless, optimal four-chamber views are easy to obtain, and in addition, four-chamber longitudinal strain has good intra-and interobserver correlation (38) and thus is usually sufficient for daily use in practice. Our division of patients into subgroups may also be considered somewhat arbitrary. Yet, the limits chosen (i.e., EF > 55% and strain > −17.5%) to define clinical normal values were adapted from those, already-published and generally accepted publications, and remained well in line with our own defined −1,96 SD lower limit of values from the age-and sexmatched control population (20,33,38). Important to note, none of the patients were given the cardioprotectant dexrazoxane during the survey, eliminating its possible effect on the measured strain rate among the CCS.

CONCLUSIONS
Healthcare providers should pay special attention to the modifiable cardiovascular risk factors among the CCSs, as they play a pivotal role in developing heart failure long-term.
To date, the CCSs at risk for developing cardiac problems may not be identified early enough when using LV EF alone. Indeed, our results indicate that longitudinal strain putatively offers a potent tool for the long-term risk assessment among CCSs beyond the first decade of followup. Especially for those with modifiable cardiovascular risk factors and LV hypertrophy with normal EF, LV longitudinal strain would beneficially contribute to the final decisionmaking already in the pediatric population and among young adults.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Helsinki University Hospital, Helsinki, Finland; Tampere University Hospital, Tampere, Finland. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.

AUTHOR CONTRIBUTIONS
JN, KY, and TO designed the study, collected and analyzed the data, and wrote the manuscript. KV and TP participated in designing the study, analyzing the data, and writing the manuscript. AS, KP, SM, TS, KJ, and AE participated in analyzing the data and writing the manuscript. All authors contributed to the article and approved the submitted version.