Echocardiographic parameters of left ventricular size and function as predictors of symptomatic heart failure in patients with a left ventricular ejection fraction of 50–59% treated with anthracyclines

Abstract

Aims

The aim of this study was to assess whether baseline echocardiographic measures of left ventricular (LV) size and function predict the development of symptomatic heart failure or cardiac death (major adverse cardiac events, MACE) in patients treated with anthracyclines who have a pre-chemotherapy left ventricular ejection fraction (LVEF) between 50 and 59%.

Methods and results

Patients with an LVEF between 50 and 59% before anthracyclines were selected. In these patients, LV volumes, LVEF, and peak longitudinal strain (GLS) were measured. Individuals were followed for MACE and all-cause mortality over a median of 659 days (range: 3–3704 days). Of 2234 patients undergoing echocardiography for pre-anthracycline assessment, 158 (7%) had a resting ejection fraction of 50–59%. Their average LV end-diastolic volume (LVEDV) was 101 ± 22 mL, LVEF was 54 ± 3%, and global longitudinal strain (GLS) was −17.7 ± 2.6%. Twelve patients experienced a MACE (congestive heart failure) at a median of 173 days (range: 15–530). Age, diabetes, previous coronary artery disease, LVEDV, indexed LVEDV, LVESV, indexed LVESV, and GLS were all predictive of MACE (P = 0.012, 0.039, 0.0029, 0.012, and 0.0065 for LVEDV, LVEDVI, LVESV, LVESVI, and GLS, respectively). Indexed LVEDV and GLS remained predictive of MACE when adjusted for age. Age and GLS were also predictive of overall mortality (P < 0.0001 and 0.0105, respectively).

Conclusion

In patients treated with anthracyclines with an LVEF of 50–59%, both baseline EDV and GLS predict the occurrence of MACE. These parameters may help target patients who could benefit from closer cardiac surveillance and earlier initiation of cardioprotective medical therapy.

Introduction

Earlier cancer diagnosis and the advent of newer classes of cancer treatments have led to an increase in the survival length and thus of the number of survivors in cancer patients. As of 1 January 2022, it is estimated that the population of cancer survivors will increase to nearly 18 million in the USA.¹ Increased survival duration also leads to a larger number of older cancer survivors with more cardiovascular risk factors and more vulnerability to develop cardiovascular side effects of cancer drugs.

Anthracyclines are an integral component of many chemotherapeutic regimens; however, their benefit is limited, in part, by their acute and chronic cardiotoxicity.^2–4 Chronic anthracycline toxicity may present initially as asymptomatic left ventricular (LV) dysfunction and ultimately, symptomatic heart failure (HF) which can occur even decades after the discontinuation of the treatment.⁴

Available data suggest that almost 1/3 (27%) of the adult patients exposed to anthracyclines will show some measure of cardiac dysfunction ∼5 years post treatment, with 2–5% developing overt HF.^5–7 When HF is diagnosed, patients with anthracycline-induced cardiomyopathy have a survival rate of <50% at 1 year.^8,9

Although patients treated with anthracyclines who have an asymptomatic decrease in left ventricular ejection fraction (LVEF) can respond to HF treatment, 45% are non-responders.¹⁰ The number of non-responders increases with the time elapsed between the chemotherapy and the diagnosis of decreased LVEF.¹⁰ Hence, the early detection of anthracycline-induced cardiomyopathy is crucial, as it allows early introduction of appropriate medical therapy.

LVEF is a widely used parameter to predict the occurrence of HF in patients treated with anthracyclines. Despite its common use, there is evidence that even if the decrease in LVEF is diagnosed early after anthracycline treatment, more than a third of the patients do not recover after cardiovascular therapy.¹⁰ This is not unexpected as LVEF does not reliably detect histologically confirmed cardiomyocyte damage.¹¹ Moreover, in a broad spectrum of patients with symptomatic HF, LVEF has been shown to have limited prognostic value when >45%.¹²

Myocardial deformation (strain) and strain rate are more sensitive than LVEF for early detection of subtle cardiac dysfunction in cardiac myocardial pathologies, including in patients treated with anthracyclines.^13,14 The early decreases in peak longitudinal strain and strain rate observed when LVEF is still preserved have recently been demonstrated to predict subsequent reductions in LVEF in women with breast cancer treated with anthracyclines, taxanes, and trastuzumab.^15,16 The significance of these findings regarding the occurrence of clinical events, however, remains to be proved.

Early diagnosis of anthracycline-induced myocardial injury could be especially useful in patients with normal to low-normal ejection fraction (EF 50–59%). In this range, LVEF does not appear to provide any prognostic value in patients with HF.¹² In patients treated with anthracyclines, however, the LVEF range of 50–59% at baseline appears to be associated with higher risk of anthracycline-associated congestive HF and cardiac death than higher LVEF values (EF > 60%).^17,18 Nevertheless, it is currently estimated that because of its variability, echocardiography can only detect with certainty LVEF changes of >6–10%.^19,20 Thus, given this degree of variability, an additional predictor of risk in the low normal/mildly impaired range of LVEFs would be helpful to further stratify patients for closer monitoring.

The purpose of the present study was to evaluate the value of the echocardiographic measures of LV size and function in predicting the development of symptomatic HF and cardiac death in patients with an LVEF between 50 and 59% prior to anthracycline treatment.

Methods

Study subjects

Consecutive patients treated with anthracyclines between 2002 and 2012 at the Massachusetts General Hospital and with a baseline LVEF between 50 and 59% by 2D echocardiography were studied. The Institutional Review Board of Massachusetts General Hospital approved the study protocol.

Echocardiography and strain analysis

Transthoracic echocardiography was performed using commercially available equipment (Vivid 7 or E9, GE Medical Systems, Milwaukee, WI, USA; or iE33, Philips Medical Systems, Andover, MA, USA). Cine loops from three standard apical views (apical four-chamber, A4C, apical two-chamber, A2C, and apical three-chamber, A3C) were recorded using grey-scale harmonic imaging and saved in compressed DICOM format.

LV volumes and LVEF were measured by a blinded observer (N.M.) using Simpson biplane method in apical four- and two-chamber views as recommended by the American Society of Echocardiography.²¹ LV volumes were analysed as absolute values and were also indexed for body surface area (LVEDVI, LVESVI).

Longitudinal strain was quantified in an 18-segment model using an offline previously validated analysis program [2D Cardiac Performance Analysis (2D CPA), TomTec Imaging System, Munich, Germany].^22,23 2D CPA is a speckle tracking-based analysis software that provides measures of the endocardial strain and strain rate on DICOM loops obtained from ultrasound machines. One cardiac cycle was selected (starting at the R wave) and the endocardial borders were traced in the end-systolic frame of the 2D images from the three apical views (Figure 1). The software automatically measures maximum systolic strain in each segment and averages it. An observer (N.M.) blinded to the occurrence of symptomatic HF or cardiac death performed all strain measurements offline. Due to the difficulty in tracking the A3C in some patients (13% patients had uninterpretable tracking due to poor acoustic windows in the setting of prior chest wall surgery or breast implants), the primary measurement of interest was the LV peak global longitudinal systolic strain averaged from the A2C view and A4C views. The secondary measurement of interest included the LV peak global longitudinal systolic strain averaged from the A2C, A3C, and A4C views. The sphericity index, a measure of LV remodelling, was calculated as the ratio of short- to long-axis lengths averaged from A2C and A4C views.^24,25

Outcomes

The primary end point was the occurrence of major adverse cardiac events (MACE). Major adverse cardiac events were defined as subsequent occurrence of either New York Heart Association (NYHA) class III or IV congestive HF, cardiac arrest, or cardiac death. Outcomes were obtained through review of the institutional electronic medical records and were verified by a board-certified cardiologist blinded to all other clinical data. Loss of follow-up because of non-cardiac-related mortality was also verified. The duration of follow-up was determined by review of the patient's hospital chart.

Reproducibility

Intra-observer variability was assessed in 10 random subjects by one observer (N.M.) measuring at 1-week interval. Inter-observer variability was determined by two observers blinded to each other's results (N.M. and T.C.T.). The intra-observer variability of endocardial peak longitudinal strain reported as the mean error ± SD of 10 measurements was 1.0 ± 0.6% in absolute values (5 ± 4% in percentages), and the inter-observer variability was 1.0 ± 1.0% in absolute values (6 ± 6% in percentages).

Statistics

Continuous data are presented as mean ± SD or median and range, and categorical variables as percentages. Differences in continuous data between the patients with or without MACE were compared using Student's t-test or Wilcoxon rank comparison for non-parametric variables. Categorical variables were compared using the χ² test. Time to first MACE was defined as the number of days between the start of anthracycline therapy and the date of first MACE. Patients who had not experienced a MACE as of their last visit date were censored at this date. Cox proportional hazard analysis was used to determine significant clinical and echocardiographic predictors of MACE and overall mortality. In a multivariable analysis, the echocardiographic predictors of MACE were adjusted for age (the most significant clinical predictor of MACE). Survival without MACE as a function of LVEDVI or GLS was expressed using Kaplan–Meier analysis. The incremental value of measures of LV function over baseline clinical variables was assessed in a step-wise model. All analyses were performed using a standard statistical software program (SAS statistical package, SAS Institute Inc., Cary, NC, USA). A P-value threshold of <0.05 was considered statistically significant.

Results

Patient characteristics and follow-up

Out of 2234 patients with an available echocardiogram, 2208 had normal LVEF, including 158 (7%) with an LVEF between 50 and 59%. Compared with patients with an LVEF ≥ 60%, patients with an LVEF between 50 and 59% were slightly younger (P = 0.0048 vs. patients with EF > 60%), more predominantly male (P = 0.0003 vs. patients with EF > 60%), with more haematological malignancies (P = 0.0387 vs. patients with EF > 60%), and more MACE (8 vs. 2%, P < 0.0001, Table 1). The patients were followed over a median of 659 days (range: 3–3704 days).

Major adverse cardiac events

Twelve patients (8%) developed MACE. The comparison of clinical characteristics of patients with MACE to the patients without MACE is illustrated in Table 2. Major adverse cardiac events were detected at a median of 173 days (range: 15–530 days). All 12 patients developed symptomatic HF; no cardiac-related death was noted.

Overall mortality

There were a total of 60 deaths (38%) among all patients at a median follow-up of 273 days (range: 3–3704 days). The overall mortality was not different in patients who did or did not develop MACE (hazard ratio of 1.97, 95% CI: 0.91–3.82, P = 0.08). The length of follow-up was higher in surviving patients [1191 days (13–3507 days) compared with 270 days (3–3704 days), P < 0.0001].

Echocardiographic characteristics of the patients

The echocardiographic characteristics of the patients are presented in Table 3. The mean LVEF was 54 ± 3%, and the average (two and four chambers) peak longitudinal strain was −17.7 ± 2.6%. GLS measured on the two- and four-chamber views as well as on the three apical views were predictive of MACE (P = 0.0065 and P = 0.035, respectively).

There were no significant differences in LVEF or LV spherical index at baseline between patients who did or did not develop MACE. In contrast, the LV volumes were significantly higher (LVEDV 118 ± 31 mL vs. 99 ± 21 mL, P = 0.005, LVESV: 56 ± 16 mL vs. 45 ± 10 mL, P = 0.0010) and the peak global longitudinal strain was lower in the group of patients who developed MACE (−16.0 ± 2.5% vs. −17.7 ± 2.6%, P = 0.015). The indexed end-systolic and end-diastolic volumes were also higher in the group with MACE (P = 0.02 and 0.004, respectively).

Predictors of MACE

The predictors of the rate of occurrence of MACE are detailed in Table 4. Age was the strongest clinical predictor of the rate of occurrence of MACE. Neither the baseline LVEF as a continuous variable nor a baseline LVEF of <55% predicted the occurrence of MACE. The significant univariate echocardiographic predictors were LVEDV, indexed LVEDV (LVEDVI), LVESV, indexed LVESV (LVESVI), and GLS with hazard ratios of 1.03 (P = 0.012), 1.04 (P = 0.039), 1.07 (P = 0.0029), 1.11 (P = 0.012), and 1.36 (P = 0.0065) per unit change, respectively. MACE-free survival based on the highest quartile of LVEDVI (>61 mL/m²) and the most decreased quartile of GLS (<−16%) is presented in Figure 2. A GLS ≤ −16% was associated with a 4.7-fold increase in MACE (CI: 1.50–15.96). Of note, the LVEF of patients with a GLS ≤ −16% was not different from the patients with a GLS of greater than −16% (54 ± 3 vs. 54 ± 4%, P = 0.11). After adjustment for age, LVEDVI and GLS each remained predictive of MACE (P = 0.016 and P = 0.033, respectively). LVESV was not considered as it was correlated with LVEDVI (r² = 0.84, P< 0.0001).

The additional prognostic information of echocardiographic parameters of LV function in the prediction of MACE was assessed in a step-wise model with LVEF and GLS added to the clinical significant univariate predictors. Whereas LVEF did not change the prognostic value of clinical variables, the addition of GLS to the clinical variables independently increased the prognostic value of the model (Figure 3).

Follow-up strain analysis in patients who developed MACE

In 8 out of 12 patients with MACE, a follow-up echocardiogram was available (1350 ± 1237 days after the cardiac event). Average strain measured on the two- and four-chamber views was significantly reduced in the follow-up study compared with baseline (−15.4 ± 2.1 vs. −12.0 ± 2.5%, P = 0.01).

The predictors of overall mortality are detailed in Table 5. Age and cancer type predicted mortality. The mortality according to cancer type was as follows: breast cancer 25%, haematological 36%, other cancers 64% (P < 0.0035). Both LVEF and GLS were echocardiographic predictors of overall mortality. When performing a multivariable analysis including age, cancer type, LVEF, and GLS, LVEF and GLS did not provide additional prognostic information to clinical variables (P = 0.32 and P = 0.083, respectively).

Discussion

In the present study, we examined the role of echocardiographic parameters of LV size and systolic function in predicting the later occurrence of major adverse cardiac events (congestive HF and cardiac death) in patients with a baseline LVEF between 50 and 59% before anthracycline treatment. We report that LVEF does not have a prognostic value in this population, but LV volumes and GLS are independent predictors of MACE, even after adjusting for age.

The 2234 patients with an echocardiogram reflect the overall population treated with anthracyclines in terms of age and underlying cancer pathologies.^26,27 The occurrence of MACE (congestive HF as there were no cardiac death) in our study was higher (8%) than what has been reported previously in the literature (between 2 and 5%),^6,7 a finding that can be explained by the selection of patients with a low normal LVEF. There were no marked differences in cardiac risk factors and treatments noted in patients with an LVEF of 50–59% compared with patients with a higher LVEF. However, there was a higher proportion of previous coronary artery disease noted in the group with MACE, which may in turn induce subtle subclinical cardiac disease. The median timing of MACE occurred <6 months after initiation of anthracyclines, such an early time, has been reported previously,^6,28 and underlines the feasibility and potential usefulness of frequent cardiac monitoring throughout and early after the anthracycline treatment.

The previous history of CAD was a predictor of MACE. Subjects with CAD may be more susceptible to the effect of doxorubicin due to presence of possible underlying subclinical cardiac damage. Furthermore, it is also possible that the HF noted in this group of patients was the result of the underlying CAD and not necessarily doxorubicin therapy.

The new consensus document on deformation imaging details two mathematically equivalent methods, the measurement of the total length of the endocardium and the averaging of the strain in equidistant segments.²⁹ The measurement of the total length of the endocardium was not available when we analysed the patients. Therefore, we measured the average of every segment. Our method however differs slightly from the endocardial length, as we did not take into account the segments that were not well seen.

The mean strain value in our cohort was 17.7%, a slightly lower value than volunteers without evidence of cardiovascular disease (−18.6% in Ref. 30), which is expected in our cohort. The LV volumes, however, were within normal limits and similar to what has been reported in healthy individuals.²¹ LV volumes and GLS were independent predictors of MACE after adjustment for the strongest clinical predictor, age. LVEF was not predictive of MACE, an unsurprising finding as patients had a limited range of normal EF. Although LV volumes predict MACE in a variety of cardiovascular diseases, they are usually not taken into account in pre-chemotherapy patients. Two recent studies have reported the value of longitudinal strain in patients treated with chemotherapy in predicting a subsequent decrease of LVEF.^15,16 However, in these studies, the changes in LVEF were small (an average of 5% in the study by Sawaya et al.¹⁵) and most often LVEF remained within normal limits, setting into question the prognostic value of the detected changes in LVEF. The present study demonstrates the value of GLS in the prediction of major cardiac events, i.e. symptomatic HF. Patients who were in the lowest strain quartile had a five-fold increase in symptomatic HF, even though their LVEF was not different from the other patients. The findings of reduced average strain in subjects with available follow-up echocardiogram further support the value of strain analysis in follow-up of this group of patients.

Interestingly, LVEF and GLS were also predictors of overall mortality in this population. Very few patients if any in this cohort died of cardiac causes, and the mortality rate is consistent with that reported for breast cancers and lymphomas (the majority of haematological malignancies),³¹ suggesting that most patients died of cancer. Although neither variable was significant when adjusted for age and type of cancer, there was a trend for GLS to remain independent. An intriguing possibility would be that GLS (and to a lesser degree LVEF) reflects the tumour extension, as it has been shown that cancer itself may cause cardiac dysfunction.^32,33 Larger studies would be required to demonstrate this hypothesis.

There are several clinical implications to our findings. The present study demonstrates that LV volumes and GLS can non-invasively identify patients at high risk for symptomatic HF, before development of symptoms, and before any detectable impairment of LVEF. Identification of this increased risk, however, does not imply early termination of potentially life-saving anticancer therapy. Rather, baseline LVEDV and GLS measurements may help target patients who may benefit from closer cardiac surveillance and possibly initiation of potential cardioprotective medical therapy. Although cardioprotective treatments can be given to all patients,³⁴ they are associated with side effects, in particular in frail patients, and additional cost. Thus, a targeted approach of identifying and treating patients at high risk of MACE may be more effective.

There are limitations to the current study. This study incorporated subjects with ejection fraction between 50 and 59%, which made up only 7% of patients with normal EF. Although the patients with this range of LVEF represent a category of patients in whom strain might be most useful, the results cannot be generalizable to those subjects with EF above 59%. Although higher than in the overall population, the number of MACE was relatively small. Therefore, the multivariable analysis of the MACE prediction must be confirmed in larger studies. The A3C view was difficult to analyse in 13% of patients; therefore, the GLS measured on the A4C and A2C views was better able to differentiate patients at high risk of MACE than the GLS measured on all three views. It must be noted however, that both variables were predictive of MACE. Finally, although the follow-up period was as long as 10 years in some patients, the median follow-up was shorter and we cannot eliminate occurrence of very late events.

Conclusions

In conclusion, both LV volumes and GLS measured prior to the initiation of treatment with anthracyclines in patients with an LVEF of 50–59% predict the occurrence of subsequent symptomatic HF and overall mortality. Analysis of these parameters may help guide the clinician on the subsequent treatment plan in terms of monitoring of cardiac function and therapeutic adjustments.

Conflict of interest: None declared.

References