Cardiac and cardiometabolic phenotyping of trastuzumab-mediated cardiotoxicity: a secondary analysis of the MANTICORE trial

Abstract

Aims

An improved understanding of the pathophysiology of trastuzumab-mediated cardiotoxicity is required to improve outcomes of patients with human epidermal growth factor receptor 2 (HER2)-positive breast cancer. We aimed to characterize the cardiac and cardiometabolic phenotype of trastuzumab-mediated toxicity and potential interactions with cardiac pharmacotherapy.

Methods and results

This study was an analysis of serial magnetic resonance imaging (MRI) and circulating biomarker data acquired from patients with HER2-positive early-stage breast cancer participating in a randomized-controlled clinical trial for the pharmaco-prevention of trastuzumab-associated cardiotoxicity. Circulating biomarkers (B-type natriuretic peptide, troponin I, MMP-2 and -9, GDF-15, neuregulin-1, and IGF-1) and MRI of cardiac structure and function and abdominal fat distribution were acquired prior to trastuzumab, post-cycle 4 and post-cycle 17. Ninety-four participants (51 ± 8 years) completed the study with 30 on placebo, 33 on perindopril, and 31 on bisoprolol. Post-cycle 4, global longitudinal strain deteriorated from baseline in both placebo (+2.0 ± 2.7%, P = 0.002) and perindopril (+0.9 ± 2.5%, P = 0.04), but not with bisoprolol (−0.2 ± 2.1%, P = 0.55). In all groups combined, extracellular volume fraction and GDF-15 increased post-cycle 4 (+1.3 ± 4.4%, P = 0.004; +130 ± 150%, P ≤ 0.001, respectively). However, no significant change in troponin I was detected throughout trastuzumab. In all groups combined, visceral and intermuscular fat volume increased post-cycle 4 (+7 ± 17%, P = 0.02, +8 ± 23%, P = 0.02, respectively), while muscle volume and IGF-1 decreased from post-cycle 4 to 17 (−2 ± 10%, P = 0.008, -18 ± 28%, P < 0.001, respectively).

Conclusion

Trastuzumab results in impaired cardiac function and early myocardial inflammation. Trastuzumab is also associated with deleterious changes to the cardiometabolic phenotype which may contribute to the increased cardiovascular risk in this population.

Introduction

Human epidermal growth factor receptor 2 (HER2) is overexpressed in ∼20% of breast cancers. The anti-HER2 agent trastuzumab reduces risk of disease progression or death by 40%.¹ However, HER2 isoforms are also expressed in cardiomyocytes and appear to play a critical role in regulating cardiac structure and function through neuregulin-1 signalling.² As a result, trastuzumab is also associated with a dose-independent two-fold increased risk of left ventricular (LV) dysfunction and a five- to seven-fold increased risk of heart failure.^1,3 Furthermore, ∼22% of patients receiving trastuzumab will experience interruptions or discontinuation due to cardiac events.⁴ However, the pathophysiology of trastuzumab cardiotoxicity has not been well elucidated and requires further characterization to enable maintenance of cancer treatment efficacy and minimization of cardiovascular risk.

Circulating biomarkers have utility for the early detection of anthracyclines and/or trastuzumab-mediated cardiotoxicity⁵ and also have the potential to elucidate involved mechanisms. For example, high-sensitivity troponin, a marker of myocardial injury, is elevated following anthracycline chemotherapy for breast cancer,^5–8 but its response to trastuzumab is less characterized. Similarly, biomarkers related to cardiac energetics, metabolism, and extracellular matrix could also provide mechanistic insight into trastuzumab-mediated cardiotoxicity. Cardiac magnetic resonance imaging (MRI) is the gold standard imaging modality for cardiac structure and function and can also assess myocardial fibrosis and/or inflammation through T₁ mapping. Magnetic resonance imaging can also be used to quantify abdominal fat and muscle distribution changes. On computed tomography, these metrics predict chemotherapy tolerability and incident cardiovascular disease in patients with breast cancer.^9,10 However, the longitudinal effects of chemotherapy and trastuzumab on abdominal composition are unknown and could plausibly play an important role in cardiometabolic risk. To date, no studies have comprehensively evaluated serial changes in cardiac structure and function as well as body composition during trastuzumab therapy.

We previously conducted a randomized-controlled trial, MANTICORE 101-Breast, of bisoprolol, a beta-blocker, or perindopril, an angiotensin-converting enzyme inhibitor (ACEi), for the prevention of trastuzumab-mediated cardiotoxicity and found that these agents attenuated declines in cardiac MRI-derived LV ejection fraction (EF) compared with placebo.¹¹ In this secondary analysis of the MANTICORE trial, we aim to further characterize the longitudinal cardiac and cardiometabolic effects of trastuzumab via imaging and circulating biomarkers and the impact of pharmaco-intervention.

Methods

Design and ethics

The study was a double-blind, three-arm, parallel group, randomized, placebo-controlled trial (clinicaltrials.gov: NCT01016886). Participants were randomized 1:1:1 to bisoprolol, perindopril, or placebo during trastuzumab therapy (∼12 months). This trial complied with the Declaration of Helsinki and obtained approval from institutional health research ethics boards. All participants provided written informed consent. The protocol has been previously reported.¹²

Participants

Adults (>18 years) diagnosed with HER2-positive early-stage (I-IIIA) breast cancer and scheduled to receive adjuvant trastuzumab were recruited from two tertiary cancer hospitals in Edmonton and Winnipeg, Canada. Exclusion criteria included baseline LVEF <50%, history of heart failure, myocardial infarction or cardiomyopathy, uncontrolled hypertension, contraindications to study drugs or MRI, concurrent treatment with ACEi, beta-blockers, or angiotensin receptor antagonist, prior thoracic radiation or chemotherapy, or estimated glomerular filtration rate <30 mL/min/1.73 m².

Outcomes

Magnetic resonance imaging

Comprehensive cardiac MRI scans were performed on 1.5 T system (Siemens Healthcare, Erlangen, Germany) at three time points: (i) prior to trastuzumab therapy (baseline); (ii) after four trastuzumab treatment cycles (∼3 months from baseline); and (iii) at completion of trastuzumab therapy (typically 17 cycles in total, ∼12 months from baseline). Steady-state free precession cine imaging was acquired at end-inspiration with retrospective electrocardiographic (ECG) gating and reconstruction to 30 phases. Left ventricular, right ventricular (RV), and left atrium (LA) volume analysis was performed by an experienced cardiac MRI interpreter blinded to treatment assignment on commercial software (Syngo Argus; Siemens Healthcare, Erlangen, Germany). Left ventricular volumes and mass and RV volumes were calculated from a contiguous short-axis stack of cines using a method of disks approach. Left atrium volumes were estimated using a modified area-length biplane method. Volumes and masses at all time points were indexed to the baseline body surface area. LV global longitudinal strain (GLS) was analysed as the average endocardial longitudinal strain of the two-, three-, and four-chamber long-axis views using custom in-house software (MATLAB, The Mathworks, Natick, MA, USA). T₁ mapping was acquired from a basal short-axis slice using saturation-recovery single-shot acquisition pulse sequence prior to and 20 min following intravenous administration of gadopentetate dimeglumine (0.15 mmol/kg; Magnevist, Bayer Schering, Germany); septal extracellular volume (ECV) was calculated using measured haematocrit.¹³

In the same scan, a 20-slice, single-shot, fast-spin-echo acquisition was used to generate high fat–water contrast images of the abdomen for segmentation of fat and muscle. Images were obtained prior to the cardiac scan using a body matrix coil and the relevant spine array coils. An axial stack with 6 mm slice thickness (no gaps) and 1.04 mm² in-plane resolution was acquired with a scan time of ≤5 min. Five slices centred around the third lumbar vertebral body were analysed. The sum of the volumes of muscle, intermuscular, visceral, and subcutaneous fat for the five slices was measured using custom semi-automated software to identify the boundaries of subcutaneous fat, muscle, and visceral fat. Bones and organs were removed from the visceral fat pool. Interrater analysis reliability for these measurements in 10 random subjects was excellent, with intraclass correlation coefficients: subcutaneous fat, visceral fat, and muscle = 0.99 each; intermuscular fat = 0.77. Muscle and fat volumes from a cross-section at the third lumbar vertebra have been shown to be accurate relative to cadaver measurements and representative of whole-body composition.^14,15

Circulating biomarkers

Venipunctures were performed prior to each cardiac MRI scan. Blood samples were centrifuged for 15 min at 1000 g within 30 min of collection and cryostored at −80°C for future batch analysis. We analysed circulating biomarkers associated with myocardial injury [high-sensitivity troponin I, B-type natriuretic peptide (BNP)], inflammation (GDF-15), or matrix remodelling (MMP-2, MMP-9), as well as primary agonists of the tyrosine receptor growth factor axis (neuregulin-1, IGF-1).

B-type natriuretic peptide and high-sensitivity troponin I were measured using a two-site alkaline phosphatase-conjugated chemiluminescent immunoassay by an automated Beckman Coulter DxI 800 analyser. For the BNP assay, the detection limit is 1 ng/L with minimal cross-reactivity with other related peptides. For the troponin assay, the Beckman Coulter recommended 99th percentile was 11.6 ng/L for females, coefficient of variation of <10% at 5.6 ng/L, limit of blank of 0.27 ng/L, and limit of detection of 0.90 ng/L. Plasma GDF-15, IGF-1, neuregulin-1, MMP-2 and -9 levels were determined using commercially available human enzyme-linked immunosorbent assay kits from R&D systems (DGD150, DG100, DY377, MMP200, and DMP900, respectively, Minneapolis, MN, USA) following manufacturer instructions. Absorbance was measured at 450 nm and corrected against 540 nm using a SpectraMax microplate reader (Soft Max^®Pro 5). These kits show negligible cross-reactivity (<0.5%) and high sensitivity (detection limits by manufacturer: GDF-15, 4.39 pg/mL; IGF-1, 0.056 ng/mL; MMP-2, 0.082 ng/mL; MMP-9, 0.156 ng/mL).

Statistical analysis

To characterize the cardiac phenotype of trastuzumab-mediated cardiotoxicity and the mediating effects of pharmacotherapy, two statistical approaches were used. First, temporal changes in cardiac MRI measures and circulating biomarkers were analysed for the groups independently using generalized linear mixed models. Participant was used as a random effect and time as a fixed and repeated effect. The distribution and link function that produced normality of model residuals (determined used QQ plots) was used. When normality was not achieved with any combination of distribution and link function, the model with the best fit, determined by comparing Akaike Corrected Information Criterion between models was used. Significant time main effects were investigated using contrasts between each time point. When a histogram indicated significant deviation from a normal distribution, circulating biomarkers were log-transformed and are reported as median (interquartile range). For body composition parameters, all groups were combined for this analysis due to lack of rationale for effect of bisoprolol or perindopril on these metrics. Second, to understand intergroup differences in the cardiac MRI and biomarkers, the magnitude of the change from baseline to post-cycle 4 and post-cycle 17 in all variables were compared between groups using analysis of covariance, with the baseline values as a covariate. Baseline values of each variable were also compared across groups using a one-way analysis of variance with post hoc pairwise comparisons by either Tukey’s or Games–Howell tests, depending on equality of variances. To determine the potential influence of pre-trastuzumab receipt of anthracycline treatment, subgroup analysis was performed comparing anthracycline vs. non-anthracycline regimens at each time point using independent t-tests. A P-value of ≤0.05 was used to denote statistical significance. SPSS version 26.0 (IBM Corp, Armonk, NY, USA) was used for all analyses.

Results

Clinical data

Ninety-four participants completed the study (n = 30, placebo; n = 33, perindopril; n = 31, bisoprolol). Patients received a regimen of docetaxel, carboplatin, and trastuzumab (TCH) every 3 weeks for 6 cycles (77%), or 5-fluorouracil, epirubicin and cyclophosphamide (FEC) every 3 weeks for 3 cycles, followed by docetaxel and trastuzumab every 3 weeks for 3 cycles (DH, 23%). In both regimens, trastuzumab was given every 3 weeks concurrent with docetaxel and then alone for a total of 17 cycles. Forty-one per cent also received left-sided radiotherapy. The groups did not differ in cancer pathology, treatment, or baseline demographic or cardiovascular profiles (Table 1). The mean dose received was 7 ± 2 mg/day for perindopril and 8 ± 3 mg/day for bisoprolol and there were no study drug withdrawals. As previously reported, baseline heart and blood pressure were similar between groups, while bisoprolol significantly reduced heart rate relative to placebo and perindopril, and perindopril and bisoprolol reduced systolic and diastolic blood pressure relative to placebo.¹¹

Table 1.

Participant characteristics

	Placebo	Bisoprolol	Perindopril	P-value
	n = 30	n = 31	n = 33
Age (years), mean ± SD	51 ± 7	53 ± 10	50 ± 8	0.41
Body surface area (m2), mean ± SD	1.8 ± 0.2	1.8 ± 0.2	1.9 ± 0.2	0.13
Pre-trastuzumab receipt of anthracyclines, n (%)	7 (23%)	4 (12%)	11 (35%)	0.17
Cardiovascular risk factors
Hypertension, n (%)	2 (7%)	2 (6%)	0	0.36
Type II diabetes, n (%)	0	3 (10%)	1 (3%)	0.16
Smoking status, n (%)				0.20
Current	3 (10%)	3 (10%)	2 (6%)
Former	12 (40%)	7 (23%)	12 (42%)
Dyslipidaemia	0	1 (3%)	1 (3%)	0.63

Cardiac magnetic resonance imaging

Data for LV volumes, EF, and mass and cardiotoxicity incidence have been previously reported.¹¹ Patients who had received FEC (n = 22) prior to the baseline assessment had lower LV function compared with patients scheduled to receive TCH (n = 72) (LVEF: 60 ± 5% vs. 63 ± 4%, P = 0.02; GLS: −18.1 ± 2.8% vs. −19.9 ± 2.4%, P = 0.005). However, these regimen-based differences had dissipated by post-cycle 4 and 17. LV end-systolic volume index increased significantly and to a similar extent in all three intervention groups but declines in LVEF post-cycle 4 and 17 were attenuated in the bisoprolol and perindopril groups by increased end-diastolic volume index (Table 2).

Table 2.

Cardiac magnetic resonance imaging data

	Placebo	Bisoprolol	Perindopril	Baseline (ANOVA P-value)
	(n = 30)	(n = 31)	(n = 33)
LV end-diastolic volume (mL/m2)
Baseline	76 ± 13	69 ± 10	67 ± 14a	0.01
Post-cycle 4	76 ± 9	75 ± 11b	70 ± 15b
Post-cycle 17	79 ± 12	76 ± 14b	74 ± 16b
LV end-systolic volume (mL/m2)
Baseline	30 ± 7	26 ± 5a	25 ± 6a	0.01
Post-cycle 4	35 ± 7b	31 ± 5b	29 ± 8b
Post-cycle 17	35 ± 8b	30 ± 6b	30 ± 8b
LV ejection fraction (%)
Baseline	61 ± 5	62 ± 4	62 ± 5	0.55
Post-cycle 4	54 ± 5b	59 ± 4	59 ± 6
Post-cycle 17	56 ± 4b	61 ± 4	59 ± 6
LV mass (g/m2)
Baseline	53 ± 8	51 ± 7	52 ± 7	0.63
Post-cycle 4	55 ± 8	52 ± 8	52 ± 7
Post-cycle 17	53 ± 8	52 ± 6	52 ± 8
LV global longitudinal strain (%)
Baseline	−20.1 ± 2.5	−19.4 ± 2.6	−19.0 ± 2.5	0.24
Post-cycle 4	−18.1 ± 2.7b	−19.6 ± 2.3	−18.1 ± 2.4b
Post-cycle 17	−18.8 ± 2.5b	−20.4 ± 2.0b	−18.9 ± 2.2
RV end-diastolic volume (mL/m2)
Baseline	71 ± 16	63 ± 12	65 ± 12	0.07
Post-cycle 4	71 ± 13	69 ± 12b	69 ± 14b
Post-cycle 17	74 ± 14	71 ± 14b	71 ± 17b
RV end-systolic volume (mL/m2)
Baseline	27 ± 9	24 ± 6	26 ± 7	0.15
Post-cycle 4	32 ± 8b	27 ± 7b	30 ± 6b
Post-cycle 17	31 ± 9b	28 ± 8b	29 ± 8b
RV ejection fraction (%)
Baseline	61 ± 7	62 ± 6	60 ± 6	0.31
Post-cycle 4	55 ± 6b	60 ± 6	57 ± 6b
Post-cycle 17	59 ± 6b	61 ± 8	60 ± 6
Left atrial volume (mL/m2)
Baseline	44 ± 9	36 ± 9a	37 ± 10a	<0.01
Post-cycle 4	45 ± 10	44 ± 9b	42 ± 12b
Post-cycle 17	46 ± 14	45 ± 11b	43 ± 11b

Data are mean ± standard deviation.

ACE, angiotensin-converting enzyme; LV, left ventricular; RV, right ventricular; missing data: global longitudinal strain, n = 3; native T₁ time, n = 36; extracellular volume, n = 39.

^aDifferent than placebo (P < 0.05).

^bDifferent than baseline (P < 0.05).

Post-cycle 4, LV GLS deteriorated from baseline in both placebo (+2.0 ± 2.7%) and perindopril (+0.9 ± 2.5%) groups but not with bisoprolol (−0.2 ± 2.1%, P < 0.01 compared with other groups, Table 2, Figure 1). Post-cycle 17, LV GLS remained reduced relative to baseline in the placebo group (+1.3 ± 2.5%), recovered in the perindopril group (+0.1 ± 2.4%), and was improved in the bisoprolol group (−1.0 ± 2.0, P < 0.01 compared with other groups, Figure 1). In each group, RV volumes changed in parallel to LV volumes (Table 2). Post-cycle 4, RVEF was decreased from baseline in placebo (−6 ± 6%) and perindopril (−3 ± 7%), but not bisoprolol (−2 ± 6%), but only bisoprolol differed from placebo (P < 0.01). Similar to both ventricles, LA volume did not change in the placebo group but was increased relative to baseline post-cycle 4 and 17 in bisoprolol (+7 ± 9 and +5 ± 8 mL/m², respectively) and perindopril (+5 ± 8 and +7 ± 12 mL/m², respectively, Table 2).

Figure 1.

Boxplots and individual data for absolute changes in global longitudinal strain between treatment groups after (A) cycle 4 (∼3 months) and (B) cycle 17 (∼12 months). Box shows interquartile range and median (middle line), with whiskers extending to the rest of the distribution except for outliers (defined as 1.5*interquartile range, indicated by diamonds). The dotted lines indicate no change.

T₁ and ECV data were not available in 14% of scans and therefore analysis was performed for all patients combined. Cancer regimen had no effect on observed native T₁ and ECV at any time point. In all patients, ECV fraction increased significantly from baseline to post-cycle 4 (22.9 ± 3.3 to 24.3 ± 3.4%, P = 0.004) but native T₁ did not change (1200 ± 41 to 1207 ± 46, P = 0.28). By post-cycle 17, ECV fraction had returned to baseline (22.4 ± 3.5%, P = 0.36) and native T₁ was lower than baseline (1184 ± 46 ms, P = 0.02).

Abdominal magnetic resonance imaging—body composition

Follow-up body composition imaging was not available in 8 (9%) participants and subcutaneous fat was excluded in 17% of cases as it was outside the field of view of the scanner; otherwise image quality allowed differentiation of fat and skeletal muscle (Figure 2, Supplementary material online, File). Body mass and all fat depots were significantly higher or trended in that direction in those who had received FEC at baseline, post-cycle 4 and 17 (mass: P = 0.008, 0.007, 0.03; subcutaneous: P = 0.05, 0.02, 0.07; visceral: P = 0.009, 0.06, 0.12; inter-muscular: P = 0.04, 0.009, 0.03, respectively); while muscle volume did not differ. In the overall cohort, there was an early increase at post-cycle 4 in body mass (+2 ± 5%, P = 0.008), visceral (+7 ± 17%, P = 0.02), and inter-muscular fat (+8 ± 23%, P = 0.02; Figure 3). Between post-cycle 4 and 17, muscle volume decreased (−2 ± 10%, P = 0.008), and subcutaneous fat volume increased (+6 ± 23%, P = 0.02; Figure 3). Over the 12 months of trastuzumab therapy, the relative increase in abdominal adiposity was disproportionate to the increase in overall body weight (1 ± 8%): subcutaneous fat (6 ± 23%), visceral fat (4 ± 19%), and inter-muscular fat (7 ± 25%). Skeletal muscle decreased −2 ± 10% over the same time frame.

Figure 2.

Examples of body composition: subcutaneous fat (blue), visceral fat (green), intermuscular fat (red), and muscle (orange) across the three time points in (A) a representative case with body weight loss across trastuzumab concurrent to muscle volume loss and fat volume elevation; and (B) a representative case of weight gain across trastuzumab concurrent to muscle volume loss and fat volume elevation. IMF, inter-muscular fat.

Figure 3.

Body composition changes over time for all groups combined. *Difference from baseline (P ≤ 0.05). ^†Difference from post-cycle 4 (P ≤ 0.05).

Biomarkers

At baseline, patients who had received FEC had significantly elevated circulating biomarkers of myocardial injury and inflammation compared with those scheduled to received TCH, high-sensitivity troponin [8.0 (5.9–12.3) vs. 2.1 (1.6–3.0) ng/L, P < 0.001], GDF-15 [616 (514–888) vs. 494 (384–597) ng/mL, P = 0.001], and MMP-2 (234 ± 37 vs. 191 ± 45 ng/mL, P < 0.001), respectively. High-sensitivity troponin remained higher in FEC at post-cycle 4 [5.0 (3.9–7.5) vs. 2.4 (1.9–3.5) ng/L, P < 0.001] and post-cycle 17 [3.2 (2.4–4.0) vs. 2.5 (1.6–3.5) ng/L, P = 0.02].

Regarding temporal changes in biomarkers by pharmaco-intervention, there were no between-group differences for magnitude of change other than for BNP which transiently increased in bisoprolol treated patients and transiently decreased in those receiving placebo or perindopril (Figure 4). Also, only in the bisoprolol group, neuregulin-1 increased from baseline at post-cycle 4 [713 (653–814) pg/mL to 869 (674–1094) pg/mL, P = 0.002] and post-cycle 17 [829 (718–1010) pg/mL, P = 0.03].

Figure 4.

Violin plots of distribution of circulating biomarkers by treatment group over time. Inside box shows interquartile range and median (white dot), with whiskers extending to the rest of the distribution except for outliers (>1.5 * interquartile range). The range of the plot includes all of the observed data. The coloured shape outside of the lines is a kernel density estimation to show the distribution of the data.

The changes over time for the remaining biomarkers were very similar within each group (Figure 4). IGF-1 was markedly decreased at post-cycle 17 (all groups combined: −18 ± 2% from baseline, P ≤ 0.001 within each group, Figure 4). GDF-15 transiently increased substantially from baseline to post-cycle 4 (+130 ± 150% from baseline, P ≤ 0.002 within each group, Figure 4). Similarly, MMP-2 increased transiently in all patients. Notably, high-sensitivity troponin I did not increase post-cycle 4 or 17 in any group.

Discussion

This secondary analysis of the MANTICORE trial contributes several new findings regarding temporal changes in the cardiac phenotype and body composition during trastuzumab therapy. We found that trastuzumab-mediated cardiotoxicity is characterized by concurrent LV and RV dysfunction and reversible myocardial inflammation and oedema. While bisoprolol and perindopril preserved biventricular function relative to placebo, they had minimal effects on circulating cardiac biomarkers. Our non-cardiac imaging and circulating biomarkers results suggest unfavourable body composition and growth factor adaptations during trastuzumab that could contribute to increased cardiovascular risk in breast cancer survivors independent of direct cardiac effects.

Effect of anthracyclines prior to trastuzumab

Those patients pre-treated with anthracyclines (FEC) prior to trastuzumab had reduced cardiac function (LVEF, GLS) and increased circulating biomarkers of injury (high-sensitivity troponin), inflammation (GDF-15), and remodelling (MMP-2), and greater abdominal fat compared with chemotherapy-naïve patients at our pre-trastuzumab baseline assessment. While the differences in GDF-15, MMP-2, or LV function did not persist throughout trastuzumab, high-sensitivity troponin I and abdominal fat remained higher post-cycle 4 and 17 in those who had received initial sequential anthracycline therapy. While elevated troponin has been mechanistically linked to anthracycline-related myocardial injury, further study is required to better characterize the impact of anthracycline therapy on the cardiometabolic profile.

Changes in cardiac phenotype

We found that trastuzumab-related RV dysfunction and remodelling occurs in parallel to changes of the LV. In an observational study of 41 patients receiving trastuzumab ± anthracyclines, biventricular function also decreased at 6 months and subsequently recovered to baseline by 18 months.¹⁶ We extend these findings by showing that perindopril and bisoprolol had protective effects on depression of LV GLS post-cycle 17.

In all groups, we observed an increase in GDF-15, MMP-2, and ECV fraction from baseline to post-cycle 4 followed by a return to baseline post-cycle 17 without concurrent changes in high-sensitivity troponin I. GDF-15 is a stress-induced cytokine released from vascular smooth muscle cells and cardiomyocytes in response to injury, anoxia, and inflammation,¹⁷ while MMP-2 is involved in the proteolysis of several myocardial proteins, leading to reduced contractility and cardiac dysfunction.¹⁸ These findings suggest that trastuzumab induces early myocardial inflammation and oedema, in line with previous preclinical reports of mononuclear cell infiltration and myocardial edema^19,20 and clinical reports of myocardial oedema on T₂-weighted imaging after trastuzumab exposure.²¹ However, we found that myocardial oedema is transient, with no discernable relationship to changes in cardiac function.

In contrast to prior studies of anthracycline-mediated cardiotoxicity, we found no increase in high-sensitivity troponin during trastuzumab therapy.^5–8 A similar finding was reported in 71 patients with breast cancer undergoing serial testing during non-anthracycline, trastuzumab-based treatment.²² Therefore, cardiomyocyte injury does not appear to play a critical role in trastuzumab-mediated cardiotoxicity and, consequently, expert consensus recommendation of measuring troponin every 3 months during trastuzumab therapy²³ may not provide added information to imaging unless patients also receive anthracyclines.

Neuregulin-1 is a paracrine growth factor that is released from cardiac and vascular endothelial cells in response to physiologic stress and binds to nearby ErbB receptors to promote adaptation of cardiac structure and function.²⁴ As trastuzumab disrupts neuregulin-1/ErbB signalling, the small drop in neuregulin-1 reported by Geisberg et al.²⁴ with anthracycline or trastuzumab (37% HER2-positive) therapy was suggested to represent a loss in cardio-reparative capacity. Our finding of bisoprolol increasing neuregulin-1 by post-cycle 4 with maintenance to post-cycle 17 suggests that potentially some of the protection by bisoprolol was mediated by enhancement of endogenous neuregulin-1 levels.

Changes in cardiometabolic phenotype

To our knowledge, this is the first study to prospectively evaluate changes in body composition with MRI during trastuzumab. After just 3 months of chemotherapy and trastuzumab therapy, both visceral and intermuscular fat volume were significantly increased. Both fat deposition locations carry independent risk for metabolic dysfunction including insulin resistance, dyslipidaemia, and atherogenesis, as well as cardiovascular morbidity and mortality,^25,26 including for women with early-stage breast cancer, irrespective of pre-existing cardiovascular risk factors and cancer treatment.⁹

Another novel metabolic finding was that we observed very low circulating IGF-1 at the completion of trastuzumab therapy, with 30% of patients within the 2.5th percentile for healthy women²⁷ compared with 11% at baseline. The growth hormone/IGF-1 axis plays a key role in muscle mass regulation,²⁸ and low IGF-1 levels are associated with a substantially higher risk of cardiovascular death, possibly through adverse effects on visceral adiposity and dyslipidemia.²⁹ Taken together, our observed changes to the cardiometabolic phenotype may provide a pathophysiological basis for the increased risk of cardiovascular events in breast cancer survivors and suggest that clinicians should promote healthy lifestyle choices, including access to cardio-oncology rehabilitation during trastuzumab therapy.

Limitations

Our small sample size and low number of LV dysfunction events precluded predictive analysis for cardiotoxicity, yet the high-quality and reproducible MRI data enabled identification of several important and novel cardiac and cardiometabolic changes. T₁ mapping data were not available in 14% of assessments due to difficulty with image acquisition and we did not acquire T₂ mapping, another MRI measure of myocardial oedema. However, the early myocardial oedema suggested by the elevated post-cycle 4 ECV fraction is supported by corresponding increases in GDF-15, a biomarker of myocardial inflammation. Also, our patient cohort was relatively young, with a lower prevalence of hypertension and diabetes and therefore may represent a relatively lower-risk population. Finally, changes in cardiometabolic phenotype should be also be evaluated in patients with HER2 negative breast cancer.

Conclusions

Our MRI and circulating biomarkers data suggest that trastuzumab therapy is associated with early LV dysfunction and reversible myocardial inflammation and oedema without concurrent myocardial fibrosis or injury. Visceral and inter-muscular fat accumulate while IGF-1 and muscle volume decrease during trastuzumab, all of which have been associated with higher risk of cardiovascular events in population studies. Bisoprolol and perindopril preserve biventricular function but had no discernable effect on myocardial inflammation.

Supplementary material

Supplementary material is available at European Heart Journal – Cardiovascular Pharmacotherapy online.

Data availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Funding

The trial was supported by the Canadian Institutes of Health Research, the Alberta Cancer Foundation, and the University Hospital Foundation. Perindopril study drug was provided in-kind by Servier. A.A.K. is supported by Postdoctoral Fellowships from Susan G. Komen Foundation (PDF17483149) and the Canadian Institutes of Health Research. J.A.E. and G.Y.O. are supported by Alberta Innovates-Health Solutions, the Canadian Institutes of Health Research, and/or the Heart and Stroke Foundation of Canada. M.J.H. is supported by the Faculty of Nursing Research Chair in aging and quality of life at the University of Alberta.

Conflict of interest: K.C. is currently an employee of Siemens Healthineers but was a graduate student at the time of the study. J.A.E. reports study funding from Novartis, Servier and Pfizer and grants from Merck, Bayer, Trevena and Amgen. G.Y.O. reports study funding from Amgen. All other authors have no conflicts of interest to disclose.