Myocardial injury detected by T1 and T2 mapping on CMR predicts subsequent cancer therapy–related cardiac dysfunction in patients with breast cancer treated by epirubicin-based chemotherapy or left-sided RT

Objectives Cancer therapy-related cardiac dysfunction (CTRCD) is a relevant clinical problem and needs early prediction. This study aimed to analyze myocardial injury using serial laboratory and cardiac magnetic resonance imaging (CMR) parameters after epirubicin-based chemotherapy compared with left-sided radiotherapy and to study their value for early prediction of CTRCD. Methods Sixty-six consecutive women (53 ± 13 years) including n = 39 with epirubicin-based chemotherapy and n = 27 with left-sided radiotherapy were prospectively studied by 3 T CMR including left ventricular (LV) mass and volumes for ejection fraction (LVEF), as well as feature-tracking with global longitudinal strain (GLS) and T1/T2 mapping. CMR was performed at baseline, at therapy completion (follow-up 1, FU1), and after 13 ± 2 months (FU2). CTRCD was defined as LVEF decline of at least 10% to < 55% or a > 15% GLS change at FU2. Results T1 and T2 increased at FU1 after epirubicin-based chemotherapy, but not after left-sided radiotherapy. CTRCD occurred in 20% of patients after epirubicin-based chemotherapy and in 4% after left-sided radiotherapy. T1 at FU1 was the best single parameter to predict CTRCD with an area under the curve (AUC) of 0.712 (CI 0.587–0.816, p = 0.005) with excellent sensitivity (100%, 66–100%), but low specificity (44%, 31–58%). Combined use of increased T1 and LVEF ≤ 60% at FU1 improved AUC to 0.810 (0.695–0.896) resulting in good sensitivity (78%, 44–95%) and specificity (84%, 72–92%). Conclusion Only epirubicin-based chemotherapy, but not left-sided radiotherapy, resulted in increased T1/T2 myocardial relaxation times as a marker of myocardial injury. Combined use of CMR parameters may allow an early prediction of subsequent CTCRD. Key Points • Myocardial T1 and T2 relaxation times increased at FU1 after epirubicin-based chemotherapy, but not after left-sided radiotherapy. • Cancer therapy–related cardiac dysfunction (CTRCD) occurred in 20% of patients after epirubicin-based chemotherapy and in 4% after left-sided radiotherapy. • Combined use of increased T1 and reduced LVEF had an AUC of 0.810 (0.695–0.896) to predict CTRCD with good sensitivity (78%, 44–95%) and specificity (84%, 72–92%). Supplementary Information The online version contains supplementary material available at 10.1007/s00330-021-08260-7.


Introduction
The standard of care in early breast cancer, which has not spread beyond the breast or the axillary lymph nodes, includes systemic anthracycline-based chemotherapy and/ or radiotherapy [1,2]. A typical scenario for chemotherapy followed by postoperative radiotherapy would be a tumor > 2 cm or not feasible for optimal surgery and wish for breast conservation as recommended by current guidelines [2]. Epirubicin and doxorubicin are the most widely used anthracyclines for breast cancer treatment but are associated with increased risk for cancer therapy-related cardiac dysfunction (CTRCD) [3]. Radiotherapy of breast cancer is strongly recommended after breast-conserving therapy both in combination with preceding chemotherapy or radiotherapy alone [2]. However, radiotherapy resulted in longterm cardiac toxicity such as heart failure and myocardial infarction related to heart radiation [4,5]. Early studies showed that the risk for cardiovascular death was increased in patients with left-sided breast cancer as a result of higher cardiac radiation [4,5]. The cardiotoxic effect was presumably mediated by the progression of coronary heart disease; however, the aforementioned studies were performed several decades ago by using less well-established dose reduction strategies. The cardiotoxic effect of current radiation strategies with reduced cardiac dose such as three-dimensional CT-based planning and deep inspiration breath-hold technique are less well understood.
Cardiac magnetic resonance imaging (CMR) allows accurate detection of myocardial injury using novel T1 and T2 mapping techniques, which detect myocardial edema related to acute injury [6,7]. A recent animal study by Galán-Arriola et al underscored their value for detection of cardiotoxicity after intracoronary injection of doxorubicin [8]. This study revealed that T2 was the earliest mapping parameter of cardiotoxicity. T2 increased 6 weeks after chemotherapy initiation and correlated with intracardiomyocyte edema. Subsequently, T1 increased and was associated with the decline of LV ejection fraction [8]. We hypothesized that T1/T2 mapping CMR is able to detect cardiotoxicity in patients either treated by epirubicin-based chemotherapy or by radiotherapy and is useful for early prediction of subsequent CTRCD.
The aims to prospectively analyze myocardial injury using serial T1/T2 mapping CMR and laboratory parameters after epirubicin-based chemotherapy compared with leftsided radiotherapy and to study the value of these parameters for early prediction of CTRCD.

Patient groups and CMR scheduling
The local ethics committee approved the study (PV5292) and all women gave their written informed consent. Women with breast cancer were prospectively recruited between October 2016 and October 2017 (Fig. 1). The treatment strategy was decided by an interdisciplinary tumor board as part of clinical routine. Subsequently, patients were either treated with epirubicin-based chemotherapy followed by radiotherapy (group 1) or with left-sided radiotherapy (group 2). Patients were not randomized to the treatment strategies of group 1 or group 2. Group 1 included 39 patients (age: 51 ± 11 years, range 30-75 years) and one subject was excluded due to claustrophobia at baseline CMR (Fig. 1). Chemotherapy consisted of 4 cycles of epirubicin (90 mg/m 2 ) combined with cyclophosphamide (600 mg/m 2 ) and 12 cycles of paclitaxel (80 mg/m 2 ) according to current guidelines [2]. Five group 1 patients (13%) additionally received trastuzumab due to human epidermal growth factor receptor 2 positive (HER2 +) status. Chemotherapy lasted for 5 ± 1 months and follow-up 1 (FU1) CMR was performed at 2 ± 2 weeks after completion (Fig. 2). Subsequently, 18 group 1 patients (46%) underwent left-sided radiotherapy, whereas 21 (54%) obtained right-sided radiotherapy. Surgery was either performed before (n = 15, 38%) or after chemotherapy (n = 24, 62%). FU2 CMR was acquired in group 1 at 13 ± 2 months after BL CMR.
Group 2 included 27 women (age: 56 ± 14 years, range 28-82 years) with left-sided breast cancer, who were treated with initial surgery followed by left-sided radiotherapy lasting for 1.2 ± 0.4 months (Fig. 2). Whole breast radiation was performed using a linear accelerator (Varian TrueBeam®, Varian Medical Systems). Three-dimensional tangential treatment plans with a  Patients with epirubicin-based chemotherapy received therapy for 5 ± 1 months followed by follow-up 1 (FU1) CMR at 2 ± 2 weeks after completion of therapy. Surgery (S) was either performed before (n = 15) or after chemotherapy (n = 24). Patients with left-sided radiotherapy received initial surgery, followed by radiotherapy and FU1 CMR at 1 ± 2 weeks after completion of radiotherapy. Follow-up 2 (FU2) CMR was performed in both groups at 13 months after baseline (BL) CMR "field-in-field" technique were established to minimize the heart dose [9]. A deep inspiration breath-hold technique was applied to reduce the heart radiation dose. Respiratory motion was monitored using the Varian respiratory gating system and the patient was instructed to take a deep breath in and hold it. The treatment beam was stopped when the breathing signal dropped outside a defined threshold for the respiratory motion of the heart [10]. The total whole-breast radiation dose was 48 ± 4 Gy and the mean cardiac dose was 2 ± 2 Gy. FU1 CMR was acquired at 1 ± 2 weeks after completion and FU2 CMR was obtained 13 ± 1 months after BL CMR. Blood samples were drawn immediately before each CMR.

CMR protocol
CMR was performed on a 3.0-T scanner (Ingenia, Philips Medical Systems). The CMR protocol included a cine short-axis stack for LV volumes, mass, and function assessment using a standard steady-state free-precession (SSFP) sequence. T2 mapping was performed using a free-breathing navigator-gated black-blood prepared gradient and spin-echo (GraSE) hybrid sequence in three short-axis slices (basal, midventricular, and apical) [11]. T1 mapping was performed using a 5 s(3 s)3 s MOLLI sequence with typical imaging parameters: voxel size 2 × 2 × 10 mm 3 , echo time = 0.7 ms, time to repetition = 2.3 ms, partial echo factor = 0.8, flip angle = 35°, SENSE factor = 2, linear phase encoding, ten start-up cycles to approach steady-state prior to imaging, typical effective inversion times between 134 and 5500 ms [16]. Ten minutes after injection of 0.15 mmol/kg gadoterate meglumine (Dotarem®, Guerbet), end-diastolic late gadolinium enhancement (LGE) images were acquired using phase-sensitive inversion recovery (PSIR) sequences in short-axis and two-, three-, and four-chamber views [12]. Details are in Supplemental Material.

CMR data analysis
Two investigators (M.A. and S.S.) independently and blindly analyzed each CMR in random order using CVi42 software (Circle Cardiovascular Imaging Inc). Parameters were indexed to the body surface area (BSA). Evaluation of LV and right ventricular (RV) volumes and LV mass was performed in standard fashion [13]. To measure global native T1 and T2 relaxation times and ECV, corresponding short-axis maps were used to carefully delineate endo-and epicardial contours with 10% endo-and epicardial offsets to avoid contamination. The presence of LGE was visually analyzed [13].

Myocardial strain analysis
Myocardial strain was analyzed on cine CMR images using Segment feature-tracking software version 2.1.R.6108 (Medviso) [14]. In short, this software analyzes myocardial strain by computing interframe deformation fields using an endocardial tracking strategy based on non-rigid image registration [14]. Global peak systolic LV longitudinal (GLS) and radial strain (GRS) were measured on 3 long-axis cine series, whereas circumferential strain (GCS) was measured on three short-axis cine series. Endoand epicardial contours were manually delineated on enddiastolic images and were then automatically propagated by the software throughout the cardiac cycle generating myocardial strain [14].

Cancer therapy-related cardiac dysfunction
Presence of cancer therapy-related cardiac dysfunction (CTRCD) was defined as a decline in LVEF of at least 10 to < 55% or a > 15% GLS change at FU2 CMR [3,15].

Statistical analysis
Statistical analysis was performed using MedCalc for Windows, version 13.3.3.0 (MedCalc Software), and SPSS for Windows, version 21.0 (IBM SPSS Statistics). All CMR data are given as the mean of two observers. Continuous data are presented as mean ± SD and categorical data are presented as absolute numbers and percentages. A mixed model was applied to analyze changes in the obtained parameters with time and pairwise comparison. A receiver-operating characteristic (ROC) analysis was performed to evaluate the performance of the obtained values at FU1 to predict the occurrence of CTRCD at FU2 by calculating areas under the curve (AUC) and optimal cut-off values from the ROC curves using the Youdenindex. Statistical significance was defined as p < 0.05.

Baseline demographics of both patient groups
Patients with epirubicin-based chemotherapy had higher weight, body mass index (BMI), and body surface area (BSA) compared to patients with left-sided radiotherapy (Table 1). Cancer stage is given for both groups in Table 1. There were no differences regarding cardiovascular risk factors and medication. No patient had known heart failure. High sensitive troponin T and NT-proBNP revealed no cardiac injury at baseline (Table 1).

Therapy-induced changes of laboratory and CMR parameters after radiotherapy
High sensitive troponin T slightly increased after leftsided radiotherapy to 6 ± 3 pg/ml at FU1 (p < 0.04) and normalized at FU2 (Table 3). All other laboratory parameters were unchanged at FU1 and FU2. Mean T1 and T2 remained constant at FU1, indicating the absence of myocardial edema (Table 3, Fig. 3). This was paralleled by normal LV function measured by ejection fraction and myocardial strain (Table 3). A slight reduction in LV and RV volumes was observed at FU1 and FU2. No focal or diffuse myocardial fibrosis was observed by LGE or ECV.

Patients with and without CTRCD
CTRCD occurred at FU2 in 8 of 39 patients (20%) after epirubicin-based chemotherapy and in one of 27 patients (4%) after left-sided radiotherapy, resulting in a total number of 9 of 66 patients (14%) with CTRCD. High sensitive troponin T and T1 and T2 increased in both groups at FU1 and normalized at FU2 (Table 4). LVEF was reduced at FU2 in patients with CTRCD according to the definition of this condition, whereas all strain parameters were already reduced at FU1 and remained reduced at FU2. LV and RV volume reduction typically occurred in both patient groups at FU1 and mostly normalized at FU2.

Diagnostic accuracy of parameters at FU1 to predict later CTRCD at FU2
ROC analysis revealed that T1 and GCS at FU1 were the best single parameters to predict CTRCD at FU2 with an AUC of 0.712 (p = 0.005) and 0.712 (p = 0.07), respectively ( Table 5, Fig. 4a). T1 resulted in excellent sensitivity of 100% (66-100%), but low specificity of 44% (31-58%). Conversely, GCS resulted in an acceptable sensitivity of 67% (30-93%) and specificity of 75% (61-85%). Interestingly, blood parameters of myocardial injury such as high sensitive troponin T and NT-proBNP had low AUC (Table 5). T2 had the lowest AUC of 0.513 of all parameters to predict CTRCD. The combined use of T1 as a parameter with excellent sensitivity with a parameter with higher specificity revealed that the combined use of T1 > 1262 ms and reduced LVEF ≤ 60% following breast cancer treatment at FU1 resulted in an improved AUC = 0.810. This combined use resulted in high sensitivity of 78% (44-95%) and high specificity of 84% (72-92%). The combined use of T1 and GRS or T1 and GCS also improved the AUC to 0.800 and 0.762, respectively. Other combinations of parameters were inferior to the single use of T1 (Table 5, Fig. 4).

Discussion
This prospective study analyzed T1 and T2 mapping CMR and laboratory changes in women with breast cancer treated either by epirubicin-based chemotherapy or leftsided radiotherapy. Furthermore, the predictive value of CMR and laboratory parameters for early identification of CTRCD were analyzed. The major findings were as follows: (1) T1 and T2 relaxation times increased after epirubicin-based chemotherapy, but not after left-sided radiotherapy.

Post-chemotherapy myocardial injury in breast cancer patients
Galán-Arriola et al recently showed in an animal study that increased T2 is the earliest imaging marker of doxorubicin-related cardiotoxicity [8]. T2 increase was related to intracardiomyocyte edema, which occurred as early as 6 weeks after initiation of chemotherapy in pigs. This study showed that the T2 increase was not paralleled by a decrease of ejection fraction and discontinuing chemotherapy upon detection of T2 prolongation halted LV dysfunction development, demonstrating that early T2 prolongation occurred at reversible disease stages [8]. This study also demonstrated that T1 progressively increases after 10 weeks of doxorubicin and is coincided with a progressive ejection fraction decline [8].
Our study in humans confirms and extends the knowledge about the value of T2 and T1 to detect chemotherapy-related myocardial injury. Interestingly, we found a more pronounced increase in T1 compared to T2 in our patients with chemotherapy. This finding is most likely related to the fact that we performed the first follow-up CMR 2 weeks after chemotherapy completion, which was given for 5 months, whereas Galán-Arriola et al performed weekly CMR [8]. We presume that we imaged the patients at a later time point of myocardial injury when T2 had already normalized, but T1 was still elevated. Similar to Galán-Arriola et al, we observed that the T1 increase was paralleled by a LV dysfunction quantified by strain analysis [8].

Occurrence of CTRCD
The CTRCD frequency of 20% after chemotherapy is well in line with a recent meta-analysis, which reported a cardiotoxicity frequency of 17% with the highest risk in the high-dose chemotherapy subgroup with 19.6% [16]. The occurrence of CTRCD after radiation is currently less well analyzed. Recent studies showed that left-sided radiation results in reduced myocardial strain Table 4 Clinical and CMR parameters in patients with and without cancer therapy-related cardiac dysfunction (CTRCD) * p < 0.05, † p < 0.01 and ‡ p < 0.001 for baseline vs FU1 ll p < 0.05, ¶ p < 0.01 and # p < 0.001 for baseline vs FU2 CTRCD was defined as decline in LVEF of at least 10% to < 55% or a > 15% GLS change at FU2 With CTRCD (n = 9) Without CTRCD (n = 57)  [17,18]. However, the incidence of CTRCD was not reported. Currently, several ongoing studies will analyze early detection and prediction of cardiotoxicity after radiation therapy for breast cancer and will provide more precise data about CTRCD [19,20]. We observed a low frequency of CTRCD after radiation with 4% by using established criteria. The low frequency is potentially related to the low cardiac radiation dose of 2 ± 2 Gy, which was achieved by dedicated three-dimensional treatment plans including tangential beam orientation and deep inspiration breath-hold technique.

Diagnostic accuracy of parameters at FU1 to predict later CTRCD at FU2
Our study showed that T1 and GCS at FU1 were the best single parameters to predict CTRCD at FU2 with good AUCs of 0.712. T1 was characterized by excellent sensitivity of 100%, but low specificity of 44%. Conversely, GCS resulted in a balanced sensitivity of 67% and specificity of 75%. Interestingly, blood parameters of myocardial injury such as high sensitive troponin T and NT-proBNP had lower AUCs compared to T1 and GCS, indicating that the increase of these laboratory parameters is an epiphenomenon of chemotherapy and radiation, but they have limited value to identify CTRCD. This is in line with a recent meta-analysis on the use of troponins and NT-proBNP for cardiotoxicity prediction after chemotherapy reporting similar sensitivity of 69%, but higher specificity of 87% for troponins and limited predictive value of NT-proBNP [16]. Interestingly, T2 had the lowest AUC of 0.513 to predict CTRCD. In patients with an acute myocardial injury such as myocarditis or myocardial infarction, T1 and T2 are equally good in identifying disease [21,22]. In contrast, the current data indicated that T2 is of limited value to identify CTRCD. Altaha et al recently reported a significant overlap of temporal changes in CMR mapping parameters between chemotherapy patients and healthy subjects, whereas T1 showed less overlap than T2 [15]. Similarly, we found an T1 and T2 overlap between patients with and without CTRCD. Thus, the low specificity of T1 and low AUC of T2 might be explained by their moderate increase at the chosen time point following oncologic therapy as opposed to the more pronounced changes caused by myocarditis and myocardial infarction. It is possible, that imaging at a different time point during or after cancer therapy may result in a better performance of these parameters to predict CTRCD.

Limitations
The sample size is relatively small. However, this prospective and consecutive study adds to the body of evidence on cardiotoxic effects of cancer therapy in humans and evaluates the predictive value of CMR. Second, our findings are confined to the chemotherapy and radiotherapy applied in this study. Fig. 4 Diagnostic accuracy of parameters at follow-up 1 (FU1) to predict cancer therapy-related cardiac dysfunction (CTRCD) at followup 2 (FU2). T1 and global circumferential strain (GCS) at FU1 were the best single parameters to predict CTRCD at FU2 with an AUC of 0.712 (p < 0.01) and AUC of 0.712 (p < 0.05), respectively (a). The combined use of T1 with either left ventricular ejection fraction (LVEF) or global radial strain (GRS) resulted in an improved AUC of 0.810 or 0.800, respectively (b)