Comparison of the diagnostic and prognostic value of criteria for immune checkpoint inhibitor related myocarditis

Abstract

Background

Myocarditis is a dreaded complication of immune-checkpoint inhibitor (ICI) therapy but challenging to diagnose. There are no published data comparing the two leading diagnostic criteria for ICI-related myocarditis (ICIrM) and their association with cardiovascular events.

Methods

In this retrospective cohort study, we reviewed all patients who underwent ICI therapy and had cardiac troponin assessment for possible myocarditis across three tertiary institutions from 2011 to 2022. ICIrM was adjudicated by the Bonaca et al. criteria and the ESC-ICOS guidelines. A propensity matched control group was identified of patients treated with ICI without developing myocarditis. Baseline characteristics and long-term outcomes, including cardiac death, MACE (myocardial infarction, TIA/stroke, heart failure), and arrhythmias data were curated, and patients diagnosed with ICIrM by each criteria were compared to controls for cardiovascular events.

Results

A total of 59 patients (mean age was 73.1 ± 10.2 years, 60.1% male) were identified as having a diagnosis of ICIrM by Bonaca criteria (16 definite, 13 probable and 30 possible myocarditis). Forty-seven of these patients met the ESC-ICOS guidelines criteria, and all patients meeting either set of ICIrM criteria were treated with steroid therapy. At 3-year follow up, patients diagnosed with ICIrM by the Bonaca criteria had a high risk of cardiac mortality (HR 17.84, 95%CI 2.36-134.62, p = 0.005), MACE (HR 4.90, 95%CI 2.40-10.02, p < 0.001) and arrhythmias (HR 3.33, 95%CI 1.78–6.21, p < 0.001) when compared to matched controls. ICIrM by ESC-ICOS criteria was similarly predictive of cardiac mortality, MACE, and arrhythmias (HR 15.01, 95%CI 1.96-114.76, p = 0.009, HR 5.18, 95%CI 2.33–11.53, p < 0.001, and HR 3.41, 95%CI 1.73–6.70, p < 0.001 respectively).

Conclusion

The ESC-ICOS guidelines were more restrictive than the Bonaca et al. criteria for the diagnosis of ICIrM but similar in terms of prognostic value.

Introduction

Immune-checkpoint inhibitor (ICI) therapy is increasingly utilized due to their ground-breaking ability to effectively treat several cancers and improve prognosis [1,2,3,4]. However, ICI therapy can lead to immune-related adverse events such as colitis, dermatitis, hepatitis, nephritis, and myocarditis [5,6,7,8,9,10]. Among these, ICI-related myocarditis (ICIrM) has been associated with the highest mortality risk secondary to the development of hemodynamic or electrical instability, as well as cessation of cancer therapy [11,12,13].

The diagnosis of ICIrM has proven challenging as the clinical presentation can range from asymptomatic courses to fulminant end-organ failure or cardiogenic shock, and the central diagnostic studies of endomyocardial biopsy (EMB) and cardiac magnetic resonance imaging (CMR) may not be available or possible. Two main sets of diagnostic criteria have been proposed for ICIrM, the first being the 2019 Bonaca et al. criteria, which categorizes ICIrM diagnosis as definitive, probable and possible, and the 2022 ESC-ICOS (European Society of Cardiology-International Cardio-Oncology Society) diagnostic criteria which offers a single diagnosis of ICIrM if criteria are met [14, 15]. Other than cardiac biopsy, both criteria rely on a combination (with differences in emphasis) of symptomatic presentation, elevated cardiac biomarkers, and cardiac abnormalities detected on CMR, transthoracic echocardiogram (TTE) or electrocardiogram (ECG).

The clinical impact, both diagnostic and prognostic, of these two sets of criteria is not known. Herein, we thus sought to classify ICIrM by applying both, the Bonaca and ESC-ICOS criteria, to a large cohort of patients receiving ICI therapy, and evaluate the clinical and prognostic differences between them.

Methods

Study population

The Mayo Clinic Institutional Review Board (IRB) approved this retrospective cohort study. Requirement for consent was waived for retrospective review of the electronic charts by the IRB. Adult (age ≥ 18 years) patients who were treated with an ICI agent between January 2011 and February 2022 at Mayo Clinic sites across three states (Rochester MN, Phoenix AZ, and Jacksonville FL) were identified. For ICIrM diagnosis, electronic medical records (EMRs) were reviewed for all patients who had received ICI therapy and had at least one troponin measurement after initiation of ICI therapy. Patients identified as having potential ICIrM were adjudicated for diagnosis based on the Bonaca et al. and the ESC-ICOS diagnostic criteria. A propensity matched control group was selected for comparison from patients who had received ICI therapy but did not develop ICIrM.

Clinical data

Baseline demographics, cardiovascular risk factors and comorbidities, laboratory values including cardiac troponin, NT-proBNP, C-reactive protein (CRP), as well as echocardiographic parameters including left ventricular (LV) ejection fraction (EF), and valvular heart disease defined as moderate or severe valvular stenosis or regurgitation as reported by echocardiography were manually extracted from EMRs. Oncological characteristics including type of cancer, previous treatment with potentially cardiotoxic agents, type of ICI administered, and respective doses were also curated.

Classification of patients with ICIrM utilized clinical presentation, laboratory and ECG findings and medical imaging as per Bonaca et al. and ESC-ICOS guidelines, (Supplementary Table 1). A significant troponin elevation is defined by the ESC-ICOS guideline as “cTnI/cTnT > 99th percentile, or new significant rise from baseline beyond the biological and analytical variation of the assay used” [15], while the Bonaca et al. criteria consider a troponin value > 99th percentile as elevated [14]. Medical imaging included transthoracic echocardiography (TTE), cardiac magnetic resonance (CMR) and positron emission tomography (PET). The EMRs were reviewed for endomyocardial biopsy (EMB) results when undertaken, and other diagnostic workup that may have been conducted to exclude coronary artery disease as the cause of the presentation. The first and senior authors re-reviewed and adjudicated all patients flagged as potentially having ICIrM to ensure adherence with Bonaca et al. and ESC-ICOS guideline defining criteria, and consensus with final approval of the senior author was used to adjudicate discordant cases.

Long-term clinical outcomes were curated from the charts including all-cause and cardiovascular mortality, myocardial infarction (MI), ischemic stroke/transient ischemic attack (TIA), heart failure (HF) which was defined as new HF diagnosis and HF exacerbation, and new onset arrhythmias defined as atrial fibrillation (AF), ventricular tachycardia (VT) and complete heart block (CHB) from the first ICI dose administration until July 2024. A composite of major adverse cardiovascular events (MACE), defined as HF, MI, and ischemic stroke/TIA, was also evaluated. Cardiovascular (CV) related mortality was defined as any death due to acute MI, HF, stroke, sudden cardiac death, or death during the admission for ICIrM [16].

Statistical analysis

Continuous variables were summarized as mean and standard deviation (mean ± SD) or median and interquartile range (median [IQR]) according to distribution, while categorical variables were presented as frequencies with percentages. Independent samples t-test or non-parametric tests were used to compare continuous variables and chi-square to compare categorical variables. A propensity model was applied based on demographics (age, sex), cardiovascular risk factors (hypertension, dyslipidemia, diabetes, smoking, family history of coronary artery disease), cardiovascular diagnoses (history of MI, history of stroke/TIA), and type of cancer to select a propensity matched control group for comparison with a 1:1 ratio with a caliper width of 0.01.

Association between ICIrM and long-term clinical outcomes were summarized with hazard ratios (HR) and 95% confidence intervals (CI), derived with Cox Proportional Hazard Regression models. The proportional-hazards assumption was tested by plotting the log–minus–log survival function against the log(time). Kaplan-Meier (KM) curves were used to show survival differences between patients with and without ICIrM. P values of < 0.05 were considered statistically significant for all analyses. Statistical analyses were conducted using IBM SPSS Statistics software, version 28.0 (IBM SPSS Inc., Armonk, NY, USA).

Results

By applying the Bonaca et al. criteria, 59 (1.1%) of the 5424 patients who received ICI therapy were diagnosed with ICIrM, with 16, 13 and 30 patients being classified as definitive, probable and possible ICIrM, respectively. Of these 59 patients, three had a positive EMB, 12 had a diagnostic CMR, 10 had a suggestive diagnosis of myocarditis by CMR, 10 had a new abnormality by ECG, 8 had a new wall motion abnormality by TTE, and 5 had a positive PET. All patients had signs and symptoms consistent with or suggestive of ICIrM.

All but 12 patients with ICIrM by the Bonaca et al. criteria met also the ESC-ICOS criteria. Amongst these 47 (0.8%) patients positive for ICIrM by ESC-ICOS criteria, 3 had a positive EMB, 10 had a diagnostic and 5 had a suggestive finding of myocarditis by CMR, 9 had a new abnormality by ECG, 7 had new wall motion abnormality by TTE, and 4 had a positive PET. Only one patient with ICIrM (by both criteria) had percutaneous coronary intervention, with stent placed to the LAD at time of EMB for incidentally found coronary stenosis, however EMB was positive for ICIrM and the patient as also treated with immunosuppression.

Mean LVEF% was 56.9 ± 13.8 by TTE in those diagnosed with ICIrM by Bonaca et al. criteria, and 56.0 ± 15.1 in those diagnosed by the ESC-ICOS criteria; by MRI, mean LVEF% was 49.9 ± 14.1 and 49.0 ± 15.5 in those diagnosed by Bonaca et al. and ESC-ICOS criteria respectively. In the ICIrM group diagnosed by Bonaca et al. criteria, 40 patients (67.8%) underwent CMR, with 30.0% yielding diagnostic results, 42.5% being non-diagnostic or equivocal, and 27.5% suggestive. In the ESC-ICOS diagnosis group, 30 patients (63.8%) underwent CMR, with 33.3% diagnostic, 50.0% non-diagnostic or equivocal, and 16.7% suggestive.

Most patients with ICIrM received therapy with a single ICI agent (45 [76.3%] in those diagnosed by Bonaca et al. criteria, and 36 [76.6%] in those diagnosed by ESC-ICOS criteria). Dual ICI therapy was administered in 14 (23.7%) and 11 (23.4%), and history of radiation therapy was present in 22 (37.3%) and 17 (36.2%) respectively in those diagnosed by Bonaca et al. and ESC-ICOS criteria.

A propensity matched control group of patients (n = 59 for Bonaca criteria, n = 47 for ESC-ICOS criteria) who underwent ICI therapy and without evidence of myocarditis was obtained. In the overall cohort mean age was 69.8 ± 12.3 years and 69.5% were male. Median time from first ICI dose to diagnosis of ICIrM was 44 days (IQR 74). Baseline characteristics and comorbidities, as well as baseline cancer and ICI related characteristics are shown in Table 1. Compared to the propensity matched control group, there was no significant difference in baseline characteristics for those diagnosed with ICIrM based on either definition. In patients diagnosed with ICIrM the median troponin level was 364 (IQR 1096) by Bonaca et al. criteria and 670 (IQR 1181) by ESC-ICOS criteria, while in the propensity-matched control group the median troponin was 26 (IQR 39). The median values for each Bonaca et al. subgroup were as follows: 351 (IQR 1251) for definitive, 233 (IQR 1052) for probable, and 576 (IQR 1044) for possible. In patients diagnosed with ICIrM under Bonaca et al. criteria, AUC was 0.87 for a suggested troponin T cutoff of 99.5 ng/L with sensitivity of 79.7% and specificity of 85.4% to predict ICIrM. Additionally, in patients diagnosed with ICIrM under ESC-ICOS guidelines, AUC was 0.93 for the same suggested troponin T cutoff (99.5 ng/L) with sensitivity of 93.6% and specificity of 85.4% to predict ICIrM. In the ICIrM groups, all but one patient had Troponin T measure (with only one patient having Troponin I value), and all patients in the control group with available measure were with Troponin T.

Table 1 Baseline, cancer-related and ICI-related characteristics

Diagnostic variations

Of the 12 patients who had a diagnosis by Bonaca et al. but not by ESC-ICOS guidelines, one was classified as definitive, four as probable, and seven as possible ICIrM. A detailed description of the characteristics of these patients and the discrepancy for inclusion in each of the diagnostic criteria are presented in Table 2. The lack of new or significant change in troponins and availability or abnormalities on CMR were the main reasons for discrepancy in diagnosis between the two criteria. Seven (58.3%) patients died, of whom only one (8.3%) was due to a cardiac cause. Three (25.0%) patients suffered from MACE, while arrhythmias were seen in 5 (41.7%) patients.

Table 2 Causes of discrepancy between the 12 patients classified as non-myocarditis according to ESC-ICOS

Differences in severity and treatment

Fulminant ICIrM was identified in 19 patients (32.2%) from the Bonaca et al. criteria and 17 patients (36.1%) based on the ESC-ICOS guidelines, with detailed causes for fulminant classification provided in Table 3. Non-fulminant ICIrM was observed in 39 (66.1%) and 29 (61.7%) patients from the Bonaca et al. criteria and ESC-ICOS guidelines respectively. Only one patient was classified as steroid-refractory, and this was considered a positive ICIrM case by both criteria.

Table 3 Severity classification and treatment description of ICIrM patients according to Bonaca et al. criteria and ESC-ICOS guidelines

In the Bonaca et al. group, additional immunotherapy beyond steroids included intravenous immunoglobulin (IVIG), infliximab, antithymocyte globulin (ATG)-based immune suppressive therapy, and plasmapheresis, administered to 9 (15.3%), 3 (5.1%), 2 (3.4%), and 3 (5.1%) patients, respectively. According to the ESC-ICOS guidelines, 10 (21.2%) patients received additional therapy, with 8 (17.0%) patients receiving IVIG, 3 (6.4%) plasmapheresis, 3 (6.4%) infliximab, 2 (4.2%) ATG, and 1 (2.1%) mycophenolate.

Differences in clinical outcomes

In comparison to the propensity matched control group, clinical events of cardiac death, MACE, and HF were significantly increased in patients with ICIrM as diagnosed both by Bonaca et al. criteria (Table 4; Figs. 1A, 2A and 3A) and the ESC-ICOS guideline (Table 4; Figs. 1B, 2B and 3B). There was no statistical significance in all-cause mortality between patients with ICIrM (diagnosed by either of the criteria) and the control group. Arrhythmias occurrence was significantly increased in ICIrM patients according to both criteria compared to the control group (Table 4; Fig. 4).

Table 4 Long-term cardiac outcomes in patients diagnosed with ICI-related myocarditis according to Bonaca et al. criteria (top) and ESC-ICOS guideline (bottom)

Fig. 1

Cardiac death in ICIrM patients classified according to Bonaca et al. criteria (A) and ESC-ICOS guidelines (B) compared to propensity matched ICI treated patients without evidence of myocarditis

Fig. 2

MACE in ICIrM patients classified according to Bonaca et al. criteria (A) and ESC-ICOS guidelines (B) compared to propensity matched ICI treated patients without evidence of myocarditis

Fig. 3

Heart failure diagnosis in ICIrM patients classified according to Bonaca et al. criteria (A) and ESC-ICOS guidelines (B) compared to propensity matched ICI treated patients without evidence of myocarditis

Fig. 4

Arrhythmias in ICIrM patients classified according to Bonaca et al. criteria (A) and ESC-ICOS guidelines (B) compared to propensity matched ICI treated patients without evidence of myocarditis

Bonaca et al. criteria subtypes

A sensitivity analysis was conducted to evaluate potential differences in outcomes based on the subtypes classification of the Bonaca et al. criteria. KM analysis demonstrated an increased risk for cardiac death, MACE, HF, and arrhythmias across all ICIrM subcategories (definitive, probable, and possible) compared to patients who received ICI but did not develop ICIrM, with no statistically significant differences observed between each of the Bonaca subtypes.

Discussion

Clinically ICIrM is a challenging entity to diagnose, and may be associated with significant morbidity and mortality. This is the first study to compare the two most widely used diagnostic criteria for ICIrM and evaluate their association with longer-term cardiovascular outcomes, in a large cohort of patients receiving ICI therapy. Although the ESC-ICOS guidelines criteria was more restrictive and resulted in a smaller number of patients with diagnosis of ICIrM compared to Bonaca et al. criteria, this had no significant difference in prognostic value for adverse clinical sequalae when comparing the two diagnostic criteria. The main cause of discrepancy in diagnosis between the two diagnostic criteria was the lack of a rise in troponin compared to baseline, which permitted a diagnosis under Bonaca et al., but excluded a diagnosis under ESC-ICOS guidelines, as well as the absence of CMR or non-diagnostic changes on CMR.

Requirement for elevated troponin value for diagnosis with the ESC-ICOS guidelines may potentially miss ICIrM diagnosis in patients who present after the acute phase of troponin rise, and who may benefit from advanced imaging with CMR and potential therapy. The ESC-ICOS guidelines are nevertheless helpful in terms of clarity and being easy to follow, without perceived clinical ambiguity of ‘possible’ or ‘probable’ classification. The definite/probable/possible subclassifications although prognostic, did not always have a sequentially decreasing risk for cardiac outcomes; indeed, as demonstrated in Supplemental Fig. 1 the risk for cardiovascular outcomes was similar in magnitude in definitive, probable, and possible categories. As such, a simpler present/absent diagnostic criteria approach as applied in the ESC-ICOS guidelines does not appear to represent a prognostic disadvantage given the absence of a consistent sequentially prognostic correlation with the Bonaca et al. subclassifications.

Aligned with previous literature, ICIrM was rare, and observed to occur in 1.1% and 0.8% in this cohort based on Bonaca et al. and ESC-ICOS criteria respectively [11, 17,18,19,20]. Overall clinical outcomes including cardiac death, MACE, HF and arrhythmias had a significantly increased risk when patients included in each guideline separately were compared to the control group. This result is not unexpected and is again consistent with previous literature, noting that patients with ICIrM (regardless of diagnostic criteria) had worse overall cardiovascular outcomes than patients without evidence of ICIrM [19].

Although pathohistological diagnosis achieved by EMB remains the gold-standard for ICIrM diagnosis [19, 21], it is infrequently performed in real-world practice, which can be reflected in our 3 positive cases by EMB. This observation reflects the prevailing practices, where concerns about invasiveness of the procedure, the potential for complications, variability of presentation, and severity of illness may limit its application [19, 22]. Consequently, physicians often reserve this as a last resort diagnostic tool for cases where advanced non-invasive modalities, such as CMR or PET scans, yield inconclusive results [23, 24].

Despite its infrequency, ICIrM is associated with a heightened risk of mortality compared to other immune-related toxicities of ICI therapy [11, 25]. In addition to cardiovascular morbidity and mortality, incidence of ICIrM often requires withholding or ceasing of life-saving cancer therapy [26]. Hence, recognizing ICIrM early is crucial in patients undergoing ICI therapy, as prompt identification may enable timely intervention and potentially improve outcomes [13, 27, 28].

Bonaca et al. published a proposed definition of ICIrM in 2019, with the initial aim of establishing a standardized criterion for use in clinical trials of cancer immunotherapies, thereby enhancing identification of these events [14]. This classification system for ICIrM, which categorizes the condition into three probability-based groups, has also been adopted for clinical practice, and was subsequently used in ICOS documents [29]. More recently the ESC published a new diagnostic criteria for ICIrM in the 2022 ESC guidelines on cardio-oncology which was endorsed by ICOS [15]. The ESC-ICOS guideline sought to provide more clarity by categorizing the diagnosis solely based on the presence or absence of ICIrM with major and minor criteria, without including probable or possible categories.

The main difference from the Bonaca et al. criteria is that it emphasizes the necessity of troponin elevation for a positive ICIrM diagnosis. Additionally, it requires either a major criterion such as diagnostic CMR, or two or more minor criteria which include clinical symptoms, electrical disturbances on ECG, decline in LV systolic function, suggestive CMR, or other immune-related adverse events; the last criteria not considered as diagnostic by Bonaca et al. Another difference is that the ESC-ICOS guidelines do not include PET scans as a diagnostic imaging method.

In this study, the main reasons for differences in patient numbers between the two diagnostic criteria included the lack of troponin elevation from baseline, and the absence of or non-diagnostic CMR. Compared to Bonaca et al. criteria, diagnostic challenges may occur with the more restrictive ESC-ICOS guidelines in patients who do not show an increase in cardiac biomarkers from baseline but do have a diagnostic CMR or a combination of two minor diagnostic criteria. Furthermore, some patients were too unwell to undergo CMR or may not have been able to tolerate the test due to severe illness, claustrophobia, or significant renal impairment. Those who were more unwell and did undergo CMR may have had suboptimal images due to inability to tolerate the long duration of all the CMR sequences. This in turn would result in patients being considered under the Bonaca et al. criteria as having ICIrM if other criteria are met, but not under the ESC-ICOS guidelines. Both ESC and Bonaca et al. diagnostic criteria classify MRI findings as either ‘diagnostic’ or ‘suggestive’, with the standard approach for myocarditis assessment relying on the Lake Louise criteria.

Notably, our study found no significant differences in cardiovascular outcomes (cardiac mortality, MACE, HF, and arrhythmias) among patients with ICIrM when evaluated by any of the two guidelines, as both indicated an elevated risk for these outcomes compared to controls. The majority of patients (almost 80%) positive for ICIrM under Bonaca et al. criteria were also positive with ESC criteria, highlighting a significant overlap in clinical outcomes. Furthermore, the 12 patients that did not meet criteria for ICIrM per ESC-ICOS guidelines did receive steroid therapy clinically given they met criteria per the previous Bonaca et al. classification, and the immunosuppressive therapy may have attenuated outcomes and thus places a potential limitation on prognostic comparison for the ESC-ICOS guidelines.

The findings of this study may be limited by its retrospective nature, with potential for selection bias. Additionally, the diagnosis of ICIrM is challenging, with frequent elements of clinical uncertainty, and only a very few patients had an EMB to be able to compare to gold standard. A more standardized and systematic approach to the performance of CMR in patients suspected to have ICIrM would be helpful in clinical practice, however note that some in this sicker population may not be able to tolerate the examination. Furthermore, the relatively low cardiac event rate limits comparison for subcategories of the Bonaca et all criteria, but were sufficient to compare those who were diagnosed with ICIrM by the Bonaca et al. and ESC-ICOS guidelines overall.

Conclusion

The ESC-ICOS guidelines criteria was more restrictive than the Bonaca et al. criteria for the diagnosis of ICIrM, but both were similar in terms of prognostic value on retrospective review.