We extracted consecutive ICI-associated CAE cases (sporadic cases) referred to three French cardio-oncology units (Caen University Hospital; Hôpital Saint Antoine, Paris; and Aix-Marseille University NORD Hospital) between March 2015 and June 2019. Late ICI-associated CAE cases were defined as a CAE occurrence of ≥90 days from ICI therapy initiation, and early ICI-associated CAEs were defined as a CAE occurrence of <90 days from ICI therapy initiation. The 90-day cut-off period was based on several recent articles and reviews of ICI-associated CAEs describing higher rates of CAEs after the first and the third infusions of ICI.3 7 8 Some of these cases were previously published.3 The patients were managed according to similar protocols. Because no specific follow-up had previously been established for patients receiving ICIs during the study period, the oncologists referred patients receiving ICIs only on the basis of their clinical suspicion of a CAE. At the time of the CAE, these patients had a standardized evaluation including clinical evaluation, ECG, transthoracic echocardiography, and measurement of the brain natriuretic peptide and troponin serum levels. An additional biological sample was collected at the time of the CAE for the patients referred to Caen University Hospital. We described the patients with late ICI-associated CAE, including the demographic characteristics, cardiovascular risk factors and cardiovascular histories. We further reported the cancer type, ICIs used, the time of onset expressed as the calendar delay and as the total number of ICI injections before toxicity, the initial cardiovascular clinical manifestations, the type of cardiomyopathy, the associated cardiovascular diagnoses and the results of the complementary investigations. Informed consent was obtained from all the living subjects. The detailed list of collected variables is available in online supplementary material 1. The presence of a left ventricular systolic dysfunction (LVSD) was based on an echocardiographic or MRI left ventricular ejection fraction (LVEF) drop of <10% and LVEF of <50%, disregarding the presence of heart failure (HF) symptoms.9

VigiBase is the WHO global ICSR database. It contains more than 18 million reports from 130 countries. The database is fed with reports from healthcare professionals, pharmaceutical companies and, more recently, patients. All reports are referred to national pharmacovigilance centers within a country, which secondarily transmits them to VigiBase. The Uppsala Monitoring Centre (Uppsala, Sweden) is in charge of keeping VigiBase current by adding novel reports and removing duplicates. We used a definition of CAEs similar to that used in a recent work addressing CAEs in VigiBase.5 Details on the VigiBase search and CAE identification are available in online supplementary material 1. The request was performed from inception until January 31, 2019. The time of onset was computed as the calendar delay between the date of the most recent ICI therapy initiation and the date of the CAE onset. Reports where the TTO could not be calculated due to the lack of at least one of the dates were excluded. We classified cases as early cases when the TTO was less than 90 days and as late cases when the TTO was at least 90 days.

The definitions of the ICSR parameters are available in online supplementary material 1. Associated immune-related adverse events (irAEs) were defined according to the definitions used in a recent study in VigiBase.5

A dot blot was specifically developed for the detection of monoclonal anticardiac troponin I in late CAE cases. Controls were chosen among patients with lung cancer treated for more than 90 days with ICI who did not experience CAE and who had at least one cardiovascular risk factor at Caen University Hospital. Three hundred and eighty nanograms of the full-length recombinant human cardiac troponin I (Sigma-Aldrich) were spotted on a nitrocellulose membrane. When dried, the membrane was incubated with a blocking buffer (phosphate-buffered saline (PBS)/1% (m:v) gelatine) for 1 hour at room temperature, then sera obtained from control or patients (1/40 dilution in PBS/0.05% (v:v) TWEEN 20/1% (w:v) bovine serum albumin) or the monoclonal anticardiac troponin I (used as a positive control at 2 ng/μL, Abcam) was added. After 1 hour at room temperature under constant shaking, the membrane was washed three times with Tris-buffered saline/0.1% (v:v) TWEEN 20 (TBS/T) and was incubated for an additional 1 hour either with a mixture of antihuman IgG and IgM (1/1000, Southern Biotech) or antimouse IgG (1/1000, Abcam) antibodies, all conjugated to horseradish peroxidase. After three washes with TBS/T, immune reactive complexes were detected as present or absent using the western lightning chemiluminescence reagent plus (Perkin Elmer).

A descriptive analysis of the cohort study was performed on the available data. Qualitative variables were expressed as median and IQR. Qualitative variables were expressed in effectives and percentages, and Agresti-Coull 95% CIs of proportion were estimated for late CAE cases. Comparative analyses were performed between early and late CAE cases in the cohort population and between VigiBase early and late CAE reports. We did not merge sporadic cases and VigiBase reports to avoid duplicates because some of our cases were declared to the pharmacovigilance centers. The non-parametric Mann-Whitney-Wilcoxon test was used with quantitative variables, and χ² test or Fisher exact test was used when appropriate for qualitative variables. Statistical analyses were performed with R software V.3.5.3 for Windows (R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was defined as a p value of <0.05.

Figure 1 shows the case selection process. Nineteen late ICI-associated CAE cases were collected from our three French cardio-oncology units. Table 1 describes the late ICI-associated CAE cases. Most of the patients were men (15, 78.9%); the median age was 70 years (IQR 65–74). All the patients had cardiovascular risk factors, mostly tobacco use (13, 68.4%). A majority of the patients were treated for non-small cell lung cancer (17, 89.5%) with nivolumab monotherapy (16, 84.2%). One patient underwent a concomitant treatment with axitinib and recovered from LVSD while axitinib was continued and ICI was stopped. Thirteen patients had a prior platinum-based regimen; 3 patients had no prior treatment. No patients received myocardiotoxic drugs. Cardiovascular history was atrial fibrillation in four patients, ischemic heart disease in four patients, peripheral arterial disease in five patients, and transient ischemic stroke in one patient. Prior to the initiation of ICI therapy, 11 patients underwent an echocardiogram that was normal with an LVEF of >55% in all of them. The median TTO was 304 days (range 180–422). Patients with late CAE did not experience any early CAEs. The main clinical presentations were LVSD (14 cases, 73.7%) and symptomatic HF (9 cases, 47.4%). Late CAEs occurred on active ICI therapy in 13 patients, whereas 4 patients discontinued ICI therapy before the CAE (data missing for 2 patients). One patient presenting with HF and LVSD after 17 cycles of nivolumab had an endomyocardial biopsy. The myocardial biopsy was normal; there was no inflammatory stigmata (online supplementary figure 1). This patient had a complete LVEF recovery in 4 months with beta blockers, conversion enzyme inhibitors and ivabradine. MRI was performed in 14 patients and did not show evidence of an inflammatory process in 13 (92.9%), without late gadolinium enhancement or myocardial edema. High-dose corticosteroids were commonly used in patients with late CAE (10 patients, 52.6%). Two patients were already treated with oral corticosteroids when the CAE occurred. A second patient had improvement of his LVEF of >50% at 2 months under beta blockers and ACE inhibitor therapy. This patient was alive 10 months after the non-inflammatory LVSD diagnosis. Fatality was attributed to a cardiovascular disease in two patients (10.5%) and to malignancy in six patients (31.6%).

Figure 2 and table 1 show the comparison between early (n=19) and late (n=19) ICI-associated CAE cases from the French cardio-oncology unit database. The median TTO was 14 days (IQR 8–62) in the early CAE cases and 304 days (IQR 180–422) in the late CAE cases. Compared with the early CAE cases, the late CAE cases had significantly more cardiovascular risk factors (60% vs 100%, respectively; p=0.01), significantly higher rates of HF (5.3% vs 47.4%, respectively; p=0.01) and LVSD (36.8% vs 73.7%, respectively; p=0.049), and a significantly lower supraventricular arrhythmia rate (47.4% vs 10.5%, respectively; p=0.03). Recovery from LVSD was not different between early and late CAE cases (37.5% vs 14.3%, p=0.47). The cardiovascular-related mortality and overall mortality rates were not different between the early and late CAE cases.

Figure 1 shows the flow diagram of the case selection process for VigiBase. Of the 2216 reports, 596 had an available CAE TTO and were included in the analysis. There were 437 (73.3%) early ICI-associated CAE cases (any early event) and 159 (26.7%) late ICI-associated CAE cases (any late event). The median TTO was 21 days (IQR 8–44) in the early CAE cases and 177 days (IQR 128–317) in the late CAE cases. Seventeen percent of the late cases had a TTO longer than 1 year. Table 2 summarizes the characteristics of the included reports. Most of the patients were men (68.4%), and the completeness score of the selected cases was above 0.65 in 75% of the cases. The seriousness was available in all cases, and all cases were considered serious. The main drug indications were lung cancer (52.8%), melanoma (26.0%) and renal cancer (11.5%). A majority of the patients were treated with antiprogrammed death 1 (PD-1) therapy or anti-programmed death ligand 1 therapy alone (84.2%). Combination therapy with two ICIs was used in 46 (7.7%) patients. A comparison between the early and late CAE cases showed significantly higher rates of HF (21.1% vs 31.4%, respectively; p=0.012) and pericardial disorders (11.4% vs 18.2%, respectively; p=0.043) in the late CAE cases. Mortality rates were comparable between the early and late CAE cases (30.0% vs 27.0%, respectively; p=0.55).

Out of four late CAE cases, two (50%) exhibited the presence of IgG/IgM antibodies against cardiac troponin I compared with zero of four control patients (online supplementary table 1 and supplementary figure 2). One late CAE case with positive autoantibody detection had a history of silent myocardial infarction diagnosed 4 years before ICI initiation. The four control patients were treated with ICI anti-PD-1 monotherapy for more than 90 days for lung cancer and did not develop any CAE. These four control patients all had cardiovascular risk factors and were aged 62–75 years. Detection of IgG/IgM antibodies against cardiac troponin I before ICI initiation could not be performed.