Immune Checkpoint Inhibitors and Cardiotoxicity: An Analysis of Spontaneous Reports in Eudravigilance

Data on ICSRs with an ICI as the suspected drug were retrieved from the website of suspected ADRs (www.​adrreports.​eu) of the European pharmacovigilance database (Eudravigilance, EV). The EV is managed by the European Medicines Agency (EMA). It is a system used for the management and analyses of ICSRs related to both medicines or vaccines, which are authorized or are being studied in clinical trials in the European Economic Area. The EV contains all ICSRs reported by a healthcare professional or a non-healthcare professional to an European Union national competent authority or a marketing authorization holder. These data are publicly available for transparency through the EMA website (www.​adrreports.​eu).

Information on patient characteristics (age and sex), adverse event (outcome and seriousness), therapeutic indication, primary source qualification, primary source country for regulatory purposes, number of suspected drugs other than ICIs, and number of concomitant drugs was provided for all ICSRs and separately for each ICI. Individual case safety reports with two or more ICIs as suspected drugs were described separately.

The duration of therapy (reported in days) was retrieved from each ICSR, where available. Computations on the duration of therapy were performed for each ICI, including also data obtained from ICSRs with more ICIs as suspected drugs.

Based on our data source, we have retrieved the general information of seriousness stated as “serious” or “not serious”. In accordance with the International Council on Harmonization E2D guidelines, a case is defined as “serious” if it is life threatening, results in death, requires or prolongs a hospitalization, results in persistent or significant disability/incapacity, determines a congenital anomaly/birth defect, or results in some other clinically important conditions. Therefore, if a case is classified as serious in our data source, it means that the case matches one of the aforementioned criteria. However, the specific criteria of seriousness for all ICSRs were not retrievable.

The outcome of a cardiac ADR was classified as “Recovered/Resolved”, “Recovering/Resolving”, “Recovered/Resolved with Sequelae”, “Not Recovered/Not Resolved”, “Fatal”, and “Unknown”. The outcome with the lower level of resolution was chosen for classification whether an ICSR reported two or more cardiac adverse events with different outcomes. Individual case safety reports were classified as fatal if death occurred.

Cardiac adverse events were tabled for the 20 most reported events in the main table and for all others in the Electronic Supplementary Material (ESM). The time to event (days) was calculated only for ICSRs that reported both the duration of the therapy and the drug withdrawal as action taken after the occurrence of the ADR. Computations on the time to event were performed for each ICI, including also data obtained from ICSRs with more ICIs as suspected drugs. A boxplot of time to event was generated for the most reported ICIs.

To estimate the quota of ICSRs related to patients with pre-existing cardiovascular and metabolic diseases, we evaluated the concomitant presence of cardiovascular and metabolic agents reported as other suspected or concomitant drugs. We listed the most reported drug classes of concomitant treatments (ICSR numbers ≥ 60) in a main table and all others in the ESM. Boxplots were performed using R (Version 3.2.2; R Development Core Team).

The reporting odds ratio (ROR) and its 95% confidence interval (95% CI) were computed to assess the frequency of reporting ICSRs with ADRs belonging to the SOC “Cardiac disorders” for each ICI (ICI of interest) compared to all other ICIs. The reference group includes all ICIs excluding the one of interest. Moreover, because ICIs can be classified for their mechanism of action in anti-PD-1 (nivolumab, pembrolizumab, and cemiplimab), anti-PD-L1 (atezolizumab, durvalumab, and avelumab), and anti-CTLA-4 (ipilimumab), the ROR was also computed within the same drug class, using all other anti-PD-1 or all other anti-PD-L1 as reference groups, and for comparing anti-CTLA-4 with anti-PD-1/anti-PD-L1. The RORs were computed on ICSR numbers as these are publicly available on the EMA website (www.​adrreports.​eu). Specifically, ICSR numbers were retrieved from the section “Number of individual cases” (www.​adrreports.​eu) for the period up to 18 March, 2020. Forest plots were performed for both comparisons using R (Version 3.2.2; R Development Core Team).

During the study period, 2478 ICSRs with at least one ICI as the suspected drug were retrieved from the EV, of which 249 (10%) reported more than one ICI as suspected drugs. Specifically, 1071 (43.2%) ICSRs pertained to nivolumab, 805 (32.5%) to pembrolizumab, 234 (9.4%) to the association of nivolumab/ipilimumab, 143 (5.8%) to ipilimumab, 105 (4.2%) to atezolizumab, 78 (3.1%) to durvalumab, 23 (0.9%) to avelumab, four (0.2%) to cemiplimab, four (0.2%) to the association of nivolumab/pembrolizumab, three (0.1%) to nivolumab and atezolizumab, three (0.1%) to ipilimumab and pembrolizumab, three (0.1%) to atezolizumab and pembrolizumab, and two (0.1%) to nivolumab, ipilimumab, and pembrolizumab. The median age of patients was 69 years (interquartile range [IQR] 60.25–75) for all ICSRs (Table 1), remaining similar for all ICI groups except for ICSRs with the association of ipilimumab/pembrolizumab and atezolizumab/pembrolizumab, in which the median age was 34.5 years (IQR 32.2–36.7) and 49 years (IQR 49–49), respectively (Table 2). Most ICSRs were observed in male patients, and were reported from healthcare professionals (Tables 1, 2). Of all ICSRs, 99.4% were classified as serious. The outcome of cardiac ADRs was “recovered/resolved” in 14.3% of ICSRs, “recovering/resolving” in 10.9% of ICSRs, and “Fatal” in 30.1% of ICSRs (Table 1). In the majority of ICSRs, the ICI was the only suspected drug reported (n = 2179; 87.9%), or no concomitant drug was reported (n = 1650; 66.6%). Characteristics of cases for all ICSR groups were presented in Tables 1 and 2.

The main therapeutic indications for all ICIs used as single treatments were lung, genitourinary, and skin cancers (Table 1 of the ESM). For ICSRs reporting more than one ICI as reported in Table 1 of the ESM, the main therapeutic indication for nivolumab/ipilimumab was skin cancer (N = 106; 44.9%), followed by genitourinary cancer (N = 88; 37.3%). The median duration of therapy was 43 days for atezolizumab (IQR 2.7–69.2) and durvalumab (IQR 15–88), 30 days (IQR 1–99) for nivolumab, and 22 days for ipilimumab (IQR 1–63.5) and pembrolizumab (IQR 1–66). Because of the scant data available, the median duration of therapy was not computed for avelumab and cemiplimab.

We observed a total of 3078 cardiac ADRs (1.2 cardiac ADRs per ICSR), as more than one ADR could be reported in each ICSR. A total of 1310 (42.6%) cardiac ADRs were reported with nivolumab, 1046 (34.0%) with pembrolizumab, 287 (9.3%) with nivolumab/ipilimumab, 162 (5.3%) with ipilimumab, 127 (4.1%) with atezolizumab, 96 (3.1%) with durvalumab, 26 (0.8%) with avelumab, five (0.2%) with cemiplimab, five (0.2%) with atezolizumab/pembrolizumab, four (0.1%) with nivolumab/pembrolizumab, four (0.1%) with nivolumab/ipilimumab/pembrolizumab, three (0.1%) with ipilimumab/pembrolizumab, and three (0.1%) with nivolumab/atezolizumab. Overall, the most reported cardiac adverse events were myocarditis, cardiac failure, atrial fibrillation, pericardial effusion, and myocardial infarction. Numbers and percentages of cardiac adverse events are reported in Tables 3 and 4, and Table 2 of the ESM. The number of ICSRs for which we were able to calculate the time to event were: 188 for nivolumab, 144 for pembrolizumab, 39 for ipilimumab, 18 for durvalumab, nine for atezolizumab, two for avelumab, and one for cemiplimab. The median time to event was 36 days for atezolizumab (IQR 2–57) and durvalumab (IQR 15–81.75), 31 days (IQR 1.75–96.75) for nivolumab, and 22 days for pembrolizumab (IQR 1–64) and ipilimumab (IQR 1–36.5). Because of the scant data, the time to event was not computed for avelumab and cemiplimab. Figure 1 shows the boxplots of times to event for the most reported ICIs.

In 57 (2.3%) out of 2478 ICSRs, cardiovascular and metabolic agents were reported among suspected drugs. Specifically, 25 out of 57 ICSRs (43.9%) were related to nivolumab, 18 (31.6%) to pembrolizumab, four (7.0%) to atezolizumab, four (7.0%) to the association of nivolumab/ipilimumab, three (5.3%) to ipilimumab, two (3.5%) to durvalumab, and one (1.7%) to avelumab. A total of 78 cardiovascular and metabolic agents were identified in the ICSRs. The most reported were calcium channel blockers (ten; 12.8%), new oral anticoagulants (ten; 12.8%), statins (nine; 11.5%), beta-blockers (nine; 11.5%), and loop diuretics (seven; 9.0%). Cardiovascular and metabolic agents reported as other suspected drugs are shown in Table 5.

In 543 (21.9%) out of 2478 ICSRs, cardiovascular and metabolic agents were reported among concomitant drugs. Specifically, 259 out of 543 ICSRs (47.7%) were referred to nivolumab, 176 (32.4%) to pembrolizumab, 33 (6.1%) to ipilimumab, 31 (5.7%) to the association of nivolumab/ipilimumab, 21 (3.9%) to durvalumab, nine (1.6%) to atezolizumab, seven (1.3%) to avelumab, three (0.5%) to the association of nivolumab/pembrolizumab, two (0.4%) to the association of nivolumab/ipilimumab/pembrolizumab, one (0.2%) to the association of nivolumab/atezolizumab, and one (0.2%) to cemiplimab. A total of 1330 concomitant drugs were identified in the ICSRs. The most reported were beta-blockers (172; 12.9%), calcium channel blockers (145; 10.9%), antiplatelet agents (142; 10.7%), statins (139; 10.5%), and angiotensin receptor blockers (92; 6.9%) (Table 6). Details of cardiovascular and metabolic agents reported as concomitant drugs are shown in Table 3 of the ESM.

Compared to all other ICIs, nivolumab reported a very small increased reporting frequency of ICSRs with cardiac ADRs (ROR 1.09, 95% CI 1.01–1.18). Moreover, anti-CTLA-4 has shown a lower reporting frequency of ICSRs with cardiac ADRs compared to anti-PD-1 (ROR 0.75, 95% CI 0.67–0. 84) and anti-PD-L1 (ROR 0.79, 95% CI 0.66–0.93). All computations are reported in Figs. 2, 3, 4. The numbers of ICI reports used for the ROR computations are shown in Table 4 of the ESM.

The present study evaluated the reporting of cardiac ICSRs with ICIs using data from the EV database. Data reported in the EV database concerns suspected side effects that have been observed following the use of a medicine, but that are not necessarily related to or caused by the medicine. We found that the main suspected ICIs were nivolumab and pembrolizumab and, overall, the main cardiac events were myocarditis, cardiac failure, atrial fibrillation, pericardial effusion, and myocardial infarction. Ninety-nine percent of cardiac ADRs was classified as serious and 30.1% had a fatal outcome. Among ICIs, nivolumab reported a small increased reporting frequency of ICSRs with cardiac ADRs compared to all other ICIs.

In our study, more than 76% of all ICSRs (n = 1876) were related to nivolumab and pembrolizumab, while the remaining ICSRs reported as suspected other ICIs or their associations. In our opinion, the higher number of ICSRs reported with nivolumab and pembrolizumab might depend on their earlier marketing approval. Indeed, together with ipilimumab, which was the first authorized CTLA-4 inhibitor, nivolumab and pembrolizumab were the first PD-Lis to be approved [9].

In the present analysis, the majority of ADRs with ICIs was reported in male patients with a median age of 69 years. For all ICIs, the main therapeutic indications were lung, genitourinary, and skin cancers. These data are not surprising if we consider that ICIs are mainly indicated for the treatment of melanoma, renal cell carcinoma, bladder cancer, head and neck cancers, and lung cancers [10]. These conditions show higher prevalences in male adult patients. For instance, melanoma incidence reaches a peak at the seventh and eighth decades of life and, after the age of 40 years, it affects more men than women [11]. Similar trends and risk factors were observed for other cancers, including renal cell carcinoma [12] and lung cancers [13]. Last, the higher proportion of ADRs in male patients was previously found for all ICIs in a pharmacovigilance study that used data from the Italian spontaneous reporting system [14]. We also found that combinations of ipilimumab/pembrolizumab and atezolizumab/pembrolizumab were commonly reported in patients with a median age of 34.5 years and 49 years, respectively. These combinations are increasingly used in patients with metastatic melanoma, advanced renal cell carcinoma, and colorectal and lung cancer owing to their effect in improving the response rates and median survival time [15].

Most of the cases were serious. Taking into account that ICIs are monoclonal antibodies, these results are not surprising. Monoclonal antibodies have longer terminal half-lives and consequently even a single dose of these drugs leads to prolonged systemic exposure that increases the risk of serious adverse events [16]. Moreover, in the evaluation of seriousness, we should also consider the reporting bias of serious ADRs due to mandatory reporting obligations by regulatory authorities and that healthcare professionals are more prone to report serious than non-serious ADRs.

Almost 30% of ICSRs reported ADRs with a fatal outcome, while approximately 25% reported ADRs with a favorable outcome. It is well known that ICIs are able to induce many types of irADRs affecting the colon, liver, lungs, thyroid, skin, heart, and nervous system. Even though irADRs are usually manageable with corticosteroids or immune modulator agents, uncommon fatal events have been reported [17]. However, an important consideration that should be stated in regard to suspected adverse events retrieved from spontaneous reporting systems, with greater attention for those that are lethal, is that we cannot be sure of their causal relationship with a drug. Indeed, in the spontaneous reporting system are reported adverse events observed during a drug treatment but that are not obligatorily related to it. Therefore, their onset and severity could depend on other factors such as the frailty of the population, the nature of their underlying illness, or even to the specific reporting practices of those sending the ICSRs for this type of patient. In this study, the median duration of ICI therapy had a range of 22–43 days. In our opinion, many reasons could explain these data, including the withdrawal of the drug as an action taken after the occurrence of cardiac ADRs, the seriousness of some cancer diseases, such as lung cancer, which are associated with a high mortality rate in the short term, and last, the lack of a follow-up for ICSRs retrieved from the EV database.

As previously reported, in our study, the main identified cardiac events were myocarditis, cardiac failure, atrial fibrillation, pericardial effusion, and myocardial infarction. Many hypotheses have been formulated with regard to the association between ICIs and cardiotoxicity. Indeed, Freeman et al. reported that PD-L1 might regulate autoreactive lymphocytes playing a role in limiting activities of T cells in the heart, where PD-L1 is highly expressed [37]. It was also demonstrated that the disruption of the gene encoding for PD-1 in mice caused dilated cardiomyopathy [18]. In addition, hyperactivated CD4+/CD8+ T lymphocytes and macrophages could be microscopically found in the heart tissue of patients with ICI-related cardiac toxicity. One of the most likely explanations underlying the occurrence of this toxicity is the “shared antigen” hypothesis between the tumor and cardiac muscle, which could be related to muscle-specific antigens, desmin and troponin, detected in the tumor. Thus, a response to cardiac antigens might contribute to heart failure through an autoantibody-independent mechanism [19]. Among patients treated with ICIs, up to 60–80% experience at least one irADR and the incidence of cardiac irADRs has a range of 1.14–5%.

In line with literature data, in our study, myocarditis was the most commonly reported cardiac ADR. However, other signs and symptoms of cardiotoxicity could occur, including arrhythmias, atrial fibrillation, conduction disorders, pericardial disease, and vasculitis [20]. All these ADRs were found to be possibly related to ICIs in our study too. In our study, the majority of cardiac events were reported for nivolumab, followed by pembrolizumab, ipilimumab, and atezolizumab. In terms of the distribution of cardiac ADRs by suspected ICIs, some differences were found. Indeed, in line with literature data suggesting that the combination ICI therapy (e.g., a CTLA-4 inhibitor combined with a PD-1 inhibitor) is the predominant risk factor for myocarditis and that pericardial diseases occur more frequently in patients receiving anti-PD-1 or anti-PDL-1 antibodies than those treated with anti-CTLA-4 [21], we found that almost 30% of cardiac ADRs related to the combination nivolumab/ipilimumab concerned cases of myocarditis, while ADRs affecting the pericardium were rarely reported with ipilimumab alone compared to other ICIs.

In terms of the disproportionality analysis, we observed a very small increased reporting of cardiac events with nivolumab (9% higher frequency of reporting ICSRs than the other ICI). This result may reflect the higher number/proportion of ICSRs on nivolumab present in the database, considering also that nivolumab is approved for diverse cancer types and has been in use for a longer time compared to anti-PD-L1 ICIs. Among ICIs within the same drug class (anti-PD-1 or anti-PD-L1), no difference for the reporting frequency of cardiac ICSRs was found, while anti-CTLA-4 had a lower reporting of cardiac ICSRs compared with anti-PD-1 and anti-PD-L1.

On the contrary, literature data suggest that cardiac irADRs seem to be more common in patients treated with the CTLA-4 inhibitor compared with those receiving anti-PD-1. The risk increases with combination therapy, leading to a discontinuation rate in up to 50% of treated patients [22]. It is important to highlight that we performed a disproportionality analysis with RORs restricted to the ICI drug class for assessing the reporting frequency of cardiac ICSRs. It should be noted that this is an approach used in pharmacovigilance to assess the reporting frequency of ICSRs within a drug class [14, 23, 24]. However, this type of analysis is not useful to assess a class effect on toxicity but other studies with different methodologies are needed on this front.

However, in addition to studies describing cases of myocardial fibrosis, pericarditis, and Takotsubo cardiomyopathy with ipilimumab [25‐27], many clinical studies highlighted the occurrence of disparate cardiac ADRs during nivolumab and pembrolizumab treatments [28, 29]. Acute coronary syndrome, new‐onset heart failure, and chronic heart failure have been also reported [29‐31]. Interestingly, we also observed an increased reporting of ICI-related myocardial infarction, which underlying pathophysiology is currently unknown. However, three possible mechanisms have been hypothesized. First, the activation of ICI-related inflammation may induce the disruption of atherosclerotic coronary plaques and lead to acute myocardial infarction. Second, ICIs may cause coronary spasm causing ST elevation, which may also be related to the systemic inflammation induced by ICIs. Third, ICIs may directly activate the T-cell-mediated coronary vasculitis in the absence of atherosclerosis [32]. Further studies are needed to support these interesting pathophysiological hypotheses. Last, in our study, cardiovascular and metabolic agents (including calcium channel blockers, new oral anticoagulants, statins, beta-blockers, and loop diuretics) were reported in 2.3% and 21.4% of all ICSRs among suspected and concomitant drugs, respectively. As these drugs are used in patients with cardiovascular diseases, the role of pre-existing diseases affecting the cardiovascular system on the occurrence of cardiac ADRs could not be excluded. Indeed, a retrospective matched case-control study, evaluating 108 cases of ICI-related myocarditis and 108 controls of ICI-related ADRs other than myocarditis obtained from VigiBase, found that drugs labeled as treatment for cardiovascular conditions were more strongly reported with ICI-related myocarditis than other ICI-related ADRs. In this study, compared to our result, a slightly increased percentage (27.8%) of drugs belonging to the ATC “Cardiovascular system” was found among cases of ICI-related myocarditis. This higher percentage may be related to the different study design and to the different sample size [33]. In addition, patients diagnosed with cancer are more likely to die from cardiovascular diseases. Given this possible association, a closer cooperation between cardiologists and oncologists is strongly recommended [34].

In our study, the median time to event had a range of 22–36 days, suggesting the early occurrence of cardiotoxicity. In line with our results, literature data reported that the time to event for ICI-induced cardiac ADRs ranges from 12 days to 3 months [35].