Spectrum of cardiovascular toxicities of immune checkpoint inhibitors: A pharmacovigilance study

CV-irAE over-reporting signal detection using the World health organization (WHO) pharmacovigilance database of individual case safety reports (ICSRs), VigiBase^[7]

CV-irAE signal detected using WHO’s global database of ICSRs

Given the considerable overlap of symptoms for various CV disease etiologies as well as redundancies in reporting cardiac complications associated with oncologic therapies, we broadly categorized cardiac and vascular disease entities in the MedDRA Classification (Version 20.1) based on underlying pathophysiology (Appendix page 1). CV-AE reporting was compared to total AE (rate) in ICI subgroup (n: 31,321) versus the entire database (n: 16,343,451 for IC calculation, and n: 12,455,401 for ROR calculation). All patients were included for these analysis. IC calculation was based on all drugs and all events in VigiBase (from November 14, 1967 to January 02, 2018, as validated elsewhere),^8,9 while ROR calculation was applied to all drugs and all events after the year of first AE report associated with ICI in VigiBase (from January 01, 2008 to January 02, 2018). Details concerning number of CV-irAE by different immunotherapy regimen and CV grouping categories are available in Tables 1 and 2. Using this agnostic approach, we identified four broad CV entities where CV reporting was significantly increased after ICI treatment as compared to the entire database. ICI treatment was associated with higher reporting of myocarditis (ROR: 11.21 [9.36–13.43], IC₀₂₅: 3.2, Figure 1), pericardial diseases (ROR: 3.8 [3.08–4.62], IC₀₂₅: 1.63, Figure 1), supraventricular arrhythmias (ROR: 1.72 [1.51–1.97], IC₀₂₅: 0.56, Figure 1) and vasculitis (ROR: 1.56 [1.25–1.94], IC₀₂₅: 0.03). Further interrogation of vasculitis reporting revealed that this signal was driven primarily by temporal arteritis (ROR: 12.99 [8.12–20.77], IC₀₂₅: 2.59, Figure 1) and polymyalgia rheumatica (ROR: 5.13 [3.13–8.40], IC₀₂₅: 1.33, Figure 1). Other cardiac disease conditions, including ischemia, heart failure, and valvular disorders were not over-reported in this population (Table 1). IC values and their 95% credibility interval over time for myocarditis, pericardial disorders, temporal arteritis and polymyalgia rheumatica are represented in Figure 1.

Reported supraventricular arrhythmias on ICI (n: 222, Appendix page 3) were overwhelmingly associated with other concurrent irAE such as gastrointestinal disorders (93 (41.9%) of 222; mainly colitis and diarrhea leading to dehydration and electrolyte disorders), other cardiac conditions (69 (31.1%) of 222, mainly cardiac dysfunction, cardiac ischemia and pericardial disorders), endocrine disorders (66 (29.7%) of 222; mainly thyroid abnormalities), and neurologic disorders (28 (12.6%) of 222, including strokes and encephalitis). Since supraventricular arrhythmias are common in the cancer population and usually co-occurred with other irAE, we suspect that they were secondary to concurrent irAE complications rather than ICI treatment itself. Thus, we focused on the other CV-irAE (myocarditis, pericardial diseases and vasculitis) to determine if there was an increased signal for toxicity with various ICI regimens (Table 2) as well as to describe clinical characteristics associated with each toxicity.

We asked whether each CV-irAE was more common with anti-CTLA-4 monotherapy compared to anti-PD-1/PD-L1 monotherapy or combination ICI therapy compared to monotherapy (Table 2). ICI-associated myocarditis was over-reported for anti-PD-1/PDL1 vs. anti-CTLA-4 monotherapy (ROR: 5.62 [2.46–12.88]) and with combination vs. mono immunotherapy (ROR: 4.31 [2.86–6.38]). ICI-associated pericardial disorders were over-reported for anti-PD-1/PD-L1 vs. anti-CTLA-4 monotherapy (ROR: 2.28 [1.27–4.12]). On the other hand, temporal arteritis was over-reported for anti-CTLA-4 monotherapy versus anti-PD-1/PD-L1 monotherapy (ROR: 3.57 [1.34–9.33]).

Characteristics of patients, timing, and outcomes of CV-irAE.

Given that ICI-associated toxicities are novel clinical entities and represent new challenges for the clinician, we extracted further data to better characterize clinical characteristics of ICI-associated myocarditis (n: 122), pericardial disorders (n: 95), and vasculitis (n: 82 which included polymyalgia rheumatica n: 16 and temporal arteritis n: 18) (Tables 3 and 4). Cases of myocarditis, pericardial disorders and vasculitis were rarely overlapping (Figure 2). Diagnosis of each was made from across the globe, mainly by healthcare professionals with an increasing rate over time with most of the cases reported in 2017 and 2018. CV-irAE affected mostly men (58–67%), with a wide age range (20–90 years old). All cancer types where ICI have been utilized were affected, with melanoma, lung cancer and renal cell carcinoma being the most common (Tables 3–4). Pericardial disorders were more likely to be reported in patients with lung cancer whereas myocarditis and vasculitis were more commonly observed in patients with melanoma (p<0.001, Tables 3–4). These CV-irAE occurred early after ICI administration, as soon as after the first ICI dose, with a median time to onset of 30 days (IQR: 18–60 days) for myocarditis, 30 days (IQR: 8.5–90 days) for pericardial disorders and 55 days (IQR: 21–98 days) for vasculitis. There was no significant difference in time to onset between these CV-irAE (Figure 2, p: 0.16). CV-irAE were serious in the majority of cases (>80%), with fatalities occurring in 61 (50%) of 122 myocarditis cases, 20 (21.1%) of 95 pericardial disorders and 5 (6.1%) of 82 vasculitis (Chi-2 test for fatality rates between these latter conditions; p<0.0001). Among myocarditis cases, there were 21 deaths among 32 ICI combination therapy (65.6%) compared to 40 (44.4%) of 90 ICI monotherapy (p: 0.04,). Myocarditis was frequently associated with myositis, cardiac arrhythmias, heart failure, myasthenia gravis-like symptoms, pneumonitis and rarely pericardial disorders. Pericardial disorders were frequently associated with pulmonary disorders (mainly pleural effusions and pneumonitis) and thyroid dysfunction. Vasculitis was frequently associated with concurrent dermatologic and rheumatologic conditions (mainly arthritis) and ophthalmologic conditions (leading to impaired vision in 5 (27.8%) of 18 temporal arteritis cases). More specific details and proportion concerning characteristics of concurrent irAE associated with these CV-irAE and of ICI associated temporal arteritis and polymyalgia rheumatica are provided in Table 4.

Limitations

Several limitations need to be recognized for VigiBase analysis. Pharmacovigilance analysis allow for signal detection and generate hypothesis which need to be replicated ideally by prospective studies. Replication of disproportionality analysis in other smaller size continental pharmacovigilance databases such as European Medicine Agency or US. Food and Drug Administration AE Reporting system (FAERS) is biased by an important ICSRs overlap between databases. Alternatively, some cases of suspected drug-induced CV-irAEs are likely not reported to the national drug authorities, and therefore not submitted to VigiBase. However, a major strength is that VigiBase aggregates ICSRs collected from over 130 countries, which enables better identification of these rare AE and broader generalization of our findings. Level of causality attributed to ICI was high based on extrinsic data, chronology, symptomatology and known ICI mechanisms of action. Another limitation of Vigibase is that sources of reports are non-homogeneous, and there is limited possibility of verification of the clinical, laboratory tests or radiological findings justifying the reported diagnosis, nor completeness of reporting for age, drug dosing, time to onset, comorbid conditions and concomitant drugs. The exact denominator of patients exposed to the different ICI regimen cannot be evaluated. Instead, total number of ICSRs for each group of drug is used as denominator for this kind of disproportionality analysis in pharmacovigilance databases for signal detection.^10,11 The volume of reports for a particular medicinal product may be influenced by the extent of use of the product, publicity, the nature of the reactions and other factors such as competition bias. The value of disproportionality reporting (as the ROR) for several CV-AE and culprit drugs has already been demonstrated in various settings; nevertheless, there is still a risk that comparisons of disproportionality between medicinal products in pharmacovigilance databases may be misleading. As an example of another validated association, over-reporting for long-QT and Torsade-de-Pointes has been associated with the extent of IKr drug-induced blockade using VigiBase;¹⁰ or with identification of over-reporting for long-QT and Torsade-de-Pointes on selective-estrogen receptor modulators vs. aromatase inhibitors in women using European pharmacovigilance database.^[12]

Clinical trials are mandatory to establish efficacy but may not allow definitive conclusions on drug safety in part due to selected populations and limited power to detect balances in rare adverse events. Spontaneous notifications remain the cornerstone for ADR evaluation despite their limitations. Disproportionality analysis in pharmacovigilance databases is an important method to detect signals in drug safety research and post-marketing surveillance. Herein, we use this analysis to identify several cardiovascular complications, specifically myocarditis, pericardial disorders, and vasculitis, associated with immune checkpoint inhibitors. These types of analyses may identify other uncommon but clinically-significant complications.

While the incidence of these CV-irAE events cannot be determined using VigiBase, we can provide the following data. First, objective responses occurred in 15–60% of patients with metastatic solid cancers treated with anti-PD-1/PD-L1 agents (including 10–40% durable responses), depending on the cancer type.^23,24 Second, initial estimates suggested a rate of myocarditis ranging from 0.06% to 0.3%,⁸ although more recent studies in the setting of enhanced awareness have demonstrated up to 1.1% incidence.¹³ Pericardial and vascular disorders have been sporadically reported and thus the incidence is essentially unknown. Third, the risk of ICI induced fatal toxicities has not been studied in detail although our group have shown fatality rates ranging from 0.3% (anti-PD-1/PD-L1 monotherapy) to 1.2% (combination PD-1/CTLA-4 blockade); 17% of these fatalities were cardiac in nature (unpublished data; authors).