Myocarditis in Patients Starting Combination Checkpoint Inhibitor Therapy: Analysis of a Commercial Claims Database

Authors: Alistair C. Lindsay, Alexander M. Walker, Sebastian Schneeweiss

Abstract

Background

Immune checkpoint inhibitors have improved the clinical outcomes of several cancers but have also been associated with a greater risk of immune‐related adverse effects, especially when combined. The objective of this study was to investigate the incidence of myocarditis in relation to the use of dual concurrent versus single immune checkpoint inhibitors therapies.

Methods and Results

A cohort study was conducted using medical and pharmacy claims data (2011–2022) from a large US commercial insurer. Cox regression quantified the comparative risks of myocarditis or heart failure in patients with cancer receiving treatment with combination therapy (nivolumab and ipilimumab) in comparison to taking a single immune checkpoint inhibitor only. Mean follow‐up time in 53 018 patients was 226 days (interquartile range, 93–495 days). There were 148 cases of myocarditis (0.3%), 33 (0.7%) in patients on combination therapy, and 115 (0.2%) in patients on monotherapy. The risk of myocarditis per 1000 patients was 7.40 in the combination therapy group and 2.37 in the monotherapy group (risk ratio, 3.12 [95% CI, 2.12–4.60]). Using multivariable regression analysis, the hazard ratio for myocarditis in the combination therapy group was 2.38 (1.57–3.63). No difference in the risk of heart failure was found between combination and single therapy.

Conclusions

Therapy with 2 immune checkpoint inhibitors was associated with an increased risk of myocarditis compared with monotherapy, with most cases occurring in the first 6 months of therapy.

Clinical PerspectiveWhat Is New? Previous reports have suggested that patients with cancer taking combination checkpoint inhibitor therapy are at greater risk of myocarditis than those receiving single‐agent therapy.In this insurance database study of >53 000 patients receiving checkpoint inhibitors, those receiving a certain combination therapy were found to be at an increased risk of myocarditis (risk ratio, 3.12 [95% CI, 2.12–4.60]). What Are the Clinical Implications? Patients receiving combination checkpoint inhibitor therapy are at greater risk of myocarditis early in their treatment; increased screening for this complication may be warranted.

Immuno‐oncology drugs boost the ability of the immune system to attack tumor cells. Prominent members of the class are immune checkpoint inhibitors (ICIs), which block inhibitory proteins on the surfaces of T cells and target cells and, in doing so, activate cytotoxic T cells. ^1^ , ^2^

First introduced in 2011, several ICIs have entered clinical practice, and the use of these drugs has expanded to 24 different types of cancer. ICIs have revolutionized the treatment of these cancers by achieving remarkable improvements in overall survival rates and long‐term disease control.

Owing to the aberrant activation of autoimmune T cells, therapeutic blockade of immune checkpoints can also precipitate immune‐related adverse events, ^3^ , ^4^ 40% of which lead to treatment cessation. ^5^ Since their release, ICIs have been associated with serious immune‐related cardiovascular adverse events, ^6^ most prominently myocarditis, ^7^ in case studies and case series. ^8^ , ^9^ The simultaneous use of 2 different ICIs has been linked to a greater risk of such complications. ^10^ Still, risk estimates have varied due to the relative rarity of this complication. ^11^ The incidence of early and late cardiovascular toxicity with ICIs (alone and in combination) remains poorly described. ^5^

This pharmacoepidemiology study aimed to quantify the frequency of, and clinical outcomes from, myocarditis in patients receiving ICI, comparing combination therapy to monotherapy. The venue was a large longitudinal health insurance claims database.

METHODS

Data Source

The data that support the findings of this study are available from the corresponding author upon reasonable request. The Brigham and Women's Hospital Institutional Review Board approved this study on November 16, 2022 (Ref: 2022P002937). De‐identified data were used to ensure no individuals involved could be identified, obviating the need for consent according to the Common Rule. Data were extracted from the Clinformatics Datamart (OptumInsight, Eden Prairie, MN), a health care insurance claims dataset based in the United States with an average annual enrollment of >14 million people. At the beginning of the study, demographic information, health plan enrollment status, and inpatient and outpatient medical encounters were retrieved using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD‐9‐CM and ICD‐10 codes after October 1, 2015. CPT‐4 (Current Procedural Terminology, fourth edition) codes and filled prescriptions (including National Drug Code numbers) were used to determine drug treatments (Data S1).

Study Population

Patients entered the study cohort on their first use of any of 7 ICIs (see Exposure Definition section below). The cohort was limited to individuals with 6 or more months of continuous enrollment before drug initiation. Baseline patient characteristics, including demographics and comorbidities, were measured in the 180 days preceding and on the date of entry to the cohort and are shown in Table 1. Covariates previously identified as indicative of an increased risk of myocarditis were included in statistical models. For example, Lee et al found that preexisting autoimmune disease (eg, hypothyroidism) increases the risk of cardiovascular events after immunotherapy. ^12^ The Charlson Comorbidity Index was used to assess the baseline comorbid state of each patient.

Exposure Definition

Patients who started treatment with any of 7 ICIs (pembrolizumab, nivolumab, atezolizumab, durvalumab, cemiplimab, avelumab, or ipilimumab) at any time from the approval of the first checkpoint inhibitor (ipilimumab, March 2011) until March 1, 2022, were separated into 2 those starting either a single ICI or those starting a combination of nivolumab and ipilimumab. Patients who met the inclusion criteria could contribute to each group only once.

Figure 1 shows the longitudinal study design. Follow‐up started on the day after the first drug dispensing. The cohort members, by definition initiators of ICI, were followed until an outcome event, death, end of continuous health plan enrollment, or end of the study period, whichever came first.

Study Outcomes

The primary outcome measure was the occurrence of myocarditis. The definition of myocarditis through September 30, 2015, used the ICD‐9 codes reviewed by Idowu et al. ^13^ For follow‐up beginning on or after October 1, 2015, the same ICD‐9 codes were mapped onto ICD‐10 codes through the CMS General Equivalence Mappings. The secondary outcome was a new diagnosis of heart failure.

Statistical Analysis

Exposure propensity scores (PS) were calculated as the predicted probability of receiving combination therapy versus single therapy among all cohort entrants, conditional upon the subjects' baseline covariates, using multivariable logistic regression models. The basic covariates for adjustment were age, sex, and race. Further adjustment for potential confounders included all baseline covariates that individually predicted the occurrence of myocarditis (Table 1) and other conditions known to cause cardiac troponin increases (eg, ischemic heart disease, heart failure, and renal failure). A list of drugs considered potentially cardiotoxic is given in Data S1. Cohorts were matched on their PS on a 1 basis with a caliper of 0.01 on the propensity scale.

For the primary and secondary outcome measures, unadjusted, adjusted, and PS‐matched numbers of events, incidence rates, and hazard ratios with 95% CI were calculated using Cox proportional hazards models. Kaplan–Meier curves were used to determine the progression of risk over time.

Sensitivity analyses for the analytic method were performed using a variety of PS‐based inverse probability of treatment weights with the initially identified covariates; and matching combination therapy to monotherapy recipients according to original and high‐dimensional propensity scores, which adds empirical covariates to the investigator‐specified covariates. ^14^ Candidate covariates for high‐dimensional propensity scores consisted of all inpatient and outpatient diagnoses and procedures and pharmacy dispensings. Matching techniques remove large fractions of the comparison population, meaning matched and unmatched estimates must be compared with caution, even when they target the same underlying values.

All analyses were performed using the Aetion Evidence Platform V 4.68 using R version 3.4.2 and Stata Version 17.0.

RESULTS

Study Cohort and Patient Characteristics

Figure 2 shows the patient flow. Of the 84.9 million patients in the Clinformatics DataMart, 53 018 received therapy with 1 of the 7 ICIs studied, and a cancer diagnosis was recorded in 48 006 (90.5%). The 2 most frequently used individual ICIs were pembrolizumab and nivolumab, followed by atezolizumab, ipilimumab, durvalumab, cemiplimab, and avelumab (data not shown). A total of 4459 patients received combination therapy with nivolumab and ipilimumab. Use of checkpoint inhibitors (both as monotherapy and in combination) increased year on year (Table S1).

Table 1 compares the baseline characteristics of patients taking a single ICI to those receiving combination therapy (nivolumab and ipilimumab). Patients prescribed combination therapy were younger than those receiving monotherapy and were more likely to be men and White. Combination therapy was most frequently prescribed to patients with dermatological, respiratory, and renal cancer.

The mean follow‐up time in the overall cohort was 226 days (interquartile range, 93–495 days): 272 days in the combination therapy group (interquartile range, 108–588 days) and 223 days (interquartile range, 92–486 days) in the monotherapy group.

Absolute and Relative Rates of Myocarditis

Table 2 shows myocarditis events by treatment subgroup and Table S2 shows them by cancer type. There were 148 cases of myocarditis, representing 0.3% of the combined cohorts, 33 occurring in patients on combination therapy (0.7%) and 115 on monotherapy (0.2%).

The risk of myocarditis per 1000 patients was 7.40 in the combination therapy group and 2.37 in the monotherapy group (risk ratio, 3.12 [95% CI, 2.12–4.60]). The rate of myocarditis per 1000 patient‐years was 6.46 in the combination therapy group and 2.40 in the monotherapy group (rate ratio, 2.69 [95% CI, 1.83–3.96]; rate difference per 1000 person‐years, 4.06 [95% CI, 1.81–6.31]).

The results of the Cox proportional hazards analyses for myocarditis are shown in Figure 3. An unadjusted model provided an estimate of the increased risk of myocarditis in patients taking combination checkpoint inhibitor therapy (hazard ratio, 2.92 [95% CI, 1.98–4.29]). Estimates dropped modestly with adjustment for basic confounders (hazard ratio, 2.72 [95% CI, 1.83–4.04]) and all confounders (hazard ratio, 2.38 [95% CI, 1.57–3.63]). Alternative analysis using different PS‐based procedures all indicated a similarly elevated hazard ratio (Tables S3–S5).

Median time to onset of myocarditis was 61.5 days (interquartile range, 30–159 days). Kaplan–Meier time‐to‐event curves showed that the risk of myocarditis rose most quickly in the first 3 months after treatment initiation (Figure 4). Most events occurred within 100 days of initiating 23 of 33 (69.7%) in the combination therapy group and 77 of 115 (67.0%) in the single therapy group. At 6 months, three‐quarters of all myocarditis events had 25 of 33 (75.8%) in the combination group and 87 of 115 (75.7%) in the monotherapy group.

Risks of Congestive Heart Failure

Myocarditis can present with acute ventricular dysfunction that manifests as cardiac failure and may be incorrectly diagnosed. To determine whether myocarditis could have been misidentified as heart failure, the primary analysis was repeated using a post‐treatment first‐time diagnosis of heart failure as the outcome. There were proportionately fewer heart failure diagnoses noted in patients taking combination therapy (696/4459, 15.6%) than in patients taking monotherapy (8578/48 559, 17.7%). Under Cox regression models with covariate adjustment, there was no indication of an increased risk of heart failure over the follow‐up period in patients receiving combination therapy (Table 3). Alternative analysis using different PS‐based procedures showed no evidence of a difference in heart failure rates between treatment groups (Tables S6 through S8).

DISCUSSION

In this large population‐based cohort study, an increased risk of myocarditis was found in patients taking 2 ICIs (nivolumab and ipilimumab) simultaneously, compared with those taking a single agent, particularly in the first 100 days after treatment. This result was similar in unadjusted, adjusted, and PS models. No difference in the rate of heart failure was noted between the groups, suggesting that the outcome was identified correctly and that there was little or no residual confounding by unmeasured baseline morbidity in patients receiving combination therapy.

Randomized controlled trials may be inadequate to detect rare but serious adverse events such as myocarditis. For example, Larkin et al found no myocarditis in 313 patients with melanoma randomized to ICI combination therapy. ^15^ In 2016, Johnson et al were the first to report 2 cases of fulminant myocarditis in patients treated with ipilimumab and nivolumab, both of which were fatal. ^8^ The number of cases available has similarly limited subsequent reports investigating myocarditis. Mahmood and colleagues reported 35 ICI myocarditis cases from an 8‐site registry, ^16^ compared with the 148 events in the current analysis. Therefore, large‐scale insurance databases can be particularly informative in identifying the exact frequency of adverse effects that occur less commonly. Of note, the incidence of combination therapy myocarditis found here (0.7%) is higher than that reported by Johnson et al from pharmacovigilance data (0.27%).

The results underscore the need for early detection of ICI complications. The low risk of late‐onset myocarditis (Figure 3) suggests that biomarker screening (using troponin ^17^ ) should be more frequent in the first 6 months following the initiation of combination or single ICI therapy when 76% of events were noted to occur. Additionally, because ICI myocarditis is more common in patients on dual checkpoint therapy than on monotherapy, physicians may wish to consider the suitability of this treatment regimen in patients with autoimmune conditions. ^12^ Further investigation into the risk profile of other combination checkpoint inhibitor therapeutic approaches is warranted.

Study Limitations

The absolute number of myocarditis events in patients taking checkpoint inhibitors was low (n=148), as was the overall rate of this adverse event in the cohort (2.79 events per 1000 person‐years).

Residual confounding by unmeasured or insufficiently well‐measured patient characteristics remains possible. However, the lack of an elevated risk for heart failure in the combination therapy group suggests no critical (but difficult to quantify) general ill health associated with combination therapy.

Misclassification of myocarditis in insurance claims data is likely present but cannot account for the results. Insensitivity of the claims algorithm would generally result in fewer cases but not the observed imbalance between single and combination therapy. If the myocarditis label captured other entities, the observed association would have to be present for the conditions that were erroneously labeled myocarditis. This alternate scenario is even more implausible because the leading candidate for such an overinclusiveness of the myocarditis label, heart failure, showed no positive association with combination therapy.

Lastly, the results derive from a US population with full insurance coverage and a generally high standard of medical care. Generalization beyond this or a similar population requires caution.

Despite its limitations, this study provides an in‐depth examination of the frequency of checkpoint inhibitor myocarditis in a real‐world setting. Health insurance data offer an essential adjunct to smaller studies and case reports because they give an accurate picture of drug uptake over time and the associated risks. In this study, several sensitivity analyses were used to verify the magnitude and directionality of the results, which were confirmed using unadjusted, adjusted, and PS models.

CONCLUSIONS

In this large population‐based cohort of patients receiving checkpoint therapy, the incidence of myocarditis was low; however, the risk of this complication more than doubled in patients taking combination therapy with nivolumab and ipilimumab. Myocarditis occurred most often within 100 days of commencing treatment. No increased risk for heart failure following myocarditis was seen in patients receiving combination therapy.

This study supports the increasing role of large pharmacoepidemiologic studies based on longitudinal data routinely generated in the provision of health care for millions of patients with cancer, which can provide important information on the safety of novel therapies in clinical practice; in particular, serious adverse events with low incidence rates.

Sources of Funding

This project was funded by the Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women's Hospital. S.S., was additionally funded by FDA Research contracts (HHSF223201710186C and HHSF223201710146C), the NIH (NHLBI R01‐HL141505, NIAMS R01‐AR080194), and the Burroughs Wellcome Fund. The study was in part funded by the National Institute of Arthritis and Musculoskeletal and Skin R01‐AR080194.

Disclosures

A.C. Lindsay owns equity in GSK and Johnson & Johnson. Dr Walker is a consultant to AstraZeneca on matters unrelated to this study. Dr Schneeweiss is participating in investigator‐initiated grants to the Brigham and Women's Hospital from Boehringer Ingelheim, and UCB Pharma, unrelated to the topic of this study. He is a consultant to Aetion Inc., a software manufacturer in which he owns equity. He is an advisor to Temedica GmbH, a patient‐oriented data generation company. His interests were declared, reviewed, and approved by the Brigham and Women's Hospital in accordance with their institutional compliance policies.

Supporting information