Comment on “Risk of Pericarditis Recurrence Associated With Rilonacept and NSAID/Colchicine Therapy After Initial Pericarditis Presentation: A Retrospective Real-World Comparative Risk Analysis With TriNetX”

To the Editor: We read with interest the recent publication “Risk of Pericarditis Recurrence Associated With Rilonacept and NSAID/Colchicine Therapy After Initial Pericarditis Presentation: A Retrospective Real-World Comparative Risk Analysis With TriNetX” published in the American Journal of Therapeutics (published November 11, 2025).1 We commend the authors for attempting to evaluate the real-world efficacy of rilonacept in recurrent pericarditis (RP); however, several methodological and interpretive limitations substantially constrain the validity of the conclusions presented. The authors' statement that the incidence of a second documentation of pericarditis among patients receiving rilonacept as a monotherapy (n = 31) was 67.7% is based on a flawed analytical methodology. In addition to the limitations inherent to the small sample size, the limitations in the methodology used to derive this estimate fundamentally confound the validity of the analysis. Specifically, pericarditis recurrence was inferred solely from pericarditis-related ICD codes occurring at least 1 day after the index date. However, ICD code dates are a highly imprecise surrogate for pericarditis recurrences in RP; unlike in conditions in which diagnostic codes reliably correspond to clinical exacerbations, the ICD codes used in RP frequently reflect simple routine follow-up visits or evaluations of well-controlled patients rather than solely providing evidence of a confirmed clinical flare.2 This imprecision in ICD codes is due to the inherent lack of clinical granularity and supporting evidence (e.g., reports of chest-pain, C-reactive protein CRP elevations, imaging evidence).3 In addition, claims data derived solely from ICD codes lack the clinical documentation necessary to verify treatment exposure and dose continuity. Specifically, the analysis does not establish whether documented events occurred during active rilonacept treatment, after treatment cessation, or during off-label dose interval elongations; i.e., the analysis does not confirm that rilonacept was administered on a consistent weekly schedule (per label) when the “recurrence” took place. These distinctions are essential in assessing clinical outcomes for RP, a chronic autoinflammatory disease that requires continuous treatment for the duration of the disease, as events occurring after treatment cessation or during dose interval elongation represent the natural reemergence of the underlying autoinflammatory disease and need for continued suppressive therapy rather than failure of prior therapy.4 In the absence of data elements necessary for clinical adjudication and verified treatment exposure, recurrence estimates based on this flawed ICD-based methodology are not reliable. Consequently, the stated conclusions drawn from this analysis are not valid assessments of the real-world therapeutic effectiveness of rilonacept. Accurate classification of suspected pericarditis recurrences requires careful curation with confirmatory clinical evidence from orthogonal datasets to reliably identify and confirm true disease flares. In the phase 3 trial RHAPSODY, which demonstrated that once-weekly rilonacept provided a 96% risk reduction (rilonacept vs. placebo) in recurrence risk in the Randomized-Withdrawal (RW) Period and 98% in the 2-year Long-term Extension Period, supporting authorization as the only FDA-approved treatment for RP, only pericarditis recurrence events confirmed by the Clinical Endpoint Committee (CEC) during the RW period were used in the Primary Endpoint analysis.5–7 This approach established the gold-standard for high-quality clinical event adjudication and pericarditis recurrence quantification in RP. Consistent with that benchmark, the multicenter observational REgiStry Of the NAtural history of recurrent periCarditis in pEdiatric and adult patients (RESONANCE; NCT04687358), which has provided an informative datastream over the past 5 years in more than 500 patients with RP across 29 sites, applied a structured adjudication framework in which available contemporaneous objective clinical indicators of recurrence (e.g., patient-reported chest pain, serum CRP, and/or cardiac imaging) were collated, and suspected investigator-assessed pericarditis recurrences were reviewed in the context of the standardized RHAPSODY CEC pericarditis recurrence event adjudication criteria.8–11 For an ICD-based analyses to achieve similar validity, pericarditis recurrence adjudication would need to be supported by such orthogonal datasets. However, there is no evidence in the manuscript that such adjudication or verification was performed.1,12,13 Instead, this ICD-based study misclassified each billing code as a pericarditis recurrence, substantially conflating them with clinical outcomes, thus fundamentally undermining the validity of the reported recurrence rates and hazard estimates and rendering them unreliable and clinically uninterpretable. Contrary to the authors' assertion that the RHAPSODY-proven benefit of rilonacept in reducing pericarditis recurrences may not consistently translate to real-world populations, methodologically rigorous real-world analyses have in point of fact closely mirrored RHAPSODY results. Data from RESONANCE demonstrated a 99.5% reduction in pericarditis recurrence during long-term (>2 years) rilonacept treatment, findings that closely align with RHAPSODY results.11 Importantly, RESONANCE datasets incorporate clinical granularity and rigorous adjudication, providing substantive reliability in characterizing recurrence rates in a real-world setting. When real-world analyses are conducted using such rigorous methodological standards, they reinforce (rather than contradict) the registrational clinical trial evidence for rilonacept. By contrast, the authors' conclusion of an apparent lack of real-world effectiveness is not a quantitation of the true therapeutic efficacy of rilonacept but rather a manifestation of the fundamental methodological limitations inherent in their ICD-based approach. Collectively, these limitations constrain the interpretability of the findings and weaken the causal inferences presented. In the absence of appropriate clinical adjudication and treatment exposure verification, the authors overstated recurrence rates and mischaracterized the real-world effectiveness of rilonacept, potentially misleading clinicians and policymakers making treatment decisions for patients with RP.1 By contrast, previously published real-world analyses in RP that incorporated rigorous event adjudication and exposure verification using granular clinical data sources confirmed that the magnitude of rilonacept clinical effectiveness in the real world does in fact parallel that seen in the registrational Phase 3 trial RHAPSODY, thus reinforcing the value of robust real-world analyses in which analytical methodologies have been appropriately designed and executed.