Presented at the 41st Annual International Society for Pharmacoepidemiology Conference
Background: A postauthorization safety study (EUPAS49836) was conducted examining the potential association between Ad26.COV2.S (the previously authorized Janssen COVID-19 vaccine) and adverse events of special interest (AESIs) using claims data from 4 US-based commercial insurers. Three AESIs on the Ad26.COV2.S label were considered at higher risk for misclassification using diagnosis and laboratory codes: co-occurring thrombosis with thrombocytopenia (TTS), Guillain-Barré syndrome (GBS), and myocarditis/pericarditis (MP).
Objective: To assess the potential impact of misclassification of TTS, GBS, and MP in a self-controlled risk interval (SCRI) study using quantitative bias analysis (QBA).
Methods: The study compared case counts in a post-vaccination risk window (RW) with a subsequent control window (CW), separated by a 14-day buffer period. RW lengths were AESI-specific, with the CW equal in length to the RW. Eligible vaccinees were ≥18 years of age and received Ad26.COV2.S as their first COVID-19 vaccine, with no receipt of additional COVID-19 vaccines through the end of the CW. We required health plan enrollment for ≥1 year before Ad26.COV2.S (and covering initial COVID-19 vaccine authorization in 12/2020) and through the end of the CW or death. Lastly, individuals were required to contribute an incident AESI to the RW or CW. Relative incidence ratios (RIR) were estimated with conditional Poisson regression.
In QBA, we used positive predictive values (PPV) (from the literature for GBS and MP; we tested 3 hypothetical PPVs for TTS) to estimate the number of false positive cases likely to be observed in the data. Estimated false positive cases were assumed to either all occur in the RW (best-case scenario) or the CW (worse-case scenario). We then recalculated the RIR after removing the false positive cases.
Results: Over 419,000 vaccinees met basic eligibility criteria. We observed 5 GBS cases in the 42-day RW and 2 in the CW (RIR=2.50), 24 MP cases in the 21-day RW and 18 in the CW (RIR=1.33), and 89 TTS cases in the 28-day RW and 49 in the CW (RIR=1.82).
Using a PPV of 71.2% for GBS, we estimated the corrected case counts would be 2.98 in the RW and <0 in the CW. This resulted in a best-case RIR of 1.49, and an undefined worst-case RIR (due to dividing by 0). Using a PPV of 53.3% for MP, we estimated a best-case RIR of 0.24 (corrected RW case count of 4.39) and an undefined worst-case RIR. One or both TTS RIRs were undefined due to counts ≤0 with PPVs of 30% and 50%; the bounds of the RIR were 0.97-11.7 using a PPV of 70%.
Conclusions: For GBS, our findings indicate that the observed elevated RIR may not be attributed to differential outcome misclassification alone. It was not possible to fully assess the impact of misclassification on MP and TTS due to lower PPVs and lower power to detect smaller effects.
