BACKGROUND/PURPOSE: Biologic DMARDs used for the treatment of RA may increase the risk of tuberculosis (TB).1 Large healthcare claims databases are useful in assessing rare events such as TB, but confirmation of these events can be challenging. As part of an ongoing post-marketing safety evaluation, we sought to characterize potential TB cases in claims data among RA patients initiating biologic and non-biologic DMARDs.
METHODS: A prospective cohort study of adults who initiated biologic or non-biologic DMARDs and who had at least 6 months’ prior continuous health plan enrollment (baseline) and a baseline claim for RA (International Classification of Diseases, 9th Revision [ICD-9] 714.xx) between December 1, 2005 and March 31, 2013 was conducted using administrative data from a large United States healthcare insurer. Potential TB events following drug initiation were identified with diagnosis codes associated with healthcare claims (ICD-9 010.xx- 018.xx) and characterized using criteria from a published validation algorithm for TB detection by Calderwood et al. (2010)2, including the presence of claims suggestive of at least two TB treatments (e.g., pyrazinamide, rifampin, isoniazid, ethambutol) within 60 days, prescription for pyrazinamide, and acid-fast bacilli (AFB) testing in the preceding 60 days or subsequent 14 days. Potential TB events were also characterized by the presence of chest x-ray claims within 60 days.
RESULTS: We identified 15,183 biologic DMARD initiators and 50,492 non-biologic DMARD initiators contributing 32,230 and 100,580 person-years of follow-up, respectively. Biologic DMARD initiators were more likely than non-biologic DMARD initiators to have previously used another biologic during the baseline period (20% vs 9%) and more likely to have a baseline claim for a TB skin test (26% vs 4%). A total of 251 potential TB events were identified during follow-up, 59 among biologic DMARD initiators and 192 among non-biologic DMARD initiators. Among the potential cases, 2 (3%) biologic DMARD initiators and 2 (1%) non-biologic DMARD initiators had claims for pyrazinamide, and 4 (7%) biologic DMARD initiators and 4 (2%) non-biologic DMARD initiators had claims for at least 2 different anti-TB medications. A total of 5 (8%) potential cases among biologic DMARD initiators and 6 (3%) potential cases among non-biologic DMARD initiators had an AFB test claim. A total of 28 (47%) potential cases among biologic DMARD initiators and 103 (54%) potential cases among non-biologic DMARD initiators had a claim for chest x-ray.
CONCLUSION: Few potential TB events identified on the basis of diagnosis codes had additional supporting claims. Claims-based algorithms based on diagnosis codes alone are likely insufficient for accurate identification of TB cases in administrative data. Consideration of claims indicative of TB treatment or diagnostic work-up in combination with diagnosis codes may improve identification of TB events in safety surveillance.
