Replicating Safety Signals for Natalizumab Treatment in Multiple Sclerosis Patients Using German Administrative Health Insurance Claims Data

Introduction Multiple sclerosis (MS) is a chronic inflammatory disease of the central nervous system. Immunotherapies such as natalizumab are associated with rare, serious adverse events that were identified by the spontaneous reporting system. Objective To replicate safety signals from the spontaneous reporting system for natalizumab treatment in MS patients using two data sources (ds) of German statutory health insurance (SHI) data. Methods We analyzed administrative SHI claims data from 44 company medical funds (ds1, 2010–2018) and 12 German physicians’ associations (ds2, 2013–2018). Patients with a validated MS diagnosis (ICD‐10‐GM G35) treated with natalizumab until either subsequent other immunotherapies (natalizumab‐switcher) or no other following immunotherapy (natalizumab‐non‐switcher) were included. We searched for safety signals, based on a list of signals derived from the EMA spontaneous reporting system, and calculated odds ratios (ORs) for incident diseases using interferon therapy as a comparison. Results Overall, 36,798 MS patients were included. Although case numbers were low for most investigated diseases, a safety signal of progressive multifocal leukoencephalopathy (PML) could be replicated for natalizumab‐switcher (ds1: OR = 28.88 [6.02, 138.50]; ds2: OR = 59.94 [28.43, 147.32]) and natalizumab‐non‐switcher (ds1: OR = 28.74 [5.82, 141.88]; ds2: OR = 34.33 [15.43, 87.43]). For natalizumab‐switcher, furthermore, a safety signal for herpes simplex was reproduced (ds1: OR = 1.84 [1.27, 2.68]; ds2: OR = 1.48 [1.24, 1.77]). In ds2, increased risks were observed also for herpes zoster (natalizumab‐switcher: OR = 1.74 [1.49, 2.03]; natalizumab‐non‐switcher: OR = 1.36 [1.17, 1.58]) and “other polyneuropathies” (natalizumab‐switcher: OR = 1.42 [1.17, 1.70]). Conclusions Replication of safety signals for specific medications using German administrative SHI data is possible but requires very large datasets to obtain a sufficiently large study sample that allows the identification of rare adverse events such as PML.