Risk of natalizumab-associated progressive multifocal leukoencephalopathy in patients with multiple sclerosis: a retrospective analysis of data from four clinical studies

Background Previous estimates of risk of progressive multifocal leukoencephalopathy (PML) in patients with multiple sclerosis receiving natalizumab were stratified by three risk factors: anti-John Cunningham virus (JCV) antibodies in serum, previous immunosuppressant use, and treatment duration, which were estimated using population-based assumptions. We aimed to calculate PML risk estimates from patient-level risk-factor data and to stratify risk by concentrations of anti-JCV antibody in serum (anti-JCV antibody index). Methods Data on natalizumab-treated patients were pooled from four large, observational, open-label studies: STRATIFY-2, STRATA, TOP, and TYGRIS. Data were analysed with and without imputation for missing values of anti-JCV antibody status and previous immunosuppressant use. For anti-JCV antibody-positive patients in this pooled cohort, cumulative PML risk with or without previous immunosuppressant use was estimated using Kaplan-Meier analysis. Annual PML risks (per 12 natalizumab infusions) for patients without PML in the preceding year were estimated using conditional probability based on the life table method. For anti-JCV antibody-positive patients without previous immunosuppressant use, risk estimates were further stratified using a probability distribution for anti-JCV antibody index values, separately for patients with or without PML. Anti-JCV antibody index cutoffs were selected via sensitivity and specificity assessments for identifying PML cases in an index cohort. Findings 156 (<1%) of 37249 patients in the pooled cohort had PML. We imputed missing values on anti-JCV antibody status (3912 patients) and on previous immunosuppresant use (544 patients) using a multiple imputation method. For anti-JCV antibody-negative patients (n=13?996), estimated PML risk was less than 0.07 per 1000 patients (95% CI 0.00-0.40). In anti-JCV antibody-positive patients (n=21696), estimated cumulative PML probability over 6 years (72 infusions of natalizumab) was 2.7% (95% CI 1.8-4.0) in patients with previous immunosuppressant use and 1.7% (1.4-2.1) in those without. In patients without previous immunosuppressant use (n=18616), estimated annual PML risks per 1000 patients, conditional on having no PML before that year, ranged from 0.01 (0.00-0.03) in year 1 (1-12 infusions) to 0.6 (0.0-1.5) in year 6 (61-72 infusions) for people with an index of 0.9 or less; from 0.1 (0.0-0.2) in year 1 to 3.0 (0.2-5.8) in year 6 for those with an index of more than 0.9 up to and including 1.5; and from 0.2 (0.0-0.5) in year 1 to 10.0 (5.6-14.4) in year 6 for those with an index of more than 1.5. Interpretation: Our risk estimates calculated from patient-level clinical data allow individualised annual prediction of risk of PML in patients receiving natalizumab for multiple sclerosis, supporting yearly benefit-risk re-evaluation in clinical practice. Further, our estimates are generally consistent with previously calculated estimates. Incorporating anti-JCV antibody index allows further risk stratification for anti-JCV antibody-positive patients who have not previously taken immunosuppressants.