Importance: Risk of developing progressive multifocal leukoencephalopathy (PML) is the major barrier to using natalizumab for patients with multiple sclerosis (MS). To date, the association of risk stratification with PML incidence has not been evaluated. Objective: To describe the temporal evolution of PML incidence in France before and after introduction of risk minimization recommendations in 2013. Design, Setting, and Participants: This observational study used data in the MS registry OFSEP (Observatoire Français de la Sclérose en Plaques) collected between April 15, 2007, and December 31, 2016, by participating MS expert centers and MS-dedicated networks of neurologists in France. Patients with an MS diagnosis according to current criteria, regardless of age, were eligible, and those exposed to at least 1 natalizumab infusion (n = 6318) were included in the at-risk population. A questionnaire was sent to all centers, asking for a description of their practice regarding PML risk stratification. Data were analyzed in July 2018. Exposures: Time from the first natalizumab infusion to the occurrence of PML, natalizumab discontinuation plus 6 months, or the last clinical evaluation. Main Outcomes and Measures: Incidence was the number of PML cases reported relative to the person-years exposed to natalizumab. A Poisson regression model for the 2007 to 2016 period estimated the annual variation in incidence and incidence rate ratio (IRR), adjusted for sex and age at treatment initiation and stratified by period (2007-2013 and 2013-2016). Results: In total, 6318 patients were exposed to natalizumab during the study period, of whom 4682 (74.1%) were female, with a mean (SD [range]) age at MS onset of 28.5 (9.1 [1.1-72.4]) years; 45 confirmed incident cases of PML were diagnosed in 22 414 person-years of exposure. The crude incidence rate for the whole 2007 to 2016 period was 2.00 (95% CI, 1.46-2.69) per 1000 patient-years. Incidence significantly increased by 45.3% (IRR, 1.45; 95% CI, 1.15-1.83; P =.001) each year before 2013 and decreased by 23.0% (IRR, 0.77; 95% CI, 0.61-0.97; P =.03) each year from 2013 to 2016. Conclusions and Relevance: The results of this study suggest, for the first time, a decrease in natalizumab-associated PML incidence since 2013 in France that may be associated with a generalized use of John Cunningham virus serologic test results; this finding appears to support the continuation and reinforcement of educational activities and risk-minimization strategies in the management of disease-modifying therapies for multiple sclerosis. © 2019 American Medical Association. All rights reserved.< Réduire