Progressive multifocal leukoencephalopathy (PML) associated with natalizumab treatment continues to be a severe problem of clinically successful therapy. This is an update of risk stratification developments and discusses the current approach to depict and calculate PML incidence and PML risk. (1) PML incidence and resulting risk used in today's clinical practice are potentially outdated and the risk for patients with prior immunosuppression might have been underestimated. (2) Risk stratification according to treatment duration epochs likely suggests lower risk due to patients stopping treatment within a given epoch. PML incidence within the complete treatment epoch is statistically lowered due to the fact that patients at the beginning of an epoch presumably have a lower PML risk than the patients at the end. Periodic risk is not accurate in assessing risk for long treatment durations. (3) The JC virus (JCV) serostatus risk factor has low specificity concerning PML prediction and anti-JCV seroconversion during treatment with natalizumab further lowers its specificity over time. Specificity of the risk factor treatment duration varies depending on the average treatment duration and the number of short-term patients. These short-term patients reduce overall average treatment duration and thus enhance the specificity of the risk factor and reduce overall PML incidence. It is also suggested that short-term natalizumab patients are exclusively non-PML, even though they might still develop PML. Clinicians have to consider the cumulative risk of patients to stratify efficiently.