Objective: To compare the available diagnostic criteria for progressive multifocal leukoencephalopathy (PML) diagnosis in a real-world cohort of patients with natalizumab-associated PML and to explore opportunities for improvement of such criteria in the context of pharmacovigilance of immunosuppressive therapies. Methods: We applied the “Mentzers PML case definition” to a dataset of 28 patients with natalizumab-associated PML (many of whom were identified through MRI screening in the context of pharmacovigilance), who were previously rated according to the American Academy of Neurology (AAN) PML diagnostic criteria, and compared the response to both sets of criteria. Results: The “Mentzers case definition” resulted in a level of certainty 1–3 in patients with a positive JC virus PCR, termed ‘definite’ and ‘probable’ PML according to the AAN diagnostic criteria. Patients that tested negative for JC virus in CSF (29%) were classified level 4 by the “Mentzers case definition”, neglecting the longitudinal clinical and radiological signs of PML available, while the AAN diagnostic criteria separated these patients in ‘possible’ and ‘not PML’. Conclusions: Both the AAN PML diagnostic criteria and the “Mentzers case definition” require the positive detection of JC virus DNA in CSF to define patients at a higher degree of suspicion of PML. However, as sensitivity of JC virus PCR in CSF is limited and often returns negative in particular in early cases of PML with a mere MRI-based PML suspicion, both criteria have obvious limitations when frequent MRI is used for pharmacovigilance purposes. Thus, revision of PML diagnostic criteria is needed, including the incorporation of lesion evolution, and longitudinal CSF studies that also assess for the presences of intrathecally produced anti-JC virus antibodies.