Drug-associated progressive multifocal leukoencephalopathy: a clinical, radiological, and cerebrospinal fluid analysis of 326 cases

References were identified by means of a three-step PubMed search conducted on August 30, 2015 (Fig. 2). To obtain a first inventory of the drugs that might trigger PML occurrence, the search terms ‘progressive multifocal leukoencephalopathy’, ‘drug induced’, ‘chemically induced’, ‘medication induced’, ‘adverse event’, ‘adverse effect’, and ‘side effect’ were used in the ‘all fields’ menu without limits of time. Second, combinations of ‘progressive multifocal leukoencephalopathy’ and the names of each of the individual drugs that had been identified in the first step were applied to collect all reports of potentially drug-associated PML. Finally, reference lists of all included articles were screened for any additional relevant reports.

We included cases of possible, probable, and definite PML according to the current American Academy of Neurology criteria [3] that were published in English language journals in case reports, case series, clinical trials, editorials, letters, meta-analyses, and observational studies. In addition, articles were included when they described a simultaneous immune reconstitution inflammatory syndrome (IRIS), either occurring already during treatment with or only after withdrawal of the causative drug. With respect to the latter, IRIS was defined as deterioration of neurologic deficits following drug removal, corroborated by inflammatory changes on neuroimaging [8]. Articles were excluded if an identical case had already been described in a previous report, when PML was not related to a drug, or if an alternative diagnosis was more likely, e.g., posterior reversible encephalopathy syndrome (PRES) associated with tacrolimus, cyclosporine, cisplatin, and methotrexate. Finally, cases of medication-related granule cell neuronopathy were excluded [9, 10]. When imaging features played a pivotal role in the distinction between ‘possible PML’ and ‘not PML’, the decision of inclusion or exclusion was made after analysis by an experienced neuroradiologist with special expertise in the field of inflammatory diseases of the central nervous system (MPW).

In addition to the cases published in the literature, reports on medication-associated PML were obtained by querying the database of the Netherlands Pharmacovigilance Centre Lareb. This database contains detailed accurately verified information of adverse drug reactions reported by Dutch healthcare professionals and manufacturers. Care was taken to exclude duplicates that had already been included in the literature search.

The following data, when available, were extracted from each report: study type, number of cases, age, sex, underlying disease, immunosuppressive agent(s), time elapsed between drug introduction and either first symptoms or asymptomatic radiological findings that could be attributed to PML in retrospect, mortality rate (and time from first symptoms to death), and clinical, radiological, CSF, and histopathological features. When a combination of therapies rather than a single pharmaceutical was presumed to be responsible for the phenotype, the most likely contributing drugs—based on mechanism of action, dose, and duration of use—were noted up to a maximum of three. The localization of PML lesions was determined by the description provided by the authors of each report and the MR and/or CT images shown, if any. In addition, the presence of contrast enhancement, its distribution pattern (i.e., patchy, punctate, homogeneous, or unknown), and the dissemination pattern of lesions were listed. With respect to the latter, we distinguished between unilobar, multilobar (i.e., two or more contiguous lobes affected), and widespread (i.e., two or more noncontiguous lobes affected and/or the presence of lesions in both hemispheres). For the sake of clinical relevance and vigilance, symptoms and radiological lesions that appeared after establishing the diagnosis of PML, attributable to either progression of the disease or a consecutive emerging IRIS, were excluded.

Differences between the four subgroups, based on the underlying disease categories, were analyzed using Chi-square tests and one-way ANOVA. Post-hoc Bonferroni corrections have been applied and significance levels were set at 0.0083 (i.e., 0.05/6). Due to the conservative nature of the Bonferroni method [11], it seems likely that potential differences between the subgroups represent genuine dissimilarities rather than mere coincidence.

The literature search yielded 287 cases of drug-related PML in 184 case reports, 21 case series, 18 observational studies, three experimental studies, and two review articles [12‐239]. Furthermore, an additional 39 reports were identified in the Dutch pharmacovigilance database, resulting in a total number of 326 documented cases. In eight cases, consensus was reached after initial doubt on the diagnosis of PML; five of them have been included, while the remaining three were finally labeled as ‘not PML’. Natalizumab, predniso(lo)ne, (dimethyl) fumarate, fludarabine, rituximab, and brentuximab vedotin were the most common single agents that have been demonstrated to trigger PML (Online Resource 1). Additionally, glucocorticoids with either azathioprine or cyclophosphamide and a regime containing four or more pharmaceuticals [usually courses of chemotherapy, frequently including rituximab (R-CHOP)] comprised the most prevalent composite culprits. The multifarious underlying diseases could be subdivided into four main categories, i.e. multiple sclerosis (MS), (other) immune-mediated disorders, neoplasms (93.8 % lymphoproliferative, 3.1 % myeloproliferative, 2.1 % solid neoplasm, 1.0 % unknown), and the post-transplantation setting (Fig. 3). Because of the relatively large number of natalizumab-associated PML cases, the former was not included in the group of immune-mediated diseases, but rather considered a distinct entity. As immunosuppressive drugs are also frequently used in the prevention of relapses in neuromyelitis optica spectrum disorder after a first episode, one might expect drug-associated PML to have been described in this patient population. However, no such reports have been identified in our search. The most common drugs associated with PML per subgroup are summarized in Online Resource 2. In general, there was considerably few drug overlap between the various disease categories, an exception being the combination of azathioprine and glucocorticoids which was used in both autoimmune disorders and the post-transplantation setting.

The age and sex distribution of drug-related PML among the subgroups were in accordance with the epidemiologic features of the particular diseases, with a higher percentage of younger women and older men in the MS and neoplasm subgroups, respectively (Table 1). The mean time elapsed from introduction of the drug(s) to first symptoms was 28.9 (95 % CI 25.5–32.2) months, while the time span to the occurrence of asymptomatic lesions (only in natalizumab-induced PML) was found to be 36.4 (95 % CI 27.0–45.9) months. Significantly shorter periods were observed in the neoplasm group [mean 14.2 months (95 % CI 9.3–19.2)] compared to the three other disease categories (p < 0.001). A delay of several weeks to even months between the occurrence of first symptoms and the final diagnosis of PML was noticed in a considerable number of cases [23, 28, 50, 67, 78, 101, 103, 117, 122, 149, 158, 185, 201, 213, 223, 224, 226, 233, 235, 236]. This diagnosis was established on combined clinical, radiological, and CSF grounds in 69.4 % of cases. In the remaining 30.6 %, histopathological analysis was necessary to provide a greater degree of certainty, either because PCR for JCV in the CSF was not available or (repeatedly) negative. Brain biopsies were conducted most commonly in patients with underlying autoimmune diseases (43.8 %) and neoplasms (47.8 %), but in only 8.3 % of natalizumab-induced PML cases. The overall mortality rate was 52.2 % (95 % CI 46.5–58.0 %), with values of 12.0 % (95 % CI 5.2–18.7 %), 56.2 % (95 % CI 44.5–67.8 %), 83.3 % (95 % CI 75.5–91.2 %), and 68.4 % (95 % CI 52.9–83.9 %) in the MS, immune-mediated, neoplasm and post-transplantation subgroups, respectively.

In descending order of magnitude, the most common overall symptoms and signs of drug-related PML were motor weakness (48.6 %), cognitive deficits (43.2 %), dysarthria (26.3 %), ataxia (24.1 %), aphasia (22.7 %), and behavioral changes (21.9 %) (Table 2). Although some deviations were observed between the subgroups, motor weakness and cognitive changes constituted the two most frequently encountered symptoms in all disease categories. Behavioral changes were significantly more common in the post-transplantation group than in any of the other disease categories (MS p < 0.001, immune-mediated disorders p = 0.001, neoplasms p = 0.002). The highest prevalences of cognitive deficits and seizures were also found in this disease category (60.0 and 22.9 %, respectively). Symptoms and signs of cerebellar and brainstem involvement, i.e., ataxia, vertigo, eye movement disorders, and dysphagia tended to be more prevalent in the group of autoimmune disorders. Furthermore, dysarthria, gait abnormalities and sensory deficits (37.3, 26.9, and 20.9 %, respectively) occurred most commonly in the group of immune-mediated diseases. A notable feature of drug-related PML among the neoplasms was the high prevalence of visual deficits (25.8 %), especially when compared to the autoimmune disorders (p = 0.001). Asymptomatic PML cases were only observed in natalizumab-treated patients (10.3 %).

Whereas infratentorial lesions were present in 27.4 % of cases, drug-associated PML was considerably more prevalent in the supratentorial brain (87.7 %) (Table 3). In all disease categories, lesions were most often situated in the frontal and parietal lobes (64.1 and 46.6 %, respectively). The temporal (21.4 %) and occipital (22.7 %) lobes were less commonly affected. Involvement of the latter, however, was observed relatively frequently in the neoplasm subgroup (36.2 %), in particular when compared to the autoimmune and MS groups (p = 0.003 and p = 0.01, respectively). Infratentorial lesion localization was especially prevalent in drug-related PML associated with immune-mediated disorders, with cerebellar involvement being significantly more common than in the MS and neoplasm categories (p = 0.001 and p = 0.007, respectively). Furthermore, there was a trend towards a higher level of brainstem involvement in the group of autoimmune diseases (MS p = 0.013; neoplasm p = 0.077). In the natalizumab-induced PML subgroup, lesions were most commonly distributed in one lobe (42.9 % vs. 21.4 % multilobar and 27.1 % widespread). In contrast, widespread dissemination was by far the most frequently encountered pattern in the remaining three disease categories (autoimmune disorders 52.5 %, neoplasms 52.1 %, post-transplantation 68.0 %). Contrast enhancement was present in 27.6 % of cases, particularly in the natalizumab-induced (42.9 %) and post-transplantation (31.6 %) subgroups. In the former, punctate and patchy enhancement patterns were observed to an equal extent (19.6 %). In the other subgroups, only patchy contrast enhancement was found.

JCV DNA remained (repeatedly) undetectable in 22.7 % (95 % CI 17.3–28.1 %) of drug-associated PML cases. Furthermore, nearly one out of each seven patients (13.7 %; 95 % CI 9.3–18.1 %) initially displayed one or more negative outcomes before PCR finally converted to positive, ranging from 4.8 % in the post-transplantation (95 % CI 0.0–14.7 %) subgroup to 21.1 % (95 % CI 12.7–29.4 %) in the MS subgroup. Positive results in the first CSF sample were obtained in 63.5 %. Pleocytosis was present in 10.9 % of patients, especially in the neoplasm group (17.1 %).