Encephalitis lethargica: clinical features and aetiology.

Encephalitis lethargica, an epidemic neurological illness, typically involved a severe sleep disorder and progressive parkinsonism. A century later, our understanding relies on seminal descriptions, more recent historical research and the study of small numbers of possible sporadic cases. Theories around infection, environmental toxins, catatonia and autoimmune encephalitis have been proposed. We aimed to describe the presentation of encephalitis lethargica and test these diagnostic and aetiological theories. Subjects with encephalitis lethargica were identified in the archives of the National Hospital for Neurology and Neurosurgery, UK between 1918 and 1946. Case notes were examined to establish illness temporality, clinical features and cerebrospinal fluid results. Controls from the archives were identified for 10% of cases, matching on discharge year, sex and neurologist. Clinical presentation was compared to modern diagnostic criteria for encephalitis lethargica, catatonia and autoimmune encephalitis. In a case-control design, a multilevel logistic regression was conducted to ascertain whether cases of encephalitis lethargica were associated with febrile illnesses and with environmental exposures. Six hundred and fourteen cases of encephalitis lethargica and 65 controls were identified. Cases had a median age of 29 years (interquartile range 18) and a median time since symptomatic onset of 3.00 years (interquartile range 3.52). Motor features were present in 97.6%, cranial nerve findings in 91.0%, ophthalmological features in 77.4%, sleep disorders in 66.1%, gastrointestinal or nutritional features in 62.1%, speech disorders in 60.8% and psychiatric features in 53.9%. Of the 167 cases who underwent lumbar puncture, 20 (12.0%) had a pleocytosis. The Howard and Lees criteria for encephalitis lethargica had a sensitivity of 28.5% and specificity of 96.9%. Among the cases, 195 (31.8%, 95% confidence interval 28.1-35.6%) had a history of febrile illness within one calendar year prior to illness onset, which was more common than among the controls (odds ratio 2.70, 95% confidence interval 1.02-7.20,