Importance: Neurologic symptoms are common in COVID-19, but the central nervous system (CNS) pathogenesis is unclear, and viral RNA is rarely detected in cerebrospinal fluid (CSF).

Objective: To measure viral antigen and inflammatory biomarkers in CSF in relation to neurologic symptoms and disease severity.

Design, setting, and participants: This cross-sectional study was performed from March 1, 2020, to June 30, 2021, in patients 18 years or older who were admitted to Sahlgrenska University Hospital, Gothenburg, Sweden, with COVID-19. All patients had CSF samples taken because of neurologic symptoms or within a study protocol. Healthy volunteer and prepandemic control groups were included.

Exposure: SARS-CoV-2 infection.

Main outcomes and measures: Outcomes included CSF SARS-CoV-2 nucleocapsid antigen (N-Ag) using an ultrasensitive antigen capture immunoassay platform and CSF biomarkers of immune activation (neopterin, β2-microglobulin, and cytokines) and neuronal injury (neurofilament light protein [NfL]).

Results: Forty-four patients (median [IQR] age, 57 [48-69] years; 30 [68%] male; 26 with moderate COVID-19 and 18 with severe COVID-19 based on the World Health Organization Clinical Progression Scale), 10 healthy controls (median [IQR] age, 58 [54-60] years; 5 [50%] male), and 41 patient controls (COVID negative without evidence of CNS infection) (median [IQR] age, 59 [49-70] years; 19 [46%] male) were included in the study. Twenty-one patients were neuroasymptomatic and 23 were neurosymptomatic (21 with encephalopathy). In 31 of 35 patients for whom data were available (89%), CSF N-Ag was detected; viral RNA test results were negative in all. Nucleocapsid antigen was significantly correlated with CSF neopterin (r = 0.38; P = .03) and interferon γ (r = 0.42; P = .01). No differences in CSF N-Ag concentrations were found between patient groups. Patients had markedly increased CSF neopterin, β2-microglobulin, interleukin (IL) 2, IL-6, IL-10, and tumor necrosis factor α compared with controls. Neurosymptomatic patients had significantly higher median (IQR) CSF interferon γ (86 [47-172] vs 21 [17-81] fg/mL; P = .03) and had a significantly higher inflammatory biomarker profile using principal component analysis compared with neuroasymptomatic patients (0.54; 95% CI, 0.03-1.05; P = .04). Age-adjusted median (IQR) CSF NfL concentrations were higher in patients compared with controls (960 [673-1307] vs 618 [489-786] ng/L; P = .002). No differences were seen in any CSF biomarkers in moderate compared with severe disease.

Conclusions and relevance: In this study of Swedish adults with COVID-19 infection and neurologic symptoms, compared with control participants, viral antigen was detectable in CSF and correlated with CNS immune activation. Patients with COVID-19 had signs of neuroaxonal injury, and neurosymptomatic patients had a more marked inflammatory profile that could not be attributed to differences in COVID-19 severity. These results highlight the clinical relevance of neurologic symptoms and suggest that viral components can contribute to CNS immune responses without direct viral invasion.