To describe neural autoantibody profiles and outcomes in patients with neurologic autoimmunity associated with immune checkpoint inhibitor (ICI) cancer immunotherapy.
In this retrospective descriptive study, 63 patients with ICI-related neurologic autoimmunity were included: 39 seen at the Mayo Clinic Neurology Department (clinical cohort) and 24 whose serum/CSF was referred to the Mayo Clinic Neuroimmunology Laboratory for autoantibody testing. Serum/CSF samples were tested for neural-specific autoantibodies. Predictors of unfavorable outcome (residual adverse event severity grade ≥3) were explored (logistic regression).
Median age at neurologic symptom onset was 65 years (range 31–86); 40% were female. Neurologic manifestations were CNS-restricted (n = 26), neuromuscular (n = 30), combined (n = 5), or isolated retinopathy (n = 2). Neural-specific autoantibodies were common in patients with CNS involvement (7/13 [54%] in the unbiased clinical cohort) and included known or unidentified neural-restricted specificities. Only 11/31 patients with CNS manifestations had neuroendocrine malignancies typically associated with paraneoplastic autoimmunity. Small-cell lung cancer (SCLC)–predictive antibodies were seen in 3 patients with non-neuroendocrine tumors (neuronal intermediate filament immunoglobulin G [IgG] and antineuronal nuclear antibody 1 with melanoma; amphiphysin IgG with non-SCLC). A median of 10 months from onset (range, 0.5–46), 14/39 in the clinical cohort (36%) had unfavorable outcomes; their characteristics were age ≥70 years, female, CNS involvement, lung cancer, higher initial severity grade, and lack of systemic autoimmunity. By multivariate analysis, only age remained independently associated with poor outcome (p = 0.01). Four of 5 patients with preexistent neurologic autoimmunity experienced irreversible worsening after ICI.
Neural-specific autoantibodies are not uncommon in patients with ICI-related CNS neurologic autoimmunity. Outcomes mostly depend on the pre-ICI treatment characteristics and clinical phenotype.