LAUSR

Movement disorders can be associated with anti-neuronal antibodies. We conducted a systematic review of cases with documented anti-neuronal antibodies in serum and/or cerebrospinal fluid published in PubMed before April 1, 2020. Only patients with at least one movement disorder were included. We used random forests for variable selection and recursive partitioning and regression trees for the creation of a data-driven decision algorithm, integrated with expert’s clinical feedback. Three hundred and seventy-seven studies met eligibility criteria, totaling 844 patients and 13 antibodies: amphiphysin, GAD, GlyR, mGluR1, ANNA-2/Ri, Yo/PCA-1, Caspr2, NMDAR, LGI-1, CRMP5/CV2, ANNA-1/Hu, IgLON5, and DPPX. Stiffness/rigidity/spasm spectrum symptoms were more frequently associated with amphiphysin, GAD, and GlyR; ataxia with mGluR1, ANNA-2/Ri, Yo/PCA-1, Caspr2, and ANNA-1/Hu; dyskinesia with NMDAR and paroxysmal movement with LGI1; chorea/choreoathetosis with CRMP5/CV2, IgLON5, and NMDAR; myoclonus with GlyR and DPPX; tremors with ANNA2/Ri and anti-DPPX; and parkinsonism with IgLON5 and NMDAR. Data-driven classification analysis determined the following diagnostic predictions (with probability selection): psychiatric symptoms and dyskinesia predicted NMDAR (71% and 87%, respectively); stiffness/rigidity/spasm and ataxia, GAD (67% and 47%, respectively); ataxia and opsoclonus, ANNA-2/Ri (68%); chorea/choreoathetosis, CRMP5/CV2 (41%). These symptoms remained the top predictors in random forests analysis. The integration with an expert opinion analysis refined the precision of the approach. Breast and lung tumors were the most common tumors. On neuroimaging, cerebellar involvement was associated with GAD and Yo/PCA-1; temporal involvement with Caspr2, LGI-1, ANNA-1/Hu. Selected movement disorders are associated with specific anti-neuronal antibodies. The combination of data-driven and expert opinion approach to the diagnosis may assist early management efforts.