Horta, Erika

Abstract Purpose: Clustering of neural autoantibodies in patients with paraneoplastic neurologic disorders may predict tumor type. A mathematical analysis of neural autoantibody clusters was performed in 78,889 patients undergoing evaluation for a suspected paraneoplastic autoimmune neurologic disorder. Tumor predictive autoantibody profiles were confirmed in sera from patients with histologically proven tonsillar cancer, thymoma, and lung cancer. Patients and Methods: Of note, 78,889 patient sera were tested for 15 defined neural autoantibodies (1.2 million tests). The observed and hypothesized frequencies of autoantibody clusters were compared and their tumor associations defined. A tumor validation study comprised serum from 368 patients with a variety of tumors (thymoma, lung, or tonsil). Results: Informative oncological associations included (i) thymoma in 85% of patients with muscle striational, acetylcholine receptor antibodies plus CRMP5 autoantibodies; (ii) lung carcinoma in 80% with both P/Q-type and N-type calcium channel antibodies plus SOX1-IgG; and (iii) in men, prostate carcinoma frequency more than doubled when striational and muscle AChR specificities were accompanied by ganglionic AChR antibody. In women, amphiphysin-IgG alone was associated commonly with breast carcinoma, but amphiphysin-IgG, coexisting with antineuronal nuclear autoantibody-type 1 or CRMP5-IgG, was associated with lung cancer (P < 0.0001). In the validation cohorts, many tumor-associated profiles were encountered that matched the clusters identified in the screening study (e.g., 15% of thymoma patients had striational, acetylcholine receptor antibodies plus collapsin response-mediator protein-5 autoantibodies). Conclusions: Neural autoantibodies commonly coexist in specific clusters that are identifiable by comprehensive screening. Signature autoantibody clusters may predict a patient's cancer risk and type. Clin Cancer Res; 20(14); 3862–9. ©2014 AACR.