Serum neuronal cell-surface antibodies in first-episode psychosis–Authors' reply

Abstract

We thank Feras Mustafa, and Frank Leypoldt and colleagues for their comments on our study about the prevalence and clinical characteristics of serum neuronal cell surface antibodies in first-episode psychosis.1 We agree with Mustafa that the present divide between neurology and psychiatry services in Britain is unjustified and that training initiatives that bridge the divide are required. The Royal College of Psychiatrists' present initiative to review the neuroscience training curriculum for psychiatrists is welcome.2 Regarding the data analysis of our study, the statistical reviewer recommended using logistic regression with odds ratios reported, and likelihood ratios reported where odds ratios were not calculable. Analysis with Fisher's test is another option; however, it comes with its own controversies, such as conditioning on the margins and it being conservative in rejecting the null hypothesis. We acknowledge the low numbers of cases, and that statistical significance depends on the analysis method used. Our interpretation was appropriately cautious in stating that “Further work is required to establish the specificity and pathogenicity of antibodies in the context of psychosis”.1 The threshold that we used for antibody positivity was not lower in this study than that used previously. We reported the “dilution of serum providing a score of 1”,1 with the lowest level being 1:30 (ie, the threshold is 1·5 at a dilution of 1:20), consistent with clinical practice and recent publications.3 We acknowledge there is disagreement regarding the sensitivity of the different antibody assays in detecting anti-NMDAR encephalitis. However, the study that Leypoldt and colleagues cite as demonstrating a higher sensitivity of the fixed-cell assay defined cases on the basis of positivity with that assay, rather than assessment of clinical relevance of the antibodies in the patients tested.4 The discrepant findings between the two assays in serum testing in a proportion of patients could, therefore, reflect a higher specificity of the live assay, rather than a reduced sensitivity. We did not intend to detect anti-NMDAR antibody encephalitis. We agree that it is important that further studies should include cerebrospinal fluid analysis and that lumbar punctures should be introduced into routine British psychiatric practice for the assessment of psychosis. Finally, Leypoldt and colleagues quote a poster at the European Academy of Neurology5 showing that immunohistochemistry was positive in 29 of 35 patients with the live cell-based assay compared with 30 of 35 patients with immunohistology in patients with clinically defined NMDAR-antibody encephalitis. Clearly, there is no statistical difference between the two methods based on those data. We thank Dr Mei-Man Lee for her statistical advice.