To the Editor: Mae S. Sokol, M.D., and colleagues (1) recently reported a significantly higher rate of D8/17 positivity among 16 adolescents with anorexia nervosa (81%) than among 17 psychiatric comparison subjects (12%). The mean percentage of D8/17 positivity might "be useful in identifying PANDAS [pediatric autoimmune neuropsychiatric disorders associated with streptococcal infections] anorexia nervosa" (p. 1431). The results of that investigation are similar to findings in two previous studies (2, 3) demonstrating excellent performance of the D8/17 assay in distinguishing patients with obsessive-compulsive disorder (OCD) from healthy comparison subjects—85% to 100% of OCD patients were D8/17 positive in comparison with 5% to 15% of healthy comparison subjects. However, our subsequent experience has been less satisfactory and suggests that the assay is not yet reliable enough to be used as a diagnostic tool.

In our work, a total of 216 samples were gathered over a 4-year period from 26 subjects with Sydenham’s chorea, 42 subjects with OCD and/or tic disorders (PANDAS subgroup), and 19 healthy comparison subjects. Samples were obtained at the National Institute for Mental Health (NIMH) by workers blind to patient diagnosis and sent to Rockefeller University, where assays were performed by immunofluorescent microscopy before August 1998 and subsequently by flow cyometry, according to methods previously reported (4). Written informed consent or assent was obtained from all subjects, who were participating in studies of Sydenham’s chorea and OCD that were approved by NIMH institutional review boards.

Overall, the sensitivity of the D8/17 assay for the 68 patients with Sydenham’s chorea or OCD/tics was 61.8% (42 assays were positive). Of concern, only 12 assays (46.2%) obtained from the 26 Sydenham’s chorea patients were positive, which is significantly lower (z=5.2, p<0.001) than that previously reported for patients with rheumatic fever (89%–100% were D8/17 positive) (5, 6). Thus, the D8/17 assay failed to "diagnose" the majority of the patients with Sydenham’s chorea. Furthermore, the reliability of the assay was suboptimal. Longitudinal observations of 54 subjects tested at random intervals over the study period demonstrated test-retest agreement of 61.1% (N=132) for the 216 samples assayed (kappa=0.18). Of particular concern, agreement was observed in only 48 of 61 split samples (78.7%, kappa=0.48).

In conclusion, the sensitivity of the D8/17 assay decreased to unacceptably low levels during the period of observation. The declining sensitivity may have been due to changes in methods or the characteristics of the monoclonal antibody. If so, the performance might be improved by reversion to earlier techniques. Meanwhile, it appears premature to include the D8/17 assay in the diagnostic workup of patients with neuropsychiatric disorders such as OCD, tics, and anorexia nervosa.