Optimizing use of multi-antibody assays for Lyme disease diagnosis: A bioinformatic approach.

Multiple different recombinant and peptide antigens are now available for serodiagnosis of Lyme disease (LD), but optimizing test utilization remains challenging. Since 1995 the Centers for Disease Control and Prevention (CDC) has recommended a 2-tiered serologic approach consisting of a first-tier whole-cell enzyme immunoassay (EIA) for polyvalent antibodies to Borrelia burgdorferi followed by confirmation of positive or equivocal results by IgG and IgM immunoblots [standard 2-tiered (STT) approach]. Newer modified 2-tiered (MTT) approaches employ a second-tier EIA to detect antibodies to B. burgdorferi rather than immunoblotting. We applied modern bioinformatic techniques to a large public database of recombinant and peptide antigen-based immunoassays to improve testing strategy. A retrospective CDC collection of 280 LD samples and 559 controls had been tested using the STT approach as well as kinetic-EIAs for VlsE1-IgG, C6-IgG, VlsE1-IgM, and pepC10-IgM antibodies. When used individually, the cutoff for each kinetic-EIA was set to generate 99% specificity. Utilizing logistic-likelihood regression analysis and receiver operating characteristic (ROC) techniques we determined that VlsE1-IgG, C6-IgG, and pepC10-IgM antibodies each contributed significant diagnostic information; a single-tier diagnostic score (DS) was generated for each sample using a weighted linear combination of antibody levels to these 3 antigens. DS performance was then compared to the STT and to MTT models employing different combinations of kinetic-EIAs. After setting the DS cutoff to match STT specificity (99%), the DS was 22.5% more sensitive than the STT for early-acute-phase disease (95% CI: 11.8% to 32.2%), 16.0% more sensitive for early-convalescent-phase disease (95% CI: 7.2% to 24.7%), and equivalent for detection of disseminated infection. The DS was also significantly more sensitive for early-acute-phase LD than MTT models whose specificity met or exceeded 99%. Prospective validation of this single-tier diagnostic score for Lyme disease will require larger studies using a broader range of potential cross-reacting conditions.
Competing Interests: Richard Porwancher holds patents for bioinformatic methods for Lyme disease diagnosis (US 8,005,627 B2; US 8,694,265 B2; EP 2 084 535 B9), received previous grant support from the Centers for Disease Control and Prevention, Division of Vector-Borne Diseases, Fort Collins, CO, USA and from the National Institute of Allergy and Infectious Diseases of the National Institutes of Health, Bethesda, MD, USA, and has served as a consultant to the New Jersey Board of Medical Examiners, Trenton, NJ, USA. Richard Porwancher was previously employed by and Lisa Landsberg had been a consultant to Infectious Disease Consultants, PC, 1245 Whitehorse-Mercerville Rd., Suite 411, Mercerville, NJ 08619. The above interests do not alter our adherence to PLOS ONE policies on sharing data and materials.