2081. Building a Decision Tree with Serial Serology Measurements Improves Classification in a Flavivirus Co-circulation Region

BACKGROUND: RT-PCR (reverse transcriptase polymerase chain reaction) is often considered the “gold standard” for diagnosis of Zika Virus (ZIKV) infection; however, it has been shown to have low sensitivity. A possible remedy is to study ZIKV-specific IgG (ZsIgG) and IgM (ZsIgM) antibodies. However, the in vitro cross-reactivities of Dengue virus (DENV) and ZIKV-specific antibodies are well known, leading to diagnostic difficulties in an area with co-circulation of the two viruses. Our goal was to use Zika and Dengue serologic assays to build a classification model that improves upon the PPV of commercial kits while maintaining sensitivity. METHODS: We conducted a prospective longitudinal study in Southern Mexico where DENV and ZIKV co-circulation occurs (NCT02831699). Patients were included in two cohorts: a cohort of subjects presenting with a febrile rash meeting WHO/PAHO Zika case definition and a household cohort. After signed consent, all subjects enrolled were evaluated on study-visit Days 0, 3 and 7 (for fever rash cohort) and 28. We considered a subject “true positive” for ZIKV or DENV if RT-PCR positive at any time point. The healthy household cohort (with no positive RT-PCR) was considered “true negatives.” We fit a statistical decision tree taking as inputs serial serology measurements and outputting a predicted disease category. Funded in part by the NCI Contract No. HHSN261200800001E. Funded in part by the Mexican Ministry of Health. RESULTS: As of March 2018, we have 32 subjects in the Zika PCR+ group, 32 in the Dengue PCR+ group, and 68 in the household group. Our decision tree (Figure 1) achieved PPV of at least 90% on all three disease categories, while maintaining sensitivity above 50%. The highest PPV achieved by the kit manufacturer recommended cutoffs while maintaining a sensitivity of at least 10% on Zika PCR+ subjects is 30/114 (26%), and for Dengue PCR+ subjects is 21/30 (70%). CONCLUSION: Using serology data in a statistical decision tree improves the PPV exhibited by the kit manufacturer recommendations while still maintaining respectable sensitivity. Physicians in regions with co-circulating flaviviruses should be aware of the pitfalls of using only RT-PCR or using pre-established commercial cutoffs in the serology kits for diagnosis. [Image: see text] DISCLOSURES: All authors: No reported disclosures.