Distinguishing Zika and Dengue Viruses through Simple Clinical Assessment, Singapore

Dengue virus and Zika virus coexist in tropical regions in Asia where healthcare resources are limited; differentiating the 2 viruses is challenging. We showed in a case–control discovery cohort, and replicated in a validation cohort, that the diagnostic indices of conjunctivitis, platelet count, and monocyte count reliably distinguished between these viruses.


The Study
We conducted a case-control study at the National University Hospital with ethics approval from the hospital's Institutional Review Board. Patients infected with Zika virus and DENV who were seen at the hospital in 2016 constituted the discovery cohort. We confirmed Zika virus infection through testing for viral RNA in serum or urine, as described by Lanciotti et al. (4). We confirmed DENV infection through testing for serum DENV nonstructural protein 1 (NS1) antigen (SD BIOLINE Dengue DUO Kit; Standard Diagnostics, Kyonggi-do, South Korea) or by reverse transcription PCR (5). The clinical information collected included demographics, symptomatology, examination findings, and laboratory investigations, including complete blood count (with the monocyte count automated) and liver function test.
We compared clinical characteristics of both infections by univariate logistic regression against dichotomous symptomatology and continuous laboratory parameters. We selected predictors that could differentiate Zika virus and DENV infection as input for subsequent multivariate regression models and computed the area under the receiver operating characteristic curve (AUC) to compare model performance. We validated the results in a separate cohort of Zika virus and DENV patients from Tan Tock Seng Hospital, Singapore (5). From this validation cohort, we ascertained AUC and accuracy of the derived predictors. There were no pregnant patients in either cohort. We performed all analyses with R statistical software version 3.3.1 (http:// www.R-project.org).
Zika patients sought treatment earlier in their illness than did DENV patients. Whereas viral symptoms including fever and arthralgia were common to both, differences were discernible (Figure 1). Conjunctivitis strongly indicated Zika virus infection (odds ratio [OR] 30.1, 95% CI 9.57-94.44; p < 0.001). In contrast, fever (OR 0.05,
Our findings point to conjunctivitis, platelet, monocyte, ALT, and AST levels as candidate markers to differentiate Zika virus patients from DENV patients. Conjunctivitis alone had an AUC of 0.79 in identifying Zika virus patients; normal platelet count in addition to conjunctivitis increased the AUC to 0.92, and adding a normal monocyte count further improved the AUC to 0.95 ( Figure 2).
The use of these 3 indices (conjunctivitis and platelet and monocyte counts) had 88% sensitivity and 93% specificity in distinguishing Zika virus from DENV, with a diagnostic accuracy of 92%. Inclusion of ALT and AST, however, did not further enhance the diagnostic capability.
We applied these 3 indices to a validation cohort consisting of 25 Zika virus and 70 DENV patients (Table   1566 Emerging   Zika virus and DENV coexist in many developing nations in equatorial South America and Southeast Asia, where there is limited accessibility to health resources and virus-specific diagnostics are not readily available. Differentiating Zika virus and DENV infections early is important in the prognostication and subsequent monitoring and follow-up of these patients. Although Zika virus infection is self-limiting, concerns about its sequelae in pregnant women and birth defects are well established (6). In contrast, severe DENV infection leads to debilitating illness that can cause vascular leakage, dengue shock, and death (7).
We applied both definitions from the US Centers for Disease Control and Prevention (CDC) and World Health Organization (WHO) for suspected Zika cases (8,9) in our patient cohort and found them to be unsatisfactory in distinguishing Zika virus from DENV patients (CDC, sensitivity 100%, specificity 2%; WHO, sensitivity 71%, specificity 67%) ( Table 2). We therefore sought to develop more accurate indices to identify Zika virus among the backdrop of DENV cases in Singapore.
Our results highlight the utility of conjunctivitis and normal platelet and monocyte counts to distinguish Zika virus infection. We found conjunctivitis to be already a strong predictor of Zika virus infection. The study by Waggoner et al. in Nicaragua had reported conjunctivitis and rash in association with Zika virus infection (10). However, rash was not prominent among Zika patients in our study. Headache and myalgia were more common in DENV (7) and could help to distinguish DENV from Zika virus in our cohort.
Prior studies had not ascertained if incorporation of basic laboratory indices could further enhance diagnostic capability. In our univariate logistic regression model, thrombocytopenia, transaminitis, and monocytosis were notable in DENV infection. Conversely, Zika patients tended to have normal platelet, aminotransaminase, and monocyte levels.

Conclusions
We were able to derive 3 simple clinical predictors on the basis of our findings: in the presence of conjunctivitis and normal platelet and monocyte counts, diagnostic AUC for Zika increased from 0.79 to 0.95, with 92% accuracy (88% sensitivity and 93% specificity). The accuracy of our derived indices exceeds that of WHO's and CDC's definitions for Zika case identification, notwithstanding that performance may differ with disease prevalence or population factors. Distinguishing Zika virus from DENV infection on clinical grounds remains daunting, and it will be ideal to validate these derived indices in a prospective patient cohort. Until then, these simple clinical assessments using conjunctivitis and basic blood count parameters will be helpful in regions of the world where both Zika virus and DENV are endemic.