Predictive value of proteinuria in adult dengue severity

Abstract

Background: Dengue is an important viral infection with different presentations. Predicting disease severity is important in triaging patients requiring hospital care. We aim to study the value of proteinuria in predicting the development of dengue hemorrhagic fever (DHF), utility of urine dipstick test as a rapid prognostic tool.

Methodology and principal findings: Adult patients with undifferentiated fever (n = 293) were prospectively enrolled at the Infectious Disease Research Clinic at Tan Tock Seng Hospital, Singapore from January to August 2012. Dengue infection was confirmed in 168 (57%) by dengue RT-PCR or NS1 antigen detection. Dengue cases had median fever duration of 6 days at enrollment. DHF was diagnosed in 34 cases according to the WHO 1997 guideline. Dengue fever (DF) patients were predominantly younger and were mostly seen in the outpatient setting with higher platelet level. Compared to DF, DHF cases had significantly higher peak urine protein creatinine ratio (UPCR) during clinical course (26 vs. 40 mg/mmol; p<0.001). We obtained a UPCR cut-off value of 29 mg/mmol based on maximum AUC in ROC curves of peak UPCR for DF versus DHF, corresponding to 76% sensitivity and 60% specificity. Multivariate analysis with other readily available clinical and laboratory variables increased the AUC to 0.91 with 92% sensitivity and 80% specificity. Neither urine dipstick at initial presentation nor peak urine dipstick value during the entire illness was able to discriminate between DF and DHF.

Conclusions: Proteinuria measured by a laboratory-based UPCR test may be sensitive and specific in prognosticating adult dengue patients.

Figure 1. Box plots of daily UPCR for dengue fever (DF) (A), dengue hemorrhagic fever (DHF) before DHF onset by defervescence day (B) and by DHF onset day (C).

Black bars in the boxes indicate daily medians whereas blue diamonds indicate daily means of UPCR. The length of whiskers are 1.5 times the interquartile range from the box, the dots which are out of the whiskers are the outliers.

Figure 2. Receiver-operating characteristic curves of prediction of DHF by peak UPCR only (A), the logistic regression model using UPCR at initial presentation adjusted by age and illness day (B) and the logistic regression model using white blood cell count, serum hematocrit, platelet count, serum protein, bleeding and UPCR at initial presentation (C).

By maximizing sensitivity and specificity, the peak UPCR cut-off of 29 mg/mmol yielded sensitivity 76% and specificity 60%. From logistic regression model using UPCR at initial presentation adjusted by age and illness day, the maximum sensitivity and specificity were 76% and 76.9%. The maximized sensitivity and specificity of logistic regression model using white blood cell count, serum hematocrit, platelet count, serum protein, bleeding and UPCR at initial presentation were 91.7% and 79.6%.

Figure 3. Time course analysis of proteinuria for DF and DHF.

Overall means are indicated as solid lines with 95% credible intervals as dashed lines. The red bar on X-axis indicated days with a “significant” difference between DF and DHF. The blue line indicated UPCR level of 29 mg/mmol.