The predictive and diagnostic accuracy of vascular endothelial growth factor and pentraxin-3 in severe dengue

Abstract

This study aimed to evaluate vascular endothelial growth factor (VEGF) and pentraxin 3 (PTX-3) as predictive and diagnostic markers in differentiating severe dengue from non-severe dengue. The study was conducted in Ampang Health Clinic, Ampang Hospital and Serdang Hospital. The plasma levels of VEGF and PTX-3 were compared between severe dengue and non-severe dengue by ELISA from the day of presentation until discharged. Multiple logistic regression was used to develop predictive and diagnostic models by incorporating other clinical parameters. The receiver operating characteristics (ROC) analysis was used to assess the accuracy of the biomarkers and the developed models. Eighty-two patients were recruited, 29 with severe dengue and four died. The Area Under the Curve (AUC) was statistically significant in VEGF as diagnostic marker at Day 2 and 3 of illness with sensitivity of 80.00%-100.00% and specificity of 76.47%-80.00%. The predictive model with AUC of 0.84 (p < 0.01) has a sensitivity of 100.00% and specificity of 79.25% for predicting severe dengue. The diagnostic model with AUC of 0.71 (p < 0.01) has a sensitivity of 76.19% and specificity of 73.58% for diagnosing severe dengue. The AUC for PTX-3 was not statistically significant. VEGF may be used in combination with other clinical parameters to predict the severity of the disease. As a single biomarker, it may be used as an adjunct investigation to support the diagnosis of severe dengue. PTX-3 was not able to differentiate severe dengue from non-severe dengue.

Diagnostic accuracy and predictive value in differentiating the severity of dengue infection

“…A total of 26 included studies conducted between 1994 and 2017 were published between 2008 and 2018 . However, two studies were related, leaving a total of 25 studies of different settings .…”

“…Two studies clearly indicated that the analysis was prior to the diagnosis of SD: Low 2018 and Carrasco 2014 . Thus, these two studies could be considered as predictive models.…”

“…Thus, these two studies could be considered as predictive models. Accuracy was highest with 86% in Low 2018 with its logistic regression model predicting SD in the first three days of illness (Figure ). The sensitivity was 100% and the specificity 79.3% (Table ).…”

“…The models of Low 2018 and Nguyen 2017, developed in reference to the WHO 2009 case classification, were promising with an accuracy of 86% they could predict the SD, which has better sensitivity and specificity in comparison with the ‘warning signs’ . The current recommended prediction is by using ‘warning signs’, which has similar sensitivity but poor specificity of <50% .…”

“…Several methods have been evaluated to improve the accuracy of the currently used WHO 2009 classification, all of which involve statistical modelling . To our knowledge, there is no review to summarise and recommend a statistical model as a diagnostic tool for severity of dengue infection.…”