Host biomarkers distinguish dengue from leptospirosis in Colombia: a case-control study

Abstract

Background: Dengue fever and leptospirosis have partially overlapping geographic distributions, similar clinical presentations and potentially life-threatening complications but require different treatments. Distinguishing between these cosmopolitan emerging pathogens represents a diagnostic dilemma of global importance. We hypothesized that perturbations in host biomarkers can differentiate between individuals with dengue fever and leptospirosis during the acute phase of illness.

Methods: We randomly selected subjects from a prospective cohort study of acute febrile illness in Bucaramanga, Colombia and tested 19 serum biomarkers by ELISA in dengue fever (DF, n = 113) compared to subjects with leptospirosis (n = 47). Biomarkers were selected for further analysis if they had good discriminatory ability (area under the ROC curve (AUC) >0.80) and were beyond a reference range (assessed using local healthy controls).

Results: Nine biomarkers differed significantly between dengue fever and leptospirosis, with higher levels of Angptl3, IL-18BP, IP-10/CXCL10, Platelet Factor 4, sICAM-1, Factor D, sEng and sKDR in dengue and higher levels of sTie-2 in leptospirosis (p < 0.001 for all comparisons). Two biomarkers, sEng and IL18BP, showed excellent discriminatory ability (AUROC >0.90). When incorporated into multivariable models, sEng and IL18BP improved the diagnostic accuracy of clinical information alone.

Conclusions: These results suggest that host biomarkers may have utility in differentiating between dengue and leptospirosis, clinically similar conditions of different etiology.

Figure 1

Integrating Clinical and Laboratory Data with Biomarker Data Improves Discrimination of Dengue Fever and Leptospirosis. Logistic regression analysis was used to generate two models to discriminate between dengue fever and leptospirosis and the predicted probabilities from those models were plotted using ROC curve analysis. The first model used clinical and laboratory data (clinical model: age, sex, height, duration of illness, leukopenia, rash, dizziness) and had good discriminatory performance with a c-index (equivalent to the AUC) of 0.86 (95% CI: 0.79-0.91). By adding in the biomarker data, we generated a model with excellent discriminatory ability and a c-index of 0.979 (95% CI: 0.94-0.996). The biomarker model (clinical model with IL-18BP, sEng) had a c-index that was statistically higher than that of the clinical model (p = 0.0003).

Figure 2

Integrating Biomarker Data into a Clinical Score Improves Diagnosis of Dengue Fever. A clinical score (from -1 to 2) was created for each study participant by assigning a value of 0 (not present), -1 (more common in leptospirosis) or +1 (more common in dengue) for leukopenia, rash and dizziness. The score was used to create an area under the ROC curve (AUC) of 0.81 (95% CI, 0.73-0.86), p < 0.001. Biomarker data were then integrated into the clinical score (from -1 to 4) by assigning a value of +1 if IL-18BP (>24.5 ng/mL) and sEng (> 9.12 ng/mL) levels were higher than the assigned cut-offs to generate an AUC of 0.96 (95% CI, 0.91-0.98), p < 0.001.

Figure 3

Biomarkers Discriminate Between Dengue Fever and Leptospirosis. Aligned dot plots and median of serum biomarker levels in dengue fever (n = 113) and leptospirosis (n = 47) measured at time of presentation during the acute phase of febrile illness. A population derived healthy range for adults in Bucaramanga, Colombia (n = 15) is represented by the shaded area with the median and 5-95% shown by the horizontal limits. All biomarkers were significantly different between cases with dengue fever and controls with leptospirosis (p < 0.001) following Bonferonni adjustment for multiple comparisons (19 pair-wise comparisons).