Continuous variables were compared using Wilcoxon rank sum test (since non-normally distributed). Characteristics of the tests were analysed in 2 x 2 tables using the ‘diagt’ routine.20 The McNemar test was used to compare sensitivities of the tests. Two hundred and twenty nine patients were examined. All patients met the WHO clinical case definition for acute dengue infection.21 One hundred and sixty two patients (71%) returned for the follow-up specimen collection visit and were considered further. Of the 162 patients

included in the current analysis, 72 patients (44%) were given a laboratory confirmed diagnosis of dengue infection by paired serology (Table 1). Four patients were found to have evidence of recent dengue infection. The demographic and clinical data of the patients are shown in Table 2. Patients with confirmed dengue infection had lower white cell counts (4.8 vs 7.2, P<0.0001) ICG-001 chemical structure and platelets (147 vs. 162, P=0.03) than patients with non-dengue infection. Twenty six patients (16%) were found to have non-dengue causes for their illness: four patients (2%) grew Salmonella typhi from blood cultures, by serology nine patients (6%) had murine typhus, seven (4%) had scrub typhus, and six (4%) had leptospirosis.

Only one patient had evidence of dual infection: acute secondary dengue and typhoid; this patient had positive NS-1 antigen and dengue rRT-PCR and also grew Salmonella typhi from a blood KU-57788 cell line culture. The serotype of Casein kinase 1 dengue virus was sought by performing nested RT-PCR on the acute specimens from the 72 serologically

confirmed cases. Seventy one of the cases (99%) were serotype 3, with the remaining case serotype 2. The four patients with serological evidence of recent dengue infection were negative in the nested RT-PCR. The performance characteristics of NS-1 antigen detection, rRT-PCR, and IgM antibody detection in the acute plasma specimen against our ‘gold standard’ of paired serology are shown in Table 3. NS-1 antigen or IgM antibody test alone had low sensitivity compared with the paired serology result, 54% and 17% respectively. rRT-PCR sensitivity was 89%, significantly higher than both NS-1 antigen and IgM antibody tests (P<0.00001). Specificities of NS-1 antigen detection, rRT-PCR, and IgM antibody detection were 100%, 96% and 88% respectively. The effect of fever duration at presentation on assay sensitivity is shown in Figure 1. NS-1 antigen sensitivity peaked in the early stages of fever (three days of fever at presentation). IgM antibody sensitivity rose later, peaking in patients presenting with five days of fever. The sensitivity of rRT-PCR remained high throughout. However, as a result of the relatively small number of specimens on each day, particularly for four and five days of fever, the 95% confidence intervals around the sensitivities are wide. The performance characteristics of combinations of the acute specimen tests are shown in Table 3 (combined by an ‘OR’ operator).