rRT-PCR had the best operating characteristics (sensitivity 89%, specificity 96%, PPV 94%, NPV 92%) and would be potentially sufficient

as a single assay for confirmation of dengue infection, since it allows for accurate confirmation or refuting of infection. The combinations of NS-1+rRT-PCR or NS-1+IgM+rRT-PCR resulted in the highest sensitivity (93%), although this was associated with an inevitable fall in specificity (96% and 83% respectively). Compared to previous LDK378 concentration studies on NS-1 antigen ELISA we report a slightly lower sensitivity. Dussart et al. found the Panbio NS-1 antigen ELISA to have a sensitivity of 60% when used on stored serum specimens from French Guiana14 and, in a similar Veliparib order study from Puerto Rico, Bessoff et al. reported a sensitivity of 65%.13 On prospectively collected specimens from clinically suspected dengue cases in Laos, Blacksell et al. reported a sensitivity of 63%.24 The sensitivity of rRT-PCR was slightly better than reported by the original authors who found that PCR detected viral RNA 83% of acute specimens from patients with confirmed dengue.11 Comparing operating characteristics of assays between studies can be difficult, since there are many potential confounding factors. Firstly, in the current study, specimens were collected prospectively on patients with illness broadly compatible with dengue whereas several of the previous

evaluations of NS-1 antigen ELISA have been retrospective, using well characterised serum specimen collections. We feel that the results presented here are likely to more accurately reflect the operating characteristics of the tests in a routine clinical setting. Secondly,

infections due to dengue serotype 3 predominated in our study, and previous work has noted that the Panbio NS-1 antigen ELISA may miss infections caused by this serotype.24 Thirdly, timing of presentation and specimen collection may affect assay performance: in our study, most patients presented very early in the course Cyclic nucleotide phosphodiesterase of their infection. Although we demonstrated trends in the sensitivity of each assay, the small number of patients presenting with more than three days of fever limited our ability to perform statistical analysis. Previous studies have demonstrated the effect of timing of presentation on NS-1 antigen and IgM antibody24 or PCR11 assays, but no comparison between antigen detection, PCR, and serology on the same patient population has been described. Finally, infection status (primary infection versus secondary infection) may also make study-to-study comparisons difficult. We identified very few patients with acute primary infection (3/72, using Panbio kit criteria), resulting in an inability to determine potential differences in test characteristics between primary and secondary infections. We plan to perform further work to delineate the optimum sampling ‘window’ for each assay for patients with primary and secondary dengue infection.