Our study sought to evaluate the usefulness of a positive TT and leucopenia alone and in combination in identifying patients with dengue among children and adults with AFI in an ED of a dengue endemic area. Previous studies examining the use of one or both of these tests have primarily been performed in Asia among children suspected to have dengue and who required hospital admission 
. While direct comparisons with the results of Asian studies are difficult because of differences in disease epidemiology and study design, our study supports the findings of previous studies conducted in Thailand 
, which found that the presence of either leucopenia, a positive TT, or both is helpful in distinguishing between patients with and without dengue. In our study and others 
, leucopenia alone was found to be an especially good indicator of dengue among adults. Future studies examining the predictive value of the tourniquet test for the diagnosis of dengue in adults should evaluate a positive TT in combination with leucopenia.
As described in previous studies, we found that a positive TT alone was specific but not sensitive in distinguishing dengue from other AFI 
. This is especially true when the WHO cut-off of 20 or more petechiae per 2.5 cm2
is used. As in other studies, we chose to maximize detection of positive dengue cases (or the sensitivity of the tourniquet test) by using the cut-off of 10 or more petechiae per 2.5 cm2 
. The presence of either a positive TT or leucopenia correctly identified 94% of patients who had dengue, and the absence of a positive TT or leucopenia was highly predictive of the absence of disease with a NPV >98%. That is, less than 2% of enrolled patients with neither a positive TT nor leucopenia had dengue.
Compared with the data reported from Thailand on the combined performance of the TT and leucopenia in identifying dengue patients, our results showed a lower sensitivity (45% vs.74%) and PPV (52% vs.73%–83%) but similar specificity (94% vs. 86%) 
. Our finding of lower sensitivity most likely reflects differences in study design. These previous studies primarily enrolled hospitalized patients who then received daily TTs until the day of defervescence. Our study participants had a single TT performed at initial presentation to the ED. Previous research has demonstrated that the sensitivity of the TT depends on repeated testing and the timing of the test with respect to the day of illness with sensitivity increasing as a patient nears defervescence 
. Most (68%) of our participants were enrolled within three days of symptom onset so that the sensitivity of the TT found in our study corresponds to the day −4/day −3 values found in the study from Thailand (52% versus 46%–56%) 
. We feel that using values solely from the time of initial patient contact is more useful, as it uses only information that is available to physicians at the time of initial triage.
Given that the study was performed in the setting of a concomitant influenza pandemic (CDC, unpublished data) and we did not restrict enrollment to suspected dengue cases that required hospitalization, our study also provides data on the performance of these triage criteria for dengue during periods of high influenza transmission. The lower PPV in our study largely reflects the effect of the study being conducted among patients with AFI (versus only those suspected of having dengue) at the time of a major outbreak of another AFI. While the specificity and PPV of the combination of tests was lower when cases with pandemic influenza A H1N1 were included, overall performance was still good. Reanalyzing the data after excluding influenza cases resulted in a PPV for the combination of leucopenia and positive TT of 74%, similar to previously published estimates 
. As only 12% of patients without dengue or influenza had leucopenia in our study, there was only a minor decrease in the NPV upon excluding influenza patients for the combination of leucopenia and positive TT (92% vs 84%).
Our study has some important limitations. It was performed at an ED in a tertiary-level referral hospital with a large catchment area, and patients seeking care at this facility are more likely to have severe disease than patients who seek care at lower-level facilities. Thus, dengue cases in our study were likely not representative of the whole spectrum of dengue occurring in Puerto Rico. Data were missing on TT and white blood cell count results for approximately 12% and 2% percent of patients, respectively. The 35 participants who did not have a TT performed did not differ significantly from those who did have a TT performed in terms of sex or median age. No cases of dengue were diagnosed among the patients who did not have a TT performed. In most cases, only one serum specimen was available for dengue testing, leading to the possibility that some true dengue cases were not detected and misclassified as dengue negative. A comprehensive testing strategy, including a highly sensitive, single-plex anti-DENV RT-PCR, a NS-1 assay, and a MAC ELISA, was used to try to limit the amount of this misclassification bias.
The effects of dengue serotype and immunological status (primary versus secondary dengue infection) on the incidence of a positive TT or leucopenia have not yet been sufficiently investigated. Few studies have examined the role of these variables on the performance of these tests and these studies have usually been relatively small. Even fewer studies have had an adequate number of cases with both virological confirmation and immunological data to look at both variables simultaneously and stratify serotype specific results by immunological status, an important potential confounder. A positive TT was more frequently seen in children with dengue in Nicaragua when DENV-1 was the predominant serotype compared to an era when DENV-2 was circulating widely 
. However this difference was not evident in the sub-group of virologically confirmed cases and the study did not stratify between primary and secondary infections. No difference in white blood cell count was seen between 385 adults with DENV-2 or DENV-3 infection in Taiwan 
, but significantly more DENV-3 patients had secondary infections making accurate comparisons between groups difficult. No differences were seen in the proportion of European travelers with a positive TT between patients with a primary or secondary immune response 
, but the small number of patients with a secondary immune response limited the power of the study. No significant differences for TT positivity or leukocyte count in primary and secondary infections were seen among 89 hospitalized DENV-1 patients in Fiji 
Overall our study indicates that a combination of two rapid, widely available tests can assist clinicians in distinguishing dengue from other AFIs that have similar clinical signs and symptoms. Previous investigators have reported that the combination of the TT and leucopenia is more accurate in identifying patients with dengue than the World Health Organization's 1997 clinical case definition 
, and our data supports the contention that few patients with dengue are likely to be missed when these criteria are used. Twenty-nine of the 31 laboratory-positive dengue patients had either a positive TT or leucopenia. The two dengue patients that were not detected by these two tests were thrombocytopenic.
Our study and others 
suggest that the TT may be useful in identifying dengue patients before a major decrease in platelet count, a group for whom dengue is often overlooked as a diagnostic possibility in Puerto Rico. Patients in our study population with AFI who had a negative TT and normal WBC appear to be at low risk of having dengue, and patients with both leucopenia and a positive TT have a high likelihood of having dengue, even in the setting of an outbreak of another clinically similar illness. Increased emphasis should be placed on determining the usefulness of the TT in combination with white blood cell count in identifying patients with dengue in Puerto Rico and elsewhere in the Americas. Further exploration of the sensitivity, specificity and predictive value of these tests by day of illness would be of particular benefit for clinical decision making. Additional larger studies should also be conducted to explore the effect of dengue serotype and immune status on the diagnostic performance of these tests.