This was a prospective diagnostic cross-sectional study that evaluated accuracy of the NS1 Bioeasy™ rapid immunochromatographic test and inter- and intra-observer reliability in whole blood specimens at point of care (POC) and serum specimens in the laboratory. The study was conducted according to the Standards for Reporting Studies of Diagnostic Accuracy (STARD) .
The study population consisted of patients over 18 years of age with up to 4 days of acute febrile syndrome without an established diagnosis, treated consecutively and by spontaneous demand at an emergency hospital of the Resende Municipal Health Secretariat, Rio de Janeiro State, Brazil, during a dengue epidemic in March 2015.
The selected patients answered a questionnaire on clinical signs and symptoms associated with dengue according to WHO 2009 criteria for Dengue diagnosis [4, 23]. All of them signed the free and informed consent form.
Three blood specimens were drawn in vacuum tubes by a nurse and four medical students. Prior asepsis was performed with 70% alcohol, followed by brachial venipuncture with a 25 × 7 mm BD Vacutainer® needle. One tube with EDTA K2 was sent to the hospital laboratory for performing a complete blood count, another with heparin for ICT with whole blood at POC and the third with clotting activator and gel separator for obtaining serum. The latter was transported to the Flavivirus Laboratory, where it remained frozen at −70 °C for subsequent characterization of the specimen and ICT with serum.
The ICTs with whole blood were performed at point of care in the hospital by a nurse, and the ICTs with serum by a biologist and the same nurse (that performed ICT with blood at POC), in the Flavivirus Laboratory of the Oswaldo Cruz Institute, FIOCRUZ, the regional reference laboratory in Rio de Janeiro State. The observers were blinded to the reference test.
The index test used was Dengue Eden Test NS1 ICT from the Bioeasy™ company. This is a rectangular cassette with an orifice to add whole blood, serum, or plasma, and which contains an immunochromatographic membrane coated with NS1 antigen on the test line (T) and a window to view the result. When only the control line (C) is visible, the test is considered negative; if lines C and T are visible, the result is positive; and when no line or only the T line is visible, the test result is invalid .
The tests were performed according to manufacturer’s instructions and with readings at 15 and 30 min. Tests with invalid results at 15 min were reread at 30 min.
All samples were tested using Reverse Transcription Polymerase Chain Reaction (RT-PCR) for dengue and Platelia™ Dengue NS1 Ag-ELISA (Bio-Rad™ Laboratories, France) as reference, performed according to the protocol described by Lanciotti et al.  and the manufacturer’s specifications, respectively . Dengue cases were defined by a positive result in at least one of them and non-dengue by negative results in both. Positive specimens according to the reference test were tested by Dengue Serion ELISA classic dengue virus IgM and IgG tests (Virion/Serion, Würzburg, Germany) , according to manufacturers’ specifications for characterization of primary and secondary cases, respectively.
Non-dengue patients were also tested for Zika virus with RT-PCR  to investigate co-circulation of these viruses.
The reference tests were performed at the Flavivirus Laboratory by two biologists blinded to the index test.
The qualitative variables were described by simple frequencies and the quantitative variables by the median with interquartile range (IQR) or minimum and maximum. Shapiro-Wilk test showed rejection of normality for the quantitative variables, indicating the use of a non-parametric test. Verification of association in the dengue and non-dengue groups used the Mann-Whitney non-parametric test and Pearson’s chi-square test for quantitative and qualitative variables, respectively.
To assess the performance of the index test, dengue NS1 ICT, in each setting using whole blood at point of care (POC) and serum in the laboratory in readings at 15′ and 30′, compared to the standard reference, the following measures of accuracy were considered: sensitivity (Sn), specificity (Sp), positive and negative predictive values (PPV, NPV), and positive and negative likelihood ratios (LR+) and (LR-) with the respective 95% confidence intervals (95% CI). Post-test probabilities were calculated for scenarios with prevalence rates of 30%, 50%, and 83.3%, using the Fagan nomogram .
The sensitivity and specificity of the index test in different settings were compared for using the χ2 McNemar test .
Reliability was assessed by the variability of the ICT results in whole blood/POC and serum/laboratory with readings at 15 min. Inter-observer reliability was analyzed by the readings done by the two professionals (nurse and biologist) and intra-observer reliability by the same professional (nurse), in blinded fashion. The proportions of positive agreement (Ppos) and negative agreement (Pneg) were calculated, as well as simple and prevalence-adjusted bias-adjusted kappa (PABAK) coefficients with respective 95% CI . Kappa values (k) were interpreted according to the classification by Landis & Koch : poor (k < 0.0), slight (0.0 < k < 0.2), fair (0.2 < k < 0.4), moderate (0.4 < k < 0.6), substantial (0.6 < k < 0.8), almost perfect (0.8 < k < 1.00), and perfect agreement (k = 1).
Statistical significance was set at p < 0.05.
The software packages used were R Commander 3.2.1, WinPepi 11.50, and MedCalc 15.8 [32,33,34].