Multicenter Pivotal Clinical Trial of Urine Malaria Test for Rapid Diagnosis of Plasmodium falciparum Malaria

ABSTRACT The need to expand malaria diagnosis capabilities alongside policy requirements for mandatory testing before treatment motivates exploration of noninvasive rapid diagnostic tests (RDTs). We report the outcome of the first cross-sectional, single-blind clinical performance evaluation of a urine malaria test (UMT) for diagnosis of Plasmodium falciparum malaria in febrile patients. Matched urine and finger-prick blood samples from participants ≥2 years of age with fever (axillary temperature of ≥37.5°C) or with a history of fever in the preceding 48 h were tested with UMT and microscopy (as the gold standard). BinaxNOW (Pf and Pan versions) blood RDTs were done to assess relative performance. Urinalysis and rheumatoid factor (RF) tests were conducted to evaluate possible interference. Diagnostic performance characteristics were computed at 95% confidence intervals (CIs). Of 1,800 participants screened, 1,691 were enrolled; of these 566 (34%) were febrile, and 1,125 (66%) were afebrile. Among enrolled participants, 341 (20%) tested positive by microscopy, 419 (25%) were positive by UMT, 676 (40%) were positive by BinaxNOW Pf, and 368 (22%) were positive by BinaxNow Pan. UMT sensitivity among febrile patients (for whom the test was indicated) was 85%, and specificity was 84%. Among febrile children ≤5 years of age, UMT sensitivity was 93%, and specificity was 83%. The area under the receiver-operator characteristic curve (AUC) of UMT (0.84) was not significantly different from that of BinaxNOW Pf (0.86) or of BinaxNOW Pan (0.87), indicating that the tests do not differ in overall performance. Gender, seasons, and RF did not impact UMT performance. Leukocytes, hematuria, and urobilinogen concentrations in urine were associated with lower UMT specificities. UMT performance was comparable to that of the BinaxNOW Pf/Pan tests, making UMT a promising tool to expand malaria testing in public and private health care settings where there are challenges to blood-based malaria diagnosis testing.

T he current WHO recommendation for parasitological confirmation of suspected malaria cases prior to treatment is a paradigm shift from entrenched presumptive diagnostic practices. Annually, there are 214 million estimated cases of malaria globally (1). Despite increasing parasitological diagnosis capabilities in countries of endemicity, universal access to testing in cases of fever suspected of being induced by malaria remains a major challenge, particularly in community and private health care settings. This is due in part to limited access to quality-assured microscopy, laboratory infra-Performance of the UMT. (i) Febrile participants. UMT sensitivity among the 566 febrile participants was 85% (95% confidence interval [CI], 79, 89), and specificity was 84% (95% CI, 80, 88) ( Table 3). Among febrile children of Յ5 years of age, UMT sensitivity was 93% (95% CI, 80, 98), and specificity was 83% (95% CI, 75, 89). Compared to UMT, BinaxNOW Pf sensitivity was high (99% [95% CI, 97, 100]), while specificity was low (69% [95% CI, 64, 74]) (both P Ͻ 0.001). The sensitivity and specificity of BinaxNOW Pan did not differ significantly from those of UMT, while PPVs and NPVs did not differ significantly between UMT and either of the BinaxNOW tests. The negative likelihood ratio (NLR) for UMT differed from that of BinaxNOW Pf and was closer to that of BinaxNow Pan, while the positive likelihood ratio (PLR) for UMT was also closer to that of BinaxNow Pan than to that of BinaxNow Pf (Table 3).
(iii) Overall study population (febrile and afebrile). Overall UMT sensitivity and specificity for all participants (febrile and afebrile combined) were 79% (95% CI, 75, 84) and 89% (96% CI, 87, 91), respectively. UMT sensitivity did not differ from that of BinaxNOW Pan (80%) but was significantly lower than that of BinaxNow Pf (98% [95% Other performance characteristics. Graded PDs by microscopy compared with UMT performance showed that increased PD was associated with increased probability of fever (P Ͻ 0.001). UMT, BinaxNOW Pf, and BinaxNOW Pan sensitivities increased with higher PDs among all patients, as well as with stratified subpopulations of febrile and afebrile patients (Table 5). While BinaxNOW Pf sensitivity was highest across PDs and stratified subpopulations, the UMT and BinaxNOW Pan specificities were higher than the specificity of BinaxNOW Pf. The sensitivities of all tests decreased with increasing age and increased with higher PDs. In contrast, specificities of all tests were significantly higher in adults than in younger participants in the overall study population (all, P Ͻ 0.001). The PPVs of UMT did not differ between age categories, but while NPVs of children and adolescents were significantly lower than those of adults (P Ͻ 0.001), all values were Ն90% in each category. Upon stratification by fever, the majority of the performance characteristics of the three tests did not differ significantly between the oldest and youngest participants.
Gender and season (dry and rainy) did not affect UMT performance. Urinalysis revealed that a higher number of leukocytes (Ͼ15) was associated with a significantly higher specificity (P ϭ 0.010), while hematuria was associated with significantly lower specificity (P ϭ 0.001). UMT specificity was lower among participants with urobilinogen concentrations of Ն1 mg/dl than in those with levels of Ͻ1 mg/dl (P Ͻ 0.001). Specific gravity, which indicates urine concentration, was not associated with UMT performance. All of the 15 rheumatoid factor-positive (RF ϩ ) participants tested negative for malaria and were negative by UMT.

PCR correction of microscopy results.
The original microscopy results showed that 365 individuals were infected with Plasmodium parasites: 362 (99%) were infected with P. falciparum, and 3 (1%) were infected with P. malariae. A further analysis of these samples by PCR showed that seven study samples originally found to harbor P. falciparum by microscopy were negative by PCR. Additionally, 58 microscopy-negative samples were positive by PCR; of these 56 were infected with P. falciparum, and 2 were infected with P. malariae. The PCR results of one participant sample indicated infection with Plasmodium ovale, but the source of the infection was originally identified by microscopy as P. falciparum.
The six participants infected with non-P. falciparum parasites as determined by PCR were removed from the data set. There were no statistically significant differences between original and PCR-corrected microscopy results of discordant samples by UMT and BinaxNOW Pf/Pan, likely reflecting high-quality microscopy methods and welltrained microscopists ( Table 6).
The original microscopy results indicated that 361 study participants were infected with P. falciparum, while PCR-corrected microscopy results indicated that 410 participants had P. falciparum malaria, and the urine malaria test indicated that 451 partici- pants were infected with P. falciparum. UMT sensitivities, specificities, and predictive values changed by less than 3% regardless of whether all study participants, febrile participants, or afebrile participants were assessed, by between 0 and 7% for BinaxNOW Pf, and by between 0 and 11% for BinaxNOW Pan. Overall, the largest changes occurred  among the positive predictive values (PPVs), but this was expected as a mathematical relationship between prevalence and PPV value is known, whereby increasing prevalence always produces an increased PPV. Likelihood ratios also did not demonstrate clinically meaningful changes. In contrast to microscopy results, those for sensitivity, specificity, and negative predictive values changed by less than 3% regardless of which participant population, total, febrile, or afebrile, was assessed. The largest changes occurred among the positive predictive values (PPVs), with differences between the original and PCRcorrected microscopy results ranging from 3% to 12%. However, this was expected as a mathematical relationship between prevalence and the PPV value is known, whereby increasing prevalence always produces an increased PPV. Likelihood ratios also did not demonstrate clinically meaningful changes.

DISCUSSION
The UMT is the first noninvasive malaria test clinically evaluated on a large scale at the population and community levels. Given its sensitivity and specificity, a disease prevalence of 20% in the study population, and a PLR of 7.2, the posterior probability of disease in a patient who tests positive by the UMT is 59%. Given the NLR of 0.23, the posterior probability of disease in a patient who tests negative by UMT is 4%. This indicates that the UMT could aid in the clinical management of suspected malaria cases. For example, upon receiving a negative UMT result for a suspected malaria case, the clinician now knows that this patient's probability of having malaria parasites detectable by microscopy is unlikely, i.e., only 4%. In contrast, a positive UMT result would indicate that the probability of detecting malaria parasites in this patient by microscopy is 59%.
The performance of UMT was higher among febrile participants (who are the intended use population) than among afebrile individuals. High rates of false positives and lower specificities reported among HRP-2-based blood RDTs corroborate our findings (15)(16)(17)(18)(19)(20) and are worrisome in spite of their high sensitivity. Broader clinical utility of a rapid malaria test should be based on its overall diagnostic value using the AUC of the receiver-operator characteristics. Our data support a role for the UMT as a diagnostic or screening tool at the control phase in countries with high to medium malaria transmission rates. However, test performance may be lower in lowtransmission regions or preelimination countries, as reported for blood RDTs (21)(22)(23)(24)(25). Nevertheless, field evaluation studies in different transmission settings would provide performance data on UMT to demonstrate its utility.
It should be noted that P. falciparum parasitemia and antigenemia do not correlate directly due to sequestration, while there are no data on the correlation between parasitemia and antigenuria. Therefore, UMT performance metrics must be established de novo and not on the basis of antigenuria reported by RDTs.
The availability of the blood-based RDTs has greatly increased malaria testing particularly in the public sector, where about 53% of patients with malaria-like symptoms are tested. In the formal and informal private sector health care settings, only 36% and 6%, respectively, of patients are tested (26). The UMT could potentially expand malaria testing in private health care settings, particularly in hard-to-reach locations or health care facilities where blood draw is difficult or impractical for microscopy, and advance the current global effort toward universal diagnosis in cases of fever suspected of being malaria.
Viral hemorrhagic fevers (such as Ebola viral disease [EVD]) are major global health concerns where the UMT could be invaluable. The Ebola viral proteins connected with infection are present in blood and other body fluids at the time of fever and onset of symptoms (27,28) but are rarely found in patient urine (29,30). Thus, in outbreak situations, the UMT could potentially offer a safer and rapid diagnostic tool to rule out malaria as the cause of fever, particularly in rural communities with severely limited laboratory facilities.
MBFs. Exact binomial confidence intervals were calculated for sensitivity, specificity, and positive and negative predictive values (PPVs and NPVs, respectively). McNemar's test was used to compare sensitivities and specificities between diagnostic tests. The weighted generalized score method was used to compare predictive values (37). The chi-square test was used to compare the number of participants testing positive by each rapid test across categorical variables. Fisher's exact test was used to compare the number of participants testing positive by each rapid test across dichotomous variables when appropriate. A Cochran-Mantel-Haenszel test was used to compare the odds of detecting malaria parasites by microscopy between 24-and 48-h fever durations among participants who tested positive by each rapid test. All analyses were conducted in R, version 3.0.1 (http://www.R-project.org/), using the packages epicalc, pROC, and ROCR.