Evaluation of diagnostic tests for Wuchereria bancrofti infection in Brazilian schoolchildren.

INTRODUCTION
Since the launch of the Global Programme to Eliminate Lymphatic Filariasis, more than 70% of the endemic countries have implemented mass drug administration (MDA) to interrupt disease transmission. The monitoring of filarial infection in sentinel populations, particularly schoolchildren, is recommended to assess the impact of MDA. A key issue is choosing the appropriate tools for these initial assessments (to define the best intervention) and for monitoring transmission.


METHODS
This study compared the pre-MDA performance of five diagnostic methods, namely, thick film test, Knott's technique, filtration, Og4C3-ELISA, and the AD12-ICT card test, in schoolchildren from Brazil. Venous and capillary blood samples were collected between 11 pm and 1 am. The microfilarial loads were analyzed with a negative binomial regression, and the prevalence and associated 95% confidence intervals were estimated for all methods. The accuracies of the AD12-ICT card and Og4C3-ELISA tests were assessed against the combination of parasitological test results.


RESULTS
A total of 805 schoolchildren were examined. The overall and stratified prevalence by age group and gender detected by Og4C3-ELISA and AD12-ICT were markedly higher than the prevalence estimated by the parasitological methods. The sensitivity of the AD12-ICT card and Og4C3-ELISA tests was approximately 100%, and the positive likelihood ratios were above 6. The specificity of the Og4C3-ELISA was higher than that of the AD12-ICT at different prevalence levels.


CONCLUSIONS
The ICT card test should be the recommended tool for monitoring school-age populations living in areas with ongoing or completed MDA.

The Global Programme to Eliminate Lymphatic Filariasis (GPELF), launched by the World Health Organization (WHO) in 1997, has targeted the elimination of the disease as a public health problem by 2020.Of the 73 endemic countries, 56 have implemented mass drug administration (MDA) programs to interrupt disease transmission.By 2012, more than 4.4 billion doses of antifi larial drugs had been delivered to almost 984 million people worldwide 1 .
Effective monitoring and evaluation of the impact of MDA programs are important for achieving the goal of disease elimination.A number of diagnostic tools designed to assess fi larial infection are standardized and currently recommended by the WHO 1 .These tools include both parasitological methods for detecting microfi laremia and tests for detecting the circulating antigens of Wuchereria bancrofti, such as the immunochromatographic test (AD12-ICT card test) and enzyme-linked immunosorbent assay (Og4C3-ELISA) 1 .The choice of which diagnostic tools to use is usually driven by characteristics such as accuracy (sensitivity and specifi city), feasibility of use in the fi eld, technical skills required, and cost 1 .It is critical to defi ne the most appropriate diagnostic tools for the initial assessments of lymphatic fi lariasis (LF) infection in the population (to defi ne the best intervention), as well as to monitor transmission in areas undergoing MDA, because the performance of such tools can vary substantially according to the epidemiologic and demographic characteristics of the study population 2,3 .
The WHO has recommended the monitoring of fi larial infection in sentinel populations, particularly schoolchildren [4][5][6] , due to children's lower duration of exposure (many of these children were born following or during the start of the intervention), which allows for a more accurate assessment of transmission 5 .However, the pediatric population usually exhibits lower microfilarial loads and higher proportions of amicrofi laremic and asymptomatic infections compared with adults [7][8][9] .As these features will most likely infl uence the operating characteristics of diagnostic tests, it is important to evaluate them in children.This study evaluated the pre-MDA performance of several diagnostic methods, namely, the thick fi lm test, Knott's concentration method, fi ltration,

METHODS
Og4C3-ELISA, and the AD12-ICT card test, in children from an endemic area of Brazil prior to the initiation of an MDA program.

Study setting
The study was conducted in the three neighborhoods of the City of Olinda, Metropolitan Region of State Pernambuco, Brazil, with the highest prevalence of microfi laremia, according to a survey conducted in 1998 (unpublished data), between 2007 and 2010.The city has an area of 41,659km 2 and a population of 377,779, as reported in the last census 10 , and is considered one of the remaining foci of fi lariasis in Brazil 11,12 .
Children between the ages of 4 and 15 years who attended fi ve public schools in these neighborhoods (Alto da Conquista, Alto da Bondade, and Sapucaia) participated in this study.

Data collection
Parents or guardians were informed of the main study objectives, and consent was sought for the children's participation.After obtaining written consent, each child's data (name, address, age, and gender) were obtained using a standardized form.The blood samples were obtained between 23:00 and 1:00.Initially, capillary blood samples were obtained by fi nger prick to perform the thick blood fi lm (~60μL of blood) and the AD12-ICT card (100μL) tests.A venous blood sample (7mL) was then collected to perform the fi ltration, Knott's concentration, and Og4C3-ELISA tests.This sample was partitioned into tubes containing ethylenediamine tetraacetic acid (EDTA), tubes containing 2% formalin, and dry tubes to perform the fi ltration, Knott's concentration test, and Og4C3-ELISA test, respectively.The blood samples were transported in a refrigerated container to the laboratory, stored at 2-8°C, and processed on the following day.

I) Parasitological methods: A) Thick blood fi lm (TBF) test:
The smear was left to dry overnight at room temperature.The slides were then dehemoglobinized in water, fi xed in methanol for 3 min, stained according to previously described procedures, and read with an optical microscope 13 .B) Knott's concentration method: Venous blood (~1mL) was drawn into a tube containing 9mL of 2% formalin and processed according to the methodology described by Knott 14 .C) Filtration technique: Blood (~1mL) was fi ltered through a polycarbonate membrane with a width of 13mm and pore size of 3µm, as described by Dennis and Kaen 15 .These three parasitological tests rely on the microscopic visualization of microfi lariae.Thus, their specifi city, although not their sensitivity, can be considered to be very high.II) Immunological methods: A) AD12-ICT card test: This test detects circulating fi larial antigen (CFA) using the monoclonal antibody AD12, which recognizes a 200-kDa fi larial antigen from either adult worms or microfi lariae 16 .The test was performed according to the manufacturer's instructions and read by trained technicians in the fi eld after 10 min.The visualization of the two lines (test and control) was interpreted as a positive result.B) Og4C3-ELISA: The test was performed as recommended by the manufacturer (TropBio®, Townsville, Australia).Samples with an antigen concentration ≥128mL units (UA) were interpreted as positive 17 .This test detects CFA using a monoclonal antibody against antigens of the bovine parasite Onchocerca gibsoni, which has no known cross-reactivity with other human helminthes.The test provides quantitative results that enable the observation of variations in antigenemia levels after treatment [18][19][20][21][22][23][24] .

Data analysis
The reading and interpretation of all laboratory tests were performed by observers who were blinded to the other test results.Data entry and analysis were performed using EpiInfo and R software.Mean microfi larial densities were estimated and compared with a negative binomial regression 25 .The parasite load, determined by the three parasitological techniques (thick smear, Knott's concentration, and membrane fi ltration), and antigenemia, determined by the Og4C3-ELISA in terms of unit antigens (UA), were stratifi ed by age group and gender.Antigenemia was compared between groups with a Kruskal-Wallis test.The mean antigenemia was quoted for descriptive purposes only, as most subjects were negative and, therefore, the median values were zero.
The infection prevalence identifi ed by each technique was stratifi ed by age group and gender.Statistical comparisons between groups were made using chi-square (χ2) and Kruskal-Wallis tests for qualitative and continuous results, respectively.A p-value of 0.05 was considered statistically signifi cant.
The sensitivity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR+), negative likelihood ratio (LR-), and their respective 95% confi dence intervals (CIs) for the AD12-ICT card test and Og4C3-ELISA were estimated against the gold standard of the combination of the parasitological techniques (thick blood fi lm, Knott's concentration, and fi ltration).The LR+ measures how many times more likely it is that a positive result is obtained for individuals with the disease than for individuals without the disease, whereas the LR-reveals how many times more likely it is that a negative result is obtained for individuals with the disease than for individuals without the disease.Values of LR+ above 10 and LR-below 0.1 provide strong evidence to confi rm or rule out the diagnosis of disease, respectively 26 .
The parasitological tests are assumed to have 100% specifi city but are of unknown sensitivity, as many individuals can harbor adult worms without demonstrating microfi laremia 12 .Consequently, the sensitivity of the ICT card test and the Og4C3-ELISA can be estimated, but not their specifi city.Because the specifi city cannot be directly evaluated, we estimated the specifi cities of the Og4C3-ELISA and AD12-ICT card tests as a function of the prevalence of fi lariasis using the equation derived by Staquet et al. 27 .This calculation allows for the comparison of the specifi cities of the Og4C3-ELISA and AD12-ICT card tests for different prevalences 27 of samples with viable worms (adult or microfi laria) or circulating antigen.The performances of the AD12-ICT card and thick fi lm tests were compared according to sections 2.2 and 3.2.1 of Hayen et al. 28 , with the fi ltration test used as the reference standard.There are eight possible combinations of the three tests, but we considered only the four cells in which the fi ltration test is positive.Due to the high specifi city of fi ltration, these cells can reliably be considered to be truly positive.The Table 1 was adapted from Table 2 of Hayen et al. 28 .The true positive fractions in the AD12-ICT card and thick blood fi lm tests can be estimated as (r + t)/n + and (r + s)/n + , respectively, and the McNemar test can be used to assess the difference between these values.

Ethical considerations
This study was approved by the Ethics Committee of the Aggeu Magalhães Research Center (CAEE: 0069.0.095.000-06).All of the participants with a positive result in any of the tests were treated and underwent clinical and laboratory examinations.
A total of 805 schoolchildren participated in the study and agreed to be tested via fi nger prick.Overall, 783 children were tested by the thick blood fi lm test and 797 by the AD12-ICT card test.Consent for taking venous blood samples was obtained from only 554 participants, all of whom were examined by Knott's concentration method, 546 by the fi ltration technique, and 545 by the Og4C3-ELISA test.In 487 schoolchildren, fi larial infection was assessed with all laboratory tests.
The microfi laria (MF) loads ranged from 0 to 216 MF/60μL, 0 to 1,423 MF/mL, and 0 to 1,834 MF/mL according to the thick blood fi lm, Knott's concentration, and fi ltration tests, respectively.No statistically signifi cant variation in MF load was observed between age groups or genders when assessed by any of these three methods (Table 2).The mean values of fi larial antigenemia also did not differ according to age group or gender (Table 2).
The estimated prevalence of fi larial infection according to both fi larial antigen tests (overall and stratifi ed by age group and gender) was similar and markedly higher than the prevalence estimated by parasitological methods.This difference in estimates was more pronounced in the younger age group (4-9 years), in whom the prevalence of filarial infection estimated by the ICT and Og4C3-ELISA tests was approximately fi ve times higher than that obtained by parasitological methods (thick blood fi lm, Knott's concentration, or fi ltration test) (Table 3).
The MF prevalence estimated by either Knott's (4.6%; 95%CI: 3.1-6.8)or the fi ltration (5.2%; 95%CI: 3.5-7.5%)method was slightly higher than the prevalence estimated by the thick fi lm method.The prevalence of fi lariasis according to the parasitological methods was signifi cantly higher in the 10-15-year-old age group than in the 4-9-year-old age group, whereas no difference between age groups was found by the immunological methods (AD12-ICT card test and Og4C3-ELISA) (Table 3).The prevalence of fi larial infection was higher in boys than in girls according to all techniques, although a statistically signifi cant difference in the prevalence between the groups was observed only with the AD12-ICT card test (Table 3).
The comparative analysis of the performance of the AD12-ICT card test and Og4C3-ELISA against the gold standard (the combination of all parasitological test results) showed that both immunological tests exhibited 100% sensitivity.The PPVs were approximately 30%, and the NPVs were close to 100%.The LR+ was 6.89 and 7.89 for the AD12-ICT card test and Og4C3-ELISA, respectively.Both LR-values were zero (Table 4).The true positive fractions (TPFs) of the AD12-ICT card and thick blood fi lm tests, calculated with the Hayen formula, were 100% (27+0/27) and 85.2% (23+4/27), respectively.This difference of 14.8% between the TPF values was statistically signifi cant, suggesting that for each 1,000 schoolchildren with the infection, the AD12-ICT card test was able to detect 148 more cases than the thick blood fi lm test.The TPFs for the AD12-ICT card test and Og4C3-ELISA were both 100%.Accordingly, no difference could be found in the ability of the tests to detect cases of fi larial infection.

DISCUSSION
This study assessed the accuracy of tests in detecting bancroftian filariasis in a large and representative sample of schoolchildren from an endemic area in northeastern Brazil before the initiation of MDA.The performance evaluation of these tests in children is critical to evaluate the impact of the clinical and epidemiological characteristics of this population that may infl uence test performance 2,3 .This evaluation is of particular importance, as this demographic group has been recommended as a sentinel population by the WHO in whom to assess MDA effectiveness in areas undergoing this intervention.
The prevalence of fi lariasis estimated by the filarial antigen tests (AD12-ICT card test and Og4C3-ELISA) was far higher than that estimated by parasitological methods.The difference between the results of these methods was as much as fi vefold in the group aged below 10 years.The comparative evaluation of the performance of the thick fi lm test in relation to the AD12-ICT card test using the Hayen et al. 28 formula also showed that in each group of 1,000 schoolchildren with filarial infection, the AD12-ICT card test was able to detect approximately 150 more cases than the thick fi lm test.These results are in agreement with previous studies conducted in the school-age population [29][30][31][32][33][34][35][36][37][38][39] and reinforce the usefulness of the AD12-ICT card test in screening for fi larial infection in children for the assessment of MDA effi cacy 6,40 .The greater sensitivity of the antigen tests can be explained by their ability to detect low-level microfi laremia and amicrofi laremic cases that are typically more common in children than in adults 7 .
However, it is worth noting that both the AD12-ICT and Og4C3-ELISA tests may yield positive results after treatment even when there is no viable infection 16 , which could overestimate prevalence estimates.Conversely, because this study population was tested before MDA, it is probable that any prevalence excess due to false-positive tests would be of low magnitude.Additionally, the data draw attention to the low sensitivity of the parasitological methods, particularly the thick blood fi lm test, in the screening of fi larial infection at younger ages.
Several studies have shown that the sensitivity of both parasitological and antigen tests decreases as the MF loads decrease 21,31,33,41 .This decrease may result in differential performance across population subgroups.In this study, the MF loads were slightly higher in subjects younger than 10 years old and in girls, when measured by the parasitological methods, although these differences were not statistically   signifi cant.Similarly, we found no signifi cant differences in antigenemia according to age or gender.Thus, any variation in test accuracy across these subgroups of this particular population may be limited.
The prevalence estimated by Knott's test and the fi ltration test was, in general, slightly higher than that estimated by the thick film test.This finding implies that the former tools are most appropriate for fi lariasis screening of school-age populations before antifi larial treatment.Knott's technique is more laborious than the fi ltration technique but is usually less expensive 42,43 , and it showed a similar performance in detecting microfi laremia (15.3 vs. 12.0 MF/mL).Thus, it may be preferred in areas with limited fi nancial resources.A further advantage of this test is its ability to differentiate species of fi laria, which makes it useful in other endemic areas where co-infection with other fi larial parasites occurs.
The analysis according to age showed that the prevalence was signifi cantly higher in the 10-15-year-old age group than in the 4-9-year-old age group, as evaluated by all parasitological methods.However, no difference in the prevalence of antigenemia between age groups according to the ICT and Og4C3 tests was observed.Moreover, the prevalence of infection estimated by the antigen tests (ICT and Og4C3) was approximately seven times higher than that estimated by the parasitological methods in the younger age group (4-9 years).In the older group (10-15 years), although the prevalence estimated by the antigen tests was again higher than that estimated by the parasitological methods, this difference was less pronounced.These results demonstrate the lower sensitivity of the parasitological methods in the younger age groups compared with the older population, as a larger proportion of individuals in the prepatent stages or who have amicrofilaremic infections (with adult worms but with no or few reproducing females) are usually found in this population, as previously described 7,33 .
The performance evaluation of the AD12-ICT card test and Og4C3-ELISA against the gold standard of the combination of the parasitological test results demonstrated 100% sensitivity for both tests.Additionally, the values of LR+ and LR-were compatible with the high accuracy of these tests in the diagnosis of infection in this study population, even when considering that these measures may be distorted due to the imperfection of the gold standard due to the combination of parasitological results.The LR+ for both tests may represent an underestimate because subjects with an amicrofi laremic infection could have been misclassifi ed as free of disease according to the gold standard.Similarly, the low PPVs of the AD12-ICT card and Og4C3-ELISA tests can also be explained by the gold standard, which is assumed to have a lower sensitivity 27 than the antigen-based tests (AD12 and OG4C3 tests).
In regions with ongoing elimination programs, it is important to use high-specifi city tests to ascertain the actual status of transmission in the population with reasonable  certainty.Although the parasitological methods are 100% specifi c, they lose their usefulness after the initiation of MDA due to the potent microfi laricide effect of the antifi larial drugs, which yield a dramatic decrease in microfi laremia 16,44 .However, studies have shown that patients can remain infected with detectable levels of antigenemia even after several courses of treatment 16,24,45,46 .These issues limit the use of the antigen-based tests, such as the AD-12 and Og4C3 tests, in the assessment of treatment effectiveness, both at the individual and population levels.
The comparative evaluation demonstrated the high specifi city of the AD12-ICT and Og4C3-ELISA tests in the study population, although a slightly higher specifi city of the Og4C3 test was observed when the different levels of prevalence were taken into account.Similar results were also found by Gass et al. 40 , who demonstrated that the Og4C3-ELISA test showed a greater specifi city.These results suggest that the Og4C3-ELISA test is more appropriate for assessing the interruption of fi larial transmission in areas with ongoing elimination programs because it yields more accurate negative results, in addition to providing quantitative results that enable the monitoring of MDA effectiveness 18 .Furthermore, the Og4C3-ELISA test can be performed using fi lter paper, which facilitates the attainment of samples in fi eld studies as well as in venous blood samples.However, there is a need for better standardization of the technique, as well as better quality control of the kits by the manufacturer, particularly with regard to the plates and reagents.
Based on these results, we conclude that for an initial assessment of possible transmission of W. bancrofti, the ICT card test is the ideal tool for surveys and for mapping infection in the school-age population.The high accuracy of the AD12-ICT card test, in addition to other features such as rapidity and the simplicity of execution, makes this test the method of choice for the screening of fi larial infection in this population prior to the initiation of antifi larial treatment.
Due to their low sensitivity, the parasitological methods, particularly the thick fi lm technique, are very limited in their usefulness for the estimation of the burden of infection in this population even before treatment.The ICT card test is considered the most suitable test due to its high sensitivity and, as previously mentioned, its lower specifi city in populations that have already been treated 16,[46][47][48] .Finally, in areas with ongoing or completed MDA, the interruption of transmission, as assessed by the tracking of incident cases of fi lariasis by the ICT card test in the school-age population without prior antifi larial treatment, is the preferred approach.

Figure 1
Figure 1 shows the curves comparing the specifi cities of the AD12-ICT and Og4C3-ELISA tests over the feasible range of values of infection prevalence (5% to 20%) based on the method of Staquet et al.Although the exact values are unknown, we can conclude from these ranges that a) both tests have relatively high specifi cities (84.4% [95% CI: 80.8-87.4]for the AD12-ICT and 86.5% [CI 95%: 83.0-89.3]for the Og4C3-ELISA), and b) the Og4C3-ELISA has a slightly higher specifi city than the AD12-ICT test.

FIGURE 1 -
FIGURE 1 -Estimated specifi city of the AD12-ICT card test and Og4C3-ELISA according to the prevalence of microfi laria determined by the combination of all parasitological tests.AD12-ICT: immunochromatographic test with the monoclonal antibody AD12; Og4C3-ELISA: enzyme-linked immunosorbent assay using the monoclonal antibody 4C3, targeting Onchocerca gibsoni.

Rev Soc Bras Med Trop 47(3):359-366, May-Jun, 2014 RESULTSTABLE 1 -Contingency table of two tests in a paired design among those positive for the reference standard (fi ltration). Test N is a 'new' test, e.g., the AD12-ICT card test, and test E is an existing test, e.g., the thick blood fi lm test. Adapted from Table 2 of Hayen et al. 28 .
AD12-ICT: immunochromatographic test using the monoclonal antibody AD12.Note: n + = r + s + t + u, where n + : number of people with the disease; r: true positive; s: false negative; t: false positive, and u: true negative.

TABLE 2 -Means of microfi laria density and antigenemia measured by the thick fi lm, Knott concentration, membrane fi ltration, and Og4C3-ELISA tests according to age group and gender.
Og4C3-ELISA: enzyme-linked immunosorbent assay with the 4C3 monoclonal antibody targeting Onchocerca gibsoni; 95%CI: 95% Confi dence Interval; LR χ2: Likelihood ratio chi square; n: number of children examined; pos: number of positive results with the technique used; *Determined by a negative binomial regression model.