Accuracy of Enzyme-Linked Immunosorbent Assays (ELISAs) in Detecting Antibodies against Mycobacterium leprae in Leprosy Patients: A Systematic Review and Meta-Analysis

IgM against Mycobacterium leprae may be detected by enzyme-linked immunosorbent assays (ELISAs) based on phenolic glycolipid I (PGL-I) or natural disaccharide octyl bovine serum albumin (ND-O-BSA) as antigens, and the IgG response can be detected by an ELISA based on lipid droplet protein 1 (LID-1). The titers of antibodies against these antigens vary with operational classification. The aim of this study was to compare the accuracy of ELISAs involving PGL-I and ND-O-BSA with that involving LID-1. We included studies that analyze multibacillary and paucibacillary leprosy cases and evaluate the diagnostic accuracy of ELISAs based on LID-1 and/or PGL-I or ND-O-BSA as antigens to measure antibody titers against M. leprae. Studies were found via PubMed, the Virtual Health Library Regional Portal, Literatura Latino-Americana e do Caribe em Ciências da Saúde, Índice Bibliográfico Espanhol de Ciências de Saúde, the Brazilian Society of Dermatology, National Institute for Health and Clinical Excellence, Cochrane Library, Embase (the Elsevier database), and Cumulative Index to Nursing and Allied Health Literature. The Quality Assessment of Diagnostic Accuracy Studies served as a methodological validity tool. Quantitative data were extracted using the Standards for Reporting of Diagnostic Accuracy. Sensitivity, specificity, and a diagnostic odds ratio were calculated, and a hierarchical summary receiver-operating characteristic curve and forest plots were constructed. The protocol register code for this meta-analysis is PROSPERO 2017: CRD42017055983. Nineteen studies were included. ND-O-BSA showed better overall performance in terms of sensitivity, specificity, positive and negative likelihood ratios, and diagnostic odds ratio when compared with PGL-I and LID-1. The multibacillary group showed better performance on these parameters (than the paucibacillary group did), at 94%, 99%, 129, 0.05, and 2293, respectively. LID-1 did not provide any advantage regarding the overall estimate of sensitivity in comparison with PGL-I or ND-O-BSA.


Introduction
Leprosy is a chronic infectious disease caused by Mycobacterium leprae (a microorganism that mainly affects the skin and peripheral nerves) and is considered one of the six most dangerous diseases in developing countries by the World Health Organization (WHO) [1].Leprosy diagnosis is based on the presence of at least one of the following three cardinal signs: definite loss of sensation in a pale or reddish skin patch, a thickened or enlarged peripheral nerve with loss of sensation and/or weakness of the muscles innervated by that nerve, and the presence of acid-fast bacilli in a slitskin smear [2].
A continual and slow reduction in the number of new cases has been observed during the last decade, even though more than 200,000 new cases are diagnosed every year.India, Brazil, and Indonesia represent 80% of all cases [3].
A case of leprosy is defined as an individual that has a skin lesion consistent with leprosy, with definite sensory loss, with or without thickened nerves, and/or with positive skin smears [4].Cases of leprosy are classified operationally as either paucibacillary leprosy (PB) or multibacillary leprosy (MB) depending on the number of skin lesions [5].Early diagnosis and appropriate treatment of leprosy patients are essential conditions for stopping the transmission and reducing the physical and social consequences of the disease [4].
e discovery of phenolic glycolipid I (PGL-I) in 1980, a specific component of M. leprae, and its use in serological assays in patients with leprosy in 1981 are major advances in serological research on the disease [6][7][8].Due to the glycolipid nature of PGL-I, the humoral immune response of leprosy patients predominantly involves IgM [9].e detection of these IgM antibodies represents the best-evaluated and standardized serological test for leprosy [10][11][12][13][14][15].In addition to native PGL-I, IgM levels can be measured by means of a synthetic mimotope: a natural disaccharide linked to bovine serum albumin (ND-O-BSA) [15][16][17].e IgG response to M. leprae can be measured using LID-1 as an antigen: a chimeric protein generated by the fusion of antigens ML0405 and ML2331 [18].Because it was reported early on that individuals with a high bacillary load have a high IgM titer against PGL-I [19], even in the chronic stage of the disease, the accuracy of the tests based on PGL-I (native or synthetic) and LID-1 has been compared previously in several studies, with the aim of identifying an adequate test for serological diagnosis [15][16][17][18][20][21][22][23][24].
e titers of antibodies against PGL-I, ND-O-BSA, and LID-1 vary, with the clinical presentation being the strongest in MB patients and the weakest or absent in PB patients.e bacterial index may also correlate with antibody titers [20][21][22].
e Guidelines for the Diagnosis, Treatment and Prevention of Leprosy (WHO) warn that studies of the most commonly used ELISA and lateral flow tests show low sensitivity for PB leprosy, which is often harder to diagnose clinically than MB leprosy.Based on currently available evidence, newer ELISA and other laboratory tests do not represent a clear advantage over current standard diagnostic methods [25].
To date, a number of studies have used ELISAs based on PGL-I, ND-O-BSA, or LID-1 as antigens.
e successful implementation of these methods reflects the good performance of these tests.Nonetheless, sensitivity and specificity of these assays vary depending on the geographic origin of the population studied [21].erefore, our aims were to conduct a meta-analysis of studies on the accuracy of the available serological tests and to summarize the accuracy of these tests in detecting antibodies against M. leprae.e aims were achieved successfully.

Methods
e protocol for this meta-analysis was published in the international prospective register of systematic reviews (PROSPERO 2017: CRD42017055983) before its implementation and is described in Supplementary Materials (Text S1).
e protocol and final report were developed based on the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy [26].
2.1.e Review Question/Objective.What is the diagnostic accuracy of the commercially available ELISA based on antigen LID-1 as compared to ELISAs based on native antigen PGL-I or synthetic antigen ND-O-BSA for the detection of antibodies against M. leprae in patients with leprosy?
More specifically, we performed a meta-analysis of studies on the diagnostic test accuracy of PGL-I, ND-O-BSA, and LID-1 ELISAs to obtain summary points for the accuracy values of the assays for antibodies against M. leprae.

Inclusion Criteria.
e mnemonic PIRD (participants, index test, reference test, and diagnosis of interest) was employed for the inclusion criteria as recommended for systematic reviews of diagnostic test accuracy [26].Studies were included that dealt with MB and PB leprosy cases and evaluated the diagnostic accuracy of ELISAs based on LID-1 and/or PGL-I or ND-O-BSA antigens to measure antibody titers against M. leprae.
e gold standard for the diagnosis of leprosy is based on clinical diagnosis.erefore, only studies that selected and classified patients with leprosy on the basis of clinical diagnosis were included.

Types of Included Studies.
e studies had to have any epidemiological design that afforded a detailed measure of sensitivity, specificity, and receiver-operating characteristic (ROC) curves.

2.4.
e Search Strategy.is study was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol standard proposed by the Cochrane Collaboration ® [27].A three-step search strategy was utilized in this review.First, an initial limited search of Medline was performed by searching for MeSH index terms and related keywords.is search involved an analysis of words contained in the title and abstract and index terms used to describe the article.Second, another search involving all the identified keywords and index terms was performed across all the included databases.ird, a reference list of all dissertations with clearly detailed accurate values was considered.Studies published since 1982-the year when the first ELISA based on the PGL-I antigen was developed to detect antibodies against M. leprae-until February 2018 were considered for inclusion in this review.Moreover, only published studies were included because these studies were evaluated by external reviewers.e search strategy can be found in Supplementary Materials (Text S1).

2
Canadian Journal of Infectious Diseases and Medical Microbiology Database searching was carried out in PubMed, which includes Medline and other health databases; in the Virtual Health Library Regional Portal (VHL Regional Portal), which includes Medline, Literatura Latino-Americana e do Caribe em Ciências da Saúde (LILACS), Índice Bibliográfico Espanhol de Ciências de Saúde (IBECS), and other health databases; via the Brazilian Society of Dermatology; at the National Institute for Health and Clinical Excellence (NICE); in the Cochrane Library; in Embase, the Elsevier database; and in the Cumulative Index to Nursing and Allied Health Literature (CINAHL).e databases used to search dissertations as a source of gray literature were Google Scholar and EVIPNet (WHO).
e keywords were LID-1, PGL-I, ND-O, NDO, IDR1, Specificity, Sensitivity, and Measurement Accuracy.e terms were combined via the boolean operators "AND" and/or "OR" to compose the search strings.

Assessment of Methodological Quality.
e documents selected for retrieval were assessed by two independent reviewers for methodological validity prior to inclusion in this study, by means of standardized critical appraisal instruments from the Quality Assessment of Diagnostic Accuracy Studies (QUADAS 2), which was released in 2011 after revision of the original QUADAS.e QUADAS tool consists of four key domains that evaluate patient selection, an index test, reference standard, and flow and timing (flow of patients through the study and timing of the index tests and reference standard).Each domain is assessed in terms of the risk of bias, and the first three domains are also evaluated in terms of concerns about applicability [28,29].Any disagreements between the reviewers were resolved either through discussion or based on the opinion of a third reviewer.
2.6.Data Extraction.Quantitative data were extracted from papers according to the Standards for Reporting of Diagnostic Accuracy (STARD) [30,31].A 2 × 2 table was compiled to classify the data as true positive, false positive, true negative, and false negative.Sensitivity, specificity, positive and negative likelihood ratios (LR+ and LR−), and the diagnostic odds ratio (DOR), with a confidence interval (CI) of 95%, were calculated for each study, and subsequently, the results were combined.
Two forest plots were generated side by side: one for sensitivity and the other for specificity showing the means and 95% CIs of each selected primary study.rough summary receiver-operating characteristic (SROC) curves, the presence or absence of heterogeneity was identified.e meta-analysis was performed based on the hierarchical model of summary receiver-operating characteristic (HSROC) curves [32].
e HSROC curve provides information on the overall performance (sensitivity, specificity, LR+, LR−, and DOR) of a test via different thresholds.
To evaluate the potential of publication bias, Deeks' funnel plot was constructed, with p < 0.05 indicating the presence of publication bias [33].Fagan's nomogram, conceived to provide posttest probability, was employed to estimate clinical utility of the test values and is based on LR+ and LR− obtained from the meta-analysis [34].

Results
Our search yielded 968 citations related to leprosy through the combined application of descriptors in the databases described above.After the eligibility criteria (duplicate texts, articles related to other topics, and text excluded for review criteria or quality methods), 19 baseline studies remained.
e results of our search strategy are shown in a PRISMA flowchart (Figure 1).Excluded studies are summarized in Table S1.
e evaluation of methodological quality revealed that the studies included in this meta-analysis had a "low risk of bias" in patient selection and flow and time domains.Some of the selected studies were "at risk of bias" in the index test (10.5%) and the reference standard domains (15.7%).On the contrary, patient selection and the reference standard showed a "low applicability concern."Only 5.2% of the selected studies yielded "applicability concerns" in the index test (Figure 2).e methodological quality summary bias risk concern and applicability of each domain for each included study are presented in Figure S1.e data extracted from the final selection are given in Table S2.

Effects of Clinical Manifestations of Leprosy on the Accuracy of Tests.
To verify whether leprosy patient groups varied significantly in the performance of the M. leprae antigen ELISAs (PGL-I, ND-O-BSA, and LID-1), we carried out a global estimate of the accuracy of each test by group (MB and PB).
For all ELISA antigens, the combined specificity in the MB group was the same as that in the PB group.

Publication Bias and Heterogeneity. Deeks' funnel plot
was constructed to analyze the potential publication bias for each antigen-specific ELISA in both patient groups.PGL-I Deeks' funnel plots did not reveal any publication bias in the two groups (p � 0.63 and 0.69 for groups MB and PB, respectively).For the ND-O-BSA antigen, only Deeks' funnel plot in the MB group did not show publication bias (p � 580).On the contrary, studies on the LID-1 antigen showed a publication bias risk in both the MB and PB groups (Figure S2).
e SROC curves for each ELISA antigen revealed a range of 87-100% for the area under the curve (AUC), with a 95% confidence contour and 95% prediction contour for each population studied (Figure S3). e SROC curves did not show heterogeneity among the included studies.

Accuracy of ELISAs in Detecting M. leprae.
By means of the HSROC curves, the accuracy of each type of M. leprae ELISA based on different antigens was evaluated, and a summary point was generated for each population under study (Table 2).When we evaluated the accuracy of PGL-I  S4C).
Specificity values were between 97% and 99% in each type of ELISA and in each group analyzed.

Discussion
Most of the ELISAs performed in the included studies in this systematic review and meta-analysis were performed in Brazil (59%) and Asia (28.2%, predominantly in China): countries that have different epidemiological profiles [4].ese studies revealed variations in sensitivity and specificity depending on the ELISA antigen and the patient group (MB or PB).ese variations may be related to the strains found in each region and immune responses of the patients.
In the present meta-analysis, studies that analyzed ELISA tests involving the ND-O-BSA antigen indicated sensitivity (77-100% for group MB and 15-93% for group PB) and specificity (97-100%) ranges that are more favorable than did studies on PGL-I and LID-1 ELISAs.Sensitivity values among studies from different regions and among studies from the same regions showed great differences, for both the MB and PB groups, as reported previously [21].Even studies that were designed by the same authors and conducted in the same regions produced different sensitivity values [16,17].Specificity values were more similar among the studies analyzed in both groups, MB and PB.
A general diagnostic test accuracy estimate was carried out for each ELISA antigen in both leprosy groups.e HSROC curves showed better sensitivity (94% (95% CI 78-98) for MB and 56% (95% CI 17-81) for PB) and specificity (99% (95% CI 97-100)) for the ND-O-BSA antigen.In our results, the ELISAs using PGL-I were not subject to conclusive publication bias in either of the groups studied (MB or PB).As for the ELISA involving the ND-O-BSA antigen, only the MB group showed publication bias.Sensitivity and specificity found for each ELISA matched the accuracy reported by other authors [15,43,44,49].On the contrary, when we compared the performance reported in these studies with our results, there was no consensus regarding a superior antigen for leprosy ELISAs. is finding may be due to the fact that most of the studies were conducted with conventional ELISAs made in-house and due to other factors like sampling time, sample transport, and sample preservation, which may cause test performance variations.Additionally, there is no standardized cutoff value for any of these ELISA antigens or for either group of patients with leprosy.Nevertheless, we can conclude that all the analyzed antigens have better diagnostic accuracy for MB leprosy, as reported elsewhere in the literature [15,21,23,24,43,45,49,50]. Very divergent accuracy results in the group of PB patients were found.Based on the estimated median sensitivities found in this patient group, negative tests are not that useful for ruling out PB leprosy patients.Using these serologic tests, PB patients can be diagnosed as negative when they really are not.e absence of added value for the use of LID-1 was also observed in a recent study, where the researchers detected antibodies against the PGL-I antigen in patients with leprosy by rapid tests [51].Owing to the presence of anti-BSA antibodies, which may interfere with the test results [52], the ND-O antigen conjugated to human serum albumin (HSA) has been used in ELISAs [15,43,49].Nevertheless, these studies suggest that the BSA or HSA carrier protein of the antigen does not significantly influence the anti-PGL-I seropositivity of the groups under study [15,43].e primary literature on ELISAs using the ND-O-HSA antigen was not included because only three studies were found, and at least four are required for a reliable meta-analysis.
e DORs obtained here were higher on average in the MB group than in the PB group.e DOR varies from zero to infinity, with higher values denoting a better discriminatory diagnostic test.Additionally, posttest probability (Fagan's nomogram) was high, specifically within the MB patient group; for each antigen included in our analysis, this result is indicative of good clinical utility of an ELISA as a supplement to clinical diagnosis.
ere are some limitations of our study.First of all, there is no standard cutoff value for any ELISA antigen analyzed here and for either group of patients under study (MB or PB).Although this situation did not hamper implementation of our meta-analysis, accuracy variations can occur due to differences in the number of true or false positives or negatives among the studies.Management of outliers varies among authors and affects the measurement of accuracy parameters, resulting in a wide range of sensitivity and specificity estimates across studies, as shown in our study.Second, although the specificity was almost 100%, only a few authors included groups of controls of patients with other diseases such as tuberculosis, and most authors used endemic control samples and a few samples of nonendemic controls.Nevertheless, there was a risk of publication bias in the studies on the ND-O-BSA antigen in the PB group.Few studies indicated whether the samples were from primary or secondary infections, or whether the patients received treatment.Finally, due to different amounts of patient information in the included studies, the data were not divided into additional groups based on other variables, e.g., gender or age.

Conclusion
In this meta-analysis, in the MB group, the LID-1 ELISA did not show any advantage with respect to the overall sensitivity estimate (79% (95% CI 66-89)) when compared to native antigen PGL-I (78% (95% CI 60-90)) or to synthetic antigen ND-O-BSA (94% (95% CI 81-97)).Specificity of all the ELISAs in this group was close to 100% for all antigens, whereas in the PB group, all the assays showed lower sensitivity values as compared with the MP group in terms of detection of antibodies against M. leprae.
Our results confirm that traditional ELISAs have good accuracy in detecting MB leprosy and poor accuracy in detecting PB leprosy.
e WHO research priorities for leprosy include new tools for early detection.To achieve this goal, it is important to have a standardized serological, molecular, or immunological assay that is applicable to different geographic regions with different epidemiological profiles and pathogen strains.In the future, these laboratory tools are expected to become important for the diagnosis of leprosy (MB and PB), for surveillance of household contacts, and for establishing health policy interventions.

Figure 3 :
Figure 3: A forest plot of sensitivity of ELISAs by antigen: (a) PGL-I, (b) ND-O-BSA, and (c) LID-1, according to each studied group (MB (A.1, B.1, and C.1) and PB (A.2, B.2, and C.2)). e same specificity was found in both groups for the three ELISAs (A.3, B.3, and C.3). e circle in a square represents sensitivity and specificity, and the horizontal line represents the point estimate (95% CI for each study).Diamonds represent the combined value estimate (95% CI).
Figure 1: A flowchart of the steps performed in the systematic review and meta-analysis.

Table 1 :
A summary of the included studies.

Table 2 :
Accuracy of ELISAs for detection of leprosy using different M. leprae antigens.Summary points of the HSROC curve accuracy for each M. leprae antigen used in the ELISAs for each population studied.Op. class � operational classification; 95% CI � 95% confidence interval; LR− � negative likelihood; LR+ � positive likelihood ratio; DOR � diagnostic odds ratio. * Figure 4: Fagan's nomogram and the posttest probabilities.Fagan's nomogram was built with a prevalence of 30% for household contacts of leprosy patients in an endemic area.If a patient tests positive, the posttest probability that they truly have leprosy would be (a) PGL-I ELISA: 98% for group MB and 88% for group PB; (b) ND-O-BSA ELISA: 98% for group MB and 97% for group PB; and (c) LID-1 ELISA: 90% for group MB and 78% for group PB (solid red line).On the contrary, if this patient tests negative, the posttest probability of having the disease and not being detected would be (a) PGL-I ELISA: 9% for group MB and 23% for group PB; (b) ND-O-BSA ELISA: 3% for group MB and 16% for group PB; and (c) LID-1 ELISA: 8% for group MB and 26% for group PB (dotted blue line).