This single-centre diagnostic accuracy study was carried out using well-characterized frozen sera available from the DITM Tropica biobank. The inclusion criteria were as follows: (i) the sera had been donated for research purposes by migrants from sub-Saharan Africa following routine analyses; (ii) matching results for APC and/or polymerase chain reaction (PCR) were available for S. stercoralis; and (iii) results from both a commercial enzyme-linked immunosorbent assay (ELISA) and an in-house immunofluorescence test (IFAT) for strongyloidiasis routinely performed at DITM (see below) were available. When available, the result of the microscopic examination of formol-ether concentrated faeces was also reported in the study database. Exclusion criteria were as follows: (i) treatment with ivermectin in the previous 6 months; and (ii) the impossibility of executing the ICT, for any reason.
All available serum samples in the biobank were considered eligible for inclusion in the study. Consecutive samples that met the inclusion criteria were included in the study until the required sample size was achieved. The results of this study are reported following the Standards for Reporting of Diagnostic Accuracy Studies checklist [10].
Index test
The ICT is a prototype lateral flow assay that detects IgG4 against S. stercoralis [9]. The test line is lined with the recombinant antigen NIE [11]. After thawing, 35 µl of serum is added to the bottom square well. The serum flows up the strip to the control area; three drops of a buffer are then added to the oval top well. After 15 min, the result is visible on the cassette pad. A positive control line should be observable to verify that the test is valid. The presence of the test line in addition to the control line defines a positive result. The procedure for the implementation of this test has been described in detail previously [9].
Tests used for the evaluation of the index test
Due to the absence of a diagnostic gold standard, the results of multiple tests for S. stercoralis were used to estimate the accuracy of the ICT, as explained below.
Faecal tests
Both APC and PCR for S. stercoralis are routinely used at the study site. For the APC, 30 g of faeces mixed with charcoal is cultured on 10 agar plates (3 g faeces per plate). The plates are incubated at 26 °C for 48 h, followed by macroscopic examination. The plates are then washed with 2.5% formalin, which is centrifuged, and the sediment examined microscopically. In a previous retrospective study, the sensitivity of the APC method performed at the DITM was 45.4% [95% confidence interval (CI) 30.4–61.1] [12].
The real-time PCR assay is based on the method of Verweij et al. [13]. Briefly, for DNA extraction, about 200 mg of faeces is suspended in 200 µl of phosphate-buffered saline containing 2% polyvinylpolypyrrolidone (Sigma-Aldrich) and frozen overnight at − 20 °C until the extraction is performed. After thawing and boiling, the samples are processed by an automated extraction system (Magna Pure LC.2; Roche). The real-time PCR assay is performed as described previously [13]. The amplification target is the small-subunit ribosomal RNA gene sequence of S. stercoralis. Appropriate positive and negative controls are included in all runs. The PhHV-1 control DNA is amplified with the appropriate primers/probe mix in the same PCR reaction as the control for PCR inhibitors and amplification quality. The thermocycling comprises 40 cycles, and cycle threshold values < 40 are considered positive. The reactions, detection and data analysis are performed with the CFX96 detection system (Bio-Rad Laboratories). In a previous retrospective study, the method demonstrated a sensitivity of 56.8% (95% CI 41.0–71.6) [12].
Serologic tests
The IFAT used in this work was an in-house method implemented at the DITM, based on antigens retrieved from S. stercoralis filariform larvae obtained from positive APC. The detailed procedure has been described previously [14]. A positive result is defined as a titre ≥ 1:20. A previous retrospective study demonstrated good accuracy of the test, with a sensitivity of 94.6% (95% CI 90.7–98.5) and a specificity of 87.4% (83.4–91.3) [15]. It is routinely used for screening and individual diagnosis at the study site.
The commercial ELISA was an ELISA assay based on Strongyloides ratti antigens, and was developed by Bordier Affinity Products. A previous retrospective study estimated its sensitivity and specificity as 90.8% (95% CI 85.8–95.7) and 94.0% (95% CI 91.2–96.9), respectively [15]. The test was performed as per the manufacturer’s instructions. However, as the cut-off varies between runs, we used a normalized optical density ratio to compare the results obtained in different sessions. A ratio ≥ 1 defines a positive result. The test is widely available and deployed for routine screening and diagnostic activities across Europe.
Study procedures
One investigator retrieved the sera from the biobank and re-coded them. The laboratory staff performing and reading the ICT were blinded to the results of the previous reference tests. After the test had been performed by one member of laboratory staff, the ICT was read independently by two operators. The two readings were then given to an investigator not involved in the execution of the test, who sought an independent reading from a third member of laboratory staff in cases where there was disagreement between the readings of the first two operators. A test result was considered final when there was agreement on it between at least two readers. The results were reported as positive or negative. If the control band was not present on a test, the result was considered invalid and the test was repeated. In the case of a second invalid test, the final result was reported as such.
Statistical analysis
Primary endpoint
The accuracy of the ICT was first assessed against the results of the faecal tests that matched each serum sample. The faecal tests were classified as positive or negative, as reported in the laboratory records corresponding to the sampling time point. In detail, the result of the faecal test panel was classified as positive in cases where at least one of the faecal tests available for that sample (formol-ether concentrated, APC, PCR) was positive. The panel result was considered negative in cases where all the available faecal tests for that sample were negative.
Secondary endpoint
Since there is no gold standard for the detection of S. stercoralis, the accuracy of the index test was also assessed by latent class analysis (LCA) [16, 17], which combined the results of the faecal and serological tests that were already available.
The ICT results were classified as positive or negative. Indeterminate results were not used for the calculation of accuracy [18]. Cohen’s κ [19] was used to assess the agreement between readers. Estimates were reported along with the 95% CI. For the primary and secondary endpoints, the test results were displayed in contingency tables from which the sensitivity and specificity were calculated. LCA is a statistical method that uses a probabilistic model to estimate the unobserved true disease status of patients (positive or negative) based on the results of observed tests that are considered imperfect classifiers of disease status. For this study, four imperfect tests were used, with the assumption of conditional independence. The data analysis was performed using SAS software, version 9.4 (SAS Institute, Cary, NC). The level of statistical significance level was set at 0.05.
Sample size
In Yunus et al. [8], the sensitivity and specificity of the ICT by INFORMM were 91.3% and 100.0%, respectively. A sample size of 160 produced a two-sided CI with a width equal to 0.10 when the sample sensitivity was 0.90. For specificity, the sample size was 54, assuming a specificity of 0.99 with a width equal to 0.10. As an undefined number of samples were classified as positive or negative only at the moment of data analysis (following application of the LCA, as described above), 60 additional samples were analysed to reasonably meet the expectation that the minimum target number of positive cases had been identified.