Efficacy of LAMP assay for Mycobacterial spp. detection to prevent treatment delays and onset of drug resistance: a systematic review and meta-analysis

Search strategy

Diverse scientific databases, for example, PubMed, Google Scholar, Science Direct, Scopus, BioRxiv, and MedRxiv, were searched to screen for studies performed using LAMP for TB diagnostics from the year 2000 till March 2022. The terms such as LAMP, Tuberculosis, Mycobacterium and mycobacteria were used in various combinations during our research without any limitations: “LAMP + Tuberculosis” or “LAMP + Mycobacterium” or “LAMP + mycobacteria” or “LAMP + TB” or “LAMP + Tuberculosis + Mycobacterium” or “LAMP + Tuberculosis + mycobacteria” for PubMed, Science Direct, and Google Scholar without using any language restriction. The retrieved results were screened for duplication and conformity with the prespecified eligibility criteria.

Study eligibility criteria

Data extraction

Potential articles after reviewing titles and abstracts followed by full text for inclusion were extracted by two authors (G.S.B. and Z.H.). Consultation from two independent authors (S.J. and S.H.) was made to eliminate the doubt about any discrepancy. The extracted information from included studies had authors, year of publication, location of study, sample size, types of specimens, target genes, detection method, and standard reference method. The data extracted for evaluation of diagnostic accuracy for LAMP were performed by using either respiratory or non-respiratory specimens with any of the reference methods such as smear microscopy, culture, and GeneXpert. The important parameters in this meta-analysis such as TP, TN, FP, and FN of all studies were either extracted or calculated to provide their sensitivity and specificity values. The included studies (n = 30) were then assessed for their methodological quality to reduce systematic biases and inferential errors from the collected data.

Statistical analysis

The quantitative analysis of the included studies (n = 30) from the data extracted such as the values of TP, FP, TN, FN and sample size was performed. Furthermore, the values of sensitivity and sensitivity were mined or calculated from the available data. Moreover, pooled sensitivity and specificity of LAMP associated with 95% confidence interval (CI) were estimated. To maintain the accuracy and precision, the formulas: Accuracy = [TP + TN/TP + TN + FP + FN]* 100 and Precision = [TP/TP+FP]* 100 (12,13) were used. Accuracy and precision are important characteristics of any measurement. Accuracy is the degree of closeness of measured value to a standard value. However, precision provides the information regarding the closeness of multiple measured values to each other. Accuracy and precision are independent of each other. Forest plot for sensitivity and specificity were plotted using R-software along with summary receiver operating characteristic (SROC) for the given study.

Quality assessment

Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool was used to assess the methodological quality of the eligible studies. The risk of bias in the included studies (n = 30) was assessed from four areas of bias, for example, patient selection, index test, reference standards, and flow timing (14,15). For each QUADAS-2 domain specific yes/no questions were tailored. Following these criteria, the eligible studies were then refereed for low, unclear, or high risks of bias. Furthermore, we also judged to generate low, unclear, or high-risk applicability.

Literature survey

We followed the PRISMA guidelines (11) to search the literature for the present study (Fig. 1). The major scientific databases viz. PubMed, Science Direct, Scopus, BioRxiv, and MedRxiv have been extensively searched applying the above inclusion criteria and around 1,600 articles were extracted. From the 1,600 articles, we included the ones that were published after the year 2000 since the inception of LAMP technology (6) and thus excluded 22 articles. Further, only articles written in English language were considered and thus excluded 44 articles. Reading the titles and abstracts of these studies allowed to exclude further 1,029 articles comprising the review articles, editorials, proceedings etc. Following this exclusion, we removed the duplicated articles and further excluded 390 articles. Additional 73 articles were irrelevant as they didn’t use LAMP technology for the diagnosis of any mycobacterial species and were excluded, leaving a panel composed of 42 eligible studies. Lastly, from the 42 included articles, further 12 articles were also eliminated because their TP, FP, TN, and TN values were either not specified in these articles or the sensitivity and specificity values could not be calculated. Altogether, we observed that only 30 articles were eligible for detailed meta-analysis (Fig. 1) considering all the exclusion criteria.

Study characteristics and meta-analysis

Table I shows the data extracted from the eligible studies mentioning the details of authors, year of publication, country of study, types of specimens, target genes, detection method, and reference methods. Figure 2 shows the country-wise distribution of 30 identified articles included in the present study. Most of the studies (43.3%; n = 13) were conducted in the high TB burden nations such as India followed by Thailand and Japan (each 10%; n = 3). Two studies each were also conducted in countries such as China, Korea, and Switzerland (6.3%; n = 2). Apart from this, one study each, that is, 3.3%, was from countries included such as Gambia, Nepal, South Africa, Sri Lanka, and Vietnam. Although most of the included articles do not mention about the patient details, the type of specimen (Fig. 3) used in most of the studies was sputum (42.8%; n = 21). In addition, some studies have been tested on other specimens such as cerebrospinal fluid (n = 4), fecal samples (n = 1), urine (n = 3), blood (n = 2), and pleural fluid (n = 3) for the detection of mycobacteria by using LAMP. Furthermore, the standard smear microscopy, culture assay, and PCR-based methods were used as references either alone or in combination (n = 30). Of note, radiology was also used (n = 4) to validate LAMP results as a reference standard (Tab. I), with one study using immunochromatography (16). Next, we examined the various target genes used for the eligible studies. Ten different types of target genes including hspX, IS900, mpt64, Pab, sdaA, rimM, 16SrRNA, MPB64, gyrB, and IS6110 were used in the included studies (n = 30). IS6110 gene was most frequently used in the included studies (n = 14; 31.18%) followed by gyrB (n = 9, 20.45%), 16SrRNA (n = 6, 13.63%), and MPB64 (n = 6, 13.63%) genes (Fig. 4). Furthermore, while analyzing detection methods used for these 30 studies, fluorescent method (n = 19, 32.39%) was the most frequently performed followed by colorimetry (n = 14, 25.35%), gel electrophoresis (n = 11, 20.00%) and turbidity (n = 9, 16.36%) methods (Fig. 5). In 53.33% (n = 16) of studies, more than one detection method was used. In 16.66% (n = 5) of studies, combination of three methods was used while in only two studies (6.66%), combination of four different methods was reported (17,18).

Among all the eligible studies, 4 studies showed 100% sensitivity, while for 16 studies this parameter was higher than 90%. Similarly, 6 studies exhibited 100% specificity while 90% or more specificity was observed in 24 studies (Tab. I). Furthermore, upon analysis of sensitivity and specificity using forest plot at 95% CI, we found that the sensitivity values varied between 0.26 and 1.00 and the specificity values ranged from 0.67 to 1.00 (Fig. 6). A total of 27 out of the 30 included studies showed pooled sensitivity greater than 70%. Only three studies reported sensitivity values of 26% and 45% each (19-21). In terms of FP rate (1-specificity), 27 included studies showed a pooled FP rate higher than 80% (Fig. 7). Additionally, the accuracy and precision rates of included studies were calculated and varied between 37.73% and 99.33%. The analysis proved that 22 studies displayed more than 90% accuracy with only 4 studies depicting less than 80% accuracy (Tab. II). Likewise, the precision rates varied between 39.47% and 100%. The analysis showed that 21 studies exhibited more than 90% precision rate with only 3 studies depicting less than 80%. Of note, we observed that six studies displayed 100% precision rate.

Quality assessment of the study

Almost two-thirds of the included studies (22 out of 30 studies) have a high risk of patient selection bias due to non-random patient selection and case-control study design (Fig. 8, Tab. I). Around 26% (8 out of 30) of the included studies have low risk of patient selection bias because these studies provided sufficient details about patient inclusion/exclusion criteria; 86% of included articles (26 out of 30 studies) present low risk of index test bias because these tests clearly stated the quantitative detection read-outs with reported thresholds. Moreover, these studies explicitly declared that their index and reference tests were done simultaneously in parallel to each other or that testing was blinded from each other. Two studies (19,22) were reported without defined detection thresholds. One study (23) had unclear risk of index test bias as the quantitative detection thresholds were not explained. It was either unclear whether index test results were interpreted with knowledge of reference test results or if only qualitative read-out was used for reading the results. Hence, index test bias of these studies was unclear. For the rest of the included studies, almost all (n = 30) have low risk of reference standard bias because they provided enough information about the standard reference test used in the study.

Half of the studies (15 out of 30) have an unclear risk of flow and timing bias as there is not enough information, whether reference standard results were interpreted with the knowledge of the results of the index test. One study (24) was at high risk as it did not provide any information on whether the samples for a reference test and the index test were taken at the same time. Our review question did not focus on any patient demographics. None of the included studies attempted to exclude patients based on demographics and thus had no “concern of patient selection applicability” (Fig. 8, Tab. I). Index tests of all studies have generally been used for Point-of-care test (POCTs) and thus have low concern of index test applicability. Reference standard tests of nearly all studies were culture, smear microscopy, Xpert test, PCR, or combinations of them. Thus, we graded these studies as having low concern of standard test applicability.