Background

A century after the discovery of Mycobacterium tuberculosis (Mtb) as the etiological agent of tuberculosis (TB) by Robert Koch, the disease is still one of the leading causes of death (after AIDS) worldwide [1, 2]. According to the World Health Organization (WHO) report in 2020, approximately 10 million (range, 8.9–11.0 million) people became infected with Mtb in 2019; of these, approximately 1.2 million (range, 1.1–1.3 million) deaths occurred among HIV-negative people [3]. Furthermore 208,000 deaths (range, 177,000–242,000) were related to HIV-positive individuals [4]. According to the use of comprehensive treatment programs, statistics show that the number of TB deaths among HIV-positive and HIV-negative people fell by 31% and 69% between 2000 and 2019, respectively [4]. Drug-resistant tuberculosis (DR-TB) is a global concern for TB control programs, and close to half a million people have been diagnosed with rifampicin-resistant TB (RR-TB); of which 78% had MDR-TB (resistant to both rifampicin and isoniazid). Three countries, India (27%), China (14%), and Russian Federation (8%) had the largest share of the global MDR-TB burden, respectively [4, 5]. Therefore, prompt diagnosis and appropriate treatment of both TB and DR-TB cases are considered as key factors in reducing mortality from this disease [6, 7]. According to WHO guidelines, diagnosis of both resistance and susceptible isolates requires methods such as molecular techniques, cultivation, sequencing, and bacteriological confirmation; although methods e.g. culture (in LJ, 7H10, 7H11 and MGIT media) and phenotypic drug susceptibility testing (DST) are considered as gold standard methods for diagnosis of this bacterium, however, these methods are time-consuming, laborious, and sometimes ambiguous results are obtained [4, 8]. Therefore, reducing the detection time leads to reduce the risk of transmission of resistant strains [9]. In recent years, several rapid methods for diagnosing TB or DR-TB isolates such as PCR, real-time PCR, line probe assay (LPA), Xpert MTB/RIF assay, and BACTEC MGIT 960 liquid culture have been introduced [8]. Xpert MTB/RIF assay is applicable in high-prevalence and low-income settings; although this method has recently been approved by WHO for rapid detection of both Mtb and RR-TB isolates, it has several disadvantages, for example, it is only used for rifampin and is not affordable in all regions of developing countries [8,9,10,11]. BACTEC system is based on the generation of radioactive CO2 from substrate palmitic acid; this system is high specific and helpful in distinguishing Mtb from other mycobacteria, as well as is used as a comparative method versus DST. In general, in this system growth can be detected in 5–10 days; however, the detection time of bacteria in this method is faster than the conventional culture method (more than two weeks), but it takes more time to detect than molecular methods [12]. Overall, molecular methods are more rapid compared to other methods, nevertheless, these methods are costly, and also require standardization and a multi-step process; moreover, the use of these methods increases the risk of cross-contamination and misdiagnosis [13]. High resolution melting (HRM) assay is a new method to detect point mutations, single nucleotide polymorphism (SNP), and internal tandem duplications [14, 15]. HRM technique is based on differences in the melting profiles of test and reference DNAs; in this method, using real-time PCR assay, first, immunofluorescent dyes bind to double-stranded DNA, and during the denaturation step of PCR amplicons, differences in melting curves indicate that a mutation occurs in the test DNAs compared to reference DNA [16]. The recent method is relatively inexpensive and requires only unlabeled primers and dsDNA binding dyes; the process of detection is too fast, and due to the reaction in a closed tube, the risk of contamination is low; unlike other molecular methods, this method has no post-PCR process [17, 18]. To date, various studies have examined the sensitivity and specificity of HRM, and the aim of this meta-analysis was to determine the overall accuracy of HRM in identifying DR-TB cases.

Methods

Literature search strategy

The systematic review was performed based on preferred reporting items for systematic reviews and meta-analysis (PRISMA) guideline [19]. At the first, a comprehensive literature search‏ ‏was‏ ‏conducted‏ ‏using global databases such as‏ ‏Web of Science, PubMed, Scopus, Cochrane Library, and Google scholar. Relevant studies were retrieved according to keywords such as “Mycobacterium tuberculosis”, “Tuberculosis”, “TB”, “Drug-resistant TB”, “High resolution melting”, “Specificity”, and “Sensitivity”. Studies were selected regardless of publication date, and we also evaluated the bibliography of all candidate articles to identify duplicate studies.

Study screening and selection

To determine the eligibility of studies, title, abstract, and full text of potential studies were evaluated. This process was conducted by two authors separately, and disagreements were resolved through discussion. Our inclusion criteria included: (1) original articles on HRM value for diagnosis of DR-TB; (2) evaluated studies by reference tests (antibiogram, BACTEC MGIT 960 system, sequencing) and HRM; (3) studies containing information such as true positive (TP), false positive (FP), true negative (TN), and false negative (FN); (4) English studies. In the present study, excluded criteria were including: (I) congress abstracts, letters, and review articles; (II) articles with unclear results and insufficient data; (III) studies on extrapulmonary TB; (IV) non-English studies. Finally, 24 articles met our included criteria [20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43]. ‏

Data extraction and quality assessment

In this step, we used the Newcastle–Ottawa Scale (NOS) to assess the quality of included studies. Information such as first author, publication year, country, subjects, number of Mtb (resistant and susceptible) isolates, studied genes, and number of TP/FP/TN/FN for each participant are listed in Table 1.

Table 1 Characteristic of included studies
Full size table

Statistical analysis

The accuracy and reliability of HRM method were measured using indexes such as sensitivity, specificity, and diagnostic odds ratio (DOR) with 95% confidence interval (CI). Subsequently, we employed summary receiver operating characteristic (SROC) curve to measure the area under the curve (AUC) [44]. In addition, Chi-squared and I-squared tests (p < 0.01 or I2 > 50%) were used to measure heterogeneity between studies. Based on significant heterogeneity between studies, we conducted a pooled analysis (fixed-effects model or random-effects model). Furthermore‏,‏ subgroup analysis was performed‏ ‏individually‏ ‏to assess the accuracy of HRM assay for two types of DR-TB including, mono-resistant TB (rifampin, isoniazid, ethambutol, streptomycin, pyrazinamide, or fluoroquinolone) and MDR-TB‏.‏ All statistical‏ ‏tests‏ ‏were two-sided with a significant cut-off (p < 0.05), and were fulfilled by Meta-DiSc software.

Results

Characterization of included studies

After an initial evaluation of the retrieved articles, finally 24 studies were accordance with our criteria (Fig. 1). Included studies had been published in Austria, Korea, Singapore, United States, China, India, Iran, Japan, Egypt, Africa, Peru, Russia, and Thailand during 2009–2020. In this study, data from 4732 Mtb isolates (both drug-resistant and drug-susceptible strains) were studied. In all eligible studies, authors had assessed the accuracy and integrity of HRM assay for both first-line (isoniazid, rifampin, streptomycin, ethambutol, pyrazinamide, and MDR-TB) and second-line (particularly ofloxacine) treatments, using standard methods such as sequencing, MGIT 960, and proportional method (Table 1). In these studies, different primers were designed related to antibiotic resistance including katG, mab-inhA, rpoB, rrs, rpsL, eis, embB, pncA, gyrA‏, and gyrB. Quality assessments showed that most included studies had a low risk of bias, which confirmed the proportion of selected studies.

Fig. 1
figure 1

Flowchart of the selection of excluded and included studies

Full size image

Meta-analysis results

The results obtained from forest plots showed that the sensitivity and specificity of HRM were acceptable for distinguishing DR-TB strains from susceptible strains. According to statistical analysis, the sensitivity and specificity of DR-TB cases were 89% (95% CI: 88–90) and 95% (95% CI: 94–96), respectively. However, the I2 values for sensitivity and specificity were higher than 50%, which indicated a significant heterogeneity. The diagnostic accuracy of HRM was measured by SROC curve, so that the AUC and Q* values were 0.9754 and 0.9289, respectively.

Subgroup analysis

Due to a significant heterogeneity in the estimated results, we performed subgroup analysis to discover the source of heterogeneity. Each of 47 studies independently had measured Mtb resistance to isoniazid, rifampin, ethambutol, streptomycin, pyrazinamide, fluoroquinolones MDR-TB, and first line. Parameters such as sensitivity, specificity, diagnostic ORs, and SROC for these antibiotics are listed in Table 2.

Table 2 Pooled means of sensitivity and specificity, DOR, and SROC for each antibiotics
Full size table

According to the results of subgrouping analysis, HRM is a reliable method for diagnosis of DR-TB cases, in particular the sensitivity and specificity for isoniazid, rifampin, and MDR-TB was higher than 90%. Furthermore, this method accurately discriminates resistance and susceptibility to other antibiotics such as ethambutol, streptomycin, pyrazinamide, and fluoroquinolones. The SROC curves represented the maximum polymerization spots of sensitivity and specificity associated with each of HRM assay (Fig. 2).

Fig. 2
figure 2figure 2

Summary Receiver Operating Characteristic curves of each antibiotic for discriminating drug-resistant TB cases (2.1: ethambutol, 2.2: first-line TB, 2.3: fluoroquinolone, 2.4: isoniazid, 2.5: MDR-TB, 2.6: pyrazinamide, 2.7: rifampin, and 2.8: streptomycin)

Full size image

Discussion

In recent years, the emergence of DR-TB strains is accounted a serious threat to TB control worldwide [45]. Diagnosis and screening of patients infected with MDR and extensively drug-resistant (XDR) strains is an important strategy to monitor and prevent the geographical spread of DR-TB strains [46]. Although conventional drug susceptibility testing such as proportional method, absolute concentration method, and resistance ratio method are known as reference methods, these are time-concuming and laborious [47]. To date, conventional drug susceptibility testing by liquid culture system has been introduced as colorimetric redox-indicator method, which in turn reduces the detection time to less than a week, but this system requires expensive tools and cannot be used in all developing countries [48, 49]. The Xpert MTB/RIF assay is a relatively new method recommended by WHO, but it should be noted that this method is only able to detect rifampin-resistant cases [50]. Among the molecular methods that can be used to determine TB antibiotic resistance, HRM is a fast, inexpensive, and simple method, which can detect different mutations using a small number of probes or even without a special probe [51]. In their meta-analysis study (using 7 articles), Yin et al. showed that the sensitivity and specificity of HRM in the diagnosing of rifampin-resistant TB cases were 94% and 99%, respectively [18]. A summary of common methods for detecting drug-resistant TB are listed in Table 3.

Table 3 Comparison of popular methods for the detection of drug-resistant tuberculosis
Full size table

To date, many studies have examined the accuracy of HRM for the diagnosis of DR-TB, however, no study has systematically evaluated the effectiveness of this method. In the present meta-analysis, we estimated the sensitivity and specificity of HRM in detecting DR-TB cases at 89% and 95%, respectively. According to our results, accuracy of HRM was slightly higher than other molecular methods, especially PCR-SSCP, while it was lower compared to the conventional drug susceptibility testing method. HRM method is credible method for detecting of mutations in specific genes; approximately 95% of rifampin-related mutations occur in the rifampicin resistance determining region (RRDR), while, 5% of those are outside this locus and cannot be detected by conventional molecular methods [65]. Nevertheless, some mutations such as nucleotide transversions (A:T and G:C) are difficult to distinguish, since they have very little influence on the overall thermal denaturation profile [66]. Sharma et al. showed that HRM could be a rapid and reliable method for the diagnosis of MDR-TB in cases of extrapulmonary TB [67]. In another study, Mu et al. examined this method for detecting DR-TB in formalin-fixed or paraffin‑embedded tissues; they found that the sensitivity of HRM for antibiotics such as rifampin, isoniazid, levofloxacin, and moxifloxacin was was 95.00%, 96.00%, 100%, and 100%, respectively, while its specificity was 95.15%, 95.92%, 94.69%, and 89.92%, respectively [68]. The stability of our findings are confirmed by similar studies and our results also showed that HRM is a trustworthy method for detecting resistance to some antibiotics, especially isoniazid and rifampin. In the present study, the heterogeneity was significant in some cases so that it was not eliminated by subgroup analysis; this issue requires a comprehensive program. This phenomenon may be due to differences in some factors such as study design, low sample size, references methods, examining tools, and dyes used. Despite its reliability, HRM has disadvatages including: (1) this method is performed only on cultured isolates; (2) HRM results are strongly depend on the quality of the extracted DNA; (3) researchers use a variety of protocols to set up HRM (self-design program). Our study had several limitations: (1) low sample size; (2) potential heterogeneity due to differences in study design and also unreliability of results; (3) inaccessibility to raw data for further analysis; (4) failure to assess the publication bias.

Conclusion

Despite the limitations of the present study, according to previous studies, we have shown that HRM is an accurate method for diagnosing and monitoring DR-TB cases. This method, together with the results of drug susceptibility testing, seems to be a suitable strategy for rapid and inexpensive diagnosis of DR-TB cases.