Previous studies have evaluated MeltPro as a follow-on test for detecting drug resistance using cultured isolates (Hu et al. 2014), smear-positive specimens (Pang et al. 2016), and formalin-fixed, paraffin-embedded (FFPE) tissues from TB patients (Mu et al. 2021), and as an initial test for diagnosing spinal TB with pus specimens (Wang et al. 2018), and exhibited their reasonably diagnostic accuracy. In this study, we have evaluated the performance of MeltPro as an initial test for diagnosing PTB with drug-resistance detection. We used sputum, BALF, and pulmonary puncture fluid specimens from patients with presumptive PTB at two clinical centers. This is the first time such a retrospective and large sample analysis has been conducted.

Regarding the diagnosis of PTB, the overall sensitivity and specificity of MeltPro were 69.0% and 97.1%, respectively, while the overall sensitivity and specificity of Xpert MTB/RIF were 72.7% and 98.0% in the present study. There were no significant differences between the two tests. Xpert MTB/RIF is a WHO-approved molecular rapid diagnostic test and is recommended as an initial test for diagnosing TB and RIF resistance. A Cochrane Review on the diagnostic accuracy of Xpert MTB/RIF for PTB found a pooled sensitivity and specificity of 84.7% and 98.4%, respectively, using bacteriological culture as the reference standard (Zifodya et al. 2021). However, the sensitivity of Xpert MTB/RIF, based on a composite reference standard considering clinical and radiographic findings, was 72% in a previous study (Berhanu et al. 2018), which is similar to the sensitivities of MeltPro and Xpert in the current study, based on a reference standard of bacteriologically/histopathologically confirmed PTB and clinically diagnosed PTB.

In smear-positive patients, the sensitivity of MeltPro for diagnosing PTB was 97.9%, aligning closely with the WHO target product profile for this population (98%) (WHO 2021a). In smear-negative patients, both MeltPro and Xpert demonstrated similar sensitivities of 50.9% and 56.5%, respectively, in the current study. Previous research has shown that Xpert sensitivity in smear-negative, culture-positive participants ranges from 41 to 77% (Berhanu et al. 2018; Dorman et al. 2018; Mishra et al. 2020; Wang et al. 2019). The sensitivities of MeltPro for diagnosing PTB and smear-negative cases were observed to be suboptimal compared to Xpert in this study, implicating that MeltPro was incapable of identifying a greater number of true positive cases. The primary rationale behind this difference in detection capability could potentially be traced to the sample processing procedures. Notably, Xpert integrates DNA extraction, PCR amplification, and detection into a single self-enclosed test unit, and all steps are automated following sample loading (WHO 2021c). On the contrary, MeltPro involves a manual DNA extraction process preceding PCR and melting curve analysis, and this manual intervention may lead to incomplete DNA extraction and subsequent loss of DNA (Mu et al. 2021). Furthermore, the bacterial loads in smear negative patients are relatively low, thus true positive cases that cannot be identified mainly appear in these populations.

In this study, the sensitivity of MeltPro for diagnosing PTB in sputum, BALF, and pulmonary puncture fluid was 65.4%, 81.3%, and 71.0%, respectively; all of which were comparable to Xpert's performance. Prior studies have shown that both conventional detection methods and new molecular rapid diagnostic tests for PTB diagnosis in BALF specimens offer promising diagnostic potential compared to sputum specimens (Badr et al. 2022; Uddin et al. 2022). The current study's findings further support this notion, emphasizing the potential benefits of using bronchoscopy to collect BALF from patients with presumptive PTB for MeltPro testing to enhance diagnostic sensitivity. For patients with mass or nodular lesions, CT-guided percutaneous puncture of the lung was found to be a viable option for specimen collection.

Regarding the detection of RIF resistance, the sensitivity and specificity of MeltPro based on the MYCOTB results were 98.0% and 97.1%, respectively, in the current study. WHO recommends Xpert MTB/RIF as the initial test for detecting RIF resistance, and it has demonstrated high overall sensitivity (96%) and specificity (98%) when compared to phenotypic DST (WHO 2020). In the present study, based on the Xpert MTB/RIF and/or MYCOTB results, MeltPro's sensitivity and specificity increased to 98.3% and 99.2%, respectively. WHO also recommends moderate complexity automated NAATs as initial tests for detecting RIF resistance, showing overall pooled sensitivity and specificity of 96.7% and 98.9% (WHO 2021d), respectively. The data indicate that MeltPro performed with higher sensitivity as an initial test for detecting RIF resistance. The difference in phenotypic DST methods used may contribute to this variation, but the ability of MeltPro to detect heteroresistance is likely the major factor. A previous study demonstrated that a high melting curve assay can detect RIF resistance mutations down to a concentration of 5% mutant DNA (Van Rie et al. 2020), which is challenging for other tests to detect at such low levels of RIF heteroresistance.

In the present study, the sensitivity and specificity of MeltPro for detecting INH resistance were 96.4% and 96.0%, respectively, while the overall pooled sensitivity and specificity of moderate complexity automated NAATs were 86.4% and 99.2% (WHO 2021d), respectively. MeltPro's higher sensitivity may be attributed to its inclusion of the ahpC promoter region in addition to the inhA promoter region (17 to 8) and katG 315, which are the only two regions covered by moderate complexity automated NAATs. Previous studies have shown that mutations in the ahpC promoter account for 8.9–13% of total INH-resistant cases in China (Hu et al. 2014).

An important advantage of MeltPro is its ability to detect not only RIF and INH resistance but also resistance to FQ. According to the WHO's updated definition of pre-extensive drug-resistant tuberculosis (pre-XDR-TB) in 2021, it includes TB caused by M. tuberculosis strains that meet the criteria for multidrug-resistant or rifampicin-resistant tuberculosis (MDR/RR-TB) and are also resistant to any FQ (WHO 2020). A study on the prevalence of XDR-TB in a Chinese MDR-TB cohort after redefinition revealed that among a total of 425 MDR-TB isolates, 311 (73.2%) were FQ-resistant (Yao et al. 2021). These changes in definitions underscore the importance of detecting FQ resistance in high-burden countries like China. In this study, the sensitivity and specificity of MeltPro for detecting FQ resistance were 90.3% and 99.4%, respectively, while the overall pooled sensitivity and specificity of the Xpert MTB/XDR Assay (low complexity automated NAAT) were 93% and 98% (WHO 2021d), respectively, with MeltPro showing slightly lower sensitivity. It is worth noting that the low complexity automated NAAT is recommended by WHO as a follow-on test in specimens determined to be MTB-positive, but not as an initial test. In a study using sputum specimens from patients with smear-positive TB, the sensitivity and specificity of MeltPro for ofloxacin resistance against a reference standard based on phenotypic MGIT 960 DST were 83.3% and 98.1% (Pang et al. 2016), respectively. Several factors may account for the differences. Firstly, the frequencies of mutations conferring FQ resistance may vary between geographic regions, which could be the primary reason. Secondly, the relatively poor sensitivity of molecular methods in detecting heteroresistance compared to phenotypic DST methods (Torrea et al. 2019) might contribute to the variable performance of molecular methods in clinical samples, particularly in regions with high TB prevalence like China where FQ heteroresistance is relatively common (Hu et al. 2023). MeltPro also demonstrated high sensitivity (97.5%) and specificity (95.2%) for detecting STR resistance in this study. Hence, we suggest that MeltPro can serve as the initial test for detecting resistance to second-line injectable drugs in individuals with signs and symptoms of PTB.

It is worth noting that there could potentially exist other mechanisms contributing to drug resistance. Nevertheless, MeltPro possesses the capacity to exclusively identify resistance resulting from mutations encompassed within its assay. Furthermore, akin to other molecular tests, MeltPro focuses on the detection of nucleic acid sequences, rather than amino acid sequences. This approach means that even silent mutations, which do not lead to alterations in amino acids, might still identified as drug-resistant mutations.

In the current study, 12 MeltPro positive and 8 Xpert positive results were diagnosed as false positives based on the final diagnosis. Increased false positive results have been reported when using ultrasensitive molecular assays, such as Xpert, in individuals with a recent episode of tuberculosis (Huo et al. 2020). On the other hand, false positive results may arise in patients who have previously been treated with anti-TB drugs, given the fact that both Xpert and MeltPro cannot distinguish between alive and dead bacilli (Zainabadi et al. 2022).

Unlike Xpert, which utilizes a single self-enclosed real-time PCR test unit, MeltPro is compatible with nearly all mainstream real-time PCR machines. Consequently, MeltPro emerges as more convenient option compared to most other molecular tests, providing a shorter turnaround time for generating diagnostic results (David et al. 2023). Furthermore, the throughput of MeltPro exceeds that of Xpert in a single assay (96 vs 80), while its cost for detecting RIF and INH resistance is only half of Xpert's. The notable advantage of MeltPro lies in its capability to detect resistance to a broader range of anti-TB drugs, facilitating the timely initiation of an effective treatment regimen. This advantage position MeltPro as a viable alternative for the rapid simultaneous detection of MTB and drug resistance, particularly in resource-limited settings. However, the patient outcomes and cost-effectiveness associated with using MeltPro as an initial diagnostic test require further evaluation.

This study had several limitations. Firstly, although Xpert Ultra has a better TB detection capability (Chakravorty et al. 2017), we did not compare its diagnostic accuracy with MeltPro as Xpert Ultra has not yet been launched in the two centers of this study. Secondly, sequencing was not used to confirm the mutation types and to clarify discrepancies among MeltPro, Xpert, and MYCOTB. Thirdly, the fact that a proportion of patients diagnosed with PTB were culture-negative hindered further phenotypic DST analysis. Notably, this was a retrospective study conducted within the confines of two TB-specialized hospitals, which may inherently restrict the broader applicability of our results. Hence, a multicenter prospective study is deemed essential to affirm and validate the diagnostic performance of MeltPro.