Authors: Bo Xiang, Wenchuang Zeng, Minzhi Wu, Junkai Ye, Zhijian Lin, Yueting Jiang
Categories: Research, Mycobacterial infections, DNA microarray chip, Diagnostic performance, Risk factors, Clinical diagnosis
Source: BMC Infectious Diseases
Authors: Bo Xiang, Wenchuang Zeng, Minzhi Wu, Junkai Ye, Zhijian Lin, Yueting Jiang
China has a high burden of Mycobacterium tuberculosis (MTB), with increasing infections from non-tuberculous mycobacteria (NTM). Rapid and accurate identification is crucial for effective clinical management. This study aimed to evaluate the diagnostic performance of DNA microarray chip and identify risk factors associated to mycobacterial infections.
Between January 2021 and June 2023, 973 patients with presumptive mycobacterial infections were recruited from the First Affiliated Hospital of Guangzhou Medical University. Various detection methods were used for etiological diagnosis, and their performance was compared. Multivariable logistic regression analysis determined clinical risk factors for mycobacterial positivity.
Mycobacterial infection was confirmed in 438 patients, with 254 (58.0%) infected with MTB, 183 (41.8%) with NTM, and one co-infected with two. DNA microarray chip demonstrated the highest true positive rate (TPR) for mycobacterial detection at 96.8% and a true negative rate (TNR) of 90.8%. For MTB, the TPR was 94.5% with a TNR of 99.3%. Risk factors included cough/expectoration, pulmonary consolidation, cavitation, and bronchiectasis for mycobacterial infection, while smoking history and diabetes mellitus were independent risk factors for MTB infection.
DNA microarray chip offers a sensitive and specific diagnostic approach for mycobacterial infections, potentially enhancing clinical diagnosis and informing targeted interventions.
The online version contains supplementary material available at 10.1186/s12879-025-12402-3.
The genus Mycobacterium is categorized into the MTB complex, NTM, and M. leprae [1]. Tuberculosis, an infectious disease caused by MTB complex, was the leading cause of death from a single infectious agent before the COVID-19 pandemic [2]. Although primarily affecting the lungs (pulmonary TB), TB can also impact other organs. NTM encompasses a broad class of mycobacteria excluding the MTB complex and M. leprae. Out of over 200 identified species, only a select few are pathogenic to humans, presenting conditional pathogenicity [1]. The prevalence of NTM disease, which affects various organs and tissues, is rapidly rising, thus posing significant public health challenges. In China, the situation is particularly dire, with NTM infections surpassing those in many global regions [3].
NTM diseases, often clinically indistinguishable from TB, frequently result in misdiagnoses, leading to inappropriate treatments. Thus, accurate and swift identification of Mycobacterium species is crucial for effective management of MTB and NTM infections. However, the current diagnostic methods have significant limitations. Diagnosis of MTB: Traditional diagnostic approaches, such as AFB (Acid-fast bacilli) smears, lack sensitivity and specificity, and fail to distinguish between MTB and NTM. Culture is time-consuming, and rapid culture takes 1–2 weeks, sometimes even 4–8 weeks. The Xpert MTB/RIF recommended by the WHO can rapidly detect MTB and rifampicin resistance, but it cannot identify NTM strains. Alternative methods, such as the p-nitrobenzoic acid selective medium and MPB64 antigen tests, also depend on prior culturing [4, 5]. Despite these advancements, there are still challenges in differentiating MTB from detecting drug-resistant strains in resource-limited areas. Identification of NTM Molecular diagnostics, like Line Probe Assay (LPA), although capable of detecting multiple indicators, has limited detection sensitivity for samples with low loading [6]. Real-Time PCR has high detection sensitivity, but the instrument and cost are relatively high. Next-generation sequencing, offer high-resolution identification but at a high cost [7, 8]. MALDI-TOF-MS provides rapid and accurate results but requires cultured colonies and sophisticated equipment [9]. Currently, direct homologous gene or sequence comparisons are considered the gold standard for NTM species identification [10, 11]. Indirect homologous gene or sequence comparison methods, such as DNA microarray chip, is a rapid, accurate diagnostic technique for Mycobacterium species identification and simultaneous detection of 17 common mycobacterial infections [12, 13]. Clinical microbiology laboratories urgently require cost-effective, high-throughput methods capable (1) Distinguishing MTB from NTM rapidly. (2) Accurately identifying NTM species to guide species-specific therapy. (3) Overcoming the limitations of culture and complex molecular assays. Given its cost efficiency, ability to simultaneously detect 17 mycobacterial species, and relatively short turnaround time (6.5 h), large-scale clinical validation of DNA microarray chip is critical for improving mycobacterial diagnosis and management. Here, we summarized 973 clinical samples of suspected mycobacterial infections and evaluated the clinical value of DNA microarray chip in Mycobacterium identification through a large sample cohort.
This retrospective study analyzed 7,668 samples from patients with suspected mycobacterial pulmonary infection at the First Affiliated Hospital of Guangzhou Medical University, a referral hospital, between January 2021 and June 2023. Mycobacterial species identification was performed on all patients using DNA microarray chip technology. Exclusion To enhance the accuracy of our study, we excluded outpatients and instances where the same patient underwent multiple tests during the study period. A total of 473 patients with positive mycobacterial species identification from DNA microarray chip were included in the study group. For comparative analysis, we randomly selected 500 patients with negative mycobacterial species identification from DNA microarray chip, who were admitted to our hospital during the same period, to form the control group (Additional file 2). Data on demographic characteristics, clinical symptoms (e.g., fever, cough), underlying comorbidities (e.g., hypertension, diabetes), radiological manifestations (e.g., nodules, cavitation), and laboratory parameters (e.g., CRP, WBC) were retrospectively collected from the hospital’s comprehensive electronic medical record system. A standardized data extraction form was utilized to ensure systematic and uniform data acquisition across all study participants.
A sample was spread evenly across the slide to form an oval film measuring 10 mm by 20 mm and left to air dry. The slide was fixed by brief flaming and stained with carbolic acid red, heated until steaming (non-boiling), and maintained for 5 min. After rinsing, it was decolorized with acid-alcohol for 1–2 min, repeated as needed, and rinsed again. Finally, it was stained with methylene blue for 30–60 s. Microscopically, acid-fast bacilli appeared red against a light blue background.
Samples for mycobacteria isolation culture included sputum, deep sputum, bronchoscopic sputum, and bronchoalveolar lavage fluid. Each sample was treated with 1–2 volumes of 4% NaOH, vigorously shaken and mixed for 1–2 min, and then allowed to stand at room temperature for 15–20 min. Subsequently, PBS (pH 6.8) was added, thoroughly mixed and centrifuged. The pellet was resuspended in 0.5-1 mL of PBS and cultured on Lowenstein-Jensen solid media or in liquid media using Bactec-MGIT960 (BD, USA) systems at 37 °C.
M. tuberculosis ATCC 27,294 and M. abscessus BNCC 267,506 were used as positive control in the culture processes. Negative culture results should only be reported after a minimum incubation period of eight weeks for solid media and six weeks for liquid media.
The sample types and preprocessing protocols for real-time PCR identification of mycobacterial species were consistent with those detailed in Section “Culture”. The PCR fluorescent probe detection of mycobacterial genes was conducted following the manufacturer’s protocol, involving sample preparation, DNA extraction, PCR amplification, and data interpretation. Reaction mix (40 µL) included PCR mix (38 µL) and template DNA (2 µL). Amplification: 37 °C/5 min, 94 °C/1 min, followed by 40 cycles of 95 °C/5 s and 60 °C/30 s. The detection utilized FAM (465–510 nm) fluorescence channels (Ct < 37 for positive).
The kit (careTB PCR Assay, QIAGEN, ShenZhen, China) included a positive control (plasmid DNA) and a negative control (TE buffer).
5 mL peripheral venous blood samples were collected and placed in heparin anticoagulant tubes. The testing process followed the manufacturer’s instructions (X.DOT-TB, TB Healthcare, Guangdong, China). If the number of spots in the test panel minus the number of spots in the negative control panel was ≥ 11, and the number of spots in the negative control panel was ≤ 10, the result was considered positive for MTB. Alternatively, if the number of spots in the negative control panel was 11–20, and the number of spots in the test panel was at least twice the number of spots in the negative control panel, the result was also considered positive for MTB. Although T-SPOT.TB cannot be used to diagnose active tuberculosis, it may serve as an auxiliary tool to assess the host’s tuberculosis infection status.
The sample was placed into a treatment tube, followed by shaking after the treatment solution was added, and then allowed to stand for 15–20 min. Subsequently, 2 mL of the treated sample solution was absorbed and added into the cartridge. The cartridge was inserted into the GeneXpert instrument (Cepheid, USA), and the results were interpreted approximately 2 h later.
The sample types and preprocessing protocols for DNA microarray chip identification of mycobacterial species were consistent with those detailed in Section “Culture”. The nucleic acid from the aforementioned samples was extracted using the 36 nucleic acid rapid extractor (CapitalBio, China). Subsequent steps including PCR amplification, chip hybridization, chip scanning, and result interpretation were performed following the standard procedures outlined in the mycobacterial species identification kit (CapitalBio, China). The biochip could identify 17 mycobacterial species, including MTB, M. intracellulare, M. avium, M. gordonae, M. kansasii, M. fortuitum, M. scrofulaceum, M. gilvum, M. terrae, M. chelonae/M. abscessus, M. phlei, M. nonchromogenicum, M. marinum/M. ulcerans, M. aurum, M. szulgai-M. malmoense, M. xenopi, and M. smegmatis [12].
The kit included a positive control (plasmid DNA containing a fragment of the target gene segment) and a negative control (sequence independent plasmid DNA).
The sample types and preprocessing protocols for mNGS identification of pathogens were consistent with those detailed in Section “Culture”. The sample was placed in a sterile sputum container, stored at 4 °C, and sent to CapitalBio (China) for mNGS detection.
MTB positivity (meeting any one criterion):
Microbiological Positive for culture or AFB staining, and positive for GeneXpert, or TB-DNA detection.Two or more positive Including AFB staining, T-SPOT.TB, mNGS, or DNA microarray chip.Clinical Only DNA microarray chip positive, with supporting clinical symptoms and imaging findings consistent with MTB infection.
NTM positivity (meeting any one criterion):
Microbiological Positive for culture or AFB staining, and positive for DNA microarray chip or mNGS detection.Clinical Only DNA microarray chip positive, with supporting clinical symptoms and imaging findings consistent with NTM infection.
All statistical analyses were performed using SPSS 26.0 statistical software. Statistical data are presented as cases or percentages, and Pearson’s chi-squared test or Fisher’s exact test are used for comparison between groups. Measurement data conforming to normal distribution by t-test. P < 0.05 was considered significant, and all tests were two-tailed. For the comparison of diagnostic performance, the areas under the receiver operating characteristic (ROC) curves of different tests were compared using the DeLong test for correlated ROC curves. Prior to constructing the logistic regression model, this study completed systematic data First, three-category variables were properly transformed into dummy variables with clearly defined reference groups; subsequently, multiple imputation was employed to handle the data with a missing proportion of less than 40%, while for the data with a missing proportion greater than 40%, the variables were deleted. Collinearity analysis demonstrated that all variables exhibited variance inflation factor (VIF) < 5, meeting modeling requirements. Multivariable logistic regression analysis was performed to identify independent risk factors associated with mycobacterial infection (vs. other etiologies) and to differentiate between MTB and NTM infections. All biologically plausible and clinically relevant variables, including demographic, clinical, radiological, and laboratory indicators, were included in the initial model. The ‘Enter’ method was employed, whereby all selected variables are simultaneously entered into the model, ensuring that the effect of each variable is adjusted for all others. This approach allows for the evaluation of the independent contribution of each factor based on clinical rationale rather than relying solely on statistical significance from univariate screening. The results are presented as odds ratios (ORs) with corresponding 95% confidence intervals (CIs).
A total of 973 patients with suspected mycobacterial infection were enrolled in this study (Table 1, Additional file 1, 2). There were 560 (57.6%) males and 413 (42.4%) females. Their average age was 55.0 ± 15.7 years (To ensure anonymity, the ages in the additional file 1 are shown as age groups). Of these patients, 438 were diagnosed with mycobacterial infection through clinical symptoms, imaging and etiology. Characteristics and baselines of the patients were listed in Table 1. Statistical analyses were performed for the characteristics and baselines.
Table 1Characteristics and baselines of the patients with mycobacterial infectionsMycobacterial (n = 438)Other etiological (n = 535)P-ValueMTB (n = 254)NTM (n = 183)P-ValueAge (years)54.6 ± 15.655.3 ± 15.70.46952.9 ± 16.957.0 ± 13.40.005 > 60174(39.7%)217(40.6%)0.792100(39.4%)73(39.9%)0.913Sex Male, n(%)233(53.2%)327(61.1%)0.013164(64.6%)68(37.2%)< 0.001 Female, n(%)205(46.8%)208(38.9%)90(35.4%)115(62.8%)Smoking history126(28.8%)183(34.2%)0.07099(39.0%)26(14.2%)< 0.001Clinical manifestationsn = 438n = 535n = 254n = 183 Fever39(8.9%)69(12.9%)0.04920(7.9%)19(10.4%)0.364 Cough/Expectoration323(73.7%)287(53.6%)< 0.001172(67.7%)150(82.0%)0.001 Short of breath90(20.5%)119(22.2%)0.52248(18.9%)41(22.4%)0.369 Panting8(1.8%)11(2.1%)0.7976(2.4%)2(1.1%)0.539 Hemoptysis70(16.0%)54(10.1%)0.00623(9.1%)47(25.7%)< 0.001 Chest pain29(6.6%)30(5.6%)0.51022(8.7%)7(3.8%)0.045 Dyspnea24(5.5%)20(3.7%)0.19317(6.7%)7(3.8%)0.194Underlying diseasesn = 438n = 535n = 254n = 183 Malignant tumor84(19.2%)176(32.9%)< 0.00143(16.9%)41(22.4%)0.152 Diabetes mellitus71(16.2%)79(14.8%)0.53561(24.0%)10(5.5%)< 0.001 Hypertension61(13.9%)114(21.3%)0.00340(15.7%)21(11.5%)0.204 Heart disease24(5.5%)28(5.2%)0.86511(4.3%)13(7.1%)0.209 COPD43(9.8%)67(12.5%)0.18522(8.7%)21(11.5%)0.330CT findingsn = 357n = 452n = 199n = 157 Nodule256(71.7%)291(64.4%)0.027154(77.4%)102(65.0%)0.010 Consolidation108(30.3%)134(29.6%)0.85264(32.2%)44(28.0%)0.399 Cavitation100(28.0%)64(14.2%)< 0.00170(35.2%)29(18.5%)< 0.001 Ground-glass opacity51(14.3%)137(30.3%)< 0.00132(16.1%)19(12.1%)0.287 Mottling/patchy shadow294(82.4%)364(80.5%)0.509160(80.4%)133(84.7%)0.290 Stripe shadow161(45.1%)232(51.3%)0.07893(46.7%)67(42.7%)0.445 Reduced lung volume69(19.3%)73(16.2%)0.23823(11.6%)46(29.3%)< 0.001 Translucent shadow92(25.8%)167(36.9%)0.00151(25.6%)40(25.5%)0.974 Bronchostenosis54(15.1%)62(13.7%)0.57039(19.6%)15(9.6%)0.009 Bronchiectasis209(58.5%)147(32.5%)< 0.00180(40.2%)128(81.5%)< 0.001 Tree-in-bud sign76(21.3%)35(7.7%)< 0.00134(17.1%)42(26.8%)0.027 Pleural thickening92(25.8%)121(26.8%)0.74944(22.1%)47(29.9%)0.093 Pleural effusion67(18.8%)155(34.3%)< 0.00142(21.1%)24(15.3%)0.161 Pericardium involvement18(5.0%)66(14.6%)< 0.0018(4.0%)10(6.4%)0.315 Lymphadenopathy151(42.3%)204(45.1%)0.42090(45.2%)60(38.2%)0.184Laboratory tests ESRn = 332n = 3880.886n = 183n = 1480.784 Normal107(32.2%)127(32.7%)58(31.7%)49(33.1%) Elevated225(67.8%)261(67.3%)125(68.3%)99(66.9%) CRPn = 154n = 2150.039n = 86n = 680.113 Normal79(51.3%)87(40.5%)49(57.0%)30(44.1%) Elevated75(48.7%)128(59.5%)37(43.0%)38(55.9%) PCTn = 351n = 466< 0.001n = 200n = 1500.673 Normal280(79.8%)298(63.9%)161(80.5%)118(78.7%) Elevated71(20.2%)168(36.1%)39(19.5%)32(21.3%) WBCn = 406n = 523< 0.001n = 235n = 1700.559 Normal323(79.6%)359(68.6%)191(81.3%)131(77.1%) Elevated54(13.3%)131(25.0%)28(11.9%)26(15.3%) Decreased29(7.1%)33(6.3%)16(6.8%)13(7.6%)Immune proteins IgGn = 262n = 2850.063n = 150n = 1110.302 Normal187(71.4%)211(74.0%)102(68.0%)84(75.7%) Elevated70(26.7%)60(21.1%)44(29.3%)26(23.4%) Decreased5(1.9%)14(4.9%)4(2.7%)1(0.9%) IgAn = 262n = 2850.002n = 150n = 1110.473 Normal231(88.2%)243(85.3%)135(90.0%)95(85.6%) Elevated28(10.7%)22(7.7%)14(9.3%)14(12.6%) Decreased3(1.1%)20(7.0%)1(0.7%)2(1.8%) IgMn = 262n = 2850.018n = 150n = 1110.605 Normal206(78.6%)202(70.9%)116(77.3%)89(80.2%) Elevated19(7.3%)16(5.6%)13(8.7%)6(5.4%) Decreased37(14.1%)67(23.5%)21(14.0%)16(14.4%) C3n = 262n = 2850.021n = 150n = 1110.175 Normal146(55.7%)138(48.4%)90(60.0%)56(50.5%) Elevated1(0.4%)9(3.2%)0(0.0%)1(0.9%) Decreased115(43.9%)138(48.4%)60(40.0%)54(48.6%) C4n = 261n = 2850.252n = 149n = 1110.336 Normal165(63.2%)162(56.8%)94(63.1%)70(63.1%) Elevated13(5.0%)13(4.6%)5(3.4%)8(7.2%) Decreased83(31.8%)110(38.6%)50(33.6%)33(29.7%) CH50n = 261n = 2840.249n = 149n = 1110.186 Normal67(25.7%)91(32.0%)32(21.5%)35(31.5%) Elevated189(72.4%)187(65.8%)114(76.5%)74(66.7%) Decreased5(1.9%)6(2.1%)3(2.0%)2(1.8%)COPD: Chronic Obstructive Pulmonary Disease; ESR: Erythrocyte Sedimentation Rate; CRP: C-reactive protein; PCT: Procalcitonin; WBC: White Blood Cell; IgG/A/M: Immunoglobulin G/A/M; C3/ Complement 3/4; CH50: 50% hemolytic unit of complement. P-values < 0.05 were considered significant
Among the 438 patients with mycobacterial infection, 254 patients (58.0%) were infected with MTB, 183 patients (41.8%) were infected with NTM, and one patient (0.2%) had a co-infection of MTB and M. avium-intracellulare complex. The NTM infections included 115 cases (62.8%) of M. avium-intracellulare complex, 56 cases (30.6%) of M. chelonae/M. abscessus, seven cases (3.8%) of M. kansasii, two cases (1.1%) of M. fortuitum, one case (0.5%) of M. gilvum, and two cases (1.1%) of unclassified mycobacteria (identified by culture method) (Fig. 1A). According to the infection of patients, MTB infections were more common in the age group of 20 to 79 years, while NTM were more common in the age group of 40 to 79 years (Fig. 1C).
Fig. 1Epidemiological characteristics of mycobacteria. (A). Mycobacterial species distribution in pulmonary infection. (B). Mycobacterial species distribution identified as positive by DNA microarray chip but discordant with clinical outcomes. (C). Age group distribution of mycobacterial infections
Additional file 1, 2 detailed the etiological detection outcomes and clinical manifestations of the 973 patients enrolled in the study with suspected mycobacterial infections. The TPR, TNR, positive predictive value (PPV), and negative predictive value (NPV) for mycobacteria detection by DNA microarray chip were 96.8%, 90.8%, 89.6%, and 97.2%, respectively. Although culture is considered the gold standard for mycobacterial infections [14], it exhibited a relatively low TPR of 32.1%. DNA microarray chip demonstrated the highest TPR (96.8%) albeit with a slightly lower TNR (90.8%) (Table 2). In the specific context of MTB detection, the DNA microarray chip not only maintained the highest TPR (94.5%) but also achieved a high TNR (99.3%) (Table 2).
Table 2Diagnostic performance of mycobacterial and MTB infection detection methodsDiagnostic formycobacterialinfectionTPRTNRPPVNPV+LR-LRDNA microarraychip (n = 973)96.8% (95.2%-98.5%)90.8% (88.4%-93.3%)89.6% (86.9%-92.4%)97.2% (95.8%-98.6%)10.57 (8.09–13.81)0.035 (0.021–0.059)AFB (n = 973)42.2% (37.6%-46.9%)100.0% (99.3%-100.0%)100.0% (98.0%-100.0%)67.9% (64.6%-71.2%)∞ (N/A)0.578 (0.534–0.626)Culture (n = 559)32.1% (26.7%-37.4%)100.0% (98.6%-100.0%)100.0% (96.1%-100.0%)57.7% (53.3%-62.2%)∞ (N/A)0.679 (0.627–0.735)mNGS (n = 634)88.8% (84.0%-93.7%)97.7% (96.3%-99.0%)92.9% (88.8%-96.9%)96.3% (94.6%-98.0%)38.12 (21.18–68.46)0.115 (0.074–0.177) Diagnostic for
MTB infection
TPR
TNR
PPV
NPV
+LR
-LR DNA microarraychip (n = 973)94.5% (91.7%-97.3%)99.3% (98.7%-99.9%)98.0% (96.2%-99.7%)98.1% (97.1%-99.1%)135.01 (56.5-323.5)0.055 (0.033–0.092)TB-DNA (n = 949)79.0% (73.9%-84.0%)100.0% (99.5%-100.0%)100.0% (98.2%-100.0%)92.9% (91.1%-94.8%)∞ (N/A)0.210 (0.166–0.267)Xpert (n = 889)79.3% (74.1%-84.5%)100.0% (99.4%-100.0%)100.0% (98.0%-100.0%)93.2% (91.3%-95.1%)∞ (N/A)0.207 (0.161–0.266)T-SPOT.TB (n = 569)71.2% (63.9%-78.6%)83.7% (80.2%-87.2%)60.1% (52.8%-67.4%)89.4% (86.4%-92.4%)4.37 (3.44–5.55)0.344 (0.266–0.445)mNGS (n = 634)81.8% (73.8%-89.9%)99.5% (98.8%-100.0%)96.0% (91.6%-100.0%)97.1% (95.8%-98.5%)148.76 (47.9–460.0)0.183 (0.117–0.285)TPR: True Positive Rate; TNR: True Negative Rate; PPV: Positive Predictive Value; NPV: Negative Predictive Value; +LR: Positive Likelihood Ratio; -LR: Negative Likelihood Ratio
The AUC for mycobacterial species detection by the DNA microarray chip, AFB, culture, and mNGS were 0.945 (95% confidence interval [CI] 0.917 to 0.973), 0.687 (95% CI 0.641 to 0.733), 0.664 (95% CI 0.619 to 0.708), and 0.938 (95% CI 0.907 to 0.968) (Fig. 2A). Analysis of the differences of AUC values in paired samples revealed that the DNA microarray chip outperformed AFB (P < 0.001) and culture (P < 0.001), while it was comparable to mNGS (P = 0.662) in detecting mycobacterial species. For MTB detection, the AUC values for the DNA microarray chip, TB-DNA, GeneXpert, T-SPOT.TB, and mNGS were 0.898 (95% CI 0.836 to 0.961), 0.866 (95% CI 0.795 to 0.937), 0.884 (95% CI 0.817 to 0.951), 0.744 (95% CI 0.668 to 0.820), and 0.889 (95% CI 0.825 to 0.954), respectively (Fig. 2B). Analysis of the differences of AUC values in paired samples indicated that the DNA microarray chip was superior to T-SPOT.TB (P = 0.001), and equivalent to TB-DNA (P = 0.395), GeneXpert (P = 0.705), and mNGS (P = 0.743) in the detection of MTB.
Fig. 2ROC curve analysis of the detection performance of different mycobacterial detection methods. (A). ROC curve analysis evaluating the diagnostic performance of DNA microarray chip, AFB, culture, and mNGS for mycobacteria detection. (B). ROC curve analysis evaluating the diagnostic performance of DNA microarray chip, culture, TB-DNA, T-SOPT.TB, Xpert, and mNGS for MTB detection
DNA microarray chip not only enable rapid and accurate identification of MTB and various NTM infections, but also contribute to clinical diagnosis through their high detection sensitivity. In this study, five cases (2.0%, 5/255) of MTB and 44 cases (23.9%, 44/184) of NTM, which were found to be positive only detected by DNA microarray chip, and were ultimately determined to be Mycobacterium positive after combining with clinical symptoms, thus altering the final diagnosis. Of the 44 NTM cases, 30 were identified as M. avium-intracellulare complex, nine as M. chelonae/M. abscessus, three as M. kansasii, and two as M. fortuitum (Additional file 1).
In summary, the results of DNA microarray chip testing were consistent with the final clinical diagnosis in 93.5% of cases (910/973), with 424 cases being double positive and 486 cases being double negative. In inconsistent results, DNA microarray chip failed to detect 14 cases of mycobacterial infection, and 49 cases were considered to be false positives or colonization rather than infection in clinical diagnosis. The 49 cases included five cases of MTB and 44 cases of NTM, M. avium-intracellulare complex was the most common (18.4%), followed by M. fortuitum (12.2%) and M. chelonae/M. abscessus (8.2%) (Fig. 1B).
Univariate analyses revealed significant differences between mycobacterial infections and other etiologies across multiple demographic, clinical, and radiological features (Table 1). The mycobacterial infection group demonstrated a lower proportion of males (P = 0.013) and a lower incidence of fever (P = 0.049), but a higher frequency of cough/expectoration (P < 0.001) and hemoptysis (P = 0.006). Furthermore, underlying conditions such as malignant tumor (P < 0.001) and hypertension (P = 0.003) were less prevalent in this group. Radiological manifestations, including nodules (P = 0.027), cavitation (P = 0.004), bronchiectasis (P < 0.001), and tree-in-bud appearances (P < 0.001), were more common in the mycobacterial infection group. In contrast, ground-glass opacities (P < 0.001), translucent shadows (P = 0.001), pleural effusion (P < 0.001), and pericardial involvement (P < 0.001) were observed less frequently. Significant inter-group differences were also noted in several laboratory markers, including CRP, PCT, WBC, IgA, IgM, and C3.
Multivariable logistic regression analysis identified several independent factors associated with mycobacterial infection (vs. other etiologies), with all variables demonstrating acceptable collinearity (VIF < 5; Table 3). The presence of cough/expectoration, consolidation, cavitation, and bronchiectasis were significantly associated with increased odds of mycobacterial infection (OR = 2.222, P < 0.001; OR = 1.511, P = 0.042; OR = 2.245, P = 0.005; OR = 2.279, P < 0.001). Conversely, malignant tumor, hypertension, ground-glass opacity, translucent shadow, pericardium involvement, elevated PCT, and decreased WBC were associated with decreased odds of mycobacterial infection (OR = 0.587, P = 0.003; OR = 0.615, P = 0.023; OR = 0.53, P = 0.001; OR = 0.646, P = 0.016; OR = 0.458, P = 0.01; OR = 0.621, P = 0.024; OR = 0.574, P = 0.018) (Table 3).
Table 3Multivariable logistic regression analysis of clinical characteristics
Comparative analysis between MTB and NTM infection groups also revealed distinct profiles (Table 1). Patients with MTB infection were younger (P = 0.005) and included a lower proportion of females (P < 0.001). This group also presented with lower rates of cough/expectoration (P = 0.001), hemoptysis (P < 0.001), reduced lung volume (P < 0.001), bronchiectasis (P < 0.001), and tree-in-bud appearances (P = 0.027). In contrast, a history of smoking (P < 0.001), chest pain (P = 0.045), diabetes mellitus (P < 0.001), and radiological findings of nodules (P = 0.010), cavitation (P < 0.001), and bronchostenosis (P = 0.009) were more frequent in the MTB group. No significant differences were observed in laboratory tests or immune proteins between the two groups.
In the corresponding multivariable model for distinguishing MTB from NTM infections (Table 3), a smoking history and the presence of diabetes mellitus were independent factors associated with increased odds of MTB infection (OR = 2.173, P = 0.049; OR = 3.833, P = 0.003). Female, fever, hemoptysis, and bronchiectasis were associated with decreased odds of MTB infection (OR = 0.512, P = 0.029; OR = 0.343, P = 0.027; OR = 0.438, P = 0.023; OR = 0.293, P = 0.001).
Treatment strategies for pulmonary diseases caused by various mycobacterial infections, including MTB and NTM, vary significantly. Notably, even among pulmonary diseases induced by different NTM species, the therapeutic choices differ. Approximately 30% of patients suspected of having multidrug-resistant TB (MDR-TB) were later found to have NTM infections, suggesting a potentially higher prevalence of NTM [3, 15]. The rising incidence of these infections can also be attributed to the diversification of diagnostic methods and enhanced diagnostic accuracy [3]. Therefore, accurate species identification of mycobacteria is crucial before treatment to facilitate targeted interventions. DNA microarray chip technology is a novel diagnostic approach recognized for early mycobacterial infection diagnosis and has been widely used in clinical settings to analyze sputum samples for both pulmonary and extrapulmonary mycobacterial infections [16–19].
The M. avium-intracellulare complex and M. abscessus complex are significant contributors to lung diseases globally [20, 21]. Additionally, M. kansasii is clinically significant and often transmitted via aerosols [22]. Geographical variations in the distribution of NTM across mainland China are influenced by environmental factors such as temperature and humidity. For instance, in Shanghai, M. kansasii is the most common NTM (45.0%) [23]. In addition, by reviewing the relevant reports in China, it was found that M. avium-intracellulare (46.2%) was most common in Beijing. However, in Shenzhen, M. chelonae/M. abscessus (43.0%) was most common. In this study, we conducted DNA microarray chip assays on 7,668 lower respiratory tract samples suspected of mycobacterial infection, identifying 515 positive cases of MTB and NTM, resulting in a positivity rate of 6.7%. The M. avium-intracellulare complex was most frequently detected (62.8%), followed by M. chelonae/M. abscessus (30.6%) and M. kansasii (3.8%). The lower M. kansasii infection rate may be due to COVID-19 control measures reducing its aerosol transmission. Of the 98 cases identified only using DNA microarray chip technology in this study, 49 were confirmed as mycobacterial infections by clinicians, and the remaining 49 were classified as false positives. The observed false-positive results could potentially stem from either bacterial colonization or methodological inaccuracies, such as cross-contamination. On the other hand, 14 false-negative findings may be attributed to low bacterial load, which falls below the detection threshold of the assay, or inherent technical limitations of the diagnostic methodology. This discrepancy underscores the need for further verification through PCR or alternative pathogen detection methods to ascertain the accuracy of results were mycobacterial species identification conflicts with clinical diagnosis.
The study involved 973 patients suspected of mycobacterial infections with pulmonary lesions. DNA microarray chip technology showed superior detection performance compared to AFB staining and culture, and comparable performance to mNGS. It also demonstrated high efficacy in diagnosing MTB infections, superior to culture and T-SPOT.TB, and equivalent to other advanced tests such as TB-DNA, GeneXpert and mNGS (Fig. 2; Table 2) [19, 24–26]. Compared with AFB/culture diagnosis for MTB/NTM: DNA microarray directly detects MTB and NTM nucleic acids, eliminating the requirement for viable bacteria (uninfluenced by antibiotics) and achieving higher sensitivity than microscopy. Compared to T-SPOT diagnosis for MTB: As an immunological assay, T-SPOT is prone to false positives. In contrast, nucleic acid tests specifically identify MTB DNA. Comparable to TB-DNA/GeneXpert/mNGS diagnosis for MTB: All these methods rely on nucleic acid amplification, targeting conserved MTB genes, thus presenting similar sensitivity/specificity. The findings underscore the efficacy of DNA microarray chip technology in the precise diagnosis of mycobacterial infections, highlighting its crucial role in the management and prevention of these diseases.
Compared to other diagnostic methods, DNA microarray chip technology has high consistency with DNA sequencing results, a simplified nucleic acid extraction process, a shortened diagnostic cycle, reduced manual burden, accurate and simple result interpretation, and lower cost [12, 19]. Cost-benefit analysis revealed DNA microarray (~ 37), T-SPOT.TB (~ 50), and mNGS (~ 570). Technically, it detects more mycobacterial species than GeneXpert and T-SPOT.TB. Operationally, its 6.5 h TAT exceeds GeneXpert’s 2 h but outperforms mNGS (24 h), T-SPOT.TB (24–48 h), and MALDI-TOF (culture-dependent). Therefore, considering comprehensively, the DNA microarray chip assay is a simple, rapid, high-throughput, and economical diagnostic detection method that is ideal for the clinical screening of large numbers of samples.
Although the pathophysiology of MTB and NTM diseases exhibit similar phenotypes, differences exist in some disease manifestations [27]. Our multivariable logistic regression analysis confirmed that a history of smoking and diabetes mellitus are risk factors for MTB infection, while female, bronchiectasis and hemoptysis were associated with NTM infections, consistent with previous findings [27–30]. Additionally, this study identified fever was risk factors for NTM infection (Table 3). Smoking induces immunosuppression and disrupts the respiratory tract barrier [31], diabetes leads to metabolic dysfunction and immune decline [32]. These factors significantly increase susceptibility to MTB infection through distinct pathological mechanisms. Bronchiectasis impairs mucociliary clearance and promotes mucus stasis, and hemoptysis-a common clinical feature of bronchiectasis, further elevates the risk of NTM infection [33, 34]. Beyond comparing MTB and NTM infections, this study also examined risk factors distinguishing mycobacterial infections from other etiologies. Multivariable logistic regression analysis revealed cough/expectoration, pulmonary cavitation, consolidation and bronchiectasis as significant correlates of mycobacterial infections. Analyzing risk factors and clinical characteristics associated with different infections improves diagnostic accuracy and treatment strategies, ensuring that patients receive timely and appropriate care.
However, this study has several limitations. First, the DNA microarray chip used in this study targets only a predefined set of mycobacterial species, thereby limiting its ability to identify subtypes outside the detection panel. A negative result indicates the absence of the 17 common Mycobacterium species included in the assay but cannot entirely rule out mycobacterial infection. In cases where clinical and radiological findings remain suggestive of such infection, unbiased detection methods such as mNGS may be employed for further clarification. Second, the retrospective design of the study may lead to incomplete data collection and potential selection bias, despite efforts to minimize these issues. Moreover, by including only definitively diagnosed cases, the study cohort may overrepresent patients with clearer clinical manifestations, possibly resulting in an overestimation of the diagnostic accuracy of the DNA chip. Future prospective studies involving consecutive, unselected patient cohorts undergoing both the DNA chip assay and gold standard testing are warranted to validate its real-world performance.
This retrospective study demonstrates the significant utility of DNA microarray chip technology in the accurately diagnosing of mycobacterial infections within a clinical setting. Although culture remains the gold standard for mycobacterial infections identification within current medical systems, DNA microarray chip serve as a valuable complementary method for mycobacterial infections. Incorporating such advanced diagnostic tools into routine practice is essential to enhance the control and prevention of mycobacterial diseases, enabling more targeted and timely healthcare interventions.
Below is the link to the electronic supplementary material.
Supplementary Material Additional file 1 Clinical information for 973 patients with presumptive mycobacterial infection
Supplementary Material Additional file 2 Flowchart of the enrolled cases and summary of the test results of different mycobacterial detection methods