In this prospective, cross-sectional, double-blind diagnostic accuracy study, presumed tuberculosis patients from the Affiliated Infectious Diseases Hospital of Zhengzhou University from November 2023 to February 2024 were included. Collection of samples and diagnosis were performed before treatment. The sample size was calculated using PASS 21.0 based on the sensitivity and specificity of the AIMLAM test. According to our previous research, the prevalence of microbiological reference standard (MRS) TB in presumed TB patients was approximately 40%, the lowest acceptable sensitivity was 50%, and the expected sensitivity was 70%. Both acceptable and expected sensitivity were 95% and α was 0.05. Power (1 − β) was set at 0.9, and the dropout rate was 10% (Gao et al., 2024; Huang et al., 2023). Results showed that we needed to include 180 presumed cases of TB.

This study followed the Standards for Reporting of Diagnostic Accuracy (STARD) guideline. The study adhered to the Helsinki Declaration and was approved by the Medical Ethics Committee of the Sixth People’s Hospital of Zhengzhou. The ethics number is IEC-KY-2023-48.

We included presumptive TB patients meeting the following criteria: 1) aged ≥18, men and women; 2) with symptoms of TB (cough of at least 2 weeks, unexplained fever, weight loss, night sweats); and 3) able to produce sputum and collected urine. Patients were excluded if they 1) had HIV and 2) were receiving tuberculosis treatment.

The definition of definite TB and possible TB was based on the “Diagnosis for pulmonary tuberculosis (WS 288-2017)” (Commission, 2018) and the WHO tuberculosis guideline (World Health Organization, 2022). Definite TB was defined as Active tuberculosis (ATB), through TB culture or acid fast bacillusor or GeneXpert. Possible TB was defined as patients not meeting the criteria of definite TB but had X-ray positive, and they 1) had symptoms and signs or 2) were TST or IGRA or TB antibody positive. Non-TB was defined as patients not eligible for definite or possible TB.

Microbiological reference standard (MRS) is a benchmark diagnostic criterion based on direct microbiological evidence of infection. In tuberculosis diagnostics, MRS typically includes methods such as sputum smear microscopy, culture, and nucleic acid amplification tests (e.g., GeneXpert), which detect the presence of Mycobacterium tuberculosis directly (Shreffler and Huecker, 2020).

Composite reference standard (CRS) involves a composite of clinical assessments, including patient history, physical examinations, radiological findings, and sometimes histopathological evidence, to diagnose an infection (Shreffler and Huecker, 2020).

Sputum, urine, and blood were collected before any treatment was involved. Sterile, dry containers were used to collect three samples of sputum, including those obtained immediately, at night, and in the early morning. These were combined into one specimen, dissolved thoroughly in 5 mL of saline solution, and stored at 4°C for examination. Approximately 5 mL of midstream urine from the patient was collected in the morning and stored at 4°C for examination. Approximately 5 mL of fasting venous blood was collected and stored at room temperature.

Urinary LAM detection was conducted using AIMLAM kits (Leide Biosciences Co., Ltd, Guangzhou, China) according to the user’s manual (Zhang et al., 2023; Gao et al., 2024; Huang et al., 2023). The performers were blinded to the clinical information of the participants. This kit utilizes a chemiluminescent immunoassay to detect LAM in urine. LAM-specific antibodies immobilized on magnetic beads capture LAM, forming a complex of magnetic bead–antibody–antigen. Subsequently, the complex binds to a luminescent label, resulting in the formation of a magnetic bead–antibody–antigen–aminoluciferin label complex. Following separation and washing, a pre-triggering solution and a triggering solution were introduced to the reaction mixture, with the level of LAM in the test sample being directly proportional to the relative light unit value obtained.

The acid-fast stain (AFB) and microscopy were performed on sputum based on the Ziehl-Neelsen stain (van Deun et al., 2008). Sputum culture was performed according to protocols (Hanna et al., 1999) previously established based on the BACTEC™ MGIT™ 960 TB system (BD Diagnostic Systems, Sparks, MD, USA). GeneXpert was performed on 1.0 mL of sputum sample according to the user’s instruction (Cepheid, Inc., Sunnyvale, CA, USA) (Giang Do et al., 2015). The performers of the above tests were masked to the AIMLAM results and clinical information of the participants.

White blood cell count (WBC), granulocyte count, lymphocyte count, erythrocyte sedimentation rate (ESR), procalcitonin (PCT), and C-reactive protein (CRP) were determined using a fully automated hematology analyzer (Mindray BC-5180 CRP).

All statistical analyses were conducted using SPSS 26.0. The diagnostic accuracy of AIMLAM compared with MRS and CRS was calculated. Continuous data were transformed into categorical data according to the following methods. Age was categorized using two cutoff values: 30 and 60 years. Blood cell count, ESR, PCT, and CRP were categorized based on commonly used clinical reference standards, as follows: WBC, 4–10 × 10⁹/L; lymphocytes, 0.8–4 × 10⁹/L; neutrophils, 1.8–6.3 × 10⁹/L; monocytes, 0.1–0.6 × 10⁹/L; ESR, 0–10 mm/h for men and 0–20 mm/h for women; CRP, 0–10 mg/L; and PCT, 0.5 ng/L. Because there were only three cases (1.21%) and one case (0.88%) in the high lymphocyte count group with LAM positive (LAM+) or LAM negative (LAM−), we grouped patients with normal and high lymphocyte counts into the ≥0.8 × 10⁹/L category. Categorical data were expressed as frequencies and percentages.

Chi-square tests or Fisher’s exact probability method was used for analysis. The sensitivity and specificity of each detection method were calculated based on MRS and CRS (Shreffler and Huecker, 2020; Kraef et al., 2022; Broger et al., 2020b).

MRS-positive patients were divided into two groups: LAM+ and LAM−. General data and blood indicators of the two groups were analyzed for differences. Logistic regression was employed to examine the association between lymphocyte count and AIMLAM. Model 1 included only lymphocyte count as the independent variable. Model 2 was adjusted for gender and age. Model 3 was further adjusted for factors in model 2 as well as diabetes and autoimmune diseases. Model 4 was additionally adjusted for variables in model 3 and blood indicators showing statistically significant differences between LAM− and LAM+ individuals. Fagan nomography was used to demonstrate the clinical utility of AIMLAM in diagnosing TB patients from presumed TB patients.

The cost per detected TB patient was calculated as the total testing cost divided by the number of true positive cases. For all presumed TB patients, the average cost per person was determined using the formula (Aliyu et al., 2014; Gao et al., 2024):

(1) Average cost for all presumed TB patients:

Average cost = (total number of individuals × unit price)/number of true positive cases

For patients with low lymphocyte counts, the average cost per person was also calculated.

The prices of all test reagents in this article are only applicable to the time of testing in this study. Currently, there is no confirmed price for AIMLAM. All costs were calculated based on the average exchange rate for the year 2023 (1 USD = 7.075 CNY), with the exchange rate sourced from the IRS website (https://www.irs.gov/individuals/international-taxpayers/yearly-average-currency-exchange-rates). A significance level of P <0.05 was considered statistically significant.

Initially, 335 patients were enrolled. Thirty-three people with HIV were excluded, 15 were excluded for not providing consent or withdrawing consent, 23 were excluded for extrapulmonary disease, 3 were excluded for receiving TB treatment, and 13 were excluded for lost follow-up. Finally, a total of 248 patients were included in the analysis. Among them, 113 were definite TB cases, 53 were possible TB cases, and 135 were non-TB ( Figure 1 ). There were 131 men and 117 women. Fifty-three were less than 30 years old, 118 were 30–60 years old, and 77 were older than 60 years. Fifty-three cases (21.47%) had diabetes mellitus (DM), and 24 cases (9.68) had a history of TB. Sixty-one cases (24.6%), 184 cases (74.19%), and 3 cases (1.21%) had low, normal, and high lymphocyte counts, respectively. The distribution of sex, DM, monocytes, lymphocytes, ESR, PCT, and CRP was significantly different between groups ( Supplementary Table 1 ).

Using MRS, the sensitivity rates of AFB, culture, GeneXpert, and AIMLAM were 27.43%, 45.13%, 74.34%, and 71.68%, respectively. The sensitivity of AIMLAM was significantly higher than that of the AFB smear and culture but showed no significant difference compared to GeneXpert. The specificity of AFB, culture, and GeneXpert was 100%, as they compromised the MRS ( Figure 2B ). The specificity of AIMLAM was 95.56% ( Figure 2A ). Using CRS, the sensitivity rates of the AFB smear, culture, GeneXpert, and AIMLAM were 18.67%, 30.72%, 50.60%, and 51.20%, respectively. The specificity of the AFB smear, culture, and GeneXpert was 100%. The specificity of AIMLAM was 97.56% ( Supplementary Figure 1 ).

Among the 113 definite TB patients, 81 were LAM-positive. Age, sex, comorbidity, and other serum variables were not statistically different between LAM+ patients and LAM− patients. The proportion of low neutrophils, low lymphocytes, high ESR, high PCT, and high CRP in the LAM+ TB group was significantly higher than that in the LAM− TB group. ( Supplementary Table 2 ). In the unadjusted model (model 1), patients with low lymphocytes were more likely to have positive LAM results (OR = 9.431, 95% CI: 2.659–33.447, P = 0.001). When adjusted for sex and age (model 2), the OR was 9.562 (95% CI: 2.476–36.928, P = 0.001). When adjusted for factors in model 2 plus DM and TB history (model 3), the OR was 10.021 (95% CI: 2.555–39.297, P = 0.001). When adjusted for factors in model 3 plus neutrophils, ESR, and CRP (model 4), the OR was 5.992 (95% CI = 1.421–25.262, P = 0.015). Among the 135 MRS− cases, 6 were LAM positive. None of the factors listed were significantly different between LAM+ and LAM− patients ( Table 1 ).

All presumed TB patients were grouped into <0.8 × 10⁹/L (64 out of 248) and ≥0.8 × 10⁹/L (184 out of 248) according to their lymphocyte counts. In the <0.8 × 10⁹/L group, the sensitivity rates were 32.56%, 44.19%, 81.40%, and 93.02% for AFB, culture, GeneXpert, and AIMLAM, respectively. In the ≥0.8 × 10⁹/L group, the sensitivity rates were 24.29%, 45.71%, 70.00%, and 58.57%, respectively. The sensitivity of AIMLAM in low lymphocyte patients was much higher than that in high lymphocyte patients (P < 0.001) ( Table 2 ).

The Fagan nomogram was employed to evaluate the clinical value of AIMLAM in diagnosing TB. The Fagan nomogram showed posttest positive and negative probabilities of 93% and 20%, respectively, under a pretest probability set at 46% ( Supplementary Figure 2 ).

For the cost-effectiveness analysis, without considering lymphocyte counts, the cost to detect a positive TB case was $48.40, $136.10, $180.27, and $129.82 for the AFB smear, culture, GeneXpert, and AIMLAM, respectively. After stratifying by lymphocyte counts, for patients with lymphocyte counts <0.8 × 10⁹/L, AIMLAM testing emerged as the most cost-effective option (excluding the AFB smear), achieving a sensitivity of 93% and a specificity of 95%. In this group, the cost of AIMLAM to detect a positive TB case decreased to $67.84 (a 47.74% reduction), which was lower than GeneXpert ($111.65, a 38.06% reduction) and culture ($94.28, a 30.73% reduction). Conversely, for patients with lymphocyte counts ≥0.8 × 10⁹/L, the difficulty of achieving a positive detection increased, and the most cost-effective diagnostic method was the AFB smear, followed by culture, AIMLAM, and finally GeneXpert (Table 3).

This stratification by lymphocyte counts highlights the optimal sequence of diagnostic methods for minimizing costs while maintaining diagnostic accuracy, offering valuable guidance for clinical decision-making. Since CBC testing is routinely performed in clinical practice, it does not contribute additional costs, further enhancing the practical utility of this approach.