Risk Factors for False-Negative T-SPOT.TB Results in Bacteriologically Confirmed Pulmonary Tuberculosis: A Retrospective Cohort Study of 4,931 Patients

Abstract

Purpose

T-SPOT.TB is an interferon-γ release assay that aids in detecting Mycobacterium tuberculosis infection but cannot distinguish latent infection from active tuberculosis. False-negative results may delay diagnosis and treatment in patients with active pulmonary tuberculosis (PTB). This study investigated factors associated with false-negative T-SPOT.TB results in bacteriologically confirmed PTB.

Patients and Methods

This retrospective cohort study included 4,931 adults with bacteriologically confirmed PTB hospitalized at Fuzhou Pulmonary Hospital, Fujian Province, China, between January 2018 and September 2025. The primary outcome was a false-negative T-SPOT.TB result. Demographic characteristics, comorbidities, radiological findings, inflammatory markers, and T-lymphocyte subset counts were collected. Multivariable logistic regression was used to identify independent predictors, while restricted cubic spline (RCS) analyses explored potential nonlinear dose–response relationships.

Results

False-negative T-SPOT.TB results occurred in 437 patients (8.86%). Compared with patients with positive results, those with false-negative results were older (median, 61 vs 57 years), had higher neutrophil-to-lymphocyte ratios (NLR; 3.85 vs 3.25), and had lower CD4⁺ T-cell counts (540 vs 594 cells/μL) (all P<0.001). Older age (OR=1.007, 95% CI: 1.001–1.014; P=0.024), chronic obstructive pulmonary disease (OR=1.848, 95% CI: 1.273–2.682; P=0.001), and elevated NLR (OR=1.018, 95% CI: 1.002–1.035; P=0.026) were independent risk factors. Diabetes mellitus, cavitary lesions, airway lesions, and higher CD4⁺ T-cell counts were independently associated with a reduced risk of false-negative results. RCS analyses demonstrated nonlinear associations for NLR (P for nonlinearity <0.001) and CD4⁺ T-cell count (P for nonlinearity=0.002), with a marked increase in false-negative risk above an NLR of 2.96 and below a CD4⁺ T-cell count of 490.7 cells/μL.

Conclusion

Advanced age, chronic obstructive pulmonary disease, elevated NLR, and reduced CD4⁺ T-cell counts increase the likelihood of false-negative T-SPOT.TB results. In high-risk patients with TB-compatible imaging, a negative T-SPOT.TB result should not preclude further microbiological evaluation or diagnostic investigation.

Introduction

Tuberculosis (TB), caused by the Mycobacterium tuberculosis complex, is an airborne infectious disease. Transmission occurs when individuals with infectious pulmonary TB expel bacilli in airborne particles, particularly during coughing, speaking, sneezing, or singing.¹ TB most commonly involves the lungs and is termed pulmonary tuberculosis (PTB). However, TB can also affect sites outside the lung parenchyma, including lymph nodes, Pericardium, bones and joints, the genitourinary tract, meninges, and other organs, resulting in extrapulmonary tuberculosis (EPTB).² Mycobacterium tuberculosis may disseminate haematogenously from an active primary pulmonary focus to affect extrapulmonary sites, with concurrent pulmonary and extrapulmonary tuberculosis frequently coexisting.²

TBremains one of the most formidable public health challenges worldwide. According to the Global Tuberculosis Report 2025 published by the World Health Organization, approximately 10.7 million new TB cases were reported globally in 2024, with 1.23 million deaths attributed to the disease, making TB the leading cause of mortality from a single infectious pathogen.³ As one of the 30 high-burden countries, China accounted for approximately 6.5% of global incident cases in 2024, reflecting a substantial and persistent disease burden.³ The WHO End TB Strategy aims to end the global TB epidemic by 2035, with targets of reducing TB incidence by 90%, TB deaths by 95%, and the proportion of TB-affected households facing catastrophic costs to zero, compared with 2015 levels.⁴ Its 2030 milestones are reductions of 80% in TB incidence and 90% in TB deaths, together with the elimination of catastrophic costs.⁴ China aims to reduce TB incidence to below 43 per 100,000 by 2030, through integrated active case-finding, preventive treatment, standardized and drug-resistant TB care, financial protection, and community-based Zero-TB initiatives.⁵ In this context, early and accurate diagnosis of TB is central to interrupting transmission and improving clinical outcomes.

The interferon-γ (IFN-γ) release assay (IGRA) represents one of the most important in vitro diagnostic tools currently available for detecting specific immune responses to Mycobacterium tuberculosis (MTB) infection.⁶ Among available IGRA platforms, T-SPOT.TB stimulates peripheral blood mononuclear cells (PBMCs) with the TB-specific antigens ESAT-6 and CFP-10 and quantifies infection status by enumerating effector T cells secreting IFN-γ, expressed as spot-forming cells (SFCs).^7,8 Compared with the conventional tuberculin skin test (TST), T-SPOT.TB is unaffected by prior Bacille Calmette-Guérin vaccination or infection with most non-tuberculous mycobacteria, thereby conferring superior diagnostic specificity.⁸ Nevertheless, a non-negligible false-negative rate has been consistently reported across multiple studies.⁹ In our retrospective cohort of 4,931 bacteriologically confirmed PTB patients, 437 patients had negative T-SPOT.TB results, corresponding to a false-negative rate of 8.86%. The clinical implications of false-negative T-SPOT.TB results warrant particular attention: in patients with bacteriologically negative suspected pulmonary TB, such results can readily lead to misdiagnosis or diagnostic delay, increase the risk of deferred initiation of anti-TB therapy, and consequently accelerate disease progression while facilitating community-level transmission. Despite these concerns, a systematic understanding of the host-related determinants of T-SPOT.TB false negativity remains lacking, as the existing literature is largely confined to small-sample studies that limit the reliability and generalisability of reported findings. Although advanced age has been identified as the most prominent risk factor in several studies,^10,11 the influence of comorbidities (eg, chronic obstructive pulmonary disease [COPD], pneumoconiosis, and diabetes mellitus [DM]), T-lymphocyte subset counts, and inflammatory markers (eg, the neutrophil-to-lymphocyte ratio [NLR]) on T-SPOT.TB sensitivity remains either contentious or insufficiently evidenced. Furthermore, the potential nonlinear dose–response relationships between these continuous variables and the risk of false negativity, as well as any clinically meaningful threshold values, have yet to be systematically characterised.

Against this background, this study aimed to systematically identify and quantify independent clinical risk factors for false-negative T-SPOT.TB results in a cohort of patients with bacteriologically confirmed pulmonary tuberculosis. Restricting the analysis to bacteriologically confirmed cases allowed us to identify host-related and radiological factors that were genuinely associated with false-negative results. Although these factors were identified in patients with confirmed disease, they may help clinicians identify patients in whom a negative T-SPOT.TB result should be interpreted with caution, particularly those with suspected pulmonary tuberculosis but negative bacteriological findings. In such patients, the presence of these risk factors should prompt further microbiological investigations, radiological evaluation, or close clinical follow-up, rather than exclusion of active tuberculosis solely on the basis of a negative T-SPOT.TB result.

Materials and Methods

Study Design and Patient Selection

This retrospective cohort study was conducted at Fuzhou Pulmonary Hospital and exclusively enrolled inpatient individuals diagnosed with bacteriologically confirmed pulmonary tuberculosis between January 2018 and September 2025. Patients with concomitant EPTB were not excluded, provided that PTB was bacteriologically confirmed on respiratory specimens; thus, EPTB was recorded as a concomitant clinical characteristic rather than as the basis for study inclusion. For patients with multiple hospitalizations during the study period, only the data from their first laboratory-confirmed diagnosis of pulmonary tuberculosis were included to avoid duplicate entries and ensure independent observations. For patients with a history of tuberculosis who subsequently developed a bacteriologically confirmed recurrent infection, no additional selection or stratification was applied. They were included according to the same eligibility criteria as other patients, and only the first hospitalization with a confirmed PTB diagnosis during the study period was retained for analysis. The diagnostic criteria for tuberculosis were based on the World Health Organization (WHO) Consolidated Guidelines on Tuberculosis, Module 3: Diagnosis.¹² The inclusion criteria were as follows: (1) a confirmed diagnosis of PTB based on a positive sputum acid-fast bacilli smear, positive mycobacterial culture, or positive molecular diagnostic test (eg, Xpert^® MTB/RIF Ultra); and (2) age ≥ 18 years. Xpert^® MTB/RIF Ultra amplifies an 81-bp core region of the rpoB gene and the multicopy insertion sequences IS1081 and IS6110. Detection of IS1081/IS6110 contributes to MTBC identification, while melt-curve analysis of four rpoB probes is used to identify rifampicin-resistance-associated mutations. Patients were excluded if they met any of the following criteria: (1) incomplete clinical data, including missing admission height or weight measurements (rendering BMI incalculable), complete blood count, serum albumin, T-lymphocyte subset counts, or T-SPOT.TB results; (2) human immunodeficiency virus (HIV) infection; (3) rheumatic or autoimmune diseases; or (4) use of immunosuppressive agents or systemic corticosteroids prior to complete blood count, serum albumin, T-lymphocyte subset testing, or T-SPOT.TB assay. A total of 4,931 patients were ultimately included in the analysis. All clinical data were extracted from the hospital’s electronic medical record system, encompassing demographic characteristics, laboratory findings, radiological features, and comorbidity information.

Variables

In routine clinical practice, bacteriological examinations (sputum acid-fast bacilli smear, mycobacterial culture, and/or molecular testing) and T-SPOT.TB testing were requested concurrently as part of the diagnostic work-up at admission. The primary outcome variable was the T-SPOT.TB test result. Peripheral venous blood samples were collected within 24 hours of hospital admission and processed within 4 hours of collection for the isolation of PBMCs, using the T-SPOT.TB assay (Oxford Immunotec Ltd., Abingdon, UK) according to the manufacturer’s instructions. After stimulation with tuberculosis-specific antigens ESAT-6 and CFP-10, the effector T-cell spots secreting IFN-γ were enumerated, and results were interpreted as reactive (positive) or non-reactive (negative) based on the number of SFCs in accordance with the manufacturer’s criteria. Among bacteriologically confirmed tuberculosis patients, a non-reactive T-SPOT.TB result was defined as a false-negative, which served as a binary outcome variable for subsequent analyses. Only definitive positive or negative T-SPOT.TB results were included in the present analysis; borderline results were not included.

The exposure variables comprised demographic characteristics, laboratory parameters, and comorbidities. Demographic variables included age and sex. BMI was calculated from height and weight measured at admission (kg/m²). Laboratory parameters included complete blood count, T-lymphocyte subsets, and serum albumin; all specimens were collected within 24 hours of admission and processed immediately. Peripheral blood T-lymphocyte subset counting by flow cytometry is a routine baseline test for hospitalized patients with suspected or confirmed pulmonary tuberculosis, performed to assess immune status. Complete blood count parameters included white blood cell count (WBC), neutrophil count (NEU), neutrophil percentage (NEU%), lymphocyte count (LYM), platelet count (PLT), and monocyte count (MONO). The following systemic immune-inflammatory indices were derived from the raw blood count data: (1) NLR (NLR = absolute neutrophil count / absolute lymphocyte count); (2) platelet-to-lymphocyte ratio (PLR = platelet count / absolute lymphocyte count); (3) lymphocyte-to-monocyte ratio (LMR = absolute lymphocyte count / absolute monocyte count); and (4) systemic immune-inflammation index (SII = platelet count × absolute neutrophil count / absolute lymphocyte count). Lymphocyte subset and related variables included CD3+, CD4+, and CD8+ T-lymphocyte counts, CD45+ cell counts, and the CD4+/CD8+ T-lymphocyte ratio. Comorbidities included DM, hypertension, COPD, pneumoconiosis, EPTB, and tuberculous meningitis (TBM), all of which were confirmed based on prior medical history and admission diagnosis records and were incorporated into the analysis as binary variables (present/absent). Radiological features included cavitary lesions, airway lesions, bronchiectasis, and pleural effusion. Airway lesions were defined as radiological evidence of tracheal or bronchial involvement, primarily manifesting as airway wall thickening, irregular luminal narrowing, or occlusion attributable to endobronchial tuberculosis. All pulmonary imaging findings were independently reviewed by two radiologists, with discrepancies resolved through consensus discussion, and were treated as binary variables. All clinical data were complete, with no missing values.

Statistical Analysis

Statistical analyses were performed using R software (version 4.3.3, R Foundation for Statistical Computing, Vienna, Austria). Continuous variables were tested for normality using the Shapiro–Wilk test. Normally distributed data were expressed as mean ± standard deviation (SD) and compared using Welch’s t-test; non-normally distributed data were presented as median (interquartile range, IQR) and compared using Mann–Whitney U-test. Raincloud plots were used to visualize the distribution of continuous variables, integrating raw data points, box plots, and kernel density estimates. Categorical variables were expressed as frequencies (percentages) and compared using Chi-square test or Fisher’s exact test when appropriate. Univariate logistic regression was performed to identify potential risk factors associated with false-negative T-SPOT.TB results. Variables with P < 0.05 in univariate analysis were included in multivariate logistic regression. Results were reported as odds ratios (ORs) with 95% confidence intervals (CIs). Restricted cubic spline (RCS) regression with 4 knots was employed to explore potential non-linear relationships between continuous variables (Age, NLR, and CD4⁺ T-cell count) and the probability of false-negative results, adjusting for the same set of covariates that showed statistical significance (P < 0.05) in the univariate analysis to ensure consistency with the multivariable regression model. The overall association and non-linearity were evaluated by Wald tests. Forest plots were generated to visualize the multivariable regression results. A two-sided P value < 0.05 was considered statistically significant.

Results

Baseline Characteristics

Among 4,931 bacteriologically confirmed PTB patients, 4,494 (91.14%) had positive and 437 (8.86%) had false-negative T-SPOT.TB results. Among the 4,931 included patients, a total of 1,437 patients (29.14%) had concomitant EPTB, including 50 patients with concomitant tuberculous meningitis; no patient was included on the basis of EPTB alone. False-negative patients were older (median 61 [IQR 49–70] vs 57 [42–68] years; P < 0.001). Compared with T-SPOT.TB-positive patients, those with false-negative results showed higher neutrophil percentage (0.70 vs 0.68, as a proportion of total leukocytes; P = 0.002) and NLR (3.85 vs 3.25; P < 0.001), and lower lymphocyte count (1.26 vs 1.37 × 10⁹/L; P = 0.001), platelet count (256 vs 270 × 10⁹/L; P = 0.015), and LMR (2.49 vs 2.74; P = 0.018). All circulating lymphocyte-related counts were significantly reduced in the false-negative group, including T-lymphocyte subsets such as CD3⁺ (935 vs 1,025 cells/μL; P < 0.001), CD4⁺ (540 vs 594 cells/μL; P < 0.001), and CD8⁺ (354 vs 373.5 cells/μL; P = 0.004), as well as CD45⁺ lymphocytes (1,386 vs 1,494 cells/μL; P < 0.001). Sex, BMI, total white blood cell count, and CD4⁺/CD8⁺ ratio did not differ between groups (all P > 0.05).

COPD (9.2% vs 4.0%; P < 0.001) and pneumoconiosis (27% vs 21%; P = 0.003) were more prevalent in the false-negative group, while DM was less frequent (14% vs 21%; P < 0.001). Cavitary lesions (39% vs 51%; P < 0.001) and airway lesions (8.7% vs 22%; P < 0.001) were less common on chest imaging in false-negative patients. Baseline characteristics for both groups are detailed in Table 1, with key parameters compared in Figure 1.

Table 1.

Baseline Characteristics of Bacteriologically Confirmed PTB Patients Stratified by T-SPOT.TB Results

Variables	T-SPOT.TB (+) n = 4,494	T-SPOT.TB (-) n = 437	p
Age,M (Q1, Q3)	57.00 (42.00, 68.00)	61.00 (49.00, 70.00)	<0.001
BMI,M (Q1, Q3)	20.25 (18.37, 22.43)	20.52 (18.10, 22.68)	0.643
ALB,M (Q1, Q3)	38.60 (35.30, 41.50)	38.80 (35.20, 41.80)	0.992
WBC,M (Q1, Q3)	6.70 (5.44, 8.39)	6.77 (5.42, 8.62)	0.359
NEU,M (Q1, Q3)	4.44 (3.36, 6.05)	4.62 (3.36, 6.38)	0.131
NEU%,M (Q1, Q3)	0.68 (0.60, 0.75)	0.70 (0.61, 0.78)	0.002
LYM,M (Q1, Q3)	1.37 (1.02, 1.81)	1.26 (0.91, 1.76)	0.001
Hb,M (Q1, Q3)	125.00 (113.00, 136.00)	125.00 (112.00, 139.00)	0.848
PLT,M (Q1, Q3)	270.00 (218.00, 341.00)	256.00 (204.00, 346.00)	0.015
MONO,M (Q1, Q3)	0.50 (0.38, 0.65)	0.49 (0.37, 0.68)	0.615
NLR,M (Q1, Q3)	3.25 (2.12, 5.12)	3.85 (2.19, 6.19)	<0.001
PLR,M (Q1, Q3)	197.05 (136.22, 296.26)	201.11 (131.54, 327.06)	0.209
LMR,M (Q1, Q3)	2.74 (1.81, 4.06)	2.49 (1.54, 3.92)	0.018
SII,M (Q1, Q3)	881.70 (495.51, 1,622.16)	936.58 (505.72, 2,027.97)	0.055
CD3,M (Q1, Q3)	1,025.00 (746.00, 1,373.00)	935.00 (641.00, 1,296.00)	<0.001
CD3%,M (Q1, Q3)	0.71 (0.64, 0.76)	0.70 (0.62, 0.76)	0.015
CD4,M (Q1, Q3)	594.00 (422.00, 818.00)	540.00 (331.00, 749.00)	<0.001
CD4%,M (Q1, Q3)	0.41 (0.35, 0.47)	0.39 (0.33, 0.46)	0.003
CD8,M (Q1, Q3)	373.50 (251.00, 527.00)	354.00 (231.00, 487.00)	0.004
CD8%,M (Q1, Q3)	0.26 (0.20, 0.32)	0.25 (0.20, 0.32)	0.915
CD4/CD8,M (Q1, Q3)	1.59 (1.16, 2.18)	1.53 (1.11, 2.12)	0.104
CD45,M (Q1, Q3)	1,494.00 (1,097.00, 1,962.00)	1,386.00 (974.00, 1,874.00)	<0.001
Sex, n (%)			0.136
Male	3,302 (73%)	336 (77%)
Female	1,192 (27%)	101 (23%)
Treatment, n (%)			0.061
Retreatment	489 (11%)	61 (14%)
Initial Treatment	4,005 (89%)	376 (86%)
DM, n (%)			<0.001
No	3,532 (79%)	376 (86%)
Yes	962 (21%)	61 (14%)
Hypertension, n (%)			0.636
No	3,828 (85%)	368 (84%)
Yes	666 (15%)	69 (16%)
Concomitant EPTB, n (%)			0.999
No	3,184 (71%)	310 (71%)
Yes	1,310 (29%)	127 (29%)
Concomitant TBM, n (%)			0.126
No	4,445 (99%)	436 (100%)
Yes	49 (1.1%)	1 (0.2%)
COPD, n (%)			<0.001
No	4,312 (96%)	397 (91%)
Yes	182 (4.0%)	40 (9.2%)
Cavitary lesions, n (%)			<0.001
No	2,224 (49%)	268 (61%)
Yes	2,270 (51%)	169 (39%)
Airway Lesions, n (%)			<0.001
No	3,508 (78%)	399 (91%)
Yes	986 (22%)	38 (8.7%)
Pneumoconiosis, n (%)			0.003
No	3,544 (79%)	317 (73%)
Yes	950 (21%)	120 (27%)
Bronchiectasis, n (%)			0.956
No	3,733 (83%)	364 (83%)
Yes	761 (17%)	73 (17%)
Pleural_Effusion, n (%)			0.150
No	3,405 (76%)	317 (73%)
Yes	1,089 (24%)	120 (27%)

Notes: All enrolled patients had bacteriologically confirmed PTB. “Concomitant EPTB” indicates EPTB diagnosed in addition to PTB. TBM was recorded as a specific form of concomitant EPTB. Data are presented as median (interquartile range, IQR) for continuous variables. M represents median, and Q₁ and Q₃ represent the 25th and 75th. Categorical variables are presented as n (%). Age, years.

Abbreviations: BMI, body mass index (kg/m²); ALB, albumin (g/L); WBC, white blood cell (×10⁹/L); NEU, neutrophil (×10⁹/L); NEU%, neutrophil percentage; LYM, lymphocyte (×10⁹/L); Hb, haemoglobin (g/L); PLT, platelet (×10⁹/L); MONO, monocyte (×10⁹/L); NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; LMR, lymphocyte-to-monocyte ratio; SII, systemic immune-inflammation index (calculated from neutrophil, platelet, and lymphocyte counts). Lymphocyte subset analysis: CD3⁺ indicates total T lymphocytes (cells/μL); CD3⁺% indicates the percentage of CD3⁺ T lymphocytes relative to total lymphocyte population; CD4⁺ indicates helper T cells (cells/μL, %); CD4⁺% indicates the percentage of CD4⁺ T lymphocytes relative to total lymphocyte population; CD8⁺ indicates cytotoxic T lymphocytes (cells/μL, %); CD8⁺% indicates the percentage of CD8⁺ T lymphocytes relative to total lymphocyte population; CD4⁺/CD8⁺ indicates the ratio of CD4⁺ to CD8⁺ T lymphocytes; CD45⁺ indicates total lymphocytes (cells/μL). PTB, pulmonary tuberculosis; T-SPOT.TB, T-cell spot test for tuberculosis. DM, Diabetes Mellitus; EPTB, Extrapulmonary Tuberculosis; COPD, Chronic Obstructive Pulmonary Disease.

Figure 1.

Comparative analysis of clinical characteristics between T-SPOT plus and T-SPOT minus groups.

Notes: The distribution of clinical characteristics between the T-SPOT positive group (red) and the negative group (blue) is presented. Raincloud plots were employed to visualize the data, integrating raw observation points (left), box plots indicating the median and interquartile range (middle), and kernel density plots representing the distribution density of the data (right).

Independent Risk Factors for False-Negative T-SPOT.TB

Variables that showed statistically significant differences in baseline characteristics between groups, along with sex (included a priori), were considered as candidate predictors. Given the potential collinearity among certain candidates, pre-consolidation of related indices was performed prior to model construction. Specifically, NEU and LYM are the direct arithmetic components of the NLR and are therefore mathematically dependent on it; accordingly, NLR was selected as the representative inflammatory marker. Although the LMR and NLR both incorporate lymphocyte counts, they capture distinct dimensions of immune homeostasis by reflecting the balance of lymphocytes to monocytes and neutrophils to lymphocytes, respectively. Consequently, both were retained as independent variables in our analysis. With respect to lymphocyte subsets, CD3⁺ T cell count represents the arithmetic sum of CD4⁺ and CD8⁺ T cell counts, CD45⁺ cell count exhibits a high degree of correlation with CD3⁺ T cell count, and the percentage-based indices (eg, CD3%, CD4%) carry redundant information relative to their corresponding absolute counts. Accordingly, absolute CD4⁺ and CD8⁺ T cell counts were selected as representative T-lymphocyte subset variables. Following this consolidation process, the variables ultimately entered into univariable logistic regression analysis were: age, sex, DM, COPD, pneumoconiosis, pulmonary cavitation, airway lesions, NLR, LMR, and absolute CD4⁺ and CD8⁺ T cell counts.

On univariable analysis, age, DM, COPD, pneumoconiosis, cavitary lesions, airway lesions, NLR, CD4⁺ T-cell count, and CD8⁺ T-cell count were each significantly associated with false-negative risk, whereas sex and LMR were not (P = 0.122 and P = 0.981, respectively).

On multivariable logistic regression (Table 2; Figure 2), older age (OR 1.007, 95% CI 1.001–1.014; P = 0.024), COPD (OR 1.848, 95% CI 1.273–2.682; P = 0.001), and elevated NLR (OR 1.018, 95% CI 1.002–1.035; P = 0.026) independently increased the odds of false-negative T-SPOT.TB. DM (OR 0.636, 95% CI 0.475–0.852; P = 0.002), cavitary lesions (OR 0.628, 95% CI 0.509–0.774; P < 0.001), airway lesions (OR 0.364, 95% CI 0.258–0.512; P < 0.001), and CD4⁺ T-cell count (OR 1.000, 95% CI 0.999–1.000; P = 0.035) were independently associated with reduced odds of false-negative results. Pneumoconiosis did not reach significance in the multivariable model (P = 0.090,).

Table 2.

Univariable and Multivariable Logistic Regression Analyses of Factors Associated with False-Negative T-SPOT.TB Results Among Patients with Bacteriologically Confirmed Pulmonary Tuberculosis

Variables	Univariable Analysis		Multivariable Analysis
	OR (95% CI)	P	OR (95% CI)	P
Age	1.012 (1.006–1.018)	<0.001	1.007 (1.001–1.014)	0.024
Sex	0.833 (0.66–1.05)	0.122	1.055 (0.817–1.363)	0.681
DM	0.596 (0.45–0.788)	<0.001	0.636 (0.475–0.852)	0.002
COPD	2.387 (1.67–3.413)	<0.001	1.848 (1.273–2.682)	0.001
Pneumoconiosis	1.412 (1.131–1.763)	0.002	1.223 (0.969–1.543)	0.090
Cavitary lesions	0.618 (0.505–0.756)	<0.001	0.628 (0.509–0.774)	<0.001
Airway lesions	0.339 (0.241–0.476)	<0.001	0.364 (0.258–0.512)	<0.001
NLR	1.025 (1.01–1.041)	0.001	1.018 (1.002–1.035)	0.026
LMR	1 (0.967–1.033)	0.981	–	–
CD4	0.999 (0.999–1)	<0.001	1 (0.999–1)	0.035
CD8	0.999 (0.999–1)	0.010	1 (0.999–1)	0.838

Notes: The outcome was a false-negative T-SPOT.TB result (T-SPOT.TB negative) among patients with bacteriologically confirmed pulmonary tuberculosis. Reference categories were female sex, no diabetes mellitus, no chronic obstructive pulmonary disease, no pneumoconiosis, absence of cavitary lesions, and absence of airway lesions. ORs for age, NLR, LMR, CD4⁺ T-cell count, and CD8⁺ T-cell count were estimated per one-unit increase.

Abbreviations: OR, odds ratio; CI, confidence interval.

Figure 2.

Forest Plot of Multivariable Logistic Regression Analysis Identifying Independent Factors Associated with False-Negative T-SPOT.TB Results in Bacteriologically Confirmed PTB.

Notes: Each point represents the OR estimate for the corresponding variable, with horizontal lines indicating the 95% CI. The vertical dashed line at OR = 1.0 represents the null effect. Variables to the right of the dashed line indicate increased odds of false-negative T-SPOT.TB results, while variables to the left indicate decreased odds (ie, independent protective factors).

Dose–Response Relationships: RCS Analysis

RCS analyses were conducted to examine the continuous dose–response relationships of age, NLR, and CD4⁺ T-cell count with false-negative T-SPOT.TB risk (Figure 3).

Figure 3.

Dose-Response Relationships Between Age, NLR, CD4⁺ T-Cell Count, and the Risk of False-Negative T-SPOT.TB Results: A RCS Analysis.

Notes: (A) Age; (B) NLR; (C) CD4⁺ T-cell count. Solid lines represent the estimated odds ratios (ORs), and shaded areas denote the 95% confidence intervals (CIs). Reference values (OR = 1.0) were set at the median of each respective variable. Knots were placed at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles. In panel B, the red dot indicates the nadir (the point of lowest risk) at an NLR of 2.79. In panel C, the red dot highlights a cut-off point at 588 cells/μL. The gray density plots at the bottom of each panel illustrate the distribution of the study population. P-overall indicates the overall statistical significance; P-non-linear reflects the significance of the non-linear test.

For age, no statistically significant overall association was found (P-overall = 0.145), and the trend was strictly linear (P-non-linear = 0.835), indicating a marginal, proportional increase in false-negative risk across the age spectrum without evidence of a threshold or inflection.

For NLR, a highly significant and non-linear association was observed (P-overall < 0.001; P-non-linear < 0.001). Risk remained relatively stable at low NLR values, with the nadir at NLR = 2.96, beyond which false-negative odds escalated steeply.

For CD4⁺ T-cell count, a strong non-linear inverse association was identified (P-overall < 0.001; P-non-linear = 0.002), with a risk-transition point at 490.7 cells/μL. Below this threshold, false-negative risk rose sharply as CD4⁺ counts declined; above it, risk decreased progressively and stabilized at lower odds with higher cell counts.

Discussion

In this retrospective cohort of 4,931 patients with bacteriologically confirmed PTB, the false-negative rate of T-SPOT.TB was 8.86%, a finding that aligns with the range established in the current literature. For instance, a systematic review and meta-analysis of 17 studies reported a pooled sensitivity of 92% for T-SPOT.TB in culture-confirmed cases, implying a false-negative incidence of approximately 8%,¹⁰ which is highly consistent with our observation. Conversely, substantially higher false-negative rates have been documented elsewhere; for example, Sato et al reported a rate of 22.3% in a Japanese cohort of 193 PTB patients.¹³ Such variations are likely attributable to differences in study populations, immune status, and methodology.

Older age was a consistent and independent predictor of false-negative T-SPOT.TB (OR 1.007, 95% CI 1.001–1.014, P = 0.024). RCS analysis confirmed a strictly linear dose–response relationship (P-non-linear = 0.835), indicating that false-negative risk increases proportionally across the age spectrum without evidence of a threshold effect. This linear age-dependent pattern is mechanistically sound, as immunosenescence progressively impairs Mycobacterium tuberculosis-specific CD4+ T cell responses and impairs IFN-γ secretion,^14–16 thereby reducing SFCs to levels below the diagnostic positivity threshold. The systematic review by Yamasue et al, which included 14 studies evaluating age as a risk factor, identified advanced age as the most consistently reported predictor of false-negative IGRA results (pooled OR 2.06, 95% CI 1.68–2.52), irrespective of regional TB incidence.¹⁰ The TBNET study similarly identified advanced age as the sole independent predictor of false-negative QuantiFERON-TB Gold In-Tube (QFT-GIT) results in multivariable analysis (OR 1.04, 95% CI 1.02–1.07).¹¹ In summary, findings across multiple scales, including our own study, the TBNET cohort (OR 1.04), and the systematic review by Yamasue et al (OR 2.06), collectively identify advanced age as a universal and independent predictor of false-negative results in TB-specific immune assays. However, the relatively lower effect size observed in our study (OR 1.007) suggests a more gradual risk escalation compared to QFT-GIT-based reports. This implies that T-SPOT.TB’s standardized PBMC concentration may partially buffer the impact of age-related lymphopenia, offering enhanced diagnostic robustness in the elderly.

COPD was among the strongest independent risk factors identified (OR 1.848, 95% CI 1.273–2.682, P = 0.001). COPD is characterised by chronic airway inflammation with persistent neutrophilic infiltration and systemic inflammation.¹⁷ In this cohort, the false-negative group exhibited significantly elevated NLR (3.85 vs 3.25, P < 0.001) and neutrophil percentage (0.70 vs 0.68, P = 0.002), consistent with the neutrophil-driven immune imbalance characteristic of COPD. In patients with COPD complicated by PTB, the reduced frequencies of natural killer cells and CD4+ T lymphocytes¹⁸ can substantially impair MTB-specific T-cell responses. This impairment, in turn, increases the likelihood of false-negative T-SPOT.TB results. Given that few studies have focused on the association between COPD and IGRA false-negativity, our findings provide further evidence from a large clinical population that may serve as a useful reference for managing such diagnostic challenges.

Pneumoconiosis was more prevalent in the false-negative group (27% vs 21%, P = 0.003) and was significant on univariable analysis (OR 1.412). Although it did not reach statistical significance in the multivariable model (OR 1.223, P = 0.090), its OR retained clinical relevance. Yu et al reported that the sensitivity of QFT-GIT in the TB-silicosis group was 54.3%, which was significantly lower than the 93.0% observed in the TB-only group.¹⁹ Chronic silica exposure drives the upregulation of programmed cell death protein 1 (PD-1) in peripheral CD4⁺ T-cell, which subsequently attenuates IFN-γ synthesis and liberation via the PD-1/ Programmed death-ligand 1 signaling axis, thereby undermining the diagnostic sensitivity of the QFT-GIT assay.¹⁹ This immune exhaustion mechanism provides biologically plausible grounds for the borderline association observed in this study. The failure to reach significance in the multivariable model may partially reflect competitive collinearity with COPD or limited statistical power to detect small effect sizes.

Cavitary lesions (OR 0.628, 95% CI 0.50914–0.774, P < 0.001) and airway lesions (OR 0.364, 95% CI 0.258–0.512, P < 0.001) were independently protective against false-negative results. The T-SPOT.TB-positive group had substantially higher rates of both cavitation (51% vs 39%) and airway lesions (22% vs 8.7%) compared with the false-negative group. Cavitation and endobronchial spread are morphological correlates of higher mycobacterial burden and more intense local immune activation. Greater antigen exposure drives more vigorous MTB-specific T-cell proliferation and IFN-γ secretion, favouring assay positivity.^20–22 Sato et al provided direct CT-based evidence supporting this mechanism: patients with only minimal infiltrates (nodules or micronodules without consolidation or cavitation) had a false-negative rate as high as 86%, attributed to insufficient antigen-driven immune stimulation.¹³ The concordant protective direction observed across both studies substantiates a mechanistic relationship between imaging-defined disease burden and T-SPOT.TB diagnostic sensitivity, suggesting that antigen-driven immune activation is a pivotal determinant of assay reactivity. Given this mechanistic framework, the absence of cavitation or airway dissemination on chest imaging should not be interpreted as evidence against active tuberculosis; rather, it should prompt heightened clinical vigilance for false-negative T-SPOT.TB results in patients with otherwise compatible clinical presentations.

Elevated NLR was an independent risk factor (OR 1.018, 95% CI 1.002–1.035, P = 0.026), with RCS analysis revealing a highly significant nonlinear association (P-overall < 0.001; P-non-linear < 0.001) and a critical inflection at NLR = 2.96, above which false-negative risk escalated steeply. This threshold effect implies that the immunosuppressive impact of neutrophil-driven inflammation on T-SPOT.TB becomes substantially amplified when inflammatory imbalance surpasses a critical level. Biologically, elevated NLR may suppress T-SPOT.TB assay positivity through two complementary mechanisms: on one hand, lymphocytopenia directly diminishes the pool of IFN-γ-producing effector cells available for antigen stimulation;²³ on the other hand, activated neutrophils suppress T-cell proliferation through multiple convergent mechanisms, including Mac-1-mediated immunological synapse formation with localized H₂O₂ generation at the cell–cell contact interface, and secretion of immunosuppressive cytokines IL-10 and TGF-β.²⁴ Zhang et al similarly identified elevated NLR as an independent predictor of IGRA false negativity using propensity score matching;²⁵ the present study extends this evidence by characterising the nonlinear dose–response relationship and defining a clinically applicable threshold.

Lower CD4⁺ T-cell count was independently associated with increased false-negative risk (P = 0.035), with RCS analysis revealing a strongly nonlinear inverse relationship (P-non-linear = 0.002) and a critical inflection at 490.7 cells/μL, below which false-negative risk rose sharply. The diagnostic principle of T-SPOT.TB is fundamentally predicated on the IFN-γ secretion capacity of MTB-specific CD4⁺ T-cell in response to ESAT-6 and CFP-10 stimulation;²⁶ when CD4⁺ T-cell counts are inadequate to sustain a sufficient pool of SFCs, assay readouts are liable to fall below the established positivity threshold. Yamasue et al’s meta-analysis confirmed low lymphocyte counts as a major independent risk factor for IGRA false negativity (pooled OR 2.68, 95% CI 2.00–3.61).¹⁰ Notably, CD8⁺ T-cell count was significant on univariable analysis (P = 0.010) but lost significance in the multivariable model (P = 0.838), suggesting its effect is mediated or confounded by CD4⁺ T-cell count, which emerges as the more decisive cellular determinant of T-SPOT.TB performance. Given that the per-unit OR for CD4⁺ T-cell count is close to 1.000, its clinical relevance is more meaningfully captured by the nonlinear threshold identified on RCS analysis than by the OR alone.

Furthermore, lymphocyte count was found to be significantly lower in the false-negative group than in the true-positive group in the present study (1.26 vs 1.37 × 10⁹/L, P = 0.001). This hematological finding is consistent with the meta-analysis reported by Yamasue et al,¹⁰ in which a reduced peripheral blood lymphocyte count was identified as the second most significant risk factor for false-negative results, surpassed only by advanced age (pooled OR 2.68, 95% CI 2.00–3.61). Lymphocytopenia directly reduces the pool of IFN-γ-responding effector T cells, an effect that is particularly pronounced in whole-blood-based QuantiFERON-TB Gold (QFT) assays;²³ although T-SPOT.TB partially mitigates this limitation by normalizing to a fixed number of PBMCs, severe lymphocyte depletion may nonetheless result in a significant decline in spot counts.²⁶

DM was independently associated with reduced false-negative odds (OR 0.636, 95% CI 0.475–0.852, P = 0.002), with the T-SPOT.TB-positive group having a higher prevalence of diabetes (21% vs 14%). Walsh et al reported that among bacteriologically confirmed TB patients, those with diabetes or chronic hyperglycaemia (glycated haemoglobin [HbA1c] ≥ 6.5%) were significantly more likely to have a positive QFT-G result (adjusted OR 2.4, 95% CI 1.1–5.0), with an estimated QFT sensitivity of approximately 81% in diabetic versus 63% in non-diabetic patients; T-SPOT.TB sensitivity was 93% overall and did not differ significantly by diabetes status.²⁷ Chronic hyperglycaemia, especially when glycemic control remains suboptimal, may predispose T-SPOT.TB assay readouts toward positivity, potentially attributable to an elevated mycobacterial burden in diabetic hosts that drives more potent antigen-specific IFN-γ responses.²⁷ Corroborating evidence from a large cross-sectional study (n = 5,302) found that persons with undiagnosed diabetes had significantly higher QFT TB antigen values than normoglycaemic individuals (0.96 vs 0.50 IU/mL, P < 0.05), with fasting plasma glucose independently and positively associated with QFT antigen levels (P = 0.017).²⁸ However, the generalisability of the findings by Walsh et al may be constrained by the limited T-SPOT.TB sample size (n = 43). Moreover, the modulatory role of DM in IGRA performance is far from conclusive, with conflicting data reported in diverse clinical settings. For instance, while some evidence suggests a heightened immune response under hyperglycemic conditions, others, such as Choi et al,²⁹ have reported a paradoxical decline in QFT sensitivity in diabetic individuals. Given the potential confounding effects of diabetes subtype, glycaemic control, and bacillary burden, this finding warrants cautious interpretation and further prospective investigation.

Neither sex nor LMR reached statistical significance in either univariable or multivariable analyses. Regarding sex, several prior studies including the TBNET study and the systematic review by Yamasue et al have consistently failed to identify sex as an independent determinant of false-negative IGRA results.^10,11 This observation is further corroborated by the findings of the present study. The lack of significance for LMR may reflect the functional duality of monocytes in tuberculosis immunity: monocytes can act as antigen-presenting cells to drive T-cell responses,³⁰ yet simultaneously suppress local inflammation through the secretion of immunosuppressive cytokines such as IL-10.³¹ This bidirectional immunomodulatory role may attenuate the net predictive utility of LMR as a marker of false-negative results. Although both NLR and LMR incorporate a lymphocyte component, NLR demonstrated superior predictive value in the present study, potentially because the neutrophil-to-lymphocyte imbalance captured by NLR exerts a more directional and biologically specific suppressive effect on T-cell IFN-γ responses.

HIV-associated immunosuppression can impair IGRA performance, particularly in individuals with low CD4+ T-cell counts, and may increase the likelihood of indeterminate or negative results.³² Previous studies have associated younger age with false-negative T-SPOT.TB results,³³ and younger children have been reported to produce lower average levels of IFN-γ than older children.³⁴ These observations raise the possibility that the ongoing maturation of cellular immune responses during childhood may influence T-SPOT.TB reactivity, thereby potentially contributing to false-negative or indeterminate results. Therefore, these populations were excluded to minimize the influence of established causes of impaired T-SPOT.TB responses and to better identify other factors genuinely associated with false-negative results in relatively immunocompetent adults with bacteriologically confirmed PTB.

A key translational implication of this study is that a negative T-SPOT.TB result should not be used alone to exclude PTB in patients with clinical or radiological findings suggestive of tuberculosis. In such patients, advanced age, COPD, elevated NLR, lower CD4⁺ T-cell count, and the absence of cavitary or airway lesions on CT may indicate an increased likelihood of a false-negative T-SPOT.TB result. When these features coexist, further microbiological or histopathological evaluation, such as bronchoalveolar lavage with Xpert MTB/RIF or tissue biopsy where appropriate, should be considered to reduce diagnostic delay.

The novelty of this study lies in its large cohort of bacteriologically confirmed PTB patients and its integrated assessment of demographic characteristics, comorbidities, CT findings, inflammatory markers, and T-lymphocyte subsets associated with false-negative T-SPOT.TB results. Restricted cubic spline analyses further characterized nonlinear associations and identified clinically informative thresholds for NLR (approximately 2.96) and CD4⁺ T-cell count (approximately 490.7 cells/μL). These findings provide an evidence-based framework for interpreting negative T-SPOT.TB results in suspected PTB and may also inform the assessment of patients with both bacteriologically and immunologically negative results, who require comprehensive diagnostic evaluation.

Several limitations of this study warrant acknowledgement. First, as a single-centre retrospective cohort study, generalisability to populations with differing TB epidemiology may be limited. Second, although patients receiving immunosuppressive agents and those with HIV co-infection were systematically excluded, the severity of glycaemic control in diabetic patients was not fully captured and may represent a residual confounder. Third, the genetic determinants of IGRA performance were not ascertained. This is a potential limitation because markers such as HLA-DRB1*0701 are independently associated with false-negative results¹⁰ and may therefore represent important unmeasured confounders. Fourth, this study was restricted to adults with bacteriologically confirmed PTB based on respiratory specimens. Although patients with concomitant EPTB were included, the findings should not be directly extrapolated to patients with isolated EPTB. These limitations highlight the need for multicentre, prospective studies to externally validate the risk stratification framework proposed herein.

Conclusion

This study involving 4,931 patients demonstrates a false-negative rate of 8.86% for T-SPOT.TB in bacteriologically confirmed pulmonary tuberculosis, suggesting that a negative result alone is insufficient to rule out the disease. Advanced age, comorbid COPD, and an elevated neutrophil to lymphocyte ratio above 2.96 significantly increase the risk of false-negative outcomes, while higher CD4 positive T cell counts and the presence of cavitary or airway lesions are associated with improved assay sensitivity. These findings highlight the critical influence of host immune status and localized lung pathology on the performance of interferon gamma release assays. When clinical or radiological evidence strongly supports a tuberculosis diagnosis, a negative T-SPOT.TB result must be interpreted with caution, particularly in elderly or immunocompromised patients, to avoid diagnostic delays and ensure timely initiation of treatment.

Funding Statement

This study was supported by the Fuzhou Municipal Science and Technology Bureau (Grant No. 2025-S-036), and by the Fuzhou Municipal Key Specialty Improvement and Upgrading Project for Municipal Hospitals (Fuzhou Key Specialty of Tuberculosis and Non-Tuberculous Mycobacterial Diseases).

Abbreviations

ALB, Albumin; BMI, Body mass index; CFP-10, Culture filtrate protein 10; CI, Confidence interval; COPD, Chronic obstructive pulmonary disease; CT, Computed tomography; DM, Diabetes mellitus; ESAT-6, Early secreted antigenic target 6 kDa; EPTB, Extrapulmonary tuberculosis; HbA1c, Glycated haemoglobin; IFN-γ, interferon-γ; IGRA, interferon-γ release assay; IQR, Interquartile range; LMR, Lymphocyte-to-monocyte ratio; LYM, Lymphocyte count; MONO, Monocyte count; MTB, Mycobacterium tuberculosis; NEU, Neutrophil count; NEU%, Neutrophil percentage; NLR, Neutrophil-to-lymphocyte ratio; OR, Odds ratio; PBMCs, Peripheral blood mononuclear cells; PD-1, Programmed cell death protein 1; PLR, Platelet-to-lymphocyte ratio; PLT, Platelet count; PTB, Pulmonary tuberculosis; QFT / QFT-GIT, QuantiFERON-TB Gold In-Tube; RCS, Restricted cubic spline; SD, Standard deviation; SFCs, Spot-forming cells; SII, Systemic immune-inflammation index; TB, Tuberculosis; TBM, Tuberculous meningitis; TST, Tuberculin skin test; WBC, White blood cell count; WHO, World Health Organization.

Data Sharing Statement

The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.

Ethics Statement

This retrospective study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board/Ethics Committee of Fuzhou Pulmonary Hospital of Fujian Province (approval number: 2023-010[Research]-02). The Institutional Review Board/Ethics Committee of Fuzhou Pulmonary Hospital of Fujian Province waived the requirement for individual informed consent due to the retrospective nature of the study and because all patients had previously provided written consent at admission for their clinical data to be used in future research. All patient data were anonymized and deidentified prior to analysis to ensure confidentiality.

Disclosure

The authors report no conflicts of interest in this work.