Abstract

Objective:

To compare the clinical and Imaging characteristics between children with pediatric pulmonary tuberculosis (PTB) and children with pneumonia who were initially suspected of PTB, and to explore valuable indicators for differential diagnosis.

Methods:

This retrospective study enrolled pediatric inpatients initially suspected of PTB at Guangzhou Chest Hospital from March 2020 to May 2021. All participants were divided into the PTB group and the control group. Clinical characteristics, laboratory results, and chest computed tomography (CT) imaging findings were compared between the two groups. On the basis of univariate analysis, variables were further screened via the least absolute shrinkage and selection operator (LASSO) regression with reference to clinical significance. A stable and reliable predictive model was finally established using Firth penalized logistic regression.

Results:

A total of 63 children were enrolled in this study. Among them, 48 were finally diagnosed with PTB, including 23 cases with etiological confirmation and 25 cases with clinical diagnosis, which constituted the PTB group. The remaining 15 children were diagnosed with pneumonia and assigned to the control group. Compared with the control group, the PTB group had a significantly higher rate of tuberculosis contact history, longer pre-hospital symptom duration, and higher rates of pre-hospital antibiotic and systemic glucocorticoid use. In terms of chest CT manifestations, multiple pulmonary nodules, hilar/mediastinal/axillary lymphadenopathy, calcification, and fine linear or reticular opacities were significantly more common in the PTB group. The final predictive model was constructed based on the two most differentiating imaging features, namely multiple pulmonary nodules and hilar/mediastinal/axillary lymphadenopathy, and presented excellent diagnostic efficacy. The area under the receiver operating characteristic curve was 0.99. The corresponding sensitivity, specificity and overall accuracy were 93.75%, 100% and 95.24%, respectively.

Conclusion:

This study preliminarily developed an imaging model based on multiple pulmonary nodules and hilar/mediastinal/axillary lymphadenopathy, which may assist in distinguishing pediatric PTB from pneumonia. However, given the single-center, specialized-hospital setting of this study and the exclusion of certain patient subgroups, the generalizability of these findings is limited. Therefore, the conclusions require validation through future multicenter, large-scale prospective studies.

1 Introduction

Pulmonary tuberculosis (PTB), a chronic infectious disease caused by Mycobacterium tuberculosis (MTB) lung infection, remains a major global public health threat, and children are a highly vulnerable population. The World Health Organization (WHO) has reported a substantial burden of pediatric tuberculosis, which is frequently complicated by delayed diagnosis and unfavorable clinical outcomes (1). The diagnosis of pediatric PTB is challenging due to non-specific clinical manifestations that often overlap with common childhood respiratory diseases such as pneumonia (2). In addition, etiological confirmation is difficult in children because of the paucibacillary feature of this disease and difficulties in collecting qualified respiratory specimens. Clinicians therefore rely heavily on clinical, radiological and epidemiological evidence for diagnosis (3).

Such diagnostic uncertainty leads to a common clinical difficulty: distinguishing pediatric PTB from community-acquired pneumonia with tuberculosis-like manifestations. In clinical practice, these two diseases are easily confused with each other. Bidirectional misdiagnosis may cause disease progression, increase the risk of complications including miliary tuberculosis and intracranial tuberculosis, and adversely affect long-term prognosis. Therefore, identifying reliable clinical, laboratory, and imaging markers that can accurately distinguish between these entities at the point of initial suspicion is a critical unmet need in pediatric respiratory medicine.

In clinical practice at Guangzhou Chest Hospital, many children admitted with an initial suspicion of PTB are finally diagnosed to have pneumonia after systematic assessment, indicating a high rate of diagnostic overlap between the two diseases. Previous studies have separately summarized the clinical features of pediatric PTB and pneumonia. However, few studies have performed a targeted comparative analysis focusing on children with initial suspected of PTB, namely the population with prominent diagnostic confusion.

Therefore, this study aimed to compare clinical manifestations, laboratory indicators and chest imaging findings between children with diagnosed PTB and children initially suspected of PTB but finally diagnosed with pneumonia. The extracted differential indicators may assist clinicians in early identification of the two diseases, reduce the rate of misdiagnosis, and optimize clinical management strategies for affected children.

2 Subjects and methods2.1 Study subjects

Children admitted to the Pediatric Department of Guangzhou Chest Hospital between March 2020 and May 2021 and initially suspected of PTB were enrolled in this study.

2.1.1 Inclusion and exclusion criteria

Etiologically confirmed PTB was defined as a positive result for Mycobacterium tuberculosis nucleic acid testing, culture, or acid-fast bacilli smear (any single positive item). Clinically diagnosed PTB was determined based on clinical and imaging manifestations, significant improvements in symptoms and imaging findings after effective anti-tuberculosis therapy, and the exclusion of other competing diseases. All diagnoses referred to the national health industry standard Diagnosis of Tuberculosis (WS 288-2017) (4).

Inclusion criteria for the PTB group were as follows: (1) Meeting the above PTB diagnostic criteria; (2) No anti-tuberculosis treatment before admission.

The control group consisted of children initially suspected of PTB but finally diagnosed with pneumonia. Pneumonia diagnosis complied with the Diagnostic Criteria for Community-Acquired Pneumonia in Children (2019 Edition) (5). Additional inclusion requirements for the control group included: (1) No empirical anti-tuberculosis treatment after admission; (2) Significant absorption of pulmonary lesions on follow-up chest computed tomography (CT) after antibacterial treatment, and no evidence of PTB at the 6-month follow-up; (3) Absence of concurrent active tuberculosis, confirmed by negative results for sputum or gastric fluid acid-fast bacilli smear, negative Mycobacterium tuberculosis molecular testing and rapid culture, a tuberculin purified protein derivative (PPD) induration less than 10 mm, and a negative interferon-gamma release assay (IGRA).

Exclusion criteria for both groups were: (1) Presence of genetic or metabolic diseases; (2) Primary or secondary immunodeficiency, malignant tumors, or other severe chronic underlying disorders; (3) Rheumatic immune diseases or administration of biological immunosuppressive agents within one month before admission; (4) For the PTB group, patients with other concomitant pathogenic infections requiring targeted anti-infective treatment during hospitalization were excluded; for the control group, patients with fungal infection were excluded.

2.2 Study methods2.2.1 Data collection

Comprehensive clinical characteristic data of children in both groups were collected, including the following specific contents:

General information: Gender, age, height, weight, place of residence (urban/rural), history of tuberculosis exposure, duration of disease before admission, use of antibiotics and hormones before admission.

Clinical manifestations: Fever, cough, expectoration, shortness of breath, hemoptysis/blood—streaked sputum, chest pain, extrapulmonary clinical manifestations (such as digestive tract symptoms, headache and dizziness, disturbance of consciousness), and results of lung auscultation

Results of routine laboratory tests: White blood cell count, hemoglobin level, high-sensitivity C-reactive protein (hs-CRP), procalcitonin (PCT), total protein, albumin, alanine aminotransferase (ALT), aspartate aminotransferase (AST), PPD test results, blood IGRA results, and etiological test results, including pathogenic bacterial culture of sputum or bronchoalveolar lavage fluid, sputum or gastric juice acid-fast bacilli smear, rapid M. tuberculosis culture, M. tuberculosis molecular biological assay, Fungal culture, and detection of eight common respiratory pathogens.

Imaging examination results: chest CT findings. Bronchoscopy findings were documented for children with PTB who were complicated by tracheobronchial tuberculosis(TBTB). Head magnetic resonance imaging(MRI) findings were recorded for children with PTB who were complicated by Central Nervous System Tuberculosis(CNS-TB).

2.2.2 Etiological detection

The following etiological examinations were performed in all children:Sputum (or bronchoalveolar lavage fluid) samples were collected for pathogenic bacterial and fungal culture;Sputum (or gastric juice) samples were collected for acid-fast bacilli smear, rapid M. tuberculosis culture (liquid method), and M. tuberculosis molecular biological detection;Venous blood was collected for detection of eight respiratory pathogens (including Legionella pneumophila, Mycoplasma pneumoniae, Rickettsia burnetii, Chlamydia pneumoniae, influenza A virus/ influenza B virus IgM, parainfluenza virus, respiratory syncytial virus, and adenovirus) and fungal serum antigen detection; During hospitalization, bronchoscopy was performed according to the children's condition, and bronchoalveolar lavage fluid was obtained for etiological detection (high-throughput sequencing was performed if necessary).

2.2.3 Imaging assessment protocol

All chest CT scans were performed using the same 64-slice spiral CT scanner (Siemens Somatom Definition AS) with a unified protocol: tube voltage 100–120 kV, tube current 50–100 mAs, slice thickness 1.0 mm, reconstruction interval 0.5 mm, lung window (window width 1,500–2,000 HU, window level −600 to −700 HU) and mediastinal window (window width 300–400 HU, window level 30–50 HU) reconstruction. Two radiologists with more than 10 years of experience in chest imaging independently read the CT images in a blinded manner (unaware of the clinical diagnosis and laboratory results). The diagnostic criteria for key CT findings were as follows: (1) Lymphadenopathy: lymph node short-axis diameter ≥10 mm; (2) Multiple nodules: ≥3 pulmonary nodules with diameter <10 mm; (3) Fine linear/reticular opacities: thin linear or reticular shadows distributed in the pulmonary interstitium; (4) Patchy shadows with blurred edges: flocculent or sheet-like high-density shadows with unclear boundaries in the pulmonary parenchyma. Discrepancies in reading results were resolved through joint discussion by the two radiologists and a third senior radiologist. The inter-rater reliability was evaluated using the Kappa coefficient, and the Kappa values of all key CT findings were >0.8, indicating good consistency.

2.2.4 Statistical analysis

All statistical analyses were conducted using R software (version 4.2.1). Continuous variables are presented as mean ± standard deviation or median with interquartile range according to their distribution. Categorical variables are expressed as frequencies and percentages. Between-group comparisons were performed using the independent-samples t test, Mann–Whitney U test, chi-square test, or Fisher exact test, as appropriate. A P value <0.05 was considered statistically significant. Descriptive data are reported to one decimal place, and P values less than 0.001 are presented as P < 0.001.

For multivariable predictive modeling, variables with P < 0.05 in univariate analysis were initially enrolled. To reduce overfitting and multicollinearity, variable selection was conducted via the least absolute shrinkage and selection operator (LASSO) regression. Two clinically meaningful and statistically stable predictors were finally screened and retained. Considering the limited and unbalanced sample size, multivariable regression analysis was performed using Firth penalized logistic regression. Bootstrap resampling with 1,000 iterations was used to obtain stable odds ratios (OR) and 95% confidence intervals (CI). Model diagnostic efficiency was assessed by the area under the receiver operating characteristic curve (AUC), sensitivity, specificity and overall accuracy.

3 Results3.1 Comparison of general information between the two groups

A total of 63 children were enrolled in this study. Of these, 48 were ultimately diagnosed with PTB (23 etiologically confirmed and 25 clinically diagnosed) and constituted the PTB group; the remaining 15, diagnosed with pneumonia, formed the control group.

There were no statistically significant differences in gender, age, height, weight, and place of residence between the two groups (all P > 0.05), indicating good comparability. However, the number of children with a history of tuberculosis exposure, disease duration before admission >14 days, and use of antibiotics and hormones before admission in the PTB group was significantly higher than that in the pneumonia control group, with statistically significant differences (P < 0.05 or P < 0.001). The specific data are shown in Table 1.

General InformationPTB group (n = 48)Control group (n = 15)POR (95%CI)Gender (Male/Female) (cases)24/248/70.820.88 (0.28, 2.73)Age (years) [M (P25, P75)]10.9 (8.0, 13.0)8.5 (6.5, 12.5)0.42-Height (dm) [M (P25, P75)]14.5 (11.5, 15.8)13.4 (11.2, 14.8)0.81-Weight (kg) ()31.4 ± 16.333.7 ± 15.90.65-Urban/Rural (cases)35/1310/50.641.35 (0.39, 4.69)History of tuberculosis exposure [cases (%)]26 (54.1)3 (20.0)0.024.76 (1.18,18.87)Disease duration before admission >14 d [cases (%)]38 (79.1)7 (46.6)0.024.34 (1.27,14.93)Use of antibiotics and hormones before admission [cases (%)]40 (83.3)6 (40.0)<0.0017.50 (2.08,26.93)

Baseline demographic and clinical history characteristics of pediatric patients with PTB vs. those with pneumonia initially suspected of PTB.

-, Not statistically analyzed.

3.2 Comparison of clinical manifestations between the two groups

With the exception of fever accompanied by chills (P < 0.001), there were no significant differences between the groups in the incidence of other major respiratory symptoms (all P > 0.05). In terms of lung auscultation results, a statistically significant difference was observed between the two groups (P < 0.001). Regarding extrapulmonary manifestations, statistically significant differences were observed between the two groups (P < 0.05), except for findings related to digestive tract symptoms. The detailed clinical manifestations and their frequencies for both study groups are summarized in Table 2.

Clinical ManifestationsPTB Group (n = 48)Control Group (n = 15)POR (95%CI)No respiratory symptoms6 (12.5)1 (6.6)0.671.98 (0.23,16.87)Respiratory symptoms----Fever40 (83.3)14 (93.3)0.590.36 (0.04, 2.85)Cough and expectoration42 (87.5)14 (93.3)0.880.60 (0.08,4.55)Shortness of breath17 (35.4)5 (33.3)0.881.10 (0.34,3.52)Hemoptysis/blood—streaked sputum9 (18.7)1 (6.6)0.483.16 (0.38,26.10)Chest pain5 (10.4)1 (6.6)0.671.62 (0.19,13.85)fever accompanied by chills0 (0.0)11 (73.3)<0.0010.02 (0.00, 0.15)Fatigue, weight loss, night sweats5 (10.4)1 (6.6)0.671.62 (0.19,13.85)Lung auscultation----Abnormal10 (20.8)14 (93.3)<0.0010.02 (0.00,0.13)Extrapulmonary manifestations----Digestive tract symptoms8 (16.6)3 (20.0)0.330.80 (0.20,3.22)Headache and dizziness10 (20.8)8 (53.3)0.020.23 (0.07,0.77)Disturbance of consciousness5 (10.42)0 (0.00)<0.001-

Comparison of clinical manifestations, respiratory symptoms, lung auscultation findings, and extrapulmonary features between the PTB group and control group [cases (%)].

-, Not statistically analyzed.

3.3 Comparison of routine laboratory test results between the two groups

The PTB group had significantly lower proportions than the control group for: elevated ALT/AST (6.2% vs. 26.6%; P < 0.05) and PCT >0.5 ng/ml (0.0% vs. 86.6%; P < 0.001).

A comparison of the detailed laboratory findings between the two groups is presented in Table 3.

Routine laboratory itemsPTB group (n = 48)Control group (n = 15)POR (95%CI)White blood cells (×109 /L) [M (P25, P75)]8.3 (6.1, 12.1)11.3 (8.1, 13.3)0.25-Hemoglobin (g/L) ()112.1 ± 21.0123.2 ± 15.80.07-hs-CRP (mg/L) [M (P25, P75)]15.3 (3.7, 24.8)22.2 (2.0, 48.4)0.31-Elevated ALT or AST at admission [cases (%)]3 (6.2)4 (26.6)0.030.18 (0.04, 0.82)Albumin (g/L) ()36.1 ± 6.939.7 ± 5.20.08-Total protein (g/L) ()70.9 ± 8.670.8 ± 4.90.98-PCT >0.5 ng/ml [cases (%)]0 (0.0)13 (86.6)<0.001-Eight types of respiratory pathogens [cases (%)]----Legionella pneumophila0 (0.0)0 (0.0)--Mycoplasma pneumoniae18 (37.5)5 (33.3)0.771.20 (0.35,4.07)Chlamydia pneumoniae5 (10.4)1 (6.6)0.671.63 (0.19,14.12)Rickettsia burnetii0 (0.0)0 (0.0)--Influenza----Influenza A virus IgM3 (6.2)2 (13.3)0.380.43 (0.07,2.63)Influenza B virus IgM4 (8.3)1 (6.6)0.831.27 (0.14,11.51)Parainfluenza virus15 (31.2)7 (46.6)0.270.52 (0.16,1.70)Respiratory syncytial virus0 (0.0)0 (0.0)--Adenovirus0 (0.0)0 (0.0)--M. tuberculosis etiological examination [cases (%)]----Acid—fast bacilli smear6 (12.5)1 (contaminated)--M. tuberculosis culture3 (6.2)0 (0.0)--M. tuberculosis molecular microbiology20 (41.6)1 (contaminated)--Fungal culture [cases (%)]0 (0.0)0 (0.0)--Positive pathogenic bacteria [cases (%)]0 (0.0)7 (46.6)--Positive blood IGRA46 (95.8)0 (0.0)--PPD >10 mm28 (58.3)0 (0.0)--

Comparison of routine laboratory parameters, inflammatory markers, and etiological test results between children diagnosed with PTB and the control group.

hs-CRP, High-sensitivity C-reactive protein; PCT, Procalcitonin; ALT, Alanine aminotransferase; AST, Aspartate aminotransferase; Blood IGRA, Interferon-gamma release assay; PPD, Purified protein derivative of tuberculin; -, Not statistically analyzed.

In the control group, 7 children had positive pathogenic bacteria detection (3 cases with positive pathogenic bacteria culture and 5 cases with positive high-throughput sequencing), and the pathogenic bacteria involved were 2 cases of Streptococcus pneumoniae, 2 cases of Pseudomonas aeruginosa, 1 case of Acinetobacter baumannii, 1 case of Klebsiella pneumoniae, 1 case of Serratia marcescens, and 1 case of Moraxella catarrhalis, In the control group, one case exhibited a positive sputum smear for acid-fast bacilli, and another case had a positive M. tuberculosis molecular microbiology result. Based on clinical context, these specimens were comprehensively judged to be contaminated.

3.4 Comparison of imaging examination results between the two groups

No statistically significant differences were observed for cavity and pleural effusion on chest CT; however, all other features differed significantly between the groups (P < 0.05). The specific results of the chest CT comparison are shown in Table 4.

Imaging examinationsPTB group (n = 48)Control group (n = 15)POR(95%CI)Cavity9 (18.7)1 (6.6)0.483.16 (0.38, 26.10)Pulmonary consolidation17 (35.4)9 (60.0)0.050.36 (0.12, 1.08)Fine linear/reticular opacities30 (62.5)2 (13.3)<0.00110.50 (2.20,50.06)Patchy shadows with blurred edges20 (41.6)11 (73.3)0.030.28 (0.08, 0.94)Multiple pulmonary nodules47 (97.9)3 (20.0)<0.001235.00 (28.60, 1930.00)Lymphadenopathy (hilar/mediastinal/axillary)46 (95.8)4 (26.6)<0.00169.00 (17.10,280.00)Pleural effusion6 (12.5)1 (6.6)0.881.98 (0.23, 16.87)Calcification35 (72.9)1 (6.6)<0.00135.00 (4.40, 278.00)

Comparison of chest CT imaging features between the PTB group and the control group [cases (%)].

Within the PTB group, 20 patients were identified as having concurrent TBTB, and 7 patients were diagnosed with concurrent CNS-TB. The specific subtypes and their respective frequencies among these complications are detailed in Table 5.

ItemsCases (n %)Bronchoscopic findings of TBTB-Inflammatory infiltration type1 (2.0)Ulcerative necrosis type4 (8.3)Granulation proliferation type4 (8.3)Lymph node fistula type10 (20.8)Scar stricture type1 (2.0)Cranial MRI findings of CNS-TB-Meningeal type5 (10.4)Cerebral parenchymal type1 (2.0)Mixed type2 (4.1)

Subgroup analysis of children with PTB complicated by TBTB or CNS-TB: findings from bronchoscopy and cranial MRI [cases (n %)].

-, Not statistically analyzed; TBTB, Tracheobronchial tuberculosis; CNS-TB, Central nervous system tuberculosis.

3.5 Robust predictive model

Univariate analysis revealed that 14 independent variables were statistically significant. Following screening based on clinical significance, seven independent variables were included: Fine linear/reticular opacities, Patchy shadows with blurred edges, Multiple nodules, Lymphadenopathy(hilar/mediastinal/axillary), Calcification, History of tuberculosis exposure, Fever with chills.

Based on Lasso regression, two variables were included in the final model: multiple pulmonary nodules, lymphadenopathy(hilar/mediastinal/axillary).The results indicated that all two variables were independent influencing factors for distinguishing pediatric PTB from pneumonia initially suspected of PTB (all P < 0.05).The specific data are shown in Table 6.

VariablesCoefficient (β)Standard error (SE)OR (Bootstrap 95%CI)PIntercept−4.610.820.00 (0.00,0.05)P < 0.001Multiple pulmonary nodules5.620.83276.18 (55.39,1430)P < 0.001Lymphadenopathy (hilar/mediastinal/axillary)4.821.13124.21 (10.58,867.35)P < 0.001

Multivariable firth penalized logistic regression model for the differentiation between pediatric PTB and pneumonia initially suspected of PTB.

The combined model achieved an AUC of 0.99 (95% CI: 0.97,1.00), with 100% specificity and 93.75% sensitivity, The data are shown in Table 7.

Diagnostic modelAUC (95% CI)Optimal cut-offSensitivity (%)Specificity (%)Accuracy (%)Combined Model0.99 (0.97,1.00)0.4093.75100.0095.24Multiple pulmonary nodules0.89 (0.81,0.97)0.7897.9280.0090.48Lymphadenopathy (hilar/mediastinal/axillary)0.85 (0.75,0.95)0.6995.8373.3387.30

Diagnostic performance of the combined predictive model and individual imaging indicators for differentiating pediatric PTB from pneumonia initially suspected of PTB.

AUC, Area under the curve; The combined model includes both “Multiple pulmonary nodules” and “Lymphadenopathy(hilar/mediastinal/axillary)” as predictors; Optimal cut-off was determined by maximizing Youden's index (J = Sensitivity + Specificity−1).

4 Discussion

PTB continues to pose a severe global public health threat, with a persistent and challenging epidemiological situation. According to the Global Tuberculosis Report 2025 released by the WHO, an estimated 10.7 million individuals developed tuberculosis worldwide in 2024. Children accounted for 12% of these cases, equivalent to approximately 1.2 million new pediatric tuberculosis patients. Although tuberculosis is preven