Chinese Expert Consensus on Diagnosis and Treatment of Pulmonary Nodules (2024 Edition)

I. Definition and Classification of Pulmonary Nodules

Management principles

(1) Definition of pulmonary nodule

Imaging shows a focal, roughly round, solid or subsolid opacity with density higher than lung parenchyma and a maximum diameter ≤3 cm; it may be solitary or multiple, without atelectasis, hilar lymphadenopathy, or pleural effusion. A solitary pulmonary nodule usually has no obvious symptoms, appearing as a well-defined, dense soft-tissue shadow ≤3 cm in maximum diameter surrounded by aerated lung. Multiple pulmonary nodules usually present as one nodule accompanied by one or more additional nodules; it is generally believed that diffuse pulmonary nodules of >10 are mostly due to malignant tumor metastasis or benign lesions (inflammatory diseases caused by infectious or non-infectious factors).

(2) Classification

1. By number: a single lesion is defined as solitary; two or more lesions are defined as multiple.
2. By size: to better guide tiered diagnosis and treatment and precisely manage patients, nodules ≤5 mm in maximum diameter are defined as micronodules, and those 5-10 mm as small nodules. Nodule size is clearly related to malignancy probability. Micronodules may be followed up and managed at primary hospitals; small nodules may be managed at experienced hospitals, such as the branch centers of the China Lung Cancer Prevention and Treatment Alliance; nodules 10-30 mm should be diagnosed and treated as early as possible. Follow-up management follows the clinical workflow for solid and subsolid pulmonary nodules.
3. By density: nodules are divided into solid and subsolid; the latter includes pure ground-glass nodules and part-solid nodules: (1) Solid pulmonary nodule: a round or near-round increased-density shadow in the lung, dense enough to obscure the running vessels and bronchi within. (2) Subsolid pulmonary nodule: all pulmonary nodules containing ground-glass density are called subsolid; a ground-glass lesion refers to a CT shadow of increased density with or without clear borders, but not dense enough to obscure the vessels and bronchi running within. Subsolid nodules include pure ground-glass nodules (pGGN), mixed ground-glass nodules with both ground-glass and solid density (mGGN), the latter also called part-solid nodules. If the ground-glass lesion contains no solid component it is pGGN; if it contains a solid component it is mGGN.
4. By difficulty of early diagnosis: a "difficult-to-characterize pulmonary nodule" is one that cannot be definitively diagnosed by non-surgical biopsy and is highly suspected of being early lung cancer. Early lung cancer hidden in a nodule is too small for definite preoperative pathological diagnosis; repeated follow-up may delay treatment, while limited differential diagnosis may cause overtreatment. To solve these problems, the definition of "difficult-to-characterize pulmonary nodule" is proposed, and a multi-disciplinary team (MDT) working model and shared doctor-patient decision-making are recommended.

【Recommendation 1】 A "difficult-to-characterize pulmonary nodule" cannot be definitively diagnosed by non-surgical biopsy and is highly suspected of being early lung cancer. For such nodules, the MDT working model and shared doctor-patient decision-making are recommended (Class III recommendation).

II. Routes of Detection of Pulmonary Nodules

(1) Definition of high-risk populations for lung cancer and routine imaging screening

The 2011 National Lung Screening Trial (NLST) showed that, compared with chest X-ray, screening high-risk populations with low-dose chest CT (LDCT) reduced lung cancer mortality by 20%. Since then, China has also recommended annual LDCT screening for high-risk groups. In 2013 the U.S. Preventive Services Task Force (USPSTF) recommended annual LDCT lung cancer screening for people aged 55-80 with ≥30 pack-years of smoking who currently smoke or have quit for less than 15 years. Recent CISNET simulation modeling on cost-effectiveness showed that identifying the optimal screening population for LDCT can reduce lung cancer deaths and prolong survival, and can reduce disparities in screening criteria by sex and race/ethnicity. Screening people aged 50 or 55-80 with ≥20 pack-years brings more benefit than the 2013 USPSTF criteria. Thus in 2021 USPSTF recommended annual LDCT for people aged 50-80 with ≥20 pack-years who currently smoke or have quit for less than 15 years, lowering the minimum screening age from 55 to 50 and the smoking index from 30 to 20 pack-years.

(2) Detection due to symptoms

Most pulmonary nodules are asymptomatic; symptoms and signs appear only when a malignant nodule invades surrounding and other tissues and organs—such as progressive cough, hemoptysis, chest pain, hoarseness, and dyspnea—depending on disease course and location.

(3) Incidental detection

Found incidentally when chest CT is done for other diseases, such as tuberculosis, viral infection, or other respiratory diseases, or cardiac examination. If a pulmonary nodule is found on chest CT for some infectious disease, prior and pre-/post-treatment chest CT images must be referenced and dynamic follow-up performed to confirm whether the nodule is infection-related.

【Recommendation 2】 Define the Chinese high-risk lung cancer population as people aged ≥40 (Class II recommendation) with any of the following risk factors: (1) smoking index ≥400 year-smokers (or 20 pack-years) (Class IA); (2) history of environmental or high-risk occupational exposure (e.g., asbestos, beryllium, uranium, radon) (Class IB); (3) concurrent COPD, diffuse pulmonary fibrosis, or prior tuberculosis (Class IB); (4) prior malignancy or family history of lung cancer, especially in first-degree relatives (Class IB). Annual chest LDCT screening is recommended for high-risk groups (Class IA recommendation).

III. Routine Examination and Evaluation of Pulmonary Nodules

(1) Imaging examination

Compared with chest X-ray, chest CT provides more information—location, size, shape, density, margins, and internal features of the nodule. LDCT screening is recommended for high-risk groups; LDCT parameters refer to the Chinese Low-Dose CT Lung Cancer Screening Guideline (2023 Edition). For screen-detected or incidental nodules, thin-section CT or thin-section high-resolution CT should be performed at the lesion to better show nodule features. If prior chest CT exists, comparison with historical imaging is recommended.

(2) Imaging evaluation

Benign vs. malignant may be judged by two angles: external appearance assessment ("judging by looks") and internal exploration ("looking at substance"), including nodule size, shape, margin, tumor-lung interface, internal structure, and dynamic changes on follow-up.

1. External assessment: (1) Size: as nodule volume increases, malignancy probability rises. (2) Shape: most benign nodules are round or near-round; compared with malignant solid nodules, malignant subsolid nodules more often have irregular shapes. (3) Margin: malignant nodules are often lobulated or spiculated; pleural indentation and vascular convergence signs often suggest malignancy. Benign nodules mostly lack lobulation; margins may have sharp angles or fibrous strands, and surrounding fibrous strands or pleural thickening suggest benignity. (4) Nodule-lung interface: malignant nodules often have clear but irregular, rough or spiculated interfaces; inflammatory nodules often have blurred margins, while benign non-inflammatory nodules often have clear, smooth, even neat margins.
2. Internal assessment: (1) Density: uniformly dense pGGN, especially <5 mm, often suggests atypical adenomatous hyperplasia (AAH). Heterogeneous mGGN with solid component >50% often suggests high malignancy probability, mostly minimally invasive adenocarcinoma (MIA) or invasive adenocarcinoma (IA), though MIA or IA can also present as pGGN. Persistent GGNs are mostly malignant or have malignant potential. Mean CT value of GGN is important for differential diagnosis: higher density means higher malignancy probability, lower density means lower probability, but pGGN CT value may not correlate with pathological invasiveness; nodule size and morphological change must be combined. (2) Structure: bronchial cutoff with local wall thickening, or irregular cutoff lumen, suggests malignancy. For more accurate assessment of intranodular and perinodular vascular relations, contrast-enhanced CT with ≤1 mm slice thickness may be post-processed and reconstructed; the nodule-vessel sign helps characterize the nodule.
3. Dynamic follow-up: nodules showing the following changes on follow-up are mostly benign: (1) marked short-term change in external features, loss of lobulation or development of deep lobulation, margins becoming smooth or blurred; (2) uniform or lowering density; (3) shrinkage or disappearance without density increase; (4) rapid growth with doubling time <15 days; (5) solid nodule stable for >2 years—but this does not apply to GGN, because adenocarcinoma in situ (AIS) and MIA-stage GGN can remain stable long-term. Thus "long-term" here means more than 2 years; how long stability implies benignity still needs deeper research.

Nodules showing the following changes on follow-up are mostly malignant: (1) increasing diameter with doubling time consistent with tumor growth (malignant doubling varies widely: solid nodules ~20-400 days; subsolid 400-800 days or longer); (2) stable or enlarging lesion with new solid component; (3) shrinking lesion with new or increasing solid component; (4) angiogenesis consistent with malignant pattern; (5) appearance of lobulation, spiculation, and/or pleural indentation.

(3) AI imaging-assisted diagnostic evaluation

Both China's NMPA and the U.S. FDA have approved some AI imaging-assisted diagnosis systems to improve efficiency and performance; clinical research has focused on detection or diagnostic support of nodules manually selected by imaging experts. Computer-aided diagnosis (CAD) based on chest imaging helps improve physician sensitivity in detecting pulmonary nodules and accuracy in distinguishing benign from malignant. Compared with conventional imaging, AI-assisted nodule evaluation and management has the following advantages: (1) precise measurement of longest diameter, volume, and density; (2) more comprehensive assessment of margins and invasion; (3) precise assessment of intranodular vessels and their growth. In addition, AI has unique advantages: (1) 3D reconstruction, including vascular 3D reconstruction, to reveal 2D vs. 3D differences; (2) dynamic comparison automatically and accurately pairing the same lesion in the same patient at different times and sequences, including 3D density and volume changes, to calculate volume doubling time; (3) deep learning to reveal more benign/malignant differences; (4) deep mining of internal nodule structure, with accumulation of big data revealing more characteristic differences.

To further improve AI efficiency, Chinese experts have proposed human-machine MDT (expert-robot MDT), i.e., MDT between humans and computers, in which human experts and an AI nodule evaluation system interact to produce individualized diagnoses. Human-machine MDT both exploits AI's unmatched advantages (precise 3D longest diameter, volume, and density) and avoids AI's false positives and false negatives. AI is not a biological human and can only be an auxiliary tool, not directly responsible for clinical diagnosis and treatment. This makes it even more necessary for responsible human experts to integrate AI with clinical experience to formulate plans, so as to minimize misdiagnosis and mistreatment and maximize patient benefit. To achieve this, we rely on AI technology, expert experience, and human-machine MDT dialogue. Discussion on how to upgrade the current human-machine separated diagnostic model into an interactive one will help "simplify complex problems, digitize simple problems, program digital problems, and systematize programmed problems," improving doctors' ability to solve difficult nodule problems and ultimately making evidence-based, individualized decisions.

(4) Tumor markers

Although no universally accepted, highly sensitive and specific biomarker for early lung cancer diagnosis exists, the following may be performed where feasible to provide reference: (1) Pro-gastrin-releasing peptide (Pro-GRP): the preferred marker for auxiliary diagnosis, efficacy evaluation, and recurrence monitoring of small cell lung cancer (SCLC); (2) Neuron-specific enolase (NSE): for SCLC auxiliary diagnosis, efficacy evaluation, recurrence monitoring, and prognosis; (3) Carcinoembryonic antigen (CEA): mainly for lung adenocarcinoma auxiliary diagnosis, efficacy evaluation, recurrence monitoring, and prognosis; (4) Cytokeratin fragment 21-1 (CYFRA21-1): mainly for lung squamous carcinoma; (5) Squamous cell carcinoma antigen (SCC-Ag): mainly for lung squamous carcinoma. Combined use improves the positive rate and accuracy of lung cancer screening. The lung cancer biomarker panel (LCBP) prediction model led by Prof. Bai Chunxue of Zhongshan Hospital, Fudan University, uses combined tumor markers (Pro-GRP, CEA, SCC, CYFRA21-1) together with age, sex, smoking history, nodule diameter, and spiculation to stratify nodule risk; its sensitivity is 94.6% and specificity 94.2%. The LCBP model outperforms the U.S. Mayo ACCP model in predicting malignant risk, suggesting it is more suitable for Chinese high-risk populations, though prospective real-world studies are still needed to validate follow-up strategies.

【Recommendation 3】 For screen-detected or incidental nodules, thin-section CT or thin-section high-resolution CT of the lesion is recommended to better show features (Class IB), and comparison with historical imaging (Class IB).

【Recommendation 4】 AI-assisted imaging helps malignant risk assessment and clinical decision-making. A single tumor marker is not recommended for subcentimeter small/micronodule screening. Combined markers (ProGRP, SCC, CEA, CYFRA21-1) that keep rising or rise together on follow-up, combined with clinical information (nodule diameter, spiculation) into the LCBP model, outperform the ACCP model in predicting malignant risk (Class II recommendation).

IV. Individualized Assessment of Pulmonary Nodules

(1) Assessment of clinical malignancy probability

Collecting diagnostic and differential information—age, occupation, smoking history, chronic lung disease, personal and family tumor history, treatment course and outcome—provides reference. Although benign and malignant nodules cannot be perfectly distinguished, assessing malignancy probability from clinical and imaging features (Table 1) remains important for choosing follow-up methods.

The ACCP guideline uses the Mayo Clinic prediction model, the most widely used. It analyzed 419 patients with non-calcified nodules 4-30 mm in diameter by multivariate logistic regression and identified six independent predictors: age (OR 1.04/year), current or past smoking (OR 2.2), history of extrathoracic malignancy >5 years before nodule discovery (OR 3.8), nodule diameter (OR 1.14/mm), spiculation (OR 2.8), and upper-lobe location (OR 2.2). Prediction model: malignancy probability = e^x/(1+e^x) (Eq.1); X = -6.8272 + (0.0391×age) + (0.7917×smoking) + (1.3388×malignancy) + (0.1274×diameter) + (1.0407×spiculation) + (0.7838×location) (Eq.2), where e is the natural log; age in years; smoking=1 if currently or previously smoked (else 0); malignancy=1 if history of extrathoracic cancer >5 years (else 0); diameter in mm; spiculation=1 if nodule margin is spiculated (else 0); location=1 if in upper lobe (else 0). Note that model predictions still differ from clinician judgment; choose and build models based on target population, ease of use, and validation. Moreover, the ACCP tenet that "upper-lobe nodules are more malignant" does not fully apply to China and most Asia-Pacific regions, because the apical-posterior segment of the upper lobe is also a predilection site for tuberculosis.

(2) Functional imaging

Many high-risk nodules seen on routine thin-section CT need functional imaging for individualized assessment, mainly contrast-enhanced chest CT and PET-CT.

1. Contrast-enhanced chest CT: a multicenter study of 356 patients with 5-40 mm lung lesions found malignancy rate 48%; using an enhancement threshold >15 Hounsfield units (HU), sensitivity, specificity, and accuracy for malignant lesions were 98%, 58%, and 77%.
2. PET-CT: For solid nodules >8 mm that cannot be characterized, PET-CT may be considered. It has no clear advantage for pGGN or nodules with solid component ≤8 mm. For nodules with solid component >8 mm, PET-CT helps distinguish benign from malignant, based on the high glucose uptake and metabolism of tumor cells: after injecting 18F-FDG, uptake by the nodule is measured; malignant nodules take up more. The standardized uptake value (SUV) is a key PET-CT parameter reflecting tracer uptake. A meta-analysis found PET-CT sensitivity and specificity for pulmonary nodules of 88% and 78%. PET-CT also helps guide biopsy site.

One study compared contrast CT and PET-CT in 380 patients with 8-30 mm nodules. Among 312 matched patients, 191 (61%) had lung cancer; contrast CT sensitivity/specificity were 95.3%/29.8%, while PET-CT were 79.1%/81.8%; ROC AUC was 0.62 vs. 0.80.

(3) Circulating abnormal cells

Circulating genetically abnormal cells (CAC) are cells in peripheral blood carrying tumor-specific chromosomal sites (amplifications and deletions), similar to the genetic abnormalities of the primary tumor. CACs are found in both early (stage I) and advanced (stage IV) NSCLC blood, and their count correlates with recurrence and survival. Studies found CAC sensitivity/specificity for ≤10 mm nodules was 70.5%/86.4%, and for stage I NSCLC 67.2%/80.8%. In China's first national multicenter prospective cohort on "liquid biopsy CAC + LDCT AI analysis," FISH detected CAC abnormalities at 3p22.1/3q29 and 10q22.3/CEP10, while a deep-learning CNN AI platform analyzed LDCT (DICOM) images; the two tools were complementary in early lung cancer diagnosis. A multimodal early-diagnosis model built on "clinical features, imaging features, AI analysis, and CAC" achieved sensitivity 89.53%, specificity 81.31%, AUC 0.880. In an independent validation set, sensitivity/specificity were 82.86%/80.95%, AUC 0.895, clearly outperforming the Mayo model (AUC 0.772) and U.S. VA model (AUC 0.740). Therefore, in Chinese populations we prefer models derived from Chinese patient data.

(4) Non-surgical biopsy

1. Bronchoscopy: including direct brush cytology, biopsy, or transbronchial lung biopsy (TBLB) under fluoroscopy for cytology and histology. Autofluorescence bronchoscopy (AFB) is a newer method for early central lung cancer, exploiting differences in autofluorescence to improve detection of premalignant dysplasia or early carcinoma in situ. Endobronchial ultrasound-guided TBLB (EBUS-TBLB) uses a peripheral ultrasound probe to view peripheral lesions and biopsy under ultrasound guidance, improving peripheral-nodule yield. A randomized controlled study showed EBUS-TBLB sensitivity for ≤20 mm malignant peripheral lesions was 71% vs. 23% for conventional TBLB. Virtual bronchoscopic navigation (VBN) reconstructs 3D images from thin-section high-resolution CT and plans paths; the doctor chooses the optimal path and VBN animates the airway route for fully visual guidance. Ultra-thin bronchoscopes can reach the 5th-8th generation bronchi. Electromagnetic navigation bronchoscopy (ENB) combines physics, informatics, radiology, and bronchoscopy to detect peripheral lung lesions inaccessible to conventional bronchoscopy. Combining EBUS with VBN or ENB improves peripheral lesion diagnosis with high safety and has promise in nodule differential diagnosis and early lung cancer. A meta-analysis showed overall diagnostic yield of EBUS/ENB/VBN for peripheral pulmonary lesions (PPL) was 70.5%, higher for >20 mm nodules with bronchus sign. A recent Chinese single-center study found EBUS+ENB diagnostic yield of 82.5%.
2. Transthoracic needle biopsy (TTNB): performed under CT or ultrasound guidance, with high sensitivity and specificity for peripheral lung cancer. Lesions near the chest wall may be biopsied under ultrasound; lesions not abutting the chest wall may undergo CT-guided percutaneous lung biopsy.

Researchers are studying whether bronchoscopy robots can biopsy nodules; studies show they can help obtain tissue from nodules around 10 mm. Whether they achieve "strong primary-care, broad coverage" awaits health-economic studies and comparison with advanced techniques such as AI-empowered early lung cancer diagnosis.

(5) Surgical biopsy

Preoperative evaluation by senior experts and/or MDT is recommended to avoid overtreatment. Consider thoracoscopy only when senior experts and/or MDT assess high malignant risk: it is suitable for nodules that cannot be sampled by bronchoscopy or TTNB, especially small peripheral nodules where thoracoscopic lesion excision both diagnoses and treats. Surgery for diagnosis alone of a solitary nodule is not recommended.