The research team for this study includes experts from institutions such as Guangzhou University of Chinese Medicine and Nanjing University of Chinese Medicine, with defined roles such as principal investigator and study coordinators responsible for data collection, patient management, and analysis. The trial sponsor, Guangzhou University of Chinese Medicine First Affiliated Hospital, oversees the design, management, data analysis, and interpretation of the trial, including the publication of results. The coordinating center for the trial, also located at Guangzhou University of Chinese Medicine First Affiliated Hospital, ensures the trial’s overall operation, managing data collection, patient follow-up, and protocol compliance in collaboration with various teams. Contact information for the principal investigators and coordinators is provided to facilitate clear communication. This study protocol was developed in accordance with the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) 2013 Statement. The completed SPIRIT checklist is provided in Checklist 1.

The clinical trial protocol (version 3.0, March 20, 2023) was approved by the Research Ethics Committee of the First Affiliated Hospital of Guangzhou University of Chinese Medicine (approval number: K-2022‐131-XZ-01). The trial is registered in the Chinese Clinical Trial Registry (ChiCTR2300069570) and will be conducted in accordance with the Declaration of Helsinki and Good Clinical Practice principles.

Written informed consent will be obtained before any trial-specific procedures. Investigators will explain the study’s purpose, interventions, expected benefits, foreseeable risks, alternative management options, confidentiality protections, voluntary participation, and the right to withdraw without affecting usual medical care. Additional consent will be obtained for the storage and exploratory analysis of plasma samples.

Participant data will be entered into an electronic data capture (EDC) system using coded study identifiers. Direct identifiers will be stored separately from research data, access will be restricted to authorized study personnel, and analyses will use deidentified datasets. Biological samples will be labeled with coded identifiers and used only for approved analyses described in the protocol or subsequent ethics-approved amendments.

Participants will not be denied clinically indicated diagnostic evaluations or standard treatments. Study medication and protocol-specified assessments will be provided by the study. No direct financial incentives are planned. Medical management and compensation for trial-related injuries will be provided according to institutional and national regulations.

Chest CT will be performed using a thin-slice protocol harmonized across participating centers as far as feasible. CT acquisition parameters, reconstruction kernel, slice thickness, and scanner model will be recorded. Radiologists, blinded to treatment allocation, will assess nodule type, longest diameter, solid-component size, morphology, and progression status.

The AI-assisted nodule analysis system, an independently developed deep learning–based automatic lung nodule segmentation technology (patent application number: 202510952658.0), will segment the target nodule and extract quantitative features, including nodule volume, longest diameter, solid-component size, density, and malignancy risk index. The AI output will not replace clinical judgment or standard reporting. Discrepancies between radiologist assessment and AI output will be reviewed by a blinded adjudication committee or a senior radiologist according to a prespecified adjudication procedure.

Baseline data will include demographics, smoking history, comorbidities, prior cancer history, medication history, physical examination, vital signs, blood routine, liver function, kidney function, urinalysis, stool examination, tumor markers, ECG, chest CT, AI-derived nodule variables, TCM symptom score, HADS, and PSQI. Follow-up visits will occur at 3 and 6 months during the intervention period and every 6 months thereafter until nodule progression, surgery, withdrawal, death, loss to follow-up, or study completion. The trial protocol is version 3.0 (March 20, 2023), and the schedule of study procedures is shown in Table 2.

Plasma samples will be collected at baseline and after 6 months of medication. A 10-mL peripheral blood sample will be collected in ethylenediaminetetraacetic acid tubes, centrifuged at 1500g for 10 minutes, aliquoted, and stored at −80 °C until analysis. The phrase “phlegm-heat-toxin pathogenesis” refers to a TCM theoretical pattern characterized clinically by phlegm accumulation, heat-related symptoms, toxin-like pathogenic stagnation, and impaired qi regulation. In this trial, the concept will be operationalized using TCM symptom scores and exploratory plasma markers rather than treated as a directly observable biomedical entity. The exploratory biological analysis will evaluate whether YFSJ changes biomarker patterns associated with inflammatory activation, immune dysregulation, oxidative stress, ER stress, and bile-acid metabolism. Candidate assays include multiplex or enzyme-linked immunosorbent assay measurement of interleukin-6 (IL-6), IL-1-β, tumor necrosis factor-α, IL-8, interferon-γ, C-reactive protein, vascular endothelial growth factor, and other prespecified cytokines; targeted liquid chromatography-mass spectrometry or mass spectrometry analysis of bile-acid metabolites such as deoxycholic acid, taurochenodeoxycholic acid, and glycocholic acid; and validated assays for ER stress–related markers such as glucose-regulated protein 78 or binding immunoglobulin protein, CCAAT/enhancer-binding homologous protein, activating transcription factor 6, spliced X-box-binding protein 1, and related pathway proteins when measurable in plasma or extracellular vesicles. Untargeted metabolomics or proteomics may be used to generate hypotheses, with false discovery rate control for multiple testing.

Participants will receive phone reminders before scheduled visits. Participants who miss a visit will be contacted by phone, text message, or home visit, as appropriate, to determine the reason and encourage continued participation. Medication adherence will be assessed through returned packet counts and participant report. Available imaging, laboratory, and questionnaire data will be collected at withdrawal whenever consent permits.

Data will be recorded in standardized CRFs and entered into an EDC system. Research staff will receive uniform training in eligibility confirmation, informed consent, CT data collection, AI output recording, questionnaire administration, adverse event reporting, and data entry. Data queries will be resolved before the database lock. The sponsor and centralized clinical management team will perform regular internal monitoring to ensure participant safety, protocol adherence, and data quality.

Because the intervention is noninvasive and the risk is considered low, an independent data monitoring committee is not planned. Serious adverse events will be reported to the ethics committee and sponsor according to the protocol and regulatory timelines. Emergency unblinding will be allowed only when knowledge of the treatment assignment is essential for clinical management.

Data will be collected at baseline and during follow-up visits using standardized CRFs. Key data include demographics, clinical history, imaging findings (chest CT and AI-assessed nodule features), laboratory results (tumor markers, blood, urine, stool, liver and kidney function, and ECG tests), and patient-reported outcomes (TCM symptom scale, provided in Multimedia Appendix 1, HADS, and PSQI). To ensure data quality, assessors will receive uniform training, and duplicate measurements will be conducted where appropriate. All instruments used have been validated in prior clinical practice. Data collection forms are included in the protocol and managed through an EDC system.

Based on the actual number of participants enrolled in the 2 groups, dropout and exclusion cases, as well as demographic and other baseline characteristics, corresponding efficacy and safety analyses will be conducted. Descriptive statistical analysis will be used, with qualitative indicators described as frequency tables, percentages, or proportions; quantitative indicators will be described as mean (SD), median (IQR), and minimum and maximum values.

For primary end points and clinical efficacy analysis, the main and sensitivity analyses will be performed according to the full analysis set and per-protocol set, respectively. Safety analysis will be conducted using the safety set. The main analysis of the primary end point will use the log-rank test to compare the 2-year progression rates between groups, with Cox regression adjusting for covariates (eg, center) as a sensitivity analysis. Subgroup analyses will be conducted based on the center and whether >8 mm nodules were surgically removed. The malignancy risk index, quality-of-life scores at different time points, and other multiple measurement indicators will be analyzed using mixed-effects models. Eligibility-related protocol deviations, including any deviations from the retained immune disease exclusion criterion, will be documented. Participants with major eligibility deviations will be handled according to the prespecified analysis-set definitions, with sensitivity analysis based on the per-protocol set. Safety end points will summarize adverse reactions and events occurring during the study and compare rates between groups. Hypothesis testing will uniformly use 2-sided tests, providing test statistics and corresponding P values, with P<.05 considered statistically significant.

This trial is designed to test the hypothesis that YFSJ, when added to guideline-based surveillance, can reduce the 2-year progression rate of high-risk pulmonary nodules compared with placebo combined with the same surveillance strategy. If the intervention reduces progression without increasing adverse events, YFSJ may provide a noninvasive adjunctive option for patients with small, high-risk, and mixed ground-glass nodules who do not yet meet the criteria for invasive treatment.

The primary end point is objective radiologic progression. Patient-reported psychological and sleep outcomes are included because prolonged surveillance may create distress, and symptom improvement is relevant to integrative care; however, these outcomes will be interpreted as secondary and supportive.

Existing pulmonary nodule guidance focuses on risk stratification, CT follow-up, and surgical or diagnostic intervention when features suggest malignancy [4-6]. Recent AI-based studies further show that integrating quantitative imaging, clinical features, and longitudinal data can refine malignancy risk classification [8,9]. However, these approaches mainly improve detection and decision-making; they do not provide pharmacological prevention of nodule progression.

YFSJ has been studied in non–small cell lung cancer models and adjunctive treatment settings, with evidence suggesting regulation of apoptosis, autophagy, and inflammatory pathways [14-18]. While those studies focused on established lung cancer, this trial extends this prior work to a prevention-oriented population with high-risk pulmonary nodules, using a randomized, double-blind, and placebo-controlled design to reduce bias.

Key strengths of this study include the multicenter, randomized, double-blind design; matched placebo control; objective 2-year radiologic primary end point; standard-of-care surveillance in both groups; AI-assisted quantitative imaging; patient-reported outcomes; and planned exploratory plasma analyses. The trial also explicitly distinguishes clinical efficacy end points from mechanistic hypotheses, which improves interpretability.

Several limitations should be considered. First, a 2-year follow-up may not capture all late-growing ground-glass nodules, which can progress after longer periods of stability. Second, AI-derived malignancy indices depend on the software, scanner protocol, and validation population; therefore, AI outputs will be interpreted as secondary quantitative measures rather than definitive diagnoses. Third, the placebo contains 5% active YFSJ ingredients to support blinding, which could attenuate between-group differences if low-dose exposure has biological activity. Fourth, the mechanistic biomarker analyses are exploratory and require confirmation in independent studies. Finally, the findings may be most generalizable to Chinese clinical settings, where TCM use and CT surveillance workflows are similar to those in participating centers.

The final results will be disseminated through peer-reviewed publications, conference presentations, trial registry updates, and communication with participating centers and relevant stakeholders. The full protocol, statistical analysis plan, deidentified dataset, and statistical code will be made available after the completion of the trial when permitted by ethics approval, consent, and applicable data governance rules.

This protocol describes a multicenter randomized trial evaluating YFSJ for the prevention of high-risk pulmonary nodule progression. By combining guideline-based surveillance, objective CT progression end points, AI-assisted imaging, patient-reported outcomes, safety monitoring, and exploratory plasma biomarkers, this study will provide evidence on whether YFSJ is a safe and effective adjunctive strategy for patients with high-risk pulmonary nodules.