Union-LC-MPE Cohort:A Multicenter Cohort Study on Diagnosis and Prognosis of Malignant Pleural Effusion in Lung Cance

The Union-LC-MPE project aims to achieve the highest performance in precision medicine through Artificial Intelligence/Machine Learning (AI/ML) modelled on multimodal patients' data, together with implementing an AI/ML model in a real-life setting. Lung cancer remains the malignant tumor with the highest incidence and mortality rates in China, and malignant pleural effusion (MPE) is one of its most common complications: approximately 15%-50% of patients with non-small cell lung cancer develop MPE at initial diagnosis or during disease progression, which usually indicates advanced-stage (stage IV) disease, with a median survival of only 3-12 months and a cytological diagnostic yield of only 40%-60%. A set of patient-centered ML tools designed and validated for the project will therefore be developed for the differential diagnosis of benign versus malignant MPE and for predicting progression and outcome in LC-MPE patients. To maximize its impact, the use of Trustworthy explainable AI methodology will integrate the AI's inherent performances with the input of human intuition to construct a responsible AI application able to fully implement truly individualized diagnostic and therapeutic decisions in lung cancer with MPE, interpretable and trustworthy for clinicians. The final objective is the establishment of a multicenter Data Sharing and Elaboration Platform (DSEP) for LC-MPE. The DSEP will provide guiding tools for patients and clinicians, generating awareness on early diagnosis and individualized treatment of MPE, and lastly giving access to researchers and the general scientific community to the most up-to-date data sources on lung cancer with MPE.

Within the Union-LC-MPE project, an ad-hoc decision support system for LC-MPE patients will be developed. Patient decision aids are tools that might be used by patients either before or within a consultation with physicians. They explicitly represent the decision to be made and provide patients with user-friendly information about each diagnostic and treatment option by focusing on harms and benefits. This AI-assisted diagnostic and prognostic tool could allow patients to explain and clarify the high complexity of the information provided by the AI/ML approach, supporting the choice between a confirmatory invasive diagnostic workup and expectant management, as well as among the different treatment strategies. These decisional support systems have been demonstrated to be effective in empowering patients, improving their knowledge, promoting their active participation in clinical decision-making about treatments, and improving overall patient satisfaction with care while decreasing decisional conflict and decisional regret.

Finally, within the Union-LC-MPE project, it will be assessed whether using the decision support tool during the clinical consultation would foster the quality of the shared decision-making as well as the quality of the doctor-patient communication. Alongside the evaluation of the impact of the tool, it will also be evaluated whether the inclusion of the AI/ML predictive models in clinical practice will be an added value in supporting pulmonologists' and oncologists' clinical decision-making and decreasing cognitive fatigue and decisional conflict.

Union-LC-MPE adopts a two-pronged approach to develop a medical device through the creation and validation of retrospective and prospective AI-based models for the differential diagnosis of MPE and the prediction of prognosis in LC-MPE patients, using the integration of multisource data (real-world and multi-omics data) through a retrospective - setting up a multicenter platform of available data from 3,000 patients - and a prospective - multi-omics prospective data collection in 200 patients - study phase.

The retrospective cohort consists of patients with primary lung cancer complicated by malignant pleural effusion (LC-MPE). Data from an estimated 3,000 LC-MPE patients, diagnosed between January 1, 2022 and September 30, 2026, will be collected from Union Hospital of Tongji Medical College and all the collaborating clinical centers. For this retrospective cohort, an application for exemption of informed consent will be submitted to the ethics committee, given that the study uses already generated medical data and cannot identify specific individuals.

In the prospective phase, the study cohort consists of 200 patients with primary lung cancer complicated by MPE, diagnosed between October 1, 2026 and October 1, 2028. Baseline data will be collected including complete clinical information, multi-OMICs analysis, chest imaging, pathological reports, and genomic testing results, and survival and disease progression information will be supplemented through telephone follow-up, outpatient review and hospitalization records. The primary endpoint is overall survival (OS); secondary endpoints include pleural effusion progression-free survival (PE-PFS), duration of pleural effusion control, and 60-day and 30-day MPE recurrence rates. Informed consent for the study will be obtained before enrolment.