Advancing Precision Lung Cancer Surveillance and Outcomes in Diverse Populations

Protocol for an Observational Cohort Study Integrating Real-World Data and Microsimulation to Assess Imaging Surveillance Strategies in Stage I-IIIA NSCLC Patients in OneFlorida+

This observational study evaluates the effectiveness of computed tomography (CT) imaging surveillance after curative-intent treatment for stage I-IIIA non-small cell lung cancer (NSCLC) in a diverse U.S. population.

The main questions are:

How do CT surveillance use and adherence vary by race, ethnicity, and socioeconomic status?

Does semi-annual CT surveillance improve outcomes compared with annual surveillance?

Adults ages 20-90 with stage I-IIIA NSCLC treated between 2012 and 2026 will be identified using OneFlorida+ electronic health records, tumor registry data, claims, and clinical notes. Patients will be followed for up to five years after curative-intent therapy to evaluate surveillance patterns, recurrence, second primary lung cancers, complications, and survival.

This study, Advancing Precision Lung Cancer Surveillance and Outcomes in Diverse Populations (PLuS2), is an observational cohort study designed to evaluate real-world computed tomography (CT) imaging surveillance strategies following curative-intent treatment for early-stage non-small cell lung cancer (NSCLC). The study uses existing clinical data and does not assign interventions.

The study population includes adults ages 20-90 with pathologically confirmed stage I-IIIA NSCLC who completed curative-intent therapy. Patients treated between 2012 and 2026 will be identified within the OneFlorida+ Clinical Research Consortium and followed for up to five years after treatment to assess surveillance patterns and outcomes.

UF Health serves as the data coordinating site in collaboration with the OneFlorida+ Data Trust. Data sources include structured electronic health records, tumor registry data, selected claims, and unstructured clinical notes. Data extraction and cohort identification are conducted centrally using standardized definitions aligned with the PCORnet Common Data Model.

The primary objectives of the study are to:

1. Describe utilization and adherence to guideline-recommended CT surveillance and evaluate determinants of use by race, ethnicity, socioeconomic status, and clinical factors.
2. Compare recurrence, second primary lung cancers, complications, and survival among patients undergoing semi-annual versus annual CT surveillance.
3. Use observational findings to inform microsimulation models that project long-term outcomes under alternative surveillance strategies.

Clinical natural language processing (NLP) methods are used to classify CT scans as routine surveillance versus symptom-directed diagnostic imaging and to extract recurrence indicators, smoking history, and selected social determinants of health from unstructured clinical notes. NLP outputs are integrated with structured data to improve classification accuracy and completeness.

Quality assurance procedures include automated checks for completeness, range, and internal consistency, validation against source EHR and tumor registry records, and maintenance of a standardized data dictionary defining all variables and coding systems. Standard operating procedures govern cohort assembly, quarterly data refreshes, data management, analysis, and reporting.

The expected cohort size is approximately 1,700 patients, providing sufficient power to evaluate surveillance utilization patterns and outcome differences by surveillance interval. Missing data will be addressed using multiple imputation and sensitivity analyses.

Statistical analyses include descriptive methods, multivariable and mixed-effects regression models, and time-to-event and competing-risk analyses. Results from the observational analyses will be used as inputs for microsimulation modeling to estimate long-term population-level outcomes associated with different CT surveillance strategies.

Inclusion Criteria:

• Adults diagnosed with primary, pathologically confirmed stage I-IIIA non- small cell lung cancer (NSCLC).
• Received curative-intent therapy, including:

Surgical resection, or Stereotactic body radiation therapy (SBRT), with or without adjuvant therapy.

• Completed curative-intent therapy and survived at least 5 months after therapy initiation.
• Received care within the University of Florida (Gainesville or Jacksonville) health system, and are included in the OneFlorida+ clinical data network.
• Have available:

Structured EHR data (including tumor registry linkage, imaging, procedures, and vital statistics), and Unstructured clinical documents (clinic notes, radiology reports, pathology reports) enable NLP extraction.

- Have ≥1 follow-up encounter in the EHR after treatment completion to allow surveillance assessment.

Exclusion Criteria:

• Histology other than NSCLC, including:

Small-cell lung cancer Carcinoid tumors Rare non-NSCLC primary tumors

• Metastatic disease (stage IV) at diagnosis.
• No curative-intent therapy received (e.g., palliative treatment only).
• Patients with incomplete treatment records prevent the determination of the treatment completion date.
• Patients with no surveillance-relevant follow-up data, including:

No post-treatment clinical notes No radiology/pathology records No linked tumor registry or vital status data

• Patients with recurrence or who die within the first 5 months after starting curative-intent therapy (insufficient window for guideline-defined surveillance).
• Patients with missing identifiers prevent linkage to cancer registries or the death index.
• Individuals whose imaging cannot be classified as surveillance vs. diagnostic due to the absence of relevant structured or unstructured data.