Decoding Recurrence in Early-Stage and Locoregionally Advanced Non-Small Cell Lung Cancer: Insights From Electronic...

Area of Science:

• Oncology
• Medical Informatics
• Computational Biology

Background:

• Recurrence after curative resection is common in early-stage and locoregionally advanced non-small cell lung cancer (NSCLC).
• Understanding recurrence risk factors is crucial for improving patient outcomes.
• Existing methods for data curation can be time-consuming and may not capture longitudinal recurrence patterns effectively.

Purpose of the Study:

• To develop and validate a natural language processing (NLP) system for efficient curation of recurrence data in NSCLC.
• To analyze longitudinal recurrence patterns and identify associated risk factors using the NLP system.
• To establish a foundation for predictive models for NSCLC recurrence.

Main Methods:

• A deep learning-based NLP system was developed to process over 700,000 clinical notes from 6,351 NSCLC patients.
• The system identified patients experiencing recurrence, with performance metrics including precision (94.3%), recall (93%), and F1 score (93.5%).
• Kaplan-Meier and Cox proportional hazards analyses were used to assess recurrence-free survival (RFS) and distant metastasis-free survival (DMFS) in relation to clinical features.

Main Results:

• The NLP system identified 336 recurrences (25.9%) among 1,295 patients with resected stage I-IIIA NSCLC.
• Local/regional recurrences accounted for 52.4%, distant metastases for 44%, and unknown recurrence for 3.6%.
• Stage IB patients showed a higher recurrence likelihood than stage IA (aHR, 1.63; P = .02). Clinically significant TP53 alterations were linked to lower RFS and DMFS in stage IA/IB patients (aHRs 1.89 and 2.47, respectively).

Conclusions:

• A scalable NLP system effectively curates real-world NSCLC recurrence data.
• The study identified specific clinical factors, including tumor stage and TP53 alterations, associated with recurrence.
• This NLP approach facilitates the development of predictive models for preventing, diagnosing, and treating NSCLC recurrence.