A Deep-Learning-Enabled Workflow to Estimate Real-World Progression-Free Survival in Patients with Metastatic Breast...

Area of Science:

• Oncology
• Medical Informatics
• Natural Language Processing

Background:

• Real-world progression-free survival (rwPFS) is critical for cancer drug research, but manual extraction from clinical notes is time-consuming.
• Traditional methods like RECIST are impractical for real-world data analysis.
• Natural Language Processing (NLP) offers a promising solution for accelerating the extraction of tumor progression data.

Purpose of the Study:

• To configure a pre-trained healthcare NLP framework for extracting structured progression events from free-text clinical notes and radiology reports.
• To study rwPFS in metastatic breast cancer (mBC) cohorts using this NLP framework.

Main Methods:

• Developed and validated a semi-automated workflow using de-identified EHR data from the nference nSights platform.
• Validated the workflow on 316 patients with hormone receptor-positive, HER2-negative mBC treated with Palbociclib and Letrozole.
• Defined outcome events as NLP-captured progression or therapy change, and censoring events as death, loss to follow-up, or study end.

Main Results:

• The NLP engine achieved 98.2% accuracy at the sentence level and 88% accuracy at the patient level for capturing initial progression within ±30 days.
• Median rwPFS for the cohort was 20 months; manual curation in a subset yielded 25 months, closely aligning with the workflow's 22 months.
• Sub-analysis integrating radiology reports and clinical notes showed high accuracy, with patient-reported outcomes (PHQ-8) significantly associated with disease progression.

Conclusions:

• The developed workflow enables rapid and reliable determination of rwPFS in mBC patients on combination therapy.
• Further validation across diverse external datasets and other cancer types is recommended for broader applicability and generalizability.