ASCO 2025: Dr. Paul Conkling Shares Insights on PD-L1 Testing Practices in Metastatic NSCLC

Paul R. Conkling, MD, senior medical director at Ontada, joined Lung Cancers Today at the 2025 American Society of Clinical Oncology (ASCO) Annual Meeting to discuss his research on programmed cell death-ligand 1 (PD-L1) testing practices in metastatic non–small cell lung cancer (mNSCLC). The study investigated insights from US community oncology practices to “identify opportunities for augmenting personalized medicine in mNSCLC care.”

“PD-L1 testing is essential for guiding treatment decisions in mNSCLC, particularly when considering immunotherapy,” Dr. Conkling explained. “We’ve seen increasing adoption of PD-L1 testing at the community level, which is fantastic. However, we wanted to get a clearer picture of actual testing practices in a contemporary real-world setting.”

The observational study focused on adults with mNSCLC who were diagnosed with de novo stage IV disease or disease that had progressed from an earlier stage, who began first-line treatment between November 1, 2022, and August 31, 2024.

“Our primary data source was the iKnowMed electronic health record (EHR) system, developed and owned by Ontada,” Dr. Conkling said. “We started by extracting data from the structured fields in the EHR; however, we knew that important information might also be the unstructured data, so we supplemented the structured fields by using a validated natural language processing (NLP) algorithm to capture PD-L1 documentation.”

Dr. Conkling explained that unstructured data are “rich in detail and context, but can be more difficult to analyze directly.” By incorporating unstructured data into the NLP algorithm, investigators identified 313 additional patients who received PD-L1 testing, which increased the rate of testing from 75% to 89% in the community oncology setting.

“The key finding was that almost 90% of patients with mNSCLC in our study received PD-L1 testing,” Dr. Conkling said. “It suggests that PD-L1 testing is becoming increasingly integrated into routine clinical practice in community oncology settings.”

Dr. Conkling highlighted that the investigators compared the results of the NLP approach to “manual abstraction of a subset of 100 patient records where PD-L1 testing was not initially evident in the structured data.” The sample revealed that the NLP algorithm performance was 89% for sensitivity and 90% for the F1 score.

“The high testing rate we observed suggests that physicians are generally adhering to guidelines recommending PD-L1 testing for patients with mNSCLC,” Dr. Conkling said. “The fact that approximately 10% of patients did not receive testing could be due to a variety of factors, including patient preference, comorbidities, or selection of hospice care. Our results suggest that PD-L1 testing has been largely implemented in clinical practice for lung cancer care in the community oncology setting. The almost 90% testing rate is encouraging, but it highlights the need for continued efforts to ensure access to testing for all patients who could benefit.”

Dr. Conkling emphasized the key takeaways from his research, which included both the high rate of PD-L1 testing rates in community oncology practices and how “leveraging NLP algorithms can significantly improve the accuracy of real-world data analysis.”

“Future research should focus on understanding the strategies and processes that successful community oncology practices used to implement PD-L1 testing, with the goal of applying those learnings to the implementation of forthcoming actionable biomarkers,” he said. “Ultimately, our goal is to ensure that all patients receive the personalized care they deserve.”