With a 5-year survival rate of ∼20%, lung cancer is one of the most prevalent and deadly cancers. About 70% of cases have lower survival rates than localized cancers because they are discovered when the disease is regional or far away. Therefore, better patient outcomes might result from increasing the number of cases found in a specific area.
After being isolated and analyzed with multiplex ELISA, extracellular vesicles (EVs) from non-small cell lung cancer (NSCLC) plasma samples were tested for protein biomarkers carried by these EVs. Six hundred and eighty-three control subjects (no known cancer diagnosis) and 161 lung cancer cases (stages I = 88, II = 34, III = 20, and IV = 19) were combined for analysis. While the control cohort was made up of 12% smokers with a median age of 58 years and 52% female population, the case-cohort was made up of 41% smokers with a median age of 63 years.
A machine learning system first found thirty-five relevant EV biomarkers to distinguish NSCLC patients from controls. The ROC displayed an AUC of 0.987 (95% CI: 0.981-0.993) with 97% (CI: 93%-99%) sensitivity and 92% (CI: 90%-94%) specificity. Researchers discovered identical performance across all stages examined when performance was broken down by stage, with sensitivities of 98%, 94%, 95%, and 100% for stages I through IV, respectively. Then, after narrowing the field of potential biomarkers down to 11, investigators achieved a 92% sensitivity and 92% specificity. They conducted an in-silico experiment, randomly dividing the dataset into ten training and validation sets (67%/33%) to better understand performance in a future independent validation. It revealed that across the 10 held-out validation sets, the mean AUC was 0.982 (range: 0.968–0.991), the mean sensitivity was 95% (range: 85%–100%), and the mean specificity was 92% (range: 86%–95%).
Through the use of a liquid biopsy test based on EV biomarkers, our pilot study suggests a potential method for identifying lung cancer at an early stage, when treatment may be more successful. For biomarker and algorithm validation, our findings call for more research using different datasets. To better assess the best use in the clinical setting, future planned studies will also assess the impact of controls with confounding disorders, such as benign lung nodules.
Reference: annalsofoncology.org/article/S0923-7534(22)01899-3/fulltext
