CERSI Collaborators: Feng Jiang, MD, PhD (PI); Charles S, White, MD (Co-I); Sanford A. Stass, MD (Co-I); Nevins W. Todd, MD (Co-I); Whitney M. Burrows, MD (Co-I); Hongjie Liu, PhD (Co-I); Hua He, PhD (Consultant); Pushpa Dhilipkannah, MS, (PM)

FDA Collaborators: Oladimeji Akinboro, MD, Office of Oncologic Diseases, Center for Drug Evaluation and Research

Project Start: September 16, 2021

Regulatory Science Challenge

Low-dose CT scan (LDCT) is an effective method detecting lung cancer in smokers early, but a significant number of smokers screened with LDCT have indeterminate lung nodules that indicate an elevated risk of cancer relative to smokers with smaller nodules. Indeterminate nodules are not large enough for clinicians to recommend surgery to remove them immediately, but individuals with indeterminate nodules may receive unnecessary diagnostic procedures and treatments, which can cause harm.

This project aims to address the need for more accurate diagnostic tools that can classify between cancerous and non-cancerous lung nodules in smokers, especially those who are screened by LDCT. This project provides an intended benefit to public health by developing a new method that includes different biomarkers. This method has the potential to improve the accuracy of lung cancer diagnoses, reduce the number of false positive test results, and minimize over-treatment, thus improving patient outcomes and reducing healthcare costs.

Project Description and Goals

Researchers aim to precisely distinguish between cancerous and non-cancerous lung nodules in positive LDCT findings, and thus reduce the false-positive rate and over-treatments. This study sets out to investigate this research question by incorporating new biomarkers with the previously developed integrated biomarker panel to develop a new method.

This study aims to develop a combined approach for separating cancerous and non-cancerous pulmonary growths. The goal is to achieve a sensitivity (i.e., positive test results among those who have cancer) of ≥95% and a negative predictive value (in other words, the likelihood that those who have a negative result do not actually have cancer) of up to 99%.

The study will be conducted in two phases:

1. Development of a combined approach for distinguishing cancerous from non-cancerous lung nodules by analyzing existing samples and using machine learning.
2. Prospective validation of the approach by collecting new samples and blindly validating it, as well as comparing this approach with the Mayo Clinic Model and Vancouver Model for dividing cancerous from non-cancerous lung nodules on baseline LDCT scans.

Long-term goals include but are not limited to: Informing personalized decision-making about lung cancer screening, reducing false-positive test results and invasive procedures for false-positive test results, and using cost-effective assays at the first level of screening to pre-identify smokers for subsequent screening using LDCT and other expensive or invasive procedures.