Skip to Main Content

An official website of the United States government

Blood based biomarkers for early detection of lung cancer in those who have never smoked

Principal Investigator
Name
Edwin Ostrin
Degrees
MD, PhD
Institution
University of Texas MD Anderson Cancer Center
Position Title
Associate Professor
Email
About this CDAS Project
Study
PLCO (Learn more about this study)
Project ID
2025-0031
Initial CDAS Request Approval
Jun 12, 2025
Title
Blood based biomarkers for early detection of lung cancer in those who have never smoked
Summary
While most cases of lung cancer are related to tobacco use, with effective tobacco control, lung cancer in those with light or no smoking history is becoming a greater public health concern. It is estimated that lung cancer in those who have never smoked (LCINS) would be the fifth-leading cause of cancer death worldwide and the seventh-leading cause in the United States. No widely accepted early detection strategy exists for those who are not high risk for lung cancer based on a tobacco use history. Efforts to broaden low-dose CT based lung cancer screening to those at lower risk have been likely met with increasing rates of overdiagnosis. Our team has an extensive track record of developing and validating blood-based biomarkers for cancer risk, including a 4-protein biomarker panel (4MP) that has been extensively validated to improve prediction of lung cancer risk over clinical risk factors alone. Here we demonstrate that the 4MP and a newly validated 4-metabolite biomarker panel (4MetP) have similar performance in LCITNS. Our translational objective of this proposal is to test the predictive performance of the 4MP and 4MetP, individually and in combination with clinical risk models, for 1-year risk prediction of lung cancer among those no smoking history. We hypothesize that a composite panel of the 4MP, 4MetP, with addition of clinical risk characteristics, can identify those with no smoking history and at a high enough risk that they may benefit from lung cancer screening. Samples from those with no smoking history from several MD Anderson cohorts and the Prostate, Lung, Colorectal, and Ovarian (PLCO) trial will be utilized to build a composite panel consisting of blood-based biomarkers and clinical characteristics to identify those at risk for lung cancer.

Aim 1: Evaluate the predictive performance of the 4MP and 4MetP for 1-year risk of lung cancer among never-smoking individuals. We will build on preliminary data and evaluate the performance of the 4MP and 4MetP in cohorts of LCINS and matched controls from MD Anderson. We will measure the 4MP and 4MetP in 12-month prediagnostic LCINS samples from the PLCO (n=70) as well as ten times the number of unmatched controls to externally validate the biomarker panels. We will assess the performance of these panels and any composite biomarker panel to predict 1-year risk of developing lung cancer as well as risk of lung cancer mortality.

Aim 2: Assess whether an algorithm that consider repeat biomarker testing improves sensitivity and lead-time detection of lung cancer among never-smoking participants in the PLCO cohort. We will obtain all available longitudinally collected samples from the cases and controls specified in Aim 1. We will evaluate whether an individualized algorithm accounting for biomarker trends can improve upon a single threshold measurement to improve diagnostic performance.
Aims

-Aim 1: Test the predictive performance of the 4MP and 4MetP for LCINS risk.
-Aim 1A: Evaluate the predictive performance of the 4MP and 4MetP for detecting lung cancer among never smoker individuals in samples from MD Anderson. In preliminary experiments, we have tested the performance of our biomarker panels in a subset of samples from individuals with newly diagnosed cases of lung cancer versus never-smoking controls from three cohorts at MD Anderson. Here, we will measure the 4MP and 4MetP and clinical risk on the rest of these samples (n=326 cases, n=404 controls) and additional controls. We will evaluate the performance of these tools individually and in combination, with the ability to retrain panels to optimally perform in LCINS. Our focus will be on building a biomarker panel for LCINS with high specificity to minimize false-positive results and identify a group of high-risk individuals who may benefit from LDCT.
-Aim 1B: Evaluation of combined clinical and biomarker scores for 1-year risk prediction of LCINS in the PLCO cohort. We will externally validate the 4MP, the 4MetP, and combined, retrained model from Aim 1A in samples from PLCO. The specimen set will consist of pre-diagnostic serum collected within five years of preceding a diagnosis of lung cancer (n=70) and ten times the number of randomly selected non-case control (n=700) participants that did not develop any cancer during the PLCO study follow-up period. For all cohorts, time-dependent (e.g. 0-1 year, 1-2 years, etc) performance estimates include area under the Receiver Operating Characteristics curve (AUC), sensitivity, and specificity as well as population-adjusted positive predictive value (PPV), and negative predictive value (NPV) will be determined. We will additionally establish clinical decision-making cut points for 1-year risk prediction of lung cancer amongst never-smoking individuals using a threshold that that corresponds to the same level of lung cancer risk as those that are currently eligible for screening in the US, suggesting that this may be a higher risk group who may benefit from LDCT.

-Aim 2: To assess whether an algorithm that considers repeat biomarker testing improves sensitivity and lead-time detection of lung cancer among never-smoking participants in the PLCO cohort. In the PLCO cohort, 49 lung cancer cases had 2 or more serial blood draws available. Here, we will assess whether biomarker panel scores are stable in non-cases, and stable or increasing in lung cancer cases. We will use the Parametrical Empirical Bayes (PEB) model as well as a novel improved PEB (iPEB) algorithm developed by trial co-investigator Dr. Irajizad to evaluate the benefit of repeat biomarker testing over a single-threshold method. Model performance estimates will include AUC, sensitivity, specificity, and lead-time.

Collaborators

Edwin Ostrin (University of Texas MD Anderson Cancer Center)
Ehsan Irajizad (University of Texas MD Anderson Cancer Center)
Johannes Fahrmann (University of Texas MD Anderson Cancer Center)