Paul Lampe/McGarry Houghton

Project Title/Research Areas: Hybrid Plasma Markers that Complement CT Imaging for Early Lung Cancer Detection
Principal Investigator/Institution:  Paul Lampe, Ph.D. / McGarry Houghton, M.D., Fred Hutchinson Cancer Research Center

Description of Current Activity

Lung cancer is the leading cause of cancer deaths worldwide with >159,000 deaths annually in the US alone. Making matters worse, 5-year survival rates remain a dismal ~18%, since most lung cancer cases are identified at an advanced stage, which confers a poor prognosis. In contrast, 5-year survival rates are much higher (~56%) when lung cancer is diagnosed at its earliest stages. The National Lung Screening Trial (NLST) employed low-dose Computed Tomography (CT) imaging of the chest to screen for lung cancer in a high-risk population (smokers aged 55-74). This study demonstrated a 20% reduction in mortality in the group receiving CTs when compared to standard care. Unfortunately, pulmonary nodules are a relatively common finding in the lungs of smokers, such that 26% of the subjects required additional work-up for their pulmonary nodules, with <5% of them representing cancer. Analysis of several screening cohorts indicates that 25-50% of smokers >50 years of age have CT identifiable pulmonary nodules but very few of them (~2.5%) are caused by lung cancer. Implementation of recommended screening in VA populations found 56% with nodules but only 1.5% with cancer. Current guidelines for incidentally detected pulmonary nodule evaluation call for additional imaging and/or invasive biopsy procedures depending upon the size/characteristics of the nodule and key clinical parameters (e.g., age, smoking history). There are clinical and economic needs for innovative risk stratification strategies as we begin this new era of recommended screening.

Our project involves the combination of plasma biomarkers with CT image data to improve lung cancer screening. We created a high-density antibody array containing 3200 different antibodies that we use to interrogate pre-diagnostic sample sets from observational trials in a nested case-control design study to evaluate proteomic, glycomic and autoantibody differences. Using pre-diagnostic lung cancer samples from the Cardiovascular Health Study (CHS), we found 30 proteomic, glycomic or autoantibody biomarkers that were significantly increased (p<0.002) in people that are subsequently diagnosed with lung cancer.


As part of this grant, we have used plasma samples from 297 lung nodule positive subjects that have been screened via CT and have known cancer/nodule status (147 were cancer) to test these 30 markers and potentially discover additional candidates (Aim 1, finished). We are currently combining these data with CT imaging parameters and clinical data to create a risk prediction model that we will test in a similar sized prospectively collected cohort (Aim 3). We have also purified B-cells from lung tumor tissue and examined whether autoantibodies present in plasma are tumor-derived to assess their utility for the detection of cancerous nodules (Aim 2, results submitted for publication). We have also begun to add “radiomic” approaches to supplement biomarker and conventional CT image features to improve lung cancer early detection (see figure below). 

Risk Predication Model for Pulmonary Nodule Diagnosis in LDCT Lung Cancer Screening Chart with a footnote below describing the graphic
Lung cancer nodule detection by low-dose computed tomography (LDCT) imaging and subsequent treatment can save lives but at considerable financial and quality of life costs, thus making the right choice between benign or malignant status is a critical part of the choice to treat or observe. We are combining two novel approaches to improve risk stratification for subjects with pulmonary nodules detected by LDCT imaging. The first is based on a large format antibody array we created containing >3200 different antibodies to investigate plasma sample sets. We can utilize the same antibody array platform for proteomic, glycomic, and autoantibody-antigen complex interrogation by implementing three distinct probing strategies. The second novel approach is the use of quantitative nodule features extracted from CT images using the methods of 'radiomics‘. Currently, we are validating our 24-parameter model (AUC > 0.9) that combines semantic imaging factors, with both plasma biomarkers and radiomic CT features for classification of CT-detected nodules as malignant or benign.