Circulating Proteome for Pulmonary Nodule Malignancy

ABSTRACT Background While lung cancer low-dose computed tomography (LDCT) screening is being rolled out in many regions around the world, differentiation of indeterminate pulmonary nodules between malignant and benign remains to a challenge for screening programs. We conducted one of the first systematic investigations of circulating protein markers for their ability to assess the risk of malignancy for screen-detected pulmonary nodules. Methods Based on four LDCT screening studies in the United States, Canada and Europe, we assayed 1078 unique protein markers in pre-diagnostic samples based on a nested case-control design with a total of 1253 participants. Protein markers were measured using proximity extension assays and the data were analyzed using multivariate logistic regression, random forest, and penalized regressions. Results We identified 36 potentially informative markers differentiating malignant nodules from benign nodules. Pathway analysis revealed a tightly connected network based on the 36 protein-coding genes. We observed a differential mRNA expression profile of the corresponding 36 mRNAs between lung tumors and adjacent normal tissues using data from The Cancer Genomic Atlas. We prioritized a panel of 9 protein markers through 10-fold nested cross-validations. We observed that circulating protein markers can increase sensitivity to 0.80 for nodule malignancy compared to the Brock model (p-value<0.001). Two additional markers were identified that were specific for lung tumors diagnosed within one year. All 11 protein markers showed general consistency in improving prediction across the four LDCT studies. Conclusions Circulating protein markers can help to differentiate between malignant and benign pulmonary nodules. Validating these results in an independent CT-screening study will be required prior to clinical implementation.