Emerging Strategies in Lung Cancer Screening: Blood and Beyond

Abstract

Background

Although low dose computed tomography (LDCT)-based lung cancer screening (LCS) can decrease lung cancer-related mortality among high-risk individuals, it remains an imperfect and substantially underutilized process. LDCT-based LCS may result in false-positive findings, which can lead to invasive procedures and potential morbidity. Conversely, current guidelines may fail to capture at-risk individuals, particularly those from under-represented minority populations. To address these limitations, numerous biomarkers have emerged to complement LDCT and improve early lung cancer detection.

Content

This review focuses primarily on blood-based biomarkers, including protein, microRNAs, circulating DNA, and methylated DNA panels, in current clinical development for LCS. We also examine other emerging biomarkers—utilizing airway epithelia, exhaled breath, sputum, and urine—under investigation. We highlight challenges and limitations of biomarker testing, as well as recent strategies to integrate molecular strategies with imaging technologies.

Summary

Multiple biomarkers are under active investigation for LCS, either to improve risk-stratification after nodule detection or to optimize risk-based patient selection for LDCT-based screening. Results from ongoing and future clinical trials will elucidate the clinical utility of biomarkers in the LCS paradigm.

Introduction

Despite considerable progress in both diagnosis and treatment, lung cancer remains the leading cause of cancer death among men and women in the United States (1). While treatment for early-stage lung cancer is highly effective and potentially curative, most patients are diagnosed at late stage, which carries poor prognosis and 5-year overall survival <25% (2). Therefore, recent research efforts have focused on improving screening and early detection of lung cancer.

Large randomized trials, such as the National Lung Screening Trial, have demonstrated that lung cancer screening (LCS) using chest low dose computed tomography (LDCT) decreases lung cancer mortality by 20% in high-risk individuals (3). As a result, in 2013, the United States Preventive Task Force (USPSTF) recommended yearly LDCT-based LCS for eligible individuals. Despite this recommendation, LDCT remains underutilized, with less than 6% of at-risk individuals screened in 2021 (2). Furthermore, fewer than 25% of these individuals continue annual screening beyond the initial LDCT (4). Numerous factors contribute to low uptake of LCS, including patients’ and providers’ lack of knowledge about screening, stigma associated with lung cancer, as well as concerns about potential harm and unnecessary testing due to false-positive imaging results (5).

In addition, the stringent smoking history and age eligibility criteria of USPSTF guidelines may leave many individuals, particularly those from racial minorities who are at increased risk for developing lung cancer, ineligible for screening (6). Although USPSTF guidelines were expanded in 2021 to address these disparities by lowering age and smoking history requirements, further evidence-based strategies to identify individuals at high risk for lung cancer who fall outside existing LCS criteria are needed. For instance, never smokers, who account for approximately 15% of lung cancer cases in the United States (7), are automatically excluded from any smoking-based eligibility assessment.

Based on this critical unmet need, numerous biomarkers, utilizing blood, airway epithelia, breath, sputum, and urine, are under study to complement LDCT and improve LCS (Tables 1 and 2). Biomarkers with high specificity may reduce unwarranted invasive procedures associated with false-positive LDCT results. Conversely, biomarkers may also identify individuals at a high risk for lung cancer who may benefit from CT screening but are excluded by current eligibility criteria. In this review, we examine emerging molecular biomarkers in clinical development for early detection of lung cancer.

Blood-Based Biomarkers

Blood-based biomarkers offer key advantages, including the relative noninvasive nature of sample collection and the availability of established laboratory procedures for specimen preparation, component isolation, and assay performance. Biomarkers under investigation encompass a range of analytes, including proteins, autoantibodies, microRNAs (miRNAs), circulating DNA, and methylated DNA, alone or in combination with clinical features such as demographic and radiographic characteristics. Several biomarker panels have undergone extensive validation for classification of indeterminate pulmonary nodules, with other biomarkers under ongoing investigation.

Autoantibodies and Proteins

Tumor-associated autoantibodies reflect the initial humoral immune response against a tumor. Increase in tumor-associated autoantibodies can be observed months to years prior to clinical or radiographic evidence of cancer. A potential disadvantage of this approach is limited sensitivity; while healthy individuals typically have undetectable or low antibody titers (thereby conveying specificity), antibodies may also be absent in patients with disease (8).

The Early Cancer Detection Test (EarlyCDT-Lung), an enzyme-linked immunoassay measuring 7 autoantibodies to tumor-associated antigens (tumor protein P53, cancer-associated gene, New York esophageal squamous cell carcinoma 1, SRY-box 2, GBU4-5, Hu antigen D, melanoma-associated antigen 4) has been validated in multiple settings (9). Among patients with symptomatic lung cancer and patients at increased risk for lung cancer, the EarlyCDT-Lung test demonstrated a specificity of 91%; however, sensitivity was only around 40%. EarlyCDT classifies test results based on a proprietary scoring algorithm and autoantibody-specific cut-off values. Positive results are reported as “moderate level” if the levels of one or more autoantibodies in the panel are above the “low” cut-off value but all are below the “high” cut-off value. Positive results are reported as “high level” if the levels of one or more autoantibodies are above the “high” cut-off value (10).

The Early Diagnosis of Lung Cancer Scotland trial, which started prior to the routine use of LDCT for LCS, randomized over 12 000 individuals at risk for developing lung cancer to receive the EarlyCDT-Lung test vs standard clinical practice at that time (imaging pursued if patients developed symptoms). If the EarlyCDT-Lung test was positive, patients underwent LDCT every 6 months for up to 2 years. While EarlyCDT-Lung-directed surveillance did not increase the frequency of lung cancer detection, lung cancers detected in the intervention arm tended to be earlier stage than those diagnosed in the control arm. Specifically, 56 lung cancers were detected in the intervention arm (41% early stage), and 71 in the control arm (27% early stage) (11). While promising, longer follow up and prospective evaluation within a contemporary LDCT-based LCS program would be needed to determine the potential clinical utility of this assay.

Nodify XL2 is a plasma-based test integrating 2 plasma proteins linked to cancer-related inflammation (LG3BP and C163A) and clinical risk factors (12). Initially derived from a 13-protein classifier, the test uses multiple reaction monitoring mass spectrometry and has been refined over the years to its current focus on LG3BP, C163A, and 5 clinical risk features (age, smoking status, nodule diameter, nodule edge characteristics, and nodule location). Nodify XL2, previously referred to as Xpresys Lung 2, was prospectively validated in the PANOPTIC study of 685 patients presenting with 8–30 mm nodules. In a subgroup of 178 patients with a clinician-assessed pretest probability of cancer ≤50% and lung cancer prevalence of 16%, the test had sensitivity of 97%, specificity of 44%, and negative predictive value (NPV) of 98%. The classifier performed better than positron emission tomographic (PET) imaging, validated lung nodule risk models, and physician cancer probability estimates (P < 0.001). The authors suggested that this test could result in a potential 40% reduction in invasive procedures for benign nodules (13). A multicenter, randomized control trial of Nodify XL2 in the management of low- to moderate-risk lung nodules is ongoing (NCT04171492). Currently, the Nodify XL2 test is available commercially and covered by Medicare for patients diagnosed with a lung nodule between 8 and 30 mm and who have a pretest cancer risk of 50% or less assessed by the Mayo Clinic model (12).

miRNAs

Circulating miRNAs have emerged as promising biomarkers for lung cancer diagnosis. Large retrospective studies demonstrated that the use of 2 miRNA-based strategies, a 24-miRNA signature classifier as well as a 13-miRNA signature test, resulted in a 4–5-fold reduction in the LDCT-false-positive rate with high specificity (81%–75%) and sensitivity (87%–78%) (14, 15).

The BioMILD study, a prospective trial of >4000 individuals with heavy smoking history, assessed the value of a 24-miRNA signature classifier at the time of baseline LDCT with the goal of personalizing LCS intervals. CT positive individuals had a 16-fold higher 4-year lung cancer incidence than CT negative individuals, and miRNA signature classifier positive individuals had a 2-fold higher 4-year lung cancer incidence than miRNA signature classifier negative individuals. Lung cancer incidence at 4 years was 0.8% for CT-/miRNA−, 1.1% for CT-/miRNA+, 10.8% for CT+/miRNA−, and 20.1% for CT+/miRNA+ participants. The authors concluded that the combined use of LDCT and miRNA testing at baseline can predict individual lung cancer incidence and may guide personalized screening intervals (16). The ongoing COSMOS II study is evaluating prospectively the 13-miRNA signature test along with LDCT in high-risk individuals (15).

Circulating Tumor Nucleic Acids

The role of ctDNA (circulating tumor DNA) as a biomarker for selection of molecularly targeted therapies in advanced non-small cell lung cancer (NSCLC) is well established. Recently, its value in early-stage lung cancer detection has also gained attention as improving next-generation sequencing technologies have markedly increased the sensitivity of ctDNA detection in plasma. CancerSEEK, a pan-cancer early detection blood test assessing tumor-specific mutations across 16 genes and select protein biomarkers (e.g., cancer antigen 125, carcinoembryonic antigen [CEA], cancer antigen 19–9), was studied in >1000 patients previously diagnosed with solid tumors and 850 healthy controls. Specificity was >99% and sensitivity was 59% in 104 patients with lung cancer (17). Notably, individuals could have had symptomatic stage I–III disease and thus, this study set did not reflect a true screening cohort. A prospective, interventional study of over 10 000 women without cancer coupled an earlier version of the CancerSEEK assay with PET/CT (18). Obtaining PET/CT for individuals with positive blood test increased specificity and positive predictive value (PPV) for cancer detection; however, the overall accessibility and cost implications of PET/CT in this setting must be considered. Further validation of CancerSEEK is ongoing in a prospective, observational study (NCT04213326).

Lung cancer-specific approaches, such as Lung Cancer Likelihood in Plasma (Lung-CLiP), are also under development. This biomarker test integrates ctDNA detection via Cancer Personalized Profiling by Deep Sequencing with machine learning to predict the presence of NSCLC-derived cell-free DNA (cfDNA) in blood. Lung-CLIP had 98% specificity, with sensitivity of 41% for stage I, 54% for stage II, and 67% for stage III cases (19). In this training cohort, most cases were incidentally diagnosed, early-stage lung cancers and were not identified by LDCT-based LCS screening. Prospective studies evaluating this test’s performance in a screening population are currently lacking.

Other recent cancer screening assays are focusing primarily on methylation and fragmentation patterns. For instance, the Galleri test measures methylated patterns of cfDNA to detect multiple cancer types. In a study of >6600 individuals with and without cancer, the test demonstrated 99% specificity and 55% sensitivity for cancer detection across tumor types and stages. In training and validation cohorts of early-stage lung cancer patients, the Galleri test had a sensitivity of approximately 20% (20). Updated results from this study demonstrated positive test results in 1.4% (92/6621) of participants, of which 38% (35/92) were confirmed cancer cases. The PPV of the test was approximately 40% (21). Of note, this study’s overall cohort included symptomatic participants with cancer, as well as >80% White, non-Hispanic patients. Although the clinical validity of this screening assay is being tested in the SUMMIT study of 13 000 adults without known cancer diagnosis (NCT03934866), the Galleri test is already available in the United States with a prescription from a licensed healthcare provider. At this time, there is limited evidence-based data to guide practitioners through potential clinical scenarios, such as when subsequent imaging studies do not explain a positive Galleri test result. As of July 2023, the test is not covered by most payers (22).

DNA evaluation of fragments for early interception (DELFI) employs genome-wide analysis of patterns of cfDNA fragmentation to detect cancer (23). Because cfDNA fragmentomes represent both genomic and chromatin characteristics, they have the potential to identify a spectrum of tumor-derived changes. In a study of 365 symptomatic patients at risk for lung cancer, analysis of cfDNA using DELFI, clinical risk factors, CEA levels, and CT imaging detected 94% of patients with cancer across stages and subtypes with 80% specificity. Using Monte Carlo simulations for a theoretical population of 100 000 high-risk individuals, pre-screening for LDCT with DELFI would detect 394 additional lung cancer cases (roughly 8-fold increase) compared to LDCT alone and would reduce the number of unnecessary procedures by approximately 50% (23). A multicenter, prospective study to determine the sensitivity and specificity of the DELFI Lung Cancer Screening Test among LDCT-based LCS eligible patients is underway (NCT05306288).

Biomarkers from Other Sources

Airway Epithelia

Smoking-induced alterations in airway epithelial gene expression may serve as potential biomarkers for lung cancer detection. The Percepta Genomic Sequencing Classifier analyzes 80 genes from bronchoscopic brushings. In a population of smokers with histologically normal bronchial airways, the test identified lung cancer cases with 80% sensitivity and 84% specificity, with preserved sensitivity (90%) for stage I tumors (24). Based on these findings, 2 large prospective studies (AEGIS I and II) enrolled patients undergoing bronchoscopy due to lung cancer suspicion (25). Airway brushings from normal appearing mainstem bronchi underwent RNA expression profiling. In these trials, the classifier had high sensitivity (88% and 89%) yet lower specificity (47% in both). To increase clinical utility, the test was combined with physician-assessed pre-bronchoscopy pretest probability of cancer (low: < 10%, intermediate: 10%–60%, high: > 60%). In a combined group of low- and-intermediate probability lesions with nodules <3 cm, the sensitivity of the test was 88%, with a negative predictive value (NPV) of 94%. Further, the test was used to re-classify indeterminate lung lesions among 412 patients recruited in AEGIS I and II. Almost 30% of intermediate-risk lung lesions were down-classified to low risk with 91% NPV, and over 50% of low-risk lesions were down-classified to very low risk with 99% NPV (26). As such, a negative test in patients may allow physicians to avoid unnecessary invasive procedures. While the Percepta Genomic Sequencing Classifier is commercially available and covered by Medicare (27), clinical utilization remains limited. To improve feasibility, this methodology is also being applied to nasal swabs, which may be more accessible and less resource intensive than airway sampling.

Exhaled Breath

The use of exhaled breath analysis as an alternative approach for LCS has been increasingly discussed in recent years due to its noninvasive nature and cost-effectiveness of collection, and the possibility of real-time analysis. These biomarkers are based on volatile organic compounds (VOCs), which are carbon-containing compounds originating from endogenous metabolic processes and are already established for medical use as urea breath tests for Helicobacter pylori and hydrogen-methane tests for small-bowel bacteria overgrowth (28).

In a study of 43 patients with NSCLC and 41 healthy control participants, solid phase micro-extraction and gas chromatography and mass spectrometry (GC/MS) analysis demonstrated that VOCs, such as 1-butanol and 3-hydroxy-2-butanone, were found at significantly higher concentrations in the breath of lung cancer patients (29). Another VOC analytic technique, ion mobility spectrometry (IMS) offers a shorter processing time compared to GC/MS (approximately 8 minutes compared to 1 hour). In a pilot study, this method distinguished 32 patients with lung cancer from 54 healthy controls using a combination of peak regions within IMS chromatograms (30). Limitations of VOC detection and analysis will need to be considered, including lack of standardized sample collection and storage, as well as low VOC concentrations (generally nanomolar or picomolar) in clinical samples.

Sputum

Although sputum can be collected conveniently and may feature morphological abnormalities in lung cancer, it has historically demonstrated poor sensitivity for diagnosis of malignancy. Recent advances in cytologic techniques and analytic algorithms may improve the utility of sputum-based biomarkers. One potential area of interest is detection of lipophilic and amphiphilic porphyrins, which have a high affinity for neoplastic tissue. In a proof-of-concept study, a tetra (4-carboxyphenyl) porphyrin (TCPP)-based assay was positive more often in individuals with cancer compared to those who were at high risk (US military veterans and ≥20 pack smoking history). The assay, combined with years smoked, had an overall accuracy of 81% in the test population (31). Building on this discovery, the CyPath Lung test combines automated flow cytometry using TCPP and antibody profiling with machine learning. In a validation cohort of 150 patients with lung cancer or at high risk for lung cancer, the test demonstrated 82% sensitivity, 88% specificity, NPV 96%, and PPV 61%. However, cancer prevalence in this data set was approximately 19%–far higher than the expected cancer prevalence in a LCS population (32). Additionally, sputum miRNA has been investigated for the characterization of indeterminate pulmonary nodules, with one study finding that the expression of 13 sputum miRNAs identified malignant pulmonary nodules with a sensitivity of 83% and specificity of 88% (33). Prospective studies of sputum-based biomarker testing are necessary to understand the clinical validity and applicability of this approach.

Urine

Urine represents another potential approach for biomarker development; it is abundant, easily sampled, and does not require extensive processing. In a study of 225 patients, a three-protein panel of insulin like growth factor binding protein 1, soluble interleukin-1 receptor antagonist, and CEA-related adhesion molecule 1 discriminated NSCLC cases from healthy control participants with a sensitivity of 84% and specificity of 95% (34). In a separate study of 1005 patients, elevated urinary creatine riboside and N-acetylneuraminic acid were significantly associated with lung cancer risk and prognosis before clinically detectable disease appeared (35). Importantly, these and other urine metabolic analytes may be affected by metabolism-altering factors such as diet, medication, and comorbidities.

Future Directions

Despite tremendous advances in biospecimen acquisition and molecular testing techniques, ongoing challenges have limited the clinical utility of existing biomarkers for LCS. Blood and sputum-based testing has been historically limited by low sensitivity. Exhaled breath-based testing is cumbersome and lacks standardization. Airway epithelial testing requires invasive, bronchoscopic sampling. Urine-based tests may be confounded by diet, medications, and medical comorbidities. Furthermore, an overall limitation of many LCS biomarker studies to date, including CancerSEEK, the Galleri, and DELFI tests, is that sample sets consisted of symptomatic individuals. Accordingly, results may not reflect performance in a true screening setting. Rigorous validation in an asymptomatic, diverse screening population is pending.

While certain assays, including Nodify XL2, the Galleri test, and the Percepta Genomic Sequencing Classifier, are currently available commercially, broad adoption among clinicians remains limited, likely due to concerns regarding feasibility, cost, as well as lack of evidence-based recommendations to guide practitioners utilizing these tests. Further, uptake of biomarker testing for LCS remains limited by the modest reach of LCS itself. As of 2023, fewer than 10% of eligible individuals in the United States undergo LDCT (2). Ongoing outreach and education efforts directed toward health care professionals and the general public are needed to increase uptake to levels comparable with more established cancer screening modalities such as mammography for breast cancer. Although the noninvasive nature of LDCT makes it feasible and convenient, any biospecimen-based biomarker—whether blood, nasal swab, exhaled breath, or the like—will inherently require procedure modification. By contrast, other cancer screening methods such as Pap smear for cervical cancer and colonoscopy for colorectal cancer feature built-in opportunities for tissue acquisition. Finally, biomarkers for LCS will require testing in newly eligible populations. The reduction in minimum age and smoking history in the 2020 USPSTF LCS guidelines expands eligibility particularly for under-represented minority populations, who have historically accounted for very small proportions of participants in earlier LCS trials (6).

Radiomics

Not covered in this review is the emerging field of imaging-based deep learning, often referred to as radiomics. Radiomics has been successfully utilized to classify benign and malignant lung nodules, predict tumor growth, and stratify patients based on risk. Recent studies have shown that a radiomic-based prediction model had a higher PPV, lower false-positive rate, and overall higher prediction value than thoracic radiologist readings alone (36). Indeed, integrated prediction models including clinical, biomarker, and radiographic characteristics are under study and may improve noninvasive diagnostic accuracy among high-risk patients. For example, such models had a higher sensitivity and specificity than serum miRNA biomarkers alone among smokers (37). Another biomarker model including clinical variables, serum CYFRA 21-1 level, and a radiomic signature improved diagnostic accuracy over the Mayo Clinic model, thus reducing the rate of invasive procedure from 63% to 51% among intermediate-risk individuals (38).

Generalizability

The use of noninvasive biomarkers as either a substitute or adjunct to LDCT has the potential to optimize LCS for minority and vulnerable populations, who often have limited access to screening facilities and are at higher risk for developing and dying from lung cancer (6). However, the applicability of these emerging tests must be validated among diverse cohorts. Although many biologic and socioeconomic factors—such as diet, lifestyle, ethnicity, and environmental exposures—can influence epigenetic markers, most early detection DNA methylation biomarker studies to date have included homogeneous populations or have not been controlled for race. Concerted efforts to include racial and ethnic minorities in all stages of biomarker testing are essential.

Cost-Effectiveness and Feasibility

As new LCS biomarkers continue to emerge, consideration of cost-effectiveness and large-scale implementation becomes critical. A recent study found that LCS that incorporates a hypothetical diagnostic biomarker, in place of Lung Imaging Reporting and Data System (Lung-RADS) 3 (probably benign nodule) and 4A (suspicious nodule) guidelines, could improve cost-effectiveness if the biomarker had at least medium sensitivity, 90% specificity, and a cost of $250 or less, with an incremental cost-effectiveness ratio less than $100 000 per quality-adjusted life-year gained (39). A Chinese population-based model found that a strategy incorporating LDCT and the 13-miRNA signature test among individuals with a ≥20-pack-years smoking history was more cost-effective than LDCT alone (40). Although such analyses have clear limitations—such as assuming 100% uptake and adherence—they provide early insights into economic considerations.

Conclusions

Although LDCT-based LCS can decrease lung cancer-related mortality among high-risk individuals, it remains an imperfect and underutilized process. To address these limitations, multiple potential biomarkers are under active investigation to improve early lung cancer detection. These biomarkers tend to fall into one of 2 categories: (a) improving risk-stratification after lung nodule detection; or (b) optimizing risk-based patient selection for imaging. Blood-based biomarkers, focusing on a range of protein, miRNA, circulating DNA, and methylated DNA panels, are furthest in clinical development. Invasive strategies, such as gene expression profiles of airway epithelia, and noninvasive strategies utilizing breath, sputum, and urine are also under study. Ultimately, optimized LCS may integrate a combination of molecular biomarkers, imaging findings, and clinical features to increase LCS accuracy and accessibility. Such changes to clinical practice will require high-quality, prospective data from diverse populations and ongoing clinician education regarding the importance of early lung cancer detection.

Nonstandard Abbreviations

LDCT, low dose computed tomography; LCS, lung cancer screening; USPSTF, United States Preventive Task Force; miRNAs, microRNAs; EarlyCDT-Lung, Early Cancer Detection Test; NPV, negative predictive value; PET, positron emission tomographic; NSCLC, non-small cell lung cancer; CEA, carcinoembryonic antigen; PPV positive predictive value; cfDNA cell-free DNA; DELFI, DNA evaluation of fragments for early interception; VOCs, volatile organic compounds; GC/MS, gas chromatography and mass spectrometry; TCPP, tetra (4-carboxyphenyl) porphyrin.

Author Contributions

The corresponding author takes full responsibility that all authors on this publication have met the following required criteria of eligibility for authorship: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; (c) final approval of the published article; and (d) agreement to be accountable for all aspects of the article thus ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved. Nobody who qualifies for authorship has been omitted from the list.

Sheena Bhalla (conceptualization—equal, data curation—equal, formal analysis—equal, investigation—equal, visualization—equal, writing—original draft—lead, writing—review and editing—equal), Sofia Yi (conceptualization—equal, data curation—equal, formal analysis—equal, investigation—equal, visualization—equal, writing—original draft—supporting, writing—review and editing—equal), and David Gerber (conceptualization—equal, funding acquisition—equal, investigation—equal, visualization—equal, writing—review and editing—equal).

Authors’ Disclosures or Potential Conflicts of Interest

Upon manuscript submission, all authors completed the author disclosure form.

Research Funding

Funded in part by the Cancer Prevention and Research Institute of Texas (CPRIT; RP160030, PP190052, PP230041).

Disclosures

S. Bhalla has served in a consulting/advisory role for Takeda, Mirati, and Merus. D.E. Gerber has research funding from Astra-Zeneca, BerGenBio, Karyopharm, Novocure, Sagimet; has served in a consulting/advisory role for BeiGene, Catalyst, Daiichi-Sankyo, Elevation Oncology, Jansen, Mirati, Regeneron, Sanofi; is a shareholder in Gilead, Medtronic, Walgreens; has pending U.S. patents 16/487,335, 17/045,482, 63/386,387, 63/382,972, 63/382,257; is co-founder and Chief Scientific Officer, OncoSeer Diagnostics, LLC.

References