Optimizing Lung Cancer Screening With Risk Prediction: Current Challenges and the Emerging Role of Biomarkers

Abstract

The Oncology Grand Rounds series is designed to place original reports published in the Journal into clinical context. A case presentation is followed by a description of diagnostic and management challenges, a review of the relevant literature, and a summary of the authors' suggested management approaches. The goal of this series is to help readers better understand how to apply the results of key studies, including those published in Journal of Clinical Oncology, to patients seen in their own clinical practice.

Lung cancer screening has been demonstrated to reduce lung cancer mortality, but its benefits must be weighed against the potential harms of unnecessary procedures, false-positive radiological findings, and overdiagnosis. Individuals at highest risk of lung cancer are more likely to maximize benefits while minimizing harm from screening. Although current lung cancer screening guidelines recommended by the US Preventive Services Task Force (USPSTF) only consider age and smoking history for screening eligibility, National Comprehensive Cancer Network and other society guidelines recommend screening on the basis of individualized risk assessment including family history, environmental exposures, and presence of chronic lung disease. Risk prediction models have been developed to integrate various risk factors into an individualized risk prediction score. Previous evidence showed that risk prediction model-based screening eligibility could improve sensitivity for detecting lung cancer cases without reducing specificity. Furthermore, recent advances in lung cancer biomarkers have enhanced the performance of risk prediction in identifying lung cancer cases relative to the USPSTF criteria. These risk prediction models can be used to guide shared decision-making discussions before proceeding with lung cancer screening. This study aims to provide a concise overview of these prediction models and the emerging role of biomarker testing in risk prediction to facilitate conversations with patients. The goal was to assist clinicians in assessing individual patient risk, leading to more informed decision making.

CASE PRESENTATION

A 66-year-old sheep farmer and former smoker presents to you for a routine maintenance examination. He identifies as non-Hispanic White. His brother died of lung cancer, and he wishes to know whether he should have lung cancer screening. The US Preventive Services Task Force (USPSTF) criteria for lung cancer screening include smoking pack-years and years since smoking cessation (for former smokers). The patient smoked one pack per day for 21 years (ie, 21 smoking pack-years) and quit smoking 13 years ago. Seven years ago, he underwent surgery and adjuvant chemotherapy for stage IIA colon cancer, without recurrence. You engage in shared decision making and discuss the risks and benefits of lung cancer screening. The patient is apprehensive about needing a biopsy for suspicious findings since his surgery was complicated, and he does not want unnecessary procedures. However, he wishes to know if his personal and family history of cancer places him at a higher chance of getting lung cancer that would outweigh his concerns about procedural complications.

CLINICAL CHALLENGES IN EVALUATION AND RELEVANT LITERATURE

Challenges in Evaluation

Lung cancer remains the leading cause of cancer mortality with 5-year survival of 23% in the United States. This poor survival largely reflects that most lung cancer is diagnosed at advanced or metastatic stages.¹ Low-dose computed tomography (LDCT) screening for lung cancer has been demonstrated to reduce mortality by detecting lung cancer at an earlier, more treatable stage. In the National Lung Screening Trial (NLST), LDCT was associated with a 20% reduction in lung cancer mortality compared with chest X-ray.² On the basis of these findings, the USPSTF established the first national screening guidelines in 2013, which recommended annual LDCT screening for asymptomatic individuals age 55 to 80 years who have smoked for at least 30 pack-years and have been actively smoking for the previous 15 years.³ These guidelines were updated in 2021 to lower the starting age from 55 to 50 years and the minimum cumulative smoking exposure from 30 to 20 pack-years relative to its 2013 recommendation.⁴

Although lung cancer screening offers benefits, it also carries potential harms, including radiation exposure, high false-positive rates, and overdiagnosis. Diagnosing lung cancer typically involves an invasive biopsy, which can lead to procedural complications such as pneumothorax. Furthermore, treating confirmed lung cancer could potentially lead to morbid complications, especially in patients with other health concerns. Because of these tradeoffs, informed, shared decision-making between patients and providers is a crucial component of lung cancer screening. The Centers for Medicare and Medicaid Services require shared decision making for reimbursement,⁵ and multiple society guidelines recommend discussing the benefits and harms of screening on the basis of individual risk factors.^6,7 Although smoking history and age are two important risk factors for lung cancer that are taken into account by the USPSTF guidelines, the amount of lung cancer risk still differs across eligible individuals because several known risk factors are not currently considered in the guidelines. For example, a previous study showed that 90% of the mortality benefit observed in NLST could have been achieved by screening the highest 60th risk percentile calculated using a risk prediction model that includes additional risk factors for lung cancer.⁸

Risk stratification and individualized risk assessment could improve the performance of lung cancer screening by identifying high-risk individuals who may benefit from screening. Multiple risk prediction models have been proposed, incorporating various risk factors into a single risk score.⁹ Risk model-based eligibility for lung cancer screening has been shown to outperform the USPSTF criteria in identifying high-risk individuals, with improved sensitivity, larger mortality benefits, and higher cost-effectiveness.^9,10 However, there are several challenges to incorporating risk factors in shared decision making in clinical practice, which include defining key risk factors, choosing models to use that integrate these factors into an individual risk assessment, and incorporating risk assessment into clinical workflow.

Summary of Relevant Literature

Several lung cancer risk prediction models incorporate demographic, clinical, and environmental risk factors for lung cancer (Table 1).^11-13 The dominant risk factors for lung cancer include smoking history, male sex, African American race, age, personal and family history of cancer, exposure to asbestos and radon, and the presence of chronic lung disease. Various combinations of these factors are used in different models, many of which have been externally validated in independent data sets (Table 2). To inform the 2021 USPSTF recommendation update, the Cancer Intervention and Surveillance Modeling Network (CISNET) modeling group compared the performance of three major risk prediction models^11-13—the PLCO_m2012 model,¹¹ LCDRAT,¹² and Bach¹³—versus the current 2021 USPSTF guidelines in a simulated US birth cohort.¹⁵ The modeling study showed that risk model-based screening strategies would reduce lung cancer mortality more effectively than the 2021 USPSTF criteria but increased overdiagnosis (Table 2).

TABLE 1.

Risk Factors Included in Selected Major Models

TABLE 2.

Evaluation of Selected Major Models Versus Current Standard for Screening Eligibility by USPSTF

Biomarkers, which include DNA, proteins, and hormone levels, could potentially be useful in the early detection of cancer^9,16 and may serve as a useful adjunct to risk prediction models on the basis of clinical factors. However, most biomarker risk prediction models for lung cancer lack validation.⁹

In companion to this article, Irajizad et al¹⁴ showed how incorporating a panel of circulating protein biomarkers can improve the prediction of lung cancer mortality. In their previous work, the authors conducted a proof-of-principle case-control pilot (INTEGRAL) to show that integrating a blood-based panel of circulating protein biomarkers (called 4MP) could improve the detection of patients with lung cancer compared with a smoking history–based risk model alone.¹⁷ The 4MP biomarker panel includes surfactant protein B, cancer antigen 125, cytokeratin-19 fragment, and carcinoembryonic antigen. By expanding their approach, they demonstrated that incorporating the 4MP circulating biomarker panel into an established risk prediction model (PLCO_m2012) enhances performance with respect to predicting lung cancer mortality. This evaluation was based on the model they had previously developed (called 4MP + PLCO_m2012) for predicting lung cancer risk¹⁸; this model combines the 4MP biomarker panel and the PLCO_m2012 model that includes age, smoking, race/ethnicity, body mass index, education, presence of chronic lung disease, personal history of cancer, and family history of lung cancer. Prediagnostic case and noncase sera from the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial cohort were used to train the 4MP + PLCO_m2012 model for lung cancer risk¹⁸ and to validate the model for lung cancer mortality.¹⁴ Of the 552 lung cancer cases, 387 (70%) died of lung cancer. The combined 4MP + PLCO_m2012 model had improved sensitivity, specificity, and positive predictive value for lung cancer death compared with the USPSTF 2021 criteria.¹⁴

An ideal biomarker-based risk stratification may be implemented independently, without relying on other existing risk prediction models; this is how prostate-specific antigen is used in prostate cancer screening. This may simplify the future implementation of risk-based screening as interpreting a single blood test may be easier than collecting multiple clinical data elements from patient histories and entering them into a calculator to obtain a risk score. However, given the significance of smoking as a risk factor for lung cancer, sole reliance on biomarkers for risk prediction will not be as accurate as incorporating biomarkers with other key risk factors, such as smoking.

Key Remaining Questions for Risk Stratification in Lung Cancer Screening

Implementation

Ideally, individuals would decide on screening on the basis of an understanding of the risks and benefits of screening and probabilities of cancer. Such expertise, however, is not often accessible in clinics nor among the general population, and historically, decision aids on the basis of risk prediction models are infrequently used in clinical practice. Conflicting priorities, time constraints, trouble obtaining decision aids, poor patient comprehension, and anticipated patient emotions have all been cited as factors that limit the use of decision aids.⁷ One major barrier has been the need for additional data elements beyond than those required by the USPSTF. For instance, calculating smoking pack-years alone has proven a significant barrier to uptake.¹ Multiple efforts are underway to determine the best risk prediction implementation strategy.¹ Finally, although prediction tools can assist physicians in identifying patients for screening, they do not help address other barriers, such as challenges in follow-up care, limited access to imaging facilities, and competing priorities during patient visits. Therefore, prediction tools alone are unlikely to address the challenge of low screening uptake.

Competing Comorbidities

Risk-based screening tends to screen individuals at older ages with smoking-related comorbidities, but the life expectancies of these groups is likely different. Heavy smokers with smoking-related comorbidities, such as chronic obstructive lung disease, may die of a smoking-related death before developing lung cancer. Therefore, evaluating optimal risk-based screening strategies requires examining more comprehensive metrics, such as life-years gained versus the number of lung cancer averted, which takes into account differential life expectancies across individuals screened. In the CISNET modeling study, risk model-based screening strategies at a threshold selected for similar sensitivity as the 2021 USPSTF criteria (ie, 6-year threshold of ≥1.0 using PLCO_m2012) still resulted in higher life-year gain compared with the 2021 USPSTF criteria (Table 2).¹⁵

Moreover, patients with chronic lung disease are also at higher risk of complications from biopsies and subsequent lung cancer treatment. Thus, clinical judgment should also be used to evaluate competing comorbidities and risks of complications and deaths.

Risk Threshold

The effectiveness of screening could depend on the risk threshold used to select people for screening.¹⁰ However, it is unclear at what risk threshold screening should be recommended. At a population level, multiple considerations, such as cost-effectiveness and population smoking prevalence, may drive threshold selection. At a local level, considerations related to resource allocation may also play an important role. Access to appropriate screening and reliable follow-up is essential to realize the benefits of screening, but because of resource limitations, high-risk patients may need to be given priority among the eligible population. At an individual level, patient preferences, competing comorbidities, and risk tolerance may drive personal risk threshold.

Time Point for Risk Prediction

Risk assessment can be used at multiple time points along the screening timeline: before screening (to evaluate eligibility), workup of screening findings (to evaluate the likelihood of malignancy of a nodule identified through LDCT), and decision for intervention (to evaluate the aggressiveness of malignancy). For example, an indolent lesion may be monitored, whereas a fast-growing lesion requires intervention. Although assessing lung cancer risk can help evaluate the likelihood of malignancy, it may not account for the aggressiveness of malignancy. However, estimating lung cancer mortality⁸ could address the aggressiveness of lung cancer. It is unknown at which time point risk stratification will be most effective or if risk stratification should be used at multiple points in the screening timeline.

Equity

Risk prediction models can identify high-risk individuals who are currently ineligible for screening, such as Black individuals with low cumulative smoking exposure. However, the factors contributing to lung cancer risk in individuals with no smoking history are not fully understood, and further research is ongoing to evaluate the potential of LDCT screening for Asian never-smokers.¹⁹ This raises important equity considerations, as currently, these USPSTF-ineligible individuals would need to cover the cost of screening themselves

Available Prospective and Clinical Trial Validation

Trials in the United Kingdom, Australia, and Canada are prospectively enrolling participants for lung cancer screening on the basis of eligibility using the PLCO_m2012 risk prediction model.⁴ However, it is impractical to assess all the risk prediction models available because of the time frame and scale required for randomized controlled trials to demonstrate the definitive benefits of screening (such as NLST). Currently, the combination of randomized controlled trials and modeling simulations are used for policy decisions, but there is an emerging role for alternative epidemiologic and real-world evidence designs.

OUR APPROACH TO MANAGEMENT

The individual in our case had originally presented to his primary care provider (PCP) in 2014 and at that time, did not meet lung cancer screening criteria according to the 2013 USPSTF guidelines because of his lower cumulative smoking exposure (21 pack-years v 30 pack-years). Later that year, the patient was diagnosed with stage IIB squamous cell carcinoma and underwent a lobectomy as an initial primary lung cancer treatment. The patient was compliant with routine surveillance but stopped scans in 2020, 6 years after his initial diagnosis, because he was deemed to be no longer at risk for recurrence. In 2022, he presented with hemoptysis, and a second primary squamous cell carcinoma was discovered in the left mainstem bronchus (stage IB). Because of the location, he was considered inoperable. He underwent hypofractionated radiation to the left hilum and tolerated the treatment well.

Although risk prediction models were not used at the time of our patient's initial presentation, it is instructive to work through what his calculated risk would have been. According to the 2013 USPSTF criteria, the patient was ineligible for lung cancer screening at the time of his initial PCP visit but would have met the 2021 USPSTF criteria. Using the PLCO_m2012 model,¹¹ his 6-year lung cancer risk was 3.1% at the time of his PCP visit (Table 3). Given his family history and personal history of cancer, this risk assessment shows he was at higher risk for lung cancer than his smoking history alone would suggest. Notably, the PLCO_m2012 6-year risk threshold of 1%-1.1% yielded similar screening coverage as the 2021 USPSTF criteria in the 1960 US birth cohort,^10,15 and his PLCO_m2012-calculated risk was above the threshold. Additional biomarker testing results (eg, 4MP), in addition to PLCO_m2012, could have changed his risk category and informed a more detailed discussion about the risk-benefit of lung cancer screening.

TABLE 3.

PLCO_m2012¹¹ Risk Calculation for Case Presentation

It is noted that he had a history of lung cancer at the time of the diagnosis of his second primary lung cancer (SPLC) in 2022, making him ineligible for screening under the current National Comprehensive Cancer Network guidelines, which have separate recommendations for lung cancer survivors.⁶ We previously developed a risk prediction model for SPLC for lung cancer survivors, called SPLC-RAT,²¹ and applied it to this patient. The SPLC-RAT model yielded a 5-year risk of 4.9% and a 10-year risk of 8.7% when assessed at the time of his initial diagnosis in 2014. This model can be used to specifically counsel lung cancer survivors about their risk of developing another lung cancer and could have informed this patient about whether to continue screening after surveillance for recurrence ended.

At Stanford, we have a dedicated lung cancer screening program equipped with sufficient imaging equipment to screen eligible individuals and supported by a multidisciplinary team and staff. Such infrastructure requires investment and resources that may not be available in all settings, such as an advanced practice provider dedicated to screening and subspecialty support. Our patients have two options for undergoing lung cancer screening. Either their primary care physician or another provider can directly order the scan, or individuals can be referred to our dedicated screening clinic. Within the clinic, an advanced practice provider is responsible for verifying eligibility, conducting the shared decision-making visit, providing smoking cessation counseling, ordering the scan, coordinating with a multidisciplinary team to manage the results, and ensuring appropriate follow-up on the basis of any findings.

The USPSTF guidelines are currently used to decide who qualifies for lung cancer screening. Although risk prediction models are used for eligibility determination in pilot implementation studies, they are not yet used for assessing screening eligibility in most centers, including ours. Biomarker testing for lung cancer screening requires further validation in prospective studies before clinical implementation.

Currently, the main clinical use of risk prediction models is in the context of shared decision making (Fig 1).⁵ Risk prediction models can help inform shared decision making between the provider and the patient by providing the patient with their personal likelihood of benefit. The relative weight of screening benefits versus harms will subsequently be determined by clinical judgment and patient preferences. To fully reap the benefit of lung cancer screening, it must be carried out reliably in a setting with a patient who is both fit enough to receive treatment and to adhere to follow-up. While it remains to be seen how risk prediction models and biomarkers will ultimately be incorporated into lung cancer screening, it is evident that they have great potential to improve personalized shared decision making.

FIG 1.

Clinical flowchart of an initial lung cancer screening visit at Stanford. APP, advanced practice provider; LC, lung cancer; LCS, lung cancer screening; PCP, primary care provider.

AUTHOR CONTRIBUTIONS

Conception and design: All authors

Collection and assembly of data: Julie Tsu-yu Wu, Summer S. Han

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Optimizing Lung Cancer Screening With Risk Prediction: Current Challenges and the Emerging Role of Biomarkers

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).