Clinical Adjuncts to Lung Cancer Screening: A Narrative Review

BACKGROUND

With the 2021 update of the U.S. Preventive Services Task Force (USPSTF) lung cancer screening (LCS) guidelines, the population eligible for LCS with low-dose computed tomography (LDCT) expanded by 87%, enabling more women and racial/ethnic minority individuals to benefit from screening.^1,2 The guidelines recommend screening based on age (50-80 years) and smoking history (≥20 pack-years and smoked within the past 15 years).² However, among the screening-eligible population, individual risk of lung cancer is highly heterogeneous, and the mortality benefit associated with screening varies based on lung cancer risk.³ Without considering other factors, including comorbidities and life expectancy, this approach inherently includes individuals who are unlikely to benefit from screening and excludes individuals who are more likely to benefit from screening.⁴ An alternate approach is to tailor screening based on more personalized risk assessment to improve the balance of benefits to harms.

In this context, clinical adjuncts, or additional information and tools to guide clinical decision-making, may optimize LCS effectiveness and efficiency. Proposed adjunctive approaches integrate clinical history, risk prediction models, shared decision-making (SDM) tools, and biomarker tests, with varying applicability to current practice. These clinical adjuncts can guide key steps in LCS, including selecting screening candidates, supporting SDM, evaluating screening-detected nodules for cancer, and determining screening intervals (Figure 1). Herein, we review the most promising clinical adjuncts and highlight where in the screening process they may be most useful.

Figure 1.

Potential Use of Clinical Adjuncts at Key Steps in Lung Cancer Screening

DISCUSSION

Clinical History

Perhaps most notably missing from LCS guidelines is consideration of family history and medical history. Family history of lung cancer is known to increase risk for the disease, independent of smoking history.⁵ However, family history in the clinical context tends to include only first-degree relatives and often suffers from accurate recall, leading to less precise calculation of risk. In a study estimating lung cancer risk based on complete family history from linked statewide cancer registry data and genealogy records, individuals at two to five-fold higher risk for lung cancer were identified.⁶ Therefore, with additional family history data, better risk calculations could be performed to inform decisions about LCS.

Beyond recommending against screening individuals with a health condition that substantially limits life expectancy or the ability to undergo curative lung surgery, the USPSTF provides little guidance on screening individuals with comorbitidies.² Chronic obstructive pulmonary disease (COPD) is a major risk factor for lung cancer, noted in 60-85% of cases.⁷ Screening-eligible individuals at greatest risk for lung cancer have the highest prevalence of COPD and a higher risk of dying from other causes, due to a high prevalence of cardiovascular and other chronic diseases.^7,8 Accordingly, those with COPD at utmost lung cancer risk may not actually benefit from LCS. In a National Lung Screening Trial (NLST) sub-analysis, the benefit of annual LCS was greatest in those with normal lung function or mild-to-moderate COPD, not in those with severe or very severe COPD.⁸ A limitation of using degree of COPD severity to recommend LCS is that individuals would need spirometry prior to LDCT. However, routine in-office spirometry may help to identify those with airflow limitation who are more likely to benefit from LCS.

Other lung-specific comorbidities, including emphysema and interstitial lung disease, also confer lung cancer risk.^9,10 Even non-primary lung disease, such as peripheral arterial disease, has been associated with increased lung cancer risk, likely because smoking history is a shared risk factor. The extent to which existing comorbidities may be incorporated into estimating benefits and harms of LCS is relatively unknown, warranting further investigation.¹¹

Risk Prediction Models

Proposed risk prediction models estimate the probability of developing or dying from lung cancer or the probability of pulmonary nodule malignancy, based on age, smoking history, and other clinical and non-clinical factors. These models can be applied at each step in Figure 1 to optimize screening efficiency and outcomes.

To be clinically useful, a risk prediction model should be accurate, reliable, and generalizable to its target population. A model may underperform when applied to populations independent from which it was developed, because of shortcomings in its development or variation in population composition or predictor variable measurement.¹² External validation of a model’s predictive performance is therefore necessary before it is used clinically. Predictive performance is commonly evaluated by discrimination and calibration. Discrimination, often quantified by the area under the curve (AUC), indicates the ability of a model to accurately classify individuals with or without the event of interest, while calibration indicates the closeness between the probabilities of model-predicted or expected (E) versus observed (O) events. Even if a model discriminates well (high AUC), it is impractical if it is insufficiently calibrated (E/O ratio far below or above 1.0). Accordingly, we focus on models that have been externally validated in independent populations.

Predicting lung cancer risk before screening.

Applying risk prediction models to select high-risk individuals for LCS could improve the balance of benefits to harms by screening fewer individuals, discovering fewer false-positive results, and detecting more early-stage lung cancers. Over 30 distinct models have been proposed to predict individual risk of developing or dying from lung cancer within a specified timeframe.¹³ The majority predict risk based on age, smoking history, and other conventional risk factors, including sex, family history of lung cancer, and COPD or emphysema. Some incorporate additional factors requiring clinical assessment (e.g., lung function measures) or biospecimen collection and analysis (e.g., molecular and genetic markers).

To date, less than half of these models have been externally validated, primarily those incorporating only conventional risk factors, of which five have performed relatively well in US, European, and Australian cohorts (Table 1).^14-21 Four of these models are derived from US trial populations, specifically the Bach model in the Carotene and Retinol Efficacy Trial (CARET) population of high-risk smokers and the PLCO_M2012 model, Lung Cancer Risk Assessment Tool (LCRAT), and Lung Cancer Death Risk Assessment Tool (LCDRAT) in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Screening Trial and NLST populations of ever-smokers. The Bach model is the simplest, incorporating age, sex, asbestos exposure, and smoking intensity, duration and quit-years to estimate annual lung cancer incidence.²² The PLCO_M2012 model incorporates age, race/ethnicity, education, body mass index (BMI), COPD, history of cancer, family history of lung cancer, and smoking status, intensity, duration, and quit-years to predict six-year lung cancer incidence.²³ LCRAT and LCDRAT include the same set of predictors (age, sex, education, race/ethnicity, BMI, emphysema, family history of lung cancer, and smoking intensity, duration, and quit-years) to estimate five-year lung cancer incidence or mortality, respectively.²⁴ In contrast, the Liverpool Lung Project (LLP) model was derived from a matched case-control study population in the United Kingdom, including never-smoking individuals, to predict five-year lung cancer incidence based on smoking duration, asbestos exposure, history of pneumonia, history of cancer, and family history of lung cancer.²⁵ It has been since updated (LLP_v2) to incorporate history of other respiratory diseases and limited to ever-smoking individuals.²⁶

Table 1.

Externally validated top-performing models predicting risk of lung cancer incidence or mortality

Model [reference]	Location Data Source(s)	Outcome	Target Population	Predictors	Predictive Performance
Bach [22]	United States CARET: 18,172 high-risk smokers	Lung cancer incidence in 1 year	Ever-smoking adults	Age; sex; smoking intensity, duration, and quit-years; asbestos exposure	AUC: 0.72 Calibration assessed graphically
PLCOM2012 [23]	United States Development: - 36,286 PLCO control arm smokers ages 55-74 Validation: - 37,332 PLCO intervention arm smokers ages 55-74 - 51,033 NLST participants	Lung cancer incidence in 6 years	Ever-smoking adults	Age; education; race/ethnicity; BMI; smoking status, intensity, duration, and quit-years; COPD; history of cancer; family history of lung cancer	AUC: 0.797 (PLCO), 0.701 (NLST) Calibration: 90th percentile absolute error=0.042
LCRAT [24]	United States Development: - 39,180 PLCO control arm smokers ages 55-74 Validation: - 39,822 PLCO intervention arm smokers ages 55-74 - 26,554 NLST control arm participants	Lung cancer incidence in 5 years	Ever-smoking adults	Age; sex; education; race/ethnicity; BMI; smoking intensity, duration, pack-years, and quit years; emphysema; family history of lung cancer	AUC: 0.80 (PLCO), 0.70 (NLST) E/O: 0.94 (PLCO), 1.06 (NLST)
LCDRAT [24]	United States Development: - 39,180 PLCO control arm smokers ages 55-74 Validation: - 39,822 PLCO intervention arm smokers ages 55-74 - 26,554 NLST control arm participants - 29,091 NHIS smokers ages 50-80	Lung cancer mortality in 5 years	Ever-smoking adults	Age; sex; education; race/ethnicity; BMI; smoking intensity, duration, pack-years, and quit years; emphysema; family history of lung cancer	AUC: 0.81 (PLCO), 0.73 (NLST), 0.78 (NHIS) E/O: 1.08 (PLCO), 1.31 (NLST), 0.94 (NHIS)
LLP [25]	United Kingdom Age- and sex-matched case-control study: 579 lung cancer cases, 1157 controls	Lung cancer incidence in 5 years	General population	Smoking duration; asbestos exposure; history of pneumonia; history of cancer; family history of lung cancer	AUC: 0.70 Calibration not assessed
LLPv2 [26]	United Kingdom Age- and sex-matched case-control study: 579 lung cancer cases, 1157 controls	Lung cancer incidence in 5 years	Ever-smoking adults	Smoking duration; asbestos exposure; history of pneumonia; history of emphysema; history of bronchitis; history of tuberculosis; history of COPD; history of cancer; family history of lung cancer	Not reported

Using well-validated risk prediction models to select ever-smoking individuals for LCS is more effective in preventing lung cancer deaths compared to USPSTF screening criteria.^23,24 In LCS trials outside the US, the PLCO_M2012 and LLP_v2 models have and are being applied to optimize selection of high-risk individuals.^26-28 Applying risk prediction models may additionally reduce lung cancer disparities. In retrospective analyses, the PLCO_M2012 model had a higher sensitivity for detecting lung cancer, especially among African Americans and women, over USPSTF LCS criteria.^29,30

From a cost-effectiveness perspective, however, the value of risk-based screening appears modest.³¹ Despite efficiency in averting more lung cancer deaths per person screened, risk-based screening preferentially selects those at highest risk, who are generally older, smoke heavier, and have more comorbidities. Those at highest risk, in turn, are costlier to screen and may benefit less due to shorter life expectancy. To address these concerns, novel strategies that consider estimated gains in life expectancy and individual preferences have been proposed.^32-34 While screening high-risk individuals with long life expectancy is preferable, the minimum gain in life expectancy by which to recommend screening remains less clear.

Predicting nodule malignancy risk.

Using risk prediction models that accurately distinguish malignant from benign pulmonary nodules may facilitate earlier lung cancer diagnosis and treatment and reduce harms and costs from unnecessary follow-up. Currently, the Lung CT Screening Reporting and Data System (Lung-RADS^®) is used in the US to standardize interpretation and management of LDCT screening results.³⁵ Results are classified visually into assessment categories that define whether a screening exam is negative (category 1 or 2) or positive (category 3 or 4A/B/X).

Table 2 presents the most commonly validated models for predicting pulmonary nodule malignancy. These models are derived from patient populations across diverse settings and malignancy prevalence and incorporate different combinations of patient and imaging characteristics. The Mayo Clinic and VA models were originally constructed to estimate the pre-test probability of malignancy for solitary pulmonary nodules (SPN) detected incidentally by chest radiography.^36,37 The Peking University People’s Hospital (PKUPH) model was later developed and shown to predict nodule malignancy risk more accurately than the Mayo Clinic and VA models among Chinese patients.³⁸ In comparison, the Brock (Pan-Can) model is the only one developed using trial data to estimate malignancy risk of screen-detected pulmonary nodules among ever-smoking patients.³⁹

Table 2.

Externally validated models predicting risk of pulmonary nodule malignancy

Model [reference]	Location Data Source(s)	Malignancy prevalence	Predictors	Predictive Performance
Mayo [36]	United States 629 Mayo Clinic patients with indeterminate 4-30 mm SPN found by chest radiography	23%	Age; smoking status; history of extrathoracic cancer; nodule diameter, spiculation, location	AUC: 0.80 Hosmer-Lemeshow goodness of fit, p=0.62
VA [37]	United States 375 VA patients with 7-30 mm SPN found by chest radiography	54%	Age; smoking status and quit-years; nodule diameter	AUC: 0.78 Hosmer-Lemeshow goodness of fit, p=0.61
PKUPH [38]	China Development: 371 patients with pathologically diagnosed SPN from 2000 to 2009 Validation: 62 patients with pathologically diagnosed SPN from 2009 to 2010	53%	Age; family history of cancer; nodule diameter, spiculation, border, calcification	AUC: 0.89 Calibration not assessed
Brock (PanCan) [39]	Canada Development: 1871 high-risk smokers with nodules from PanCan Study Validation: 1090 high-risk smokers with nodules from BCCA chemoprevention trials	PanCan: 5.5% BCCA: 3.7%	Age; sex; family history of lung cancer; emphysema; nodule diameter, spiculation, location, type, count	AUC >0.90 Calibration: 90th percentile absolute error=0.003

These models have generally performed less well than originally reported across independent external validation studies. In most studies comparing the Brock model to others in non-screening populations, the discriminatory accuracy of the Brock model has been greater or equivalent to that of the Mayo Clinic model, yet greater than the VA and PKUPH models.^40-44 In the largest and most recent evaluation, however, the Mayo Clinic model exhibited greater accuracy than the Brock model in discriminating malignancy risk of large (>8 mm) nodules.⁴⁵ Across studies that also assessed calibration, both models underestimated or overestimated the actual probability of malignancy.^42,43,45 In the only study to compare these four models in a screening population, the Brock model performed best, exhibiting excellent discrimination and acceptable calibration.⁴⁶ Yet, data are inconsistent on whether the Brock model outperforms Lung-RADS in discriminating malignancy risk for screening-detected nodules.^47,48 Overall, existing prediction models appear to be of limited utility in optimizing nodule management at present.

Predicting lung cancer risk to optimize screening intervals.

Several lung cancer risk prediction models have been extended by incorporating LDCT results to determine optimal screening intervals. Using COSMOS trial data, Maisonneuve et al. recalibrated and extended the Bach model to include lung nodule characteristics and emphysema identified at the baseline screening exam as predictors.⁴⁹ When externally validated in ever-smoking individuals receiving annual LCS for 10 years, this model showed accurate prediction of lung cancer for the first two screening rounds, but overprediction thereafter.⁵⁰ Using data from independent sets of NLST participants, Tammemägi et al. developed and validated the PLCO_M2012results model, which added Lung-RADS results to the PLCO_M2012 model.⁵¹ Also with NLST data, Robbins et al. developed the LCRAT+CT model to continuously predict short-term lung cancer risk following a negative or positive LDCT exam, as a function of pre-screening risk factors (LCRAT) and LDCT features;⁵² this model has not yet been externally validated. Applying risk prediction models to personalize follow-up based on LDCT results could increase LCS efficiency and cost-effectiveness, but this remains to be proven.

Further Considerations.

Although evidence suggests integrating risk prediction models into LCS can avert more lung cancer deaths and reduce unnecessary follow-up procedures and costs, additional considerations are important to recognize before implementation (Table 3).

Table 3.

Further Considerations for Clinical Implementation of Risk Prediction Models

Determining benefits and harms of using risk prediction models. Simulation studies indicate that risk prediction models may be more effective in determining screening eligibility than existing age and smoking history criteria; yet, empirical data are limited on whether a risk-tailored approach improves lung cancer detection and outcomes.2 Prospective studies enrolling individuals using risk prediction models will offer insight into the effectiveness and feasibility of this approach. The International Lung Screening Trial (ILST) is evaluating the accuracy of the PLCOM2012 model versus USPSTF eligibility criteria in detecting lung cancers and efficiency of the PanCan model versus Lung-RADS in managing screen-detected nodules.28 Based on interim and economic analyses of ILST data, the PLCOM2012 model appears to be more efficient and cost-effective than the 2013 USPSTF criteria in selecting screening candidates.53,54 The Yorkshire Lung Screening Trial has similar plans to compare outcomes between individuals selected using the PLCOM2012 model, LLPv2 model, and 2013 USPSTF criteria.27 Demonstrating clinical feasibility and acceptability. The USPSTF has called for studies to identify implementation barriers associated with risk-tailored screening, given uncertainty about how using risk prediction models could hinder LCS implementation in primary care.2 Model-based risk assessment is more complex, requiring information beyond age and smoking history. While yet unproven, automating risk model calculations within electronic health records may feasibly support personalized LCS programs.55 Models requiring any information from clinical or biomarker assessment, especially if not readily available or measurable, may thus be impractical, even if they demonstrate better predictive performance. Also as some models include race and socioeconomic status, their acceptability in calculating lung cancer risk and life expectancy should be examined.55 Identifying optimal thresholds. No consensus exists on the cost-effective threshold for personalized selection of ever-smoking individuals in LCS. The proposed PLCOM2012 risk threshold of 1.51% was based on where lung cancer mortality was consistently lower among NLST intervention versus control participants and not on cost-effectiveness.56 Since risk models appear to yield different absolute risk estimates for the same individual,17 optimal risk thresholds must be specified for each model. Broadening scope to never-smoking individuals. In Western countries, the potential application of personalized LCS has focused on ever-smoking individuals. To broaden LCS to the general population, more extensive research would be needed to determine the accuracy of risk assessment and cost-effectiveness of screening never-smoking individuals.57

Shared-Decision Making (SDM) Tools

Individual preferences must be respected, given the US Centers for Medicare and Medicaid Services coverage requirement for SDM before LDCT screening.³⁴ Conveying the risks, benefits, and uncertainties of LCS to individuals requires a complex SDM conversation. SDM tools vary in method of delivery, from classroom exercises, handouts, surveys, or web-based tools.^58,59 Overall, a systematic review of fifteen such tools demonstrated increased patient knowledge and decreased decisional conflict.⁵⁸

Web-based SDM tools are increasingly common. The important factors in determining if such tools will improve decision-making include ease of use, integration into the electronic medical record (EMR), and comprehensibility. LCSDesTool, one web-based tool in the VA system, provided usability but faltered in comprehensibility secondary to medical jargon.⁶⁰ Another tool used in the VA population, DecisionPrecision, suffered from poor EMR integration, worsening time constraints for primary care providers.⁶¹ Some tools have integrated lung cancer risk calculators, e.g., shouldiscreen.com that calculates risk with the PLCO_M2012 model, although individual risk calculated and depicted by these tools varies considerably.⁶² In a prospective randomized control trial, shouldiscreen.com was compared to Options Grids, a one-page summary table to compare the options, with frequently asked questions for each option. Though evidence suggested both tools facilitated a high-quality SDM process, Option Grids was associated with decreased decision regret and increased patient knowledge.⁶³ As these tools are integrated in clinical practice, their constant improvement and evaluation will serve to improve the quality of SDM conversations going forward.

Biomarkers

Significant efforts also focus on discovering and validating biomarkers as a complement to LDCT in identifying individuals who are more likely to benefit from LCS and discriminating benign from malignant nodules seen at imaging. Although numerous different biomarkers have been evaluated, their clinical utility remains insufficiently investigated. We highlight those biomarker tests used in clinical practice currently, as many others are at various stages of research and development.^64,65

Post-Screening Biomarkers

More biomarker tests have been validated and become commercially available in the diagnostic setting, namely in determining malignancy risk of indeterminate pulmonary nodules. One example is EarlyCDT-Lung (Onc-Immune). This blood test measures autoantibodies against p53, CAGE, NY-ESO-1, SOX2, GBU4-5, HuD, and MAGE-A4 to assess pulmonary nodule malignancy risk, although it has been evaluated for use in early detection of lung cancer.⁶⁶ In a double-blinded randomized trial comparing use of EarlyCDT-Lung followed by LDCT to usual care, EarlyCDT-Lung detected lung cancers at an earlier stage, but did not increase the frequency of lung cancer detection over two years.⁶⁷ Another protein-based assay is “PAULA’s” test (Protein Assays Utilizing Lung cancer Analytes), which measures CEA, CA-125, CYFRA 21-1, and NY-ESO-1 to detect lung cancer in high-risk individuals, was able to distinguish cases from controls with 77% sensitivity, 80% specificity, and 0.85 AUC in the independent validation phase.⁶⁸ While serum protein-based assays show great promise, further study is needed to determine their clinical validity.

Exhaled breath condensate (EBC) is another source to sample the airway specifically. EBC contains cells, DNA fragments, and volatile organic compounds. These compounds can be analyzed by mass spectrometry, nano-sensors, and colorimetric sensors. EBC analysis has been shown to be helpful in discerning between benign and malignant pulmonary nodules and predicting response to therapy, but has not demonstrated utility in determining who to screen.^69,70

Bronchoscopy samples are another avenue to detect genetic alterations caused by cigarette smoking damage to the respiratory tract epithelium. A 2007 study first demonstrated an 80-gene expression set measured from histologically normal bronchial airway brushings could differentiate smokers with and without lung cancer at 90% sensitivity and 84% specificity.⁷¹ In two large prospective multicenter trials of patients undergoing bronchoscopy for concern of lung cancer, this airway gene expression classifier showed similar sensitivity (88%-89%), which increased when combined with bronchoscopy (96%-98%), but lower specificity (47% in both).⁷² The test proved useful when bronchoscopy was nondiagnostic, as the classifier had a 91% negative predictive value in patients with an intermediate pre-bronchoscopy probability of cancer and a negative bronchoscopy. Compared to other tests, the Percepta Genomic Sequencing Classifier, introduced clinically in 2015 and updated in 2019, primarily assists with lung cancer risk stratification when bronchoscopy is inconclusive.^73,74 Its clinical importance will be elucidated even more with ongoing use, including in the LCS context.

Pre-Screening Biomarkers

There are also biomarker tests for determining lung cancer risk, but none are currently approved for clinical use. An avenue for increasing early lung cancer detection is analyzing circulating genetic material. Among the most promising is the use of cell-free DNA (cfDNA) to non-invasively detect lung cancer. One novel approach called DELFI (DNA evaluation of fragments for early interception) applies machine learning algorithms to identify genome-wide cfDNA fragmentation profiles associated with cancer.⁷⁵ A large, multisite prospective validation study (CASCADE-LUNG, NCT05306288) is currently evaluating its performance in detecting lung cancer among screening-eligible individuals.

Other biomarkers include microRNAs (miRNAs), a family of molecules that help regulate gene expression. Two different tests, the miRNA signature classifier and the Mi-R test, were shown to decrease LDCT-false positive rates by five-fold and four-fold (while maintaining specificity and sensitivity above 75%) respectively in two large retrospective studies from Italy.^76,77 Since malignant cells are likewise found in sputum, a sputum-based test (LuCED) has been developed that detects abnormal bronchial epithelial cells using a novel imaging technology followed by cytopathologist review.⁷⁸ Also as whole genome sequencing costs decline, more attention has shifted toward using genetics to risk stratify individuals using polygenic risk scores (PRS). In a recent analysis, trajectories of five-year and cumulative absolute risk for lung cancer varied between individuals at different PRS deciles, suggesting that genetic background could be used to more efficiently tailor LCS.⁷⁹

SUMMARY

Clinical adjuncts represent a promising opportunity to improve LCS effectiveness and efficiency by tailoring and guiding clinical decision-making. Adjunctive use of clinical history, risk prediction models, and SDM tools has been considered most in optimizing selection of screening candidates and SDM. While currently, biomarker tests are only available clinically for pulmonary nodule evaluation. Although many proposed adjuncts to LCS appear beneficial, further evidence is needed regarding their clinical utility and implementation to support their use in practice.

Key Points.

• The benefit-to-harm ratio associated with LCS varies considerably among screening-eligible individuals.

• Clinical adjuncts can be used to tailor and guide decision-making at key steps to improve LCS effectiveness and efficiency.

• Although many proposed adjuncts to LCS appear beneficial, further evidence is needed regarding their clinical utility and implementation to support their use in practice.

Synopsis.

The updated USPSTF guidelines on lung cancer screening has significantly expanded the population of screening-eligible adults, among whom the balance of benefits and harms associated with lung cancer screening varies considerably. Clinical adjuncts are additional information and tools that can guide decision-making to optimally screen individuals who are most likely to benefit. Proposed adjuncts include integration of clinical history, risk prediction models, shared decision-making tools, and biomarker tests at key steps in the screening process. Although evidence regarding their clinical utility and implementation is still evolving, they carry significant promise in optimizing screening effectiveness and efficiency for lung cancer.

CLINICAL CARE POINTS.

• Lung cancer risk is heterogeneous among screening-eligible individuals.

• The mortality benefit associated with LCS is greater among individuals at higher lung cancer risk. However, those at highest risk may not benefit from LCS, given lower life expectancy and higher mortality from other causes.

• Decisions to recommend LCS should assess lung cancer risk, life expectancy, and personal preferences. Well-validated risk prediction and SDM tools can support this process, although best practices have yet to be established.

• While biomarkers may be integrated with LDCT to improve selection of individuals who are likely to benefit from LCS and discrimination of indeterminate pulmonary nodules, biomarker tests are only presently available to aid with pulmonary nodule evaluation.

• USPSTF guidelines recommend LCS based on age and smoking history. Individuals identified using clinical adjuncts who do not meet USPSTF criteria may incur issues with LCS reimbursement.

Abbreviations:

AUC

area under the curve

BMI

body mass index

CARET

Beta-Carotene and Retinol Efficacy Trial

COPD

chronic obstructive pulmonary disease

E/O

expected to observed ratio

EMR

electronic medical record

ILST

International Lung Screening Trial

LDCT

low-dose computed tomography

LLP

Liverpool Lung Project

LCRAT

Lung Cancer Risk Assessment Tool

LCDRAT

Lung Cancer Death Risk Assessment Tool

Lung-RADS

Lung CT Screening Reporting and Data System

NLST

National Lung Screening Trial

PAD

peripheral arterial disease

PKUPH

Peking University People’s Hospital

PLCO

Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial

SDM

shared decision-making

SPN

solitary pulmonary nodule

USPSTF

United States Preventive Services Task Force