Key Points

Lung cancer is the leading cause of cancer deaths in the US.1 From 2013 to 2021, the US Preventive Services Task Force recommended lung cancer screening (LCS) via low-dose computed tomography for adults aged 55 to 80 years with a greater than 30 pack-year history who are current smokers or quit within 15 years.² The US Preventive Services Task Force additionally highlights the need to screen only those willing and able to undergo lung surgery,^2,3 and all national guidelines recommend against screening those with poor baseline health status, whether quantified as less than 10 years’ life expectancy (American College of Chest Physicians),⁴ “life limiting comorbid conditions” (American Cancer Society),⁵ or “as long as patient functional status and comorbidity allow consideration for curative intent therapy” (National Comprehensive Cancer Network).⁶ Although national guidelines are consistent in recommending that baseline health should be considered prior to screening, how to define poor health is unclear, and the American Thoracic Society recognizes that there is an urgent need for more research in this area.⁷

The mortality benefit of LCS is thought to be largely achieved through surgical resection of early-stage cancers to achieve a cure, or improve survival after adjuvant therapies.8^,9 Two pivotal randomized clinical trials were powered to detect a mortality benefit of LCS: the National Lung Screening Trial (NLST)⁸ in the US and the Nederlands-Leuvens Longkanker Screenings Onderzoek (NELSON) trial¹⁰ in Europe. The NLST excluded those who were not healthy enough to be considered surgical candidates per individualized clinical evaluation, and the NELSON trial excluded individuals who rated their health as moderate or bad or could not climb 2 flights of stairs, as a proxy of health.¹⁰ Self-rated health and ability to climb stairs are thus key metrics that are broadly available and can provide a point of comparison to general population data.

However, little is known about whether individuals in fairly good health are more or less likely to be screened than those in poor health. The general US population of screen-eligible individuals has a higher mortality rate than NLST participants.11 In fact, those in poor health may be paradoxically more likely to undergo LCS.¹² Individuals with chronic health conditions are more likely to engage in regular health visits, which has been associated with LCS.¹³ This raises concern that increased mortality associated with lung surgery and/or competing causes of death could attenuate the 16% mortality benefit associated with LCS.¹⁴

Another important consideration is that LCS may have better outcomes among non-Hispanic Black individuals. In secondary analyses of NLST trial data, non-Hispanic Black participants experienced a greater reduction in lung-cancer specific mortality (hazard ratio [HR] 0.61 [95% CI, 0.37-1.01] vs 0.86 [95% CI, 0.75-0.98] among non-Hispanic White participants) and all-cause mortality (HR, 0.81 [95% CI, 0.65-1.00] vs 0.95 [95% CI, 0.89-1.02]; P < .05).15 If LCS is less likely to reach non-Hispanic Black individuals, then racial disparities could further diminish the actual benefit of LCS by missing this important group. Racial disparities have been observed in single centers in Rhode Island,¹⁶ Pennsylvania,¹⁷ Indiana,¹⁸ and North Carolina,¹⁹ but population-level analyses have failed to show LCS disparities by race and ethnicity.^13,20–23 However, poor self-rated health and all-cause mortality are still associated with race and ethnicity in the US,^24,25 with non-Hispanic Black individuals experiencing a 20% higher all-cause mortality than non-Hispanic White individuals,²⁶ in large part due to structural racism.²⁷ If poor health is associated with increased screening, and individuals from historically minoritized racial and ethnic groups are more likely to be in poor health, then racial or ethnic disparities could be obscured unless comparisons account for health status. Indeed, prior population-based analyses were limited by small samples of individuals who were not non-Hispanic White, and did not account for health status.

Data for this cross-sectional, population-based study were drawn from the 2017 to 2020 data sets of BRFSS, an annual telephone survey of noninstitutionalized US adults conducted by the US Centers for Disease Control and Prevention.28 BRFSS conducts more than 400 000 interviews with residents of all 50 states, the District of Columbia, and 3 territories to ascertain behaviors associated with health and health service utilization. Core questions are asked of all respondents; states may opt to add additional topic-specific modules. LCS eligibility and use are included in an optional module that was conducted in 11 states in 2017 (Florida, Georgia, Kansas, Kentucky, Maine, Maryland, Missouri, Nevada, Oklahoma, Vermont, and Wyoming),²⁹ 8 states in 2018 (Delaware, Maryland, Maine, New Jersey, Oklahoma, South Dakota, Texas, and West Virginia),³⁰ 20 states in 2019 (Arizona, Idaho, Kansas, Kentucky, Maine, Maryland, Minnesota, Missouri, Montana, Nebraska, North Carolina, North Dakota, Oklahoma, Pennsylvania, Rhode Island, South Carolina, Utah, Vermont, West Virginia, and Wisconsin),³¹ and 5 states in 2020 (Delaware, Maine, New Jersey, North Dakota, and South Dakota),³² in total representing 28 unique states. We identified US adults eligible for LCS by age and smoking history according to 2013 to 2021 USPTF guidelines and limited the sample to individuals aged 55 to 79 years, with a greater than 30 pack-year smoking history, and current smoker or former smoker who quit within 15 years prior. Exclusion criteria were personal history of lung cancer and age less than 55 or equal to or greater than 80 years. Individuals with missing values for these variables were excluded from the analysis; among 155 562 respondents aged 55 to 79 years with no personal history of lung cancer, 92.7% had complete data on tobacco history to determine LCS eligibility.

This study used publicly available, deidentified data and, thus, was exempt from institutional review board review and the need for informed consent, in accordance with 45 CFR §46. Reporting conformed to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.³³

We chose covariates a priori on the basis of prior research. In general, cancer screening tests are associated with socioeconomic status and smoking behavior,34 which holds true for LCS.¹² Correlates of LCS also include health care access,²³ COPD, prior cancer,^13,22,35 state of residence,³⁶ receipt of prior vaccination,¹² demographic characteristics (eg, marital status and employment), and smoking history in pack-years.¹²

We conducted univariable analyses for variables of interest, calculating P values from t tests, χ² tests, or linear tests for trend as appropriate to evaluate differences in LCS prevalence. We used multivariable logistic regression to examine the association between LCS and self-rated health status and LCS and race and ethnicity. We created a model that included both self-rated health status and race and ethnicity to isolate the association between each of these variables and LCS. Finally, for the main analysis, we constructed a model with all the aforementioned variables plus functional limitations and comorbidities as covariates. We assessed multicollinearity using the variance inflation factor, with a conservative threshold of 5 indicating collinearity.37

We observed significant differences by demographic factors (Table 1). LCS was associated with older age, with a screening rate of 19.2% (1338 individuals; 95% CI, 16.5%-22.0%) among those aged 65 to 79 years vs 15.2% (910 individuals; 95% CI, 12.5%-17.8%) among those 55 to 64 years (P = .04). Screening was less likely among those with difficulty paying for medical care (200 individuals who had difficulty paying for health care; 10.1% [95% CI, 7.4%-12.8%] vs 2041 individuals who did not; 17.9% [95% CI, 15.8%-20.1%]; P < .001). Sex, marital status, education, and body mass index were not significantly associated with LCS.

Worsening self-rated health status was associated with screening, as seen in Table 1: screening prevalence was 25.2% (95% CI, 20.6%-29.9%) among those in poor health vs 7.6% (95% CI, 5.0%-10.3%) among those in excellent health (P < .001 for trend). Overall, 48.5% (95% CI, 42.2%-54.9%) of those screened were in fair or poor health. Unadjusted rates of screening increased between successively poorer health categories in an approximately linear manner (Figure, panel A). Non-Hispanic Black individuals and those of other races and ethnicities reported lower rates of screening compared with non-Hispanic White individuals, though these differences were not significant (12.0% [95% CI, 4.3%-19.7%] vs 17.5% [95% CI, 15.6%-19.5%]; P = .57) (Table 1).

In a logistic regression model adjusted for demographics and health behaviors associated with screening likelihood (age, sex, marital status, body mass index, education, smoking history in pack-years, insurance status, receipt of influenza vaccine in prior 12 months, or difficulty paying for medical care), self-rated health was linearly associated with LCS: each 1-step decline in health status, such as from good to fair health, was associated with a 36% (95% CI, 20%-55%) higher likelihood of LCS (Table 2, model 1). There was no significant association between race and ethnicity and LCS in an unadjusted model or in a model adjusted for the demographics and health behaviors but not health status (model 2). In a maximally adjusted model that included specific functional limitations and diagnoses, as well as self-rated health status and race and ethnicity (model 3), the association between self-rated health status and LCS was attenuated in magnitude but remained significant (adjusted odds ratio [aOR], 1.19; 95% CI, 1.03-1.38). This association was approximately linear across all categories of self-rated health (Figure, panel B). However, the association between race and ethnicity and LCS was more precise in this model, as shown in Table 2, with non-Hispanic Black individuals 53% less likely to report LCS than non-Hispanic White individuals (aOR, 0.47; 95% CI, 0.24-0.90). Those who were not non-Hispanic Black or non-Hispanic White had similar likelihood of LCS as non-Hispanic White individuals, although precision was poor (aOR, 1.03; 95% CI, 0.35-3.01).

No substantial collinearity between variables in model 3 was identified, with variance inflation factors for self-rated health status, difficulty walking, difficulty dressing, difficulty running errands alone, chronic kidney disease, arthritis, COPD, asthma, vascular disease, diabetes, and personal history of cancer ranging from 1.03 to 1.59, well below the accepted threshold of 5.37 In this model, COPD and personal cancer history persisted as independent factors significantly associated with LCS (COPD: aOR, 2.42 [95% CI, 1.81-3.24]; personal cancer history: aOR 2.35 [95% CI, 1.76-3.14]). In an exploratory analysis to test whether the observed association between self-rated health status and LCS were mediated by physician visit history, we included a variable capturing most recent physician check-up (less than 1 year, 1 to less than 2 years, 2 to less than 5 years, 5 or more years, or never) in model 3; the association between LCS and self-rated health did not change in magnitude or significance (aOR, 1.19; 95% CI, 1.02-1.37). In another exploratory analysis, no deviations from a linear association between self-rated health status and LCS was noted; when indicator variables for each health status category were included in model 3, in addition to the ordered categorical variable, none had significant coefficients (P values from .31 to .81).

In sensitivity analyses in which the primary exposure variable of self-rated health was replaced by number of comorbidities (0, 1, 2, or ≥3), the number of comorbidities was significantly associated with LCS. Individuals with 3 or more comorbidities were nearly 3 times as likely to undergo LCS as those with none of the ascertained health conditions (aOR, 2.75; 95% CI, 1.68-4.50) (Table 3). This association persisted after adjustment for demographic and health characteristics (model 4) and further adjustment for race and ethnicity (model 5). Worsening self-rated health was associated with increasing prevalence of each functional limitation and diagnosis (eFigure in the Supplement).

These results are concerning, as the long-term benefits of screening those with frail health are unknown; these individuals were excluded from the randomized clinical trials that demonstrate a mortality benefit for LCS, and it is thought that surgical resection of early-stage cancers is associated with the demonstrated mortality benefit of screening.2^,3,9 Although stereotactic body radiotherapy is used for inoperable early-stage cancers, often for individuals who are poor surgical candidates, we know of no published randomized clinical trial that has been able to recruit enough patients to adequately compare the benefit of these 2 approaches.³⁸ It may be that nonsurgical treatment such as stereotactic body radiotherapy, given before the onset of signs or symptoms of lung cancer, provides a mortality benefit even for those in frail health. Alternatively, competing causes of death such as cardiovascular and/or pulmonary disease may overwhelm any benefit of early treatment of lung cancer. Individuals with stage I lung cancer who were healthy enough to receive surgery in national registry data were 41% more likely to die for any reason than NLST participants with stage I disease⁹; if NLST participants were compared with those who were too ill for surgery, the difference in all-cause mortality would be even larger.

In general, the use of tests and treatments of unknown or limited benefit is widespread in the US health care system and is associated with health care costs and harms.39^,40 However, LCS has been largely absent from this discussion, perhaps because of its recency or because of the absolute low uptake of LCS, which was observed in this study and others. Efforts are certainly needed to increase outreach to LCS-eligible individuals, who may be particularly difficult to reach with preventive health interventions,^41–43 while also minimizing guideline-discordant screening among individuals in poor health. In the future, eligibility criteria based on individualized risk may mitigate this challenge to some degree.⁴⁴Substantial efforts are under way to increase LCS rates,⁴⁵ which is a priority for Healthy People 2030,⁴⁶ yet prior experience with other forms of cancer screening⁴⁷ and the results presented here suggest that LCS is likely to be overused among those in frail health unless there is a more concerted effort to emphasize health status in initial screening decisions.⁴⁸

This large population-based study corroborates findings from single centers of racial disparities in LCS rates.16^–19 Prior analyses of BRFSS data to investigate racial or ethnic disparities have been limited by low numbers of respondents who are not non-Hispanic White³⁵ and lack of adjustment for health status.²⁰ LCS may offer a greater mortality benefit for non-Hispanic Black individuals than non-Hispanic White individuals,¹⁵ and non-Hispanic Black individuals are at higher risk of dying from lung cancer.⁴⁹ Therefore, the racial inequalities in screening rates observed in this study and others^16–19 as well as those in terms of LCS eligibility²⁰ and adherence^50–52 merit urgent attention. Recent changes to US Preventive Services Task Force eligibility criteria to lower the age and pack-year history of screening eligibility (from 55 to 50 years, and from 30 to 20 pack-years, respectively) are in part intended to equalize screening access across racial and ethnic groups, as non-Hispanic Black individuals experience higher rates of lung cancer than non-Hispanic White individuals with the same cigarette usage and same age.^53,54 However, changing eligibility requirements alone will not fix the observed disparities in patterns of use. Widespread implementation of LCS will not achieve its maximal impact without understanding and addressing reasons for these disparities.

Strengths of the study include that data were drawn from a population-based sample with a standardized assessment of smoking history to ascertain LCS eligibility; self-reported smoking history is reliable55 and often unavailable from the electronic medical record.⁵⁶ Self-rated health is strongly associated with mortality,⁵⁷ performing more accurately than a 30-item objective measure of health and functional status in estimating 5- and 10-year all-cause mortality.⁵⁸ Furthermore, survey data on self-rated health perform similarly across different racial and ethnic groups in the US relative to objective measures of health.⁵⁹ Self-rated health status along with the ability to climb stairs was selected by the NELSON trial investigators as a key metric to gauge baseline health.¹⁰ Furthermore, our sensitivity analyses, which relied on objective comorbidities and functional limitations, provided consistent results.

This study has multiple limitations that deserve comment. Demographics, health status, and receipt of LCS were all self-reported; the validity of self-reported LCS has not yet been quantified, to our knowledge, but would likely be similar across health and racial and ethnic categories as LCS is not a particularly stigmatized or desired behavior. Thus, misclassification of screening status would tend to bias our results toward the null, if present. A minority of BRFSS respondents identified as non-Hispanic Black (3%, unweighted), which likely limits the generalizability of our findings and highlights the need to respectfully engage racial and ethnic minority groups in future LCS research. We could not directly ascertain whether a given individual was willing and able to undergo lung surgery; this is a limitation of all publicly available data sets60 and, indeed, routine health data, as clinicians have no recommended tool to systematically assess health status prior to ordering LCS. Although evaluation before thoracic surgery includes well-established metrics such as forced expiratory volume in 1 second,⁶¹ it would be infeasible to require formal pulmonary function tests or other objective diagnostic studies for all individuals considering LCS. Functional ability is a commonly used preoperative measure of cardiovascular fitness.⁶² We could only include diagnoses ascertained by BRFSS; a notable exception is hypertension, which is assessed only in odd survey years.⁶³ Residual confounding could be present; if so, the true association between non-Hispanic Black individuals and LCS would be even lower than observed (ie, aOR even further from 1, meaning larger racial or ethnic disparities). The cross-sectional nature of the survey precludes inferences about the time sequence of reported associations; however, as the survey only asked about LCS in the prior 12 months, many reported comorbidities likely predated receipt of LCS. Similar to prior studies,^{12,13,20–22,35,36,64,65} we were unable to assess whether scans reported by respondents were truly screening tests or ordered for diagnostic or surveillance reasons. This is a major limitation of using publicly available or routine data to study LCS and must be addressed in future research to validate these findings.

Published: March 31, 2022. doi:10.1001/jamanetworkopen.2022.5318

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2022 Rustagi AS et al. JAMA Network Open.

Corresponding Author: Alison S. Rustagi, MD, PhD, Medical Service, San Francisco Veterans Affairs Health Care System, 4150 Clement St, Bldg 18, Rm 118, San Francisco, CA 94122 (alison.rustagi@ucsf.edu).

Author Contributions: Dr Rustagi had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: All authors.

Acquisition, analysis, or interpretation of data: Rustagi, Byers.

Drafting of the manuscript: Rustagi.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Rustagi.

Obtained funding: Rustagi.

Administrative, technical, or material support: Rustagi, Byers.

Supervision: Keyhani.

Conflict of Interest Disclosures: Dr Byers reported receiving a Research Career Scientist award (IK6 CX002386) from the Department of Veterans Affairs outside the submitted work. No other disclosures were reported.

Funding/Support: Dr Rustagi’s work was supported by the Grunfeld Scholars Research Development Initiative. Dr Keyhani’s work was supported by R01AG068678-01A1.

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Information: All data are freely and publicly accessible from the US Centers for Disease Control and Prevention (https://www.cdc.gov/brfss/index.html).