Burden and Impact of Frailty and Comorbidity in Individuals Screened for Lung Cancer.

Introduction
The National Lung Screening Trial (NLST) was a pivotal study demonstrating a 20% reduction in lung cancer mortality among high-risk individuals screened annually with low-dose computed tomography (LDCT; ref. 1), resulting in a paradigm shift in lung cancer screening (LCS). However, trial participants’ sociodemographic, clinical, and underlying health characteristics may not accurately represent the broader screen-eligible population, challenging the generalizability of LCS trial findings. Compared with the general US population, NLST participants had a similar distribution of sex and pack-year (P-Y) smoking history; however, they were younger, were less likely to be currently smoking, and had more years of education (2).
Individuals at risk for lung cancer and LCS-eligible are also at increased risk for other smoking-related health conditions, such as cardiovascular disease (CVD) and chronic obstructive pulmonary disease (COPD; ref. 3). Prior studies reported a high prevalence of comorbidities among populations undergoing LCS (4–6). A recent meta-analysis found a higher burden of comorbidities in individuals enrolled in LCS in the general population than reported in breast and colorectal cancer screening populations (4). Frailty is a less well-studied health-related measure among those eligible for and undergoing LCS; however, it may be important when assessing the benefits and harms of LCS. Frailty is commonly defined as a syndrome influenced by stressors and aging that significantly increases the risk of adverse health outcomes, including mortality (7). Smoking is associated with developing and worsening frailty (8).
Comorbidities and frailty affect life expectancy (9, 10); thus, characterizing these measures and understanding the benefits and harms associated with LCS across levels of both comorbidity and frailty is needed to inform LCS research gaps. Additionally, the extent to which comorbidity and frailty burden influence downstream imaging and invasive procedures following the baseline LCS imaging exam is unknown. In this study, we sought to describe the number and types of comorbidities [Charlson comorbidity score (CCS)] and characterize the predicted probability of being frail, as measured by the Faurot frailty index (FFI), in a screened population, overall and by age groups. We also evaluated baseline LCS exam results by comorbidity and frailty to determine whether those with higher CCS or FFI were more likely to experience a positive LCS exam result. Finally, we compared downstream imaging and invasive procedure rates by LCS exam results and levels of comorbidity and frailty.

Materials and Methods

Study population and setting
Data come from the North Carolina Lung Screening Registry (NCLSR), a NCI-funded registry of individuals undergoing screening LDCT at imaging sites across North Carolina. Collected NCLSR data include information on sociodemographics; lung cancer risk factors; LCS examinations, including the radiologists’ reported American College of Radiology Lung CT Screening Reporting & Data System (Lung-RADS) assessment; and follow-up procedures. We linked NCLSR data with public and private administrative data from 2013 to 2020, including Medicare, Medicaid, and private payer insurance claims in North Carolina (11). Data were linked using the identifiers of first, middle, and last name; date of birth; sex; Social Security number; or ZIP code, using a probabilistic record linkage framework with Match*Pro software version 2.1 (IMS). Each of these identifiers was present in more than 99% of individuals, with the exception of the Social Security number, which was present in 88% of individuals. Prior to linking, each identifier was cleaned and standardized across data sets (e.g., punctuation was removed from names and dates were converted to the same format). After the probabilistic linking weight was calculated for each potential pair, a coauthor (D.D. Durham) implemented a standard operating protocol to manually review instances in which matches were ambiguous to determine if pairs were a match or not, ensuring consistency and accuracy (12).
We included individuals undergoing LCS in the NCLSR who met the 2013 US Preventive Services Task Force screening criteria: aged 55 to 80 years with at least a 30 P-Y smoking history, who currently or formerly smoked and, if they quit, had done so in the past 15 years. We used the 2013 USPSTF criteria, as the study data were collected when these criteria were in use in clinical practice. We excluded individuals in the NCLSR who did not link to claims data with continuous enrollment for 12 months before the LCS exam so that we could assess comorbidity and frailty during this period. Among individuals with more than one screening exam, we assessed comorbidity and frailty in the 12 months preceding the first LCS exam.

Data measures
From the NCLSR data, we determined the screened individual’s age, sex, race, and ethnicity; smoking status (current or former); smoking P-Y; body mass index; type of residence (urban-focused, large rural, and small or isolated rural based on RUCA 3.0); the screening exam year; and the reported Lung-RADS assessment.

FFI
Using claims data in the 12 months before the LCS exam, we calculated the FFI, a validated Medicare claims–based proxy measure for frailty (13, 14). The FFI is a weighted index that includes demographic variables and 20 claims-based indicators predictive of frailty (e.g., home oxygen use, diagnosis of dementia, wheelchair use), representing an individual’s predicted probability of being frail. We classified the FFI into five predicted probability of frailty categories: low (<0.05), low-medium (0.05 to <0.10), medium (0.10 to <0.20), medium-high (0.20 to <0.40), and high (≥0.40), similar to prior studies (13, 15, 16).

CCS
Using claims data from the 12 months before the LCS exam, we calculated the CCS. The CCS is a weighted index that considers both the number and seriousness of comorbid diseases. It has been used to predict 1- and 10-year mortality rates (10). We examined the number of comorbid conditions using the CCS.

LCS results and downstream healthcare utilization
We categorized LCS exam results based on Lung-RADS into positive (Lung-RADS 3 or 4) and negative (Lung-RADS 1 and 2). We determined receipt of any subsequent imaging (LDCT, chest CT with or without contrast, MRI, or PET) or invasive procedures (lung biopsy, bronchoscopy, mediastinoscopy, mediastinotomy, thoracoscopy, and pleural procedures) within 12 months after the LCS exam, using Rendle and colleagues’ methods (5).

Statistical analyses
We describe the number and components of comorbidities in the CCS, the FFI-predicted probability distribution, and the components of the FFI among those screened by age group (<65 and ≥65 years). Results were stratified by age group as the burden of comorbidity and frailty increases with age in the general population. We dichotomized the CCS into those with CCS 0 or 1 (low CCS) versus CCS 2 or more (high CCS). We dichotomized the FFI into those with low predicted probability of frailty (FFI < 0.05) and those with low-medium, medium, medium-high, or high predicted probability of frailty (greater than low FFI, i.e., FFI ≥ 0.05). We compared the distribution of CCS among individuals with low versus greater than low FFI using χ2 tests. We also compared the reported Lung-RADS assessment of the baseline LCS exam result by dichotomized FFI and CCS using χ2 tests. We calculated and compared rates of downstream imaging and invasive procedures [with 95% confidence intervals (95% CI)] among those with low versus greater than low FFI and among those with low versus high CCS, stratified by Lung-RADS (positive and negative), using χ2 tests. Following data-use agreements, we suppressed all cell sizes <11.
Statistical analyses were conducted using SAS version 9.4. Match*Pro 2.1 (https://seer.cancer.gov/tools/matchpro/) was used to link the NCLSR and insurance claims data. This study was conducted in accordance with recognized ethical guidelines. The University of North Carolina at Chapel Hill Institutional Review Board approved this study under a waiver of the Health Insurance Portability and Accountability Act and informed consent because the study demonstrated no more than minimal risk to participants or their privacy. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline, provided in Supplementary Table S1.

Results
Of the 3,923 NCLSR individuals linked to claims data, 42.4% (n = 1,663) were aged <65 years, and 57.6% (n = 2,260) were aged 65 years and older, with a similar distribution of females and males (Table 1). Most individuals were White, non-Hispanic, and lived in an urban area. Individuals <65 years were more likely to be currently smoking but less likely to have a ≥40 P-Y smoking history compared with those ≥65 years, 63.9% versus 52.4% and 55.9% versus 64.7%, respectively. Younger patients were more likely to be obese compared with older patients (40.3% vs. 34.3%). Data abstracted from radiology reports showed that older versus younger individuals had higher Lung-RADS scores.
The range of FFI among those undergoing LCS was 0.01 to 0.98, with a median of 0.03 and an interquartile range (IQR) of 0.03–0.04 (Table 2). Most individuals had a low predicted probability of frailty (82.1%), and <1% had a medium-high or high probability of frailty. The most common FFI components increasing frailty were ambulance use (11.7%), home oxygen use (8.3%), lipid abnormalities (5.9%), and arthritis (4.1%). In contrast, the most common FFI component that decreases frailty was cancer screening (3.7%), excluding LCS, as the index was developed before LCS claims codes were available. Of these components, home oxygen use was higher in those ≥65 versus those <65 years (9.2% vs. 6.7%, respectively).
Among the cohort screened for lung cancer, 27.4% had a CCS of 0, 28.0% had a CCS of 1, 18.3% had a CCS of 2, 10.7% had a CCS of 3, and 15.6% had a CCS of 4 or more (Table 2). The most common comorbid conditions contributing to the CCS were COPD (49.8%), diabetes with and without complications or any diabetes (28.4%), peripheral vascular disease (18.4%), CVD (11.5%), and congestive heart failure (11.0%). Similar patterns of CCS score distribution and prevalence of comorbid conditions were observed for those <65 vs. ≥65 years.
To assess the combination of comorbidity and frailty, we examined the number of comorbid conditions among those with low versus greater than low FFI (Table 3). Approximately 31.2% of those with low FFI had CCS = 0 compared with 10% of those with greater than low FFI. A higher proportion of those with greater than low FFI versus those with low FFI had CCS of 2 or more (P value < 0.0001). CCS of 6 or more occurred in 4.4% of those with low FFI versus 12.2% of those with greater than low FFI.
The distribution of Lung-RADS categories by dichotomized FFI and CCS is shown in Table 4. There were no differences in Lung-RADS categories, that is, the level of suspiciousness for lung cancer, among those with low versus greater than low FFI (P value = 0.5858). In addition, Lung-RADS categories did not differ among those with lower versus higher CCS (P value = 0.4921).
We also compared the rates of any downstream procedures (imaging and invasive) following negative (Lung-RADS 1 or 2) and positive (Lung-RADS 3, 4A, 4B, or 4X) baseline LCS exam results by dichotomized FFI and CCS (Table 5). Among individuals with a negative LCS exam, the downstream imaging rate was 17.2 per 100 persons (95% CI, 15.9–18.5); the downstream invasive procedure rate was 1.5 per 100 persons (95% CI, 1.1–2.3; Table 5). Downstream imaging rates were higher among those with greater than low versus low FFI (22.3 per 100 persons, 95% CI, 18.8–25.8 vs. 16.1 per 100 persons, 95% CI, 14.7–17.6, respectively). Moreover, those with higher CCS had higher downstream imaging rates compared with those with lower CCS (20.2 per 100 persons, 95% CI, 18.2–22.3 vs. 14.8 per 100 persons, 95% CI, 13.1–16.5, respectively). In individuals with a positive LCS, the downstream imaging rate was 64.3 per 100 persons (95% CI, 60.6–67.9), and the downstream invasive procedure rate was 16.8 per 100 persons (95% CI, 13.9–19.7). There were no differences in downstream imaging or invasive procedure rates among those with positive LCS results by FFI or CCS.

Discussion
To our knowledge, this is one of the first studies to report on frailty among individuals undergoing LCS and, more specifically, to assess the overlap of comorbidities, CCS, and frailty and evaluate their relationship with LCS exam results, including downstream imaging and procedures. Our novel finding of a low predicted probability of frailty among the majority of individuals undergoing LCS (median FFI = 3%) is similar to the predicted probability of frailty in a general population of Medicare fee-for-service beneficiaries (median FFI=3.4–3.8%; ref. 16). Almost 97% of LCS recipients in our study had a low or low-to-medium frailty score, providing some reassurance that very few individuals undergoing LCS in North Carolina would be unable to tolerate surgical resection for the treatment of early-stage lung cancer based on frailty criteria. Several claims-based frailty measures exist (13, 17–19); the performance of these measures in predicting the frailty phenotype is similar (20). A validation study of the FFI reported that individuals with higher FFI-predicted probabilities had a higher one-year mortality rate than those with lower FFI-predicted probabilities (14).
Our study found a higher burden of comorbid conditions among individuals undergoing LCS than NLST participants (1). Common comorbidities included COPD, diabetes, and CVD, with variations in certain comorbidities across age groups. Our findings are similar to previous reports citing higher rates of COPD and CVD among LCS populations than among individuals enrolled in LCS RCTs (2). A meta-analysis of 69 studies in LCS revealed a high burden of comorbidities among populations undergoing LCS, with the most common comorbidities being hypertension (35.2%), COPD (23.5%), severe COPD (10.7%), ischemic heart disease (16.6%), and peripheral vascular disease (13.1%; ref. 4). We found that most individuals in our cohort (72.3%) had a CCS of ≥1 and more than a quarter had a CCS of ≥3. Notably, we found no difference in the CCS between those younger or older than 65.
Our finding that those with a higher predicted probability of frailty are more likely to have more comorbid conditions demonstrates that individuals screened for lung cancer who are frail also have a higher comorbidity burden as measured by the CCS. Analysis of 2022 Behavioral Risk Factor Surveillance System data indicated that 16% of those screened for lung cancer self-reported poor health (21). Similar to our findings that 15.6% of screened individuals had a CCS of ≥4, a recent retrospective study of 31,795 screened individuals showed that 19% had a CCS of ≥4 (22).
Frailty is a significant cause of premature functional decline and early mortality among older adults. Cancer and its treatments pose significant stressors, leading to a high incidence of frailty among older individuals (23, 24). Frailty may affect both the ability to tolerate cancer treatments and the prognosis of patients with cancer, especially those with lung cancer who tend to be older and have increased comorbidities (25). A systematic review of 16 studies evaluating frailty (eight different frailty assessments were used, and frailty definitions varied widely) and outcomes in patients with lung cancer who were aged 65 years and older (mean age, 76.6 years) revealed a strong correlation between frailty and mortality (HR, 3.5–11.9), and frailty was also associated with increased treatment-related toxicity (26). In a retrospective study of 1,667 patients with lung cancer, 17.8% were classified as being in a frail state at the time of diagnosis using a frailty index based on laboratory tests (27). All-cause mortality was 61.1% for all patients. However, a significantly higher rate of all-cause mortality was noted in frail patients versus those who were robust (frail vs. robust, HR, 1.62, 95% CI, 1.35–1.94; ref. 27). The NLST did not evaluate frailty and excluded individuals who were on oxygen and could not tolerate surgery from enrollment. Given these exclusions, the benefits of LCS among those with significant frailty are unknown.
We found an overall rate of downstream imaging and invasive procedures of 17.2% and 1.5%, respectively, following a negative baseline LCS exam, with downstream imaging rates higher among those with higher FFI and higher CCS scores. Although no other studies have examined downstream imaging or procedure rates by frailty or comorbidity, the PROSPR cohort of 7,243 patients reported a higher overall downstream imaging rate [20.8% (95% CI, 19.9–21.7) vs. 17.2% (95% CI, 15.9–18.5), respectively] and a lower downstream invasive procedure rate [0.8% (95% CI, 0.6–1.0) vs. 1.5% (95% CI, 1.1–2.0), respectively] in LCS individuals with a negative baseline LCS exam (5). Downstream imaging and procedures in individuals with negative LCS exam results may be performed for radiologic findings suspicious for inflammation or infection. This may be more common in those with a higher burden of comorbidities and more frailty. Our study data were collected while Lung-RADS version 1.1 was in effect, which did not include a category for such findings (https://edge.sitecorecloud.io/americancoldf5f-acrorgf92a-productioncb02-3650/media/ACR/Files/RADS/Lung-RADS/Lung-RADS-v1-1-Assessment-Categories.pdf). In Lung-RADS 2022, a Lung-RADS 0 category was introduced for “findings suggestive of an inflammatory or infectious process,” with a recommended repeat LCS exam in 1 to 3 months (https://www.acr.org/-/media/ACR/Files/RADS/Lung-RADS/Lung-RADS-2022.pdf). As expected, our results show that downstream imaging and invasive procedure rates were higher among individuals with positive LCS exam results; however, we found no difference in these rates based on FFI or CCS for those with positive results. In comparison with the PROSPR study, our study found that the overall rate of downstream imaging was higher [45.3% (95% CI, 41.5–49.1) vs. 64.2% (95% CI, 60.6–67.9), respectively], whereas the rate of downstream invasive procedures was similar [15.6% (95% CI, 12.8–18.4) vs. 16.8% (95% CI, 13.9–19.7), respectively; ref. 5].
The efficacy of LCS hinges on a delicate balance of multiple factors, including an individual’s inherent risk of developing lung cancer, the benefit of early lung cancer detection, potential adverse outcomes from the screening process itself (false-positive results leading to procedures), and the likelihood of mortality from underlying comorbidities, especially cardiopulmonary conditions (3). Understanding this intricate relationship is essential for accurately assessing the benefits and harms of LCS.
The mortality benefit of LCS observed in LCS trials is due to the detection and surgical resection of early-stage lung cancer (28, 29). Frailty and comorbid conditions are important factors that may affect the ability to undergo curative-intent surgical resection, thereby affecting the net benefits of LCS. In a recent analysis of 1,614 patients with stage I to IIIA non–small cell lung cancer from 11 International Lung Cancer Consortium studies, patients with respiratory comorbidities were less likely to undergo surgical resection (stage IA adjusted OR = 0.54; 95% CI, 0.35–0.83 and stages IB–IIIA adjusted OR = 0.57; 95% CI, 0.46–0.70; ref. 30). Cardiometabolic comorbidities increased the risk of death from competing causes (adjusted HR, 1.34; 95% CI, 1.12–1.69), and both respiratory and cardiometabolic comorbidities were associated with worse overall survival (31). Frailty has been shown to independently predict worse outcomes after lung resection, even after controlling for comorbidity (31). LCS trials excluded individuals with poor health (1, 29, 32); however, frailty has not been previously evaluated in the context of LCS outcomes.
Our study has several limitations. First, the study population is derived from one geographic location, so the findings may not be generalizable to the entire US population. We include individuals screened at seven North Carolina imaging locations, including academic and community sites, so observed patterns are broader than at a single site. Given the use of claims data to assess comorbidity and frailty burden, our study population excludes individuals who lack health insurance and includes those with consistent healthcare access who may have different frailty or comorbidity profiles. Although the FFI was developed in a population of those aged 65 and older, we also applied the FFI to those below 65. To account for this potential measurement error, we stratified our results by age group and found similar distributions of comorbidity and frailty. However, future work is needed to evaluate and assess frailty in younger populations.
Our study provides novel findings on frailty among individuals undergoing LCS. The downstream imaging and invasive procedure rates after positive LCS did not differ by FFI or CCS. The observation that among those with a negative LCS exam who have greater than low frailty, the rates of downstream imaging and invasive procedures are higher than in those with low frailty may suggest that individuals with greater frailty are more likely to have infectious or inflammatory findings that necessitate further follow-up. This observation warrants further investigation, particularly because the majority of those who undergo LCS will have a negative exam. Furthermore, we demonstrate that in general, screened individuals are in the low probability of frailty category, with a predicted probability of frailty distribution similar to that of the general older population. Hence, lung cancer-screened individuals are not likely to be too frail to undergo treatment, especially surgical resection, if a highly suspicious nodule is detected. Additional research is needed to evaluate downstream complications and other outcomes related to frailty to further understand and assess the net benefits of LCS.

Supplementary Material
Supplementary Table 1Supplementary Table 1 shows the STROBE reporting checklist