Association of Stage Shift and Population Mortality Among Patients With Non–Small Cell Lung Cancer

IMPORTANCE Early detection by computed tomography and a more attention-oriented approach to incidentally identified pulmonary nodules in the last decade has led to population stage shift for non–small cell lung cancer (NSCLC). This stage shift could substantially confound the evaluation of newer therapeutics and mortality outcomes. OBJECTIVE To investigate the association of stage shift with population mortality among patients with NSCLC. DESIGN, SETTING, AND PARTICIPANTS This retrospective cohort study was performed from October 2020 to June 2021 and used data from the Surveillance, Epidemiology, and End Results (SEER) registries to assess all patients from 2006 to 2016 with NSCLC. MAIN OUTCOMES AND MEASURES Incidence-based mortality was evaluated by year-of-death. To assess shifts in diagnostic characteristics, clinical stage and histology distributions were examined by year using χ 2 tests. Trends were assessed using the average annual percentage change (AAPC), calculated with JoinPoint software. Kaplan-Meier survival analysis assessed overall survival according to stage and compared those missing any stage with those with a reported stage.


Introduction
Lung cancer remains among the leading causes of cancer death in the United States. 1,2 Based on the Annual Report to the Nation, incidence rates for lung cancer have significantly declined from 2012 to 2016 (average annual percent change [AAPC] of −2.6% for male individuals and −1.1% for female individuals). 1 Moreover, mortality for lung cancer from 2013 to 2017 has decreased at a greater rate compared with the incidence and has experienced one of the largest declines in death rates compared with other common cancer deaths (AAPC of −4.8% among male individuals and −3.7% among female individuals). 1 The improved outcomes with lung cancer are quite multifactorial and can be attributed to advances in prevention, early detection, and treatment of lung cancer. 3,4 In recent years, the many advances in medical therapeutics, such as targeted therapy, immunotherapy, and chemotherapy, have garnered interest in the role they may play in affecting lung cancer population-level mortality. 5,6 In particular, Howlader et al, 5 using data from Surveillance, Epidemiology, and End Results (SEER) registries from 2001 to 2016 concluded that the observed lung cancer mortality during the study timeframe was likely to be explained by the approval and use of targeted therapies especially for non-small cell lung cancer (NSCLC). However, Howlader et al 5 do not provide direct evidence to support the direct effect of targeted therapies in affecting mortality but rather cite how the lack of other major advances in prevention or treatment explain the substantial decline in incidence-based mortality.
Many randomized clinical trials have demonstrated the significant survival benefit of targeted therapies among patients harboring driver variants among the druggable oncogenes, such as epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK). [7][8][9] Moreover, immune-based therapies, in particular programmed cell death protein 1-programmed death ligand 1 (PD-1-PD-L1) inhibitors have substantially improved survival among patients with NSCLC even if they do not carry EGFR and ALK variants. [10][11][12][13] Since 20% of patients with NSCLC have a substantial and sustained response to targeted and immunotherapies, these therapies certainly play a role in affecting population mortality. 14 Nevertheless, in associating the role of targeted and immunotherapies in decreasing mortality, Howlader et al 5 minimize the role of early detection and stage shift on mortality by suggesting that "patients moved from unknown stages to more specific stage categories (as a result of the availability of better imaging) rather than shifting from late to early stages." 5 However, the role of stage shift due to early detection in affecting NSCLC population-level mortality may not be as minimal as suggested. As a matter of fact, computed tomography (CT) screening for lung cancer is the only modality demonstrating decreased disease-specific mortality. As per the National Lung Screening Trial (NLST), screening with low-dose CT resulted in a positive screening rate of 24.2% compared with 6.9% with chest radiography. 6 The increased rate of early detection has translated into decreased mortality among patients diagnosed with lung cancer. The NLST, Nederlands-Leuvens Longkanker Screenings Onderzoek (NELSON) trial, and International Early Lung Cancer Action Project (I-ELCAP) have demonstrated that CT screening can identify early stage disease in 4 of 5 patients unknowingly harboring lung cancer 6,15,16 . Given these survival benefits, the United States Preventive Services Task Force (USPSTF) has recommended since 2013 annual screening with low-dose CT as standard of care with further expansion of the screening criteria in 2021 to include younger adults with a smaller pack-year smoking history in order increase early detection and subsequently decrease mortality from lung cancer. 17 To better understand the association of early detection with lung cancer mortality, it is pertinent to evaluate the extent of stage shift in the last decade and its effect in contributing to NSCLC incidence-based mortality. We hypothesize that lung cancer early detection by CT, both intentional and nonintentional (back-alley screening), such as cardiac CT angiograms screening for coronary disease, and a more attention-oriented approach to incidentally identified pulmonary nodules, is associated with a stage shift and subsequent decreased mortality from earlier surgical intervention.
We performed a more systematic exhaustive stage and histology evaluation over the same period (2006-2016) using the same SEER data set and methodology as Howlader et al. 5

Data Source and Study Selection
The SEER Program compiles information on cancer incidence, including patient demographics, selfreported race and ethnicity, tumor characteristics, treatment, and vital status from population-based cancer-registries. 18 The registries included in SEER currently cover approximately 35% of the United States population. 18 This retrospective cohort study analysis was conducted from October 1, 2020, to June 30, 2021, using data extracted from SEER. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. 19 Since the SEER database is public and deidentified, the study was deemed as exempt research by the institutional review board at the Icahn School of Medicine at Mount Sinai and informed consent was waived.
Based on when the International Early Lung Cancer Action Program began recommending lung cancer screening, 16

Statistical Analysis
Clinical stage and histology distributions were examined by year to assess shifts in diagnostic characteristics over time using χ 2 tests. Additionally, the AAPC in tumor characteristics over time was other analyses were conducted using SAS software version 9.4 (SAS Institute). 21 Incidence-based mortality within 5 years of diagnosis, which was defined as the number of deaths among those with a NSCLC diagnosis in SEER, divided by the total population residing in the geographic areas of the SEER registries, was calculated in SEER*Stat for 2006 to 2016. 22 Figures were created in R version 3.4.0 (R Project for Statistical Computing). The significance threshold was P < .05. Significance was based on χ 2 tests (for the association between characteristics and year), and t tests for the joinpoint trends. All testing was 2-sided. Statistical analysis was performed from October 2020 to June 2021. had a tumor size less than 2 cm, whereas 26.9% (n = 84 150) had tumor size greater than or equal to 5 cm (Table).

Results
Incidence and incidence-based mortality for NSCLC according to sex are presented in Figure 1.   (Figure 2). A sensitivity analysis was conducted through 2015, before the staging change in SEER and found similar results.
There was a large, statistically significant increase in those diagnosed with adenocarcinomas, from  (Figure 3).  Those missing any staging information had a mean (SD) age of 71.5 (0.

Discussion
In this cohort study, we examined 312 382 patients in SEER diagnosed with NSCLC between 2006 and 2016 to describe trends in incidence, incidence-based mortality, stage, and histology at diagnosis. Over the last decade, lung cancer population mortality has decreased. This decline has been driven by many factors, including smoking cessation, medical therapies, CT screening, and earlier therapeutic interventions. While prior studies have explored the association of smoking cessation, earlier interventions, and targeted therapies to NSCLC mortality, the role of stage shift due to early detection has not been adequately studied. 3,4,23 With limited precedent, this study wished to evaluate the extent of stage shift on the population level in the last decade and its association with NSCLC incidence-based mortality through a more exhaustive evaluation of stage and histology.
Consistent with prior literature, we found that the incidence-based mortality for NSCLC from 2006 to 2016 has declined at a faster rate compared with the incidence. 1,2 To better elucidate the improvements in incidence-based mortality during the study period, we assessed the trends in stage at diagnosis. We detected a significant association between year of diagnosis and stage. Patients with  our findings support patients from later stages being shifted to earlier stages, rather than unknown stages being better classified to earlier stages.
The stage and histology shifts we describe in the SEER data are consistent on a population-level with prior studies assessing and corroborating the efficacy of low-dose CT screening for lung cancer. [24][25][26][27][28] In particular, in the NLST, of 649 positive screening tests with low-dose CT, 70.2% were stage I and II with III and IV accounting for only 29.8%. 6 Moreover, the majority of the early stage lung cancer tumors were indolent histologies, such as bronchioloalveolar carcinoma (BAC) and adenocarcinoma, which have an overall 5-year survival of approximately 89% and 70%, respectively,   [24][25][26][27][28] Taken together, the increase in early-stage NSCLC with a corresponding increase in adenocarcinoma that we detected in the SEER data mirrors what prior clinical trials on CT-based lung cancer screening have described on a smaller scale. Our findings in context with these prior studies seem to suggest that awareness of CT lung cancer screening is associated with an earlier detection of NSCLC (back-alley CT screening). The greater decline in incidence-based mortality compared with the incidence of NSCLC over the past decade may be partially explained by stage and histology shifts.
We realize that patient adherence to lung cancer screening with low-dose CT remains limited. [34][35][36] According to the National Cancer Institute (NCI), uptake of CT screening has been limited and stable since 2010, with 4.5% and 5.9% of adults aged 55 to 80 years in 2010 and 2015 respectively, who met the USPSTF criteria for lung cancer screening, received a CT scan within the prior year. 19 Thus, we cannot only attribute the trends in NSCLC incidence and incidence-based mortality over the past decade to purposeful lung cancer screening with CT.

Limitations
The findings from this study should be interpreted within the context of its limitations. This was a retrospective study using a database that only contained preselected demographic and clinical variables. We did not have available data for smoking status, family history of lung and/or bronchial cancer, occupational exposure to carcinogens, and driver variations, such as EGFR, which would provide increased insight, especially when evaluating the trends in incidence-based mortality.
Moreover, for the participants in the SEER database, we lack detailed information on the diagnostic method, thus limiting us from measuring the rate of CT scan uptake among the study population.
Additionally, 2.3% of study population were either unstaged or their staging information was unknown. Although we performed an exhaustive analysis to better define these participants, we realize that the lack of staging for this portion of study population prevents us from recategorizing to a well-defined stage.

Conclusions
This cohort study found that population-level mortality for NSCLC has decreased from 2006 to 2016.
Although advances in treatments, particularly targeted therapeutics, have played a role in affecting mortality, our analysis suggests that decreased mortality is also associated with a diagnostic shift from later to earlier stage lung cancer and a histology shift to adenocarcinoma. Studies investigating the population impact of treatment on lung cancer mortality must take into account the confounding association of stage shift with survival and mortality outcome.

ARTICLE INFORMATION
Accepted for Publication: October 12, 2021.