Population-Level Trends in Lung Cancer Mortality and Stage Distribution During the USPSTF LDCT Guideline Era: A Multi-Database Quasi-Experimental Study

Abstract

Introduction

In 2013, the United States (U.S.) Preventive Services Task Force recommended low-dose computed tomography (LDCT) for lung-cancer screening. We assessed temporal, population-level changes in incidence, stage at diagnosis, and mortality before and after this recommendation among adults aged 55–79 years.

Methods

Using the US Centers for Disease Control and Prevention Wide-ranging Online Data WONDER (CDC WONDER), Global Burden of Disease (GBD), and SEER data, we examined age-adjusted incidence (AAIR) and mortality (AAMR) trends. Joinpoint regression and interrupted time-series (ITS) models evaluated pre- and post-2014 changes in population-level trends; counterfactual analyses estimated differences between observed outcomes and projections based on pre-2014 trends. SEER was used to characterize stage distributions (localized, regional, distant).

Results

Post-2014 declines accelerated across data systems. Annual AAMR decreased by 3.03% in CDC WONDER and 2.58% in GBD; AAIR declined by 1.85% and 2.37%, respectively. SEER showed a higher proportion of localized disease and fewer distant-stage diagnoses after 2014; joinpoint analyses identified a 2013 inflection with steeper declines in distant-stage incidence thereafter. In models extrapolating pre-2014 trends to subsequent years, observed deaths were lower than counterfactual projections by 28,168 in CDC WONDER and 6,906 in GBD. As an exploratory indicator, mortality-to-incidence ratios declined nationally and in both sexes.

Conclusions

During the USPSTF LDCT guideline era, independent national datasets showed accelerated mortality declines and a shift toward localized-stage diagnoses. These convergent findings are temporally consistent with real-world benefits of LDCT screening, but they should be interpreted within a broader context that includes continued reductions in smoking, gradual screening uptake, and major therapeutic advances. The findings are hypothesis-generating and do not estimate the isolated causal effect of LDCT screening.

Plain Language Summary

Why was this study done? Lung cancer is the leading cause of cancer death worldwide. To reduce these deaths, United States health officials issued a major recommendation in 2013: adults aged 55–79 with a history of heavy smoking should undergo annual screening using Low-Dose Computed Tomography (LDCT). The goal of this policy was to catch tumors early when they are curable. We wanted to see if this national recommendation actually led to better outcomes for patients across the country. What did we do? We analyzed data from three massive national databases (CDC WONDER, Global Burden of Disease, and SEER) covering millions of Americans. We compared trends in lung cancer cases and deaths from before the recommendation (1999–2013) to the trends after it was implemented (2014–2021). We used statistical models to determine if the “speed” of improvement changed after the policy began. What did we find? We found that after 2014, lung cancer death rates declined more rapidly than in earlier years. We also observed a population-level shift toward more localized-stage diagnoses and fewer distant-stage diagnoses. Compared with projections based on pre-2014 trends, observed deaths were approximately 28,000 lower in CDC WONDER data during the post-2014 period. What does this mean? These findings are consistent with improved population-level lung cancer outcomes during the guideline era. However, they cannot determine how much of the observed change was attributable to LDCT screening, as smoking reductions, diagnostic changes, and therapeutic advances also occurred during this period.

Keywords

lung cancer LDCT counterfactual analysis AAMR CDC WONDER

Introduction

Lung cancer remains the leading cause of cancer-related mortality worldwide, posing substantial medical, economic, and social challenges.¹ Although tobacco control measures, early detection, and systemic therapies have advanced since the 1990s, lung cancer persisted as a major contributor to cancer-related deaths in the early 2010s.² Among high-risk individuals aged 55–79 years, smoking prevalence, imaging screening quality, and innovations in targeted and immunotherapies collectively influence incidence and survival trends, underscoring the public health significance of population-level epidemiological shifts.^3,4,5

Randomized trials such as NLST and NELSON have established that annual low-dose computed tomography (LDCT) screening reduces lung cancer-specific mortality and enhances early-stage detection in high-risk groups.^5,6 Prompted by this evidence, the United States (U.S.) Preventive Services Task Force (USPSTF) issued a Grade B recommendation for LDCT screening in 2013, later adopted nationwide through Medicare and private insurance policies.^7-10 However, real-world outcomes remain contentious: while screening may improve survival through stage migration, its benefits could be offset by false positives, overdiagnosis, radiation risks, and regional implementation disparities.¹¹ Moreover, population-based surveys suggest that LDCT uptake among eligible adults has remained modest in the United States, potentially limiting the immediate impact of screening at the national level. Concurrent declines in smoking and advances in molecular therapies further complicate the attribution of mortality trends to any single intervention.¹²

Several national analyses have already described temporal trends in lung cancer incidence and mortality before and after the USPSTF LDCT recommendations.^13-15 However, most prior work has relied on a single data source or focused on either incidence or mortality alone, with less emphasis on integrating stage distribution, mortality-to-incidence ratio (MIR), and counterfactual projections within a unified framework. Crucially, distinguishing the effects of screening from substantial concurrent advances in systemic therapies (targeted agents and immunotherapy) and long-term tobacco control remains a major methodological challenge in ecological studies. Critical unresolved questions include whether mortality declines accelerated post-2013, whether stage distributions shifted concurrently, and how overlapping screening uptake, tobacco control, and treatment innovations interact in population-level outcomes.^13-15 The MIR, though widely used as a proxy for lethality and care quality, has important limitations and is not a validated surrogate for individual-level survival, underscoring the need for cautious interpretation.^16-18 Additionally, with U.S. population aging, rate-based declines may not reflect absolute disease burden, necessitating counterfactual frameworks to estimate deviations between observed and projected event counts.¹⁹

This study uses the 2013 USPSTF LDCT recommendation as a pre-specified policy-era anchor for a quasi-experimental trend analysis, integrating three independent national data sources (CDC WONDER mortality data, GBD model-based estimates, and SEER registry data) to assess population-level changes in incidence, mortality, stage distribution, and MIR among adults aged 55–79 years. Because this ecological design cannot separate the effects of screening from concurrent changes in smoking patterns and lung cancer treatment, the study was designed to evaluate temporal associations and hypothesis-generating patterns rather than the isolated causal effect of LDCT screening. First, we evaluated whether AAIR and AAMR for lung/tracheal cancer declined more rapidly post-mid-2010s. Next, we analyzed concurrent stage migration and its relationship with overall epidemiological changes. Finally, we compared observed post-2014 case and mortality counts with projected trajectories based on pre-2014 trends to quantify deviations from expected population-level patterns. Throughout, we interpret these findings as population-level associations rather than causal effects of LDCT screening. Cross-database and methodological consistency checks ensured robust and generalizable conclusions.

Methods

Data Sources and Study Population

The analysis drew on three principal data sources: the US Centers for Disease Control and Prevention (CDC) Wide-ranging Online Data (WONDER) system (specifically, the Underlying Cause of Death files),²⁰ the Global Burden of Disease (GBD) database (GBD 2021 results),²¹ and the US National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) database (SEER 21 Registries, 2000–2022).²² The choice of 2014 as the breakpoint for ITS analysis was based on the subsequent national adoption of the 2013 USPSTF recommendations by Medicare and private insurers, representing the approximate onset of broad population-level exposure. CDC WONDER and GBD data have been extensively employed in lung cancer studies to evaluate incidence, mortality, and regional disease burden.^23,24 Between 2004 and 2022, SEER registry coverage increased from 28% to 47% of the US population, enhancing the representativeness of its lung cancer incidence and mortality data^25,26. The study cohort comprised patients aged 55–79 years with tracheal, bronchial, or lung cancer identified in SEER using International Classification of Diseases for Oncology, Third Edition (ICD-O-3) topography codes C33–C34 (including C34.92) and in mortality datasets using underlying-cause International Classification of Diseases, Tenth Revision (ICD-10) codes C33–C34²⁰, with all datasets providing population-level incidence and mortality metrics. For CDC WONDER, we analyzed national death certificate data; for GBD, we used country-level modeled estimates for incidence and mortality; and for SEER, we used registry-based incidence data with information on summary stage.

To address methodological disparities across databases, we harmonized CDC WONDER and GBD incidence and mortality data (1999–2021) with SEER stage-specific incidence data (2004–2022). Age-adjusted incidence and mortality rates (AAIR and AAMR per 100,000) were calculated using direct standardization, weighted by the GBD 2019 US reference population aged 55–79 years.²⁷ The mortality-to-incidence ratio (MIR) was calculated as an exploratory, population-level indicator of cancer lethality and access to effective care rather than a validated survival surrogate.¹⁶ Stratification variables included sex (CDC WONDER, GBD) and cancer stage (localized, regional, distant) alongside sex (SEER). SEER summary stage was classified as localized, regional, distant, or unknown; cases with unknown or missing stage were examined descriptively, and stage-specific incidence rates were calculated using known-stage cases in the denominator. Temporal variation in the proportion of unknown stage was evaluated to assess potential bias in apparent stage shifts.

This study relied exclusively on publicly available, de-identified, aggregate data from CDC WONDER, GBD, and SEER and did not involve direct contact with participants, individual-level records, human tissue, or protected health information. Therefore, institutional review board approval and informed consent were not required. Consistent with applicable guidance for analyses of publicly available de-identified data, no formal ethics approval was sought. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki, as applicable to research using anonymized secondary data. The data from CDC WONDER, GBD, and SEER were accessed on August 16,2025.

Statistical Analysis

Joinpoint regression modeled temporal trends (1999–2021), estimating average annual (AAPC) and annual (APC) percentage changes to identify inflection points in incidence, mortality, and MIR trajectories. Interrupted time series analysis (ITS) with a log-linear model assessed pre- and post-2014 trends: we fit ordinary least squares regression models to the natural logarithm of annual age-adjusted rates, using the specification log (y_t) = β₀ + β₁t + β₂I(t ≥ 2014) + β₃(t − t₀)I(t ≥ 2014) + ε_t, where t denotes calendar year, t₀= 2014, I(·) is an indicator function, β₂ captures the immediate level change at the breakpoint, and β₃ represents the change in slope after 2014. Population denominators were used to compute rates but were not included as offsets in the ITS models. Newey–West robust standard errors (lag=2) were applied to correct for observed positive autocorrelation (serial correlation) in the residuals, verified via Durbin-Watson tests and ACF1. We did not include time-varying covariates such as smoking prevalence, LDCT uptake, or dissemination of novel systemic therapies because comparable year-specific data stratified by age group, sex, geography, stage, and outcome were not consistently available across CDC WONDER, GBD, and SEER. For example, LDCT uptake data are typically available from surveys or claims-based sources rather than harmonized annual national series, smoking prevalence estimates are not directly linkable to lung cancer outcomes in these aggregate datasets, and treatment uptake for targeted therapy and immunotherapy cannot be uniformly captured across the study period. Therefore, ITS estimates reflect net temporal changes during the guideline era and should not be interpreted as fully adjusted causal effects of LDCT screening.

Counterfactual projections based on pre-2014 trends quantified the difference between observed cases or deaths and the numbers expected under continuation of pre-2014 trends during 2014–2019 and 2014–2021. Specifically, we used the pre-2014 segment of the ITS model to forecast post-2014 rates under a continuation of the prior trend, multiplied these predicted rates by age-specific population counts, and summed the differences between predicted and observed events to obtain the difference between projected and observed cases and deaths, with positive values indicating fewer observed events than expected under the pre-2014 trend. These differences capture the combined influence of all concurrent changes, including screening, treatment, smoking, and diagnostic practice, rather than the isolated effect of LDCT screening. Event study analyses with relative year dummies (k = year – 2014, k=−1 as reference) extended the ITS specification by replacing the single post-2014 slope term with indicator variables for each year before and after 2014, estimated annual effects with Newey–West 95% CIs, and were used to visualize the timing and magnitude of deviations from pre-2014 trends. Sensitivity analyses included alternate cutoff years (2013/2015), placebo tests (2007/2009/2011), and exclusion of 2020–2021 data, with results aligning with primary findings.

The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline for observational studies.²⁸ The completed STROBE checklist was submitted as supplementary material not for review and not for publication. All patient-level details were de-identified in the source databases, and no identifiable individual information was accessed. All analyses were performed using R 4.5.0, with statistical significance set at p<0.05 (two-tailed).

Results

Current Status and Trends in Lung Cancer Incidence and Mortality

Between 1999 and 2021, lung cancer AAIR and AAMR in US adults aged 55–79 years exhibited a pronounced downward trend, with a consistent inflection point occurring during 2013–2015 followed by an accelerated decline (Figure 1; eTable 1). Although males maintained higher incidence and mortality rates than females, their steeper decline rates progressively narrowed this gender disparity. The MIR as an exploratory population-level indicator also decreased during this period: CDC WONDER data showed an overall AAPC of −1.18% (−1.39 to −1.00), with greater declines among females (−1.36%) than males (−0.98%), while GBD data indicated a more modest overall decrease of −0.20% (−0.22 to −0.18), most pronounced during 2013–2019 (Figure 1; eTable 2). Notably, while age-adjusted rates declined, absolute case counts increased slightly due to demographic shifts, whereas death counts decreased in CDC WONDER data but remained stable in GBD records (Table 1). Despite differences in magnitude and inflection timing between datasets, both consistently demonstrate a long-term decline with accelerated reduction post-2013, with 2020–2021 fluctuations not altering this fundamental trend.

Figure 1.

Joinpoint analysis of AAIR, AAMR, and MIR trends (1999–2021). AAIR for both (A), female (B), and male (C); AAMR for both (D), female (E), and male (F); MIR for both (G), female (H), and male (I). Abbreviation: AAIR, age-adjusted incidence rates; AAMR, age-adjusted mortality rates; CDC WONDER, Centers for Disease Control and Prevention Wide-Ranging Online Data for Epidemiologic Research; GBD, Global Burden of Disease

Table 1.

AAR in Incidence and Mortality for Lung Cancer in the U.S. (1999–2021)

		1999		2021		AAPC
		Counts	AAR	Counts	AAR	AAPC
CDC WONDER	Incidence
	Both	116108	294.94 (293.23, 296.65)	128729	198.47 (197.38, 199.55)	-1.85 (-2.05 to -1.64)
	Male	67496	385.50 (382.58, 388.42)	65333	215.07 (213.42, 216.72)	-2.71 (-2.91 to -2.51)
	Female	48612	222.79 (220.78, 224.80)	63396	184.19 (182.75, 185.63)	-0.95 (-1.17 to -0.69)
	Mortality
	Both	90512	225.24 (223.76, 226.72)	76906	116.07 (115.25, 116.89)	-3.03 (-3.10 to -2.97)
	Male	54180	304.37 (301.80, 306.95)	42099	135.94 (134.64, 137.24)	-3.66 (-3.75 to -3.58)
	Female	36332	162.38 (160.68, 164.07)	34807	98.72 (97.68, 99.75)	-2.30 (-2.37 to -2.22)
GBD	Incidence
	Both	154281	306.32 (304.77, 307.86)	162343	181.93 (181.04, 182.82)	-2.37 (-2.45 to -2.28)
	Male	90233	398.34 (395.72, 400.95)	87514	209.70 (208.31, 211.09)	-2.89 (-2.96 to -2.82)
	Female	64048	231.43 (229.60, 233.25)	74829	157.61 (156.48, 158.74)	-1.76 (-1.88 to -1.65)
	Mortality
	Both	120753	238.43 (237.07, 239.79)	120626	135.04 (134.28, 135.80)	-2.58 (-2.67 to -2.5)
	Male	72509	319.17 (316.83, 321.50)	67276	161.23 (160.01, 162.44)	-3.10 (-3.17 to -3.03)
	Female	48244	172.77 (171.20, 174.34)	53350	112.08 (111.13, 113.03)	-1.99 (-2.07 to -1.91)

Abbreviation: CDC WONDER, Centers for Disease Control and Prevention Wide-Ranging Online Data for Epidemiologic Research; GBD, Global Burden of Disease; AAPC, average annual percent change; AAR, age- adjusted rate.

Interrupted Time Series Analysis

Using 2014 as the breakpoint, ITS analysis revealed divergent patterns between datasets. For incidence rate, CDC WONDER data showed minimal immediate change at the breakpoint, whereas GBD data exhibited an abrupt decline (approximately −4% to −6%, particularly among females). Mortality rates in both datasets demonstrated significant immediate reductions (CDC WONDER: −3% to −4% across genders), with post-breakpoint slopes steepening substantially (eTable 3). CDC WONDER’s AAMR decline accelerated from pre-breakpoint rates of −2.2%, −1.3%, and −2.9% annually to −4.9%, −4.5%, and −5.3%, while GBD’s acceleration was more modest (from −2.4%, −1.7%, and −3.2% to −2.9%, −2.9%, and −3.5%). AAIR declines also intensified post-2014: CDC WONDER rates shifted from −1.2%, −0.3%, and −2.1% to −3.2%, −2.4%, and −3.9%, while GBD rates progressed from −2.1%, −1.4%, and −2.7% to −3.2%, −2.8%, and −3.4% (Figure 2; eTable 4). MIR trends paralleled these findings, with CDC WONDER showing immediate −2% to −5% reductions followed by slope changes from −0.9%, −1.0%, and −0.8% to −1.8%, −2.2%, and −1.4%, while GBD exhibited minimal immediate effects with subsequent −0.1% annual declines (eFigure 1; eTable 4).

Figure 2.

Immediate level changes and post-2014 slope changes in lung cancer rates. ITS for AAIR (A); ITS for AAMR (B). Abbreviation: CDC WONDER, Centers for Disease Control and Prevention Wide-Ranging Online Data for Epidemiologic Research; GBD, Global Burden of Disease

Newey–West corrected analyses (lag=2) confirmed these patterns while addressing autocorrelation. AAIR maintained its slope acceleration without immediate breakpoint effects in CDC WONDER (additional −2% annual decline across groups), while GBD data showed breakpoint declines (especially among females) with modest subsequent slope changes. AAMR demonstrated immediate reductions with accelerated declines in CDC WONDER, whereas GBD showed weaker post-breakpoint slope changes. Residual diagnostics indicated positive autocorrelation, though corrected estimates aligned with primary findings (eTable 5). Event study plots revealed no pre-2014 anticipatory effects, with post-2014 deviations from pretrend emerging within 1–2 years and accumulating to approximately 15%–20% lower mortality in CDC WONDER by years 6–7, alongside 10%–18% lower incidence primarily through slope changes. GBD showed immediate mortality reductions (greater in women) but less sustained acceleration, particularly for males (eFigure 2).

Counterfactual analyses showed that observed deaths during 2014–2021 were lower than projections based on pre-2014 trends by 28,168 in CDC WONDER and 6,906 in GBD, supporting the presence of accelerated mortality declines despite database discrepancies (Figure 3; eTable 6). These differences represent the gap between the observed number of deaths and the number predicted had pre-2014 trends continued unchanged, and therefore capture the combined influence of all concurrent changes rather than the isolated effect of LDCT screening. Since the policy change in 2014, both CDC WONDER and GBD data have shown that lung cancer incidence and mortality rates have significantly deviated from counterfactual predictions. Although the databases differ in magnitude, they consistently indicate an accelerated post-2014 decline across sexes (eFigure 3). The CDC WONDER showed pronounced MIR improvements (average ΔMIR 2014–2021: 0.043 overall, 0.045 females, 0.040 males), contrasting with nonsignificant GBD changes (−0.015, −0.016, −0.014) (eTable 7). Among US adults aged 55–79 years, lung and tracheal cancer mortality rates exhibited an immediate post-2014 decline that progressively accelerated, with incidence rates following a similar accelerated pattern and CDC WONDER MIRs demonstrating significant reductions in our exploratory analyses.

Figure 3.

Observed lung cancer incidence and mortality compared with counterfactual projections based on pre-2014 trends. CDC WONDER cohort incidence (A) and mortality (B); GBD cohort incidence (C) and mortality (D). Abbreviation: CDC WONDER, Centers for Disease Control and Prevention Wide-Ranging Online Data for Epidemiologic Research; GBD, Global Burden of Disease

Shifting the cutoff point to 2013 or 2015 with 0–2 year lags yielded consistent results: both datasets revealed immediate mortality rate declines that accelerated after the alternative cutoff points, with incidence rates showing comparable acceleration. The gender-specific patterns persisted, with MIR improvements largely confined to the CDC WONDER. Model variations remained within 1–2 percentage points or percentage points per year of the primary analysis (eTable 8; eFigure 4). Placebo tests using 2007/2009/2011 cutoff points showed no strata matching the 2014 breakpoint’s magnitude or slope steepness, with estimates clustering near zero and lacking directional consistency—findings that reduce the likelihood that our results are driven by spurious structural breaks, while still falling short of establishing causal effects (eTable 9; eFigure 5).

SEER Staging Analysis

SEER staging trends since 2004 revealed declining distant metastasis AAIR alongside slowly increasing localized and decreasing regional rates, with synchronous shifts around 2014: localized cases rose from 20% to 30%, distant metastases fell from 56% to <50%, and regional cases declined slightly (Figure 4). Joinpoint analysis confirmed these patterns, with distant metastasis APC accelerating to −3.8% (−4.6 to −3.3) post-2013, localized stage showing 2014–2017 APC increases of 7% to 9%, and regional stage declines intensifying after 2016 (eTable 10; eFigure 6). ITS quantification using 2014 as the breakpoint revealed immediate −1.5% to −2.0% distant metastasis reductions with subsequent −1.5% to −2.5% annual declines, localized stage jumps of +10% followed by +2% to +4% sustained increases, and regional stage slope changes to −2% annually (eFigure 7). Counterfactual analysis showed 4,333 fewer distant-stage cases and 27,093 more localized-stage cases than expected under continuation of pre-2014 trends during 2014–2021 (eTable 11; eFigure 8). These patterns are consistent with a shift toward earlier-stage diagnosis at the population level; however, they must be interpreted cautiously given SEER’s geographic expansion over time, potential changes in staging and coding practices, and temporal variation in the proportion of unknown-stage cases. In addition, some fraction of the excess localized cases may reflect overdiagnosis of indolent tumors rather than clinically meaningful early detection, which cannot be distinguished using registry data alone. These staging shifts are consistent with population-level movement toward earlier-stage diagnosis during the guideline era, but they should not be interpreted as definitive evidence that LDCT screening alone caused improved survival. Registry data cannot distinguish clinically meaningful early detection from detection of indolent tumors, and stage shift does not necessarily equate to improved individual-level survival.

Figure 4.

Trends in AAIR and proportions by SEER summary stage. AAIR by stage (A); stage proportions among known-stage cases (B). AAIR, Age-adjusted incidence rates; SEER, Surveillance, Epidemiology, and End Results

Discussion

This study provided a comprehensive, cross-database evaluation of age-adjusted incidence and mortality trends for lung and tracheal cancers among U.S. adults aged 55–79 between 1999 and 2021, focusing on the period following the 2013 USPSTF LDCT recommendation. Our primary contribution is the integration of independent data sources (CDC WONDER, GBD, SEER) within a unified quasi-experimental framework (ITS and counterfactual analysis), which allowed us to concurrently examine accelerated mortality/incidence declines, shifts in stage distribution, and trends in the exploratory MIR. The analysis revealed an overall decline in both incidence and mortality rates, with an accelerated reduction occurring between 2013 and 2015—particularly for mortality after 2014—indicating a notable temporal change in population-level lung cancer trends. This temporal pattern coincides with the 2013 USPSTF recommendation for LDCT screening in high-risk populations and the subsequent integration of targeted therapies and immunotherapies into clinical practice. Stage-specific SEER data corroborate these findings, demonstrating increased detection of localized disease alongside decreased distant-stage diagnoses post-2014. Together, these observations suggest that population-level lung cancer outcomes improved during the USPSTF LDCT guideline era. The direction of the findings is compatible with a meaningful real-world contribution of LDCT screening, in line with randomized trial evidence, but the observed trends likely reflect the combined influence of screening uptake, tobacco control, diagnostic intensity, and therapeutic innovation. Therefore, the results should be interpreted as hypothesis-generating temporal associations rather than evidence that the USPSTF recommendation or LDCT screening alone caused the observed changes.

Despite overall declining rates, demographic shifts contributed to rising case numbers in GBD datasets, whereas CDC WONDER showed substantial mortality reductions not reflected in GBD records. These divergent patterns suggest database-specific variations in population coverage and mortality ascertainment methods may complicate cross-study comparisons.²⁹ The larger absolute declines in mortality inferred from CDC WONDER compared with GBD likely reflect differences in source data (vital registration versus statistical modeling), age standardization, and handling of incomplete or non-specific cause-of-death codes. Accordingly, our primary inferences emphasize the direction and timing of trends across data sources, ensuring the consistency of the observed inflection point, rather than solely relying on the exact magnitudes of AAPCs.

Male lung cancer mortality consistently exceeded female rates throughout the study period, though the disparity diminished progressively after 2014 as male rates declined more sharply.^14,30 This convergence may reflect multiple overlapping factors, including historical differences in smoking cessation patterns, changing eligibility for screening, gradual uptake of early detection technologies, and sex-specific differences in access to or response to modern therapies.³¹ Historical smoking patterns further explain these trends: while male smoking prevalence peaked earlier and declined steadily, female rates remained elevated until the 1980s, delaying corresponding mortality reductions in women. MIR trends additionally appeared more favorable among men in CDC WONDER, which is consistent with—but does not prove—faster improvements in prognosis or access to effective therapies in males compared with females.^32,33 Although the gender gap continues to narrow, its persistence highlights the need for sex-specific prevention strategies—particularly in optimizing screening accessibility and therapeutic interventions for high-risk women.^34,35 Future investigations should elucidate how biological, behavioral, and healthcare utilization factors mediate sex disparities in lung cancer outcomes to inform precision public health approaches.

Robustness analyses with Newey–West corrections supported the presence of post-2014 changes in trend despite residual autocorrelation, although these changes should be interpreted as temporal associations rather than proof of a single causal intervention. Post-2014, CDC WONDER revealed a significant immediate decline in mortality rates, which further accelerated after this inflection point.³⁶ This finding is consistent with 2014 being a period when multiple factors—including dissemination of LDCT screening guidelines, continued reductions in smoking, and rapid uptake of targeted and immune therapies—were acting simultaneously. Counterfactual analyses further illustrated that observed post-2014 mortality counts were lower than expected under continuation of pre-2014 trends, particularly in CDC WONDER.^37,38 These differences should be understood as descriptive deviations from projected trends and not as deaths causally prevented by LDCT screening alone. Comparisons with counterfactual scenarios suggest that this decline is substantially different from what would have been expected had earlier trends continued, but they do not isolate the effect of LDCT from other concomitant changes. Although both CDC WONDER and GBD data generally agree on the accelerated decline post-2014, differences in the magnitude of the change warrant consideration. Disparities in case identification methods, statistical modeling approaches, and survival reporting between these datasets may contribute to some discrepancies.³⁹ For instance, GBD data are derived from statistical modeling of cancer incidence and mortality rates, adjusted for global population structure and based on cross-country comparisons. In contrast, CDC WONDER cancer data are collected through national cancer registries, healthcare facility reports, death certificates, and other sources, processed and updated according to strict standardization procedures, primarily aimed at supporting public health research and policy decision-making. Although the two datasets have different focal points, they demonstrate high consistency at key time points, enhancing the credibility of research conclusions. These differences emphasize the need for standardized data collection and reporting to ensure the reliability of future trend analyses.

Several contemporaneous factors may have contributed to the observed post-2014 changes. First, long-term declines in cigarette smoking in the United States would be expected to reduce lung cancer incidence and mortality, especially among birth cohorts with earlier and more pronounced reductions in tobacco exposure. Second, LDCT screening uptake among eligible adults increased gradually after the USPSTF recommendation and insurance coverage expansion, but uptake remained incomplete and varied by region, health system, insurance status, and access to preventive care. Third, the study period coincided with major improvements in systemic therapy, including targeted therapies for molecularly defined lung cancer and immune checkpoint inhibitors, which may have improved survival independently of screening. Because harmonized year-specific data on smoking prevalence, LDCT use, and treatment uptake were not available across all data systems and strata analyzed here, the relative contribution of these factors could not be separated in our models. These factors may bias the magnitude of the estimated post-2014 trend changes if their timing coincided with the policy era.

Staging analysis of the SEER database revealed significant shifts in lung cancer staging post-2014, particularly with an increase in the proportion of localized cases and a decrease in distant metastasis cases. This change may be attributed in part to the widespread adoption of early screening technologies, particularly LDCT screening.^40,41 The rise in localized cases is consistent with increased detection of earlier-stage lung cancers, but this pattern requires cautious interpretation. LDCT screening can detect clinically important early cancers, but it can also identify indolent lesions that may not have progressed to symptomatic or fatal disease during a patient’s lifetime.⁴² Therefore, a shift toward localized-stage diagnosis does not necessarily translate into improved individual-level survival and cannot be interpreted as direct proof of screening benefit using registry data alone. Counterfactual analysis showed approximately 27,000 more localized-stage cases and more than 4,000 fewer distant-stage cases than expected under continuation of pre-2014 trends, reflecting a population-level change in stage distribution. However, these estimates should be interpreted as deviations from projected trends rather than as cases directly caused or prevented by LDCT screening. At the same time, LDCT screening is known to be susceptible to overdiagnosis, and a substantial fraction of the additional localized cases may represent indolent tumors that would not have become clinically significant in the absence of screening.^43,44 Furthermore, the observed stage shifts must be interpreted cautiously due to known structural limitations, including SEER’s geographic expansion, evolving staging and coding practices, and temporal variation in the proportion of unknown-stage cases over time, all of which may contribute to the observed stage distribution. The absence of individual-level data on LDCT use prevents us from directly linking stage shifts to screening participation. However, despite the decline in distant metastasis cases, mortality rates remain unstable in certain male subgroups, suggesting that the effectiveness of screening and treatment may vary by gender or region, warranting further attention in future public health strategies.

Although this study provides a comprehensive analysis of trends in lung and tracheal cancer mortality and incidence using multi-source data, it has certain limitations. First, the data were sourced from public databases such as CDC WONDER, GBD, and SEER. While efforts were made to standardize the data using a consistent age structure, differences in data collection methods and reporting across these datasets may have impacted cross-source comparisons. Second, our ITS and event-study models did not incorporate time-varying covariates such as smoking prevalence, LDCT uptake, diagnostic intensity, or adoption of novel systemic therapies. These factors are major determinants of lung cancer incidence, stage distribution, and mortality, and their temporal overlap with the USPSTF guideline era may bias both the magnitude and interpretation of the estimated trend changes. As a result, the estimated changes represent aggregate population-level temporal patterns and should not be interpreted as causal effects of screening alone. Third, while counterfactual analysis offers a useful descriptive tool for visualizing deviations from pre-existing trends, it relies on model-based extrapolation and assumes that pre-2014 trends would have continued in the absence of subsequent changes. This assumption may not hold in the presence of ongoing tobacco control, evolving diagnostic practices, treatment advances, and demographic change. Therefore, the counterfactual estimates capture a composite difference between observed and projected outcomes rather than the number of deaths or cases directly prevented by LDCT screening. Fourth, MIR was used only as an exploratory population-level summary; it does not directly measure survival and may be affected by changes in diagnostic intensity and coding. Fifth, SEER stage information is subject to missingness, potential misclassification, geographic expansion of registry coverage, and changes in staging or coding practice over time. Registry data also cannot distinguish clinically meaningful early detection from overdiagnosis of indolent tumors. Accordingly, the observed increase in localized-stage disease should be viewed as a population-level stage distribution shift rather than definitive evidence of individual-level survival improvement. Lastly, although the sensitivity analysis addressed the impact of the COVID-19 pandemic, the effects of COVID-19 on policy implementation and the management of lung cancer treatments warrant further in-depth research. Future studies should explore disparities across populations and regions, particularly with regard to screening effectiveness across racial, gender, and socioeconomic groups. As data accumulation increases and screening technologies evolve, extended trend analyses and causal inference models will improve the accuracy of long-term assessments of lung cancer prevention and control strategies.

Conclusions

In summary, this study identified accelerated declines in lung and tracheal cancer mortality and incidence rates among US adults aged 55–79 years during the USPSTF LDCT guideline era, concurrent with a shift toward localized-stage diagnosis. These findings are temporally consistent with real-world benefits of LDCT screening but do not establish that LDCT screening alone caused the observed changes. The observed patterns likely reflect the combined effects of tobacco control, evolving treatment, diagnostic practice, and gradual uptake of screening. Nevertheless, the convergence of declining rates, shifting stage distribution, and favorable MIR patterns across multiple data systems underscores the importance of maintaining comprehensive lung cancer control strategies, including evidence-based screening, smoking cessation, and equitable access to modern therapies. Persistent gender disparities and evolving mortality trends warrant further investigation to optimize prevention strategies.

Supplemental Material

Supplemental Material -Population-Level Trends in Lung Cancer Mortality and Stage Distribution During the USPSTF LDCT Guideline Era: A Multi-Database Quasi-Experimental Study

Supplemental Material for Population-Level Trends in Lung Cancer Mortality and Stage Distribution During the USPSTF LDCT Guideline Era: A Multi-Database Quasi-Experimental Study by Kaide Xia, Junwen Wang in Cancer Control

Footnotes

Acknowledgements

We thank the CDC, SEER, and the Institute for Health Metrics and Evaluation (IHME) for providing access to publicly available datasets.

ORCID iD

Junwen Wang

Ethical Considerations

Ethics approval and consent to participate: This study used publicly available, de-identified secondary data from SEER, CDC WONDER, and the Global Burden of Disease (GBD) study; therefore, it did not involve human subjects research as defined by applicable regulations. Institutional Review Board approval and informed consent were not required.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

CDC WONDER data are publicly available (https://wonder.cdc.gov). SEER incidence data are accessible via SEER*Stat (https://seer.cancer.gov/seerstat/). GBD 2021 results are publicly available (). Analysis code and aggregated outputs are available from the corresponding author upon reasonable request.

Disclaimer

The views expressed are those of the authors alone and do not necessarily reflect those of the Centers for Disease Control and Prevention (CDC), the National Cancer Institute (NCI)/National Institutes of Health (NIH), the Institute for Health Metrics and Evaluation (IHME), or any affiliated organizations.

Conflicts of AI

Generative AI (ChatGPT, OpenAI) was used only for grammar and wording edits. The authors reviewed and take full responsibility for all content.

Supplemental Material

Supplemental material for this article is available online.

Appendix

References

Jemal

Bray

Center

Ferlay

Ward

Forman

. Global cancer statistics. CA Cancer J Clin. 2011;61(2):69–90. doi:10.3322/caac.20107.

Siegel

Zou

Jemal

. Cancer statistics, 2014. CA Cancer J Clin. 2014;64(1):9–29. doi:10.3322/caac.21208.

US Preventive Services Task Force Krist

Davidson

, et al. Screening for lung cancer: US preventive services task force recommendation statement. JAMA. 2021;325(10):962–970. doi:10.1001/jama.2021.1117.

Moyer

. US Preventive Services Task Force. Screening for lung cancer: US Preventive Services Task Force recommendation statement. Ann Intern Med. 2014;160(5):330-338. doi:10.7326/M13-2771

National Lung Screening Trial Research Team Aberle

Adams

, et al. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med. 2011;365(5):395–409. doi:10.1056/NEJMoa1102873.

de Koning

van der Aalst

de Jong

, et al. Reduced lung-cancer mortality with volume CT screening in a randomized trial. N Engl J Med. 2020;382(6):503–513. doi:10.1056/NEJMoa1911793.

Mulshine

D’Amico

. Issues with implementing a high-quality lung cancer screening program. CA Cancer J Clin. 2014;64(5):352–363. doi:10.3322/caac.21239.

Goodwin

Nishi

Zhou

Kuo

. Use of the shared decision-making visit for lung cancer screening among medicare enrollees. JAMA Intern Med. 2019;179(5):716–718. doi:10.1001/jamainternmed.2018.6405.

Seiler

Malcarney

Horton

Dafflitto

. Coverage of clinical preventive services under the affordable care act: From law to access. Public Health Rep. 2014;129(6):526–532. doi:10.1177/003335491412900611.

10.

Internal Revenue Service, Department of the Treasury . Employee Benefits Security Administration, Department of Labor, Centers for Medicare & Medicaid Services, Department of Health and Human Services. Coverage of certain preventive services under the affordable care act. Final rules. Fed Regist. 2013;78(127):39869–39899.

11.

Jonas

Reuland

Reddy

, et al. Screening for lung cancer with low-dose computed tomography: Updated evidence report and systematic review for the US preventive services task force. JAMA. 2021;325(10):971–987. doi:10.1001/jama.2021.0377.

12.

Cornelius

Loretan

Jamal

, et al. Tobacco product use among adults - united states, 2021. MMWR Morb Mortal Wkly Rep. 2023;72(18):475–483. doi:10.15585/mmwr.mm7218a1.

13.

Henley

Ward

Scott

, et al. Annual report to the nation on the status of cancer, part I: National cancer statistics. Cancer. 2020;126(10):2225–2249. doi:10.1002/cncr.32802.

14.

Jani

Singh

Abdallah

, et al. Trends in lung cancer incidence and mortality (1990-2019) in the united states: A comprehensive analysis of gender and state-level disparities. JCO Glob Oncol. 2023;9:e2300255. doi:10.1200/GO.23.00255.

15.

Islami

Ward

Sung

, et al. Annual report to the nation on the status of cancer, part 1: National cancer statistics. J Natl Cancer Inst. 2021;113(12):1648–1669. doi:10.1093/jnci/djab131.

16.

Choi

Lee

Nhung

, et al. Cancer mortality-to-incidence ratio as an indicator of cancer management outcomes in organization for economic cooperation and development countries. Epidemiol Health. 2017;39:e2017006. doi:10.4178/epih.e2017006.

17.

Asadzadeh Vostakolaei

Karim-Kos

Janssen-Heijnen

MLG

Visser

Verbeek

ALM

Kiemeney

LALM

. The validity of the mortality to incidence ratio as a proxy for site-specific cancer survival. Eur J Public Health. 2011;21(5):573–577. doi:10.1093/eurpub/ckq120.

18.

Ellis

Belot

Rachet

Coleman

. The mortality-to-incidence ratio is not a valid proxy for cancer survival. J Glob Oncol. 2019;5:1–9. doi:10.1200/JGO.19.00038.

19.

Fan

Jiang

Gong

, et al. Epidemiological and demographic drivers of lung cancer mortality from 1990 to 2019: Results from the global burden of disease study 2019. Front Public Health. 2023;11:1054200. doi:10.3389/fpubh.2023.1054200.

20.

Centers for Disease Control and Prevention . CDC WONDER. Published online June 13, 2025. https://wonder.cdc.gov/

21.

GBD 2021 Diseases and Injuries Collaborators . Global incidence, prevalence, years lived with disability (YLDs), disability-adjusted life-years (DALYs), and healthy life expectancy (HALE) for 371 diseases and injuries in 204 countries and territories and 811 subnational locations, 1990-2021: A systematic analysis for the global burden of disease study 2021. Lancet. 2024;403(10440):2133-2161. doi:10.1016/S0140-6736(24)00757-8.

22.

National Cancer Institute, DCCPS, Surveillance Research Program . Surveillance, Epidemiology, and End Results (SEER). Incidence - SEER Research Limited-Field Data, 21 Registrie. Published online April 2025. Program. www.seer.cancer.gov)SEER*StatDatabase

23.

Jani

Marshall

Singh

, et al. Lung cancer mortality in europe and the USA between 2000 and 2017: An observational analysis. ERJ Open Res. 2021;7(4):00311–02021. doi:10.1183/23120541.00311-2021.

24.

Jani

Kareff

Morgenstern-Kaplan

, et al. Evolving trends in lung cancer risk factors in the ten most populous countries: An analysis of data from the 2019 global burden of disease study. EClinicalMedicine. 2025;79:103033. doi:10.1016/j.eclinm.2024.103033.

25.

Goldstraw

Chansky

Crowley

, et al. The IASLC lung cancer staging project: Proposals for revision of the TNM stage groupings in the forthcoming (eighth) edition of the TNM classification for lung cancer. Journal of Thoracic Oncology. 2016;11(1):39–51. doi:10.1016/j.jtho.2015.09.009.

26.

Bytnar

McGlynn

Nealeigh

Shriver

Zhu

. Cancer incidence in the US military: An updated analysis. Cancer. 2024;130(1):96–106. doi:10.1002/cncr.34978.

27.

Global Burden of Disease Collaborative Network . Global Burden of Disease Study 2019 (GBD 2019) Population Estimates 1950-2019. Published online 2020. doi:10.6069/7EGY-0354.

28.

von Elm

Altman

Egger

, et al. The strengthening the reporting of observational studies in epidemiology (STROBE) statement: Guidelines for reporting observational studies. J Clin Epidemiol. 2008;61(4):344-349. doi:10.1016/j.jclinepi.2007.11.008.

29.

Minhas

AMK

Al-Kindi

Van Spall

HGC

Abramov

. Comparing cardiovascular mortality estimates from global burden of disease and from the centers for disease control and prevention wide-ranging online data for epidemiological research. Circ Cardiovasc Qual Outcomes. 2025;18(5):e011459. doi:10.1161/CIRCOUTCOMES.124.011459.

30.

Liu

Chen

Liu

Xia

. Gender disparities in lung cancer incidence in the united states during 2001–2019. Sci Rep. 2023;13(1):12581. doi:10.1038/s41598-023-39440-8.

31.

Luo

Wampfler

, et al. 5-year overall survival in patients with lung cancer eligible or ineligible for screening according to US preventive services task force criteria: A prospective, observational cohort study. Lancet Oncol. 2019;20(8):1098-1108. doi:10.1016/S1470-2045(19)30329-8.

32.

Randhawa

Sferra

Das

Kaiser

Erkmen

. Examining gender differences in lung cancer screening. J Community Health. 2020;45(5):1038-1042. doi:10.1007/s10900-020-00826-8.

33.

May

Shows

Nana-Sinkam

Landry

. Sex differences in lung cancer. Cancers. 2023;15(12):3111. doi:10.3390/cancers15123111.

34.

Sagerup

CMT

Småstuen

Johannesen

Helland

Brustugun

. Sex-specific trends in lung cancer incidence and survival: A population study of 40 118 cases. Thorax. 2011;66(4):301-307. doi:10.1136/thx.2010.151621.

35.

DeBolt

Harris

. The impact of social determinants of health on gender disparities within respiratory medicine. Clinics in Chest Medicine. 2021;42(3):407-415. doi:10.1016/j.ccm.2021.04.003.

36.

Siegel

Miller

Fuchs

Jemal

. Cancer Statistics, 2021. CA Cancer J Clin. 2021;71(1):7-33. doi:10.3322/caac.21654.

37.

Khouzam

Wood

Vigneswaran

, et al. Impact of federal lung cancer screening policy on the incidence of early-stage lung cancer. The Annals of Thoracic Surgery. 2023;115(4):827-833. doi:10.1016/j.athoracsur.2022.11.021.

38.

Hoffman

Atallah

Struble

Badgett

. Lung cancer screening with low-dose CT: A meta-analysis. J GEN INTERN MED. 2020;35(10):3015-3025. doi:10.1007/s11606-020-05951-7.

39.

Mattiuzzi

Lippi

. Cancer statistics: A comparison between world health organization (WHO) and global burden of disease (GBD). Eur J Public Health. 2020;30(5):1026-1027. doi:10.1093/eurpub/ckz216.

40.

Pinsky

Miller

. Use and outcomes of low-dose CT scan lung cancer screening in the medicare population. CHEST. 2022;162(3):721-729. doi:10.1016/j.chest.2022.03.031.

41.

Potter

Bajaj

Yang

CFJ

. The 2021 USPSTF lung cancer screening guidelines: A new frontier. The Lancet Respiratory Medicine. 2021;9(7):689-691. doi:10.1016/S2213-2600(21)00210-1.

42.

Shin

Yoon

Marwaha

. Progress in the treatment and outcomes for early-stage non-small cell lung cancer. Lung. 2018;196(3):351-358. doi:10.1007/s00408-018-0110-1.

43.

Singareddy

Flanagan

Samson

, et al. Trends in stage I lung cancer. Clin Lung Cancer. 2023;24(2):114-119. doi:10.1016/j.cllc.2022.11.005.

44.

Rossetti

Eaton

Tohmasi

, et al. Early-stage lung cancer in the era of lung cancer screening: Veterans health administration outperforms other insurance models. The Annals of Thoracic Surgery. 2024;118(5):1117-1125. doi:10.1016/j.athoracsur.2024.07.041.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

1.69 MB