Stage shifts in national lung adenocarcinoma and the impact of opportunistic self-initiated LDCT screening in Taiwan: a nationwide population-based cohort study

Summary

Background

Lung cancer in Taiwan is never-smoking and adenocarcinoma (LUAD) predominant. This study evaluated national LUAD stage-shifts and their association with opportunistic, self-initiated low-dose computed tomography (LDCT) screening.

Methods

A retrospective study was conducted using Taiwan Cancer Registry (TCR) data (2011–2023), and self-initiated LDCT screening records from five medical centers (2011–2021). Patients were classified as early-stage (stage I) and late-stage (stage II–IV). Outcomes included age-standardized incidence rate (ASIR), age-standardized incidence-based mortality (ASIBM), and average annual percentage change (AAPC) of national LUAD, and screening volume of self-initiated LDCT.

Findings

Among 126,753 LUAD cases 71,159 (56.1%) were females and 87,830 (69.3%) never smoked. The opportunistic self-initiated LDCT screening revealed 70,803 individuals, most never smoked (83.7%) and were male (56.6%). ASIRs of stage 0 (AAPC, 49.6%; 95% confidence interval [CI], 43.3–71.0) and early-stage (AAPC, 13.0%; 95% CI, 12.3–13.7) increased significantly. Conversely, late-stage ASIRs declined in selected subgroups, including never-smoking females aged <55 years (AAPC, −1.9%; 95% CI, −0.3 to −0.9) and in smoking patients aged ≥75 years (males AAPC, −2.3%; 95% CI, −3.1 to −1.5; females AAPC, −3.9%; 95% CI, −6.3 to −1.3). Rising LDCT screening volume descriptively paralleled this stage-shift in never-smoking females. ASIBM declined significantly over time, particularly in advanced-stage disease.

Interpretation

The coexistence of increasing early-stage incidence, largely stable late-stage disease, and declining incidence-based mortality suggests a transition phase characterized by both detection expansion and emerging clinical benefit. The increase in self-initiated LDCT screenings paralleled the stage-shift among never-smoking females, and further research is needed to assess long-term outcomes and cost-effectiveness.

Funding

Health Promotion Administration, Ministry of Health and Welfare; the Ministry of Health and Welfare and Formatting the Risk Prediction Models for Never-Smoking Lung Cancer (FORMOSA).

Research in context.

Evidence before this study

We searched PubMed for studies published between Jan 1, 2000, and Jan 25, 2026, using terms including “lung cancer screening,” “low-dose computed tomography (LDCT),” “never smoking,” and “stage shift.” Lung cancer in individuals who have never smoked (LCINS) is a growing global health burden, particularly among Asian females. While randomized trials like NLST and NELSON established the mortality benefit of LDCT screening in individuals who smoked heavily, evidence for those who had never smoked has been limited. In Taiwan, the prospective TALENT study recently demonstrated a high detection rate of lung cancer in never-smoking individuals, particularly those with a family history, providing crucial evidence for high-risk groups beyond smoking status. However, in many Asian countries, LDCT screening has largely been opportunistic and self-pay based. Recent population-based studies from South Korea and China raised concerns that such widespread screening leads to overdiagnosis—increasing early-stage incidence without reducing late-stage disease. Whether this holds true across all demographic subgroups, particularly when distinguishing between individuals with a history of smoking and those who had never smoked, remains unclear due to a lack of nationwide data with detailed smoking history.

Added value of this study

This is the first nationwide study to analyze lung adenocarcinoma stage-shifts stratified by detailed smoking history, and to ecologically compare them with real-world trends in self-initiated LDCT screening. We analyzed 126,753 adenocarcinoma cases from the Taiwan Cancer Registry (2011–2023) and 70,803 unique screened individuals from five major medical centers. We found a notable demographic mismatch: individuals seeking self-initiated screening were predominantly younger and without a history of smoking, whereas the absolute disease burden remains high in older populations. Despite this, a stage-shift characterized by a significant decline in late-stage incidence was observed specifically in never-smoking females aged <55 years and those who ever smoked aged ≥75 years. Our exploratory analysis revealed that the rise in screening volume descriptively paralleled the stage-shift observed in young never-smoking females. This ecological pattern contrasts with the stable late-stage incidence seen in most other subgroups, highlighting the heterogeneity of screening impacts.

Implications of all the available evidence

Our findings suggest that the massive uptake of opportunistic LDCT screening descriptively parallels a population-level stage-shift in younger never-smoking female, a group traditionally excluded from Western screening guidelines. However, the discordance between rising early-stage and stable late-stage disease across most other never-smoking subgroups strongly supports concerns regarding detection inflation and overdiagnosis. These findings underscore the critical need for developing tailored, risk-stratified screening strategies for never-smoking populations in Asia. Such strategies should integrate multi-dimensional risk factors beyond smoking status, such as family history or genetic susceptibility, to maximize the benefits of early detection while minimizing the harms of overdiagnosis and resource misallocation.

Introduction

Lung cancer is the leading cause of cancer-related mortality worldwide.¹ While smoking is a major risk factor, recent epidemiological data indicate approximately 30% of patients with lung cancer had never smoked.^2,3 In Taiwan, previous analysis based on the Taiwan Cancer Registry (TCR) revealed that between 1995 and 2015, 53% of patients with lung cancer had never smoked, with over 90% of female patients falling into this category.⁴ Additionally, among these patients who never smoked, high prevalence of lung adenocarcinoma (LUAD) (62%) were reported, and nearly 60% were diagnosed with stage IV disease.^4,5

Low-dose computed tomography (LDCT) is used for the early detection of lung cancer and has reduced lung cancer mortality among individuals who ever smoked.^6,7 Currently, based on the National Lung Screening Trial (NLST) criteria, fewer than 20% of patients with lung cancer in Taiwan qualify for screening—a common limitation in East Asia.⁴ This highlights the need for novel screening criteria tailored to populations with high lung cancer rates and low tobacco exposure. While our previous studies revealed a family history of lung cancer (LCFH) as an important risk factor for lung cancer in never-smoking individuals,^8,9 only 12.2–14.7% of never-smoking patients with lung cancer possess such a history.¹⁰ Although the specific etiologies for the remaining majority require further investigation, a more immediate clinical challenge is evaluating the impact of the widespread, opportunistic LDCT screening that is already occurring among these individuals.

Over the past two decades, across East Asia, several trends have occurred in lung cancer epidemiology, including declining smoking rates, a younger age at diagnosis, and a rising incidence of early-stage LUAD.^4,11 In Japan, a long-term growth trends has been reported in localized lung cancer from 1993 to 2015, which suggested the effectiveness of nationwide screening programs.¹² Meanwhile, recent population-based studies from South Korea and China have reported massive increases in early-stage lung cancer diagnoses following the unregulated uptake of opportunistic LDCT screening.^13,14 These surges in early-state disease were not accompanied by proportional declines in late-stage disease, raising concerns regarding overdiagnosis and detection inflation.

Globally, while several studies evaluating LDCT screening in never-smoking individuals have been conducted in countries such as China and Netherlands,^15,16 Taiwan remains uniquely positioned by pioneering a nationwide, government-reimbursed screening program for never-smoking individuals.^17,18 In Taiwan, LDCT screening was not reimbursed until July 2022 for both smoking and never-smoking populations.¹⁸ Prior to this, individuals seeking LDCT scans were required to pay out-of-pocket for opportunistic, commercial examinations. In East Asia, previous studies have raised concerns about lung cancer overdiagnosis resulting from LDCT screening among never-smoking individuals, particularly among females.^13,14,19 Recent studies investigating hospital-based LDCT volumes and local cancer registry data have further suggested that screening uptake may contribute to observed increases in early-stage diagnoses and gender disparities in Taiwan.²⁰ Conversely, other studies have reported that increased detection of early-stage disease following LDCT screening may be associated with improved survival, consistent with a stage-shift effect.¹¹ A key issue is that changes in stage distribution—such as an increasing proportion of stage I disease and a decreasing proportion of stage IV disease—do not necessarily indicate a true reduction in advanced disease burden. Such patterns may reflect either a stage-shift or detection inflation. Therefore, evaluation based on stage proportions alone is insufficient to distinguish between these mechanisms. Accordingly, examining how opportunistic LDCT uptake is associated with stage-specific incidence patterns is essential for interpreting these trends and informing the development of risk-stratified screening strategies for never-smoking individuals.

In this regional context, Taiwan provides a valuable setting for investigation. While facing a high burden of lung cancer among never-smoking individuals, Taiwan has a comprehensive national cancer registry with detailed smoking information to our knowledge, no prior nationwide study has examined trends in lung adenocarcinoma (LUAD) by detailed smoking history in parallel with real-world opportunistic LDCT screening volumes in Taiwan.^11,19, 20, 21 This retrospective study aims to characterize national stage-specific trends in LUAD and to evaluate whether observed changes are consistent with a true stage-shift, reflected by a reduction in advanced-stage incidence, or with increased detection of early-stage disease without a corresponding decline in late-stage incidence.

Methods

Patients and study design

This retrospective study comprised three cohorts: two national databases from Taiwan for incidence and mortality, and a third database for self-initiated LDCT screening. First, to examine trends in the incidence of lung cancer and LUAD from 2011 to 2023, data were obtained from the TCR. TCR is a database that records all cancer cases in Taiwan, with the completeness exceeded 98% and a detailed smoking history.^22,23 Lung cancer cases were identified using the International Classification of Diseases for Oncology, Third Edition (ICD-O-3 C34), and histological subtypes were classified based on morphology codes (Supplementary Text 1; Table S1).

Second, the lung cancer mortality from 2011 to 2023 was estimated using death records obtained from the Taiwan National Death Certificates Database (TNDCD) (Supplementary text 2). Causes of death were classified according to the International Classification of Diseases, 10th Revision. Inclusion was restricted to individuals whose primary underlying cause of death was coded as lung cancer (ICD-10 C34).

Third, this study explored the characteristics and temporal trends of individuals who underwent opportunistic self-initiated LDCT screening. Real-world data from 5 medical centers between 2011 and 2021 were analyzed to assess the uptake and patterns of self-initiated LDCT screening (Supplementary Text 3).

Ethics approval

The study protocol was approved by the Institutional Review Board of the National Taiwan University Research Ethics Committee, Chung Shan Medical University Hospital, Taichung Veterans General Hospital, China Medical University Hospital, Kaohsiung Medical University Hospital, and Shuang Ho Hospital (IRB No. 202405HM031, CS1-24164, KMUHIRB-E(I)-20240431, CE24544A, CMUH110-REC1-244, and N202503013). The requirement for informed consent was waived.

Data records and definition of variables

The clinical variables from TCR and TNDCD analyzed included date of diagnosis, age at diagnosis, sex, histological subtype, lung cancer stage, smoking history, age at death, and data of death. Lung cancer staging was determined according to the American Joint Committee on Cancer (AJCC) guidelines, using 7th edition before 2018 and the 8th edition thereafter.^24,25 Data on self-initiated LDCT screening were obtained from participating hospitals and were available at the aggregate level, including distributions of screened individuals by age, sex, and smoking history. To enable comparison between datasets with different data structures, analyses were conducted at the ecological (aggregate) level. Data from TCR/TNDCD and LDCT screening were aligned across four common strata: calendar year, sex, age group (<55, 55–64, 65–74, and ≥75 years), and smoking status (never-smoked vs ever-smoked).

Primary trend analyses focused on LUAD. For stage-specific classifications, Patients with stage I disease were classified as early-stage disease, whereas those with stages II–IV were categorized as late-stage, consistent with recent epidemiological classifications of screening-detected cohorts.^13,19 Stage 0 (adenocarcinoma in situ, AIS) was analyzed separately. Regarding smoking history, patients were categorized into those who had never smoked and those who had ever smoked. Patients who had never smoked were defined as individuals who had never smoked in their lifetime; all others were classified as ever smoked. The ever-smoking cohort was further subdivided by cumulative smoking intensity (<20 and ≥ 20 pack-years) to present detailed baseline demographic profiles.

Statistical analysis

The crude incidence rate (CR) and age-standardized incidence rate (ASIR) of LUAD and lung cancer were calculated using data from the TCR database. Subgroup analysis was performed to compare LUAD incidence by age, sex, and smoking history. Additionally, the characteristics and the annual number of unique individuals undergoing self-initiated LDCT screenings were evaluated. For the screening volume analysis, repeat annual screens were not counted, and the index date was defined by the date of screening of each individual. The age-standardized mortality rate (ASMR) of lung cancer was derived from the TNDCD. To better capture temporal changes in disease-specific mortality while accounting for underlying incidence, age-standardized incidence-based mortality (ASIBM) was calculated using a fixed 5-year look-back approach and analyses of temporal trends were restricted to deaths occurring from 2015 to 2023 (Supplementary text 4). All age-standardized rates were computed using the direct method, based on the 2000 World Health Organization world standard population²⁶ and were expressed as cases per 100,000 population. Trends in ASIR, ASMR, and ASIBM of each LUAD subgroup were assessed using the average annual percentage change (AAPC), and the significant changes evaluated through Joinpoint regression analysis (Supplementary text 5). The LUAD stage-shifts were paralleled with real-world, multicenter opportunistic LDCT screening volumes by exploratory scatter plot without claiming statistical significance. Statistical analyses were conducted using SAS software (version 9.4; SAS Institute Inc), and AAPC estimates were obtained using the Joinpoint Regression Program (version 5.1.0.0, April. 2024; Statistical Research and Application Branch, National Cancer Institute).

Role of the funding source

The funders had no role in study design, data collection and analysis, the decision to publish, or preparation of the manuscript.

Results

Characteristics of lung cancer and LUAD in Taiwan and screened individuals of self-initiated LDCT screening in 5 medical centers

Between 2011 and 2023, a total of 190,715 patients with the diagnosis of first primary lung cancer were identified. After excluding cases with invalid stage and unknown smoking status, 174,078 patients were analyzed (Figure S1A). Among the 174,078 patients with lung cancer, 126,753 (72.2%) had LUAD (Table 1; Table S2). Overall, incidence of LUAD increased during the 13-year study period (Table 1). By 2023, the annual number of lung cancer cases among females continued to increase and exceeded that of males (Tables S3 and S4; Figure S2). Patients who had never smoked accounted for 57.8% (100,570/174,078) of all lung cancer cases (Fig. 1A), and 92.2% (73,462/79,706) of female patients had never smoked. Among patients with lung cancer, 44,949 (25.8%) had a smoking history of ≥20 pack-years, representing 44.7% (42,197/94,372) of the male patients and 3.5% (2752/79,706) of female patients.

Table 1.

Incidence and case number of lung cancer diagnosed in Taiwan from 2011 to 2023.

	All	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022	2023
Lung cancer
CRa	57.2	41.4	43.6	44.8	47.1	49.9	52.4	55.9	60	63.4	63.8	66.3	71.4	80.4
ASIRa	35.6	30.3	31.1	31.1	31.9	32.7	33.5	34.9	36.7	37.8	37	37.6	39.9	43.8
Cases	174,078	9614	10,140	10,443	11,017	11,722	12,317	13,159	14,149	14,975	15,063	15,773	16,639	18,747
LUAD
CRa	41.6	25.6	28.5	29.7	32.1	34.5	37.4	40.5	44.9	48.1	48.1	50.9	55.3	63.3
ASIRa	26.2	18.8	20.5	20.8	21.9	22.8	24.2	25.5	27.7	29.0	28.3	29.3	31.4	35.1
Cases (%)	126,753 (72.2%)	5943 (61.8%)	6632 (65.4%)	6934 (66.4%)	7513 (68.2%)	8091 (69.0%)	8790 (71.4%)	9535 (72.5%)	10,595 (74.9%)	11,349 (75.8%)	11,352 (75.4%)	12,114 (76.8%)	12,892 (77.4%)	14,768 (78.8%)
SCC
CRa	7.1	7.4	7.1	7.0	7.2	7.3	6.9	7.0	7.0	7.0	7.0	6.6	7.0	7.2
ASIRa	4.2	5.4	5.0	4.8	4.8	4.7	4.3	4.3	4.2	4.0	3.8	3.5	3.6	3.6
Cases (%)	21,523 (12.4%)	1724 (17.9%)	1642 (16.2%)	1630 (15.6%)	1678 (15.2%)	1713 (14.6%)	1628 (13.2%)	1657 (12.6%)	1662 (11.7%)	1654 (11.1%)	1640 (10.9%)	1567 (9.9%)	1632 (9.8%)	1670 (8.9%)
LEC
CRa	0.2	0.2	0.1	0.2	0.2	0.2	0.2	0.3	0.3	0.3	0.3	0.3	0.3	0.2
ASIRa	0.2	0.1	0.1	0.1	0.1	0.2	0.1	0.2	0.2	0.2	0.2	0.2	0.2	0.1
Cases (%)	721 (0.4%)	39 (0.4%)	34 (0.3%)	43 (0.4%)	39 (0.4%)	56 (0.5%)	45 (0.4%)	67 (0.5%)	59 (0.4%)	73 (0.5%)	75 (0.5%)	64 (0.4%)	76 (0.5%)	50 (0.3%)
Other NSCLC
CRa	4.4	4.7	4.6	4.3	4.2	4.3	4.1	4.3	4.1	4.0	4.2	4.2	4.5	4.9
ASIRa	2.6	3.4	3.1	2.9	2.8	2.7	2.5	2.6	2.4	2.3	2.3	2.3	2.4	2.5
Cases (%)	13,274 (7.6%)	1088 (11.3%)	1062 (10.5%)	1004 (9.6%)	991 (9.0%)	1019 (8.7%)	959 (7.8%)	1010 (7.7%)	966 (6.8%)	948 (6.3%)	1002 (6.6%)	1011 (6.4%)	1041 (6.3%)	1140 (6.1%)
SCLC
CRa	3.9	3.5	3.3	3.6	3.4	3.6	3.8	3.8	3.7	4.0	4.2	4.3	4.3	4.8
ASIRa	2.4	2.6	2.4	2.5	2.3	2.3	2.4	2.3	2.2	2.3	2.4	2.3	2.3	2.5
Cases (%)	11,807 (6.8%)	820 (8.5%)	770 (7.6%)	832 (8.0%)	796 (7.2%)	843 (7.2%)	895 (7.3%)	890 (6.8%)	867 (6.1%)	951 (6.3%)	994 (6.6%)	1017 (6.5%)	998 (6.0%)	1119 (6.0%)

ASIR, age-standardized incidence rate; CR, crude incidence rate; LEC, pulmonary lymphoepithelial carcinoma; LUAD, lung adenocarcinoma; NSCLC, non-small cell lung cancer; SCC, squamous cell carcinoma; SCLC, small cell lung cancer.

^aCases per 100,000.

Fig. 1.

Smoking behavior of LC (A) and LUAD (B) patients in Taiwan from 2011 to 2023. LC, lung cancer; LUAD, lung adenocarcinoma; PKY, pack-year.

Among the 126,753 patients with LUAD, 71,159 (56.1%) were female, 63,480 (50.1%) were aged ≥65 years, and 87,830 (69.3%) had never smoked (Table 2 and Fig. 1B). Those diagnosed with stage 0 and stage I disease were predominantly younger, female, without a history of smoking, compared to those with late-stage disease (Table 2). Among 42,349 patients with stage I LUAD, 7240 (17%) of them were minimally invasive adenocarcinoma.

Table 2.

Characteristics of patients with stage 0, early-stage, or late-stage lung adenocarcinoma.

	All	Stage 0	Early stage	Late stage	Stage II	Stage III	Stage IV
Case (%)	126,753	6146 (5.1%)	42,349 (33.4%)	77,988 (61.5%)	4384 (3.5%)	10,770 (8.5%)	62,834 (49.6%)
Sex
Male	55,594 (43.9%)	1887 (29.4%)	15,110 (35.7%)	38,597 (49.5%)	2106 (48.0%)	5694 (52.9%)	2106 (48.0%)
Female	71,159 (56.1%)	4529 (70.6%)	27,239 (64.3%)	39,391 (50.5%)	2278 (52.0%)	5076 (47.1%)	2278 (52.0%)
Age group
<55 years	26,693 (21.1%)	2984 (46.5%)	10,791 (25.5%)	12,918 (16.6%)	788 (18.0%)	1882 (17.5%)	10,248 (16.3%)
55–64 years	36,580 (28.9%)	2082 (32.5%)	13,874 (32.8%)	20,624 (26.4%)	1269 (28.9%)	3036 (28.2%)	16,319 (26.0%)
65–74 years	35,382 (27.9%)	1141 (17.8%)	12,282 (29.0%)	21,959 (28.2%)	1375 (31.4%)	3245 (30.1%)	17,339 (27.6%)
≥75 years	28,098 (22.2%)	209 (3.3%)	5402 (12.8%)	22,487 (28.8%)	952 (21.7%)	2607 (24.2%)	18,928 (30.1%)
Smoking history
Never smoked	87,830 (69.3%)	5618 (87.6%)	33,399 (78.9%)	48,813 (62.6%)	2826 (64.5%)	6077 (56.4%)	39,910 (63.5%)
Ever smoked	38,923 (30.7%)	798 (12.4%)	8950 (21.1%)	29,175 (37.4%)	1558 (35.5%)	4693 (43.6%)	22,924 (36.5%)

Categorical data are presented as number (%).

Early stage: stage I; Late stage: stage II, stage III, and stage IV.

Regarding the opportunistic screening cohort, a total of 115,013 individuals who underwent self-initiated LDCT screening across the five medical centers were included in the initial pool (2011–2021) (Figure S1B). Of these, 44,210 were excluded due to missing smoking status, yielding a final cohort of 70,803 individuals for the analysis. The screening population was predominantly composed of younger individuals (62.4%, 44,183/70,803, aged <55 years) and those who had never smoked (83.7%, 59,245/70,803) (Table 3). A sensitivity analysis was conducted to compare the demographic profiles of the excluded and included populations. The age and sex distributions of the excluded individuals (e.g., 57.5% male (25,428/44,210) and 64.6% aged <55 years (28,562/44,210)) were comparable to those of the final study cohort (Table S5). The characteristics of screened individuals were presented year-by-year in Table S6 (for overall cohort), and in Tables S7–S11 (for each respective medical centers).

Table 3.

Characteristics of individuals undergoing self-initiated LDCT from 2011 to 2021.

	All
Screened individuals (%)	70,803 (100%)
Age group
<55 years	44,183 (62.4%)
55–64 years	17,165 (24.2%)
65–74 years	6545 (9.2%)
≥75 years	2910 (4.1%)
Sex
Male	40,107 (56.6%)
Female	30,696 (43.4%)
Smoking history
Never smoked	59,245 (83.7%)
Ever smoked	11,558 (16.3%)

Categorical data are presented as number (%).

Trends in incidence of LUAD in Taiwan and LDCT screenings in 5 medical centers

Over the 13-year period, the annual case of early-stage and late-stage LUAD increased over time (Table 4; Fig. 2A). The trend of ASIR stratified by age and sex were presented in Figure S3. A stage-shift toward early-stage disease was observed across the population. The overall ASIR of LUAD nearly doubled, driven primarily by a significant increase in early-stage (AAPC, 13.0%; 95% CI, 12.3–13.7) (Fig. 2B; Table S12) and stage 0 disease (AAPC, 49.6%; 95% CI, 43.3–71.0) (Figure S4A). This rapid growth in early-stage LUAD occurred concurrently across all smoking statuses (Fig. 2B; Tables S12–S14). By contrast, while the overall proportion of late-stage disease declined (82.1–47.4%; 4481/5943–6997/14,768), this might be resulted from the inflation of early-stage disease. Crucially, the absolute burden of late-stage disease did not decrease; the ASIR of late-stage LUAD showed no significant reduction (AAPC, −0.1%; 95% CI, −0.5 to 0.3), among either those with a history of smoking or those who had never smoked (Fig. 2B; Table S12). To ensure the observed stage-shift was not an artifact of the staging system transition, a sensitivity analysis restricted to stable staging periods (AJCC 7th edition: 2011–2017; AJCC 8th edition: 2018–2023) was conducted (Table S15). The trends remained consistent within both eras. Early-stage LUAD increased significantly in both 2011–2017 (AAPC, 15.5%; 95% CI, 13.5–18.0) and 2018–2023 (AAPC, 9.5%; 95% CI, 5.2–14.0), while late-stage LUAD remained stable during 2011–2017 and trended downward during 2018–2023.

Table 4.

Characteristics and incidence of lung adenocarcinoma from 2011 to 2023.

Year of diagnosis
	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022	2023
Total
ASIR	18.8	20.5	20.8	21.9	22.8	24.2	25.5	27.7	29.0	28.3	29.3	31.4	35.1
Case	5943	6632	6934	7513	8091	8790	9535	10,595	11,349	11,352	12,114	12,892	14,768
Male sex
ASIRa	19.0	20.1	20.9	21.1	22.0	22.9	23.4	25.2	26.1	25.3	25.5	27.0	29.7
Case	2927	3144	3356	3469	3724	3963	4167	4608	4890	4857	5031	5346	6016
%	(49.3%)	(47.4%)	(48.4%)	(46.2%)	(46.0%)	(45.1%)	(43.7%)	(43.5%)	(43.1%)	(42.8%)	(41.6%)	(41.5%)	(40.7%)
Female sex
ASIRa	18.7	20.9	20.7	22.7	23.6	25.5	27.6	30.1	31.8	31.2	32.9	35.6	40.3
Case	3016	3488	3578	4044	4367	4827	5368	5987	6459	6495	7083	7546	8752
%	(50.7%)	(52.6%)	(51.6%)	(53.8%)	(54.0%)	(54.9%)	(56.3%)	(56.5%)	(56.9%)	(57.2%)	(58.4%)	(58.5%)	(59.3%)
Age <55
ASIRa	4.7	5.3	5.6	6.0	6.2	7.0	7.7	8.8	9.3	9.4	10.1	11.5	12.7
Case	1218	1381	1460	1571	1616	1823	1996	2249	2364	2372	2567	2862	3167
%	(20.5%)	(20.8%)	(21.1%)	(20.9%)	(20.0%)	(20.7%)	(20.9%)	(21.2%)	(20.8%)	(20.9%)	(21.2%)	(22.2%)	(21.4%)
Age 55–64
ASIRa	57.8	63.1	64.6	69.6	74.2	77.2	80.8	92.8	98.0	92.7	96.2	104.9	123.9
Case	1533	1790	1929	2169	2395	2556	2720	3184	3421	3275	3434	3701	4406
%	(25.8%)	(27.0%)	(27.8%)	(28.9%)	(29.6%)	(29.1%)	(28.5%)	(30.1%)	(30.1%)	(28.8%)	(28.3%)	(28.7%)	(29.8%)
Age 65–74
ASIRa	106.8	118.3	119.5	124.6	125.9	135.9	137.0	146.0	154.5	147.0	149.7	154.6	170.7
Case	1503	1696	1766	1920	2026	2318	2502	2859	3232	3311	3666	3936	4557
%	(25.3%)	(25.6%)	(25.5%)	(25.6%)	(25.0%)	(26.4%)	(26.2%)	(27.0%)	(28.5%)	(29.2%)	(30.3%)	(30.5%)	(30.9%)
Age ≥75
ASIRa	150.5	154.0	149.9	152.7	163.4	162.3	174.8	169.7	169.0	169.7	169.6	166.2	177.1
Case	1689	1765	1779	1853	2054	2093	2317	2303	2332	2394	2447	2393	2618
%	(28.4%)	(26.6%)	(25.7%)	(24.7%)	(25.4%)	(23.8%)	(24.3%)	(21.7%)	(20.5%)	(21.1%)	(20.2%)	(18.6%)	(17.7%)
Ever smoked
ASIRa	6.8	7.2	7.3	7.4	7.6	7.8	7.8	8.1	8.5	8.3	8.2	8.6	9.4
Case	2163	2329	2457	2534	2698	2834	2918	3134	3366	3350	3416	3593	4045
%	(36.4%)	(35.1%)	(35.4%)	(33.7%)	(33.4%)	(32.3%)	(30.6%)	(29.6%)	(29.6%)	(29.5%)	(28.2%)	(27.9%)	(27.4%)
Never smoked
ASIRa	12.0	13.3	13.4	14.5	15.2	16.4	17.7	19.6	20.5	20.0	21.2	22.8	25.7
Case	3780	4303	4477	4979	5393	5956	6617	7461	7983	8002	8698	9299	10,723
%	(63.6%)	(64.9%)	(64.6%)	(66.3%)	(66.6%)	(67.7%)	(69.4%)	(70.4%)	(70.4%)	(70.5%)	(71.8%)	(72.1%)	(72.6%)
Stage 0b
ASIRa	0.03	0.1	0.2	0.2	0.6	0.9	1.2	1.7	1.9	2.0	2.6	3.2	3.7
Case	8	14	49	66	195	297	412	600	676	718	913	1115	1319
%	(0.1%)	(0.2%)	(0.7%)	(0.9%)	(2.4%)	(3.4%)	(4.3%)	(5.6%)	(5.9%)	(6.3%)	(7.5%)	(8.6%)	(8.9%)
Early stageb
ASIRa	3.4	4.5	5.0	5.5	6.4	7.6	8.6	9.7	10.9	10.9	11.6	13.1	15.9
Case	1054	1424	1637	1834	2224	2676	3161	3612	4171	4226	4560	5208	6452
%	(17.7%)	(21.5%)	(23.6%)	(24.4%)	(27.5%)	(30.4%)	(33.2%)	(34.1%)	(36.8%)	(37.2%)	(37.5%)	(40.4%)	(43.7%)
Late stageb
ASIRa	15.4	15.9	15.6	16.2	15.8	15.8	15.7	16.3	16.1	15.3	15.4	15.0	15.6
Case	4881	5194	5248	5613	5672	5817	5962	6383	6502	6408	6641	6569	6997
%	(82.1%)	(78.3%)	(75.7%)	(74.7%)	(70.1%)	(66.2%)	(62.5%)	(60.3%)	(57.3%)	(56.4%)	(55.0%)	(51.0%)	(47.4%)
Stage II
ASIRa	0.7	0.8	0.8	0.9	0.8	0.9	0.8	1.0	1.1	0.9	0.9	0.9	1.1
Case	228	241	257	295	293	314	318	380	431	377	376	387	481
%	(3.8%)	(3.6%)	(3.7%)	(3.9%)	(3.6%)	(3.6%)	(3.3%)	(3.6%)	(3.8%)	(3.3%)	(3.1%)	(3.0%)	(3.3%)
Stage III
ASIRa	2.2	2.4	2.3	2.3	2.2	2.3	2.3	2.3	2.2	2.0	2.0	2.0	2.1
Case	686	788	753	794	791	843	875	886	883	836	849	828	933
%	(11.5%)	(11.9%)	(10.9%)	(10.6%)	(9.8%)	(9.6%)	(9.1%)	(8.4%)	(7.8%)	(7.4%)	(7.0%)	(6.4%)	(6.3%)
Stage IV
ASIRa	12.5	12.7	12.5	13.0	12.7	12.6	12.5	13.0	12.8	12.4	12.7	12.2	12.4
Case	3967	4165	4238	4524	4588	4660	4769	5117	5188	5195	5416	5354	5583
%	(66.8%)	(62.8%)	(61.1%)	(60.2%)	(56.7%)	(53.0%)	(50.0%)	(48.3%)	(45.7%)	(45.8%)	(44.9%)	(41.5%)	(37.8%)

ASIR, age-standardized incidence rate.

^aPer 100,000.

^bEarly stage: stage I; late stage: stage II, stage III, and stage IV.

Fig. 2.

Annual incidence number and proportion of different lung adenocarcinoma stages (A); age-standardized incidence rate of lung adenocarcinoma and annual volumes of self-initiated LDCT screening volumes (B) among all, those who had ever smoked and those who had never smoked from left to right columns. Solid lines refer to observed ASIR, dotted lines refer to APC slope of ASIR estimated by Joinpoint regression analysis, and bar charts refer to annual volumes of LDCT screening. Early stage: stage I; late stage: stage II, stage III, and stage IV. ∗AAPC of ASIR is statistically significant. AAPC, average annual percentage change; ASIR, age-adjusted incidence rate; LDCT, low-dose computed tomography; LUAD, lung adenocarcinoma.

Concurrently, the annual volume of individuals undergoing self-initiated LDCT screening increased from 470 to 8494 per year (2011–2021) (Fig. 2B; Table S6). However, the noticeable drop in screening volumes observed in 2020 and 2021 was a direct consequence of the COVID-19 pandemic. During this period, the proportion of screened individuals who had never smoked increased from 69.6% (327/470) to 86.2% (7323/8494). This upward trend in overall screening volume descriptively paralleled the increase in the ASIR for early-stage and stage 0 LUAD (Fig. 2B; Figure S4). Notably, subgroup analyses revealed heterogeneous patterns (Table S12). Late-stage LUAD declined significantly among ever-smoking males (AAPC, −0.3%; 95% CI, −0.6 to −0.1). Among never-smoking females, a significant decrease emerged only after 2018 (APC, −1.9%; 95% CI, −7.9 to −0.1). These findings suggest that, despite the absence of a population-wide reduction, early signals of downstaging may be present in specific subgroups.

Subgroup analysis of trends in ASIR of LUAD and LDCT screening volumes in those who had never smoked

Among patients who had never smoked, the ASIR of early-stage (Fig. 3A) and stage 0 (Figure S5A) LUAD increased significantly across all age and sex subgroups, particularly in those aged <55 years. Importantly, never-smoking females aged <55 years were the only subgroup exhibiting a decline in the ASIR of both stage IV (AAPC, −2.9%; 95% CI, −3.8 to −1.1) and late-stage LUAD (AAPC, −1.9%; 95% CI, −3.0 to −0.9) (Table S16). For other never-smoking subgroups, the late-stage ASIR either increased (e.g., males aged <55 and ≥ 75 years) or remained stable, suggesting that the burden of late-stage disease did not proportionally decrease despite increasing screening volumes.

Fig. 3.

Age-standardized incidence rate of lung adenocarcinoma and annual volumes of self-initiated LDCT screening volumes stratified by age among males and females who had never smoked (A), and males and females who had smoked (B). Solid lines refer to observed ASIR, dotted lines refer to APC slope of ASIR estimated by Joinpoint regression analysis, and bar charts refer to annual volumes of LDCT screening. Early stage: stage I; late stage: stage II, stage III, and stage IV. ∗AAPC of ASIR is statistically significant. AAPC, average annual percentage change; ASIR, age-standardized incidence rate; LDCT, low-dose computed tomography.

The annual volume of LDCT screening stratified by age, sex, and smoking history were listed in Table S17. Visualized via scatter plot, the upward trend in LDCT screening volumes among individuals who had never smoked paralleled the rise in ASIR of early-stage LUAD in both males (Fig. 4A) and females (Fig. 4B), and the decline in ASIR for late-stage LUAD in never-smoking females (Figs. 3A and 4B). However, given the aggregate nature of these data strata, these parallel trends are exploratory.

Fig. 4.

Exploratory scatter plot illustrating the relationship between opportunistic self-initiated LDCT screening volumes and AAPC in ASIR of lung adenocarcinoma among males (A) and females (B) who never smoked, and males (C) and females(D)who ever smoked. Blue color refers to male population, pink color refers to female population, light color refers to early-stage LUAD, and dark color refers to late-stage LUAD. Dashed line refers to descriptive trend lines between LDCT volumes and AAPC in ASIR. Error bars indicate the 95% confidence intervals of the AAPC. Early stage: stage I; late stage: stage II, stage III, and stage IV. AAPC, average annual percentage change; ASIR, age-standardized incidence rate; CI, confidence interval; LDCT, low-dose computed tomography.

Subgroup analysis of trends in ASIR of LUAD and LDCT screening volumes in those who ever smoked

Among patients who ever smoked, despite males undergoing LDCT screening more frequently than females, the incidence patterns differed from those of never-smoking populations. While the ASIR of early-stage LUAD increased significantly across most subgroups (Fig. 3B; Table S18), the late-stage disease only declined in the patients aged ≥75 years (males AAPC, −2.3%; 95% CI, −3.1 to −1.5; females AAPC, −3.9%; 95% CI, −6.3 to −1.3). Additionally, males aged ≥75 years was the only subgroup demonstrated a reduction in the ASIR of stage IV LUAD (AAPC, −2.0%; 95% CI, −3.0 to −1.0) (Table S18).

Overall, the screening volumes among individuals who ever smoked descriptively paralleled the upward trends with early-stage LUAD (Fig. 4C and D), but no clear parallel patterns with the change of late-stage LUAD. Similarly, these observations remain exploratory due to the limited number of age strata.

Lung cancer mortality in Taiwan

The ASMR of lung cancer declined from 2011 to 2023 (AAPC, −1.7%; 95% CI, −2.0 to −1.3) (Figure S6A; Table S19), with males exhibiting higher mortality than females across all age subgroups (Figure S6B). Females experienced a greater reduction in mortality in the younger subgroups (aged <65 years), whereas males showed a more pronounced reduction in the older subgroups (aged ≥65 years).

From 2015 to 2023, ASIBM for LUAD decreased significantly (AAPC, −2.0%; 95% CI, −2.5 to −1.4), in both patients who had smoked (AAPC: −1.5%), and those who never smoked (AAPC: −2.2%) (Table S20). The decline was greater in females (AAPC: −2.7%) than in males (AAPC: −1.2%). Stage-specific analyses showed larger reductions in advanced disease, e.g. stage III (AAPC, −5.0%) and stage IV (AAPC, −1.6%). Late-stage disease (stage II–IV) also showed a significant decline (AAPC: −2.0%, 95% CI: −2.5 to −1.5), with consistent decreases across multiple age and smoking subgroups (Table S21).

Discussion

In this nationwide study, the incidence of LUAD in Taiwan increased between 2011 and 2023, driven predominantly by a rise in stage 0 and early-stage disease. Although this pattern may suggest a stage shift toward earlier detection, the absence of a corresponding decline in the population-level incidence of late-stage disease indicates that a substantial downstaging effect has not yet been fully realized. Instead, the expansion of early-stage LUAD, occurring in parallel with increasing uptake of opportunistic LDCT screening since 2011, is more consistent with enhanced detection, potentially including overdiagnosis.

ASIBM declined significantly during the study period, including reductions in advanced-stage disease, suggesting that these trends cannot be explained by overdiagnosis alone. Notably, reductions in late-stage incidence were observed only in selected subgroups, indicating that the beneficial impact of screening on reducing advanced disease burden remained heterogeneous at the population level.

The discordance between the increase in early-stage disease and limited decline in late-stage disease, together with the observed reduction in ASIBM, suggests a complex interplay between detection expansion and clinical benefit. These patterns raise concerns about overdiagnosis and detection inflation, but also indicate that the observed trends cannot be attributed to overdiagnosis alone. A nationwide analysis from Korea reported a rapid increase in localized lung cancer among women, accompanied by a sharp increase in lung resections, suggesting overdiagnosis associated with widespread guideline-discordant LDCT screening.¹⁴ Data from the Pudong New Area of China indicated overdiagnosis in average-risk populations exposed to widespread LDCT screening.¹³ Although ASIR of late-stage lung cancer declined in China between 2011 and 2020 (AAPC, −0.3% in males; AAPC, −0.2% in females), the estimated proportion of lung cancer overdiagnosis was substantial in females (22% in 2011–2015, 50% in 2016–2020), and LUAD overdiagnosis increased in both males (8%–22%) and females (85%–89%). In Taiwan, from 2011 to 2023, the ASIR of early-stage and stage 0 LUAD among never-smoking patients increased significantly across both sexes and all age subgroups, with the steepest increase in those aged <55 years. In contrast, significant declines in late-stage LUAD were confined to females aged <55 years who had never smoked. These patterns are consistent with potential stage-shift benefits among never-smoking individuals. However, the stage-shift may be driven by the increased detection of indolent diseases, such as AIS (stage 0) or MIA. Although our analysis separated Stage 0 from Stage I disease, the parallel rise in both stages among individuals who undergo LDCT outside guideline-based indications highlights the urgent need to address overdiagnosis. Unlike structured programs, self-initiated screening often lacks standardized longitudinal follow-up and is associated with poor adherence, which can lead to either over management or loss to follow-up, severely compromising screening efficiency.^27,28 Practical strategies in real-world screening, such as adopting nodule management protocols and extending intervals for low-risk individuals, are essential.²⁹ Furthermore, lung cancer risk prediction models tailored to these populations will be required to maximize benefit and limit harms related to overdiagnosis.³⁰

We also observed decreases in late-stage LUAD among patients who ever smoked aged ≥75 years, despite this group undergoing the least LDCT screenings. Possible explanations include cohort effects stemming from declining smoking prevalence in prior decades, and demographic shifts with an aging population. The observed heterogeneity in late-stage trends may also reflect changes in staging protocols (such as the transition from the AJCC 7th to 8th edition), increased diagnostic intensity, or competing mortality risks in older cohorts. These hypotheses warrant further investigation to distinguish true stage-shift from these confounding factors.

In Asia, a high proportion of lung cancer occurs in never-smoking individuals, particularly among females (60.3–94.8%).^31,32 A recent meta-analysis suggested that LDCT screening reduced lung cancer deaths more in never-smoking populations compared with those who ever smoked (relative risk, 0.27).³³ In the present study, lung cancer ASMR declined more in females than in males among individuals aged <55 and 55–64 years; conversely, reductions were greater in males than in females among those aged 64–75 and ≥ 75 years. These observations align with the significant decrease in late-stage LUAD among females aged <55 years who had never smoked, and among patients aged ≥75 years who had ever smoked, suggesting a population-level stage-shift in these subgroups. Moreover, the ASIBM declined significantly during the study period, particularly in stage III and stage IV disease. These findings indicate that the burden of lethal disease has decreased over time. Nevertheless, the overall decline in lung cancer mortality likely reflects multiple contributors, including stage-shift due to LDCT screening and the expanding availability of targeted therapies and immunotherapies.^11,34 The interpretation of mortality trends is limited by competing risk dynamics in populations and potential cause-of-death misclassification on death certificates. Direct linkage between survival improvement and the stage-shift remains insufficiently demonstrated, requiring cautious interpretation.

Although more males than females underwent LDCT screening (56.6% vs 43.4%; 40,107 vs 30,696), females accounted for a higher proportion of LUAD (56.1% vs 43.9%; 71,159 vs 55,594), particularly among never-smoking patients. While the ASIR of early-stage LUAD increased markedly among never-smoking males aged ≤55 years, the late-stage disease in this group did not decline as that observed in females. These sex-specific differences in LUAD incidence and stage-shift are consistent with prior literature and suggest potential differences in exposure profiles, susceptibility, or tumor biology between sexes.^21,33,35,36 Additionally, this mismatch between actual disease burden and LDCT uptake highlights the complex impact of sex and smoking status on screening willingness in Taiwan. In an opportunistic setting, screening is largely driven by individual willingness and socioeconomic factors rather than absolute risk, potentially leading to inefficient resource allocation. This phenomenon underscores the need to develop risk stratification tools to empower screening programs for the Asian non-smoking population.³⁰

Compared with reports from the United States, where increases in early-stage lung cancer have been accompanied by prompt declines in late-stage disease,^34,37 findings in the current study more closely resemble observations from Japan. Nguyen et al. demonstrated an increase in localized LUAD (AAPC, 5.7%), alongside a continued rise in distant-stage LUAD (AAPC, 1.6%), predominantly among female patients who had never smoked.^12,31 Pathologic features may also differ by smoking status. Maeda et al. reported that those who ever smoked with LUAD tended to have more solid components and worse recurrence-free survival, consistent with a more aggressive histology.³⁸ A meta-analysis revealed that patients who had never smoked are diagnosed, on average, two years earlier than those who ever smoked, suggesting slower tumor growth with earlier onset.³⁹ These data point to differences in tumor growth kinetics and histologic patterns by smoking status that may influence the magnitude and timing of population-level stage-shifts.

Recent evidence supports differences in screening performance, screening intensity, and stage distribution between Western countries and East Asia, as well as by smoking history. Western screening guidelines are largely derived from structured programs with age and smoking eligibility criteria (e.g., the NLST and NELSON trials), which control the screening intensity. In contrast, East Asia, and specifically Taiwan during our study period, has experienced an increase in unstructured, high-intensity opportunistic self-initiated screening. This unrestricted access increased the detection of indolent nodules and the risk of overdiagnosis.^19,30 A meta-analysis reported significantly higher proportions of lung cancer detected at the baseline screening round (95.4% vs 70.9%) and at stage I (88.5% vs 79.7%) among individuals who had never smoked compared with those with a history of smoking.³³ In the TALENT study enrolling never- or light-smoking participants, most lung cancer cases were identified at the baseline LDCT screen (92.1%), with subsequent rounds yielding low detection rates and few interval cancers.⁹ By contrast, in the NELSON trial, 59% of lung cancer cases in the screening group were screen-detected (with 12.8% interval cancers).⁷ In the NLST, 25.5% lung cancer cases were diagnosed beyond the baseline round, and 61.2% were diagnosed following a positive screening.^6,40 Compared with the TALENT, a greater proportion of stage IV disease was observed among screen-detected cases in NLST and NELSON (21.7% in NLST, 26.7% in NELSON, vs 1.6% in TALENT).^6,7,9 While it could be hypothesized that a longer lag time might contribute to the limited decline in advanced-stage disease, which could possibly be related to younger age at diagnosis or more indolent tumor biology among patients who never smoked, other mechanisms may coexist.^33,38,41 These include overdiagnosis, lead-time bias, length bias, and a delayed stage-shift effect. Our current study cannot distinguish among these different factors. Future longitudinal cohort studies are needed to evaluate these effects.

In Taiwan, a national LDCT screening program was launched in July 2022 targeting individuals with heavy smoking exposure or with LCFH in first-degree relatives.¹⁸ However, previous studies suggested that these two high-risk categories collectively accounted for less than 40% of lung cancer in Taiwan.^4,10 Additional risk factors reflecting genetic, hormonal, and environmental contributions require evaluation.^{8,9,35,36,42,43} In the present study, individuals who underwent self-initiated LDCT screening were generally younger and included a higher proportion of individuals who never smoked compared with those diagnosed with LUAD from TCR data. Although the lower absolute screening volumes among older age groups naturally reflects the underlying population demographics and decreasing population numbers with advancing age, these differences in demographic and risk profiles between the screening population and the disease cohort underscore the need to strengthen strategies targeting high-risk individuals. At present, two ongoing prospective studies (NCT05572944 and NCT05557487) in Taiwan aimed to develop lung cancer risk-prediction models to enable individualized screening; their results may help optimize eligibility criteria. Looking forward, the transition from an opportunistic, self-initiated model to this organized, government-reimbursed national program is expected to standardize nodule management, improve adherence, and provide a more structured framework for risk stratification of lung cancer in Taiwan.

The strength of this study is that it depicts detailed LUAD incidence trends using a nationwide registry with individual-level smoking data, which is an uncommon feature in large registry-based analyses.³ Additionally, it demonstrated real-world characteristics of opportunistic, self-initiated LDCT screening uptake using a multicenter dataset comprising 70,803 individuals. While non-risk based LDCT screening has been described,^15,20 our series includes the largest number to date with smoking history information. To our knowledge, this is the first report from Taiwan showing subgroup-specific declines in late-stage LUAD incidence despite an overall increase in early-stage disease. There are several limitations in this study. First, although the TCR and TNDCD were linked at the individual level to estimate incidence-based mortality, the opportunistic LDCT screening dataset could not be linked to individual cancer registry or mortality records. Therefore, analyses relating screening volumes to LUAD incidence and mortality trends were ecological and exploratory. Variations in data collection methodologies, differing coding practices across registries, and the lack of individual-level tracking for risk factors introduce potential confounding. Observations regarding the relationship between screening volumes and changes in ASIR were based on aggregate demographic strata and are therefore subject to small-sample instability and multiple-comparison uncertainty. The results cannot infer causality from these associations, nor can the current study design definitively distinguish a genuine stage-shift from overdiagnosis, lead-time bias, or length bias. This study did not employ mathematical approaches to quantify overdiagnosis. Therefore, while the observed increase in early-stage disease without a proportional decline in late-stage incidence provides an ecological signal, the magnitude of overdiagnosis in Taiwan remains inconclusive. Second, national data on self-initiated LDCT screening were unavailable; our screening dataset derived from five medical centers may not represent the entire population. Because these are tertiary centers primarily located in metropolitan areas, the individuals undergoing opportunistic, self-initiated LDCT screening likely represent a subpopulation with higher socioeconomic status and greater health literacy compared to the general population. The screening volumes observed in our dataset may not fully capture the screening behaviors and access disparities across the entire nation. Third, government-funded LDCT screening was initiated in July 2022. By December 2023, 956 lung cancer cases had been diagnosed, approximately 810 of which were stage 0–I, following the screening of 78,000 participants.¹⁸ These cases accounted for 5.1% of the 18,747 annual diagnoses in 2023, suggesting that the program may have had a minimal impact on the observed increase in early-stage disease. Fourth, the transition from the AJCC 7th to the 8th edition in 2018 introduced different size criteria, and may act as a conservative bias, theoretically inflating late-stage counts and deflating early-stage counts. However, our sensitivity analysis within stable staging eras demonstrates consistent trends (Table S15). Because the definition of stage IV disease remained unchanged during the study period, the reduction of stage IV LUAD in never-smoking females aged <55 years and ever-smoking males aged ≥75 years was a true reduction. Lastly, this study mainly assessed the trends of incidence of LUAD, which accounted for 89.3% (71,159/79,706) of female patients with lung cancer, but only 58.9% (55,594/94,327) of male patients. Consequently, the findings of this study could not provide a comprehensive picture of male lung cancer in Taiwan, and cannot be generalized to regions with higher smoking rates. While our analyses were stratified by sex and smoking status, potential confounding by differences in underlying age distributions across these subgroups was not fully adjusted for, which warrants caution when interpreting the comparative trends.

Conclusion

In this nationwide, population-based analysis, we observed a heterogenous stage-shift in LUAD across demographic subgroups in Taiwan. While early-stage LUAD increased across all subgroups, significant declines in late-stage LUAD were confined to never-smoking females aged <55 years and patients who ever smoked aged ≥75 years. The growth of self-initiated LDCT screenings paralleled the increase in early-stage diagnosis, particularly among those who never smoked; however, the lack of a corresponding reduction in late-stage disease across most subgroups suggests that detection expansion, including overdiagnosis, while the concurrent decline in incidence-based mortality indicates emerging clinical benefit. Therefore, a transition from unstructured, opportunistic screening toward organized, risk-stratified programs should be recommended. Future public health policies and prospective studies should focus on establishing targeted screening strategies for Asian never-smoking populations to reduce advanced-stage disease while minimizing the potential harms of overdiagnosis.

Contributors

All authors contributed to the study conception/design or acquisition, analysis and interpretation of the data. Patient data were collected by JJW, CJC, CHC, JST, PHF, YWY, ZRZ, CHC, CJY, IWC, CYT, WCL, JYH, KCC, PCY, SLY, and GCC. JJW, CJC, CHC, JST, PHF, YWY, ZRZ, CHC, and CJY verified the data and take responsibility for the integrity of the data and the accuracy of the data analysis. JJW, CJC, CHC, JST, PHF, CJY, IWC, CYT, and GCC had access to raw data. IWC, CYT, and GCC had final responsibility for the decision to submit for publication. All authors critically revised the manuscript for important intellectual content and have read and approved the final manuscript.

Data sharing statement

The data generated in this study are available upon reasonable request from the corresponding author.

Editor note

The Lancet Group takes a neutral position with respect to territorial claims in published maps and institutional affiliations.

Declaration of interests

The authors declare that there is no conflict of interest.