Our study leveraged data from the 2021 GBD Study. The Bayesian, regularized, and trimmed meta-regression (MR-BRT) was employed by the database to quantify the effect size of each risk-outcome pair. Nonlinear variations in relative risk (RR) functions were captured using through integrated spline functions and adjusted for systematic bias. RR estimates were derived from data obtained through systematic reviews and meta-analyses, with inclusion criteria determined by the 95% uncertainty interval (UI). Additionally, spatio-temporal Gaussian process regression (ST-GPR) and disease model meta-regression (DisMod-MR 2.1) were applied to aggregate heterogeneous data and correct for biases, thus estimating the exposure levels and distributions for each specific risk factor. Lung cancer, which is briefly described as neoplasms located in the trachea, bronchus, or lung, is classified in the International Classification of Diseases (ICD) 10 under the codes C33 and C34-C34.92 (12). The AYA age group for lung cancer includes individuals aged between 15 and 39 years.

In this study, estimates and the 95% UI for incidence, mortality, and disability-adjusted life years (DALYs) lung cancer burden among AYAs were extracted. The dataset from GBD 2021 encompasses global levels, five Socio-Demographic Index (SDI) level regions, 21 GBD regions, and 204 countries and territories from 1990 to 2021. The cancer data were categorized into age cohorts spanning five years each, specifically including aged 15-19, 20-24, 25-29, 30-34, and 35–39 years.

The SDI, developed by GBD researchers, is a composite measure used to assess socio-economic conditions affecting health outcomes across regions. In the GBD 2021 study, SDI values range from 0 to 1, with higher values indicating better socio-economic conditions and improved health outcomes. Based on 2021 SDI, regions are divided into five quintiles: low (0-0.466), low-middle (0.466-0.619), middle (0.619-0.712), high-middle (0.712-0.810), and high (0.810-1) (12).

To evaluate the attributable burden of lung cancer among AYAs associated with risk factors, two metrics (mortality or DALYs) and their corresponding population attributable fractions (PAFs) were assessed and calculated (11). The risk factors chosen, in accordance with the World Cancer Research Fund’s standards of convincing or probable evidence, encompassed smoking, exposure to secondhand smoke, ambient particulate matter pollution, household air pollution from solid fuels, diets deficient in fruits, elevated fasting plasma glucose levels, residential radon, and occupational exposure to substances such as asbestos, arsenic, beryllium, cadmium, chromium, diesel exhaust, nickel, polycyclic aromatic hydrocarbons, and silica. The detailed methodology of the GBD 2021 is presented in a series of key publications. The detailed methods of the GBD 2021 study and the risk assessment involving the lung cancer burden were comprehensively detailed in the collaborator group’s article (11).

The estimates were age-standardized to research data using the direct age-standardization. The purpose of age-standardization is to remove the effect of differences in age structure (13). The joinpoint regression model was used to analyze the temporal trends of the age specific rate and age-standardized rate (ASR) from 1990 to 2021. The trend was quantified using the annual percentage change (APC) and the average annual percentage change (AAPC). APC is an indicator used to describe the year-to-year change, while AAPC is an indicator used to describe the average change over a long period of time. The ASRs and numbers for mortality and DALYs were extracted for inequality analysis.

Decomposition analysis was used to identify the drivers of changes in the lung cancer burden among AYAs from 1990 to 2021. The relative contributions of three factors—population growth, aging, and changes in age-specific rate—were evaluated.

According to the World Health Organization Health Equity Assessment guidelines, slope index of inequality (SII) and concentration index serve as two fundamental measures for evaluating absolute and relative income-related inequalities, which are used to assess the health inequalities of disease burden across countries. The SII reflects the absolute difference in the above indicators between the lowest-SDI and highest-SDI countries, with higher absolute values of SII indicated greater inequality. The Concentration Index was calculated using the Lorenz curve based on per capita SDI and the corresponding national burden metrics. It reflects the area between the 45° line and the Lorenz curve. A negative index indicates the burden is higher in low-income countries. Conversely, a positive index indicates a higher burden in high-income countries.

ASR was reported per 100,000 population, with data presented as values with 95% UI. Temporal trends were assessed using joinpoint software (version 5.0.2) from the National Cancer Institute. The values of the AAPC and APC are expressed as percentages, with data presented as values accompanied by 95% confidence intervals (CI). All statistical analyses and mapping were performed using R statistical software (version 4.3.3). A two-sided P value < 0.05 was set as the significance threshold.

The numbers and rates of lung cancer among AYAs are presented in Supplementary Table 1 . In 2021, there were 28,004 incident cases, 22,622 mortality cases, and 129.1 million DALYs lost due to lung cancer among AYAs worldwide. The age-standardized incidence rate (ASIR), age-standardized mortality rate (ASMR), and age-standardized DALY rate (ASDR) of lung cancer among AYAs were 0.9 cases per 100,000 population, 0.7 cases per 100,000 population, and 42.1 cases per 100,000 population, respectively. With sex-specific classification, male AYAs experienced a higher burden of lung cancer than females. In 1990, the number of incident cases among males was 1.8 times that of females, but this declined to 1.5 times in 2021. Likewise, the ASIR in males was 1.8 times higher than in females in 1990, but decreased to 1.4 times in 2021. Similar patterns were observed for mortality and DALYs. These trends suggest a narrowing gender gap in the lung cancer burden among AYAs over the past three decades.

The joinpoint regression analysis identified that the ASIR, ASDR, and ASMR of lung cancer among AYAs continued to decrease globally from 1990 to 2021 (ASIR: AAPC = -1.2, 95% CI: -1.3 to -1.1; ASMR: AAPC = -1.4, 95% CI: -1.6 to -1.3; ASDR: AAPC = -1.4, 95% CI: -1.7 to -1.1). The joinpoint regression analysis revealed substantial change points in ASIR in 2003 and 2013. The ASIR decreased between 1990 and 2003 (APC= -0.8, 95% CI: -0.9 to -0.6), decreased at a greater rate between 2003 and 2013 (APC= -2.4, 95% CI: -2.6 to -2.1), and decreased slowly between 2013 and 2021 (APC= -0.4, 95% CI: -0.7 to -0.1). The ASDR and ASMR of lung cancer among AYAs followed the same pattern, that is the decline was larger between 2003 and 2013, but the downward trend became slow between 2013 and 2021. Global trends of lung cancer among AYAs can be found in Figure 1 . With sex-specific classification, the ASIR with a greater reduction in men than in women (men: AAPC = -1.4, 95% CI: -1.6 to -1.3; women: AAPC = -0.8; 95% CI: -0.9 to -0.7). The trends in the ASDR and ASMR were similar to that of ASIR with respect to sex. Global trends of lung cancer among AYAs by sex can be found in Figure 1 ; Supplementary Table 1 .

Among SDI regions, increases in the ASIR, ASDR, and ASMR were solely observed in the low-middle SDI and low SDI regions. Conversely, regions classified as high SDI, middle SDI, and high-middle SDI exhibited declining trends from 1990 to 2021. Interestingly, in 2021, the high-middle SDI region recorded the highest ASIR, ASDR, and ASMR, with rates of 1.1, 0.9, and 53.6 cases per 100,000 population, respectively. In addition, the middle SDI region had the highest number of incidence, mortality, and DALYs ( Supplementary Table 1 ).

At the regional level, Western Sub-Saharan Africa experienced the most significant rise in the ASIR, ASDR, and ASMR from 1990 to 2021 (ASIR: AAPC = 0.53, 95% CI: 0.47 to 0.58; ASMR: AAPC = 0.50, 95% CI: 0.45 to 0.55; ASDR: AAPC = 0.51, 95% CI: 0.46 to 0.56). The largest decrease in the ASIR, ASDR, and ASMR was observed in High-income North America and Central Asia from 1990 to 2021. In the past 32 years, it is worth noting that East Asia was the region with the highest number and ASIR, ASDR, and ASMR from lung cancer among AYAs ( Supplementary Table 1 ). Lesotho experienced the most significant increase in the ASIR, ASDR, and ASMR between 1990 and 2021 (ASIR: AAPC = 3.5, 95% CI: 3.0 to 3.9; ASDR: AAPC = 3.5, 95% CI: 3.0 to 3.9; ASMR: AAPC = 3.4, 95% CI: 3.0 to 3.9). Contrariwise, Spain was the highest national rate of decline in the ASDR and ASMR (ASMR: AAPC = -4.5, 95% CI: -5.0 to -4.0; ASDR: AAPC = -4.5, 95% CI: -4.9 to -4.1), while Kazakhstan was the highest decline for the ASIR (AAPC = -4.0, 95% CI: -4.5 to -3.4). In 2021, China recorded the highest number of new cases, deaths, and DALYs, whereas Palau and Monaco had the highest ASDR, ASMR, and ASIR, respectively. ( Supplementary Table 2 ).

The changes in the numbers of incident cases, mortality, and DALYs of lung cancer among AYAs from 1990 to 2021 can be attributed to population growth, population ageing, and changes in age-specific rate. During this period, the number of incident cases of lung cancer among AYAs increased from 26,790 in 1990 to 28,004 in 2021, while the number of deaths decreased from 23,309 in 1990 to 22,622 in 2021. Similarly, DALYs decreased from 1,335,868 in 1990 to 1,291,020 in 2021. These shifts provide important demographic context for interpreting the decomposition results. Globally, the overall difference of incident, mortality, and DALYs associated with lung cancer among AYAs have been vanishingly small. Population growth and population aging have contributed to increases, but changes in age-specific rate have led to decreases, which partially offset the overall increase ( Figure 2 ). While the overall global changes in case numbers were relatively modest, some regions experienced marked increases, particularly Western Sub-Saharan Africa and low SDI countries. For instance, the number of incident cases with lung cancer among AYAs increased by 4.5%, with 31.8% attributable to population growth, 11.7% from population aging, and 39.0% due to declining age-specific incidence rate. For most of the region, changes in age-specific rate were the primary drivers for the decrease of lung cancer burden among AYAs, whereas population growth was the most significant factor for the increase, as seen in Andean Latin America, Australasia, Central Asia, Central Latin America, Central Sub-Saharan Africa, Southern Latin America, Southern Sub-Saharan Africa. At the regional level, the most significant increasing overall difference of incident cases, mortality, and DALYs in Western Sub-Saharan Africa (Incident cases: 214.9%; Mortality: 212.2%; DALYs: 212.9%). Among the five SDI regions, the most significant increasing changes occurred in the low SDI region (Incident cases: 168.8%; Mortality: 166.7%; DALYs: 167.4%), while the significant decreasing changes was observed in the high SDI region (Incident cases: -38.0%; Mortality: -47.7%; DALYs: -47.4%). Specific numerical values can be found in the Supplementary Table 3 .

Significant absolute and relative disparities in the lung cancer burden among AYAs were identified, in the ASIR, ASDR, and ASMR, the slope index has shown a significant decrease over time, whereas the concentration index has exhibited significant increase over the same period ( Figure 3 ). From 1990 to 2021, the SII for the ASIR were 1.12 and 0.50; the SII for the ASMR were 0.89 and 0.26; the SII for the ASDR were 50.04 and 14.51, indicating positive correlation between the ASIR, ASDR, and ASMR and SDI. The substantial reduction observed points to a diminishing disparity in the age-standardized health burden of lung cancer among AYAs across the highest SDI and the lowest SDI countries during this period. For example, in 1990, the SII for the ASDR was 50.04, indicating an excess mortality of 50.04 per 100,000 population in the country with the highest SDI compared to the country with the lowest SDI. By 2021, this disparity had decreased to 14.51, reflecting a notable reduction in inequality over time.

By contrast, the analysis of relative inequalities revealed that the concentration indices for the ASIR, ASDR, and ASMR showed consistent upward trends from 1990 and 2021. For example, the concentration index of the ASDR was 0.16 (95% CI 0.12 to 0.21) in 1990 and 0.20 (95% CI 0.15 to 0.25) in 2021. Although the absolute inequality in the burden of lung cancer among AYAs has decreased regionally between high SDI and low SDI countries, inequality remains prevalent. This suggests that the burden of lung cancer among AYAs was excessively focused in wealthier countries. In addition, the decline in the absolute inequality was greater in men than in women, the relative increase in the relative inequality was greater among women than among men ( Supplementary Figures 1 , 2 ).

In 2021, we assessed the contribution of 16 risk factors to lung cancer burden among AYAs in 21 GBD-defined regions. Globally, smoking, secondhand smoke, household air pollution from solid fuel use, and ambient particulate matter pollution were identified as the main contributor to lung cancer burden among AYAs. The primary risk factors affecting both the general and male populations have exhibited relative stability over recent decades. Nevertheless, the main risk factors in women generally shifted to ambient particulate matter pollution and secondhand smoke, compared to household air pollution from solid fuel use and smoking ( Figure 4 ; Supplementary Figure 3 ). In regions characterized by high, high-middle, and middle SDI levels, smoking emerged as the predominant risk factor. Conversely, in low and low-middle SDI regions, the primary risk factor was household air pollution resulting from the use of solid fuels. Ambient particulate matter pollution was the main risk factor in women worldwide. While in men worldwide, the main risk factor remained smoking ( Figure 5 ; Supplementary Figure 4 ).Taken together, these findings highlight the varied contributions of risk factors to lung cancer burden among AYAs globally, setting the stage for deeper interpretation in the Discussion section.

In this study, firstly, the sources of GBD data largely depends on the availability and quality of GBD 2021 data. The GBD database relies on various data sources worldwide, including hospital records, death registries, censuses, health surveys, and academic research. However, the availability and quality of data vary significantly across countries and regions, particularly in resource-limited countries or regions, where incomplete or low-quality data are particularly prominent issues. This can affect the accuracy of the model’s estimation results. Secondly, the GBD 2021 database did not differentiate between bronchus and lung cancer, resulting in a lack of specific clinical information. Thirdly, the study used manual calculations to determine age-standardized rates for lung cancer among AYA population. While this method removes the impact of variations in age structure, it also restricts the ability to calculate uncertainty intervals. Fourth, the GBD risk attribution model also has certain methodological limitations. The GBD employs a counterfactual comparative risk assessment framework, which assumes that the health effects of individual risk factors are independent. However, in real-world settings, risk factors such as smoking, ambient air pollution, occupational exposures, and secondhand smoke frequently co-occur within the same populations, and may exhibit synergistic or interactive effects. The current model does not fully account for the nonlinear or non-additive impact of multiple simultaneous exposures, which may lead to either overestimation or underestimation of the disease burden attributable to a specific risk factor.

Fifth, we acknowledge that data quality and availability have a substantial impact on the estimation of lung cancer burden, particularly in LMICs, where these issues are especially prominent. Estimates of lung cancer burden may be influenced by missing or poor-quality data, which, in turn, can affect the accuracy of health inequality analyses. Although the GBD framework incorporates various statistical methods to address data sparsity and reporting heterogeneity, these approaches cannot fully eliminate the uncertainty and potential bias introduced by insufficient or low-quality data. One such important source of bias is survivorship bias, which is particularly relevant in populations with relatively low incidence and mortality rates, such as AYAs. In settings with limited resources, constrained case capture, and inadequate follow-up infrastructure, individuals with longer survival times are more likely to be recorded in cancer registries or health databases. In contrast, those with rapidly progressing disease or poor access to care may be systematically underrepresented or missed. Finally, although the GBD database provides a wealth of global and regional health data, there is still insufficient detail in terms of regional granularity and population stratification in some cases. For example, GBD data often focus on the overall disease burden of a country or region, overlooking differences between subpopulations within a country, such as ethnic minorities or specific genders, which prevents the reflection of inequalities at the micro level and leads to insufficiently targeted policy interventions.