A closed-loop comprehensive analysis of global burden data, future projections, risk factors, and molecular mechanisms on nasopharyngeal carcinoma

Abstract

Background:

Nasopharyngeal carcinoma (NPC), a frequent malignant tumor, exhibits a high propensity for both metastasis and recurrence. This study provided a comprehensive analysis of the burden, risk factors, and potential molecular mechanisms of NPC.

Methods:

Utilizing data from the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) 2021, we analyzed NPC burden in relation to age, sex, risk factors, and the sociodemographic index (SDI). The average annual percent change (AAPC) was computed through joinpoint regression. To assess causal links between risk factors and NPC, we employed two-sample Mendelian randomization (MR). Future NPC incidence trends (2022–2050) were projected using the Bayesian age–period–cohort model. We also elucidated the molecular mechanisms that trigger NPC by a bioinformatics analysis.

Results:

Global NPC mortality (AAPC: −1.817) and disability-adjusted life years (DALYs) rates (AAPC: −1.932) declined from 1990 to 2021. However, rising prevalence (AAPC: +0.317) and persistent high burden in East/Southeast Asia [e.g., Malaysia age-standardized incidence rate (ASIR): 6.09/100 000]. Socioeconomic disparities are widening, with low-SDI regions experiencing slower burden reduction. Projections suggest continued ASIR decline to 1.08/100 000 by 2050. MR analysis confirms smoking as a causal risk [current smokers odds ratio (OR) = 3.10; former smokers OR = 7.00, compared to never smoking], while Secretoglobin family 1A member 1 (SCGB1A1) downregulation emerges as a key mechanistic link-driving immunosuppression via Th2/Treg infiltration and hub genes (CXCL11, CD19).

Conclusion:

We innovatively constructed a closed-loop on NPC by multiple bioinformatic techniques. Global NPC mortality and DALYs have significantly declined, while this progress is counterbalanced by a rising prevalence and persistently high burden in East/Southeast Asia. Smoking was detected as a strong causal risk factor for NPC, while SCGB1A1 downregulation served as a novel key molecular driver of NPC pathogenesis.

Introduction

At present, nasopharyngeal carcinoma (NPC) exhibits significant geographical variations in its incidence and mortality rates worldwide. The incidence of NPC has declined significantly in South Asia and East Asia, which may be associated with tobacco control, dietary changes, and economic development[1]. However, despite advances in diagnostic and radiotherapy techniques, recurrence and metastasis remain the primary causes of treatment failure in NPC[2]. Epstein–Barr virus (EBV)-associated biomarkers, gene mutations, liquid biopsy, and DNA methylation technologies have significantly improved the accuracy of early detection, treatment selection, and disease monitoring for NPC molecular diagnosis[3]. Furthermore, metabolic reprogramming plays a critical role in the initiation and progression of NPC, involving significant alterations in glycolysis, lipid, and amino acid metabolism, which provide new targets for personalized therapy[4]. For NPC treatment, radiotherapy is the primary option, with chemotherapy–immunotherapy combination as an emerging trend. Immune checkpoint inhibitorss in recurrent/metastatic NPC have made breakthroughs, especially with chemotherapy[5]. Additionally, dysregulation of the proteasome signaling pathway is closely associated with tumor growth, invasion, distant metastasis, and immune evasion in NPC, suggesting that proteasome inhibitors may play a role in NPC immunotherapy[6]. Despite certain advances in treatment approaches, the management of NPC still faces numerous challenges. Local recurrence and distant metastasis remain major therapeutic difficulties, especially following radiotherapy[7]. Moreover, since the tumor microenvironment (TME) of NPC plays a crucial role in tumorigenesis, progression, and metastasis, understanding the complexity of the TME is essential for developing effective anticancer drugs[8].

HIGHLIGHTS

• Declining global mortality/DALYs: Global NPC mortality (AAPC: −1.817) and DALY rates (AAPC: −1.932) significantly declined from 1990 to 2021.

• Future projection: Projections suggest a continued decline in ASIR, reaching 1.08/100 000 by 2050.

• Smoking as a strong causal risk factor: MR confirmed smoking as causal.

• Novel molecular mechanism identified: Bioinformatics analysis revealed SCGB1A1 downregulation as a key molecular driver of NPC pathogenesis. This drives immunosuppression via Th2/Treg infiltration and involves hub genes (CXCL11, CD19).

• Innovative integrated approach: The study innovatively combined multiple techniques (epidemiology, MR, projection modeling, bioinformatics) to construct a “closed-loop” understanding of NPC burden, risks, and mechanisms.

At present, the burden trends, diagnosis, and treatment of NPC have garnered remarkable attention from clinicians and researchers[9]. The Global Burden of Disease (GBD) studies quantify NPC impact through metrics like disability-adjusted life years (DALYs), revealing persistent challenges despite diagnostic and therapeutic advances. For instance, in 2019, global NPC incidence reached 0.96 per 100 000, with mortality at 0.20 per 100 000, reflecting significant regional disparities[10]. However, this study was based on data from GBD 2019 and focused on the disease burden in specific age groups (adolescents and young adults). To have an up-to-date and comprehensive understanding of the NPC disease burden, we used the latest data from the GBD to analyze the burden of NPC in different locations, ages, sexes and sociodemographic index (SDI) levels from 1990 to 2021. To gain a current and all-encompassing grasp of the disease burden associated with NPC, we utilized the most recent data from the GBD study to examine the NPC burden across various regions, age groups, genders, and SDI levels between 1990 and 2021. Additionally, we employed Mendelian randomization (MR) to offer complementary perspectives on the risk factors that play a role in the burden of NPC. Last, we explored the molecular mechanisms and therapeutic targets of NPC, focusing on this potential risk factor to improve treatment efficacy and patient prognosis. In this study, a closed-loop analysis that intentionally bridged macroscopic epidemiological trends with microscopic molecular mechanisms of NPC was conducted by studying NPC’s “Epidemiology & Projection,” “Risk Factors & Causal Inference,” and “Molecular Biology.”

Methods

Data sources

This work has been reported in line with the CHEERS criteria[11] (Supplemental Digital Content Materials, available at: http://links.lww.com/JS9/H101). This study was conducted without using artificial intelligence (AI) tools in accordance with the TITAN Guidelines 2025[12] (Supplemental Digital Content Materials, available at: http://links.lww.com/JS9/H101). During the bioinformatics analysis (e.g., initial data processing with GEO, and differential expression analysis using R packages), no generative AI models were used to execute, interpret, or generate code or results. Standard, validated bioinformatics software and scripts were employed. Besides, AI was not used to generate scientific content, formulate hypotheses, interpret data, or draw conclusions. Utilizing the GBD database, this study conducted an in-depth assessment of health loss attributable to 371 diseases, injuries, and impairments, as well as 88 risk factors, across 204 countries and territories from 1990 to 2021. In the present research, the prevalence, incidence, mortality, and DALYs associated with NPC, along with their corresponding 95% uncertainty intervals (UIs), were extracted from the Global Health Data Exchange (GHDx) query tool (https://vizhub.healthdata.org/gbd-results/). All estimates are reported as counts and age-standardized rates (ASRs) per 100 000 population. Geographically, the GBD categorizes all countries into 21 distinct regions. Age groups encompass the entire population, segmented into 20 five-year intervals from under 5 to 95 years and older. The database also employs the SDI, a composite indicator quantifying the sociodemographic development of a region based on income per capita, educational attainment, and total fertility rate. Based on their SDI values (ranging from 0 to 1), the 204 countries and territories were further stratified into five quintiles: low SDI (<0.46), low-middle SDI (0.46–0.60), middle SDI (0.61–0.69), high-middle SDI (0.70–0.81), and high SDI (>0.81).

The detailed information related to the methodology of this study, including the age–period–cohort (APC) and the Bayesian age–period–cohort (BAPC) analysis, has been listed in Supplemental Digital Content Materials, available at: http://links.lww.com/JS9/H101.

MR analysis

The preceding GBD database analysis indicated that smoking is a risk factor for NPC, demonstrating a correlation between them. We obtained the percentage of NPC DALYs attributable to smoking as a risk factor. These attributable fractions were utilized to analyze regional variations in the impact of smoking on the NPC burden. Furthermore, a two-sample MR analysis was employed to investigate the causal relationship between the risk factor smoking (exposure) and NPC (disease outcome). The details of the methodology of the MR analysis were displayed in Supplemental Digital Content Materials, available at: http://links.lww.com/JS9/H101. Having established tobacco use as a major population-level risk factor and provided genetic evidence supporting its causal role in NPC, we next sought to investigate the potential molecular mechanisms through which smoking may contribute to NPC pathogenesis. We employed a comparative bioinformatics approach to identify genes dysregulated both by smoking exposure and in NPC tissue.

GEO analysis: investigating mechanisms underlying smoking’s impact on NPC

GEO data source and processing

Gene expression datasets relevant to smoking and NPC were retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) based on the following criteria: homo sapiens studies, RNA expression profiling by microarray, and samples including normal tissue and either smoking-exposed or tumor tissue. One qualifying smoking-related dataset: GSE16008 (gene expression from nasal epithelial cell samples of current smokers and never-smokers) was included. Four NPC datasets were applied, including GSE40290, GSE12452, GSE34573, and GSE53819. The four selected represent all the nonoverlapping, human, microarray-based NPC datasets in GEO that contained both primary NPC tumor samples and noncancerous nasopharyngeal/nasal control tissues at the time of our analysis. Datasets containing only cell lines, metastatic samples, or without proper normal controls were excluded.

Data analysis

GEO2R (https://www.ncbi.nlm.nih.gov/geo/geo2r) was used for data normalization and initial processing. Differential gene expression (DGE) analysis was performed on the aforementioned datasets using the R packages limma and GeoQuery. Genes meeting the thresholds of absolute log2 fold change |log2(FC)| ≥ 1 and adjusted P-value (adj. P) < 0.05 were considered differentially expressed. The Venn diagram online tool Jvenn (https://jvenn.toulouse.inra.fr/app/example.html) was employed to identify common differentially expressed genes (DEGs) across the resulting gene lists from the DGE analyses.

The methodological analyses of the Expression of the intersection gene in pan-cancer analysis and the Impact of the common differentially expressed gene (SCGB1A1) on NPC-associated gene expression (including the Immune cell infiltration estimation, Statistical validation & visualization, and the Identification of SCGB1A1-associated immune-related gene sets) were shown in Supplemental Digital Content Materials, available at: http://links.lww.com/JS9/H101.

Identification of hub genes

Protein–protein interaction (PPI) network construction

The STRING database analyzed protein–protein interactions (PPIs) in the Secretoglobin family 1A member 1 (SCGB1A1)-associated immune-related gene set (score > 0.4 retained). Cytoscape 3.10.1 visualized the network; CytoHubba maximal clique centrality (MCC), MCODE, and CytoNCA identified top hub genes, with overlaps as final hub genes. The methodological details were summarized in Supplemental Digital Content Materials, available at: http://links.lww.com/JS9/H101.

Functional enrichment analysis

Gene Ontology (GO) enrichment analysis was performed to characterize the SCGB1A1-associated immune-related gene set across three domains: Biological Processes (BPs), Cellular Components (CCs), and Molecular Functions (MFs). The Kyoto Encyclopedia of Genes and Genomes (KEGG) database was used to identify enriched pathways and elucidate the higher-level biological functions associated with the SCGB1A1 immune-related gene set. GO annotations were sourced from the Sangerbox platform. Up-to-date KEGG pathway gene annotations were retrieved via the KEGG REST API (https://www.kegg.jp/kegg/rest/keggapi.html). Enrichment analysis was conducted using the Sangerbox platform. The maximum and minimum gene set sizes were set to 5000 and 5 genes, respectively. The thresholds for statistical significance were defined as a nominal P-value < 0.05 and a false discovery rate < 0.05.

Results

Analysis of the global burden of NPC from 1990 to 2021

The global and regional trends in NPC burden from 1990 to 2021 are summarized in Figure 1 and Supplemental Digital Content Table S1, available at: http://links.lww.com/JS9/H94, with detailed country-level data provided in Supplemental Digital Content Table S2, available at: http://links.lww.com/JS9/H95, Supplemental Digital Content Table S3, available at: http://links.lww.com/JS9/H96, Supplemental Digital Content Table S4, available at: http://links.lww.com/JS9/H97, and Supplemental Digital Content Table S5, available at: http://links.lww.com/JS9/H98. Between 1990 and 2021, the global burden of NPC demonstrated a downward trend. The average annual percent change (AAPC) for the age-standardized incidence rate (ASIR), age-standardized mortality rate (ASMR), and age-standardized DALY rate (ASDR) were −0.752, −1.817, and −1.932, respectively. Conversely, the prevalence rate showed an upward trend, with an AAPC of 0.317. In 2021, the global ASIR was 1.38 per 100 000 population, the age-standardized prevalence rate (ASPR) was 6.14 per 100 000, resulting in an ASMR of 0.87 per 100 000, and an ASDR of 28.91 per 100 000 population (Supplemental Digital Content Table S1, available at: http://links.lww.com/JS9/H94).

Figure 1.

The ASPRs of nasopharynx cancer and the AAPC of ASPRs from 204 countries and territories. Notes: (A) ASPR in 2021. (B) AAPC of the ASPR from 1990 to 2021. AAPC, average annual percent change.

At the regional level in 2021, East Asia exhibited the highest ASIR and ASPR, along with relatively high levels of ASMR and ASDR. Southeast Asia had the highest ASMR and ASDR, along with relatively high levels of ASMR and ASDR. In contrast, Andean Latin America and Southern Latin America exhibited low levels of ASIR, ASPR, ASMR, and ASDR (Supplemental Digital Content Table S1, available at: http://links.lww.com/JS9/H94).

In 2021, among 204 countries and territories, the highest ASIRs were observed in Malaysia (6.09 per 100 000), Greenland (4.66 per 100 000), and Taiwan (Province of China) (4.27 per 100 000) (China: 3.42 per 100 000). The highest ASPRs were found in Greenland (15.33, 95% UI 12.21–19.21), Taiwan (Province of China) (22.60, 95% UI 17.22–29.29), Malaysia (22.53, 95% UI 19.02–26.60), China (18.43, 95% UI 14.73–22.97), and Singapore (10.95, 95% UI 8.46–13.97) (Fig. 1A). The highest ASDRs were in Equatorial Guinea (9.87 per 100 000), Uruguay (9.78 per 100 000), and Serbia (9.75 per 100 000). Notably, increasing ASIR trends from 1990 to 2021 were observed in countries such as Cabo Verde and Lesotho in Africa; Belize, Jamaica, and Paraguay in the Caribbean and Central America; and Bulgaria and Romania in the Balkan Peninsula (Supplemental Digital Content Table S2, available at: http://links.lww.com/JS9/H95, Supplemental Digital Content Table S3, available at: http://links.lww.com/JS9/H96, Supplemental Digital Content Table S4, available at: http://links.lww.com/JS9/H97, and Supplemental Digital Content Table S5, available at: http://links.lww.com/JS9/H98, Fig. 1B).

Disparities in the global burden of NPC by age and sex

The global distribution of age-standardized incidence, prevalence, mortality, and DALY rates for NPC by sex and age in 2021 is illustrated in Figure 2A–D, respectively. Males exhibited significantly higher NPC incidence, prevalence, mortality, and DALY rates than females. The peak age for NPC prevalence in both males and females was 50–54 years. The highest DALY rates occurred at 55–59 years for both sexes. Consequently, the 50–59 year-age group represented the interval with the highest absolute number of incident cases, prevalent cases, deaths, and DALYs for both males and females.

Figure 2.

The number and ASRs of incidence (A), prevalence (B), death (C), and DALYs (D) of NPC in 2021 by sex. Notes: DALYs, disability-adjusted life years. Dotted lines indicate 95% upper and lower UIs, respectively.

In age-specific patterns, among males, the incidence rate increased with advancing age until around 70 years, after which it declined. Among females, the incidence rate stabilized after approximately 55 years. Incidence rates rose with age, peaking in males aged 50–54 years (5.28 per 100 000, 95% UI 4.26–6.64), followed by a slight decline in older age groups. The peak incidence in females occurred at 55–59 years (1.95 per 100 000, 95% UI 1.65–2.30). The peak age for NPC prevalence in both males and females was 50–54 years (Fig. 2B), while the highest DALY rates occurred at 55–59 years (Fig. 2D).

In the aspect of sex differences, incidence rates were markedly higher in males than in females, particularly between the ages of 30 and 80 years. For example, incidence in males aged 50–54 years was 5.28 per 100 000 (95% UI 4.26–6.64), while incidence in females aged 50–54 years was 1.73 per 100 000 (95% UI 1.44–2.10).

The highest number of incident cases in males and females occurred at 50–54 years (11 720 cases, 95% UI 9448–14 731) and 55–59 years (3910 cases, 95% UI 3313–4622), respectively. Incidence was extremely low in children aged 5–9 years (males: 0.045 per 100 000; females: 0.028 per 100 000). No documented cases were observed in children under 5 years (Fig. 2A–D).

Disparities in the NPC burden by SDI

Analysis of NPC burden across countries/regions by SDI

Incidence

The correlation between age-standardized NPC incidence rates (per 100 000 population) and the SDI was analyzed across 204 countries and territories. Spearman correlation analysis revealed a weak, statistically nonsignificant negative correlation between SDI and NPC incidence [r = −0.0430, 95% confidence interval (CI): −0.1827 to 0.1036, P = 0.5412]. This indicates that NPC incidence globally is not significantly influenced by SDI. Notably, substantial geographical variation in NPC incidence was observed. For instance, high rates were seen in Malaysia (6.09 per 100 000), Taiwan (Province of China) (4.27 per 100 000), and mainland China (3.42 per 100 000), while low rates were found in Egypt (0.07 per 100 000), Peru (0.12 per 100 000), and Montenegro (0.13 per 100 000) (Fig. 3A). This distribution pattern likely reflects the influence of genetic susceptibility, environmental factors (e.g., EBV infection), and lifestyle (e.g., dietary habits), rather than socioeconomic development level alone.

Figure 3.

Relationship between SDI and NPC burden in 204 countries and regions. Incidence (A), prevalence (B), death (C), and DALYs (D).

Prevalence

A weak but statistically significant positive correlation was found between age-standardized NPC prevalence and SDI (Spearman r = 0.1372, 95% CI: 0.0025 to 0.2812, P = 5.044 × 10⁻²). This suggests that while a positive association exists between SDI and NPC prevalence, the strength of this association is weak, indicating that socioeconomic factors are unlikely to be primary drivers of NPC prevalence. Notably, high NPC prevalence was observed in both high-SDI regions [e.g., Taiwan (Province of China), Malaysia] and low-SDI regions (e.g., Algeria, Uganda). This further suggests that the distribution of the disease is likely more influenced by genetic, environmental, or infectious factors (e.g., EBV) than by SDI. Exceptionally high prevalence rates in regions like Greenland (15.33 per 100 000) and Tunisia (8.60 per 100 000) underscore the importance of region-specific risk factors (Fig. 3B).

Mortality and DALYs

Furthermore, Spearman correlation analysis showed a significant negative correlation between NPC mortality and SDI (r = −0.2130, 95% CI: −0.3532 to −0.0656; P = 2.271 × 10⁻³) (Fig. 3C). A significant negative correlation was also observed between SDI and the age-standardized NPC DALY rate (r = −0.2488, 95% CI: −0.3851 to −0.1103, P = 3.470 × 10⁻⁴) (Fig. 3D).

Temporal trends in NPC Burden by SDI, 1990–2021

Incidence

Between 1990 and 2021, the global burden of NPC was significantly influenced by SDI. High-SDI regions: Male incidence remained stable at 1.08 per 100 000 (95% UI 1.00–1.17), while female incidence was 0.35 per 100 000 (95% UI 0.31–0.38) (Fig. 4A). Overall, a slow declining trend was observed (average annual decrease of 0.3%). High-middle-SDI regions: Male incidence increased from 3.50 (95% UI 2.86–4.20) to 3.69 per 100 000 (95% UI 2.84–4.73), while female incidence rose from 1.27 (95% UI 1.02–1.56) to 1.13 per 100 000 (95% UI 0.87–1.49). This indicates a significant increase in male incidence (P < 0.05; Fig. 4A). Middle-SDI regions: Male incidence decreased from 3.25 (95% UI 2.71–3.76) to 2.35 per 100 000 (95% UI 1.97–2.78) by 2021, stabilizing at this level (Fig. 4A).

Figure 4.

Variation of ASR of NPC incidence (A), prevalence (B), death (C), and DALYs (D) over time in regions with different SDI.

Prevalence

High-SDI regions: Prevalence was relatively low and stable (2021: 3.62 per 100 000, 95% UI 3.38–3.87). High-middle-SDI regions: Prevalence was the highest and showed significant growth, peaking at 12.36 per 100 000 (95% UI 9.86–15.03) in 2019. Low-SDI regions: Prevalence was the lowest (2021: 2.49 per 100 000, 95% UI 2.03–3.06) (Fig. 4B).

Mortality

High-SDI regions: Mortality was the lowest and showed a continuous decline (1990: 0.59 per 100 000 vs. 2021: 0.35 per 100 000). High-middle-SDI regions: Mortality was highest in 1990 (1.93 per 100 000), declining to 1.03 per 100 000 (95% UI 0.87–1.25) by 2021. However, male mortality remained high (1.66 per 100 000). Low-middle-SDI regions: Mortality in 2021 was 0.92 per 100 000 (95% UI 0.82–1.04), with a slower rate of decline compared to high-SDI regions (Fig. 4C).

DALY rates

High-SDI regions: Consistently maintained the lowest burden, with a DALY rate of only 11.11 per 100 000 (95% UI 10.34–11.92) in 2021. High-middle-SDI regions: Bore the highest burden in 1990 (66.62 per 100 000, 95% UI 57.27–76.20), which declined to 34.84 per 100 000 (95% UI 29.05–41.64) by 2021. Middle-SDI regions: DALY rate in 2021 was 33.13 per 100 000 (95% UI 29.12–37.66), showing a steadily declining trend. Low-middle-SDI regions: DALY rate in 2021 was 30.78 per 100 000 (95% UI 27.10–34.87). Low-SDI regions: DALY rate remained at 30.29 per 100 000 (95% UI 24.58–37.47) in 2021 (Fig. 4D).

Overall, the magnitude of decline in NPC burden was greatest in high-SDI regions and smallest in low-SDI regions. This pattern suggests a significant correlation between the NPC disease burden and the level of socioeconomic development, highlighting the ongoing need for strengthened prevention and control efforts in low-middle and low-SDI regions.

Impact of risk factors on the burden of NPC

Temporal trends in attributable DALY rates

GBD data indicate that the burden of NPC is associated with several risk factors, including high alcohol use, occupational exposure to carcinogens, and tobacco use. This study analyzed trends in the percentage of NPC DALYs attributable to these risk factors globally and across SDI regions from 1990 to 2021 (Fig. 5A).

Figure 5.

Impact of different attributable risk factors on the attribution proportion of NPC DALYs (A) and deaths (B) over time.

Key global trends

High alcohol use: The global attributable DALY rate decreased from 10.98 per 100 000 in 1990 to 6.44 per 100 000 in 2021, representing a reduction of 41.3%.

Tobacco use: The attributable DALY rate declined from 8.79 per 100 000 in 1990 to 4.43 per 100 000 in 2021, a reduction of 49.6%.

Occupational exposure: Attributable risk remained relatively low and stable (1990: 0.45 per 100 000 vs. 2021: 0.30 per 100 000).

Regional variations

High alcohol use: High-SDI regions showed a significant reduction (1990: 6.75 per 100 000 → 2021: 3.45 per 100 000; reduction of 48.9%), while Low-SDI regions exhibited a slower decline (1990: 4.35 per 100 000 → 2021: 4.56 per 100 000), even showing a slight increase.

Tobacco use: The highest burden was in high-middle-SDI regions in 1990 (12.71 per 100 000), but this saw a marked decline by 2021 (7.15 per 100 000). Low-SDI regions had the smallest reduction (1990: 2.78 per 100 000 → 2021: 1.56 per 100 000).

Occupational exposure: High-middle-SDI regions had the highest attributable rate (1990: 0.66 per 100 000), remaining elevated in 2021 (0.45 per 100 000), highlighting the need for enhanced occupational protection measures.

Temporal trends in attributable death rates

Global trends (Fig. 5B): High alcohol use was the leading attributable factor for NPC deaths globally (attributable mortality rate: 0.311 per 100 000 in 1990 → 0.188 per 100 000 in 2021), showing a significant declining trend (average annual decrease of 1.3%).

Tobacco-related mortality was the second leading factor (1990: 0.289 per 100 000 → 2021: 0.153 per 100 000), declining at a faster rate (average annual decrease of 2.1%). The proportion attributable to occupational exposure was the lowest (1990: 0.010 per 100 000 → 2021: 0.007 per 100 000), with a reduction of 32%.

Regional variations

High-SDI regions: Had the lowest attributable mortality rates for both tobacco (0.069 per 100 000) and high alcohol use (0.105 per 100 000) in 2021. Occupational exposure risk was very low (0.002 per 100 000).

High-middle-SDI regions: Bore the heaviest attributable burden from high alcohol use (0.276 per 100 000 in 2021). Tobacco-related mortality was also relatively high (0.240 per 100 000).

Middle-SDI and Low-SDI regions: Occupational exposure risk was relatively higher (e.g., middle-SDI: 0.009 per 100 000 in 2021), and the decline in high alcohol use-related mortality was slower.

Comparative analysis of global and regional risk attributions for NPC burden: 1990 vs. 2021

1990

Global level: Tobacco was the leading risk factor for NPC DALYs (8.79 per 100 000), followed by high alcohol intake (10.98 per 100 000). Occupational exposures contributed the least (0.45 per 100 000) (Fig. 6A).

Figure 6.

Risk attribution in 21 regions in 1990 (A) and 2021 (B).

Regional variations: East Asia exhibited the most significant attributable burden, with high alcohol intake causing a DALY rate of 32.52 per 100 000 and tobacco-related DALYs at 25.23 per 100 000 – both substantially higher than global averages.

Sub-Saharan Africa: High alcohol intake DALY rates were notably elevated in Eastern Sub-Saharan Africa (8.88 per 100 000). The impact of occupational exposures was relatively higher in Southeast Asia (0.53 per 100 000) and Eastern Sub-Saharan Africa (0.33 per 100 000).

High-income regions (e.g., Western Europe, North America) showed alcohol and tobacco as the primary attributable factors, but absolute rates were lower than in East Asia (e.g., the high alcohol DALY rate in Western Europe was 7.13 per 100 000 compared to East Asia’s 32.52 per 100 000).

2021

Global trends: Alcohol consumption (6.44 DALYs per 100 000) and tobacco use (4.43 DALYs per 100 000) remained the primary risk factors for global nasopharyngeal cancer DALYs, with occupational exposures having a lower impact (0.30 DALYs per 100 000) (Fig. 6B).

Regional variations

East Asia had the highest alcohol-attributable DALY burden (14.13 DALYs per 100 000), significantly exceeding the global average.

East Asia and Southeast Asia: These regions had the highest DALY burdens for both tobacco and alcohol (East Asia: tobacco 10.57, alcohol 14.13; Southeast Asia: tobacco 6.79, alcohol 9.31 per 100 000). Sub-Saharan Africa: Alcohol-attributable DALY burden was exceptionally high in Eastern Sub-Saharan Africa (9.44 DALYs per 100 000), potentially linked to alcohol consumption patterns or synergistic interactions with EBV infection.

High-income regions: DALY burdens were lower in high-income Asia Pacific and Western Europe, though alcohol remained a major factor (e.g., high-income Asia Pacific: 2.54 DALYs per 100 000).

Occupational exposures: The attributable DALY burden from occupational exposures was consistently low globally but relatively higher in Southeast Asia (0.55 DALYs per 100 000) and Eastern Sub-Saharan Africa (0.38 DALYs per 100 000), possibly associated with exposures to industrial or agricultural chemicals.

Analysis of socioeconomic inequalities in NPC

This analysis employed regression-based curves to calculate the Slope Index of Inequality (SII) to assess global trends in socioeconomic inequalities in nasopharyngeal cancer incidence from 1990 to 2021.

Incidence

The SII in 1990 was −0.14 (95% CI: −0.38, 0.11), indicating a slightly higher incidence rate among low-income groups, although the difference was not statistically significant (as the CI includes zero). The SII in 2021 was −0.21 (95% CI: −0.40, −0.02), implying a statistically significantly higher incidence rate among socioeconomically disadvantaged groups compared to high-income groups (as the CI does not include zero) (Fig. 7A).

Figure 7.

Analysis of socioeconomic inequalities in NPC. Incidence rate (A), prevalence rate (B), and death rate (C).

Prevalence

1990: SII = 0.22 (95% CI: −0.50, 0.93); 2021: SII = 0.19 (95% CI: −0.40, 0.77) (note: Positive SII values for prevalence suggest higher prevalence in higher socioeconomic groups, but neither result reached statistical significance as both CIs include zero) (Fig. 7B).

Mortality

1990: SII = −0.26 (95% CI: −0.49, −0.03) – indicating significantly higher mortality in lower socioeconomic groups. 2021: SII = −0.36 (95% CI: −0.54, −0.18) – indicating a statistically significant and increased disparity in mortality favoring higher socioeconomic groups compared to 1990 (Fig. 7C).

DALYs

1990: SII = −9.27 (95% CI: −16.84, −1.71) per 100 000. This indicates a significantly higher DALY burden of 9.27 per 100 000 among the lowest socioeconomic group compared to the highest group. 2021: SII = −12.21 (95% CI: −17.84, −6.58) per 100 000. This indicates a statistically significantly higher burden and a widening of the disparity gap compared to 1990, with an absolute increase of 2.94 per 100 000 burden on socioeconomically disadvantaged groups.

Projected global burden of NPC, 2022–2050

Age-standardized incidence rate (ASIR)

Using GBD data and BAPC modeling, this study analyzed global NPC ASIR trends from 1990 to 2050. For NPC, where early-life and lifelong environmental exposures (like EBV, salted food) are critical etiological factors, accounting for cohort effects is essential for understanding long-term trends and making plausible future projections. The BAPC framework is specifically built to solve this identification problem using Bayesian constraints, making it the gold-standard for APC analysis.

Historical trends (1990–2020): Global ASIR initially declined (1990: 1.74; 95% CI: 1.73–1.75) to a nadir in 2000 (1.58; 95% CI: 1.57–1.59), followed by a mild increase post-2010 (2010: 1.31; 95% CI: 1.31–1.32). The 2020 ASIR was 1.38 (95% CI: 1.37–1.38), marginally higher than 2019 (1.37; 95% CI: 1.36–1.38).

Projections (2022–2050): ASIR is projected to decline slowly from 1.38 (2021; 95% CI: 1.38–1.39) to 1.08 (2050; 95% CI: 0.98–1.17). Uncertainty widens substantially over time, with the 2050 prediction interval spanning 0.80–1.35 (Fig. 8A).

Figure 8.

Projected global burden of NPC, 2022–2050. (A) ASIR, (B) ASPR, (C) ASMR, and (D) ASDR.

Age-standardized prevalence rate (ASPR)

Historical trends (1990–2021): ASPR increased from 5.57/100 000 in 1990 to 6.14/100 000 in 2021 (95% CI: 6.13–6.15), reflecting a 0.3% average annual growth. Acceleration post-2013 (2013: 5.36 vs. 2021: 6.14) may be associated with improved diagnostics and increased exposure to risk factors (e.g., EBV infection, smoking, salted food consumption).

Projections (2022–2050): ASPR is expected to peak at 6.34/100 000 in 2025 (95% CI: 5.99–6.68) before gradually declining to 5.64/100 000 by 2050 (95% CI: 4.09–7.20). Progressively widening prediction intervals (2025 CI width: 0.69; 2050: 3.11) indicate increasing uncertainty in long-term projections, warranting caution regarding extrapolation validity (Fig. 8B).

Age-standardized mortality rate (ASMR)

Historical trends (1990–2020): ASMR demonstrated a sustained decline from 1.69/100 000 to 0.97/100 000 (95% CI: 0.96–0.98), averaging a 1.8% annual reduction. This reflects advancements in treatment and improved control of risk factors (e.g., EBV prevention, smoking reduction).

Projections (2021–2050): The decline is expected to decelerate post-2021, reaching 0.49/100 000 by 2050 (95% CI: 0.37–0.62). Expanding CIs suggest growing uncertainty in distant forecasts, potentially linked to emerging risks (e.g., population aging, regional healthcare disparities) (Fig. 8C).

Age-standardized DALY rate

Historical trends (1990–2021): DALY rates declined significantly from 52.29/100 000 to 28.95/100 000 (95% CI: 28.91–29.00), averaging a 1.45% annual decrease. This underscores global progress in NPC control (e.g., EBV prevention, early screening, treatment advances).

Projections (2022–2050): Rates are projected to fall further to 14.15/100 000 by 2050 (95% CI: 12.93–15.37), albeit at a reduced pace (1.03% annual decline). Notably, prediction intervals widen substantially post-2022 (e.g., 2023: 26.12–29.01), indicating heightened uncertainty potentially attributable to emerging risks (e.g., aging populations, environmental changes) or model extrapolation limitations (Fig. 8D).

Region-specific risk factors

As shown in Supplemental Digital Content Figure S1, available at: http://links.lww.com/JS9/H92, the global burden of NPC has declined significantly, largely attributable to the management of behavioral risk factors, particularly reduced smoking and alcohol consumption. This underscores the effectiveness of public health interventions, such as tobacco control policies and alcohol restrictions. However, the higher burden observed in middle and high SDI regions highlights the need for targeted intensification of risk factor management. This includes promoting early screening and health education in high-incidence areas of Asia. Although occupational risks remain relatively low, they should not be overlooked, and occupational safety regulations should be strengthened to prevent associated cancers.

As shown in Supplemental Digital Content Figure S2, available at: http://links.lww.com/JS9/H93, based on ASR data from the GBD database, this study analyzed trends in NPC mortality attributable to various risk factors across Southeast Asia, East Asia, and Oceania from 1990 to 2021. The results demonstrated significant declines in ASMRs for NPC deaths attributable to alcohol use, smoking, occupational risks, and all combined risk factors, although these trends have shown stabilization in recent years. Our study reveals substantial improvements in the NPC burden across Southeast Asia, East Asia, and Oceania, primarily attributable to the effective control of smoking and alcohol-related risk factors. However, the slowing decline observed after 2010 suggests that current prevention and control strategies may be facing a plateau. Although the burden from occupational risks remains relatively low, its limited reduction warrants targeted protective interventions for specific occupational groups. These findings provide a critical foundation for optimizing NPC prevention strategies, underscoring the need for sustained, differentiated interventions tailored to distinct risk factors. In summary, this study reveals positive developments in NPC mortality risk attribution within the region, while underscoring the necessity for continued monitoring and optimized intervention strategies to achieve global disease control targets.

MR analysis confirms smoking as a risk factor for NPC

Since the aforementioned GBD results indicated that tobacco was the leading risk factor for NPC DALYs, we further conducted MR analysis to evaluate the causal relationship between smoking and NPC risk. Using the inverse-variance weighted method as the primary MR analytical approach, the results demonstrated current smokers had a significantly higher risk of developing NPC compared to never-smokers [odds ratio (OR) = 3.10; 95% CI: 1.00–9.61] (Fig. 9A). Even former smokers exhibited a substantially elevated risk relative to never-smokers (OR = 7.00; 95% CI: 1.65–29.76) (Fig. 9B). Analyses for both comparisons are detailed in the supplementary materials (Supplemental Digital Content Table S6, available at: http://links.lww.com/JS9/H99 and Supplemental Digital Content Table S7, available at: http://links.lww.com/JS9/H100). Based on the GBD database, our study identified tobacco use, occupational exposures, and alcohol intake as relevant risk factors associated with residential environmental factors, occupational environmental risk factors, genetic (or metabolic) factors, and behavioral factors. Consequently, MR analysis was further conducted using databases such as IEU, UK Biobank, and the GWAS Catalog to confirm that smoking increases the incidence of NPC in large population samples. Additionally, we utilized various GWAS databases to investigate the relationship between EBV infection and various preserved foods (such as bacon intake, salt, salty food liking, sausage intake, spreads and sauces consumed, chutney pickle, and vegetarian sausage burgers intake) with NPC. The MR results indicated no direct association between EBV infection and the consumption of various preserved foods. According to the findings from the GBD and MR analyses, we conclude that the pathogenesis of NPC is multifactorial, involving synergistic interactions among genetic susceptibility, EBV infection, and environmental exposures.

Figure 9.

MR analysis confirms smoking as a risk factor for NPC (forest plot of MR analysis). (A) MR-forestplots of the exposure (ever vs. never smoked, ieu-a-962) with four analysis methods; and (B) MR-forestplots of the exposure of smoking initiation in ieu-b-4877 with four analysis methods.

Exploring the mechanism of smoking on human NPC based on bioinformatics analysis

Finally, we investigated the potential mechanisms underlying smoking’s influence on NPC development. Having established tobacco use as a leading modifiable risk factor for NPC burden (GBD results) and confirmed its potential causal role through MR, we sought to elucidate the underlying molecular mechanisms connecting smoking to NPC pathogenesis. To this end, we adopted a systematic, multi-step bioinformatics approach to identify and prioritize candidate genes that may mediate this relationship. Comparative gene expression analysis of nasal epithelial cells from smokers versus never-smokers, alongside NPC tumor tissue, revealed that SCGB1A1 was the only gene consistently downregulated in both smokers and NPC patients (Fig. 10A, B). We have searched the PubMed and EMBASE databases and found that the specific role of SCGB1A1, a secretoglobin with immunomodulatory functions, in the smoking–NPC axis remains largely unexplored and represents a novel biological pathway. SCGB1A1 plays multiple critical roles in humans. Pan-cancer analysis demonstrated its significant downregulation in 20 cancer types, including GBM, HNSC, LUAD, and LUSC (Fig. 10C). Upregulation was observed only in CESC, OV, and PCPG (Fig. 10C). Protein-level analysis further confirmed its low expression in HNSC, LUAD, and LUSC. These findings suggest SCGB1A1 may function as a tumor suppressor gene. In the GSE12452 dataset, samples with high SCGB1A1 expression showed 426 upregulated and 270 downregulated genes (Fig. 10D, E).

Figure 10.

Mechanisms of smoking on human NPC based on bioinformatics analysis. (A) Commonly highly expressed intersection; (B) commonly down-expressed intersection; (C) expression profile of SCGB1A1 in pan-cancer; (D) heat map of high SCGB1A1 expression showed 426 upregulated and 270 downregulated genes; and (E) volcano plot of high SCGB1A1 expression on upregulated and downregulated genes.

Analysis of the immune microenvironment in NPC

Principal component analysis (PCA) revealed significant differences (P < 0.001) in immune cell infiltration profiles between normal and NPC samples (Fig. 11A), strongly suggesting involvement of the immune microenvironment in NPC pathogenesis. Immune Cell Infiltration Profiling (XCELL algorithm) quantified the specific infiltration levels of 64 immune cell populations, visualized as a heatmap (Fig. 11B) and box plots (Fig. 11C).

Figure 11.

Analysis of the immune microenvironment in NPC. (A) PCA in immune cell infiltration profiles between normal and NPC samples; Immune Cell Infiltration Profiling (XCELL algorithm) quantified the specific infiltration levels of 64 immune cell populations, visualized as a heatmap (B) and box plots (C); (D) immune cell correlation analysis.

Differentially infiltrating immune cells: Thirty immune cell types showed the most significant differences between NPC and normal tissues, prominently including Th2 cells, epithelial cells, eosinophils, common lymphoid progenitor, megakaryocyte–erythroid progenitor, macrophages M1, CD4⁺ Tcm (central memory T cells), megakaryocytes, endothelial cells, and myocytes. Compared to normal tissue, NPC tissue exhibited significantly enriched infiltration of aDCs (activated dendritic cells), Th2 cells, astrocytes, CD4⁺ T-cells, CD4⁺ memory T-cells, and Tregs (regulatory T cells). NPC tissue showed significantly depleted infiltration of B-cells, epithelial cells, and memory B-cells (Fig. 11C).

Immune cell correlation analysis: Strong positive correlations (r > 0.6) were observed among pro-B-cells, memory B-cells, naive B-cells, and B-cells overall, as well as between aDCs and CD8⁺ Tem (effector memory T cells). The strongest negative correlation was identified between eosinophils and Th2 cells (Fig. 11D).

Integration of SCGB1A1-related genes with immune genes

Simultaneously, 3062 immune-related genes were retrieved from the ImmPort database. Intersecting the aforementioned 700 genes associated with SCGB1A1 expression with these 3062 immune-related genes yielded 120 immune-related SCGB1A1–differentially expressed genes (IODEGs). This set comprised 73 upregulated and 47 downregulated genes (Fig. 12A).

Figure 12.

Integration of SCGB1A1-related genes with immune genes. (A) IODEGs; (B) BP; and (C) CC.

GO enrichment analysis

BP: IODEGs were significantly enriched in pathways regulating immune cell activities, including “immune system process,” “regulation of immune response,” and “cell activation.” CC: IODEGs were significantly enriched in components related to the “extracellular region,” “extracellular matrix,” and “vesicles” (Fig. 12B, C). MF: IODEGs were significantly enriched in functions related to “signaling receptor binding,” “molecular function regulator,” “peptidase regulator activity,” and “cytokine activity.” KEGG pathway enrichment: The most significantly enriched KEGG pathways for IODEGs were “cytokine–cytokine receptor interaction,” “complement and coagulation cascades,” and the “IL-17 signaling pathway.”

Collectively, the functional annotation strongly indicates that IODEGs are closely linked to immune cell functions and regulation, reinforcing the connection between SCGB1A1 and the immune microenvironment in NPC pathogenesis.

Identification and analysis of hub genes within the IODEG network

Network construction and visualization: The PPI network for the IODEGs was generated using the STRING database and visualized in Cytoscape (Fig. 13A, B). Hub gene identification: Hub genes were identified using three distinct Cytoscape plugins. CytoHubba (MCC algorithm): The top 10 hub genes were CXCL11, CD1D, GZMB, CCR6, CD19, IL15, CXCL10, IL33, JAK2, and CXCL1 (Fig. 13C). MCODE: The highest-scoring module (subnetwork) contained the hub genes: CXCL11, CXCL1, MYC, JAK2, KLRB1, CXCL10, CD1C, CD19, FGF2, GZMB, CD1D, and CCL8 (Fig. 13D). CytoNCA (degree centrality): The top 10 hub genes based on degree centrality were FN1, CD19, CXCL10, CXCL1, MYC, LCN2, GZMB, MUC1, CXCL11, and IL15 (Fig. 13E). Consensus hub genes: Genes identified as hubs by all three methods were CXCL11, GZMB, CD19, CXCL10, and CXCL1 (Fig. 13F).

Figure 13.

Identification and analysis of hub genes within the IODEG network. (A) STRING database for the PPI network; (B) Cytoscape for the PPI; (C) CytoHubba (MCC algorithm) for the top 10 hub genes; (D) the highest-scoring module (subnetwork) contained the hub genes; (E) CytoNCA (degree centrality) for the top 10 hub genes; (F) consensus hub genes; and (G) functional annotation of consensus hub genes.

Functional annotation of consensus hub genes

GO enrichment analysis revealed that the consensus hub genes (CXCL11, GZMB, CD19, CXCL10, and CXCL1), along with SCGB1A1, are involved in key BPs including: immune system process, regulation of response to stimulus, cell proliferation, response to external stimulus, and regulation of MF (Fig. 13G). This functional association suggests a potential role for SCGB1A1 in the progression of NPC, potentially mediated through these hub genes and their linked immune and regulatory pathways.

Discussion

This study delineates the complex evolution of NPC burden from 1990 to 2021 and its projections to 2050, revealing persistent geographical inequities, widening socioeconomic disparities, and uncovering novel molecular mechanisms that link tobacco use to NPC pathogenesis. Globally, age-standardized incidence, mortality, and DALY rates declined (AAPCs: −0.752, −1.817, and −1.932), while prevalence rose (AAPC: 0.317). In 2021, East Asia had a high incidence/prevalence, Southeast Asia had high mortality/DALYs, and Andean/Southern Latin America had low burdens. Malaysia, Greenland, and Taiwan (Province of China) had the top incidences. Males had higher NPC rates than females, with peak prevalence at 50–54 and DALYs at 55–59. SDI weakly correlated with incidence/prevalence but negatively with mortality/DALYs. High-SDI regions saw the greatest burden decline, low-SDI the least. Incidence, mortality, and DALYs may decline by 2050, but prevalence peaks in 2025. While the precise point estimates for 2050 are subject to substantial uncertainty (as visualized by the widening CIs), the model consistently projects a continuing decline in ASIR and ASDR globally, albeit at a potentially slower rate in low-SDI regions. Socioeconomic inequalities persist, with low-SDI regions needing more support. Furthermore, we found that the risk factors, i.e., alcohol and tobacco use drove DALYs, with global reductions (41.3 and 49.6%), but slower in low-SDI areas. In addition, MR confirmed smoking’s causal role (current smokers: OR = 3.10; former: OR = 7.00). The point estimates indicated a substantially elevated risk, although the wide CIs call for cautious interpretation. Bioinformatics linked smoking to SCGB1A1 downregulation, a potential tumor suppressor, with hub genes (CXCL11, GZMB, etc.) involved in immune pathways.

The current study reveals that the NPC burden has evolved in a complex and geographically stratified manner over three decades, with divergent global trends directly reflecting disparities in healthcare access. The significant decline in the ASMR (AAPC: −1.817) and ASDR (AAPC: −1.932) globally signifies tangible progress in combating NPC. This improvement is largely attributable to advancements in key areas within higher-resource settings: enhanced screening programs for EBV, the primary etiological agent, allowing for earlier detection; refinements in radiotherapy techniques (e.g., intensity-modulated radiotherapy), improving tumor control while reducing toxicity^[13,14^]. These technological and preventive strides have demonstrably saved lives and reduced disability. However, this narrative of progress is counterbalanced by concerning counter-trends. The persistent rise in the ASPR (AAPC: +0.317) presents a critical paradox. While partly reflecting improved survival due to better treatments (leading to a larger pool of living patients), it also signals potential inadequacies in primary prevention and persistent challenges in timely diagnosis across many regions[15]. Furthermore, the burden remains stubbornly high and geographically concentrated. Regions like East Asia (e.g., Malaysia with ASIR 6.09/100 000) and Southeast Asia, alongside Eastern Sub-Saharan Africa, bear a disproportionately heavy burden. In Eastern Sub-Saharan Africa, the exceptionally high alcohol-attributable DALY rate (9.44/100 000 in 2021) exemplifies region-specific risk constellations. These persistent hotspots underscore unresolved challenges deeply rooted in local contexts: endemic EBV infection with potentially distinct viral strains; cultural and dietary factors like high consumption of salt-preserved foods; limited access to advanced diagnostics and specialized treatment centers; and weaker implementation or enforcement of tobacco and alcohol control policies^[16,17^]. The widening gap in mortality and DALY burden between low-SDI and high-SDI regions, quantified by the increasingly negative SII, starkly highlights how disparities in healthcare access, infrastructure, and public health capacity fundamentally shape the NPC landscape, perpetuating inequitable outcomes despite overall global gains. Geographical hotspots, i.e., Malaysia (ASIR: 6.09/100 000), Greenland (ASPR: 15.33/100 000), and Eastern Sub-Saharan Africa (alcohol-attributable DALYs: 9.44/100 000) reflect region-specific risks (e.g., salted diets, EBV endemicity, occupational exposures). The significantly negative SII for DALYs (−12.21/100 000 in 2021) and mortality (SII: −0.36) highlight accelerated disadvantages in low-SDI regions, where diagnostic delays and limited treatment access persist. Projected ASIR reductions by 2050 (1.08/100 000) remain uncertain in these regions due to aging populations and healthcare fragmentation. The long-term projections to 2050, while indicative of potential trajectories under current trends, are inherently speculative and highly sensitive to external shocks and policy interventions not captured by our model. The widening CIs explicitly reflect this growing uncertainty. Several key factors could cause future burden to deviate significantly from our forecasts. First, the development and widespread deployment of an effective prophylactic EBV vaccine would represent a paradigm shift, potentially drastically reducing incidence in endemic regions over several decades. Conversely, accelerated population aging in high-burden regions (e.g., East Asia) could exert upward pressure on case counts, even with stable or declining ASRs. Second, major technological or therapeutic breakthroughs (e.g., highly effective and accessible immunotherapies) could accelerate mortality declines beyond current trends. Third, significant changes in the prevalence of key risk factors, i.e., driven by public health policies, economic development, or cultural shifts, could alter incidence patterns. Finally, large-scale socioeconomic disruptions or climate change impacts on health systems could worsen outcomes in vulnerable regions. Therefore, these projections should be interpreted not as a definitive future but as a plausible baseline scenario that underscores the critical importance of proactive investments in prevention, early detection, and healthcare equity.

Our analysis indicated that disparities in NPC burden are primarily driven by differential access to healthcare resources. Higher-SDI regions benefit from advanced EBV screening and precision radiotherapy, enabling early detection and effective treatment. In contrast, low-SDI regions face diagnostic delays and limited treatment access, exacerbated by weaker public health infrastructure. Furthermore, educational gaps likely contribute to higher exposure to established risk factors, such as tobacco and occupational hazards, due to less awareness and slower adoption of preventive behaviors. Environmental factors, including endemic EBV strains and dietary patterns, further compound these inequities. These intersecting barriers, including structural, educational, and environmental, sustain the observed geographic and socioeconomic gradients in NPC outcomes. The profound and widening socioeconomic disparities in NPC mortality and DALYs, quantified by the increasingly negative SII, are not incidental but are driven by a cascade of structural and healthcare access barriers prevalent in low-SDI regions. First, diagnostic delays are pervasive. In contrast to high-SDI regions with established screening protocols and rapid access to specialists, patients in low-resource settings often present with advanced-stage disease. This delay stems from constrained healthcare infrastructure, lower symptom awareness, and financial deficiency. Second, even upon diagnosis, significant treatment gaps exist. The scarcity of advanced radiotherapy technology (e.g., IMRT), inconsistent chemotherapy supply, and limited supportive care capacity compromise treatment efficacy and completion. Data from the IAEA indicate radiotherapy machine availability in low-SDI regions may be fold lower than in high-SDI regions. Third, primary prevention is less effective in these settings. Our data show slower declines in tobacco- and alcohol-attributable burdens, reflecting weaker implementation and enforcement of public health policies. Concurrently, persistent occupational exposures highlight inadequate workplace safeguards. Together, these interlinked barriers, later diagnosis, suboptimal treatment, and unmitigated risk exposure, create a self-reinforcing cycle that perpetuates the high NPC burden in low-SDI regions and explains the attenuated rate of improvement observed over the past three decades.

Our MR confirms smoking as a causal NPC risk factor. Current smokers face a 3.10-fold higher risk (95% CI: 1.00–9.61), while former smokers retain elevated susceptibility (OR = 7.00), suggesting durable epigenetic impacts. This finding was consistent with a previous meta-analysis, which included 40 original studies, found that compared with never smokers, the pooled relative risk of NPC was 1.61 for current smokers, 1.28 for former smokers, and 1.62 for ever smokers[18]. Based on the relevant studies, smoking significantly increases the risk of NPC through the combined effects of multiple mechanisms, including direct damage by chemical carcinogens, chronic inflammation and oxidative stress, immunosuppression and synergistic effects with viruses, and abnormal signaling pathways. The risk increases with the intensity and duration of smoking, while it decreases with the extension of smoking cessation time, which is in line with the results of our study^[19,20^]. Our study indicated that former smokers have a higher OR than current smokers, which may be attributed to the following factors. First, long-term heavy smokers who develop early symptoms are more likely to quit, meaning the “former smoker” group could be enriched with individuals who already had subclinical disease. Individuals experiencing early, undiagnosed NPC symptoms may be more likely to quit smoking, artificially enriching the “former smoker” group with pre-clinical cases and inflating the observed risk. Second, the act of quitting itself might be driven by health-seeking behavior. Individuals experiencing early symptoms may cease smoking, creating a temporal sequence where quitting precedes diagnosis and inflating the risk association. Furthermore, the durable epigenetic impact of a substantial smoking history is a key factor. This complex interplay warrants further longitudinal investigation. Alcohol-attributable DALYs fell 41.3% globally (1990–2021), yet increased in low-SDI regions (e.g., Eastern SSA), indicating poor policy implementation. The exact role of alcohol in the risk of NPC is still controversial. A population-based case-control study suggested no significant association between alcohol and NPC risk[21], while some studies identified a positive association between alcohol and high risk of NPC^[22,23^]. Occupational risks remain critical in Southeast Asia (DALYs: 0.55/100 000) and Eastern SSA (0.38/100 000), demanding urgent workplace safety reforms. Previous study showed that NPC might be associated with several work-related exposure agents, such as kitchen air pollution and solvents[24]. While MR minimizes reverse causation, we cannot fully exclude the possibility of genetic pleiotropy, where variants influencing smoking cessation behavior also relate to biological pathways involved in early NPC development. Distinguishing between these mechanisms requires future longitudinal studies with detailed smoking history (duration, intensity, time since quitting) and serial molecular profiling.

Importantly, we first found that SCGB1A1 might play a novel immunological nexus linking smoking to NPC. Consistent downregulation in smokers and NPC patients (P < 0.001), validated across 20 cancers (e.g., HNSC, LUAD), indicating a tumor suppressor role in multiple cancers. SCGB1A1 may exert its pivotal role on smoking-mediated NPC by reprogramming the immune microenvironment. Based on the bioinformatics analysis, NPC tissues exhibited Th2/Treg enrichment and B-cell depletion, fostering immune evasion. SCGB1A1 has been found to be involved in the development and progression of various cancers, including non-small cell lung cancer[25], clear cell renal cell carcinoma[26], and cervical cancer[27]. IODEGs (e.g., CXCL11, CD19) regulate cytokine networks (IL-17 pathway) and extracellular matrix organization. Consensus hubs (CXCL11, GZMB, CD19, CXCL10, CXCL1) co-regulated with SCGB1A1 drive “immune system process” and “cell proliferation,” positioning SCGB1A1 as a master immune-metastatic regulator. Previous study indicated that SCGB1A1 served as an important regulator of splenic immune dysfunction in non-neoplastic diseases[28]. For its role in cancer, SCGB1A1 orchestrates multifaceted immunomodulation in cancer, such as inhibiting immunosuppressive cells, reshaping cytokine networks, and sensitizing tumors to immunotherapy^[29,30^]. Based on the current evidence related to the biological roles of SCGB1A1 in cancer^[31,32^] as well as the main findings from our study, SCGB1A1 downregulation in NPC, likely induced by smoking-related methylation or transcriptional repression, contributed to an immunosuppressive niche via Th2/Treg polarization. At present, only one study demonstrated that SCGB1A1 might serve as a promising therapeutic target for the management of head and neck squamous cell carcinoma[33]. The biological effects exerted by SCGB1A1 might be associated with its regulation on multiple key cancer-related signaling pathways, i.e., extracellular matrix receptor interaction, TGF-β, and tumor metabolism signaling pathways. In contrast, its upregulation in gynecological cancers[34] may reflect a distinct tissue-specific origin or oncogenic pathway, suggesting a dual role as either a tumor suppressor or context-dependent oncogene. Based on this evidence, SCGB1A1 may explore its potential as a pan-cancer immune regulator. While SCGB1A1 consistently correlates with immune pathways across cancers, its functional impact appears highly context-dependent, influenced by tissue microenvironment and etiology[35]. This duality proposes that SCGB1A1 is not a universal marker but a modular immune-metastatic factor whose role must be interpreted within specific tumor-immune contexts. Its dual roles as a therapeutic target and biomarker underscore potential for precision oncology, though tissue-specific contexts demand further exploration. Clinical translation of SCGB1A1-based strategies, particularly in combinatorial regimens, holds promise for overcoming immunotherapy resistance.

Research implications

Several studies showed the trends in the burden of NPC by analyzing the global, regional, and national epidemiology from 1990 to 2021. These studies focused on different epidemiological data from the GBD database, such as age and region. A previous study[36] demonstrated that the NPC incidence cases in middle-aged and elderly people increased by 58.2 %, the mortality numbers increased by 33.8 %, and the DALY increased by 42.1 %. Chen et al[37] reported that Chinese males experiencing higher incidence (5.16 per 100 000) and mortality rates (2.32 per 100 000) than females for developing NPC in 2021. Our study is different from the above research. We present an ambitious and wide-ranging study that integrates global epidemiological data by analyzing all the main countries or regions with age-stratification, risk factor assessment through MR, future incidence projections, and bioinformatic exploration of molecular mechanisms. We innovative attempt to build a “closed-loop” model linking global burden estimates with causal risk factors and molecular pathways. The strengths of the study include its comprehensive use of data sources, particularly the GBD 2021 dataset, combined with projection modeling and bioinformatics, which provides a broad yet interconnected perspective on NPC. The identification of smoking as a strong causal risk factor, supported by MR analysis, is an important confirmation of existing epidemiological evidence, while the discovery of SCGB1A1 downregulation as a potential molecular link between smoking and NPC pathogenesis adds a novel mechanistic dimension. Finally, the emphasis on socioeconomic disparities and future burden projections enhances the policy relevance of our study, highlighting areas where prevention and healthcare resources should be focused.

Our study demonstrated that tailored interventions in East Asia and SSA should combine EBV vaccination, alcohol/tobacco taxation, and occupational safety enforcement. This study establishes the distinct geographical clustering of NPC (e.g., high incidence in Malaysia and East Asia), identifies the peak age group (50–59 years), and confirms significant gender disparities (higher risk in males). These findings underscore the need to implement targeted screening programs in high-risk regions and populations (e.g., males over 50 years, individuals with a history of smoking or alcohol consumption). As a result, in these high-incidence regions, enhanced endoscopic screening and plasma EBV DNA testing could be prioritized for individuals in their 50s with a history of smoking, potentially leading to earlier stage diagnosis and more curative surgical/radiotherapeutic outcomes. For molecular mechanistic analysis, we revealed that increased infiltration of Th2 cells and Tregs, along with decreased B-cell populations, in NPC tissues. Furthermore, SCGB1A1-associated hub genes (e.g., CXCL11, GZMB) were implicated in immune regulation. These findings highlight the need to investigate the interactions between these immune cells and SCGB1A1 to elucidate the mechanisms underlying tumor immune evasion, such as how Tregs suppress antitumor immune responses. Restoring SCGB1A1 expression or blocking hub genes (e.g., CXCL10) may reverse immunosuppression. Therefore, potentially SCGB1A1 expression levels could serve as prognostic biomarkers to identify patients with a higher risk of aggressive disease or immunosuppressive TMEs. Besides, SCGB1A1 may also serve as a potential therapeutic target to overcome immunotherapy resistance. Future clinical trials could explore strategies to restore SCGB1A1 function (e.g., via demethylating agents) or to target its downstream immunosuppressive pathways (e.g., using anti-Treg therapies) in combination with existing immunotherapies, thereby potentially expanding the treatment arsenal for advanced NPC.

Limitations

Though multiple strengths were identified in our study, some limitations should be acknowledged. First, our analysis relied on the GBD dataset, which may be subject to biases due to variations in regional case reporting quality and diagnostic criteria. Second, critical risk factors such as EBV subtypes and dietary patterns were not incorporated into the analysis. Third, while the MR analysis confirmed a potential causal link between smoking and NPC risk, the sample size may have been insufficient to detect smaller effect sizes or interactions with other factors. Moreover, the analyses for EBV infection and dietary factors were critically underpowered due to the limited discovery of strong genetic instruments in the currently available GWAS. The nonsignificant results for these factors are therefore inconclusive and primarily reflect this power constraint, not necessarily the absence of a causal link. As a result, future MR investigations will require larger GWAS consortia or meta-analyses specifically focused on quantitative EBV serology and detailed dietary intake measures to generate instruments with sufficient power for definitive causal inference. Fourth, while our projections to 2050 provide valuable insights into potential future trajectories, they must be interpreted with caution. All forecasting models are simplifications of reality and are subject to considerable uncertainty. The widening CIs over time underscore the speculative nature of long-term predictions. These projections are contingent upon the continuation of current trends and do not account for unforeseen disruptions, such as the emergence of novel therapeutics (e.g., effective EBV vaccines), major shifts in healthcare systems, or large-scale environmental and socioeconomic changes. Therefore, these results should be viewed not as definitive predictions but as a plausible scenario that highlights the potential consequences of current inequalities and the urgent need for proactive policy interventions. Fifth, our forecasting model, while robust for identifying baseline trends, did not incorporate specific external variables such as granular data on future population growth, aging, or medical policy shifts. We recognize that conducting formal sensitivity analyses under different scenarios (e.g., successful implementation of EBV vaccination versus status quo) would greatly enhance the policy relevance of our projections. Sixth, in this study, we identified that SCGB1A1 might serve as a pivotal driver of immunosuppression in NPC by the comprehensive bioinformatic analyses of public datasets. While this integrative, closed-loop approach provides strong predictive insights and prioritizes high-probability targets, it lacks direct in vitro or in vivo functional validation. Consequently, the precise molecular mechanisms by which SCGB1A1 downregulation modulates the TME remain putative. Last, it is well-established that NPC comprises distinct molecular subtypes (e.g., those with differential EBV gene expression, TME features, and clinical outcomes) that likely possess unique etiological drivers and therapeutic vulnerabilities. Based on our findings, a logical next step is to assemble a cohort of NPC patients with detailed smoking history and multi-omics profiles. Unsupervised clustering could be employed to define robust subtypes, after which the distribution of our identified SCGB1A1 pathway and other risk factors across these subtypes can be systematically evaluated. This would directly test the hypothesis that environmental factors like smoking drive specific molecular etiologies within the broader spectrum of NPC.

Conclusion

This comprehensive study documents declining global NPC mortality and DALY rates from 1990 to 2021, reflecting advances in EBV control and treatment. However, rising prevalence and persistent high burden in East/Southeast Asia and Eastern Sub-Saharan Africa highlight unresolved challenges. Socioeconomic disparities are widening, with low-SDI regions experiencing slower burden reduction. MR analysis provided genetic evidence supporting smoking as a causal risk factor for NPC, while SCGB1A1 downregulation emerges as a key mechanistic link-driving immunosuppression via Th2/Treg infiltration and hub genes (CXCL11, CD19). Projections suggest continued ASIR decline in 2050, but uncertainties demand targeted interventions in high-risk regions and populations. This study offers critical guidance for developing region-specific public health policies, identifying the pivotal risk factor, and formulating prevention strategies based on underlying molecular mechanisms, which could potentially reduce NPC’s global impact.

Ethical approval

Not applicable.

Consent

Not applicable.

Sources of funding

This study was supported by the Science and Technology Plan Project of Quanzhou City (No. 2023NS060 for Yilong Xu and No. 2023N040S for Sihai Dai) and the Joint Innovation Project of the School of Medicine, Huaqiao University (No. 2024YX002).

Author contributions

YLX and SHD: Project development, Data collection; HQD and QYC: Data collection, Conceptualization; YLX: Methodology, Investigation; and YLX, XW, YYX, and MYH: Original draft, Methodology.

Conflicts of interest disclosure

The authors declare that they have no conflict of interest.

Research registration unique identifying number (UIN)

Not applicable.

Guarantor

Not applicable.

Provenance and peer review

Not commissioned, externally peer-reviewed.

Data availability statement

This study analyzed publicly available datasets. The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation, upon request.