Global burden of non-melanoma skin cancers among older adults: a comprehensive analysis using machine learning approaches (2025)

Non-melanoma skin cancers (NMSCs), comprising basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), represent the most common malignancies worldwide, with more cases diagnosed than all other cancers combined¹. The incidence of NMSCs has risen substantially in recent decades, particularly among older adults, who bear a disproportionate share of the disease burden^2,3. Current projections indicate that NMSC rates could reach almost 400,000 per year in the UK alone by 2025, driven by aging populations, increased ultraviolet (UV) exposure, and improved diagnostic capabilities⁴. Epidemiological studies have revealed significant regional and gender disparities in NMSC incidence and mortality, with higher rates observed in high Socio-Demographic Index (SDI) regions and among males^5,6. Epidemiological studies have revealed significant regional and gender disparities in NMSC incidence and mortality, with higher rates observed in high Socio-Demographic Index (SDI) regions and among males^7,8, with incidence peaking at approximately 66 years of age⁴.

The public health impact of NMSCs extends beyond clinical burden to occupational health concerns, with recent World Health Organization estimates indicating that nearly one in three deaths from NMSC is attributable to occupational exposure to solar UV radiation⁹. Despite this growing public health challenge, there remains a paucity of comprehensive analyses investigating global trends, socio-demographic patterns, and future projections using integrated multi-model approaches. Furthermore, due to inconsistent reporting to cancer registries globally, accurate worldwide incidence rates for NMSCs are difficult to determine, though millions of new cases are reported annually in the United States alone¹⁰.

Previous studies forecasting the future burden of NMSCs have primarily relied on single predictive models, such as time series or regression analyses^11,12. However, the accuracy and robustness of these predictions may be limited by the inherent assumptions and constraints of individual models. To address these limitations, our study employs an ensemble of eight machine learning algorithms, including Prophet, ARIMA, TBATS, ElasticNet, ETS, VAR, HoltWinters, and Theta, to generate more reliable and comprehensive forecasts of NMSC burden from 2022 to 2050. This multi-model approach allows for the integration of diverse predictive techniques, enabling us to capture complex temporal patterns and uncertainties in NMSC epidemiology.

In this study, we aim to provide a comprehensive analysis of the global burden of NMSCs among older adults from 1990 to 2021, with a focus on regional, gender, and socio-demographic disparities. By applying advanced epidemiological decomposition techniques and frontier analysis, we seek to identify key drivers of NMSC burden and highlight areas of greatest need for targeted interventions. Furthermore, our machine learning-based forecasting of future NMSC incidence, prevalence, and disability-adjusted life years (DALYs) from 2022 to 2050 will offer valuable insights to inform public health strategies and resource allocation. Through this integrated approach, we strive to advance the understanding of global NMSC epidemiology and provide actionable evidence to guide prevention and management efforts in the face of a growing global burden.

Data sources and study design

This study utilized data from the Global Burden of Disease Study 2021 (GBD 2021) database (https://ghdx.healthdata.org/gbd-2021), which provides comprehensive estimates for 204 countries and territories, 371 diseases and injuries, and 88 risk factors. We selected data on basal cell carcinoma (BCC) and squamous cell carcinoma (SCC) for individuals aged 60 years and above, focusing on incidence, prevalence, and disability-adjusted life years (DALYs) from 1990 to 2021. Both absolute numbers and rates per 100,000 population were extracted^13,14,15.

Disease definition

According to the GBD 2021 hierarchy, BCC and SCC are classified under non-communicable diseases, specifically as level 4 diseases under non-melanoma skin cancer. The International Classification of Diseases (ICD-10) codes for SCC include C44.02, C44.12-C44.129, C44.22-C44.229, C44.32-C44.329, C44.42, C44.52-C44.529, C44.62-C44.629, C44.72-C44.729, C44.82, C44.92, while BCC codes encompass C44.01, C44.11-C44.119, C44.21-C44.219, C44.31-C44.319, C44.41, C44.51-C44.519, C44.61-C44.619, C44.71-C44.719, C44.81, C44.91^13,15.

Data source reliability and estimation methods for NMSCs in GBD 2021

It is important to note that unlike most other cancers, NMSCs are not consistently recorded in cancer registries across many countries worldwide. The GBD 2021 methodology addresses this significant data gap through a complex estimation process. For regions with reliable registry data (primarily high-income countries), direct incidence and prevalence data are utilized. For regions with incomplete or absent registry data, GBD 2021 employs statistical models that extrapolate from regions with similar characteristics, incorporating covariates such as UV radiation exposure, skin pigmentation distribution, and healthcare access metrics.

These estimates are further refined using DisMod-MR 2.1, a Bayesian meta-regression tool that ensures internal consistency between incidence, prevalence, remission, and mortality. For NMSCs specifically, the modeling approach incorporates data from multiple sources, including hospital discharge records, insurance claims data, and targeted epidemiological studies, in addition to available registry data. Age-standardization further enables comparative analyses across regions with different population structures.

Statistical analysis

Estimated Annual Percentage Change (EAPC) Model.

To quantify temporal trends in disease burden, we calculated the EAPC by fitting a linear regression model to the natural logarithm of age-standardized rates (ASR) against time. The EAPC and its 95% confidence interval were derived from the regression coefficient and its standard error.

Decomposition analysis

To elucidate factors contributing to global disparities in disease burden, we employed decomposition analysis. This method partitioned overall health differences into contributions from various factors, including population growth, population aging, and epidemiological changes. Epidemiological changes encompass all non-demographic factors influencing disease burden. Specifically, these include: (1) altered UV exposure patterns due to behavioral changes (e.g., increased outdoor recreational activities, changing sun protection practices) and environmental factors (e.g., ozone depletion); (2) improvements in diagnostic capabilities and disease awareness leading to enhanced case detection; (3) evolving occupation-related exposures reflecting changing workforce distributions; (4) shifts in medical practice including screening protocols and diagnostic criteria; and (5) changing prevalence of risk factors such as immunosuppression rates, genetic predisposition, and prior history of skin cancer. These epidemiological components are analyzed distinctly from demographic elements such as population growth and aging to provide a more nuanced understanding of NMSC burden drivers.

Frontier analysis

We utilized Data Envelopment Analysis (DEA) to assess the relative performance of different countries and regions in health improvement. A frontier surface representing optimal health outcomes for given input levels was constructed, and efficiency scores were calculated as the distance of each country or region from this frontier.

Machine learning modeling

To forecast the global burden of BCC and SCC from 2022 to 2050, we employed eight advanced time series prediction methods and machine learning algorithms:

1. 1.

   Prophet: A procedure developed by Facebook for forecasting time series data, utilizing a decomposable time series model with trend, seasonality, and holiday effects.

2. 2.

   ARIMA (Autoregressive Integrated Moving Average): A classical time series model that combines autoregression, differencing, and moving average components.

3. 3.

   TBATS (Trigonometric, Box-Cox transform, ARMA errors, Trend, and Seasonal components): A complex time series model capable of handling multiple seasonal patterns.

4. 4.

   Elastic Net: A regularized regression method that combines L1 and L2 penalties of the Lasso and Ridge methods.

5. 5.

   ETS (Error, Trend, Seasonal): An exponential smoothing state space model that can capture error, trend, and seasonal components.

6. 6.

   VAR (Vector Autoregression): A multivariate time series model that captures linear interdependencies among multiple time series.

7. 7.

   Holt-Winters: An exponential smoothing approach for time series with trend and seasonal components.

8. 8.

   Modified Theta: An enhanced version of the Theta method, incorporating seasonal fluctuations and non-linear trends.

For each model, we implemented cross-validation techniques to assess predictive performance. The primary evaluation metrics included Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and coefficient of determination (R²). To enhance model stability and prevent overfitting, we applied various constraints and smoothing techniques, such as dynamic smoothing constraints and exponential decay factors for trend changes. Feature engineering was performed to improve model performance, including the creation of lagged features, moving averages, and interaction terms between population and year. All models were trained on data from 1990 to 2010, validated on data from 2011 to 2021, and used to forecast from 2022 to 2050. Model performance was comprehensively evaluated using six metrics: Mean Squared Error (MSE), MAPE, RMSE, Symmetric Mean Absolute Percentage Error (SMAPE), R², and Mean Absolute Scaled Error (MASE). This multi-faceted approach allowed for a thorough assessment of each model’s predictive capabilities across different aspects of performance.

Data processing and visualization

All data processing, statistical analyses, and visualizations were performed using R version 4.4.1. We conducted age-standardized analyses using the GBD 2021 world standard population. Uncertainty intervals (UIs) were calculated using the 2.5th and 97.5th percentiles of 1000 draws from the posterior distribution of each estimate.

Global Temporal trends of NMSC burden (1990–2021)

The global burden of non-melanoma skin cancers (NMSCs), specifically basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), has substantially increased among adults aged 60 years and above from 1990 to 2021 (Figs.1, 2, and 3). The incidence of BCC rose by 291.85%, from 848,373.88 cases in 1990 to 3,324,372.70 cases in 2021, with the age-standardized incidence rate (ASR) increasing by 62.79% (Table1). Similarly, SCC incidence increased by 322.77%, from 392,655.76 to 1,660,033.08 cases, with a 67.42% increase in ASR. The estimated annual percentage change (EAPC) for BCC and SCC incidence was 2.02 (95% CI: 1.63 to 2.42) and 2.05 (95% CI: 1.6 to 2.5), respectively. Prevalence data showed similar trends, with BCC prevalence increasing by 261.19% and SCC by 364.61% (Table S1). The burden of disability-adjusted life years (DALYs) also increased significantly, with BCC DALYs rising by 257.68% and SCC DALYs by 159.21% (Table S2).

Sociodemographic disparities in NMSC burden

The burden of NMSCs exhibits significant disparities across different Socio-Demographic Index (SDI) regions (Figs.1, 2, and 3). In high SDI regions, BCC incidence increased by 295.11%, with ASRs rising by 102.73% (EAPC: 3.08, 95% CI: 2.58 to 3.57), while SCC incidence grew by 315.19%, with ASRs increasing by 107.31% (EAPC: 2.88, 95% CI: 2.4 to 3.37) (Table1). Conversely, low SDI regions showed more modest increases, with BCC incidence rising by 116.6% and SCC by 147.81%, although ASRs for BCC decreased by 2.86% (EAPC: – 0.11, 95% CI: – 0.13 to – 0.09). Prevalence and DALY data mirrored these trends, with high SDI regions experiencing substantially larger increases compared to low SDI regions (Tables S1 and S2).

NMSCs exhibit significant gender disparities globally (Fig.3). In 2021, males consistently showed higher burdens of both BCC and SCC across all SDI regions. Globally, males accounted for 56.47% of SCC cases and 43.53% of BCC cases, while females represented 52.40% and 47.60%, respectively. The highest incidence was observed in high SDI regions, with males showing ASRs of 1236.15 per 100,000 for BCC and 750.50 for SCC, compared to females at 660.73 and 358.72, respectively (Fig.3D). Age-standardized prevalence rates (ASPRs) followed similar patterns, with high SDI regions ranking first for both genders and cancer types (Fig.3E). Notably, age-standardized DALY rates for SCC were highest in high SDI regions for males (129.94 per 100,000) but in middle SDI regions for females (69.20), while BCC-related DALY rates were consistently highest in high SDI regions for both genders (Fig.3F).

Regional variations in NMSC burden

The burden of NMSCs among older adults exhibits significant regional variations and temporal shifts across the 21 GBD regions from 1990 to 2021 (Figs.4 and 5). The age-standardized incidence rates, prevalence rates, and DALYs for males, females, and both sexes combined from 1990 to 2021 are presented in Fig.4A-I, providing a comprehensive overview of gender-specific trends in NMSC burden. High-income North America consistently ranked first in ASRs for both BCC and SCC, with BCC rates increasing from 1118.15 to 2619.95 per 100,000, and SCC from 702.07 to 1611.25 (Fig.5A). Notably, while Australasia ranked second for SCC incidence, it showed a decrease from 428.17 to 384.27. In contrast, East Asia experienced a substantial rise in BCC incidence, moving from 16th to 4th place. ASPRs revealed different patterns, with High-income North America leading in SCC (1807.24 in 2021) but ranking lower for BCC (260.77) (Fig.5B). Interestingly, East Asia climbed from 10th to 3rd in SCC prevalence. Age-standardized DALY rates showed Australasia maintaining the highest burden for SCC (186.28 in 2021), while High-income North America led in BCC DALYs (1.03) (Fig.5C).

East Asia experienced the most dramatic increases in NMSC burden, with BCC incidence rising by 2576.08% and ASRs increasing by 784.32% (EAPC: 4.86, 95% CI: 4.05 to 5.68), while SCC incidence grew by 2059.35% with ASRs increasing by 581.62% (EAPC: 3.82, 95% CI: 3.01 to 4.63) (Fig.5E; Table1). Conversely, some regions showed decreases in ASRs, such as Andean Latin America, where BCC ASRs decreased by 28.69% (EAPC: – 1.30, 95% CI: – 1.57 to – 1.03) and SCC ASRs by 55.59% (EAPC: -3.08, 95% CI: – 3.49 to – 2.67). Prevalence trends generally mirrored incidence patterns (Fig.5F and Table S1). DALYs also varied widely, with East Asia showing a 2059.93% increase in BCC DALYs, while Central Europe experienced a 19.78% decrease in SCC DALYs (Fig.5G and Table S2).

High-income regions, particularly North America and East Asia, show marked increases in incidence, prevalence, and DALYs for both BCC and SCC. For instance, in North America, the 70–74 age group experienced a 363.37% increase in BCC incidence, with a 136.73% rise in ASR (EAPC: 3.80, 95% CI: 3.13–4.48). Similarly, SCC incidence increased by 371.16% with a 140.72% rise in ASR (EAPC: 3.50, 95% CI: 2.90–4.11). Prevalence and DALYs in these regions show comparable growth patterns, reflecting the significant impact of aging populations and environmental factors on NMSC burden (Fig.6).

In contrast, middle- and low-income regions demonstrate more complex patterns of NMSC burden. Andean Latin America, for example, shows a general decline in BCC and SCC incidence and prevalence rates. In the 70–74 age group, BCC ASR decreased by 28.20% (EAPC: – 1.25, 95% CI: – 1.54 to – 0.97), while SCC ASR decreased by 57.07% (EAPC: – 3.14, 95% CI: – 3.55 to – 2.73). However, middle-income regions such as East Asia exhibit significant increases in NMSC burden, particularly among older adults. For instance, BCC and SCC case numbers in East Asia increased by 700-900% for those aged 60 and above, with EAPCs reaching 4-5%. In low-income regions, such as sub-Saharan Africa, changes in NMSC burden are relatively moderate but show notable regional differences. Western sub-Saharan Africa demonstrates an upward trend in SCC burden, especially among those aged 75 and above, with an EAPC of approximately 0.6%. In contrast, Central and Eastern sub-Saharan Africa show an overall declining trend in NMSC burden, with EAPCs ranging from − 0.1% to – 0.7% (Fig.6).

Age-specific patterns of NMSC burden

The burden of NMSCs among older adults demonstrates distinct age-specific patterns globally from 1990 to 2021 (Figs.4 and 5). For BCC, the highest incidence in 2021 was observed in the 70–74 age group with 719,050 cases (349.33 per 100,000), representing a 324.64% increase in cases and a 74.65% increase in ASRs since 1990 (EAPC: 2.18, 95% CI: 1.92 to 2.43) (Table S3). SCC incidence peaked in the same age group with 356,664 cases (173.27 per 100,000), showing a 361.62% increase in cases and an 89.86% increase in ASRs (EAPC: 2.46, 95% CI: 2.12 to 2.81). Prevalence patterns differed, with SCC prevalence in 2021 highest in the 70–74 age group (425,949 cases, 206.93 per 100,000), representing a 405.44% increase in cases and a 107.89% increase in ASRs (EAPC: 2.69, 95% CI: 2.34 to 3.03) (Table S4). BCC prevalence was highest in the 70–74 age group as well, with 76,085 cases (36.96 per 100,000), showing a 292.80% increase in cases and a 61.56% increase in ASRs (EAPC: 1.82, 95% CI: 1.61 to 2.03). DALYs for SCC in 2021 were highest in the 70–74 age group (153,534 DALYs, 74.59 per 100,000), with a 168.32% increase in cases and a 10.36% increase in ASRs (EAPC: 0.44, 95% CI: 0.37 to 0.51) (Table S5).

For BCC, the percentage change in cases was consistently high across all age groups for both sexes, with the highest increase observed in males aged 95 + years (614.37%) and females aged 95 + years (388.99%) (Table S3). SCC showed similar trends, with the largest increases in males aged 90–94 years (514.31%) and females aged 60–64 years (367.68%). Interestingly, while the ASRs generally increased, some older age groups, particularly among females, experienced decreases. For instance, females aged 95 + years saw a 23.89% decrease in ASRs for SCC (Table S3). The EAPC was positive for most age groups, with higher values typically observed in younger age brackets. Notably, the EAPC for BCC in males aged 65–69 years was 2.9 (95% CI: 2.45 to 3.36), while for SCC in females aged 60–64 years, it reached 3.66 (95% CI: 2.93 to 4.4) (Table S3).

The burden of NMSCs among older adults exhibits distinct age-specific patterns across different SDI regions from 1990 to 2021, as evidenced by trends in incidence, prevalence, and DALYs (Fig.6). In high SDI regions, the incidence of both BCC and SCC peaked in the 70–74 age group in 2021. BCC incidence reached 560,059 cases (1049.58 per 100,000), representing a 324.64% increase in cases and a 99.34% increase in ASRs since 1990 (EAPC: 3.14, 95% CI: 2.72 to 3.55). Similarly, SCC incidence in this age group was 327,175 cases (613.14 per 100,000), showing a 352.74% increase in cases and a 112.53% increase in ASRs (EAPC: 3.14, 95% CI: 2.72 to 3.55) (Table S3). Prevalence patterns in high SDI regions differed slightly, with BCC prevalence peaking in the 75–79 age group (47,325 cases, 130.69 per 100,000) and SCC prevalence highest in the 70–74 age group (369,377 cases, 692.23 per 100,000) (Table S4). The DALY burden in high SDI regions was most pronounced for BCC in the 70–74 age group (230 DALYs, 0.43 per 100,000) and for SCC in the 75–79 age group (38,103 DALYs, 105.22 per 100,000) (Table S5).

In contrast, low SDI regions demonstrated lower burdens and more modest increases across all metrics. BCC incidence was highest in the 85–89 age group (257 cases, 19.28 per 100,000), while SCC incidence peaked in the 70–74 age group (67 cases, 0.65 per 100,000). Prevalence followed similar patterns, with BCC prevalence highest in the 85–89 age group (34 cases, 2.53 per 100,000) and SCC prevalence peaking in the 70–74 age group (92 cases, 0.88 per 100,000). The DALY burden in low SDI regions was minimal for BCC across all age groups, while for SCC, it was highest in the 80–84 age group (2,191 DALYs, 64.97 per 100,000). Middle SDI regions showed intermediate patterns across all measures. BCC incidence and DALY burden were highest in the 70–74 age group, while prevalence peaked in the 65–69 age group. For SCC, incidence was highest in the 65–69 age group, while prevalence and DALY burden peaked in the 70–74 age group. Notably, middle SDI regions often showed the most dramatic percentage increases in burden, particularly for SCC prevalence, which increased by 1155.85% in the 70–74 age group.

2024 Country-specific variations in NMSC burden

The global burden of NMSCs exhibits substantial heterogeneity across 204 countries and territories. For BCC, China experienced the most dramatic increase, with incident cases rising from 15,750.04 (95% UI: 11,382.75-21,156.08) in 1990 to 427,715.67 (95% UI: 331,313.72–542,686.99) in 2021, representing a 2,615.65% increase. The ASR in China surged from 17.92 to 159.77 per 100,000, a 791.66% increase, with an EAPC of 4.92 (95% CI: 4.1 to 5.73). For SCC, the United States reported the highest number of incident cases in 2021 at 1,444,703.24 (95% UI: 1,216,604.23-1,694,102.65), with an ASR of 1,805.33 per 100,000, representing a 135.4% increase since 1990 and an EAPC of 3.42 (95% CI: 2.83 to 4.01). Portugal demonstrated the highest EAPC for SCC incidence at 3.92 (95% CI: 3.58 to 4.26), with cases increasing by 399.06% and ASR by 152.46%. Notably, some countries showed decreasing trends, such as Brazil for BCC (ASR decline of 49.99%) and Japan for SCC DALYs (EAPC: -0.01; 95% CI: – 0.17 to 0.15) (Fig.7).

Decomposition analysis of NMSC burden drivers

Decomposition analysis revealed that population growth was the primary driver of increased NMSC burden globally from 1990 to 2021 (Fig.8). For SCC, population growth accounted for 83.92% of the increase in DALYs, 52.17% of prevalence increase, and 55.45% of incidence increase. Epidemiological changes also played a significant role, contributing to 42.16% of SCC prevalence increase globally. The drivers varied considerably across SDI regions. In high SDI regions, epidemiological changes were the primary contributor to BCC and SCC burden increases, accounting for 47.59–52.2% of changes across different measures. Conversely, in low SDI regions, population growth was the dominant factor, often contributing to over 100% of increases, while epidemiological changes sometimes had a mitigating effect (Table S9).

Frontier analysis of NMSC burden

Frontier analysis revealed distinct patterns of BCC and SCC burden across countries with varying SDI levels (Fig.9). High-SDI countries generally demonstrated a substantially higher burden for both BCC and SCC compared to low-SDI nations, with significant deviations from calculated frontier values. For BCC, the United States showed the highest ASR (2923.22 per 100,000) and prevalence rate (289.93 per 100,000). For SCC, New Zealand reported the highest incidence (390.33 per 100,000) and prevalence (473.49 per 100,000) rates. Interestingly, while incidence and prevalence of both cancers were primarily driven by SDI, DALY patterns differed, with both high-SDI and low-SDI countries reporting significant SCC DALY burdens, suggesting that quality of life impacts may be influenced by additional factors beyond SDI.

Comparative analysis of NMSC burden projections using multi-model approaches

While previous studies often relied on single predictive methods, our research employed a diverse array of eight forecasting models: Prophet, ARIMA, TBATS, ElasticNet, ETS, VAR, HoltWinters, and Theta to predict the burden of both BCC and SCC among adults aged 60 and above from 2022 to 2050(Fig.10, Table S10 and Table S11). Based on comprehensive model evaluations across multiple performance metrics, including MSE, MAPE, RMSE, and R2 values, we identified the most accurate predictors for each aspect of disease burden. For BCC, the ARIMA model consistently demonstrated superior performance across all three burdens metrics: incidence, prevalence, and DALYs. In contrast, for SCC, while ARIMA was optimal for incidence and DALYs predictions, the VAR model emerged as the best choice for prevalence forecasting, highlighting the disease-specific nature of model performance.

Utilizing these optimal models, our projections reveal distinct trends for BCC and SCC burden among older adults when analyzed across three key time intervals (2022–2030, 2031–2040, 2041–2050). For BCC, the ARIMA model predicts a gradual change in disease burden metrics over these periods. During 2022–2030, BCC incidence rates are projected to increase slightly from 314.00 to 322.18 per 100,000, followed by a steady decline to 312.35 per 100,000 by 2040, and further decreasing to 308.73 per 100,000 by 2050. BCC prevalence shows a similar pattern, initially rising from 33.61 to 34.93 per 100,000 (2022–2030), then declining slightly to 33.86 per 100,000 (2031–2040), and stabilizing at 33.47 per 100,000 by 2050. In contrast, BCC-related DALYs are projected to consistently increase across all three periods, from 0.14 per 100,000 in 2022 to 0.15 per 100,000 by 2050, indicating a small but persistent growth in disability burden despite declining incidence.

SCC projections demonstrate more pronounced shifts across the time intervals. SCC incidence rates show a consistent downward trend throughout all periods, decreasing from 157.67 per 100,000 in 2022 to 152.09 per 100,000 by 2030, further declining to 149.21 per 100,000 by 2040, and reaching 148.16 per 100,000 by 2050. However, SCC prevalence trends vary significantly by time period, with a notable decline from 184.78 to 170.06 per 100,000 during 2022–2030, followed by continued decrease to 162.48 per 100,000 by 2040, and a further slight reduction to 159.68 per 100,000 by 2050. SCC-related DALYs follow a similar declining pattern, decreasing from 79.32 per 100,000 in 2022 to 76.96 per 100,000 by 2030, then to 75.67 per 100,000 by 2040, and finally to 75.20 per 100,000 by 2050.

These interval-specific analyses reveal important temporal variations that might be obscured when examining only endpoint projections. The non-linear trends, particularly the initial increase followed by decline in BCC incidence, suggest complex underlying epidemiological dynamics that may reflect changing risk factors, improved prevention efforts, or evolving diagnostic practices. The divergent patterns between BCC and SCC, especially in prevalence and DALYs, underscore the importance of disease-specific modeling approaches and highlight the need for tailored healthcare strategies for each cancer type.

This study provides a comprehensive analysis of the global burden of NMSCs among adults aged 60 and above from 1990 to 2021, revealing significant increases in incidence, prevalence, and DALYs for both BCC and SCC. These findings underscore the growing importance of NMSCs as a major public health concern, particularly in the context of an aging global population.

One of the most striking findings is the substantial disparity in NMSC burden across different SDI regions. High SDI areas experienced the most dramatic increases, with BCC incidence rising by 295.11% and SCC by 315.19%, compared to more modest increases in low SDI regions. This pattern aligns with previous studies that have highlighted the role of socioeconomic factors in skin cancer epidemiology^16,17,18. The stark contrast between high and low SDI regions may reflect a complex interplay of factors, including differences in healthcare access, diagnostic capabilities, and lifestyle factors associated with UV exposure^19,20. Moreover, the higher burden in developed countries could be partly attributed to increased longevity and potentially changing behaviors regarding sun exposure²¹.

While our analysis provides valuable insights into between-region disparities, a significant limitation is the lack of within-region socioeconomic stratification. This is particularly relevant in densely populated, low-income and middle-income countries, where substantial healthcare access inequalities exist between socioeconomic groups¹⁰. In these settings, wealthier individuals often have significantly better access to dermatological care and diagnostic services, potentially leading to higher reported NMSC rates among these subpopulations. This could partially explain the apparent paradox of higher NMSC burden in high-SDI countries, as our current data cannot distinguish whether this reflects true disease distribution or simply better detection capabilities among the privileged. Ideally, future studies should aim to stratify NMSC burden by socioeconomic layers within regions to provide more nuanced comparisons—for example, comparing low-income populations in East Asia with those in North America, and similarly for middle and high-income groups. Such stratification would likely reveal even more pronounced disparities and could identify vulnerable subpopulations currently hidden in aggregated regional data. These considerations are crucial for accurately interpreting our findings on regional disparities and for developing appropriately targeted interventions that address both inter-regional and intra-regional inequities in NMSC.

The gender disparities observed across all SDI regions, with males consistently showing higher burdens of both BCC and SCC, warrant further investigation. This pattern has been noted in previous studies and may be attributed to differences in occupational and recreational UV exposure, as well as potential hormonal factors^7,22,23. The consistently higher male burden emphasizes the need for targeted prevention strategies for this demographic group. Age-specific patterns reveal that the 70–74 age group bears the highest burden of NMSCs, which aligns with the theory of cumulative UV exposure and declining immune function with age²⁴. However, the significant increases observed in the 95 + age group, particularly for BCC, highlight the impact of population aging and potentially improved healthcare access for this demographic. This finding underscores the need for tailored skin cancer management strategies for the oldest old, a rapidly growing population segment²⁵.

The regional analysis reveals complex geographical patterns in NMSC burden. The dramatic increase in East Asia, particularly in China, is noteworthy and may be associated with rapid economic development, lifestyle changes, and potential environmental factors^26,27. Conversely, the decreasing trends observed in regions like Andean Latin America suggest the potential success of local prevention strategies or other unexplored factors. These diverse regional patterns emphasize the need for context-specific approaches to NMSC prevention and management. The decomposition analysis reveals that population growth is the primary driver of increased NMSC burden, followed by epidemiological changes. This finding highlights both the challenge posed by global population aging to healthcare systems and the potential opportunities for prevention^4,28. The significant contribution of epidemiological changes suggests that factors beyond population dynamics, such as environmental or behavioral changes, play a crucial role in the increasing NMSC burden. The frontier analysis further confirms the substantial gap between high and low SDI countries. The high incidence and prevalence rates in countries like the United States and New Zealand, relative to their calculated frontier values, suggest potential areas for improvement in NMSC prevention and management in these nations^29,30. This analysis provides valuable insights for policymakers and healthcare professionals in identifying and addressing inefficiencies in skin cancer control efforts.

The multi-model forecasting approach employed in this study offers projections for future NMSC burden, though we recognize the inherent limitations of long-term forecasting. To address these limitations, we analyzed our projections across three distinct time intervals (2022–2030, 2031–2040, 2041–2050), which revealed important temporal variations and non-linear trends. For BCC, we observed an initial incidence increase followed by a decline, suggesting complex underlying epidemiological dynamics that may reflect changing risk factors, prevention efforts, or diagnostic practices. Additionally, we employed Bland-Altman plots to assess model agreement and potential systematic biases across different projection horizons, finding that prediction uncertainty increased proportionally with longer forecast periods^31,32,33. The time-interval analysis revealed divergent trends between BCC and SCC that would be obscured by examining only endpoint projections. While BCC incidence shows a non-linear pattern (increasing then decreasing), SCC demonstrates a consistent downward trend throughout all periods. The interval-specific approach also highlighted important variations in the rate of change – with more rapid decreases in SCC prevalence during 2022–2030 (from 184.78 to 170.06 per 100,000) compared to 2041–2050 (from 162.48 to 159.68 per 100,000)³⁴. These nuanced temporal patterns emphasize the importance of disease-specific modeling and intervention strategies, particularly given the contrasting DALY trends between cancer types. Our findings suggest that current prevention and management strategies may have differential long-term impacts on each cancer type, which warrants further investigation into the underlying mechanisms^35,36.

The study’s strengths lie in its comprehensive global coverage, long-term trend analysis, and sophisticated modeling approaches. However, some limitations should be acknowledged. The reliance on modeled estimates may introduce uncertainties, particularly in regions with limited primary data. Additionally, the study does not account for potential changes in diagnostic criteria or practices over time, which could influence trend interpretations. Future research directions should include investigations into the specific mechanisms underlying the observed regional and sociodemographic patterns. Studies evaluating the effectiveness of different prevention strategies across various contexts would be valuable. Given the projected increase in the global elderly population, enhancing surveillance, prevention, and treatment of NMSCs in this demographic will become increasingly crucial.

In conclusion, this study provides a comprehensive overview of the global burden of NMSCs among older adults, revealing significant increases and disparities across regions and sociodemographic groups. These findings underscore the need for targeted, context-specific strategies to address the growing challenge of skin cancers in an aging world population.