Materials and Methods

Data Collection

The GBD Study 2021 [17], which was produced by the Institute for Health Metrics and Evaluation, served as the primary data source for this study. This data set offers a comprehensive and up‐to‐date epidemiological insight into the burden associated with 371 diseases and injuries across 204 countries and territories. Data on brain and CNS tumors from 1990 to 2021 period were obtained through the Global Health Data Exchange (GHDx) query tool (https://vizhub.healthdata.org/gbd-results/) [17], including annual statistics on incidence, mortality, and DALYs. The corresponding population data (1992–2021) was used for APC analysis. In addition, our study referred to publicly available data, and no additional ethical consent was required.

The study population included children and adolescents aged 0–19 years, which differs from the traditional WHO definition of children as individuals younger than 18 years. Data associated with brain and CNS tumors were also extracted from the GBD 2021 database, all estimates and the 95% uncertainty intervals (UIs) for the incidence, mortality, and DALYs are generated from the 2.5th and 97.5th percentile values of 500 draws [17]. Countries with fewer or no data sources generally have a wider range of 95% UIs, suggesting greater inaccuracy in disease estimates. This analysis used the Socio‐demographic Index (SDI) for each country [18], an indicator estimated as a composite of income per capita, average years of schooling, and fertility rate in females under 25 years old. The SDI is scaled from 0 to 1, with higher values indicating higher socioeconomic levels. All countries were categorized into one of five SDI quintiles based on 2021 SDI values (Table S10).

All maps and regional visualizations presented in this study were constructed solely based on the territorial classifications and data structure provided by the GBD 2021 database. The separate listing or graphical representation of certain regions in accordance with the GBD 2021 database scheme does not imply any political position or endorsement of any territorial claims. These representations reflect only the statistical data organization as maintained by the Institute for Health Metrics and Evaluation.

For all age‐standardization, the standard population data was extracted from the World Health Organization's new World Standard Population, which was constructed from an average world population age‐structure for the period 2000–2025 [19]. Given that our analysis was restricted to individuals aged 0–19 years, only the corresponding age‐specific weights for this population were applied. These weights were proportionally rescaled to sum 1 within the 0–19 age range to ensure appropriate standardization for a pediatric‐only population [20]. Age‐standardized rates were calculated by direct method, which assumes that the rates are distributed as the weighted sum of independent Poisson random variables. According to the International Classification of Diseases, 10th revision (ICD‐10) codes, brain and CNS tumors were coded C70–C72.9 and C75.1–C75.3.

Age‐Period‐Cohort (APC) Model Analysis

To analyze temporal trends, the APC model framework was applied, which is recognized as an advanced research methodology that extends beyond traditional analyses in health and socio‐economic development studies. This model decomposes trends into age effects (natural history of the disease across life stages), period effects (calendar time changes), and cohort effects (shared experiences among individuals born in the same time period) [21, 22, 23]. The APC model estimates the net drift (overall annual percentage change) and local drift (age‐specific annual percentage change), as well as period and cohort relative risks (RRs) with corresponding 95% confidence intervals (CIs).

Given the inherent identifiability issue (exact linear dependency among age, period, and cohort), we adopted a conventional constraint‐based parameterization implemented in the APC Web Tool. In this approach, estimable functions (including net drift, local drift, and curvature components) are derived, while non‐identifiable linear components are incorporated into the drift term. The reference period and cohort were set to the median categories and do not affect the interpretation of estimable functions.

In APC analysis, age and period intervals are typically specified with equal widths, and thus 5‐year age groups were paired with 5‐year periods. In addition, this grouping is consistent with commonly used pediatric age classifications, reflecting key stages of physiological development and differences in clinical presentation and management across childhood and adolescence. Accordingly, we included data on brain and CNS tumors and the corresponding population from 1992 to 2021 at both the global level and across five SDI quintiles. The population aged 0–19 years was categorized into four groups (0–4, 5–9, 10–14, and 15–19 years). The study period was divided into six consecutive 5‐year intervals (1992–1996, 1997–2001, 2002–2006, 2007–2011, 2012–2016, and 2017–2021), yielding nine overlapping 10‐year birth cohorts from 1972 to 1981 to 2012–2021. The model estimated both overall temporal trends and age‐specific trends in incidence and mortality. Net drift represents the overall annual percentage change (% per year), whereas local drift reflects age‐specific annual percentage changes. The statistical significance of temporal trends was assessed using the Wald chi‐squared test.

Projection

For projection, we employed a BAPC model using integrated nested Laplace approximations. The BAPC framework assumes that age, period, and cohort effects follow second‐order random walk priors, allowing smooth temporal evolution while borrowing strength across adjacent intervals, and has been widely used for disease burden forecasting [24].

Statistical Analysis

Age‐standardized rates were expressed as the estimate per 100,000 populations and its 95% UI. All statistical analyses and data visualization were conducted using R software (version 4.4.0). The APC model analysis was performed using the Epi package, while data processing and visualization were carried out using the dplyr and ggplot2 packages.