A GBD 2021-Based Analytical Workflow to Estimate the Global Burden of Neoplasms Attributable to Low Physical Activity

Summary

This protocol demonstrates a reproducible workflow for extracting, processing, analyzing, and visualizing GBD 2021 estimates of neoplasms attributable to low physical activity. The workflow enables researchers to evaluate burden indicators, temporal trends, decomposition components, socioeconomic inequalities, and model-based projections using standardized population-level data.

Abstract

This protocol describes a reproducible analytical workflow for estimating the global burden of neoplasms attributable to low physical activity using Global Burden of Disease 2021 data. The workflow includes data extraction from the GBD Results Tool, definition of eligible outcomes and exposure categories, estimation of deaths, disability-adjusted life years, and age-standardized rates, temporal trend analysis using estimated annual percentage change and joinpoint regression, decomposition analysis, socioeconomic inequality assessment, and Bayesian age-period-cohort projection. The protocol also describes how to interpret uncertainty intervals, decomposition components, and inequality metrics in the context of modeled population-level estimates. As a representative application, the workflow was applied to colorectal cancer and female breast cancer attributable to low physical activity across 204 countries and territories from 1990 to 2021, with projections to 2045. This protocol can be adapted for other GBD risk-outcome pairs when researchers need a structured approach for risk-attributable burden estimation, trend characterization, inequality assessment, and projection analysis.

Introduction

Neoplasms represent a significant category of noncommunicable diseases (NCDs) contributing to global mortality and disability, posing a critical public health challenge worldwide. Globocan 2022 data revealed 19.96 million new neoplasm cases and 9.74 million deaths worldwide, with lung, breast (female), and colorectal cancers constituting the predominant malignancies. Projected cancer incidence will reach 35 million annually by 2050 (77% increase from 2022), with nations scoring low on the Human Development Index (HDI) projected to experience a 142% surge, driven by demographic transitions. Demographic expansion, population aging, deficient screening systems, and constrained treatment access in low-HDI regions drive these trends, creating an imperative for equitable allocation of global oncology resources¹.

Low physical activity (LPA) is an established modifiable determinant for major NCDs, spanning malignancies, metabolic disorders, and neuropsychiatric conditions^2,3,4. LPA constitutes a significant global health burden, demonstrating a 6.6% age-standardized prevalence rise since 2000 and contributing to 7.2% of global all-cause mortality in 2022^5,6. Moderate physical activity (PA) confers significant protection against breast (HR 0.90; 95% CI: 0.87–0.93), colon (HR 0.84; 95% CI: 0.77–0.91), and rectal cancers (HR 0.87; 95% CI: 0.80–0.95), with parallel reductions in site-specific mortality⁷. To promote essential PA worldwide, numerous countries and regions have developed national PA guidelines tailored to their specific contexts (e.g., Canada, the UK, and the US), which have been widely disseminated through organizations such as the World Health Organization (WHO) and the European Union. Regrettably, only 72.5% of adults globally meet the minimum recommended PA levels outlined in public health guidelines⁸.

Mitigating low physical activity-attributable neoplasm burden is pivotal to achieving the UN Sustainable Development Goal (SDG) 2030 target: one-third reduction in premature NCDs mortality via prevention and care⁹. Generating robust evidence to elucidate the relationship between LPA and neoplasms is essential for guiding future healthcare policies. The Global Burden of Diseases, Injuries, and Risk Factors Study 2021 (GBD 2021) provides comparable population-level estimates of mortality, disability-adjusted life years (DALYs), and risk-attributable burden across 204 countries and territories from 1990 to 2021 using standardized epidemiological approaches. In GBD 2021, disease and injury estimates cover 371 diseases and injuries, while the risk factor analysis estimates exposure, relative risk, population attributable fractions, and attributable burden for 88 risk factors¹⁰. However, fewer studies have provided a stepwise and reproducible workflow for extracting GBD 2021 low physical activity-attributable neoplasm data and applying trend, decomposition, inequality, and projection analyses within a single protocol. This manuscript addresses this methodological gap by detailing the data extraction, variable definition, statistical modeling, and interpretation procedures for this specific GBD risk-outcome pairing. This protocol describes a reproducible secondary analysis of GBD 2021 estimates for neoplasms attributable to low physical activity across 204 countries and territories from 1990 to 2021, with projections to 2045. The objectives were to quantify deaths, DALYs, age-standardized rates, temporal trends, decomposition components, socioeconomic inequalities, and projected burden, without inferring individual-level causality or direct intervention effects.

This workflow was chosen because it integrates several complementary analytical components into a single reproducible framework. Compared with studies that report only descriptive GBD estimates or single trend metrics, this protocol combines burden estimation, EAPC, joinpoint regression, decomposition analysis, inequality assessment, and BAPC projection, allowing users to examine not only the magnitude of burden but also temporal changes, demographic drivers, socioeconomic gradients, and future trends. The protocol is applicable to researchers conducting secondary analyses of GBD risk-attributable estimates, particularly when the aim is to compare burden patterns across time, sex, age groups, countries, regions, or SDI categories.

Protocol

This study used publicly available, de-identified GBD 2021 population-level estimates. No individual-level human participant data were accessed. The GBD 2021 study received a waiver of informed consent from the University of Washington Institutional Review Board; therefore, no additional institutional ethical approval was required for this secondary analysis.

NOTE: This protocol describes a reproducible workflow for conducting a secondary analysis of GBD 2021 estimates for neoplasms¹¹ attributable to low physical activity. The workflow includes data extraction, variable definition, burden estimation, temporal trend analysis, decomposition analysis, inequality assessment, and Bayesian age-period-cohort projection. The results are presented to illustrate the application of the protocol rather than to establish a primary causal relationship. The analytical workflow consisted of six sequential steps: extraction of GBD 2021 estimates, definition of exposure, disease outcomes, and SDI strata, estimation of deaths, DALYs, age-standardized mortality rate (ASMR), and age-standardized DALYs rate (ASDR), temporal trend analysis using estimated annual percentage change (EAPC) and joinpoint regression, decomposition and inequality analyses, and Bayesian age-period-cohort projection through 2045. Each result and figure was generated from one of these linked analytical steps.

1. Data source identification and access

1. Access the Global Burden of Disease (GBD) 2021 Results Tool.
   1. Open a web browser and navigate to the Global Burden of Disease Results Tool at https://vizhub.healthdata.org/gbd-results/.
   2. Select GBD 2021 as the study version from the available dataset options.
   3. Select Risk Factors from the cause hierarchy menu.
2. Define disease outcomes and risk factors.
   1. Select Neoplasms as the Disease Category.
   2. Select Low Physical Activity as the Risk Factor.
   3. Select Deaths and Disability-Adjusted Life Years (DALYs) as Outcome Measures.
3. Specify population, geography, and time period.
   1. Select Both sexes, Male, and Female separately for sex-specific analyses.
   2. Select all available age groups ≥25 years.
   3. Select all 204 countries and territories.
   4. Set the analysis period from 1990 to 2021.
4. Extract age-standardized outcome measures.
   1. Enable age standardization using the GBD 2021 standard population.
   2. Select age-standardized mortality rate (ASMR) and age-standardized DALYs rate (ASDR) as output variables.
   3. Include 95% uncertainty intervals (UIs) for all extracted estimates.
   4. Export the extracted data in CSV format for downstream analysis.
   5. For reproducibility, use the following GBD Results Tool selections: study version: GBD 2021; context: Risk factors; risk: Low physical activity; cause: Neoplasms; measures: Deaths and DALYs; metrics: Number and Rate; ages: Age-standardized and all available 5-year age groups aged ≥ 25 years; sexes: Both, Male, and Female; locations: Global, 21 GBD regions, SDI regions, and 204 countries and territories; years: 1990–2021.
   6. Export the mean estimate, lower uncertainty interval, and upper uncertainty interval for each stratum in CSV format.
5. Obtain sociodemographic index (SDI)-stratified data.
   1. Enable stratification by Sociodemographic Index (SDI).
   2. Extract SDI-specific ASMR and ASDR data for comparative analyses.
6. Access underlying data sources.
   1. Navigate to the GBD 2021 Data Input Sources Tool at https://ghdx.healthdata.org/gbd-2021.
   2. Review publicly available documentation describing primary data sources, including vital registration systems, population surveys, medical records, and disease registries.

2. Definition of key variables and classification criteria

1. Define neoplasms.
   1. Define neoplasms as malignant tumors according to the Global Burden of Disease (GBD) 2021 cause hierarchy.
   2. Include only malignant neoplasms characterized by uncontrolled cellular proliferation resulting from genetic and epigenetic dysregulation.
   3. Exclude benign neoplasms from all analyses.
2. Define low physical activity.
   1. Define physical activity intensity using metabolic equivalent of task (MET) values, where 1 MET equals an oxygen consumption of 3.5 mL O₂/kg/min at rest.
   2. Calculate weekly physical activity levels using total MET-minutes per week.
   3. Classify physical activity levels into four categories^12,13:
      Less than 600 MET-minutes per week.
      600 to 3,999 MET-minutes per week.
      4,000 to 7,999 MET-minutes per week.
      8,000 or more MET-minutes per week.
   4. Define low physical activity as less than 4,000 MET-minutes per week, consistent with GBD 2021 risk attribution methodology.
      NOTE: This cutoff was adopted from the GBD 2021 risk attribution definition and was not selected de novo in the present study. It was used to maintain consistency with the GBD comparative risk assessment framework and comparability with other GBD-based estimates.
3. Define SDI.
   1. Define the SDI as a composite indicator ranging from 0 to 1.
   2. Calculate SDI using the geometric mean of standardized per capita income, educational attainment, and fertility rate in individuals younger than 25 years.
   3. Classify countries and regions into five SDI categories¹⁴: Low, Low-Middle, Middle, High-Middle, and High.
   4. Assign countries to one of 21 GBD regions based on SDI stratification.
4. Define neoplasms attributable to low physical activity.
   1. Identify neoplasms attributable to low physical activity using the GBD 2021 comparative risk assessment framework.
      1. Within this framework, operationalize low physical activity-attributable burden by combining exposure distributions, risk-outcome relative risks, and the theoretical minimum risk exposure level to estimate population attributable fractions. Then, apply this to deaths and DALYs for eligible neoplasm outcomes.
   2. Include colorectal cancer and female breast cancer as neoplasm outcomes attributable to low physical activity exposure.
      NOTE: Thus, in this study, neoplasms attributable to low physical activity refer only to colorectal cancer and female breast cancer as defined in the GBD 2021 comparative risk assessment framework, rather than all malignant tumors.

3. Estimation of global burden of neoplasms attributable to low physical activity.

1. Extract disease burden indicators.
   1. Import deaths and disability-adjusted life years (DALYs) data for neoplasms attributable to low physical activity from the Global Burden of Disease (GBD) 2021 Results Tool, following the standard GBD comparative risk assessment framework.
   2. Extract age-standardized mortality rates (ASMR) and age-standardized DALYs rates (ASDR) per 100,000 population using the GBD 2021 standard population.
   3. Retain corresponding 95% uncertainty intervals for all extracted estimates as provided by the GBD 2021 methodology.
2. Calculate disability-adjusted life years.
   1. Define disability-adjusted life years (DALYs) as the sum of years of life lost (YLLs) and years lived with disability (YLDs), following standard GBD definitions.
   2. Calculate YLLs by multiplying age-specific mortality by the reference life expectancy specified in the GBD 2021 life table.
   3. Calculate YLDs by multiplying incident cases by disease-specific disability weights and disease duration, as described in the GBD methodology.
   4. Verify that calculated DALY values correspond to GBD 2021 reported estimates.

4. Temporal trend analysis

1. Estimate annual percentage change^15,16.
   1. Import ASMR and ASDR data into R software (version 4.4.2).
   2. Apply natural logarithmic transformation to age-standardized rates.
   3. Fit linear regression models with calendar year as the independent variable to assess temporal trends, following established epidemiological trend analysis methods.
   4. Calculate the estimated annual percentage change (EAPC) from the regression slope using the standard exponential transformation.
   5. Classify trends as increasing, decreasing, or stable based on whether the 95% confidence interval of the EAPC is above, below, or includes zero.
2. Perform joinpoint regression analysis.
   1. Open the Joinpoint Regression Program (version 5.3.0; National Cancer Institute).
   2. Import ASMR and ASDR datasets into the software.
   3. Select a log-linear regression model for trend analysis, consistent with Joinpoint methodological guidelines¹⁷.
   4. Specify a minimum of 0 and a maximum of 6 joinpoints.
   5. Select the Data-Driven Bayesian Information Criterion (BIC) method with weighted BIC optimization to determine optimal joinpoints.
      1. Select the maximum of six joinpoints according to the length of the annual time series from 1990 to 2021 and the default recommendations of the Joinpoint Regression Program.
      2. Determine the final number of joinpoints by the data-driven BIC.
   6. Run the analysis to calculate annual percentage change (APC) and average annual percentage change (AAPC) with 95% confidence intervals. For strata with sparse data or unstable annual estimates, segment-specific APCs were interpreted cautiously and were not emphasized unless the fitted trend was supported by corresponding uncertainty intervals and was epidemiologically plausible.
   7. Export APC and AAPC results for downstream analysis.
      NOTE: EAPC and joinpoint regression were used for different purposes. EAPC provided a single summary estimate of the average log-linear trend across 1990–2021, whereas joinpoint regression identified calendar periods in which the rate of change shifted and estimated segment-specific APCs. Therefore, EAPC was used for overall comparison across strata, while joinpoint regression was used to describe temporal inflection patterns.

5. Decomposition analysis

1. Prepare decomposition inputs.
   1. Stratify mortality and DALY data by age group, calendar year, and geographic region.
   2. Import stratified datasets into JD_GBDR software (version 2.37.1).
2. Perform decomposition analysis.
   1. Decompose changes in mortality and DALYs between selected time points into aging, population growth, and epidemiological change components using established decomposition methods¹⁸.
   2. Quantify the contribution of each component to overall changes in disease burden.
      1. Interpret percentage contributions as the proportional contribution of aging, population growth, and epidemiological change to the net change in deaths or DALYs between two time points.
      2. Values exceeding 100% or negative values can occur when one component increases burden while another offsets it; interpret such values as mathematical decomposition effects rather than direct biological effects.
   3. Export decomposition results for global, SDI-level, and GBD regional analyses.

6. Health inequality analysis

1. Prepare data for inequality assessment.
   1. Rank all countries and territories according to their SDI values from lowest to highest.
   2. Import ASMR and ASDR into R software (version 4.4.2).
2. Calculate the slope index of inequality.
   1. Assign a relative socioeconomic rank to each country based on cumulative population distribution across SDI strata.
   2. Fit a weighted linear regression model with ASMR or ASDR as the dependent variable and relative SDI rank as the independent variable.
   3. Calculate the slope index of inequality (SII) as the absolute difference in disease burden between the highest and lowest ends of the SDI gradient, following established inequality analysis methodology.
3. Calculate the concentration index.
   1. Construct concentration curves by plotting cumulative proportions of ASMR or ASDR against cumulative population proportions ranked by SDI.
   2. Calculate the concentration index (CII) as twice the area between the concentration curve and the line of equality.
   3. Interpret CII values close to zero as low inequality and values approaching −1 or +1 as increasing relative inequality, consistent with standard health economics methods¹⁹.
4. Estimate uncertainty for inequality metrics.
   1. For SII, derive 95% confidence intervals from the weighted regression model using country-level population weights.
   2. For CII, estimate 95% confidence intervals using bootstrap resampling of countries and territories with replacement.
   3. When posterior draws are available, repeat SII and CII calculations across draws and report the 2.5th and 97.5th percentiles as uncertainty intervals.

7. Bayesian age-period-cohort modeling

1. Prepare input data for projection.
   1. Organize age-specific deaths, DALYs, and corresponding population data by 5-year age group and calendar year from 1990 to 2021.
   2. Ensure consistency of age-group definitions and time intervals across the full observation period.
2. Apply the Bayesian age-period-cohort model²⁰.
   1. Import the age-specific burden and population datasets into R using a Bayesian age-period-cohort modeling framework.
   2. Specify age, period, and cohort effects as structured random effects, with second-order random-walk priors used to smooth adjacent age groups, calendar periods, and birth cohorts.
   3. Generate posterior estimates of age-specific rates and then calculate projected ASMR and ASDR using the GBD 2021 standard population.
   4. Assess model fit by comparing fitted estimates with observed estimates during the historical period.
   5. Conduct internal validation by withholding the most recent observed years, refitting the model, and comparing predicted estimates with observed values.
3. Generate future projections.
   1. Forecast age-standardized mortality rates and age-standardized DALYs rates for neoplasms attributable to low physical activity from 2022 to 2045.
   2. Extract posterior mean estimates and corresponding 95% uncertainty intervals for all projected outcomes. Evaluate projection uncertainty using posterior 95% uncertainty intervals, and interpret estimates farther from the observed period with increasing caution.
   3. Export projection results for visualization and reporting.

8. Statistical analysis

1. Perform general statistical analysis.
   1. Import all cleaned datasets into R software (version 4.4.2).
   2. Use R software to conduct all statistical analyses and generate figures and tables.
   3. Set statistical significance at a two-sided p-value of less than 0.05 for all analyses.
2. Conduct specialized GBD analyses.
   1. Import GBD-formatted datasets into JD_GBDR software (version 2.37.1).
   2. Use the JD_GBDR software to perform disease burden calculations, decomposition analyses, and standardized rate visualizations according to GBD analytical conventions.
3. Perform joinpoint regression analysis.
   1. Download and install the Joinpoint Regression Program (version 5.3.0) from the National Cancer Institute website (https://surveillance.cancer.gov/joinpoint/download).
   2. Import age-standardized mortality rate and age-standardized DALYs rate datasets into the Joinpoint software.
   3. Select a log-linear regression model and execute joinpoint regression analysis to evaluate temporal trends, following the Joinpoint Regression Program methodology.
   4. Export trend estimates, including annual percentage change and average annual percentage change, with corresponding 95% confidence intervals.
      NOTE: All analyses were conducted using R software version 4.4.2, Joinpoint Regression Program version 5.3.0, and JD_GBDR software version 2.37.1. The main analytical settings included log-linear EAPC models, joinpoint regression with 0–6 joinpoints selected by the data-driven Bayesian Information Criterion, decomposition into aging, population growth, and epidemiological change components, weighted regression for SII, bootstrap resampling or posterior draws for CII uncertainty, and Bayesian age-period-cohort projection to 2045.

Results

Successful implementation of this protocol should generate a complete set of harmonized GBD-derived estimates across predefined locations, years, sexes, age groups, measures, and metrics. The expected outputs include age-standardized and age-specific burden estimates with 95% uncertainty intervals, EAPC and joinpoint trend estimates, decomposition components, SII and CII inequality metrics, and BAPC-based projections with posterior uncertainty intervals. A suboptimal implementation is indicated by missing strata, inconsistent age or location labels, failure to reproduce GBD-reported estimates after data cleaning, unstable joinpoint models in sparse strata, or implausibly narrow projection intervals.

Global burden of neoplasms attributable to LPA
In 2021, neoplasms attributable to LPA accounted for 87,440 global deaths (95% UI: 48,636–125,811), constituting a 95.73% increase from the 1990 baseline (44,675; 95% UI: 24,963–64,997) (Table 1, Supplementary Table 1). After adjustment for differences in population age structure, the ASMR declined from 1.32 (95% UI: 0.75–1.92) to 1.06 (95% UI: 0.59–1.52) per 100,000 between 1990 and 2021. The simultaneous increase in absolute deaths and decrease in ASMR indicates that population growth and aging increased the total number of deaths, whereas the age-standardized mortality rate declined over time. This significant reduction was corroborated by an EAPC of -0.79 (95% CI: -0.83 to -0.75) and AAPC of -0.70 (95% CI: -0.73 to -0.68) (Table 1, Supplementary Table 2). Joinpoint regression modeling identified five temporal inflection points in ASMR trajectories (1994, 2004, 2007, 2014, and 2018) through segmented linear regression analysis (Figure 1, Supplementary Table 2).

Similarly, DALYs related to neoplasms attributable to LPA increased by approximately 84.79% from 944,286 (95% UI: 505,737–1,389,544) in 1990 to 1,744,944 (95% UI: 908,799–2,545,345) in 2021 (Table 2, Supplementary Table 1). Notwithstanding rising absolute burdens, ASDR declined from 25.16 (95% UI: 13.67–37.09) to 20.50 (95% UI: 10.73–29.89) per 100,000 during 1990–2021, constituting an 18.51% reduction (95% UI: -26.10 to -10.09). This secular attenuation paralleled significant temporal trends (EAPC: -0.76, 95% CI: -0.82 to -0.70; AAPC: -0.68, 95% CI: -0.72 to -0.64) (Table 2, Supplementary Table 2). Joinpoint regression confirmed trend inflection points in 1994, 2004, 2007, 2014, and 2018 (Figure 1, Supplementary Table 2).

In 2021, CRC constituted the principal contributor to neoplasms attributable to LPA mortality, exhibiting an ASMR of 0.87 (95% UI: 0.55–1.19) per 100,000 population, whereas female BC manifested a comparatively lower ASMR of 0.19 (95% UI: 0.04–0.34) per 100,000. From 1990 to 2021, all neoplasms subtypes exhibited declining mortality trends, characterized by EAPC of -0.82 (95% CI: -0.86 to -0.78) and AAPC of -0.72 (95% CI: -0.74 to -0.70) for CRC, and corresponding values of -0.66 (95% CI: -0.72 to -0.61) and -0.56 (95% CI: -0.58 to -0.55) for BC. These temporal patterns in ASDR paralleled the ASMR trajectories across all disease subtypes, with CRC showing the most pronounced reductions (ASDR decline: 20.34%, 95% UI: -27.99 to -10.82) (Table 1, Table 2, Supplementary Table 3, Supplementary Figure 1). To improve readability, the main text focuses on global, sex-specific, and SDI-stratified findings, while detailed country- and region-level estimates are reported in the supplementary materials.

Global burden of neoplasms attributable to LPA: Sex-age stratification
The global disease burden of neoplasms attributable to LPA exhibited a characteristic unimodal age distribution with initial escalation followed by gradual decline, disproportionately affecting females. In 2021, mortality rates peaked at 80-84 years for both sexes, while DALYs reached their zenith earlier at 70–74 years. ASMR and ASDR demonstrated progressive elevation with advancing age, reaching maximal levels in elderly populations. Notably, deceleration of age-related increments and absolute reductions was restricted to elderly males. Females maintained elevated rate ratios across all age strata, with epidemiological patterns remaining congruent with 1990 observations (Figure 2, Supplementary Table 4, Supplementary Table 5, Supplementary Figure 2).

Neoplasms attributable to LPA exhibited rising global mortality and DALY burdens during 1990-2021. Males showed greater proportional increases [mortality: 131.61% (95% UI:94.52–176.46); DALYs: 114.23% (95% UI:79.34–157.63)] versus females' marked absolute burden growth: mortality from 31,792 (95% UI: 16,647–46,903) to 57,602 (95% UI: 29,060–85,197) (+81.18%), DALYs from 676,178 (95% UI: 336,046–992,525) to 1,170,572 (95% UI: 555,314–1,756,518) (+73.12%). Females consistently exhibited disproportionately higher disease burden than males throughout the study period (Table 1, Table 2, Supplementary Table 1, and Supplementary Figure 3). This sex difference should be interpreted in light of the GBD outcome definition, because female breast cancer is included among low physical activity-attributable neoplasms, whereas male breast cancer is not. Therefore, the higher female burden partly reflects disease composition in addition to potential sex-specific differences in exposure, susceptibility, or health care access. However, ASRs for neoplasms attributable to LPA demonstrated sex-dimorphic declines: ASMR: Males: 0.92 (95% UI: 0.57–1.30) to 0.84 (95% UI: 0.52–1.19) /100,000 [EAPC -0.32 (95% CI: -0.36 to -0.28)]; Females: 1.61 (95% UI: 0.85–2.38) to 1.24 (95% UI: 0.62–1.83)/100,000 [EAPC: -0.96 (95% CI: -1.00 to -0.91))]. ASDR: Males: 16.57 (95% UI: 10.18–22.94) to 15.03 (95% UI: 9.21–20.93)/100,000 [EAPC -0.37 (95% CI: -0.43 to -0.32)]; Females: 32.38 (95% UI: 16.26–47.60) to 25.46 (95% UI: 12.04–38.21)/100,000 [EAPC -0.89 (95% CI: -0.95 to -0.83)]. AAPC confirmed this sexual dimorphism: ASMR: males -0.29 (95% CI: -0.31 to -0.26), females -0.85 (95% CI: -0.87 to -0.83); ASDR: males -0.31 (95% CI: -0.35 to -0.29), females -0.78 (95% CI: -0.79 to -0.76) (Table 1, Table 2, Figure 1, Figure 3, Supplementary Table 2).

Global burden of neoplasms attributable to LPA: Sociodemographic disaggregation analysis
Neoplasms attributable to LPA exhibited significant socioeconomic disparities: In 2021, ASMR demonstrated a progressive gradient from 0.41 (95% UI: 0.20–0.61) to 1.35 (95% UI: 0.75–1.95)/100,000 (3.3-fold increase; low-to-high SDI), paralleled by ASDR escalating from 8.83 (95% UI: 4.16–13.49) to 26.52 (95% UI: 14.13–38.73)/100,000 (3.0-fold). During 1990–2021, while global rates declined overall [ASMR EAPC: -0.79(95% CI: -0.83 to -0.75), ASDR EAPC: -0.76(95% CI: -0.82 to -0.70)], significant heterogeneity emerged across SDI quintiles: High-SDI regions showed substantial attenuation [ASMR EAPC -1.42 (95% CI:-1.48 to -1.36); ASDR EAPC -1.40 (95% CI:-1.47 to -1.33)] contrasting with escalating burdens in low-middle-SDI [ASMR EAPC 0.86 (95% CI:0.81–0.91); ASDR EAPC 0.77 (95% CI:0.73–0.81)] (Table 1, Table 2, Figure 3).

Between 1990 and 2021, SDI showed significant positive correlations with ASMR (ρ = 0.7560, p < 0.001) and ASDR (ρ = 0.7364, p < 0.001) across 21 GBD regions, with rates increasing until SDI≈0.8 then declining at higher levels (Figure 4). In 2021, these correlations persisted for ASMR (ρ = 0.6724, p < 0.001) and ASDR (ρ = 0.5938, p < 0.001) across 204 countries/territories, mirroring GBD regional patterns. Barbados, the United Kingdom, American Samoa, and Micronesia (Federated States of) demonstrated ASMR/ASDR values substantially exceeding SDI-predictions (Figure 5).

Geographical disparities in the burden of neoplasms attributable to LPA
In 2021, neoplasms attributable to LPA demonstrated distinct regional burdens: lowest in South Asia (ASMR 0.42/100,000; ASDR 9.05/100,000), Eastern Sub-Saharan Africa (0.42; 8.95), and Central Asia (0.48; 9.94), versus highest in Australasia (1.69; 34.88) and Central Europe (1.75; 32.89). East Asia bore the greatest global burden with 19,608 deaths (95% UI: 11,453–29,247) and 396,653 DALYs (95% UI: 224,218–604,269). Between 1990 and 2021, Southern Sub-Saharan Africa and Central Latin America showed the most substantial ASMR/ASDR increases for this malignancy (Table 1, Table 2). Regional and national comparisons were interpreted descriptively, because GBD estimates may differ in precision across settings depending on the availability and quality of underlying data.

Complementing this regional perspective, national-level analysis further quantified disease burden disparities: Highest ASMR occurred in Monaco [2.47 (95% UI: 1.24–3.88)], UK [2.69 (95% UI:1.50–3.82)], and Barbados [3.16 (95% UI: 1.54–4.92)], while highest ASDR manifested in Micronesia [61.31 (95% UI: 27.37–98.65)], American Samoa [62.28 (95% UI: 28.75–101.61)], and Barbados [66.62 (95% UI:31.56–104.95)]. Conversely, the lowest ASMR was characterized in Tanzania [0.16 (95% UI:0.06-–.30)], Malawi [0.19 (95% UI: 0.08–0.33)], and Uzbekistan [0.19 (95% UI: 0.08–0.32)], with correspondingly low ASDR in Tanzania [3.36 (95% UI:1.22–6.12)], Benin [3.97 (95% UI: 1.58–7.12)], and Malawi [4.25 (95% UI:1.60–7.64)]. Notably, China bore the highest global burden with 18,624 deaths (95% UI: 10,849–27,939) and 375,360 DALYs (95% UI: 211,389–574,711). During 1990 and 2021, Austria, San Marino, and Belgium achieved substantial reductions in both metrics, whereas Cabo Verde, Lesotho, and Honduras showed marked ASMR increases, with Lesotho, Mauritius, and El Salvador demonstrating significant ASDR elevations (Figure 6, Supplementary Table 6, Supplementary Table 7, Supplementary Figure 4). Countries and territories with extreme ASMR or ASDR values were highlighted to describe the distribution of burden, but these estimates should be interpreted cautiously, particularly in small populations or settings with wide uncertainty intervals.

Decomposition analysis
Decomposition analysis (aging/population/epidemiological-change) quantified mortality and DALYs variations across the global 5-SDI and 21-GBD regions (Figure 7, Supplementary Table 8). Because decomposition percentages are calculated relative to the net change in burden, very large positive or negative values can occur when opposing components offset each other. These values indicate the relative mathematical contribution of each component and should not be interpreted as biological effect sizes.

Globally, neoplasms attributable to LPA mortality and DALYs variations were primarily driven by population (mortality: 88.44%, DALYs: 131.53%), followed by aging (46.63%; -79.71%) and epidemiological change (-35.07%; 48.19%), demonstrating population's detrimental global impact.

At the SDI region level, population consistently demonstrated negative impacts on both mortality and DALYs. In contrast, aging exhibited negative effects on mortality across all SDI regions but showed positive influences on DALYs, while epidemiological changes manifested detrimental impacts in middle SDI regions and below. Specifically, high SDI regions exhibited the most pronounced impacts from all three demographic determinants: Aging (48.02% mortality change, -104.89% DALY variation), Population (175.03%, 300.33%), and Epidemiological-Change (-123.04%, -95.44%). Regarding mortality, aging demonstrated the least negative effect in low SDI regions (28.08%), Population showed minimal detrimental impact in low-middle SDI regions (50.27%), while epidemiological change displayed maximal beneficial influence in high SDI regions (-123.04%). For DALYs, Aging revealed significant positive effects in high SDI regions (-104.89%), whereas low-middle SDI regions experienced the weakest negative impact from population (64.39%) yet suffered the most substantial adverse consequences from epidemiological change (46.31%).

Across 21 GBD regions, aging (75.74%), Population (389.67%), and epidemiological change (-298.99%) exhibited their most substantial mortality contributions in Australasia and Western Europe, while demonstrating minimal mortality impacts in Central Europe, Central Latin America, and Oceania at -9.04%, 42%, and -3.06%, respectively. Surprisingly, Western Europe manifested extreme DALY variation contributions from aging (-1036.4%), Population (1987.98%), and Epidemiological-Change (-851.58%), whereas Southeast Asia (0.7%), Central Latin America (47.99%), and Oceania (-4.83%) showed the weakest corresponding DALYs influences from these determinants.

Health inequality analysis
Health inequality metrics revealed sustained disparities in neoplasms attributable to LPA burden across SDI regions: Between 1990 and 2021, SII for ASMR declined from 1.45 (95% CI: 1.24–1.66) to 1.11 (95% CI: 0.95–1.26), while CII decreased from 0.30 (95% CI: 0.26–0.34) to 0.20 (95% CI: 0.17–0.24). Corresponding ASDR trends showed SII reduction (28.52 [95% CI: 24.21–32.83] to 20.02 [95% CI: 16.62–23.41]) and CII decline (0.27 [95% CI: 0.23–0.31] to 0.17 [95% CI: 0.13–0.21]), collectively demonstrating substantial improvements in both absolute (SII) and relative (CII) inequality metrics (Figure 8, Supplementary Table 9).

Projected global burden of neoplasms attributable to LPA: Bayesian age-period-cohort modeling analysis (2022–2045)
BAPC modeling projects a gradual global decline in neoplasms attributable to LPA burdens by 2045: ASMR decreases from 1.34 (95% UI: 1.33–1.35) to 0.95 (95% UI: 0.88–1.03)/100,000 and ASDR from 25.36 (95% UI: 25.30–25.41) to 18.86 (95% UI: 16.99–20.72)/100,000. Sex-specific analysis reveals persistent disparities with females bearing higher burdens - male ASMR decreases by 0.14/100,000 [final 0.79 (95% UI: 0.71–0.88)] versus greater female reduction (0.54/100,000) yet higher final rate [1.09 (95% UI: 1.00–1.18)]. Similarly, projected ASDR reaches 14.22/100,000 males versus 23.28/100,000 females (Figure 9, Supplementary Table 10). Because projections beyond the observed period depend on historical trend stability and model assumptions, the 2045 estimates should be interpreted as model-based projections rather than definitive forecasts.

Figure 1: Joinpoint regression of ASMR and ASDR for neoplasms attributable to low physical activity from 1990 to 2021, stratified by sex. (A) ASMR and (B) ASDR. Dots indicate observed annual estimates, fitted lines indicate segmented log-linear trends, and vertical markers indicate detected joinpoints. Segment-specific APCs and AAPCs are provided in Supplementary Table 2. Please click here to view a larger version of this figure.

Figure 2: Age-specific deaths and DALYs for neoplasms attributable to low physical activity in 2021, stratified by sex. (A) Deaths and (B) DALYs bars show absolute numbers, lines show age-specific rates, and error bars or shaded bands indicate 95% uncertainty intervals. Please click here to view a larger version of this figure.

Figure 3: Temporal trends in ASMR and ASDR for neoplasms attributable to low physical activity globally and across SDI quintiles, 1990–2021. (A–F) ASMR and (G–L) ASDR. (A,G) Global, (B,H) High SDI, (C,I) High-Middle SDI, (D,J) Middle SDI, (E,K) Low-Middle SDI, (F,L) Low SDI. Detailed numerical estimates are provided in Table 1, Table 2, and Supplementary Table 2. Please click here to view a larger version of this figure.

Figure 4: ASMR and ASDR for neoplasms attributable to low physical activity by SDI across 21 GBD regions from 1990 to 2021. (A) ASMR and (B) ASDR. The fitted curve illustrates the overall association between SDI and burden. Regions above or below the fitted curve had observed rates higher or lower than expected at a similar SDI level. Please click here to view a larger version of this figure.

Figure 5: ASMR and ASDR for neoplasms attributable to low physical activity by SDI across 204 countries and territories in 2021. (A) ASMR and (B) ASDR. Countries above or below the fitted SDI curve had observed rates higher or lower than expected at a similar SDI level. These deviations may reflect exposure distribution, population structure, cancer detection, registry completeness, or uncertainty in modeled estimates. Please click here to view a larger version of this figure.

Figure 6: Country-level ASMR and ASDR for neoplasms attributable to low physical activity in 2021. (A) ASMR and (B) ASDR. Color categories represent point estimates per 100,000 population. Estimates for countries or territories with small populations or wide uncertainty intervals should be interpreted cautiously. Please click here to view a larger version of this figure.

Figure 7: Decomposition of changes in deaths and DALYs attributable to low physical activity into aging, population growth, and epidemiological change components. (A) Deaths and (B) DALYs. Positive values indicate components contributing to increased burden, whereas negative values indicate components offsetting burden growth. Percentages may exceed 100% or be negative when components act in opposite directions. Please click here to view a larger version of this figure.

Figure 8: Health inequality analysis of neoplasms attributable to low physical activity. (A–D) Cumulative fraction of death and DALYs. The SDI represents the absolute difference in burden between the highest and lowest ends of the SDI distribution, whereas the Concentration Index represents relative inequality across the SDI-ranked population. Values closer to zero indicate lower inequality. Please click here to view a larger version of this figure.

Figure 9: Bayesian age-period-cohort projections of ASMR and ASDR for neoplasms attributable to low physical activity from 2022 to 2045. (A–F) Solid lines indicate posterior mean estimates, and shaded areas indicate 95% uncertainty intervals. Projections should be interpreted as model-based estimates conditional on historical trends and model assumptions. Please click here to view a larger version of this figure.

	1990		2021		1990-2021
	Death cases	ASR per 100,000	Death cases	ASR per 100,000	PC in ASRs（%）	EAPC of ASMR
	Number (95% UI)	ASMR (95% UI)	Number (95% UI)	ASMR (95% UI)	(95% UI)	(95% CI)
Global	44,674.70(24,963.44 to 64,996.55)	1.32(0.75 to 1.92)	87,440.12(48,636.46 to 125,810.50)	1.06(0.59 to 1.52)	-19.72(-27.11 to -11.38)	-0.79(-0.83 to -0.75)
​Sex​
Male	12,882.94(7,911.97 to 17,917.59)	0.92(0.57 to 1.30)	29,838.39(18,371.98 to 42,033.66)	0.84(0.52 to 1.19)	-8.61(-23.80 to 9.61)	-0.32(-0.36 to -0.28)
Female	31,791.77(16,646.68 to 46,902.98)	1.61(0.85 to 2.38)	57,601.73(29,060.33 to 85,196.67)	1.24(0.62 to 1.83)	-23.17(-31.34 to -14.61)	-0.96(-1.00 to -0.91)
​SDI​
Low SDI	702.55(360.97 to 1,079.18)	0.36(0.19 to 0.55)	1,755.02(845.64 to 2,662.49)	0.41(0.20 to 0.61)	11.58(-7.78 to 39.82)	0.36(0.24 to 0.49)
Low-middle SDI	2,326.69(1,203.88 to 3,443.41)	0.44(0.24 to 0.65)	7,394.12(3,580.19 to 10,989.26)	0.57(0.28 to 0.84)	29.63(9.70 to 53.53)	0.86(0.81 to 0.91)
Middle SDI	7,125.05(3,839.38 to 10,401.70)	0.84(0.46 to 1.23)	21,815.22(12,091.00 to 31,568.52)	0.89(0.50 to 1.28)	5.81(-10.55 to 25.55)	0.06(0.01 to 0.11)
High-middle SDI	11,715.52(6,575.68 to 16,808.12)	1.34(0.76 to 1.90)	24,257.36(13,536.14 to 34,810.76)	1.24(0.70 to 1.78)	-7.34(-20.01 to 6.36)	-0.30(-0.36 to -0.23)
High SDI	22,740.72(12,713.50 to 33,355.57)	2.03(1.14 to 3.00)	32,101.73(17,720.24 to 46,072.78)	1.35(0.75 to 1.95)	-33.53(-40.64 to -26.20)	-1.42(-1.48 to -1.36)
​GBD region​
Andean Latin America	94.66(46.63 to 144.02)	0.52(0.26 to 0.79)	334.93(165.85 to 522.17)	0.59(0.29 to 0.92)	14.07(-17.75 to 60.72)	0.42(0.31 to 0.53)
Australasia	622.72(330.83 to 894.24)	2.70(1.44 to 3.87)	985.92(553.20 to 1,451.29)	1.69(0.95 to 2.49)	-37.32(-48.68 to -23.42)	-1.62(-1.70 to -1.55)
Caribbean	307.15(157.94 to 442.19)	1.27(0.66 to 1.84)	718.83(375.11 to 1,074.48)	1.33(0.69 to 1.98)	4.11(-14.10 to 25.40)	0.21(0.16 to 0.25)
Central Asia	255.09(121.93 to 379.62)	0.58(0.28 to 0.86)	338.66(168.31 to 509.00)	0.48(0.24 to 0.72)	-17.80(-31.97 to -1.09)	-0.08(-0.26 to 0.10)
Central Europe	2,266.85(1,272.80 to 3,271.41)	1.62(0.91 to 2.32)	4,170.99(2,237.91 to 6,015.29)	1.75(0.93 to 2.53)	7.93(-7.79 to 24.67)	0.09(-0.03 to 0.21)
Central Latin America	384.26(183.94 to 573.19)	0.52(0.25 to 0.77)	1,556.11(800.47 to 2,334.25)	0.64(0.33 to 0.95)	23.59(2.30 to 50.21)	0.71(0.48 to 0.93)
Central Sub-Saharan Africa	83.65(40.22 to 132.35)	0.48(0.24 to 0.75)	236.51(100.45 to 393.49)	0.55(0.25 to 0.91)	14.74(-21.74 to 63.78)	0.46(0.29 to 0.62)
East Asia	6,768.09(3,653.81 to 10,056.11)	1.03(0.57 to 1.53)	19,607.89(11,453.34 to 2,9247.43)	0.98(0.57 to 1.45)	-4.73(-32.51 to 33.60)	-0.29(-0.36 to -0.21)
Eastern Europe	2,802.45(1,497.00 to 3,941.72)	1.05(0.57 to 1.47)	4,428.85(2,320.28 to 6,600.14)	1.21(0.63 to 1.80)	15.79(-8.61 to 44.43)	0.36(0.26 to 0.47)
Eastern Sub-Saharan Africa	225.93(105.51 to 360.94)	0.35(0.17 to 0.55)	588.90(269.23 to 909.90)	0.42(0.19 to 0.65)	19.13(-7.33 to 59.92)	0.53(0.40 to 0.65)
High-income Asia Pacific	2,898.71(1,630.17 to 4,145.85)	1.52(0.86 to 2.19)	7,306.42(4,018.91 to 10,907.07)	1.31(0.71 to 1.90)	-14.02(-33.96 to 15.28)	-0.55(-0.60 to -0.50)
High-income North America	5,859.88(2,968.16 to 9,058.79)	1.59(0.80 to 2.47)	7,243.90(3,643.52 to 10,755.63)	1.03(0.52 to 1.53)	-34.95(-51.35 to -13.14)	-1.52(-1.62 to -1.41)
North Africa and Middle East	1,271.47(733.69 to 1,840.36)	0.88(0.51 to 1.25)	3936.12(2155.17 to 5755.13)	0.98(0.56 to 1.43)	11.42(-5.74 to 33.24)	0.67(0.50 to 0.84)
Oceania	18.63(7.98 to 30.63)	0.67(0.31 to 1.07)	47.63(18.61 to 78.91)	0.63(0.26 to 1.03)	-5.55(-23.06 to 16.71)	-0.20(-0.29 to -0.12)
South Asia	1,737.90(882.29 to 2,628.84)	0.35(0.18 to 0.53)	5,543.21(2,719.11 to 8,491.21)	0.42(0.21 to 0.64)	20.20(-6.28 to 49.97)	0.44(0.32 to 0.56)
Southern Latin America	1,913.00(997.87 to 2,899.66)	0.87(0.46 to 1.32)	6,763.64(3,374.55 to 10,004.93)	1.16(0.59 to 1.69)	33.53(6.63 to 63.18)	0.85(0.76 to 0.94)
Southern Sub-Saharan Africa	464.42(233.50 to 723.20)	1.08(0.56 to 1.67)	860.59(417.20 to 1,323.57)	0.95(0.46 to 1.46)	-12.08(-37.87 to 27.28)	-0.21(-0.35 to -0.07)
Southeast Asia	267.72(130.05 to 417.79)	1.14(0.56 to 1.76)	753.71(364.83 to 1,110.43)	1.53(0.77 to 2.26)	34.50(5.96 to 67.10)	1.00(0.79 to 1.21)
Tropical Latin America	868.79(440.43 to 1,265.92)	1.12(0.57 to 1.63)	3,270.12(1,677.67 to 4,855.25)	1.31(0.68 to 1.94)	16.96(-6.80 to 49.74)	0.56(0.49 to 0.63)
Western Europe	15,255.70(8,488.72 to 22,265.38)	2.52(1.39 to 3.69)	17,835.97(9,801.50 to 25,806.91)	1.59(0.87 to 2.32)	-36.81(-43.81 to -29.62)	-1.56(-1.61 to -1.50)
Western Sub-Saharan Africa	307.64(131.59 to 478.58)	0.40(0.18 to 0.62)	911.22(348.96 to 1,466.65)	0.53(0.22 to 0.84)	32.49(3.30 to 74.66)	1.05(0.94 to 1.16)
​Neoplasms subcategories​
Breast cancer	8,177.91(1,662.64 to 1,4354.52)	0.23(0.05 to 0.40)	15,964.16(3,230.21 to 28,326.21)	0.19(0.04 to 0.34)	-15.99(-22.39 to -9.41)	-0.66(-0.72 to -0.61)
Colon and rectum cancer	3,6496.79(23,150.55 to 5,0771.76)	1.09(0.70 to 1.53)	71,475.96(44,830.06 to 98,195.49)	0.87(0.55 to 1.19)	-20.49(-28.57 to -11.42)	-0.82(-0.86 to -0.78)

Table 1: Neoplasms attributable to low physical activity: death burden and temporal trends (1990–2021). The table presents the main burden estimates and trend indicators by sex and SDI category. More granular regional and national estimates are provided in the supplementary tables. Please click here to download this Table.

	1990		2021		1990-2021
	DALYs cases	ASR per 100,000	DALYs cases	ASR per 100,000	PC in ASRs（%）	EAPC of ASDR
	Number (95% UI)	ASDR (95% UI)	Number (95% UI)	ASDR (95% UI)	(95% UI)	(95% CI)
Global	944,285.78(50,5737 .18 to 1,389,543.57)	25.16(13.67 to 37.09)	1,744,943.68(908,799 .35 to 2,545,344.72)	20.50(10.73 to 29.89)	-18.51(-26.10 to -10.09)	-0.76(-0.82 to -0.70)
​Sex​
Male	268,107.56(165,239 .07 to 370,163.08)	16.57(10.18 to 22.94)	574,371.32(348,379 .46 to 800,744.94)	15.03(9.21 to 20.93)	-9.33(-23.71 to 7.51)	-0.37(-0.43 to -0.32)
Female	676,178.23(336,047 .59 to 992,525.42)	32.38(16.26 to 47.60)	1,170,572.36(555,313 .54 to 1,756,517.84)	25.46(12.04 to 38.21)	-21.36(-29.62 to -12.76)	-0.89(-0.95 to -0.83)
​SDI​
Low SDI	18,578.95(9,176 .46 to 28,775.19)	8.16(4.14 to 12.53)	45,807.13(20,570.83 to 70,949.02)	8.83(4.16 to 13.49)	8.29(-10.50 to 39.10)	0.21(0.09 to 0.33)
Low-middle SDI	60,418.06(29,886 .07 to 90,000.18)	9.82(4.99 to 14.56)	181,922.35(84,591 .84 to 280,191.13)	12.56(5.94 to 19.03)	27.86(8.31 to 53.45)	0.77(0.73 to 0.81)
Middle SDI	174,555.29(91,042 .96 to 256,446.54)	17.36(9.33 to 25.39)	488,709.60(253,743 .20 to 717,931.89)	18.38(9.67 to 26.94)	5.89(-10.57 to 26.39)	0.07(-0.00 to 0.14)
High-middle SDI	248,517.72(134,358 .78 to 358,881.54)	25.99(14.17 to 37.65)	462,899.84(248,381 .01 to 674,525.73)	23.44(12.50 to 34.20)	-9.81(-22.49 to 4.59)	-0.42(-0.46 to -0.38)
High SDI	440,886.67(236,640 .98 to 641,945.58)	39.77(21.25 to 58.06)	563,339.77(306,210 .10 to 819,262.15)	26.52(14.13 to 38.73)	-33.32(-39.10 to -27.18)	-1.40(-1.47 to -1.33)
​GBD region​
Andean Latin America	2,032.06(953 .15 to 3,244.52)	10.17(4.81 to 15.75)	6,788.27(3,379.82 to 10,869.37)	11.60(5.80 to 18.58)	14.05(-16.37 to 57.15)	0.38(0.28 to 0.48)
Australasia	13,091.85(6,785.74 to 19,155.80)	56.24(29.15 to 82.31)	18,402.32(10,330 .87 to 27,196.61)	34.88(19.18 to 51.79)	-37.98(-48.06 to -25.70)	-1.64(-1.72 to -1.57)
Caribbean	6,682.48(3,257.50 to 9,803.70)	26.05(12.83 to 37.96)	14,936.71(7,467.85 to 22,741.58)	27.76(13.82 to 42.35)	6.56(-11.43 to 26.59)	0.30(0.26 to 0.33)
Central Asia	6,113.13(2,875.45 to 9,342.10)	13.08(6.17 to 19.82)	7,871.50(3,630.97 to 11,955.55)	9.94(4.74 to 14.99)	-24.04(-36.62 to -7.97)	-0.47(-0.59 to -0.35)
Central Europe	46,026.42(24,775 .41 to 66,105.09)	31.28(16.84 to 45.01)	75,115.03(39,699 .13to 107,678.56)	32.89(17.34 to 47.12)	5.15(-8.57 to 19.95)	0.04(-0.06 to 0.14)
Central Latin America	9,305.49(4,170.58 to 13,956.30)	11.02(5.09 to 16.45)	37,715.23(18,709.78 to 58,000.93)	14.87(7.41 to 22.65)	34.95(11.77 to 64.97)	1.01(0.75 to 1.26)
Central Sub-Saharan Africa	2,221.31(1,016.14 to 3,556.00)	10.23(4.88 to 16.16)	6,337.46(2,428.48 to 10,487.21)	11.63(4.95 to 19.40)	13.71(-22.36 to 57.99)	0.43(0.24 to 0.61)
East Asia	159,121.88(82,933 .80 to 241,470.88)	19.79(10.48 to 29.45)	396,653.40(224,218 .04 to 604,268.93)	18.61(10.59 to 28.17)	-5.96(-32.26 to 32.19)	-0.33(-0.43 to -0.23)
Eastern Europe	59,978.58(30,902 .71 to 87,472.34)	21.54(10.95 to 31.20)	82,590.09(40,451.64 to 123,453.92)	22.92(11.22 to 34.24)	6.38(-14.08 to 33.51)	0.00(-0.10 to 0.10)
Eastern Sub-Saharan Africa	6,144.30(2,759.22 to 9,942.73)	7.98(3.74 to 12.73)	15,726.10(6,676.96 to 24,634.43)	8.95(4.06 to 13.94)	12.10(-14.88 to 51.44)	0.29(0.16 to 0.43)
High-income Asia Pacific	68,367.61(37,187 .31 to 96,750.88)	33.95(18.43 to 48.21)	125,612.22(68,484.09 to 180,959.20)	27.96(14.89 to 39.87)	-17.66(-34.93 to 3.02)	-0.70(-0.75 to -0.64)
High-income North America	110,579.32(53,603 .23 to 168,944.28)	30.91(14.75 to 47.02)	135,048.33(65,207.39 to 206,531.98)	20.66(9.84 to 31.60)	-33.18(-48.18 to -14.30)	-1.39(-1.56 to -1.22)
North Africa and Middle East	33,502.36(18,573 .46 to 49,734.41)	19.49(11.13 to 28.52)	102,096.12(51,831.63 to 151,539.80)	21.70(11.51 to 31.87)	11.34(-6.57 to 32.99)	0.60(0.46 to 0.73)
Oceania	612.40(245.17 to 1,023.79)	17.68(7.53 to 29.11)	1,573.18(576.67 to 2,667.60)	17.19(6.65 to 28.58)	-2.72(-23.76 to 23.48)	-0.13(-0.21 to -0.05)
South Asia	46,100.94(22,425 .80 to 69,073.26)	7.89(3.95 to 11.94)	133,476.27(62,105.33 to 207,069.89)	9.05(4.30 to 14.00)	14.67(-10.43 to 43.16)	0.21(0.08 to 0.34)
Southern Latin America	48,213.16(24,344 .03 to 74,368.69)	19.00(9.75 to 29.04)	161,677.60(76,724.20 to 247,604.85)	24.67(12.01 to 37.15)	29.87(1.97 to 61.27)	0.77(0.70 to 0.83)
Southern Sub-Saharan Africa	9,529.23(4,537.37 to 14,950.84)	21.05(10.09 to 33.08)	15,906.65(7,462.34 to 24,568.27)	18.09(8.45 to 28.03)	-14.05(-36.07 to 18.51)	-0.32(-0.41 to -0.24)
Southeast Asia	6,545.75(3,015.15 to 10,197.86)	23.82(11.30 to 37.07)	18,303.10(8,337.73 to 27,183.17)	31.82(14.94 to 46.79)	33.57(7.39 to 62.74)	1.08(0.89 to 1.27)
Tropical Latin America	20,004.58(9,888.11 to 30,396.32)	22.53(11.21 to 33.34)	71,793.02(34,457.98 to 107,876.77)	27.92(13.51 to 41.78)	23.94(0.01 to 55.61)	0.71(0.66 to 0.76)
Western Europe	282,256.07(150,133 .40 to 413,719.92)	48.37(25.25 to 71.16)	292,870.10(158,018 .77 to 430,007.70)	30.13(15.78 to 43.96)	-37.71(-43.35 to -31.66)	-1.60(-1.66 to -1.55)
Western Sub-Saharan Africa	7,856.86(3,225.38 to 12,446.50)	8.86(3.71 to 13.83)	24,451.00(8,434.07 to 40,517.27)	11.81(4.42 to 19.19)	33.20(1.42 to 76.70)	1.07(0.93 to 1.21)
​Neoplasms subcategories​
Breast cancer	219,022.34(43,911 .88 to 378,430.50)	5.47(1.10 to 9.50)	415,254.34(83,318. 37 to 729,174.93)	4.81(0.97 to 8.45)	-11.93(-18.74 to -4.98)	-0.51(-0.57 to -0.44)
Colon and rectum cancer	725,263.44(458,229 .16 to 1,007,049.44)	19.69(12.45 to 27.40)	1,329,689.34(831,183 .72 to 1,829,992.61)	15.69(9.81 to 21.54)	-20.34(-27.99 to -10.82)	-0.83(-0.89 to -0.77)

Table 2: Neoplasms attributable to low physical activity: DALYs burden and temporal trends (1990–2021). The table presents the main burden estimates and trend indicators by sex and SDI category. More granular regional and national estimates are provided in the supplementary tables. Please click here to download this Table.

Supplementary Figure 1: Joinpoint regression analysis of the sex-specific (A) ASMR and (B) ASDR (per 100,000 population) for neoplasms subcategories attributable to low physical activity from 1990 to 2021. Blue curve, Joinpoint regression analysis of both genders for BC; Brown curve, Joinpoint regression analysis of both genders for CRC; Green curve, Joinpoint regression analysis of female gender for CRC; Orange curve, Joinpoint regression analysis of male gender for CRC; Red curve, Joinpoint regression analysis of female gender for BC. ASMR, Age-standardized mortality rate; ASDR, Age-standardized disability-adjusted life years rate; BC, Breast cancer; CRC, Colon and rectum cancer; DALYs, Disability-adjusted life years.Please click here to download this file.

Supplementary Figure 2: Neoplasms attributable to low physical activity: Age-specific numbers, ASMR and ASDR (per 100,000 population) by gender, 1990. ASMR, Age-standardized mortality rate; ASDR, Age-standardized disability-adjusted life years rate; DALYs, Disability-adjusted life years.Please click here to download this file.

Supplementary Figure 3: Time trends in global deaths and DALYs of neoplasms attributable to low physical activity by SDI, 1990–2021. DALYs, Disability-adjusted life years; SDI, Socio-demographic Index.Please click here to download this file.

Supplementary Figure 4: The global EAPC of ASMR (A) and ASDR (B) (per 100,000 population) for neoplasms attributable to low physical activity in 204 countries and territories from 1990 to 2021. ASMR, Age-standardized mortality rate; ASDR, Age-standardized disability-adjusted life years rate; DALYs, Disability-adjusted life years; EAPC, Estimated annual percentage change.Please click here to download this file.

Supplementary Table 1: Percentage change in global deaths and DALYs of neoplasms attributable to low physical activity by gender, 1990–2021. DALYs, Disability-adjusted life years.Please click here to download this file.

Supplementary Table 2: Joinpoint regression analysis of the sex-specific ASMR and ASDR (per 100,000 population) for neoplasms attributable to low physical activity from 1990 to 2021. ASMR, Age-standardized mortality rate; ASDR, Age-standardized disability-adjusted life years rate; AAPC, Average annual percent change presented for full period; APC, Annual percent change; CI, Confidence interval.Please click here to download this file.

Supplementary Table 3: Joinpoint regression analysis of the sex-specific ASMR and ASDR (per 100,000 population) for neoplasms subcategories attributable to low physical activity from 1990 to 2021. ASMR, Age-standardized mortality rate; ASDR, Age-standardized disability-adjusted life years rate; AAPC, Average annual percent change presented for full period; APC, Annual percent change; BC, Breast cancer; CRC, Colon and rectum cancer; CI, Confidence interval.Please click here to download this file.

Supplementary Table 4: Age-specific numbers and Age-standardized rates of deaths of neoplasms attributable to low physical activity by gender, in 1990 and 2021.Please click here to download this file.

Supplementary Table 5: Age-specific numbers and Age-standardized rates of DALYs of neoplasms attributable to low physical activity by gender, in 1990 and 2021. DALYs, Disability-adjusted life years.Please click here to download this file.

Supplementary Table 6: The death burden of neoplasms attributable to low physical activity in 204 countries and territories in 1990 and 2021 and the temporal trends from 1990 to 2021. ASMR, Age-standardized mortality rate; ASR, Age-standardized rate; CI, Confidence interval; EAPC, Estimated annual percentage change; UI, Uncertainty interval.Please click here to download this file.

Supplementary Table 7: The DALYs burden of neoplasms attributable to low physical activity in 204 countries and territories in 1990 and 2021, and the temporal trends from 1990 to 2021. ASDR, age-standardized DALY rate; ASR, age-standardized rate; CI, confidence interval; DALYs, disability-adjusted life years; EAPC, estimated annual percentage change; UI, uncertainty interval. Please click here to download this file.

Supplementary Table 8: Decomposition of changes in deaths and DALYs of neoplasms attributable to low physical activity into aging, population, and epidemiological change. DALYs, Disability-adjusted life years.Please click here to download this file.

Supplementary Table 9: Health inequality analysis of neoplasms attributable to low physical activity. DALYs, Disability-adjusted life years.Please click here to download this file.

Supplementary Table 10: Global trends in ASMR and ASDR (per 100,000 population) for neoplasms attributable to low physical activity, predicted by the BAPC, 2021–2045. ASMR, Age-standardized mortality rates; ASDR, Age-standardized disability-adjusted life year rates; BAPC, Bayesian Age-Period-Cohort model; UI, Uncertainty interval.Please click here to download this file.

Discussion

This GBD 2021 analysis shows two contrasting patterns in low physical activity-attributable neoplasm burden. While ASMR and ASDR for neoplasms attributable to LPA significantly declined from 1990 to 2021 (ΔASMR = -19.72%; ΔASDR = -18.51%), absolute deaths and DALYs increased by 95.73% and 84.79%, respectively. The increase in absolute burden despite declining age-standardized rates was mainly explained by population growth and population aging. Declining ASRs may reflect multiple factors, including improvements in cancer detection, treatment, and risk-factor control, although these contributors were not directly decomposed in the present analysis^1,21,22. Notably, LPA remained responsible for over 87,000 neoplasm-related deaths and 1.74 million DALYs globally in 2021. These findings support the continued relevance of low physical activity in the GBD attribution framework. As this study used GBD comparative risk assessment estimates, the results quantify modeled population-level burden attributable to low physical activity and do not establish individual-level causal mechanisms or direct intervention effects.

Notable SDI-based disparities persist in neoplasms attributable to LPA burden. High-SDI regions (North America, Western Europe, Australia) demonstrated the highest 2021 ASMR (1.35/100,000) and ASDR (26.52/100,000) yet achieved maximal 1990–2021 reductions (EAPC: -1.42; -1.40), reflecting concurrent advances in: (i) evidence-based screening (e.g., UK's 60% colorectal screening coverage)²³, (ii) integrated therapeutic modalities (e.g., robotic-assisted surgery adoption)²⁴, and (iii) nationwide PA initiatives (e.g., South Korea's National Fitness Award Project)²⁵. Paradoxically, low-middle SDI regions (particularly sub-Saharan Africa) showed rising ASMR/ASDR trends (EAPC: +0.86; +0.77), strongly correlating with health-system constraints and absent preventive policies²⁶. This stratification aligns with WHO's Global Report on Noncommunicable Diseases 2022, underscoring low- and middle-income countries (LMICs)' urgent need for cancer control strengthening through: population-based cancer registry implementation, evidence-based screening program scaling, and integrated palliative care service delivery²⁷.

The SDI-burden correlation (ASMR ρ = 0.7560; ASDR ρ = 0.7364) reveals socioeconomic development's paradoxical effects on neoplasms attributable to LPA. High-SDI regions achieved significant mortality reduction (EAPC = -1.42) through enhanced screening and targeted therapies, yet simultaneously faced rising obesity-related malignancies (e.g., breast/colorectal cancers) from westernized diets and sedentary occupations^21,22,28. East Asia, particularly China, recorded the highest absolute number of low physical activity-attributable deaths and DALYs, which may partly reflect its large population size and demographic structure^29,30. Country-level outliers in Caribbean and Pacific settings may reflect a combination of exposure distribution, population structure, cancer detection, registry completeness, and GBD model assumptions³¹. Because these factors were not directly tested in the present analysis, they are discussed as possible contextual explanations rather than confirmed mechanisms^32,33.

Globally, women exhibit a significantly higher burden of neoplasms attributable to LPA, with the 2021 ASMR (1.24/100,000) and ASDR (25.46/100,000) surpassing male rates (0.84/100,000; 15.03/100,000), primarily driven by the high incidence of BC³⁴. LPA increases BC mortality risk (HR = 1.22, 95% CI:1.05–1.42), while engagement in non-occupational PA, irrespective of type or intensity, demonstrates varying magnitudes of risk reduction for BC incidence^35,36. Previous studies have proposed hormonal, metabolic, and inflammatory pathways linking physical activity with breast cancer risk; however, these mechanisms were not directly examined in the present GBD-based analysis³⁷. Nevertheless, global physical inactivity prevalence remains 35.5% higher in women (31.7% vs. 23.4% non-compliance with WHO guidelines)⁸, exacerbated in low-SDI regions by cultural constraints (attire restrictions: 37% reduced PA probability) and gendered caregiving burdens (4.3-hour daily disparity in domestic work: 29% reduction)³⁸.

Neoplasms attributable to LPA peak globally among 70–84-year-olds, consistent with neoplastic progression and age-related metabolic decline^39,40. Moderated ASMR/ASDR increases in elderly males correlate with sustained engagement in low-intensity activities (gardening/walking), contrasting with heightened female susceptibility where annual sarcopenia (1%–2% muscle loss) and osteomuscular comorbidities accelerate LPA vulnerability^41,42. These findings suggest that age-specific patterns should be considered when interpreting low physical activity-attributable burden.

Population factors emerged as the principal driver of increasing mortality from neoplasms attributable to LPA (88.44%) and elevated DALYs (131.53%). While Epidemiological-Change offset -123.04% disease burden in high-SDI regions, low-SDI regions experienced aggravated burden (combined contribution rate: 36.33%) through Aging and Epidemiological-Change. These findings suggest that developing nations undergoing demographic transitions (e.g., fertility decline, increased life expectancy) risk confronting dual burdens of communicable and NCDs without parallel development of comprehensive cancer prevention systems⁴³.

Health inequality analysis indicates persistent ASMR disparities (>1/100,000) between high- and low-SDI regions despite absolute/relative gap reductions (SII: 1.45→1.11/100,000; CII: 0.30→0.20 from 1990 to 2021). Three principal barriers drive inequity: (i) Infrastructure disparity: High-SDI countries/regions maintain fivefold greater fitness facility coverage than low-SDI counterparts. In Kolar, India, per capita recreational space measures merely 0.4 m², significantly below the 3.0–4.0 m² per capita standard in high-SDI regions, with most facilities concentrated in urban centers, creating spatial distribution disparities that restrict equitable health resource accessibility⁴⁴. (ii) Health education gap: Only 30% of low-SDI schools incorporate PA in compulsory curriculum (<50% implementation rate), contrasting with high-SDI nations' standardized school health programs (150 minutes/week activity) and dedicated health supervisors⁸. (iii) Resource allocation imbalance: Low-SDI regions housing >10% global population account for merely 0.4% of worldwide health expenditure. Concurrently, development assistance for health (DAH) allocates minimal funds to noncommunicable disease prevention, including cancer control, reinforcing geographical clustering of disease burden through this differential funding pattern⁴⁵.

Neoplasms Attributable to LPA demonstrate projected global declines via BAPC modeling (2045 ASMR: 0.95/100,000, 95%UI 0.88–1.03; ASDR: 18.86/100,000, 16.99–20.72). Despite this trajectory, low-SDI regions persistently confront healthcare inequities. These findings support prioritizing low physical activity within cancer risk-factor surveillance and prevention planning⁴⁶, particularly for older adults, women, and settings with increasing age-standardized rates⁴⁷. However, specific intervention strategies should be adapted to local resources and evaluated in implementation studies before broad policy adoption⁴⁸. The present analysis can help identify populations and regions that may warrant closer monitoring, but it does not directly test the effectiveness of any specific intervention.

Nevertheless, several limitations warrant consideration. First, this study relied on GBD modeled estimates rather than primary individual-level data, and the estimate precision may differ across countries according to data availability, registry completeness, and reporting quality⁴⁹. Second, physical activity exposure in GBD is partly derived from survey-based data, which may be affected by recall bias, social desirability bias, and differences in measurement instruments across settings. Third, the GBD low physical activity-attributable neoplasm category is limited to colorectal cancer and female breast cancer⁵⁰, and therefore does not represent all malignant tumors. Fourth, the exposure definition does not fully distinguish occupational, household, transport-related, and leisure-time physical activity, which may have different associations with cancer outcomes. Fifth, socioeconomic comparisons may be affected by differences in cancer detection, diagnostic capacity, and data availability across countries. Finally, BAPC projections depend on historical trends and modeling assumptions; estimates farther from the observed period should therefore be interpreted cautiously.

In conclusion, low physical activity remains an important modifiable risk factor in the GBD attribution framework, and its estimated attributable burden varies substantially by sex, age, geography, and socioeconomic development. While high-SDI regions have reduced disease burdens through technological innovations, low-SDI regions continue to face the dual challenges of population growth and systemic health inequities. Future strategies should prioritize tiered interventions, intersectoral collaboration, and technology-driven solutions to refine global cancer prevention frameworks, with particular emphasis on addressing the unique needs of women, aging populations, and LMICs. This study provides pivotal evidence for global public health policymaking, underscoring the imperative to integrate PA promotion as a central component of national cancer control strategies.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Bayesian Age-Period-Cohort (BAPC) R Package	R Foundation / CRAN Repository	https://cran.r-project.org/	Used for Bayesian age-period-cohort projection analyses.
CSV Data Files	Institute for Health Metrics and Evaluation (IHME)	N/A	Exported burden datasets from GBD 2021 Results Tool for downstream analysis.
dplyr R Package	R Foundation / CRAN Repository	https://cran.r-project.org/package=dplyr	Used for data manipulation and cleaning in R.
GBD 2021 Data Input Sources Tool	Institute for Health Metrics and Evaluation (IHME)	https://ghdx.healthdata.org/gbd-2021	Public repository for GBD source documentation.
ggplot2 R Package	R Foundation / CRAN Repository	https://cran.r-project.org/package=ggplot2	Used for graphical visualization of temporal trends and inequality analyses.
Global Burden of Disease (GBD) 2021 Results Tool	Institute for Health Metrics and Evaluation (IHME)	https://vizhub.healthdata.org/gbd-results/	Public database used for extraction of deaths, DALYs, ASMR, and ASDR data.
INLA R Package	R-INLA Project	https://www.r-inla.org/	Used for Bayesian inference within age-period-cohort modeling.
JD_GBDR Software	Jingding Medical Technology Co., Ltd.	Version 2.37.1	Used for GBD burden estimation, decomposition analysis, and standardized rate visualization.
Joinpoint Regression Program (Version 5.3.0)	National Cancer Institute (NCI)	https://surveillance.cancer.gov/joinpoint/download	Used for joinpoint regression analysis, APC, and AAPC estimation.
Microsoft Excel	Microsoft Corporation	https://www.microsoft.com/excel	Used for organizing and cleaning extracted CSV datasets prior to analysis.
Personal Computer Workstation	Any standard computing workstation	N/A	Recommended minimum: 8 GB RAM, Windows 10/11 or macOS for statistical analyses and projections.
R Statistical Software (Version 4.4.2)	R Foundation for Statistical Computing	https://www.r-project.org/	Used for statistical analysis, EAPC calculation, inequality analysis, visualization, and Bayesian age-period-cohort modeling.
RStudio Desktop	Posit Software, PBC	https://posit.co/download/rstudio-desktop/	Integrated development environment for R analyses.