Factors influencing the prevalence and death rate of COPD: a pan-database ecological study covering 201 countries and regions from 1990 to 2021

Summary

Background

Chronic obstructive pulmonary disease (COPD) is a common, heterogeneous disease and may be influenced by diverse factors. However, gaps remain in previous studies regarding the exploration of potential influencing factors. This study aims to investigate the wide range of potential factors influencing COPD based on a pan-database ecological analysis.

Methods

We integrated data from 17 global databases (e.g., the Global Burden of Disease Study) which encompass social, environmental, and health data across various regions. Generalized linear regression was used to analyze the association of cumulative and instant exposures of factors with COPD and to rank their importance. Spearman analysis was used to assess the correlation of various factors with COPD. Heatmaps, scatter plots, and nonclassical multidimensional scaling (i.e., network graph) were employed to visualize the correlations.

Findings

The study aggregated 77 social and environmental factors, 85 lifestyle and dietary factors, 25 physiological indicators, and 28 diseases. In the cumulative exposure analysis, tobacco consumption, atmospheric pollutants (e.g., ozone, CO, and organic matter aerosol), biomass cooking, and climatic conditions (e.g., vapor pressure, solar radiation, and temperature) were found to be associated with COPD prevalence. Additionally, tobacco consumption, social factors (e.g., hunger index, gender inequality index, and education year), and climatic conditions significantly impacted death rates. The results for cumulative exposures were consistent with those for instant exposures. Network graph analysis indicated a positive correlation between COPD and chronic kidney disease, other chronic respiratory diseases, gout, and stroke.

Interpretation

Various factors (e.g., tobacco consumption, atmospheric pollution, biomass cooking, temperature, social factors, and comorbidities) significantly influence COPD. Comprehensive interventions are needed to reduce the disease burden of COPD.

Funding

The Foundation of Guangzhou National Laboratory (SRPG22-018, SRPG22-016), State Key Laboratory of Respiratory Disease (SKLRD-OP-202402), Zhejiang medical health science and technology project (2025KY1245), Multi-Center Clinical Research Project of Guangzhou Medical University (GMUCR2025-02009), and Guangzhou Municipal Science and Technology Bureau (SL2024A04J00706).

Research in context.

Evidence before this study

We searched PubMed without date restrictions, using the search terms for titles and abstracts: (“chronic obstructive pulmonary disease” OR “COPD”) AND (“burden” OR “risk factor” OR “comorbidity”), with no language restrictions. From 10,237 results, we identified 153 relevant publications. Of these, 12 studies analyzed the global burden of COPD, 64 studies discussed various risk factors (e.g., tobacco consumption, air pollution, and socioeconomic status) of COPD, and 21 studies evaluated the comorbidities of COPD. However, none of these studies have integrated multiple databases to comprehensively investigate the wide range of potential factors influencing COPD.

Added value of this study

This ecological study conducts a global pan-database synthesis using 17 global databases, collecting data on 215 social and environmental factors, lifestyle and dietary factors, physiological indicators, and diseases, with the aim of investigating the potential risk factors of COPD.

Implications of all the available evidence

This study reveals the critical roles of tobacco consumption, biomass cooking, ambient air pollution, climate factors, and social factors in the prevalence and death rates of COPD. These findings provide valuable references for the formulation of public health policies.

Introduction

Chronic obstructive pulmonary disease (COPD) is one of the leading causes of disability-adjusted life years (DALYs) and deaths worldwide, imposing a heavy burden on society and healthcare systems.^1,2 According to the World Health Organization (WHO), approximately 3.5 million people globally died from COPD in 2021, accounting for 5% of all deaths.³ The prevalence and death rates of COPD vary significantly across countries and regions,^4,5 likely due to the heterogeneity of risk factors and variations in the methods (i.e., ascertainment) used to assess the disease across areas.

Smoking is widely recognized as the most significant risk factor for COPD.⁶ However, an increasing number of studies suggest that in addition to smoking, other factors are also closely associated with the onset and progression of COPD.⁷ Indoor and outdoor air pollution, particularly biomass cooking, is considered one of the major risk factors for COPD in the developing countries.^8,9 Moreover, occupational exposure, childhood respiratory infections, socioeconomic factors, genetic susceptibility, and comorbidities are also considered to significantly influence COPD.10, 11, 12, 13

Despite the extensive research on COPD risk factors, most studies face the following limitations. Firstly, they depend on a single database and common study designs (e.g., cross-sectional and cohort studies), lacking integration with additional data sources.^14,15 Secondly, previous studies typically focus on a few factors, such as smoking and its related variables, while overlooking a broad range of other indicators, including social, environmental, and dietary factors.16, 17, 18 Lastly, previous analyses have failed to quantitatively compare and rank the importance of risk factors globally, limiting a comprehensive understanding of the global risk factors. Over the past 30 years, changes in common factors may have new impacts on the disease burden of COPD. Therefore, comprehensive research is needed, utilizing multiple databases to assess the influence of various factors on the prevalence and death rate of COPD.

We hypothesize that a range of social, environmental, lifestyle, and health factors may jointly influence the prevalence and death rate of COPD. This study utilizes 17 global databases, collecting data on 77 social and environmental factors, 85 lifestyle and dietary factors, 25 physiological indicators, and 28 diseases. By carrying out the ecological analysis of multiple datasets, we aim to explore the potential factors affecting COPD prevalence and death rate, providing a scientific basis for the development of global COPD prevention and control strategies.

Methods

Data sources and processing

Data source selection and inclusion criteria

The prevalence and death rate of COPD were derived from the Global Burden of Disease (GBD) study.¹⁹ COPD is a chronic lung disease that determined by airflow obstruction, including conditions like emphysema and chronic bronchitis. It is diagnosed using spirometry, with an FEV₁/FVC ratio <0.7 after bronchodilation according to the Global Initiative for Chronic Obstructive Lung Disease (GOLD). Data for COPD prevalence came from multiple sources, including literature reviews, hospital claims data (e.g., from the US), Burden of Obstructive Lung Disease Study (BOLD), and English Longitudinal Study of Aging. The data were split by age and sex when not directly reported, and adjusted for different case definitions. The study used meta-regression to correct biases and provided consistent estimates across varying data sources. Different definitions of COPD (e.g., GOLD vs. BOLD) can lead to discrepancies in estimates. The logit-transformation method adjusted alternative COPD case definitions to a reference standard by transforming overlapping data into logit space, calculating differences, pooling these via random effects meta-regression, and applying the pooled difference to standardize all data points.

Based on the data framework from the GBD study, we integrated datasets of influencing factors from various countries and regions. We utilized a comprehensive array of international databases to investigate the relationships between various factors and the prevalence of COPD across 201 countries from 1990 to 2021, and the death rate from 1980 to 2021. This research incorporated 17 data sources, which provided insights across four primary domains: socio-economic and environmental factors, physiological metrics, lifestyle and diets, and disease prevalence and mortality. Data on socio-economic factors such as Medical Expenditure, Education Year, Human Development Index, Urbanization, Global Hunger Index, and Unemployment Rate were obtained from the World Health Organization (WHO),²⁰ World Bank,²¹ United Nations Development Programme (UNDP),²² United Nations Department of Economic and Social Affairs Population Dynamics,²³ Concern Worldwide and Welthungerhilfe (CWW),²⁴ and International Labor Organization.²⁵ Environmental factors, including temperature, ambient air pollution, and Universal Thermal Climate Index, were sourced from the University of East Anglia Climatic Research Unit Timeseries (CRU-TS),²⁶ Copernicus Atmosphere Monitoring Service (CAMS),²⁷ Copernicus Climate Change Service (C3S),²⁸ and Yale Center for Environmental Law & Policy.²⁹ Physiological Metrics such as fasting blood glucose, body mass index, child under weight, and IQ Score were obtained from the WHO, Noncommunicable Diseases Risk Factor Collaboration (NCD-RisC),³⁰ United Nations Children's Fund (UNICEF),³¹ and R Henss – Qeios 2023.³² Data on life style, for example, smoking, alcohol consumption, time spent in sleep, and dietary habits were acquired from the WHO, Food and Agriculture Organization of the United Nations (FAO),³³ Organization for Economic Co-operation and Development (OECD),³⁴ Penn World Table (PWT) 10.01,³⁵ and UNICEF. Disease Prevalence and Mortality of other disease were derived from the GBD Study. We retrieved a list of publicly available databases, including data from international organizations and datasets published alongside articles by research teams. After evaluating their suitability for consolidation into a country-level format, we included 17 databases, with details provided in Supplement Section 2.

As not all sources covered the same period or countries, all available data for the relevant years and countries were included for each variable. To ensure consistency and maximize data utilization, analyses were limited to periods and regions where data overlapped. In cases where multiple sources provided different values or units for the same or similar variables, all such data were incorporated. Unique variable names were assigned to each dataset to distinguish them. Comprehensive descriptions of each variable, including temporal and spatial coverage, definitions, and data sources, are presented in Supplement Section 3. In the context of our study, since we analyzed the relationship between exposures and COPD using regional-level data aggregated across countries, there is an inherent risk of ecological fallacy when interpreting our findings.

Data processing techniques

Within our pan-database synthesis system, prevalence and death rate estimates must be age-standardized for comparability, whereas databases of potential influencing factors do not require standardization across sources. The analytical framework employs methods that are inherently robust to variations in measurement units and data scales. The system utilizes correlation coefficients (both Pearson and Spearman) which are standardized metrics that remain invariant to linear transformations of the original data. When examining associations between variables—such as smoking and COPD—our methodology focuses on relative patterns and relationships rather than requiring uniform units or scales. By scaling effect sizes by interquartile ranges (β_IQR), we normalize the associations of different variables to enable meaningful comparisons regardless of their original units. This is particularly advantageous when integrating data from 17 different international databases, where attempting to harmonize heterogeneous sources could potentially introduce biases or distort the original information.

P-values were adjusted for multiple comparisons using the Benjamini–Hochberg procedure to control the false discovery rate (FDR), which was implemented via the MATLAB “mafdr” function. MATLAB R2024a, Python, and R Studio were used to process the data.

Calculation of prevalence rate

The GBD reports provide data on point prevalence (PoiP), defined as the percentage of people with COPD as of December 31st each year. In this study, point prevalence was converted to period prevalence (PerP), which reflects the percentage of people with COPD at any time throughout the year. Period prevalence was calculated as follows:

$$\mathbf{PerP}\left(t\right)=\mathbf{PoiP}\left(t-1\right)+\mathbf{Inc}\left(t\right)$$

where t represents the current year, PerP(t) is the period prevalence for year t, PoiP(t−1) is the point prevalence from the previous year, and Inc(t) is the incidence during year t.

For 1990, period prevalence was estimated under the assumption that there was no recovery from the disease:

$$\mathbf{PerP}\left(1990\right)=\mathbf{PoiP}\left(1990\right)+\mathbf{Deaths}\left(1990\right)$$

The age-standardized rates estimates were calculated using the direct method of standardization and were weighted using the GBD 2021 world standard population.^36,37

Population density-weighted averaging

For datasets presented in spatial map formats (e.g., temperature, pollution levels), population density-weighted averaging was employed to assign representative values to each country, taking into account the population distribution. The population density-weighted average T for a variable t(x, y) was calculated as follow:

$$T_{country}=\frac{\underset{country}{\iint }t\left(x,y\right)·\rho \left(x,y\right)dxdy}{\underset{country}{\iint }\rho \left(x,y\right)dxdy}$$

where $\rho \left(x,y\right)$ represents the population density at location (x, y).

Adjustment for confounding factors

Generalized linear models (GLMs) were used to assess the associations between potential factors and COPD outcomes (prevalence and mortality), adjusting for confounding variables to enhance accuracy. The confounders list for each factor are provided in Supplement Section 4. The GLM was employed to estimate the effects of confounders on the outcome variable, and these effects were subsequently subtracted from observed outcomes to obtain confounder-adjusted values.

Analytical methods

Trend analysis of COPD prevalence and death rates

To assess the trends in global prevalence and mortality rates of COPD, this study utilized Joinpoint regression analysis (Supplement Section 5). Joinpoint regression software (version 5.1.0, available at https://surveillance.cancer.gov/joinpoint/) was employed to calculate the annual percentage change (APC). Trends were classified as upward (APC > 0), downward (APC < 0), or stable (95% confidence interval including 0).

Generalized linear models for effect estimation and ranking

The GLMs were applied to assess associations of influencing factors with COPD prevalence and death rate, with adjustments made for confounding variables. Considering that the prevalence and death rate are variables ranging from 0 to 1, a beta distribution was selected as the link function. The GLM included both the outcome variable and confounding variables, allowing for the estimation of the independent effect of each risk factor. Additionally, relative variances were included as the weighting factors in each regression to mitigate the impact of uncertainties on the results. To standardize effect sizes across different units of measurement, beta coefficients (β) derived from the GLM were scaled by the interquartile range (IQR) of each independent variable, referred to as scaled beta coefficients (β_IQR). Confidence intervals and P-values were calculated to evaluate the statistical significance of these associations.

Both cumulative and instantaneous exposures were modeled to evaluate their impact on COPD outcomes. The cumulative exposure analysis, designed to assess sustained, long-term effects, calculated the exposure to each independent variable for a given country and year as its mean value over a 10-year rolling window ending in that specific year of outcome assessment. The association between this 10-year average exposure and the contemporaneous COPD outcome (prevalence or death rate) was then assessed using the resulting panel data. For the instantaneous exposure analysis, the association between the annual value of each risk factor and the COPD outcome in the corresponding year was assessed across the study period.

Correlation analysis and multiple testing corrections

Pearson and Spearman correlation coefficients were respectively calculated to assess linear and rank-based correlation between impact factors and COPD prevalence. The weaker of the two coefficients (Pearson or Spearman) was retained to provide a conservative estimate of the correlation. A significance threshold (P_FDR < 0.05) and an absolute correlation coefficient threshold (|r| > 0.2) were used to filter features to ensure notable correlations:

$$r_{\alpha \beta }^{Pearson}={\sum }_i\frac{\left({\alpha }_i-\overline{\alpha }\right)\left({\beta }_i-\overline{\beta }\right)}{\left(n-1\right)S_{\alpha }{S}_{\beta }}$$

$$r_{\alpha \beta }^{Spearman}=1-\frac{6{\sum }_i{\left(R\left[{\alpha }_i\right]-R\left[{\beta }_i\right]\right)}^2}{n\left(n^2-1\right)}$$

$$r_{\alpha \beta }=\{\begin{array}{c}{r}_{\alpha \beta }^{Pearson},\text{if}\left|r_{\alpha \beta }^{Pearson}\right|\le |r_{\alpha \beta }^{Spearman}|\\ r_{\alpha \beta }^{Spearman},\text{if}\left|r_{\alpha \beta }^{Pearson}\right|>|r_{\alpha \beta }^{Spearman}|\end{array}$$

Where $r_{\alpha \beta }$ is the correlation coefficient between two variables α and β. α_i and β_i denote each observation, R[αi] and R[βi] correspond to the rank of each observation.

Nonclassical multidimensional scaling for visualization

To visualize complex spatial relationships among variables, correlation coefficients (r values) were employed to ascertain distances, thereby constructing a dissimilarity matrix. Nonclassical multidimensional scaling (MDS) was employed to convert the high-dimensional dissimilarity matrix into a two-dimensional representation, using Kruskal's normalized stress-1 criterion:

$$d_{\alpha \beta }=\sqrt{1-{r_{\alpha \beta }}^2}$$

$$Y=MDS\left(d_{\alpha \beta },2\right)$$

$$Stress=\sqrt{\frac{{\sum }_{\alpha <\beta }{\left(d_{\alpha \beta }-{\hat{d}}_{\alpha \beta }\right)}^2}{{\sum }_{\alpha <\beta }{d_{\alpha \beta }}^2}}$$

Where $r_{\alpha \beta }$ is the correlation coefficient between two variables. $d_{\alpha \beta }$ is the dissimilarity coefficient between two variables. ${\hat{d}}_{\alpha \beta }$ Euclidean distance between points α and β in the reduced space. The MDS algorithm minimized stress, which quantifies the discrepancy between the original dissimilarities and the fitted distances, to achieve an optimal representation of high-dimensional data in a two-dimensional graph. Relative positions of nodes in the graph were ascertained based on their relationships with all other nodes. Separate MDS analyses were conducted on both unadjusted and confounder-adjusted data to assess the impact of confounder adjustment on the relationships among variables. Edges were added to depict pairs exhibiting correlation coefficients exceeding predefined thresholds (|r| > 0.2 for confounder-adjusted and |r| > 0.25 for unadjusted) between risk factors and disease characteristics.

Role of the funding source

The funder had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding authors and first authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

As this study utilized publicly available, de-identified data from established databases, ethics approval was not required in our institution (the First Affiliated Hospital of Guangzhou Medical University). No informed consent was necessary as the analysis was based on aggregate, anonymized data.

Results

Global burden of COPD from 1990 to 2021

The age-standardized prevalence and mortality rates of COPD vary significantly across different countries (Fig. 1). The age-standardized prevalence of COPD ranges from 1.00% to 3.71%, while the mortality rates vary between 0.0027% and 0.23%. Over the past 30 years, the prevalence and death rates of COPD have changed across countries (Fig. 1 and Supplement Section 5). In 1990, the highest age-standardized prevalence was observed in North Korea (3.46%), while the lowest was in Cape Verde (1.05%). By 2021, United States had the highest prevalence (3.71%), and Singapore had the lowest (1.01%). During this 30-year period, Singapore experienced the largest decrease in prevalence, dropping by 36.3%, whereas Saudi Arabia saw the largest increase, rising by 36.1%. In 1990, the highest age-standardized death rate was observed in China (0.23%), while the lowest was in Antigua and Barbuda (0.0055%). By 2021, Papua New Guinea had the highest death rate (0.16%), and Kuwait had the lowest (0.0027%). During this 30-year period, Singapore experienced the largest decrease in death rate, dropping by 85.46%, whereas Norway saw the largest increase, rising by 101%.

Fig. 1.

Prevalence and death rate of chronic obstructive pulmonary disease in 1990 and 2021. (A) Prevalence in 1990 and 2021 by country, (B) Prevalence and Death Rate in 1990 and 2021 by sex, (C) Prevalence and Death Rate in 1990 and 2021 by region. Regional color code: Central Asia; East Asia; Eastern Europe; North Africa and Middle East; North America; Oceania; South America; South Asia; Southeast Asia; Sub-Saharan Africa; Western Europe. Country codes see Supplement Section 4.

Fig. 1B shows that both the prevalence and mortality rates of COPD increase with age. Prevalence of COPD was similar between males and females aged 30–50 years (e.g., in 2021, the prevalence was 0.72% for males aged 30–34 and 0.79% for females in the same age group). However, the prevalence of COPD was higher in males over the age of 50 (e.g., in 2021, the prevalence was 28.0% for males aged 80–84 and 25.8% for females in the same age group). The mortality rates for men were significantly higher than for women across all age groups (e.g., in 2021, the death rate was 1.04% for men and 0.62% for women at the age of 85). Over the past 30 years, the prevalence for both men and women has remained relatively stable (e.g., from 2.84% to 2.80% for males and from 2.69% to 2.66% for females), while the death rates have significantly decreased (e.g., from 0.098% to 0.060% for males and from 0.054% to 0.034% for females).

The significant changes in the prevalence and death rate of COPD were also observed across continents over the past 30 years (Fig. 1C). From 1990 to 2021, the age-standardized prevalence of COPD remained relatively stable across multiple continents. For example, in Western Europe, it slightly decreased from 2.86% to 2.74%. In contrast, the age-standardized mortality rate for COPD showed a downward trend in several regions. For instance, in East Asia, the death rate was 0.094% in 1990, but it had dropped to 0.044% by 2021.

Potential influencing factors from multi-database

Country-level information was collected on 77 social and environmental factors, 85 lifestyle and dietary factors, 25 physiological indicators, and 28 diseases (see Supplement Section 3). Fig. 2 selectively presents several representative influencing factors, including tobacco use, secondhand smoke, atmospheric methane, ozone, biomass cooking, temperature, vapor pressure, and the adjusted Human Development Index (HDI),³⁸ showing the levels in 1990 and the trends since then. For example, in 1990, tobacco use was relatively high in both Western Europe and Southeast Asia. Over the following 30 years, tobacco use in Western Europe decreased, while in Southeast Asia, it continued to rise. Interestingly, the geographic distribution of tobacco-use and secondhand smoke paralleled the prevalence of COPD in several regions. Similarly, the distribution of biomass cooking was aligned with COPD death rates in various areas. Additionally, temperature, vapor pressure, methane levels, and adjusted HDI have shown a consistent upward trend over the past 30 years.

Fig. 2.

Influencing factors in 2021 and rate of change from 1990 to 2021 for chronic obstructive pulmonary disease. Epidemiological data of COPD: COPD 2021 prevalence, COPD 2021 death rate; Social Economics and Environment: Methane, Ambient O₃ Pollution, Temperature, Vapor Pressure, Inequality Adjusted HDI; Lifestyle and Ethnicity: Tobacco Use, Secondhand Smoking, Biomass Cooking.

Regression analysis of influencing factors on COPD

Fig. 3 shows the forest plot visualizing the relationship between potential influencing factors and COPD prevalence and death rate. The scaled β value (β_IQR) represents the percentage change in prevalence or death rate for each interquartile range (IQR) increase in exposure factors.

Fig. 3.

Forest plots of influencing factors for COPD prevalence and death rate.(A) Forest plot of influencing factors for COPD prevalence, (B) Forest plot of influencing factors for COPD death rate. βIQR indicates the change in prevalence or death rate corresponding to an increase in the interquartile range of the influencing factor. Red dots represent cumulative exposure of risk factors. Blue dots represent instantaneous exposure of risk factors. Only variables showing a strong association with COPD prevalence (P_FDR < 0.05) or COPD death rate (P_FDR < 0.05) are presented in the forest plot.

Cumulative exposure regression analysis showed that data from multiple sources consistently indicated that tobacco is the leading factor contributing to increased COPD prevalence. Specifically, for each interquartile range (IQR) increase in tobacco consumption, the following increases in COPD prevalence were observed: 0.39% (95% CI: 0.37–0.41) from tobacco use data in the GBD, representing a 16.7% relative increase compared to the global prevalence; 0.37% (95% CI: 0.35–0.39) from secondhand smoke in GBD (15.9% relative increase); 0.35% (95% CI: 0.33–0.37) from smoking reported by the GBD (15.0% relative increase); 0.35% (95% CI: 0.33–0.37) from tobacco use data in WHO (15.0% relative increase); 0.31% (95% CI: 0.29–0.34) from tobacco smoking in WHO (13.3% relative increase); and 0.29% (95% CI: 0.26–0.31) from cigarette smoking in WHO (12.4% relative increase). Air pollution, including ozone, methane, carbon monoxide (CO), and organic aerosols, was also a significant contributor to increased COPD prevalence (e.g., for each IQR increase in ozone, the COPD prevalence increased by 0.17 percentage points (95% CI: 0.16–0.19), representing a 7.39% relative increase compared to the global prevalence). The biomass cooking associated with the increased prevalence, while the gas cooking associated with the reduced prevalence. A wide range of climatic conditions, such as vapor pressure, solar radiation, wind speed, temperature, frequency of frost days, surface pressure, and the universal thermal climate index, were all consistently associated with the prevalence of COPD. Several factors influenced the COPD death rate. Tobacco use was a key contributor; for each interquartile range (IQR) increase in tobacco use, the death rate rose by 0.017% (95% CI: 0.016–0.018, representing a 51.52% relative increase compared to the global death rate). Ambient air pollution also played a role, with the death rate increasing by 0.019% (95% CI: 0.017–0.020, representing a 57.58% relative increase) per IQR increase in particulate matter. Social factors, including the³⁹ Global Hunger Index (GHI),⁴⁰ Gender Inequality Index (GII), years of education, urbanization rate and so on, were associated with higher death rates. Climatic conditions, measured by the universal thermal climate index, affected outcomes as well. Additionally, sports time—the average daily time spent on sports and physical activities—emerged as an influencing factor.

The results from instant exposure were consistent with those from cumulative exposure. Similar data from different databases, such as tobacco use and cigarette smoking from the WHO database and tobacco, smoking, and secondhand smoke from the GBD database, confirmed the same associations with COPD prevalence and death rate.

Correlation analysis between influencing factors and COPD

Fig. 4 presents heatmaps and scatter plots that visualize the correlations between various influencing factors and the prevalence, death rate, and DALYs associated with COPD, as well as differences across gender and age subgroups. In the Social Economics and Environment category, we identified significant correlations between factors such as temperature, vapor pressure, solar radiation, potential evapotranspiration, frost day frequency, methane levels, and residential radon with COPD prevalence (|r|> 0.25, P_FDR < 0.05). The strongest correlation was observed with solar radiation (r = −0.280). We also found that urbanization, the Socio-demographic Index (SDI),⁴¹ the Human Development Index (HDI), GHI, GII, years of education, and particulate matter pollution were significantly associated with the COPD death rate (|r| > 0.40, P_FDR < 0.05). Notably, the Social Economics and Environment factors exhibited a stronger correlation with COPD mortality than with prevalence. The strongest correlation was identified with the Inequality-Adjusted HDI (r = −0.470). The correlation was observed in both male and female subgroups and across multiple age groups. Notably, the correlations between air pollution and COPD prevalence and death rates were stronger in younger populations. Several scatter plots were utilized to illustrate these significant correlations. For instance, when examining years of education, we found that the COPD death rate tends to decrease as the average education level in a country increases across all continents. The overall regression coefficient (β_IQR) is −0.015, with an r value of −0.403 and P_FDR < 0.05.

Fig. 4.

Correlation heatmaps and scatter plots of chronic obstructive pulmonary disease with different categories factors. (A) Correlation Heatmap of COPD with Social Economics and Environment Factors, (B) Relationships between Chronic Obstructive Pulmonary Disease and Social Economics and Environment Factors, (BⅠ) Global hunger index vs. Confounder adjusted COPD prevalence, (BⅡ) Global hunger index vs. Confounder adjusted COPD death, (BⅢ) Vapor pressure vs. Confounder adjusted COPD prevalence, (BⅣ) Education year vs. Confounder adjusted COPD death, (BⅤ) PM pollution vs. Confounder adjusted COPD death. (C) Correlation Heatmap of COPD with Lifestyle and Diets Factors, (D) Relationships between Chronic Obstructive Pulmonary Disease and Lifestyle and Diets Factors, (DⅠ) Cigarette smoking vs. Confounder adjusted COPD prevalence, (DⅡ) Cigarette smoking vs. Confounder adjusted COPD death, (DⅢ) Secondhand smoke vs. Confounder adjusted COPD prevalence, (DⅣ) Secondhand smoke vs. Confounder adjusted COPD death. (E) Correlation Heatmap of COPD with Physiological Metrics Factors. (F) Relationship between Confounder Adjusted COPD Death and Fasting Blood Glucose. Only variables showing a strong correlation with COPD (r > 0.2, P_FDR < 0.05) are presented with colors in the heatmaps.

In terms of Lifestyle and Diets category, significant negative correlations were observed of tobacco (i.e., tobacco, tobacco-use, smoking, tobacco smoking, cigarette smoking, and secondhand smoke) and biomass cooking with COPD prevalence (|r| > 0.25, P_FDR < 0.05). The strongest correlation was identified with the tobacco-use (r = 0.451). Tobacco was also strongly correlated with the COPD mortality rate. Additionally, we observed an inverse relationship of sports time and work time with the COPD death rate. These correlations were also confirmed in diverse sex and age groups. The correlation of sports time with COPD death rate was stronger in younger populations.

In the Physiological Metrics category, a significant correlation was found between fasting blood glucose levels and COPD death rates across all age groups. Scatter plots illustrate the positive correlation between fasting blood glucose levels and COPD mortality within multiple continents.

See unadjusted heatmaps, scatter plots, and network graphs in Supplement Sections 8–16.

Correlation analysis between COPD and other diseases

In the network graph, shorter distances suggest statistical similarity between data characteristics. For example, the proximity of prevalence rates across different age groups suggests statistical similarity. The lines represent correlations between nodes, such as the positive correlations observed of COPD with chronic kidney disease, other chronic respiratory diseases, gout, and stroke (Fig. 5, see unadjusted results in Supplement Section 5).

Fig. 5.

Adjusted network graph for COPD and other diseases.(A) Correlation heatmap of COPD with other diseases, (B) Network diagram of COPD and other diseases. Network diagram depicting the links between COPD and other diseases. Strong positive correlations are indicated by red edges (r > 0.4, P_FDR < 0.001) and strong negative correlations by blue edges (r ≤ −0.4, P_FDR < 0.001). The depth of color and thickness of the edges represent the correlation magnitude. Factors from other categories are grayed out.

Discussion

We hypothesize that a range of social, environmental, lifestyle, and health factors may jointly influence the prevalence and death rate of COPD. This ecological study aimed at integrating 17 global databases to conduct a comprehensive analysis of the potential factors influencing COPD, revealing several important findings. First, the pan-database study shows a significant association between COPD prevalence and factors that include not only tobacco consumption but also air pollution, biomass cooking, and climate conditions, providing new insights into COPD prevention and control. Second, the findings highlighted that social factors, such as the GHI and the GII, also significantly impact death rates, emphasizing the critical role of socioeconomic environments in disease burden. Furthermore, chronic diseases such as chronic kidney disease, other chronic respiratory diseases, gout, and stroke are positively correlated with COPD, underscoring the importance of comorbidity management. Compared to small-scale studies that may focus on specific risk factors or populations, our research offers a broader and more generalizable perspective that leans towards public health management. This study integrated 17 global databases and employed ecological analysis to provide robust evidence for the association between COPD prevalence and multi-aspects factors, thereby offering a more universally applicable scientific foundation for global public health management.

The significance of this study lies in its interesting comprehensive analysis of numerous global databases, filling gaps left by previous research. While existing studies have explored risk factors for COPD, most have been limited to single database, failing to account for the interactions between multiple potential factors. By integrating 17 global databases, this study offers a systematic global perspective, enabling a deeper understanding of the multidimensional influencing factors for COPD. This not only provides scientific evidence for prevention and intervention but also serves as a reference for formulating global public health policies. It underscores the importance and feasibility of adopting tailored, comprehensive measures worldwide to reduce the COPD burden. Moreover, the results of this study pave the way for further research into COPD mechanisms, identifying novel factors such as radon, methane, sea salt aerosols, thermal radiation, and vapor pressure, thereby guiding future mechanistic studies. In summary, the findings of this study provide valuable guidance for future clinical practice, public health strategies, and mechanistic research.

Smoking have been clearly identified as a major risk factor for COPD in many studies.⁴ This study further confirms this view by integrating data from GBD and WHO with regression and correlation analysis, emphasizing the most significant correlation of tobacco-use (including smoking and secondhand smoke) with COPD prevalence and death rate. Asian countries have higher levels of smoking, secondhand smoke exposure, use of biomass fuels for cooking (see the upper right corner of Fig. 4D I–IV), and PM pollution (see Fig. 4B V). These factors may help explain the higher COPD prevalence and mortality rates in the region, which is consistent with findings from numerous published studies.42, 43, 44, 45, 46

In addition to these environmental and social factors, lifestyle habits such as cooking and physical activity levels play a significant role in the development of COPD. We observed a notable relationship between biomass cooking and COPD prevalence, suggesting that cooking methods using traditional biofuels may increase exposure to harmful indoor air pollutants, further exacerbating respiratory health. Additionally, physical activity was found to be inversely related to COPD mortality, supporting the notion that exercise may improve lung function and overall health outcomes for individuals with COPD. This study underscores the importance of considering lifestyle and cultural factors in understanding the global burden of COPD. More attention should be given to promoting healthier cooking methods, balanced diets, and physical activity as part of COPD prevention and management strategies.

Climate change is a global concern, and this study also highlights the close connection between climate and COPD. For example, low temperatures and high humidity have negative effects on COPD, a finding that is partially supported by existing literature. For example, observations by Mu et al.⁴⁷ on 82 patients indicate that low temperatures and high humidity are risk factors for worsening COPD symptoms. We also identified some previously unreported associations, such as the negative correlation between vapor pressure, solar radiation, and COPD prevalence.

Hunger, shortages of healthcare resources, low education levels, and gender inequality are pressing issues in developing countries. Our study also found that these factors have a negative impact on COPD, particularly on mortality rates. As shown in Fig. 4B I and II, an increase in the GHI correlates positively with both COPD prevalence and mortality in Asia, this relationship also observed in some cohort studies.^48,49 Education level has been shown to be linked to disease burden,^50,51 and as seen in Fig. 4B IV, lower education years in sub-Saharan Africa are associated with higher COPD mortality rates. Gender inequality profoundly impacts health globally, especially in low- and middle-income countries, where women often face a higher disease burden.^52,53 Gender inequality not only affects women's access to healthcare and resources but may also exacerbate COPD risk by influencing lifestyle, nutrition, and environmental exposures. Therefore, public health policies should consider gender factors to ensure women's access to health promotion and disease prevention resources.

Regarding comorbidities, the study shows positive correlations between COPD and chronic kidney disease, other chronic respiratory diseases, gout, and stroke, echoing clinical observations that COPD patients often suffer from multiple chronic conditions.54, 55, 56 This comorbidity relationship highlights the need to holistically consider the overall health of COPD patients in clinical management. In summary, this study provides a new perspective on the multidimensional risk factors for COPD, corroborating existing literature.

Although this study presents many important findings, there are certain limitations. First, differences in ascertainment were an important and inevitable factors that may contribute to variations in the results. Differences in case diagnosis and data collection practices across countries and regions can lead to biases, affecting the robustness of the findings. The uncertainties and measurement errors inherent in the data sources, such as survey or administrative data, could not be fully addressed, which could affect the accuracy of the reported outcomes. Second, our ecological study, based on population-level aggregated data, cannot determine causal relationships or capture long-term effects. Therefore, caution is needed when inferring individual-level associations, and unmeasured confounders—such as environmental exposures (e.g., industrial emissions), ethnic or genetic background (e.g., α1-antitrypsin deficiency), or policies and interventions (e.g., the use of bronchodilators and community-based COPD screening and management)—may influence the observed results. Third, although multiple potential factors have been included, some unknown or emerging factors may have been overlooked, which could affect the study's comprehensiveness. Fourth, it is important to acknowledge that much of the risk data presented in this study represents broad estimates and are generalizable. However, these estimates may not capture the full complexity and variability of individual patient experiences, which could lead to an ecological fallacy. Fifth, although the findings of this study exhibit broad generalizability, their applicability may be influenced by differences across countries and regions. For instance, data quality varies with the sophistication of healthcare systems, with more reliable data typically available in developed countries, whereas developing nations may face issues of missing or inaccurate data. Consequently, caution is warranted when applying these findings regionally. Sixth, the current version of the GBD database does not provide specific data on multi-comorbidities, which limits our ability to directly incorporate multi-comorbidities phenomena into our analysis. Additionally, to compare disease data across countries, this study used age-standardized prevalence and death rates, which were influenced by demographic structure. Developing countries with a young population structure always have high age-standardized death rates, but their actual death rates might be less severe. Future research should prioritize longitudinal studies to validate these findings, further investigate potential causal relationships, and explore additional unmeasured influencing factors, thereby providing more definitive evidence on the temporal sequence of observed associations.

Based on the findings of this study, several clinical implications emerge. First, the strong associations identified between smoking, air pollution, biomass cooking, and COPD prevalence and mortality highlight the urgent need for clinical strategies aimed at reducing these modifiable risk factors. In addition, given the significant impact of comorbidities such as chronic kidney disease and other respiratory diseases, comprehensive management of COPD patients should include monitoring and treatment for these conditions to improve overall health outcomes. This study also emphasizes the importance of lifestyle interventions, such as promoting physical activity and healthier cooking methods, which can significantly reduce the burden of COPD. Furthermore, the disparities revealed in COPD prevalence and mortality across regions suggest that tailored public health policies are crucial to addressing the specific needs of different populations.

In conclusion, this study conducted a pan-database analysis of global data to thoroughly investigate the multidimensional risk factors of COPD, revealing the critical roles of tobacco consumption, biomass cooking, air pollution, climate factors, and social factors in the disease's prevalence and death rates. These findings provide scientific evidence for the prevention and treatment of COPD and offer valuable references for the formulation of public health policies.

Contributors

Zihui Wang, Wenhan Cao, Zhixuan You, Shaoqiang Li, Mingshan Xue, Guanghui Dong, Baoqing Sun, Nanshan Zhong, and Zhangkai J. Cheng conceptualized and administrated the project. Zihui Wang, Zhixuan You, Wenhan Cao, Shaoqiang Li, Mingshan Xue, and Zhangkai J. Cheng directly accessed and verified the underlying data in all research articles. Zihui Wang and Zhangkai J. Cheng supervised the project. Nanshan Zhong and Zhangkai J. Cheng have given the funding support. All authors drafted and revised the manuscript, discussed the results, provided critical feedback, and helped shape the research, analysis, and manuscript. Zihui Wang and Zhangkai J. Cheng were responsible for the decision to submit the manuscript.

Data sharing statement

Data of 17 global databases are available in the corresponding websites (Supplement Section 3).

Editor note

The Lancet Group takes a neutral position with respect to territorial claims in published maps and institutional affiliations.

Declaration of interests

No potential conflict of interest was reported by the authors.