Study population and data collection

The GBD 2021 database encompasses 369 diseases and injuries across 204 countries and territories from 1990 to 2021, employing rigorous methodologies and diverse data sources to ensure accurate and comprehensive estimates. ADOD diagnoses were determined using ICD-10 codes F00, F01, F02, F03, G30, and G31, as outlined by the GBD^7,8.

Estimation of disease burden

The study utilized four metrics – death rates, DALYs, YLLs, and YLDs – along with their quantified forms, including case numbers and age-standardized rates (ASR), to illustrate the disease burden associated with smoking-attributable ADOD. This study identifies eight indicators of disease burden: the number of deaths, the age-standardized death rate (ASDR), the number of DALYs, age-standardized DALYs rate (ASDAR), number of YLLs, age-standardized YLLs rate, number of YLDs, and age-standardized YLDs rate. DALYs, calculated by summing YLLs and YLDs, provide a comprehensive measure of health burden by reflecting both fatal and non-fatal health losses. YLLs are calculated by multiplying disease-specific death counts by the standard expected remaining lifespan, whereas YLDs represent the years lived with disability caused by the disease⁹. Estimates are presented as case numbers and ASR per 100000 population, each with 95% uncertainty intervals (UIs).

Sociodemographic index

The sociodemographic index (SDI) serves as a comprehensive measure of economic and developmental status, closely linked to health outcomes. The measure incorporates per capita income, average schooling, and birth rate figures to present a comprehensive evaluation of development levels in all regions covered by the GBD study. The SDI ranges from 0 to 1, classifying the world into five regions: low, low-middle, middle, high-middle, and high SDI¹⁰.

Decomposition and frontier analysis

We employed decomposition analysis to disentangle and quantify the relative contributions of three principal drivers to the observed changes in the absolute number of deaths, DALYs, YLLs, and YLDs attributable to smoking-related ADOD from 1990 to 2021. These drivers include: 1) Population growth, which pertains to changes solely attributable to increases in total population size; 2) Population aging, which involves changes resulting from shifts in the age distribution of the population, specifically an increase in the proportion of older individuals who are at higher risk; and 3) Epidemiological change, which encompasses changes in the underlying age-specific risk or disease rates, reflecting the cumulative impact of factors such as variations in smoking prevalence, exposure intensity, diagnostic practices, and treatment effectiveness over time. This analytical approach enables us to attribute the total observed increase or decrease in burden to each of these distinct demographic and epidemiological factors¹¹. Furthermore, we utilized frontier analysis to assess the efficiency of various countries in managing the smoking-related ASR of deaths, DALYs, YLLs, and YLDs in relation to their sociodemographic development, as indicated by the SDI. This method establishes a statistical ‘frontier’ that represents the optimal (lowest) ASR achievable for a given SDI level, derived from global data spanning 1990 to 2021. The frontier effectively delineates the minimum disease burden theoretically attainable for a country at its specific stage of development. Subsequently, we computed the effective difference, or distance to the frontier, for each country in 2021, defined as the absolute difference between the observed ASR and the corresponding frontier value at its SDI. A greater effective difference suggests a higher potential for improvement in mitigating the smoking-related burden of disease compared to the best-performing countries at similar development levels. This analysis facilitates the identification of countries that are underperforming relative to their SDI counterparts and highlights areas where targeted interventions could achieve the most substantial improvements¹².

Forecasting the worldwide impact of smoking-related ADOD

Autoregressive integrated moving average (ARIMA) and exponential smoothing (ES) were utilized in R software to forecast the number and ASR of deaths, DALYs, YLLs, and YLDs from 2020 to 2050. The time series of smoking-attributable ADOD from 1990 to 2021 was analyzed using R software, utilizing data on deaths, DALYs, YLLs, and YLDs to investigate trend changes. The ARIMA model, which incorporates auto-regressive, integrated, and moving average components, was employed to examine and forecast the time-series data. The optimal model was selected using the standardized Bayesian Information Criterion (BIC), and a white noise test was performed to evaluate the model at a 0.05 significance level¹³. The ES model utilizes an exponentially weighted average method for time series forecasting, estimating future values by calculating the weighted average of previous observations¹⁴.

Statistical analysis

We initially analyzed the burden and temporal trends of smoking-related ADOD, considering variables such as sex, age, SDI, region, and country. Pearson correlation and linear regression analyses were employed to examine the relationship between ASDR, ASDAR, and SDI. The trend in disease burden from 1990 to 2021 was assessed using the estimated average percentage change (EAPC) of age-standardized rates, with a 95% confidence interval (CI) derived from a linear regression model. The assessment of driving factors and potential improvements was performed using decomposition and frontier analysis. Ultimately, predictions of the trend in smoking-related ADOD burden up to 2050 were made using time-series forecasting models such as the ARIMA and ES models, utilizing RStudio (version 4.1.1).

The global smoking-attributable burden of ADOD

The global burden of smoking-attributable ADOD in 1990 and 2021 is shown in Supplementary file Tables S1–S4. Between 1990 and 2021, the number of deaths related to ADOD increased from 32165 (95% UI: 7446–8932) to 67176 (95% UI: 15695–184665), and DALYs rose from 794915 (95% UI: 344378–1839709) to 1533214 (95% UI: 662723–349642), with overall increases of 108.84% and 92.88%, respectively. The number of YLLs increased from 528022 (95% UI: 122674–1511230) to 1017331 (95% UI: 241832–2876261), and YLDs increased from 266894 (95% UI: 162663–391752) to 515882 (95% UI: 313106–759417), representing overall increases of 93.29% and 92.67%. The ASR of deaths declined from 1.08 (95% UI: 0.25–3.03) to 0.84 (95% UI: 0.19–2.29), and DALYs decreased from 23.33 (95% UI: 9.99–54.46) to 18.36 (95% UI: 7.90–42.07), reflecting overall reductions of 22.22% and 21.30%, respectively. The ASR of YLLs fell from 15.76 (95% UI: 3.62–44.13) to 12.24 (95% UI: 2.90–34.28), and YLDs fell from 7.57 (95% UI: 4.64–11.22) to 6.12 (95% UI: 3.72–9.03), with overall reductions of 22.33% and 19.15% (Figures 1A–1D).

Figure 1

Analysis of global burden indicators for smoking-related Alzheimer’s disease and other dementias from 1990 to 2021. Trends are depicted using column bar graphs and line charts illustrate the age-standardized rates per 1000000 population: A) deaths; B) DALYs; C) YLDs; and D) YLLs

Global smoking-attributable ADOD burden by sex

There was a marked difference between sexes in smoking-attributable ADOD burden, with males experiencing approximately 2–3 times the burden of females. In 2021, the age-standardized death rate (ASDR) for males was 1.43 (95% UI: 0.34–4.05) and the age-standardized DALY rate (ASDAR) was 30.56 (95% UI: 12.72–71.5), significantly higher than females (ASDR=0.43; 95% UI: 0.10–1.14; ASDAR=9.01; 95% UI: 3.94–20.16). Deaths and DALYs were 46755 (95% UI: 11061–132714) and 1110023 (95% UI: 474038–2558503) for males, versus 20420 (95% UI: 4946–53998) and 423191 (95% UI: 185085–947536) for females. A comparable sex disparity existed for YLLs and YLDs (Supplementary file Tables S3 and S4). Both ASDR and ASDAR declined in both sexes, with estimated annual percentage changes (EAPCs) of -0.95 (95% CI: -0.99 – -0.91) and -0.88 (95% CI: -0.90 – -0.83), respectively. Declines were more pronounced in females (ASDR EAPC= -1.60; 95% CI: -1.68 – -1.52; ASDAR EAPC= -1.50; 95% CI: -1.56 – -1.45) than males (ASDR EAPC= -0.79; 95% CI: -0.85 – -0.73; ASDAR EAPC= -0.73; 95% CI: -0.80 – -0.60). By 2021, despite decreased age-standardized rates (ASR) of years of life lost (YLLs) and years lived with disability (YLDs) in both sexes, males had higher absolute numbers: YLLs were 737183 (95% UI: 177427–2115307) for males versus 280148 (95% UI: 65583–771699) for females, and YLDs were 372840 (95% UI: 228453–545265) versus 143042 (95% UI: 84582–215146). Epidemiological patterns of ASR for YLLs and YLDs showed similar trends (Supplementary file: Tables S1–S4, Figure S1).

Smoking-attributable ADOD burden by age

Among individuals aged 70–89 years, smoking exposure was the primary contributor to the ADOD burden. As age increased, there was an upward trend in deaths, DALYs, YLLs, and YLDs. From 1990 to 2021, both ASDR and ASDAR exhibited a downward trend across all age groups, with the most pronounced decline in the above 95 years group, reflected by an EAPC of -1.56 (95% CI: -1.63 – -1.49) for ASDR and -1.62 (95% CI: -1.69 – -1.55) for ASDAR. Subgroup analysis by sex revealed a higher smoking-attributable burden in males compared to females. For both males and females, the number of deaths and YLDs due to ADOD from smoking increased with age, peaking at 80–84 and 70–74 years for males, and at 85–89 and 80–84 years for females, before declining annually. Furthermore, the rates of deaths, DALYs, and YLLs for ADOD increased with age in both males and females, peaking in individuals aged >95 years. For both males and females, the YLD rate increased with age, reaching its highest point among individuals aged 90–94 years before declining (Supplementary file: Tables S1–S4, Figure S2).

Smoking-attributable ADOD burden by SDI

In 2021, the ASR of deaths, DALYs, YLLs, and YLDs were highest in high-middle SDI regions, with ASDR at 0.96 (95% UI: 0.22–2.6), ASDAR at 21.96 (95% UI: 9.6–49.15), ASR of YLLs at 14.41 (95% UI: 3.4–40), and ASR of YLDs at 7.55 (95% UI: 4.5–11.14). Conversely, these rates were lowest in low SDI regions. The middle SDI regions exhibited the highest DALYs, YLLs, and YLDs among the five groups, while the high SDI regions recorded the highest number of deaths. Between 1990 and 2021, the ASR of deaths, DALYs, YLLs, and YLDs declined across the five SDI regions, with the high SDI regions experiencing the most notable decrease. Conversely, the absolute numbers of deaths, DALYs, YLLs, and YLDs increased, especially in the middle SDI region (Supplementary file: Tables S1–S4, Figure S3).

This research analyzed the decomposition of original deaths, YLLs, YLDs, and DALYs for ADOD to investigate the effects of ageing, population growth, and epidemiological transitions on the ADOD burden. Globally, the DALYs and YLDs of both sexes for ADOD demonstrated an increasing trend. The decrease in deaths and YLLs was primarily attributed to the negative impact of epidemiological changes in females. Globally, the middle SDI region experienced the highest increase in DALYs and YLDs, primarily due to population growth, followed by ageing. Population growth contributed 120.62% and ageing 73.43% to the rise in disease burden. There was a negative contribution from epidemiological changes. The impact of epidemiological change differed across SDI regions, showing a notable increase of 142.84% in high SDI regions, while low, low-middle, middle, and high-middle SDI regions experienced decreases of 24.74%, 39.62%, 54.59%, and 89.65%, respectively. The adverse effect of epidemiological change on disease burden increases with higher SDI levels. Afterwards, we conducted a stratification analysis based on sex. Both DALYs and YLDs exhibited similar traits in males and females. Nevertheless, there was an increasing trend in deaths and YLLs among males. In the central SDI region, where the disease burden increased most significantly, the impact of male population growth was most notable (Supplementary file: Figure S4).

We conducted a frontier analysis using ASR of deaths, YLLs, YLDs, and DALYs alongside SDI levels from 1990 to 2021 to better understand potential improvements in the burden of ADOD relative to national development levels. Considering national development in the context of ASDR and ASDAR, the countries with the greatest potential for improvement include Egypt, Rwanda, Tunisia, Paraguay, Lebanon, Kiribati, Jordan, Iraq, Greece, Cambodia, Algeria, Albania, Denmark, and China. Countries with a low SDI (<0.5) include Somalia, Burkina Faso, Guinea-Bissau, and Benin. Despite their high SDI (>0.85), countries such as the Republic of Korea, Sweden, the United States, the Netherlands, and Denmark still possess significant potential for further progress relative to their development level. ASR of YLLs and YLDs, showed similar characteristics (Supplementary file: Figure S5).

Smoking-attributable ADOD burden by regional and national location

The 2021 disease burden of smoking-attributable ADOD and its trends were analyzed by geographical distribution, including GBD subcontinental regions and countries, with results visualized (Supplementary file Tables S1–S4). East Asia exhibited the highest ASR for deaths, DALYs, YLLs, and YLDs at 1.38 (95% UI: 0.32–3.87), 29.95 (95% UI: 12.67–68.1), 10.09 (95% UI: 6.08–15.1), and 19.86 (95% UI: 4.66–56.37), respectively. In contrast, Western Africa and Western Sub-Saharan Africa recorded the lowest age-standardized death rates (ASDR) at 0.14 (95% UI: 0.03–0.41), age-standardized DALYs at 2.89 (95% UI: 1.12–7.07), YLLs at 2.08 (95% UI: 0.49–6.14), and YLDs at 0.81 (95% UI: 0.49–1.2). Asia reported the highest figures for deaths, DALYs, YLLs, and YLDs, with values of 300.55 (95% UI 226.03–384.96), 958325 (95% UI: 412336–2196528), 635231 (95% UI: 151073–1802835), and 323094 (95% UI: 196254–476426), respectively (Supplementary file: Tables S1–S4, Figure S6).

From 1990 to 2021, most regions exhibited a slight upward trend in ASDR and ASDAR. Eastern Europe experienced the most significant rise in ASR of deaths and YLLs, with EAPCs of 1.23 (95% CI: 0.77–1.69) and 1.09 (95% CI: 0.57–1.6), respectively. East Asia experienced the highest rise in ASR of DALYs and YLDs, with EAPCs of 1.11 (95% CI: 0.62–1.6) and 1.46 (95% CI: 1.01–1.91), respectively. Southern Sub-Saharan Africa experienced the most significant reductions in ASR of deaths, DALYs, YLLs, and YLDs, with EAPCs of -2.15 (95% CI: -3.69 – -0.58), -2.07 (95% CI: -3.03 – -1.11), -2.2 (95% CI: -3.1 – -1.29), and -2.01 (95% CI: -3 – -1.01), respectively. The extensive evaluation of the EAPC for all indicators, followed by cluster analysis, indicated a notable decrease in the ADOD disease burden associated with smoking in five regions, a significant increase in two regions, a slight increase in 31 regions, and stability or a slight decrease in 10 regions ( Supplementary file: Figure S7).

In 2021, Lebanon recorded the highest ASR for deaths, DALYs, YLLs, and YLDs at 1.59 (95% UI: 0.39–4.44), 35.63 (95% UI: 15.39–80.46), 13.37 (95% UI: 8.11–20.29), and 22.26 (95% UI: 5.41–63.5), respectively (Supplementary file: Figure S8). From 1990 to 2021, the fastest increase in ASDAR was observed in Albania (EAPC=2.51; 95% CI: 1.99–3.03), and the fastest decrease was in Madagascar (EAPC= -3.08; 95% CI: -4.93–1.19). Between 1990 and 2021, Georgia experienced the most rapid increase in ASDR, with an EAPC of 2.82 (95% CI: 2.17–3.48), while Madagascar saw the steepest decline, with an EAPC of -3.34 (95% CI: -5.56 – -1.06). Similar results can also be found in the ASR of YLLs, and YLDs (Supplementary file: Tables S3 and S4, Figure S9).

Predictions of the smoking-attributable ADOD burden

The ARIMA and ES model predictions suggest a global increase in the number and ASR of deaths, DALYs, YLLs, and YLDs due to ADOD from 1990 to 2021. By 2050, projections indicate an upward trend in global ASDR, ASDAR, YLLs, and YLD, with the most notable rise among males, with an EAPC of 2.06 (95% CI: 1.34–2.79), 46.19 (95% CI: 29.76–62.62), 24.16 (95% CI: 17.94–30.38), and 23.15 (95% CI: 12.73–33.58), respectively, according to the ARIMA model (Supplementary file: Figure S10). The ARIMA model predicted that the ASDR and ASDAR for females would remain relatively stable, with a slight variation. This trend was similarly observed in the ES model (Supplementary file: Figure S11).

Limitations

This study has several limitations. The GBD study’s methodology has recognized limitations, such as biases and gaps, particularly in low- and middle-income countries with high smoking rates and inadequate healthcare resources. Secondly, the study did not analyze smoking dose (daily cigarettes), duration (age of initiation, years since cessation), and secondhand smoke exposure, which may have had an impact on the estimate of the burden of smoking-related ADOD. Finally, the long latency period of ADOD (up to decades) may lead to underestimation of smoking’s long-term effects due to insufficient follow-up in many studies.