Global drivers of heart failure attributable to atrial fibrillation/flutter: insights from Joinpoint regression, age-period-cohort analysis, and future projections from the GBD 2021 Study

Abstract

Background

Heart failure attributable to atrial fibrillation/flutter (HF-AF/AFL) represents a growing global public health challenge. However, comprehensive analyses of its long-term trends, the independent effects of age, period, and cohort, and future projections remain scarce. This study aimed to address these gaps by characterizing transitions in HF-AF/AFL burden from 1990 to 2021.

Methods

Based on the Global Burden of Disease Study (GBD) 2021, we estimated the global prevalence and years lived with disability (YLDs) for HF-AF/AFL. Joinpoint regression was used to analyze temporal trends from 1990 to 2021. An age-period-cohort model assessed independent effects of age, period, and birth cohort, and a Bayesian age-period-cohort (BAPC) approach projected disease burden to 2050. All estimates were stratified by age, sex, and sociodemographic index (SDI).

Results

The global burden of HF-AF/AFL increased significantly from 1990 to 2021, demonstrating a distinct socioeconomic gradient, with the highest burden observed in high-SDI regions. Although the absolute burden remained greater among females, the increase was more pronounced in males. Joinpoint regression identified a recent inflection point, marked by modest global declines in age-standardized prevalence (ASPR) and years lived with disability (ASYR) rates from 2018 to 2021; this downturn was particularly evident among females and in high-SDI regions. Age-period-cohort analysis confirmed an exponential increase in risk with age, a persistent rise in risk across successive periods, and elevated susceptibility in more recent birth cohorts. Projections indicate a continued rise in burden, with the number of prevalent cases forecast to reach approximately 1.44 million globally by 2050, corresponding to an ASPR of 15.04 per 100,000, and YLDs projected to rise to around 0.13 million, with an ASYR of 1.35 per 100,000. This upward trajectory was consistent across SDI strata, although future burdens exhibited regional heterogeneity.

Conclusions

The global burden of HF-AF/AFL is substantial and increasing. Our analysis projects a continued rise in this burden over the coming decades and identifies distinct risk patterns driven by age, period, and birth cohort. These findings underscore the necessity for targeted public health strategies to address this growing challenge.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12916-026-04628-y.

Background

Heart failure (HF) represents a significant global health burden driven by multiple etiological factors, such as coronary artery disease, arrhythmias, hypertension, and cardiomyopathies [1, 2]. The condition is associated with considerable morbidity and mortality, substantially impairing patients’ quality of life [3].

While stroke is the most recognized complication of atrial fibrillation and atrial flutter (AF/AFL), the widespread implementation of anticoagulant therapy and integrated care has substantially mitigated this risk [4, 5]. However, contemporary evidence demonstrates that AF/AFL is a critical trigger for HF, with a greater contribution to HF incidence than previously acknowledged [6]. Furthermore, AF/AFL and HF share a bidirectional pathophysiological relationship, which synergistically accelerates disease progression and leads to adverse clinical outcomes [7–9]. Although prior studies have outlined the foundational overview of the global burden of HF attributable to AF/AFL (HF-AF/AFL), they have not employed more advanced analytical frameworks to investigate this issue comprehensively [10].

To address this evidence gap, our study performs a secondary analysis of the Global Burden of Disease (GBD) data to delineate the long-term epidemiology of HF-AF/AFL. This analysis characterizes historical trends, identifies principal risk factors, and projects future disease burden. The findings provide an evidence base for public health strategy, informing the evaluation of past interventions, guidance for targeted prevention, and proactive resource allocation to advance evidence-based policy.

Methods

Study data

The Global Burden of Diseases, Injuries, and Risk Factors Study 2021 (GBD 2021) dataset served as the primary data source for this epidemiological analysis of HF-AF/AFL from 1990 to 2021. Coordinated by the Institute for Health Metrics and Evaluation (IHME) at the University of Washington, this comprehensive repository provides standardized estimates for 371 health conditions and 88 risk factors across 204 nations and territories [11].

It is important to note that within the GBD framework, HF is conceptualized and modeled as a nonfatal “impairment” or sequela of underlying causes, including AF/AFL. Consequently, the GBD provides estimates for the prevalence of HF and the corresponding years lived with disability (YLDs), but it does not generate estimates for mortality (years of life lost, YLLs) or disability-adjusted life years (DALYs) for HF itself. This inherent structure of the GBD data dictated the scope of our analysis, which focuses exclusively on the morbidity burden (prevalence and YLDs) of HF-AF/AFL. All prevalence, YLDs, and age-standardized rate (ASR) estimates, together with corresponding 95% uncertainty intervals (UIs), were extracted from the publicly accessible GBD 2021 Results dataset via the IHME GBD Results Tool (http://ghdx.healthdata.org/gbd-results-tool) [12].

Definition

HF and AF/AFL

Within the GBD 2021 framework, HF is defined as a functional impairment of cardiac filling or ejection due to structural or functional abnormalities from diverse etiologies [13]. Case identification is based on epidemiological data from studies using established diagnostic criteria, such as the Framingham or the European Society of Cardiology guidelines. Since 2016, the GBD case definition has aligned with Stage C and above of the Universal Definition and Classification of HF, thereby encompassing both symptomatic patients and those with compensated, asymptomatic status.

In this study, AF/AFL are specified as etiological causes of HF. AF/AFL are defined by standard electrocardiographic criteria: irregularly irregular RR intervals, absent distinct P waves, and atrial cycle length variability < 200 ms when measurable [14]. The attribution of HF to AF/AFL within the GBD study follows a systematic methodology that integrates evidence from literature reviews, analysis of death certificate and hospital record data across multiple countries, and formal expert consensus. Comprehensive methodological and definitional details were published previously [11, 15].

Years lived with disability (YLD)

YLD quantifies years of life lost due to the disability caused by an illness, calculated through multiplicative integration of condition-specific prevalence, disability weights (reflecting health loss severity), and demographic stratifiers (age, sex, location, year) [11].

Sociodemographic index (SDI)

The SDI is a composite indicator of development, ranging from 0 (least developed) to 1 (most developed), created for the GBD study [16]. It summarizes key social and economic determinants of health by combining three equally weighted components: lag-distributed income per capita, average educational attainment in the population aged 15 years and older, and the total fertility rate under age 25 (TFU25). Each component is scaled between 0 and 1 using the theoretical minimum and maximum values relevant to population health. For the GBD 2021 Study, 204 countries and territories were categorized into 5 SDI groups: high (> 0.81), high middle (0.70–0.81), middle (0.61–0.69), low middle (0.46–0.60), and low (< 0.46) [16, 17].

Age-standardized rate (ASR)

To account for varying age structures across populations and to enable unbiased comparisons, ASRs were calculated using the GBD reference population. This method adjusts for differences in population age distributions by applying the age-specific rates observed in a study population to a standard population structure. The calculation was performed using the direct standardization method, summarized by the following formula:

$$ASR=\frac{{\sum }_{i=1}^A{a}_i{w}_i}{{\sum }_{i=1}^A{w}_i}\times \text{1000,000}$$

where a_i represents the age-specific rate in the i^th age group and w_i denotes the number of persons in the corresponding i^th age group of the GBD standard population; Σ indicates the summation across all age groups. The resulting ASR represents the theoretical rate expected if the study population possessed the age structure of the standard population. This adjustment is essential for unbiased temporal and spatial comparisons in epidemiological research [18].

Statistical analysis

Burden description

Utilizing the GBD 2021 database, we quantified the global burden of HF-AF/AFL through the following core metrics: prevalence, YLDs, and their ASRs with corresponding 95% UI. Analyses were stratified by age-sex group and SDI quintile to characterize differential burden distributions across populations.

Joinpoint regression analysis

Joinpoint regression analysis, developed by the National Cancer Institute (NCI), employs segmented linear modeling to identify inflection points (joinpoints) in temporal disease trends [19]. This method partitions the observation period into contiguous intervals bounded by statistically significant joinpoints, fitting distinct linear or log-linear functions within each segment while minimizing the residual sum of squared errors (SSE) through an iterative grid-search algorithm. The model quantifies trend dynamics through two principal metrics: the annual percent change (APC), which measures interval-specific short-term fluctuations, and the average annual percent change (AAPC), representing the geometric weighted mean of APC values across all intervals to characterize the global trend magnitude over the entire study period.

In this study, we used Jointpoint regression analysis to identify significant temporal turning points in the HF-AF/AFL burden trends from 1990 to 2021, and calculated the AAPC with 95% confidence intervals (CIs) for prevalence trends spanning 1990–2021. Model validity was established via Monte Carlo permutation testing, constraining the maximum joinpoint count to six to balance model flexibility and parsimony. Trend directionality was determined through AAPC and its 95% CI: an interval entirely above zero indicates a significant upward trajectory, while an interval entirely below zero denotes a significant downward trajectory. Intervals spanning zero were interpreted as stable trends without statistically significant change. Detailed parameter settings for this analysis are provided in the Additional file 1: Section "Study data".

Age-period-cohort modelling analysis

The age-period-cohort modeling employs Poisson regression to quantify disease burden dynamics through three interdependent temporal dimensions: Age effects, which encompass biological aging and socio-developmental transitions throughout the life course, manifest as varying disease susceptibility across age strata due to physiological changes, accumulated social experiences, and shifting social roles; period effects, referring to exogenous environmental shifts that uniformly impact all age groups, include healthcare innovations, policy interventions, economic crises, or pandemics; cohort effects, capturing health consequences stemming from historical exposures unique to specific birth years, reflect the intersection of early-life experiences and macro-societal contexts that persistently influence disease trajectories across generations. The model structure is formalized as follows:

$$In\left(R_{\mathit{apc}}\right)=In\left(I_{\mathit{apc}},/,N_{\mathit{apc}}\right)=\mu +{\alpha }_a+{\beta }_p+{\gamma }_c.$$

Where ln(Rapc) denotes the natural logarithm of the incidence rate, I_apc the case count, N_apc the population denominator, μ the intercept, α_a the age effect for the a^th age group, βp the period effect for the p^th time interval, and γc the cohort effect for the c^th birth cohort [20].

To resolve the inherent identifiability problem (due to perfect collinearity: age = period–cohort), we applied the intrinsic estimator (IE) method, which orthogonalizes parameter estimates via singular value decomposition to yield stable, interpretable solutions. Key outputs include the following: Longitudinal age curve indicates the fitted longitudinal age-specific rates in the reference cohort adjusted for period deviations. Period relative risk (RR) indicates the period relative risk adjusted for age and nonlinear cohort effects in each period relative to the reference period. Cohort RR indicates the cohort relative risk adjusted for age and nonlinear period effects in each cohort relative to the reference cohort. For both relative risk metrics, values exceeding 1.0 indicate elevated disease risk, while values below 1.0 signify protective effects against disease manifestation [20]. The selection of reference values for age, period, and cohort followed the guidance of the National Cancer Institute (NCI) Age-Period-Cohort (APC) Web Tool documentation (https://analysistools.nci.nih.gov/apc/help.html) [19].

In this study, we used the age-period-cohort model analysis to evaluate the impact of three dimensions on the burden of HF-AF/AFL. We extracted estimated population data on prevalence and YLD from GBD 2021 for the global and five SDI regions from 1992 to 2021. According to the requirements of the age-period-cohort model, which typically requires age intervals to be equal to cycle intervals, we conducted paired analysis on the 5-year age group and 5-year cycle. Therefore, in this study, we divided the continuity into 14 age groups, totaling 15–19 years old to 80–84 years old (people under 5 years of age and over 84 years of age were excluded due to the absence or rarity). The range of period effects is from 1992–1996 (median 1994) to 2017–2021 (median 2019), divided into six consecutive periods (with reference to the period from 2002 to 2006). Meanwhile, from 1908–1917 (1812 cohort) to 1998–2007 (2002 cohort), with the 1953–1962 (1957 cohort) birth cohort as the reference period, 19 partially overlapping 10-year birth cohorts were arranged. Estimate overall and local trends by calculating net drift and local drift, as well as age effects, period effects, and cohort effects, to analyze the three-dimensional effects caused by disease burden. The Wald chi-square test was used for the significance of an estimable function. All statistical tests were two-tailed. Detailed reference intervals and reference points are provided in Additional file 1: Section "Definition" [19].

Bayesian age-period-cohort analysis

The Bayesian age-period-cohort (BAPC) model was employed to project global trends in HF-AF/AFL through 2050. This method synthesizes prior epidemiological knowledge with observational data from the GBD study, using Bayesian regularization to address the structural multicollinearity between age, period, and cohort effects. The model yields probabilistic projections of age-standardized prevalence, with UIs quantifying estimation precision [21]. Fitting was performed using the integrated nested Laplace approximation (INLA), which provides deterministic approximations of marginal posterior distributions, and, unlike Markov Chain Monte Carlo sampling, this approach effectively avoids posterior convergence and model convergence [22–24]. Prior studies have demonstrated that INLA offers comparable or superior coverage and accuracy relative to alternative Bayesian estimation techniques, supporting its reliability for the projections presented here [25].

To validate model performance, a hold-out approach was applied using data from 1990–2015 for training to 2016–2021 for testing. Predictive accuracy, measured by the mean squared error (MSE) between projected and observed values, ranged from 0 to 0.66 across sexes and metrics (including ASPR and ASYR) during the test period. This indicates a close fit to the observed data and supports the model’s out-of-sample reliability. These validation results further substantiate the robustness of the projections through 2050. Additional diagnostic details, including prior distribution, effective sample sizes, and model validation results, are provided in Additional file 1: Section "Statistical analysis" and Additional file 1: Fig. S1 [26].

Analytical procedures utilized Joinpoint regression 5.4.0 (NCI) for trend segmentation and the World Health Organization Health Equity Assessment Toolkit for disparity metrics. All Bayesian computations and graphical outputs were executed in R 4.4.2 (R Foundation for Statistical Computing) using the “BAPC” and “INLA” packages, which implement integrated nested Laplace approximations for efficient posterior sampling. All P < 0.05 was considered statistically significant.

Results

Global and regional burden of HF-AF/AFL in 2021

In 2021, HF-AF/AFL affected an estimated 714,138 individuals globally, corresponding to an age-standardized prevalence rate (ASPR) of 8.85 per 100,000 population and an age-standardized years lived with disability rate (ASYR) of 0.79 per 100,000. From 1990 to 2021, both ASPR and ASYR demonstrated a consistent upward trend, with an AAPC of 1.63% (Table 1).

Table 1.

ASPR and ASYR of HF-AF/AFL and their AAPCs at the global and SDIlevels, 1990–2021

	Prevalence (95% UI)					YLDs (95% UI)
	No. of people in 1990	ASPR in 1990 (per 100,000)	No. of people in 2021	ASPR in 2021 (per 100,000)	AAPC (95% CI), 1990–2021	No. of people in 1990	Age-standardized rate in 1990 (per 100,000)	No. of people in 2021	ASYR in 2021 (per 100,000)	AAPC (95% CI), 1990–2021
Global
Overall	162,561	5.36	714,138	8.85	1.63*	14,615	0.48	63,943	0.79	1.63*
	(120,008–213,951)	(3.88–7.05)	(520,543–940,901)	(6.38–11.63)	(1.60 to 1.65)	(8848–23,114)	(0.30–0.74)	(39,058–96,196)	(0.49–1.19)	(1.60 to 1.65)
Sex
Male	60,935	5.10	288,502	8.61	1.66*	5493	0.45	25,893	0.77	1.68*
	(44,778–80,060)	(3.63–6.81)	(210,443–384,366)	(6.15–11.34)	(1.62 to 1.69)	(3328–8515)	(0.28–0.70)	(15,491–40,106)	(0.47–1.17)	(1.64 to 1.71)
Female	101,627	5.52	425,635	9.04	1.59*	9122	0.49	38,050	0.81	1.59*
	(75,073–134,101)	(4.07–7.24)	(3,094,545–555,879)	(6.59–11.80)	(1.57 to 1.60)	(5529–14,474)	(0.31–0.77)	(23,570–57,018)	(0.50–1.21)	(1.57 to 1.61)
Social-demographic index
Low SDI	5979	4.93	19,377	6.27	0.78*	534	0.43	1723	0.55	0.79*
	(4093–8546)	(3.29–6.67)	(13,270–27,678)	(4.21–8.63)	(0.77 to 0.80)	(309–894)	(0.25–0.70)	(1024–2836)	(0.32–0.88)	(0.77 to 0.80)
Low-middle SDI	13,343	3.75	54,798	5.39	1.17*	1187	0.33	4861	0.47	1.18*
	(10,149–17,411)	(2.78–4.93)	(39,832–71,596)	(3.82–7.19)	(1.16 to 1.19)	(743–1846)	(0.21–0.50)	(2946–7545)	(0.29–0.72)	(1.16 to 1.20)
Middle SDI	30,247	4.92	155,902	7.18	1.22*	2705	0.43	13,877	0.64	1.23*
	(23,494–38,992)	(3.70–6.45)	(116,416–202,196)	(5.28–9.31)	(1.21 to 1.23)	(1701–4151)	(0.27–0.65)	(8552–21,104)	(0.39–0.97)	(1.22 to 1.25)
High-middle SDI	32,289	4.14	140,138	7.24	1.82*	2909	0.37	12,562	0.65	1.82*
	(23,877–42,991)	(3.04–5.49)	(102,365–185,308)	(5.27–9.57)	(1.78 to 1.85)	(1754–4570)	(0.23–0.57)	(7527–19,534)	(0.39–0.99)	(1.77 to 1.86)
High SDI	80,490	7.29	343,217	13.97	2.10*	7260	0.66	30,856	1.26	2.11*
	(57,362–107,940)	(5.27–9.67)	(249,216–452,402)	(10.31–18.36)	(2.08 to 2.13)	(4353–11,551)	(0.40–1.04)	(18,841–46,262)	(0.76–1.91)	(2.08 to 2.13)

AAPC average annual percentage change, ASPR age-standardized prevalence rate, ASYR age-standardized years lived with disability rate, CI confidence interval, HF-AF/AFL heart failure attributed to atrial fibrillation/atrial flutter, SDI sociodemographic index, UI uncertainty interval

*P < 0.001

A clear socioeconomic gradient was observed, where high SDI regions bore the greatest burden (ASPR, 13.97; ASYR, 1.26), while low-middle SDI regions exhibited the lowest rates (ASPR, 5.39; ASYR, 0.47) (Table 1). Geographically, the burden was highest in Australasia and Western Europe and lowest in Central Asia (Additional file 1: Figs. S2 and S3 and Additional file 1: Tables S1). At the national level, significant heterogeneity was evident across 204 countries and territories, with the highest burdens observed in Sweden and France and the lowest in Tajikistan; detailed results are provided in Additional file 1: Figs. S4 and S5 and Additional file 1: Tables S2. Longitudinal analysis revealed increased in most countries over the research interval. The Republic of Korea was one of the most pronounced increased nations of ASPR (AAPC = 4.35% [95% CI 4.30% to 4.40%]). Only six territories experienced significant declines in ASPR: Antigua and Barbuda (AAPC = − 0.34%), Somalia (− 0.21%), Ghana (− 0.10%), South Sudan (− 0.07%), Tajikistan (− 0.06%), and Palau (− 0.03%) (Fig. 1 and Additional file 1: Table S2).

Fig. 1.

Global distribution of HF-AF/AFL burden in AAPCs, 1990–2021. A AAPCs in ASPR. B AAPCs in ASYR. AAPCs, average annual percentage changes. ASPR, age-standardized prevalence rate. ASYR, age-standardized years lived with disability rate. HF-AF/AFL, heart failure attributed to atrial fibrillation/atrial flutter

Sex and age patterns

The burden of HF-AF/AFL was higher in females than in males in 2021, with a female-to-male ratio of 1.48 for prevalent cases and 1.47 for YLD (Table 1). The disease burden increased exponentially with age, peaking in the 80–89-year age group (Fig. 2). Although females carried a higher absolute burden, males experienced a slightly faster annual increase in both ASPR and ASYR over the study period (AAPC: 1.66% and 1.68% in males vs. 1.59% and 1.59% in females). Further age- and sex-stratified results are available in Additional file 1: Fig. S6 and Additional file 1: Table S3.

Fig. 2.

Sex- and age-structured global HF-AF/AFL in 2021. A Sex and age-structured global prevalence. B Sex- and age-structured global YLDs. HF-AF/AFL, heart failure attributed to atrial fibrillation/atrial flutter. YLD, years lived with disability

Joinpoint regression analysis

Jointpoint regression analysis reveals important nodes of long-term changes in HF-AF/AFL disease from 1990 to 2021. Although the overall trend of HF-AF/AFL was increasing (AAPC = 1.63% [95% CI 1.60%–1.65%]), a modest decline emerged between 2018 and 2021, evidenced by APCs of − 0.36% for ASPR and − 0.39% for ASYR, suggesting potential inflection points in recent temporal dynamics (Fig. 3A, B and Additional file 1: Table S4). Regarding the changes in sex differences, this research has found that males remained relatively stable from 2018 to 2021, while females experienced a particularly significant decline from 2019 to 2021 (APC = − 2.00% for ASPR; − 1.95% for ASYR) (Fig. 3A, B and Additional file 1: Table S4).

Fig. 3.

Jointpoint analysis for different sexes and SDI regions from 1990 to 2021 for HF-AF/AFL. A ASPR by different sexes globally. B ASYR by different sexes globally. C ASPR by different SDI regions D ASYR by different SDI regions. APC, annual percent change. ASPR, age-standardized prevalence rate. ASYR, age-standardized years lived with disability rate. HF-AF/AFL, heart failure attributed to atrial fibrillation/atrial flutter. SDI, sociodemographic index. *P < 0.05

From 1990 to 2021, the burden in all SDI regions showed an upward trend. High SDI regions experienced the most significant growth (AAPC = 2.10% [95% CI 2.08–2.13], P < 0.001) for ASPR and (AAPC = 2.11% [95% CI 2.08–2.13], P < 0.001) ASYR (Table 1). This trend was particularly pronounced during 2001–2004, where the APC peaked at 4.98% in ASPR and 4.96% in ASYR. However, there was a clear inflection point in high SDI regions with a downward trend with an APC of − 2.69% since 2019 for both ASPR and YLD. Although the low-middle SDI regions had the lowest disease burden, they had maintained stable growth over the past 30 years (Fig. 3C, D and Additional file 1: Table S4).

Age, period, and cohort effects on prevalence and YLDs rate

Age-specific effects of HF-AF/AFL demonstrated consistent patterns across all SDI quintiles. Both prevalence and YLD risk increased exponentially with advancing age, with the highest burden observed in the oldest age groups (80–89 years). High-SDI regions exhibited systematically elevated burden metrics at all ages compared to lower SDI quintiles (Figs. 4A and 5A and Additional file 1: Figs. S7 and S8).

Fig. 4.

Age, period, and cohort effects of prevalence rate by SDI quintiles. A Age effects are shown by the fitted longitudinal age curves of prevalence rate (per 100,000 person-years) adjusted for period deviations. B Period effects are shown by the relative risk of prevalence rate (prevalence rate ratio) and computed as the ratio of age-specific rates from 1992–1996 to 2017–2021, with the referent cohort set at 2002–2006. C Cohort effects are shown by the relative risk of prevalence rate and computed as the ratio of age-specific rates from the 1908 cohort to the 2007 cohort, with the referent cohort set at 1957; the dots and shaded areas denote incidence rates or rate ratios and their corresponding 95% CIs. CI, confidence interval. SDI, sociodemographic index

Fig. 5.

Age, period, and cohort effects of YLDs rate by SDI quintiles. A Age effects are shown by the fitted longitudinal age curves of YLDs rate (per 100,000 person-years) adjusted for period deviations. B Period effects are shown by the relative risk of YLDs rate (YLDs rate ratio) and computed as the ratio of age-specific rates from 1992–1996 to 2017–2021, with the referent cohort set at 2002–2006. C Cohort effects are shown by the relative risk of YLDs rate and computed as the ratio of age-specific rates from the 1908 cohort to the 2007 cohort, with the referent cohort set at 1957. The dots and shaded areas denote YLDs rates or rate ratios and their corresponding 95% CIs. CI, confidence interval. SDI, sociodemographic index. YLDs, years lived with disability

Period effects revealed a sustained global increase in HF-AF/AFL risk across successive time intervals. Relative to the 2002–2006 reference period, all subsequent intervals showed progressively higher risk ratios. This upward trajectory persisted throughout the observation period (1990–2021) and was consistent across all SDI quintiles (Figs. 4B and 5B and Additional file 1: Figs. S7 and S8).

Cohort effects indicated that successive birth cohorts exhibited progressively elevated HF-AF/AFL susceptibility globally and in all SDI regions, with the most pronounced acceleration observed in high-middle SDI regions. This pattern suggests emerging cohort-specific risk exposures influencing long-term disease trajectories (Figs. 4C, 5C).

Prediction of the global disease burden

Using BAPC model, we forecasted the burden of HF-AF/AFL through 2050. Globally, prevalence cases of HF-AF/AFL are projected to reach 1,436,490 (95% UI 1,208,442–1,664,539), with an ASPR of 15.04 (95% UI 12.65–17.43) per 100,000 in 2050. Concurrently, YLD will rise to 128,840 (95% UI 106,053–151,626) with an ASYR of 1.35 (95% UI 1.11–1.59) per 100,000 males (Fig. 6 and Additional file 1: Table S5). This trajectory indicates sustained growth in HF-AF/AFL burden across both sexes, with the ASPR and ASYR remaining disproportionately elevated in females relative to males (Fig. 6 and Additional file 1: Tables S6 and S7).

Fig. 6.

Prediction to 2050 of HF-AF/AFL by sex in global. A ASPR prediction by sex. B ASYR prediction by sex. ASPR, age-standardized prevalence rate. ASYR, age-standardized years lived with disability rate. UI, uncertainty interval. HF-AF/AFL, heart failure attributed to atrial fibrillation/atrial flutter

To delineate SDI-specific trends, we modeled the highest-burden nations within each SDI quintile: Sweden (high SDI), Israel (high-middle SDI), Thailand (middle SDI), Cabo Verde (low-middle SDI), and Angola (low SDI). All five countries exhibited upward prevalence trajectories through 2050. Notably, Cabo Verde showed a divergent trend in disability burden with its ASYR declining significantly despite the rising prevalence, suggesting improvements in disability management or shifts in disease severity (Additional file 1: Fig. S9 and Additional file 1: Tables S8, S9, S10, S11, S12).

Discussion

The global burden of HF-AF/AFL increased significantly from 1999 to 2021, with substantial heterogeneity observed across sex, age, and socioeconomic dimensions. Although an overall upward trend was evident, a modest decline occurred between 2018 and 2021. Age-period-cohort analysis revealed underlying drivers that portend a continuing public health challenge. Specifically, risk escalated exponentially with age, peaking in the 80–89-year cohort, while both period and cohort risk ratios increased progressively over time relative to their reference points. Critically, BAPC projections indicate that the global burden of HF-AF/AFL is expected to continue rising through 2050, with the ASPR projected to reach 15.04 per 100,000. These findings collectively underscore the urgent need for interventions that are stratified by sex and tailored to the socioeconomic context.

In 2021, HF-AF/AFL affected an estimated 714,138 individuals worldwide, reflecting a substantial global health burden. The mortality rates for HF-AF/AFL remained a persistent upward trajectory from 1990 to 2021 [8]. This trend can be attributed to complex factors, notably population aging, alongside the rising prevalence of cardiovascular risk factors, including obesity, hypertension, and diabetes mellitus, which potentiate AF/AFL incidence and subsequent HF progression [27]. The rising burden is further explained by the well-established bidirectional pathophysiological relationship between HF and AF/AFL [28]. On the one hand, both conditions share common etiological risk factors. AF/AFL and HF share common cardiovascular risk factors. For instance, aging can lead to atrial remodelling or to the development of HF by increasing reactive oxygen species, chronic inflammation, and exacerbating atrial and ventricular cell apoptosis [28, 29]. On the other hand, HF and AF/AFL can be causally interrelated [28]. AF/AFL directly promotes the onset and progression of HF through multiple mechanisms. Accelerated and irregular heart rates lead to left ventricular dysfunction, resulting in reduced cardiac output and activation of the renin–angiotensin–aldosterone system (RAAS) [30, 31]. Concurrently, the complex myocardial structural, functional, and electrophysiological remodelling underlying atrial fibrillation frequently involves left atrial enlargement and functional mitral regurgitation, thereby increasing left ventricular filling forces [32]. These factors ultimately precipitate adverse ventricular remodelling. Conversely, HF promotes AF/AFL initiation and perpetuation through several pathways. Elevated left atrial pressure induces atrial enlargement and structural remodeling, while concomitant activation of the RAAS promotes atrial fibrosis [33, 34]. These comprehensive pathological changes increase susceptibility to AF/AFL. Concurrently, ventricular dysfunction alters atrial electrophysiological properties [35]. This self-perpetuating vicious cycle likely substantially exacerbates the persistent upward trajectory of disease burden, particularly among the elderly population.

Between 2018 and 2021, however, the rate of increase in HF-AF/AFL burden moderated globally. This deceleration may be explained by two concurrent factors. First, the decline coincided with the COVID-19 pandemic, during which healthcare disruptions and altered health-seeking behaviors may have led to under-ascertainment of both incident and prevalent cases, artificially reducing recorded burden, a phenomenon supported by reported declines in cardiovascular hospital admissions during this period [36, 37]. Additionally, increased mortality among this vulnerable population due to COVID-19 infection may have temporarily reduced the prevalent case pool. Second, genuine advances in integrated disease management likely contributed to moderating burden growth. During this period, the implementation of evidence-based strategies intensified, particularly for AF/AFL, including broader use of guideline-directed pharmacological therapies and increased adoption of rhythm control via catheter ablation [38]. These interventions, within evolving care models [39], are recognized for improving outcomes and potentially slowing disease progression. These advances have been shown to improve clinical outcomes and attenuate disease progression. Thus, the observed moderation in burden likely reflects a combination of genuine, albeit modest, therapeutic progress alongside surveillance artifacts related to the pandemic.

This study demonstrated significant geographical heterogeneity in the global burden of HF-AF/AFL, with pronounced disparities across SDI regions. High-SDI regions, particularly Australasia and Western Europe, exhibited the highest burden of HF-AF/AFL. This pattern may be influenced by socioeconomic factors [40]. From 1990 to 2021, there was a consistent upward trajectory in ASPR and YLDs of HF-AF/AFL, with the most substantial increases also observed in high-SDI regions. High-income areas, including Western Europe, Australasia, and Asia Pacific, experienced ASPR increases exceeding 100%. Conversely, western and southern sub-Saharan Africa had modest growth rates, potentially attributable to lower AF/AFL incidence in populations of African ancestry relative to groups of European descent, as established in prior epidemiological studies [41]. Despite these variations, the complex interplay of underdeveloped socioeconomic conditions, low incomes, limited educational attainment, and suboptimal healthcare infrastructure in low and lower-middle SDI regions necessitates prioritized public health policy interventions and equitable medical resource allocation [42]. The disproportionate burden in high-income regions underscores the urgent need for targeted interventions, including enhanced AF/AFL screening protocols and evidence-based HF management frameworks.

This study found pronounced sex-based disparities in the global burden of HF-AF/AFL. Similar to the female-dominated global epidemiological burden of AF/AFL in existing studies, HF-AF/AFL still appears to be female-dominated [43]. This discrepancy may stem from sex-specific biological mechanisms, including heightened myocardial susceptibility to arrhythmogenic remodeling, hormonal influences on cardiac electrophysiology, and distinct autonomic regulation [44, 45]. These findings underscore the imperative for sex-stratified approaches in HF-AF/AFL prevention and therapeutic strategies.

This study also found a characteristic age-dependent progression in the burden of HF-AF/AFL. There was a monotonic increase from an onset age of approximately 30 years, reaching a peak in the 80–89 age stratum. This trajectory aligns with established cardiovascular pathophysiological progression, reflecting cumulative exposure to arrhythmic triggers and age-related myocardial vulnerability [8]. The elevated burden observed in octogenarians correlates strongly with prolonged arrhythmia duration and enhanced susceptibility of senescent cardiac tissue to AF/AFL-induced electromechanical dysfunction. Population aging amplifies this phenomenon, underscoring the imperative for stratified interventions targeting early detection, rhythm control optimization, and HF risk mitigation in geriatric populations. Such strategies should address the unique pathophysiological interplay between AF/AFL and age-related cardiac remodeling to effectively reduce disease burden.

Age-period-cohort analysis identified the independent effects of population aging, diagnostic period, and birth cohort on the burden of HF-AF/AFL [46–49]. Longitudinal age curve indicated that increasing age is a risk factor for the disease burden of HF-AF/AFL. Cohort effects exhibited an even more profound impact: successive birth cohorts born after 1957 experienced progressively elevated disease risk, particularly in high-middle SDI regions. This generational escalation likely reflects cumulative exposure to modernized risk factors, including sedentary lifestyles, processed diets, and delayed cardiovascular prevention, which synergistically accelerate the pathophysiology of HF-AF/AFL. The convergence of these relative risk effects establishes a self-reinforcing disease trajectory, necessitating life-course interventions targeting modifiable risks within high-risk birth cohorts.

Bayesian age-period-cohort modeling projects a substantial increase in the global burden of HF-AF/AFL from 2022 to 2050. Both the ASPR and YLD are projected to rise across all populations. Although females exhibit a marginally higher baseline burden, both sexes demonstrate parallel trends in disease progression. This rising prevalence underscores the condition’s persistent public health significance and implies a growing financial strain on healthcare systems globally. Economic analyses indicate that managing AF/AFL already incurs considerable costs, driven largely by hospitalizations, procedural interventions such as catheter ablation, and long-term pharmacotherapy [50]. The projected escalation in burden suggests these economic pressures will intensify, particularly in high SDI regions where advanced diagnostics and treatments are more accessible but also more costly [51]. Addressing this growing burden, which is driven primarily by population aging and increasing AF/AFL incidence, requires a coordinated, multi-level strategy. Key priorities include enhancing early detection through expanded rhythm monitoring; strengthening primary prevention of modifiable risk factors, including hypertension, diabetes, and obesity; and optimizing integrated treatment pathways for both HF and AF/AFL. Particular emphasis should be placed on geriatric populations, in whom HF-AF/AFL prevalence peaks between 80 and 89 years of age due to the accumulation of cardiovascular insults over time [52–55]. Implementing such a comprehensive approach is essential to mitigate the projected health and economic impact of HF-AF/AFL in the coming decades.

Limitation

This study has several limitations. First, diagnostic complexity in attributing HF etiology specifically to AF/AFL may introduce misclassification bias, potentially affecting prevalence estimates [56]. While standardized ICD coding (ICD-9, 427.3–427.32; ICD-10, I48–I48.92) was applied uniformly, clinical overlap with other cardiomyopathies remains a challenge. Second, our analysis adopts the GBD comparative risk assessment framework, which attributes HF to AF/AFL as a distinct impairment. It must be acknowledged, however, that HF and AF/AFL exhibit a well-established bidirectional relationship, wherein HF may also precipitate or exacerbate AF/AFL. While the GBD framework aims to quantify the burden specifically attributable to AF/AFL as a causal factor, this inherent pathophysiological complexity suggests our estimates should be interpreted as reflecting the strength of this specific epidemiological association within a standardized analytical model. Third, because HF is classified as an impairment rather than a fatal disease entity in the GBD study, our analysis was restricted to prevalence and YLDs. Mortality data, YLLs, and DALYs were therefore not incorporated, which precludes a complete assessment of the fatal and total disease burden. This is an inherent constraint of the available data architecture. Fourth, while the GBD 2021 data contain HF severity subtypes, the present analysis examined the aggregate burden only. Future investigations could yield deeper mechanistic and clinical insights by examining the distribution of HF-AF/AFL across these different severity classifications. Fifth, general limitations of the GBD framework must be considered [57]. Despite systematic data processing and modeling, variations in source data quality and coverage across regions can lead to discrepancies in burden estimations. Underreporting is particularly probable in settings with limited structured arrhythmia surveillance, where undiagnosed AF/AFL may result in an underestimation of associated HF. Finally, the long-term projections from the BAPC model carry inherent uncertainty. While the modeling accounted for historical trends in sociodemographic and risk factor distributions, unforeseen future perturbations, such as healthcare system disruptions or the introduction of novel therapies, could alter the actual disease trajectory. These limitations underscore the need for improved diagnostic specificity, integrated mortality data, and ongoing model refinement to enhance the accuracy of future burden estimates.

Conclusions

The global epidemiological burden of HF-AF/AFL from 1999 to 2021 demonstrated a substantial and continuously increasing trend, with marked heterogeneity across sex, age, and socioeconomic strata. These findings suggest that the burden of HF-AF/AFL is likely to continue rising over the coming decades and to reveal distinct age-, period-, and cohort-related risk patterns, underscoring the need for targeted public health strategies and interventions.

Abbreviations

AAPC

Average annual percentage change

ASR

Age-standardized rate

BAPC

Bayesian age-period-cohort

CI

Confidence interval

GBD

Global Burden Disease, Injuries, and Risk Factors Study

HF-AF/AFL

Heart failure attributable to atrial fibrillation/flutter

SDI

Sociodemographic index

UI

Uncertainty interval

YLD

Years lived with disability

Authors’ contributions

B.H. and G.L. contributed equally to this work and share senior authorship. B.H. and G.L. conceptualized and supervised the study and critically revised the manuscript. J.C. and Z.L. curated the data, developed the methodology, performed the statistical analyses, and drafted the original manuscript. Y.L., L.X., B.R.S., and S.L. contributed to data curation, interpretation of the findings, and manuscript review and editing. All authors read and approved the final manuscript.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

All data underlying the findings presented in this manuscript are publicly available from the IHME Global Health Data Exchange (GHDx) and the GBD Results tool ([https://vizhub.healthdata.org/gbd-results/](https:/vizhub.healthdata.org/gbd-results); also accessible via [http://ghdx.healthdata.org/gbd-results-tool](http:/ghdx.healthdata.org/gbd-results-tool)). The analytic code and detailed data extraction procedures are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable. This study was based entirely on de-identified, aggregate, publicly available data from the GBD study; therefore, ethical approval and informed consent were not required.

Consent for publication

All authors have read and approved the manuscript and consent to its publication.

Competing interests

The authors declare no competing interests.