Circulating transthyretin with atrial morpho-functional phenotypes and atrial fibrillation risk, and the modifying role of BMI

Abstract

Background

Low circulating transthyretin (TTR) concentration has been suggested as a biomarker of transthyretin tetramer instability, a prerequisite for the development of transthyretin cardiac amyloidosis. This study aimed to evaluate the associations between circulating TTR levels with incident atrial fibrillation (AF) and other arrhythmias.

Methods

This study used data from the UK Biobank. Participants with available TTR data and without prior arrhythmias were included. The primary outcome was new-onset AF. The secondary outcomes were new-onset supraventricular arrhythmias (SVA), bradyarrhythmias, cardiac block, and ventricular arrhythmias (VA). Multivariable Cox regression was applied to evaluate the associations between circulating TTR levels with arrhythmia outcomes.

Results

A total of 40,723 participants (mean age 56.7 ± 8.2 years; 55% women) were included. After adjusting for potential confounders, one standard deviation (SD) decrease in TTR levels was associated with an increased risk of incident AF (HR 1.06, 95% CI 1.02–1.11). Furthermore, significant associations between low TTR with atrial structural remodeling were observed, manifesting as increased left atrial volume index (β 0.51, 95% CI 0.09–0.92) and right atrial volume index (β 0.87, 95% CI 0.39–1.40). In addition, there was a significant association between lower TTR levels with higher incident SVA risk, but not for bradyarrhythmias, cardiac block, or VA. A consistently significant interaction effect was identified between TTR levels and BMI for the risk of AF, SVA, bradyarrhythmias, and cardiac block (all P_interaction < 0.05), with lower TTR levels being significantly associated with a higher risk of AF (HR 1.15, 95% CI 1.06–1.26), SVA (HR 1.15, 95% CI 1.06–1.25), bradyarrhythmias (HR 1.17, 95% CI 1.05–1.30), and cardiac block (HR 1.15, 95% CI 1.02–1.29) among individuals with a BMI < 25 kg/m². In addition, carriers of likely pathogenic or pathogenic TTR variants (LP/P) had lower levels of plasma TTR compared with noncarriers, as well as higher arrhythmia risks, especially for non-Val142Ile carriers.

Conclusions

Lower circulating TTR concentrations were associated with higher risk of incident AF. Exposure to low TTR and low BMI may be associated with a higher risk of AF, SVA, bradyarrhythmias, and cardiac block.

Graphical Abstract

Low circulating transthyretin levels predict incident AF. Abbreviations: TTR, transthyretin; AF, atrial fibrillation; SVA, supraventricular arrhythmias; VA, ventricular arrhythmias

Supplementary Information

The online version contains supplementary material available at 10.1186/s12916-025-04391-6.

Background

Transthyretin (TTR), previously known as prealbumin, is a homotetrameric protein primarily synthesized in the liver that transports thyroxine and retinol [1–4]. Inherited variants in the TTR gene, or advancing age in noncarriers, have been shown to destabilize the tetramer, causing its dissociation into monomers, which subsequently undergo abnormal folding and assemble into insoluble amyloid fibrils [5]. These amyloid fibrils in turn deposit extracellularly locally or throughout the body, leading to transthyretin amyloidosis, with cardiac involvement being the most common phenotype [6].

Besides well-established heart failure (HF), various studies have demonstrated high risks of arrhythmias among individuals with transthyretin amyloid cardiomyopathy (ATTR-CM), including atrial fibrillation (AF), other supraventricular arrhythmias (SVA), ventricular arrhythmias (VA), and conduction disease [6, 7]. AF represents the most common sustained arrhythmia in patients with ATTR-CM, with a prevalence of 30% to 70% at or prior to the diagnosis [2, 8]. Arrhythmias such as these are often highly symptomatic and poorly tolerated in cardiac amyloidosis patients, which potentially result in unfavorable complications and prognosis, such as higher risk of intracardiac thrombus, stroke, permanent pacing device, or implantable cardioverter-defibrillator implantation, and may require expedited and individualized treatment [7].

Tetramer destabilization is the rate-limiting step in transthyretin amyloidosis [5]. However, currently, there are no methods available to directly measure TTR stability in large population samples, but several lines of evidence support that plasma TTR concentration decreases with increasing tetramer destabilization [5]. Specifically, the presence of a stabilizing variant in TTR, such as p.T139M, is associated with higher concentrations of plasma TTR [9], whereas the presence of a destabilizing variant in TTR, such as p.Val142Ile, is associated with lower TTR concentrations [1, 5]. Additionally, treatment with transthyretin-stabilizing drugs increases plasma TTR in both healthy individuals and in patients with ATTR-CM [5, 10]. Therefore, increasingly lower plasma TTR could be considered a surrogate marker for transthyretin tetramer destabilization. Based on these data, recent studies have identified a significant association between lower plasma TTR levels and higher risk of incident HF, atherosclerotic cardiovascular disease, cardiovascular mortality, and all-cause mortality among the general population [1, 6]. However, the predictive role of plasma TTR in cardiac arrhythmias development is still unknown.

Therefore, this study leveraged data from the UK Biobank to evaluate the associations of plasma TTR levels with AF and other cardiac arrhythmias, as well as atrial structure and function phenotypes.

Methods

Study population

The UK Biobank is a population-based cohort study of ~ 500,000 adults who were 40–70 years old at the time of recruitment between 2006 and 2010. Detailed study design has been described previously [11]. The UK Biobank was approved by the North West Multi-Centre Research Ethics Committee (No. 11/NW/0382) and conducted according to the Declaration of Helsinki. Written informed consent was obtained from all participants. This research was conducted using the UK Biobank resources (application number 79146). Among the participants from the UK Biobank, 54,219 participants underwent plasma profiling under the UK Biobank Plasma Profiling Project (UKBB-PPP), who were picked randomly and are therefore representative of the full cohort [12]. In the present study, individuals who underwent plasma profiling were included, and those with diagnosed arrhythmia at baseline were excluded.

Measurement of transthyretin levels and ascertainment of covariates

The UKBB-PPP measured baseline EDTA plasma samples by the Olink Explore 3072, using a proximity extension assay (PEA), which allows for simultaneous analysis of various proteins with high specificity and sensitivity. A total of 2923 unique proteins across 8 protein panels were measured, including TTR, and extensive quality control steps were implemented [12]. The Olink PEA technology has been validated against traditional assays in previous studies, showing good concordance and reliability [12, 13]. The detailed methods are provided in Additional file 1: supplemental methods [1, 12, 14–20]. TTR levels measured on the Olink platform were quantified using Normalized Protein eXpression (NPX) values [1, 12]. The NPX values are measured on the log₂ scale, with lower values indicating lower levels of the analyte.

Detailed methods on the ascertainment of covariates are described in the Additional file 1: supplemental methods. The diagnosis of prior comorbidities was obtained using linkages with primary care and hospital inpatient records identified by International Classification of Diseases, Ninth Revision (ICD-9), or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes (Additional file 1: Table S1).

Outcomes and follow-up

The primary outcome was new-onset AF. The secondary outcomes included new-onset SVA (including AF, premature atrial contractions, and supraventricular tachycardia), bradyarrhythmias (including atrioventricular/bundle branch block, pacemaker insertion, and sinus node dysfunction), cardiac block (atrioventricular/bundle branch block), and VA (including ventricular tachycardia, ventricular premature depolarizations, cardiac arrest, and implantable cardioverter defibrillator insertion). Incident events were defined by the first occurrence of ICD-9 or ICD-10 codes for the corresponding arrhythmias in any primary care data, hospital inpatient data, death register records, or Office of Population Censuses and Surveys Classification of Interventions and Procedures-4 (OPCS-4) code for a corresponding arrhythmia-related procedure [21]. The definitions for the outcomes are listed in Additional file 1: Table S1. Participants were followed up since the recruitment up to the occurrence of corresponding arrhythmia, death, loss to follow-up, or the end of follow-up, whichever came earlier.

Atrial morpho-functional phenotypes measured by CMR

Since 2014, a subsample of UK Biobank participants was invited back for cardiac magnetic resonance (CMR) assessment [22, 23]. Common CMR parameters were acquired and processed by the UK Biobank team and made available to approved researchers. CMR parameters used in this study included functional and structural parameters of the left and right atria, and the detailed methodology of the measurements of these parameters has been described previously [24].

Genotyping

Whole exome sequencing (WES) data from the UK Biobank was utilized to ascertain carriers of TTR variants. In brief, WES was performed on the Illumina NovaSeq 6000 platform (Regeneron Genetics Center) with ≥ 20 × coverage. The final dataset release encompassed > 470,000 participants [18]. The genotyping methodology used in this study was based on the established protocols from a recent TTR-focused study [19], detailed in Additional file 1: supplemental methods. Finally, we extracted a total of 62 candidate TTR variants, including 35 likely pathogenic or pathogenic variants (LP/P variants: 24 likely pathogenic, 11 pathogenic), and 27 variants of uncertain significance (VUS) (Additional file 1: Table S2).

Statistical analysis

Baseline characteristics were described as frequency (proportions) for categorical variables. Continuous variables were presented as mean (SD) or median (interquartile range) for normal and non-normal distributions. The differences in characteristics were compared using analysis of variance (ANOVA), the Kruskal–Wallis test, and the χ² test, as appropriate.

Incidence rates of outcomes were calculated as the number of cases per 1000 person-years. Considering the sex differences in the TTR levels (Additional file 1: Fig. S1), sex-specific values were used to categorize the cohort into individuals with low (tertile 1), intermediate (tertile 2), and high TTR levels (tertile 3) [1]. Cumulative incidences of each outcome were estimated using Kaplan–Meier analyses and compared using the log-rank test. Multivariable Cox models were used to estimate the adjusted hazard ratio (HR) for the study outcomes per standard deviation (SD) decrease in TTR levels. To enhance clinical interpretability, we applied a negative transformation to TTR values. Consequently, HRs represent the risk associated with per SD decrease in the original TTR levels [1]. The SD used for scaling TTR levels was calculated separately within each stratified subgroup. The proportional hazards assumption was assessed using Schoenfeld residuals. No significant violations were identified for any covariate, and the global tests for the overall models were non-significant. Model 1 was adjusted for age, sex, and race. Based on model 1, model 2 was adjusted for body mass index (BMI), smoking, and drinking status. Model 3 was further adjusted for total cholesterol, estimated glomerular filtration rate (eGFR), triglyceride, low-density lipoprotein cholesterol (LDL-C), heart failure, diabetes, hypertension, coronary artery disease, valvular heart disease (VHD), chronic kidney disease, obstructive sleep apnea, antihypertensive medication, and cholesterol-lowering medication. These covariates were selected a priori based on established clinical relevance and confounding potential in the existing literature [1, 19, 21, 25, 26].

Restricted cubic spline (RCS) analysis based on model 3 was performed to assess the dose–response relationship of TTR levels with AF risk (R package “rms,” version 6.8–1). According to Harrell [27], using four knots provides adequate model fit and represents a good compromise between flexibility and the loss of degrees of freedom due to overfitting. Therefore, we selected four knots positioned at the 5th, 35th, 65th, and 95th percentiles of TTR levels. Subgroup analyses were conducted according to age, sex, BMI, smoking, drinking, prior comorbidities, and the genetic risk of AF. The genetic risk of AF was based on the standard polygenic risk score (PRS) released by UK Biobank (field ID 26212) [28]. Participants were classified into groups with low (the lowest PRS tertile), intermediate (the middle PRS tertile), and high (the highest PRS tertile) genetic risk [28]. A multiplicative interaction analysis was performed by using the likelihood ratio test comparing models with and without a cross-product term [29]. Multiplicative interaction was visualized using the “visreg” R package (version, 2.7.0). A Benjamini–Hochberg false discovery rate (FDR) correction was used as the significance threshold for the interaction analyses due to multiple comparisons [30]. To test the robustness of the results, several sensitivity analyses were conducted. First, sub-distribution hazard model adjusting for the competing risk of death was applied. Second, arrhythmias which occurred during the first year of follow-up were excluded. Third, patients with baseline heart failure, coronary artery disease, and VHD were excluded. Fourth, given TTR’s role in transporting thyroxine and retinol, to address potential confounding by thyroxine levels, individuals with baseline thyroid dysfunction (hyperthyroidism or hypothyroidism) were excluded. Fifth, dietary retinol intake was further adjusted based on the fully adjusted model. Missing values of dietary retinol intake were handled via multiple imputation using chained equations (MICE) method (R package “mice,” version 3.18.0). Five imputed datasets were generated, and Rubin’s rules were applied to combine the final analysis results [20]. Finally, to verify the robustness of the primary continuous exposure model, arrhythmia risks were compared across TTR tertiles, based on the fully adjusted model.

The associations between TTR levels with CMR measurements were assessed using multivariable linear regression adjusted for the potential covariates and the time interval between baseline and CMR data acquisition [22]. In addition, an analysis investigating the interaction of the time interval between these two dates and TTR levels was conducted [31]. A p-value < 0.05 was considered statistically significant. The statistical analysis was performed with RStudio (version 4.4.1) and Python (version 3.11).

Results

Baseline characteristics

After excluding participants without available data of baseline TTR (n = 457,805), those with prior AF diagnosis (n = 881), and with missing values on any covariate (n = 2761), we finally included 40,723 participants (mean age 56.7 ± 8.2, 55% women) in the primary analysis (Additional file 1: Fig. S2).

Table 1 presents the baseline characteristics of the included participants stratified by TTR levels. The median [IQR] levels of TTR in females [− 0.1 (− 0.3, 0.1)] were significantly lower than males [0.1 (− 0.1, 0.3)] (Additional file 1: Fig. S1). Participants with lower circulating TTR had a higher prevalence of coronary artery disease and diabetes, as well as a lower risk of baseline hypertension.

Table 1.

Baseline characteristics of the included individuals for the association between TTR with atrial fibrillation

Characteristic	Overall (N = 40,723)	Low TTR (N = 13,574)	Intermediate TTR (N = 13,572)	High TTR (N = 13,577)	p-value
Demographics
Age, years	56.7 (8.2)	56.7 (8.4)	56.8 (8.1)	56.7 (7.9)	0.300
Female sex	22,231 (55%)	7410 (55%)	7409 (55%)	7412 (55%)	0.900
Race/ethnicity					< 0.001
White	37,942 (93%)	12,350 (91%)	12,691 (94%)	12,901 (95%)
Asian	901 (2.2%)	407 (3.0%)	293 (2.2%)	201 (1.5%)
Black	974 (2.4%)	452 (3.3%)	289 (2.1%)	233 (1.7%)
Mixed	267 (0.7%)	107 (0.8%)	83 (0.6%)	77 (0.6%)
Others/unknown	639 (1.6%)	258 (1.9%)	216 (1.6%)	165 (1.2%)
Body mass index, kg/m2	27.4 (4.8)	27.5 (5.3)	27.3 (4.6)	27.4 (4.3)	< 0.001
Smoking status					< 0.001
Current	4406 (11%)	1653 (12%)	1358 (10%)	1395 (10%)
Previous	14,141 (35%)	4309 (32%)	4734 (35%)	5098 (38%)
Never	22,176 (54%)	7612 (56%)	7480 (55%)	7084 (52%)
Drinking status					< 0.001
Current	37,246 (91%)	11,995 (88%)	12,476 (92%)	12,775 (94%)
Previous	1557 (3.8%)	689 (5.1%)	491 (3.6%)	377 (2.8%)
Never	1920 (4.7%)	890 (6.6%)	605 (4.5%)	425 (3.1%)
Laboratory parameters
Total cholesterol, mmol/L	5.7 (1.2)	5.4 (1.1)	5.7 (1.1)	6.0 (1.2)	< 0.001
LDL-C, mmol/L	3.5 (0.9)	3.4 (0.8)	3.6 (0.9)	3.7 (0.9)	< 0.001
Triglyceride, mmol/L	1.7 (1.0)	1.5 (0.8)	1.7 (1.0)	2.0 (1.2)	< 0.001
eGFR, mL/min/1.73 m2	91.2 (13.6)	92.2 (13.4)	91.1 (13.2)	90.3 (14.1)	< 0.001
Comorbidity and medication history
Heart failure	218 (0.5%)	73 (0.5%)	71 (0.5%)	74 (0.5%)	0.900
Hypertension	11,132 (27%)	3432 (25%)	3639 (27%)	4061 (30%)	< 0.001
Coronary artery disease	2194 (5.4%)	793 (5.8%)	720 (5.3%)	681 (5.0%)	0.009
Valvular disease	301 (0.7%)	109 (0.8%)	108 (0.8%)	84 (0.6%)	0.130
Chronic kidney disease	534 (1.3%)	140 (1.0%)	176 (1.3%)	218 (1.6%)	< 0.001
Diabetes	2170 (5.3%)	848 (6.2%)	684 (5.0%)	638 (4.7%)	< 0.001
Obstructive sleep apnea	192 (0.5%)	74 (0.5%)	48 (0.4%)	70 (0.5%)	0.046
Antihypertensive medication	8696 (21%)	2716 (20%)	2807 (21%)	3173 (23%)	< 0.001
Cholesterol-lowering medication	7269 (18%)	2143 (16%)	2331 (17%)	2795 (21%)	< 0.001

Values are mean (SD) and n (%)

Abbreviations: eGFR estimated glomerular filtration rate, LDL low-density lipoprotein cholesterol, TTR transthyretin

The association between TTR levels with incident AF and atrial morpho-functional phenotypes

Over 561,705 person-years of follow-up (median follow-up 14.6 [IQR 13.7, 15.4] years), 2930 (7.2%) participants developed new-onset AF (5.22 per 1000 person-years). The cumulative incidence of AF was higher among individuals in the lower tertiles of circulating TTR (Fig. 1A). In individuals with high, intermediate, and low TTR levels, the incidence rate of AF was 4.79 (95% CI, 4.48–5.11), 5.13 (95% CI, 4.81–5.46), and 5.75 (95% CI, 5.41–6.10) per 1000 person-years, respectively.

Fig. 1.

Cumulative curves of outcomes stratified by transthyretin tertiles. Cumulative curves of incident A atrial fibrillation, B supraventricular arrhythmias, C bradyarrhythmias, D cardiac block, and E ventricular arrhythmias in the study population. Log-rank tests were used. Abbreviation: TTR, transthyretin

A decrease in circulating TTR levels was associated with significantly increased risk of new-onset AF, consistent across three models (Table 2). In the fully adjusted model, the risk of AF increased by 6% per SD decrease in TTR levels [HR, 1.06; 95% CI (1.02, 1.11); p = 0.005], which was further supported by the RCS curve between TTR and AF (Fig. 2A). After adjusting for covariates, the RCS analyses demonstrated a steeper negative slope among individuals with lower TTR levels (P non-linear = 0.001), suggesting that lower circulating TTR concentrations were associated with higher risk of incident AF.

Table 2.

The associations between circulating TTR with incident arrhythmias

	Event/number	Incidence rate	Model 1		Model 2		Model 3
		Per 1000 person-years	HR [95% CI], p-value		HR [95% CI], p-value		HR [95% CI], p-value
Primary outcome
Atrial fibrillation	2930/40,723	5.22 (5.03, 5.41)	1.09 [1.05, 1.13], p < 0.001		1.07 [1.03, 1.11], p < 0.001		1.06 [1.02, 1.11], p = 0.005
Secondary outcomes
Supraventricular arrhythmias	3098/40,609	5.54 (5.35, 5.74)	1.10 [1.06, 1.14], p < 0.001		1.08 [1.04, 1.12], p < 0.001		1.07 [1.03, 1.12], p < 0.001
Bradyarrhythmias	2016/41,274	3.51 (3.36, 3.67)	1.05 [1.00, 1.10], P = 0.040		1.03 [0.98, 1.08], p = 0.200		1.03 [0.98, 1.08], p = 0.300
Cardiac block	1590/41,428	2.74 (2.61, 2.88)	1.05 [1.00, 1.11], p = 0.075		1.03 [0.98, 1.09], p = 0.300		1.02 [0.97, 1.08], p = 0.400
Ventricular arrhythmias	701/41,476	1.20 (1.11, 1.29)	1.05 [0.97, 1.13], p = 0.300		1.02 [0.94, 1.11], p = 0.600		1.05 [0.97, 1.15], p = 0.200

Multivariable-adjusted Cox models were used to estimate the adjusted hazard ratio of study outcomes per SD decrease in TTR levels

Model 1, adjusted for age, sex, and race

Model 2, model 1 + body mass index, smoking status, and drinking status

Model 3, model 2 + total cholesterol, estimated glomerular filtration rate, triglyceride, low-density lipoprotein cholesterol, heart failure, diabetes, hypertension, coronary artery disease, valvular disease, chronic kidney disease, obstructive sleep apnea, antihypertensive medication, cholesterol-lowering medication

Bold indicates p-value < 0.05

Abbreviations: TTR transthyretin, HR hazard ratio, CI confidence interval

Fig. 2.

Risk of incident cardiac arrhythmias by circulating transthyretin levels in the study population. Risk of incident A atrial fibrillation, B supraventricular arrhythmias, C bradyarrhythmias, D cardiac block, and E ventricular arrhythmias by circulating transthyretin levels in the study population. Orange lines and shaded areas depict restricted cubic spline curves based on model 3. The blue density function displays distribution of baseline circulating transthyretin concentration. Restricted cubic spline analyses were based on the fully adjusted model, accounting for age, sex, race, body mass index, smoking status, drinking status, total cholesterol, estimated glomerular filtration rate, triglyceride, low-density lipoprotein cholesterol, heart failure, diabetes, hypertension, coronary artery disease, valvular disease, chronic kidney disease, obstructive sleep apnea, antihypertensive medication, and cholesterol-lowering medication. Abbreviation: TTR, transthyretin

Among the individuals with TTR data, CMR data were available for 3402. The baseline characteristics between those with and without CMR data are presented in Additional file 1: Table S3. Individuals with available CMR data were younger, more likely to be male, and had a lower prevalence of comorbidities. After fully adjusting for potential confounders, per SD decrease in circulating TTR levels was associated with significantly larger volume indices for both the left and right atria, including left atrial maximum volume [LAVmax, β = 0.96; 95% CI (0.12, 1.80); p = 0.025], LAVmax indexed to body surface area [LAVimax, β = 0.51; 95% CI (0.09, 0.92); p = 0.017], right atrial maximum volume [RAVmax, β = 1.50; 95% CI (0.56, 2.40); p = 0.002], RA minimum volume [RAVmin, β = 0.78; 95% CI (0.15, 1.40); p = 0.015], RAVmax index [RAVimax, β = 0.87; 95% CI (0.39, 1.40); p < 0.001], and RAVmin index [RAVimin, β = 0.45; 95% CI (0.12, 0.77); p = 0.007] (Additional file 1: Table S4).

For atrial functional parameters, per SD decrease in circulating TTR levels was associated with increased LA total emptying volume [LATEV, β = 0.44; 95% CI (0.02, 0.85); p = 0.038] and RA total emptying volume [RATEV, β = 0.69; 95% CI (0.23, 1.20); p = 0.003], but not for LA total emptying fraction [LATEF, β =  − 0.06; 95% CI (− 0.40, 0.27); p = 0.700] or RA total emptying fraction [RATEF, β = 0.06; 95% CI (− 0.27, 0.39); p = 0.700] (Additional file 1: Table S4). Analyses investigating the interaction between TTR levels and the time interval (between baseline and imaging visit date) did not find any significant results (Additional file 1: Table S4).

Sensitivity analyses and subgroup analyses

Results from all the sensitivity analyses were consistent with the primary analyses (Additional file 1: Tables S5 and S6). Lower plasma TTR levels were associated with significantly increased AF risk after excluding AF cases within the first year of follow-up [HR, 1.06; 95% CI (1.02, 1.11); p = 0.007], those with baseline heart failure, coronary artery disease and VHD [HR, 1.06; 95% CI (1.02, 1.11); p = 0.009], and individuals with baseline thyroid dysfunction [HR, 1.06; 95% CI (1.02, 1.11); p = 0.006] or using sub-distribution hazard model [HR, 1.06; 95% CI (1.02, 1.10); p = 0.006]. Besides, the association remained significant after further adjustment for dietary retinol intake [HR, 1.06; 95% CI (1.02, 1.11); p = 0.005] (Additional file 1: Table S5). When treating TTR as a categorical variable, compared to individuals with high TTR levels, those in the lowest tertile presented with significantly higher AF risk [HR, 1.11; 95% CI (1.01, 1.22); p = 0.025] (Additional file 1: Table S6).

Subgroup analyses observed that compared to individuals with a higher BMI, the association between TTR and AF risk was greater among those with a BMI < 25 kg/m² (Fig. 3). Significant interaction of BMI on the associations between TTR and AF was observed (P_interaction = 0.003; FDR-corrected P_interaction = 0.013) (Fig. 4A). Besides, the associations between lower TTR and increased AF risk were greater among individuals with an older age (P_interaction < 0.001; FDR-corrected P_interaction < 0.001).

Fig. 3.

Subgroup analyses for the association between circulating transthyretin levels with incident AF. Multivariable cox analyses were based on the fully adjusted model, accounting for age, sex, race, body mass index, smoking status, drinking status, total cholesterol, estimated glomerular filtration rate, triglyceride, low-density lipoprotein cholesterol, heart failure, diabetes, hypertension, coronary artery disease, valvular disease, chronic kidney disease, obstructive sleep apnea, antihypertensive medication, and cholesterol-lowering medication. Abbreviations: AF, atrial fibrillation; BMI, body mass index; CAD, coronary artery disease; CKD, chronic kidney disease; TTR, transthyretin

Fig. 4.

The interaction effects of BMI and transthyretin levels for the risk of cardiac arrhythmias. The interaction effects of BMI and transthyretin levels for the risk of incident A atrial fibrillation, B supraventricular arrhythmias, C bradyarrhythmias, D cardiac block, and E ventricular arrhythmias. Abbreviations: BMI, body mass index; TTR, transthyretin

Additionally, there was a significant association between TTR and AF among individuals with low PRS, but not for those in intermediate or high PRS group (P_interaction < 0.001; FDR-corrected P_interaction < 0.001). No significant interaction was identified by sex, smoking, drinking, and other comorbidities on the associations between TTR and AF (Fig. 3).

The association between TTR levels with secondary outcomes

During follow-up, new-onset SVA, bradyarrhythmias, cardiac block, and VA developed in 3098 (7.6%), 2016 (4.9%), 1590 (3.8%), and 701 (1.7%) participants, respectively. As shown in the Kaplan–Meier curves (Fig. 1B, C, D, E), the cumulative incidence of each outcome was higher among individuals in the lower tertiles of circulating TTR (all log-rank p < 0.05).

Lower plasma TTR levels were associated with significantly higher risk for SVA [per SD decrease TTR: HR, 1.07; 95% CI (1.03, 1.12); p < 0.001], after adjusting for potential confounders. No significant associations were observed between lower plasma TTR with bradyarrhythmias, cardiac block, or VA risk (Table 2). Similar to AF, the multivariable RCS analyses for SVA (Fig. 2B), bradyarrhythmias (Fig. 2C), and cardiac block (Fig. 2D) all demonstrated a steeper negative slope among individuals with lower TTR levels (all P non-linear < 0.05), suggesting a signal that lower circulating TTR levels may be associated with increased risk of these outcomes. However, no significant linear or nonlinear association was identified between TTR with VA, with a slightly flattened negative slope among individuals with lower TTR levels (Fig. 2E).

Similar to AF, subgroup analyses suggested an association between lower TTR levels with higher SVA [HR, 1.15; 95% CI (1.06, 1.25); p < 0.001], bradyarrhythmias [HR, 1.17; 95% CI (1.05, 1.30); p = 0.003], and cardiac block [HR, 1.15; 95% CI (1.02, 1.29); p = 0.023] risk among individuals with BMI < 25 kg/m² (Additional file 1: Figures S3, S4, S5), compared to individuals with BMI ≥ 25 kg/m² (all P_interaction < 0.05). Following FDR correction for multiple testing of interaction terms, the interaction remained statistically significant for SVA (FDR-corrected P_interaction = 0.010), but not for bradyarrhythmias (FDR-corrected P_interaction = 0.072) or cardiac block (FDR-corrected P_interaction = 0.084). The differential association patterns according to BMI strata are visually depicted in Fig. 4, suggesting a potentially stronger negative association between TTR levels and SVA, bradyarrhythmias, and cardiac block risk among individuals with lower BMI (< 25 kg/m²), as evidenced by steeper negative slopes in this subgroup. No significant interaction of BMI was observed for the association between TTR and VA, either before (P_interaction = 0.161) or after FDR correction (FDR-corrected P_interaction = 0.478; Fig. 4E and Additional file 1: Fig. S6). Interaction plots of dietary retinol intake and TTR levels with cardiac arrhythmias are presented in Additional file 1: Fig. S7, suggesting no significant interaction effects. In addition, all sensitivity analyses yielded results consistent with the primary findings (Additional file 1: Tables S5 and S6).

The association between TTR variants with arrhythmia outcomes

A total of 469,835 individuals were included in the genetic analysis (Additional file 1: Table S7). Consistent with prior report [19], a TTR variant was identified in 564 participants (0.12%), including 473 (0.10%) with LP/P variants and 91 (0.02%) with VUS. The p.Val142Ile variant accounted for the majority of LP/P variants, presenting in 367 of 473 individuals (77.59%). Compared to noncarriers [0.0 (− 0.2, 0.2)], individuals with LP/P [− 0.5 (− 0.7, − 0.2)], p.Val142Ile [− 0.5 (− 0.7, − 0.3)], and non-Val142Ile variants [− 0.5 (− 0.7, − 0.1)] had a lower plasma TTR level (Additional file 1: Table S7).

After adjusting for potential confounders, compared to noncarriers, TTR LP/P carriers were at a higher risk of developing incident AF [HR, 1.54; 95% CI (1.03, 2.29); p = 0.034], bradyarrhythmias [HR, 1.80; 95% CI (1.20, 2.71); p = 0.005], and cardiac block [HR, 1.90; 95% CI (1.23, 2.95); p = 0.004], but not SVA [HR, 1.42; 95% CI (0.95, 2.12); p = 0.084] or VA [HR, 1.60; 95% CI (0.79, 3.25); p = 0.191] (Table 3). Further analyses classified LP/P variants into p.Val142Ile, and non-Val142Ile variants suggested that the associations between LP/P variants with cardiac arrhythmias were mainly driven by non-Val142Ile variants. Specifically, carriers of non-Val142Ile variants had significantly increased risks of incident AF [HR, 2.31; 95% CI (1.24, 4.29); p = 0.008], as well as secondary outcomes including SVA [HR, 2.15; 95% CI (1.16, 3.99); p = 0.016], bradyarrhythmias [HR, 2.48; 95% CI (1.24, 4.97); p = 0.010], and cardiac block [HR, 3.29; 95% CI (1.65, 6.59); p < 0.001], but not VA (HR, 1.94; 95% CI (0.49, 7.77); p = 0.348). No significant associations were observed between p.Val142Ile and any arrhythmia outcome (all p > 0.05) (Table 3).

Table 3.

The associations between TTR variants with incident arrhythmias

	Event/number	TTR LP/P variants HR [95% CI], p-value	p.Val142Ile variant HR [95% CI], p-value	Non-Val142Ile LP/P variants HR [95% CI], p-value
Primary outcome
Atrial fibrillation	30,244/434,710	1.54 [1.03, 2.29], p = 0.034	1.25 [0.75, 2.09], p = 0.400	2.31 [1.24, 4.29], p = 0.008
Secondary outcomes
Supraventricular arrhythmias	31,752/433,611	1.42 [0.95, 2.12], p = 0.084	1.15 [0.69, 1.93], p = 0.600	2.15 [1.16, 3.99], p = 0.016
Bradyarrhythmias	20,438/439,296	1.80 [1.20, 2.71], p = 0.005	1.57 [0.95, 2.59], p = 0.078	2.48 [1.24, 4.97], p = 0.010
Cardiac block	15,728/440,507	1.90 [1.23, 2.95], p = 0.004	1.49 [0.86, 2.60], p = 0.200	3.29 [1.65, 6.59], p < 0.001
Ventricular arrhythmias	6440/441,069	1.60 [0.79, 3.25], p = 0.191	1.51 [0.67, 3.42], p = 0.300	1.94 [0.49, 7.77], p = 0.348

Multivariable-adjusted Cox models were used to estimate the association between TTR variants with arrhythmia outcomes, based on the fully adjusted model, adjusting for age, sex, race, body mass index, smoking status, drinking status, total cholesterol, estimated glomerular filtration rate, triglyceride, low-density lipoprotein cholesterol, heart failure, diabetes, hypertension, coronary artery disease, valvular disease, chronic kidney disease, obstructive sleep apnea, antihypertensive medication, and cholesterol-lowering medication

Bold indicates p-value < 0.05

Abbreviations: TTR transthyretin, HR hazard ratio, CI confidence interval

Discussion

To the best of our knowledge, this is the first study investigating the associations between circulating TTR levels and the risk of cardiac arrhythmias in a population-based cohort, which uncovers a potential role for TTR in the development of arrhythmias and provides novel insights into the underlying mechanisms and the potentially modifying effects of BMI.

Several novel findings should be noted. First, lower plasma TTR levels were associated with a higher risk of incident AF in the study population, which was further supported by the significant correlation between decreased TTR levels with larger atrial volumes. Second, a significant association between lower TTR levels with increased SVA risk was identified. There was no significant association between TTR levels with bradyarrhythmias, cardiac block, or VA in the overall cohort. Third, there was an interaction effect between TTR levels and BMI for the risk of cardiac arrhythmias, with lower TTR levels being significantly associated with higher risk of AF, SVA, bradyarrhythmias, and cardiac block among individuals with a BMI < 25 kg/m². Finally, carriers of TTR LP/P variants were observed to have lower levels of plasma TTR compared with noncarriers. The associations between LP/P variants with higher arrhythmia risk were mainly driven by non-Val142Ile variants (Graphical abstract).

Prior studies also identified a significant role of low circulating TTR levels in predicting disease risks and unfavorable prognosis, whereas these studies primarily focused on individuals who had already developed cardiac amyloidosis [10], as well as outcomes such as HF and mortality [1, 5, 6]. Among individuals with wild-type cardiac amyloidosis, Hanson et al. demonstrated that low baseline TTR concentrations predict overall survival, and serial TTR measurements provide an indication of disease progression [10]. In the general population, recent studies have identified a significant association between low TTR levels with increased risk of incident HF, cardiovascular disease, atherosclerotic cardiovascular disease, cardiovascular mortality, and all-cause mortality among UK Biobank participants [1], as well as higher risk of incident HF and all-cause mortality among Danish general population [5, 6]. The current study adds to the literature by examining the associations between TTR levels with cardiac arrhythmias. Consistent with prior literature, in this study, low TTR levels were found to be associated with an increased risk of new-onset AF in healthy individuals. Specifically, each SD decrease in TTR conferred a 6% higher AF risk. Notably, individuals in the lowest TTR tertile had an 11% greater risk than those in the highest tertile. Although the effect size per unit change in TTR is small, our findings support TTR’s potential utility for population-level risk stratification.

The absence of associations between TTR levels with bradyarrhythmias, cardiac block, and VA in the overall study population may be, at least partially, due to the limited statistical power from small sample sizes and low event counts. Notably, in our study, Kaplan–Meier analyses revealed significant survival differences for these outcomes, and RCS analysis showed a steeper negative slope for lower TTR levels in relation to bradyarrhythmias and cardiac block. Additionally, significant associations were observed between TTR LP/P variants with both bradyarrhythmias and cardiac block risk, suggesting a potential role of low TTR levels in the pathogenesis of these arrhythmias.

In our genetic analysis, carriers of TTR LP/P variants, including both p.Val142Ile and non-Val142Ile carriers, had lower levels of plasma TTR compared with noncarriers, which is in line with a previous UK Biobank study and further supports our primary findings [1]. Additionally, our study suggests that among individuals with TTR LP/P variants, only those carrying non-Val142Ile variants exhibited a significantly increased risk of cardiac arrhythmias. In a prior UK Biobank study, Aung et al. similarly found that non-Val142Ile carriers had significantly higher risks of AF and all-cause mortality, but not p.Val142Ile carriers [19]. However, in a recent study of 77,767 self-identified Black individuals (2.8% with p.Val142Ile variant), the p.Val142Ile variant was associated with increased AF risk [32]. This discrepancy may be attributable to differences in genetic background of the study population, as the UK Biobank is predominantly of European ancestry [19], whereas the latter study focused mainly on individuals of African ancestry [32]. Large-scale genetic studies in diverse ancestral populations are warranted to elucidate the ethnic-specific effects of TTR variants on AF risk.

The mechanisms underlying the associations between low TTR levels with arrhythmias remain incompletely understood. Several hypotheses may help explain the observed associations. On one hand, as the main culprit of cardiac amyloidosis, an accelerated excretion of TTR can be anticipated in a state with destabilized tetramers, thus reducing the half-life of TTR. A reduced half-life of TTR and a concomitant tissue deposition of amyloid fibrils may lead to a reduction in circulating TTR levels [5]. This is supported by our genetic analyses, which showed that carriers of TTR LP/P variants had significantly lower plasma TTR levels and higher arrhythmia risk compared to noncarriers. Amyloid fibril deposition of the myocardium has been demonstrated to cause atrial and ventricular wall thickening, impaired relaxation, and restrictive filling [7, 33, 34]. Elevated filling pressure then leads to atrial dilation and predisposition to AF [7, 34, 35]. This pathophysiological link between amyloid deposition and atrial remodeling is supported by our CMR sub-study findings, which demonstrated an inverse association between plasma TTR levels and atrial volume. However, when interpreting these CMR-derived results, it is crucial to acknowledge that participants included in the CMR sub-study differed from those without CMR data in several baseline characteristics. These differences raise the possibility that the observed relationships in the CMR sub-study might not be fully representative of the entire cohort, though potential confounders have been adjusted. Nevertheless, despite this limitation, our study represents, to our knowledge, the first investigation of the association between plasma TTR concentrations and CMR measurements in a relatively large sample, which may provide some mechanistic insights into the associations between TTR concentrations with cardiac remodeling. Future large-scale imaging studies are warranted to validate and extend our results.

In addition, the toxic inflammatory effect of amyloid fibrils on cardiomyocytes results in further fibrosis and structural remodeling that potentiates AF [36]. Amyloid fibrils infiltration of the conduction system and ventricular myocardium can lead to conduction diseases and VA [7]. On the other hand, an increase in oxidative stress and inflammation may promote the destabilization of the TTR protein, thus making it possible that decreased TTR levels might be a “by-product” marker of preexisting subclinical pathology [1, 37]. The mechanism driving a reduction in circulating TTR may also contribute to the increased risk of cardiac arrhythmias, even though potential confounders have been adjusted in our study.

In general, obesity is one of the common risk factors for AF and other arrhythmias [38]. However, this study identified that lower TTR levels are associated with a higher risk of AF, SVA, bradyarrhythmias, and cardiac block among individuals with lower BMI. Though statistical significance of interactions for bradyarrhythmias and cardiac block was not maintained after FDR correction, the stronger inverse associations in the low-BMI subgroup were significant across these four arrhythmia endpoints through steeper negative exposure–response slopes, which suggests a potentially modifying effect of BMI on the associations between TTR levels with cardiac arrhythmias. Previous study has also observed a positive association between circulating TTR levels with BMI [1]. In addition, prior studies have noted that BMI is significantly lower in patients with transthyretin cardiac amyloidosis, particularly owing to the p.Val142Ile variant, compared with patients with non-amyloid HF [39–42]. Beyond overt cardiac amyloidosis, carriers of the p.Val142Ile variant are more likely to have a BMI < 25 kg/m² than noncarriers [43], and the BMI is similar between patients with transthyretin cardiac amyloidosis and p.Val142Ile carriers without clinically evident cardiac involvement [42]. However, in our study, individuals harboring non-Val142Ile LP/P variants presented with a lower BMI than other variants and manifested significantly increased cardiac arrhythmia risk.

Taken together, these findings suggest that the change of TTR levels and BMI in certain population groups may be the consequence of specific variants, and such patients may experience subclinical deposition in either the heart or gastrointestinal tract, leading to malabsorption and similar changes in BMI [42]. In other words, individuals with lower circulating TTR and lower BMI may be more likely to harbor destabilizing variants in TTR, thereby predisposing them to higher risk of arrhythmias, which may help explain the interaction effects of BMI on the association between TTR with arrhythmias. Nevertheless, this explanation remains speculative at this stage, and future investigations are warranted to elucidate the specific variant type and molecular mechanisms underlying these observations.

In addition, our study also found significant interactions by age and PRS on the associations between TTR with incident AF. Specifically, there was a pronounced association between circulating TTR with AF among older individuals, which is not surprising considering that advancing age represents a risk factor for tetramer destabilization [44]. This finding suggests that older individuals may be more prone to develop AF when expose to lower levels of TTR, compared to younger individuals. Additionally, the association between TTR and AF was more prominent among individuals with low PRS, which was in line with previous studies focusing on the associations between other risk factors and AF [28, 29].

Clinical implications

The findings of our study have important clinical implications. First, since AF is the most common sustained arrhythmia and is associated with considerable mortality and morbidity from stroke, heart failure, dementia, and hospitalizations [45], identifying risk factors for AF has been an ongoing mission to facilitate the risk stratification for AF [29]. Our study suggests a potential mechanistic association between low circulating TTR concentration as a marker of tetramer instability and incident AF in community-dwelling adults. Incorporation of TTR levels into screening programs or risk stratification may help improve AF risk prediction. Serial TTR measurements may serve as a dynamic biomarker to identify escalating risk trajectories and guide targeted interventions. In addition, individuals with low circulating TTR levels and low BMI may benefit from screening for cardiac arrhythmias and close monitoring for future arrhythmia risks.

Limitations

Several limitations should be noted in our study. First, the possibility of a selection bias exists due to the exclusion of individuals with missing data. For instance, participants included in the CMR sub-study differed from those without CMR data in several baseline characteristics, potentially introducing selection bias, which is a common challenge in resource-intensive imaging studies [31]. Second, due to the nature of observational study, the causative link between TTR and AF cannot be determined. Besides, we cannot totally exclude any residual confounding. Nevertheless, we have adjusted for potential confounders and conducted comprehensive subgroup analyses and sensitivity analyses. Third, as UK Biobank participants are predominantly of European ancestry, our findings may have limited generalizability to other populations where TTR biology or arrhythmia susceptibility may differ. Future large-scale studies are needed to validate these associations in diverse populations before broader conclusions can be drawn. Fourth, Olink-derived TTR levels in UK Biobank were reported in NPX units, a relative quantitative value, which complicates translation to routine care. Besides, direct comparison between Olink-measured TTR levels and those from traditional clinical assays has not yet been established. Future studies are needed to establish the clinical significance of absolute TTR changes and to validate the concordance between Olink-derived and conventional assay methods for TTR. Finally, the TTR levels and CMR measurements were obtained at different time points. Although we have adjusted for the time interval and found no interaction effects, the impact of this on the findings remains possible. While our findings are hypothesis-generating, future validation in an external cohort with contemporaneous data collection is warranted.

Conclusions

Lower circulating TTR concentrations were associated with higher risk of incident AF. Exposure to low TTR and low BMI may be associated with a higher risk of AF, SVA, bradyarrhythmias, and cardiac block.

Abbreviations

ATTR-CM

Transthyretin amyloid cardiomyopathy

AF

Atrial fibrillation

SVA

Supraventricular arrhythmias

VA

Ventricular arrhythmias

BMI

Body mass index

TTR

Transthyretin

LP/P

Likely pathogenic or pathogenic

CMR

Cardiac magnetic resonance

WES

Whole exome sequencing

VUS

Variants of uncertain significance

eGFR

Estimated glomerular filtration rate

LDL-C

Low-density lipoprotein cholesterol

VHD

Valvular heart disease

Authors’ contributions

NZ conceptualized and designed the study; TL revised the analytic design and provided methodological support; NZ, ZJ, JZ1, JL, XH statistically analyzed the data and developed the data visualizations; NZ and JZ1 drafted the manuscript; TL acquired funding and had full access to all the data used in present study. GT, JZ2, GYHL, KC critical reviewed of the manuscript; TL and GYHL. provided supervision. All authors read and approved the final manuscript.

Funding

This study was supported by grants from the Noncommunicable Chronic Diseases-National Science and Technology Major Project (2024ZD0521900), the National Natural Science Foundation of China (82170327, 82370332 to T. L.), the Tianjin Key Medical Discipline (Specialty) Construction Project (TJYXZDXK-3-006B, TJWJ2022XK013), and the Hong Kong Metropolitan University (Research Impact Fund to G. T.: RIF/2022/2.2).

Data availability

The dataset supporting the conclusions of this article is available on the website: https://www.ukbiobank.ac.uk/. Access to the UK Biobank resource can be obtained via an approved application.

Declarations

Ethics approval and consent to participate

The North West Multi-Centre Research Ethics Committee Study approved the UKB study (No. 11/NW/0382), and all participants provided written informed consent. This study was conducted under UK Biobank application number 79146.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.