Gut microbe-generated metabolite trimethylamine N-oxide and risk of abdominal aortic aneurysm: a cohort study

Abstract

Background and Aims

Abdominal aortic aneurysms (AAAs) are associated with increased mortality in older adults. The gut microbe-generated metabolite trimethylamine N-oxide (TMAO) has been linked to AAA risk and promotes AAA progression in animal models. Whether circulating TMAO levels in apparently healthy older adults predict AAA development and adverse AAA outcomes remains unknown.

Methods

Plasma TMAO levels were quantified using stable isotope dilution liquid chromatography–tandem mass spectrometry in 4442 community-dwelling adults (aged ≥65 years) in the Cardiovascular Health Study. Participants underwent ultrasound screening and prospective follow-up. Multivariable models assessed associations of serial TMAO levels with AAA development and incident risk for adverse AAA events, adjusting for traditional risk factors, renal function, socioeconomic status, and cardiometabolic lifestyle factors.

Results

Higher baseline TMAO levels were associated with larger infrarenal aortic diameter and increased AAA risk. Over a median 12.2-year follow-up (54 402 person-years), 79 participants experienced incident adverse AAA events (repair, rupture, or AAA-related death). Elevated TMAO levels were independently associated with higher risk of adverse AAA events, whether modelled continuously [adjusted hazard ratio (HR) 1.28, 95% confidence interval (CI) 1.07–1.54 per doubling], by quantiles (HR for tertile 3 vs tertile 1: 2.46, 95% CI 1.32–4.59), or using a clinical threshold (≥6.2 μM; HR 1.93, 95% CI 1.25–2.99).

Conclusions

In community-based older adults, higher TMAO levels were independently associated with increased risk of AAA development and adverse AAA events. Findings support TMAO as a novel risk factor for AAA and a potential therapeutic target for AAA prevention.

Structured Graphical Abstract

Key Question

Do gut microbial trimethylamine N-oxide (TMAO) levels in apparently healthy older adults predict both increased incident risk of abdominal aortic aneurysm (AAA) and adverse AAA outcomes?

Key Finding

In this observational study involving community-dwelling older adults in the Cardiovascular Health Study, higher circulating levels of TMAO were independently associated with increased risk for AAA and adverse AAA events including surgical repair, rupture, or AAA-related death.

Take Home Message

There are currently no therapeutic interventions to slow the progression of AAA disease. These findings support TMAO as a novel risk factor for AAA and adverse aortic outcomes. Randomized controlled trials targeting reduction of TMAO levels as a treatment for AAA prevention might be considered.

Introduction

Abdominal aortic aneurysm (AAA) development can lead to rupture^1-3 and is a significant cause of death in any age group, with mortality rate of a ruptured AAA ranging from 40% to over 80%.^4,5 AAAs typically remain asymptomatic until they rupture, underscoring the importance of early detection and risk stratification. Aggressive global preventive efforts (e.g. blood pressure control, lipid-lowering therapy, smoking cessation, and consideration of antiplatelet therapy) are recommended for those with AAA.⁶ Nonetheless, surgical or percutaneous interventions are the only definitive treatments offered for AAA and are performed when the aorta diameter reaches a threshold which the risk for aortic rupture significantly increases.^6,7 Thus, novel preventive and therapeutic strategies are in urgent need to better manage AAA risk.

We have found that the gut microbial trimethylamine N-oxide (TMAO) pathway can serve as a novel contributor to incident cardiovascular disease (CVD) risks,^8,9 and is mechanistically linked to thrombotic and cardiovascular processes.^8,9,10,11 In recent studies, elevated levels of TMAO were associated with AAA risk among participants with high atherosclerotic risk factors undergoing cardiovascular risk assessment,¹² as well as among participants undergoing image surveillance monitoring for aorta dilation and ectasia.¹³ Atherosclerosis is the most common risk factor for AAA development.^14,15 In multiple animal models, the TMAO pathway was shown to be both mechanistically linked to AAA progression and to serve as a pharmacological target for preventing aneurysm growth and rupture.¹² As the gut microbiota TMAO pathway is also modifiable by dietary interventions,^16,17 the gut microbial TMAO pathway could be a modifiable target to reduce AAA development and progression. To date, all investigations of TMAO pathway and AAA have focused on higher-risk cohorts—those cohorts with established CVD and/or with existing mild dilation undergoing serial imaging surveillance monitoring. Studies in a community-based cohort, however, have potential to further improve the quality of findings and offer further advantages. For example, they can increase the relevance and generalizability of findings, enhance the representativeness of observations, and permit a deeper understanding of real-world health determinants. Thus, before interventional studies are undertaken, it is important to understand whether TMAO levels are associated with AAA risks in general populations as clinical populations undergoing risk assessment may have reverse causation and selection bias. Moreover, it remains unclear whether circulating TMAO levels are prospectively associated with risk of AAA development and adverse AAA outcomes in the general population. Herein we investigate whether repeated circulating TMAO levels are associated with the risk for AAA development and incident adverse AAA-related events in a well-characterized, community-based cohort of older adults.

Methods

Study design and population

The Cardiovascular Health Study (CHS) is a prospective, multicentre, community-based cohort study designed to investigate risk factors for CVD in older adults (ClinicalTrials.gov NCT00005133).^18,19 Women and men ≥65 years of age were recruited from Medicare-eligible residents in four US communities: Sacramento County, California; Washington County, Maryland; Forsyth County, North Carolina; and Allegheny County, Pennsylvania–using systematic sampling with a random start.¹⁸ In the initial cohort, 5201 participants were enrolled in 1989 to 1990, followed by an additional 687 predominantly Black participants recruited in 1992 and 1993. In the present study, baseline and follow-up plasma TMAO levels were measured in 5019 participants from samples collected in 1989 to 1990 (CHS visit year 2) for the original CHS cohort (N = 4407) and in 1992 to 1993 (CHS visit year 5) for the minority cohort (N = 612). Additional TMAO measurements were also performed on all subjects in 1996–97 (CHS visit year 9 examination). Supplementary data online, Figure S1 illustrates the timing of TMAO measurements, the baseline AAA ultrasound screening study, and the frequency, duration and type of ongoing follow-up. AAA-related phenotypes and outcomes in the present analyses used the same criteria as outlined previously for AAA studies in CHS.^20,21,22 Participants in analyses comparing baseline TMAO level with aorta dimensions were excluded if they reported antibiotic use within 2 weeks prior to baseline blood draw (N = 148), had a history of AAA (N = 40), or lacked ultrasound image data (N = 870) from the AAA screening examination performed in CHS year 5 visit in 1992 to 1993 (Figure 1). After these exclusions, 3961 participants were included in the analysis of baseline TMAO levels in relation to AAA risk detected by ultrasound (Figure 1A). In separate time-to-event analyses, the association between repeat TMAO measurements and incident AAA-related adverse events (AAA repair, AAA rupture, or AAA-related death) were examined during longitudinal follow-up (Figure 1B, Supplementary data online, Figure S1). For these analyses, participants had to have at least one TMAO measurement prior to an adverse AAA event (N = 5418), including 5019 with baseline TMAO and 399 without baseline TMAO but with follow-up TMAO measured in 1996 to 1997. Participants were excluded if they had recent antibiotic use (within 2 weeks; N = 89), prior AAA repair or rupture (N = 9), or lacked longitudinal follow-up after aortic imaging (i.e. did not undergo ultrasound scan in 1992–93; N = 878). A final analytic sample of 4442 participants (Figure 1B) was included, of which 4401 participants underwent AAA ultrasound screening in 1992 to 1993, and an additional 41 were included in the analysis who missed the ultrasound screening in 1992 to 1993 but were monitored through follow-up (2 with AAA repair, 1 with rupture, and 38 without adverse AAA events) (Figure 1B). Sample sizes were determined by available TMAO measurements and outcome data after application of pre-specified inclusion and exclusion criteria (Figure 1). All procedures conform to the STROBE guidelines for observational studies.

Figure 1.

Flowchart of participant inclusion and exclusion criteria. (A) Participant inclusion and exclusion criteria for analyses comparing baseline TMAO levels and AAA detected by ultrasound imaging. Baseline TMAO levels were measured from CHS participants from their 1989–1990 visit for the original cohort and 1992–1993 visit for the minority cohort. AAA screening was performed in 1992 to 1993. (B) Participant inclusion and exclusion criteria for analyses comparing repeated TMAO levels and incident adverse AAA events (surgical AAA repair, AAA rupture, or AAA-related death). TMAO, trimethylamine N-oxide; AAA, abdominal aortic aneurysm; US, ultrasound; F/U, follow-up

Quantification of TMAO by mass spectrometry

To assess long-term exposure, repeat plasma TMAO levels were measured in all CHS participants with available stored blood samples at baseline [1989–90 (N = 4407, 84.7% of living participants); 1992–93 (N = 612; 89.1%)] and at a subsequent examination in 1996–97 (N = 3393; 76.9%). At these visits, 12-h fasting plasma blood samples were collected using ethylenediaminetetraacetic acid tubes and immediately processed and stored at −80°C until analysis. Plasma TMAO levels were quantified using stable isotope dilution online electrospray ionization liquid chromatography–tandem mass spectrometry (LC–MS/MS). Chromatographic separations were performed using an ultra-high performance liquid chromatography system coupled to a Shimadzu triple quadrupole mass spectrometer (LCMS-8050 and LCMS-8060; Shimadzu Scientific Instruments, Columbia, MD) at the Cleveland Clinic, as previously described.^8,9,23 Multiple quality-control samples of varying TMAO levels were run with every batch of samples (81 participant samples per batch), and all interday and intraday laboratory coefficients of variation were <6% for TMAO at both baseline and subsequent follow-up samples.

Ascertainment of AAA outcomes

B-mode greyscale ultrasonographic imaging of the abdominal aorta was performed in the longitudinal and transverse projections by trained ultrasonographers, as previously described.^20,21,22 The suprarenal measure of aortic diameter was taken 1 cm distal to the origin of the superior mesenteric artery, just above the level of the left renal artery. The infrarenal measure of aortic diameter was determined by the site of the maximum diameter of the aorta (lumen plus wall) below the renal arteries. In the CHS cohort, AAA was defined at the 1992 to 1993 ultrasound as an infrarenal aortic diameter ≥3.0 cm or an infrarenal-to-suprarenal ratio ≥1.2, or based on a documented AAA repair or rupture occurring after enrolment and before the ultrasound scan in 1992 to 1993.^20,21,22 Incident adverse AAA events were defined as a composite of AAA surgical repair, AAA rupture, or AAA-related death. Repair was identified through review of all relevant medical records and International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) procedure codes 38.44, along text fields indicating repair or resection of an AAA. Procedures explicitly involving aneurysms in alternative regions of the aorta (ascending, arch, thoracic) were excluded. AAA rupture was identified using ICD-9-CM diagnosis code 441.3 together with supporting documentation indicating a ruptured AAA. All participants had adequate clinical information in hospital records and discharge summaries to confirm AAA rupture or repair. AAA-related death was determined through the CHS cause-of-death adjudication process, which involved a detailed review of death records to confirm deaths directly attributable to AAA and to exclude deaths owing to ascending, arch, or thoracic aneurysm.²² Time to incident adverse AAA event (repair, rupture, or AAA-related death) was calculated from the date of ultrasound scan (1992–93) to the earliest qualifying event. Participants with multiple events were classified according to their first event. Those who experienced AAA repair or AAA rupture prior to the scan were excluded from time-to-event analyses and considered to have a prior history of repair or rupture. In addition to those who underwent ultrasound screening in 1992 to 1993, a small subset of participants (N = 41; three events) who did not receive imaging were also included in the analysis because they were monitored for adverse AAA events through clinical follow-up and confirmation of medical records.

Covariates

Details of data collection in CHS have been previously described, and covariates were harmonized using standardized protocols.¹⁸ Information on baseline clinical characteristics and lifestyle factors was collected at study clinic visits to assess participants’ sociodemographic characteristics, lifestyle, medical history, other health-related phenotypes, and clinical events. Medication use was assessed by a validated medication inventory. Plasma concentrations of glucose, insulin, lipids, and inflammatory biomarkers were assessed using enzymatic methods. Body mass index (BMI) was calculated by directly measuring body weight in kilograms divided by height in meters squared. Prevalent CVD in the CHS cohort was defined as previously reported.^24,25 The estimated glomerular filtration rate (eGFR) was calculated with an equation from the Chronic Kidney Disease Epidemiology Collaboration using serum creatinine and cystatin C concentrations.²⁶ Chronic kidney disease was defined as eGFR <60 mL/min/1.73 m². Physical activity was assessed at baseline (1989–90 or 1992–93) and in 1996 to 1997 using a modified Minnesota Leisure-Time Activities questionnaire.²⁷ Usual dietary habits were assessed in 1989 to 1990²⁸ using a 99-item validated food frequency questionnaire (National Cancer Institute) and in 1996 to 1997²⁹ using a validated 131-item food frequency questionnaire.³⁰ The Alternative Healthy Eating Index was calculated to evaluate diet quality.³¹

Statistical analysis

Two complementary analyses were conducted. First, we examined the association between baseline plasma TMAO levels and AAA detected by ultrasound imaging at the 1992–93 follow-up visit (Figure 1A). Second, we evaluated the prospective association between serial plasma TMAO levels and the risk of incident adverse AAA events, defined as AAA surgical repair, AAA rupture, or AAA-related death over longitudinal follow-up (Figure 1B). Continuous variables were presented as mean ± standard deviation for approximately normally distributed data and as median [interquartile range (IQR)] otherwise. Comparisons between two groups were performed using the t-test or the Wilcoxon rank-sum test, respectively. For multiple pairwise comparisons of ranked data, the Dunn test was applied; trends across ordered groups were evaluated using the Jonckheere–Terpstra test for continuous variables and the Cochran–Armitage trend test for categorical variables. TMAO levels were analysed as continuous variables, by quantiles, and using a clinically relevant cut-off (≥6.2 μM), corresponding to the 75th percentile in a prior large clinical cohort study and currently used in practice to assess CVD risk.^9,11

For ultrasound-detected AAA, odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using logistic regression. The linearity of quantitative predictors was assessed using the spline-based diagnostic approach described below. Confounders were selected a priori based on the disjunctive cause criterion,³² identifying variables associated with both TMAO levels and AAA. Models were fit using a staged approach: Model 1 was adjusted for age, sex (male vs female), race (White vs non-White), smoking status (never, former, or current), diabetes, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, triglycerides, C-reactive protein (CRP), systolic and diastolic blood pressure; Model 2 additionally adjusted for renal function (eGFR <60 mL/min/1.73 m²); and Model 3 provided full adjustment for enrolment site, education, self-reported income, self-reported health status, alcohol intake, physical activity, BMI, waist circumference, antiplatelet medication, lipid-lowering medication, antihypertensive medication, and the Alternate Healthy Eating Index. Given the low frequency of AAA in this cohort, ORs were interpreted as approximating risk ratios (RRs). In sensitivity analyses, RRs were estimated using modified Poisson regression with robust error variance.³³ Missing covariate data (<5%–10% for most variables at baseline; 1.7%–20.2% at the 1996–97 visit) were imputed using single imputation via best-subset regression incorporating demographic, lifestyle, and clinical risk variables, an approach previously shown in CHS to yield results comparable to multiple imputation.³⁴

For survival analyses, time zero was defined as the date of the 1992–93 ultrasound examination, when participants entered the risk set. Baseline exposure was defined as the first available, valid TMAO measurement obtained prior to any clinical event. For a small subset of participants missing initial TMAO data or reporting antibiotic use in the 2 weeks prior to the first blood draw (N = 433), the next available TMAO measurement was used as the baseline value. Participants who experienced an AAA-related event prior to the 1992–93 scan were excluded to ensure all participants were event-free at the start of follow-up. Participants were followed until the first occurrence of an adjudicated incident AAA event, death from any cause other than AAA, or the end of the study period (June 2015).

Associations between TMAO and incident adverse AAA events were estimated using time-dependent Cox proportional hazards regression models. TMAO levels were modelled as a time-varying exposure using a cumulative average approach: baseline TMAO was related to risk of incident adverse AAA events until the second measurement (1996–97), after which the cumulative average was related to subsequent risk. For participants without a second measurement (~39%), the baseline measurement contributed to risk estimation throughout follow-up.^{35,36,37,38,39} Primary analyses used TMAO tertile categories to capture potential nonlinear dose–response relationships. Multivariable adjustment was analogous to that used in the logistic regression models, with covariates treated as time-varying where appropriate to minimize residual confounding. The proportional hazards assumption was assessed using Schoenfeld residuals and was not violated. In primary analyses, non-AAA-related death was treated as a censoring event in cause-specific Cox models. In sensitivity analyses, the competing risk of non-AAA-related death was evaluated using Fine–Gray subdistribution hazard models,⁴⁰ and given the relatively low events-per-parameter ratio in the fully adjusted model, ridge-penalized Cox regression was performed to assess the robustness of the findings.⁴¹ To explore potential nonlinear relationships and assess the linearity assumption for quantitative predictors, restricted cubic spline models were fitted within adjusted multivariable models, with three knots placed at the 10th, 50th, and 90th percentiles of the TMAO distribution.⁴² No meaningful deviation from linearity was observed; therefore, results are presented as per-doubling (log₂-transformed) effects. The P-value for linearity corresponds to TMAO modelled as a linear term. The P-value for nonlinearity was obtained from a likelihood ratio test comparing the linear model with the spline model. Predictive performance was evaluated using the time-dependent area under the receiver operating characteristic curve, net reclassification improvement (NRI) and integrated discrimination improvement (IDI). To account for right censoring, all metrics were estimated using inverse probability of censoring weighting,⁴³ with NRI and IDI calculated using methods for censored survival data.⁴⁴ All statistical analyses were performed using R software (version 4.5.1). A two-sided P-value of <.05 was considered statistically significant.

Results

Characteristics of participants

Figure 1 summarizes the inclusion and exclusion criteria for participants from the CHS cohort in our studies. First, we evaluated the association between baseline TMAO levels and AAA risk based on AAA ultrasound screening in 1992 to 1993 (3 years after baseline sample collection for >84% of participants). In total, 3961 participants had baseline TMAO measurements, no recent antibiotic use at the time of blood draw, and subsequent infrarenal aortic diameter assessment by ultrasound (Figure 1A). Second, we examined the association between time-varying TMAO levels and incident adverse AAA events (repair, rupture, or AAA-related death) among 4442 CHS participants who were prospectively followed for over a median (IQR) of 12.2 (7.3–18.3 years) (Figure 1B).

Baseline characteristics of the participants undergoing AAA ultrasound imaging screening in 1992 to 1993 (N = 3961), stratified by AAA status, are shown in Table 1 and Supplementary data online, Table S1. Notably, the majority of participants were women (63.0%). Only 18.1% were Black. AAA was diagnosed by imaging in 335 (8.5%) individuals. Compared with participants without AAA, those with AAA were more likely to be male, current or former smokers, and to have lower levels of HDL cholesterol and eGFR, and higher levels of triglycerides and high-sensitivity CRP (Table 1). Table 2 (and Supplementary data online, Table S2) describes baseline characteristics, laboratory measures, socioeconomic status, and dietary factors of the 4442 participants monitored long term for incident adverse AAA events (repair, rupture, or AAA-related death), stratified by tertiles of baseline plasma TMAO levels. During a median follow-up of 12.2 years (IQR, 7.3–18.3 years, maximum 23.1 years; 54 402 person-years), 79 participants (1.8%) experienced a first adverse AAA event (defined as AAA surgical repair, rupture, or AAA-related death). Higher baseline TMAO levels were observed among participants who were older, male, had a history of diabetes mellitus or smoking, had reduced HDL cholesterol and elevated triglyceride level (Table 2).

Table 1.

Baseline characteristics of CHS participants stratified by AAA status at the 1992–1993 ultrasound screening

Characteristics	All participants (N = 3961)	Participants without AAA (N = 3626)	Participants with AAA (N = 335)	P-value
Age, years	72 ± 5	72 ± 5	72 ± 5	.15
Female sex, n (%)	2495 (63.0)	2348 (64.8)	147 (43.9)	<.0001
Race, n (%)				.62
White	3220 (81.3)	2945 (81.2)	275 (82.1)
Black	716 (18.1)	657 (18.1)	59 (17.6)
Othera	25 (.7)	24 (.6)	1 (.3)
Diabetes mellitus, n (%)	807 (20.4)	735 (20.3)	72 (21.5)	.6
BMI, kg/m2	26.8 ± 4.7	26.7 ± 4.7	27.0 ± 4.6	.28
Systolic blood pressure, mmHg	136 ± 22	136 ± 21	136 ± 23	.78
Diastolic blood pressure, mmHg	71 ± 11	71 ± 11	71 ± 12	.25
Smoking, n (%)				<.0001
Never smoker	1879 (47.4)	1767 (48.7)	112 (33.4)
Former smoker	1638 (41.4)	1473 (40.6)	165 (49.3)
Current smoker	444 (11.2)	386 (10.6)	58 (17.3)
HDL cholesterol, mg/dL	55 ± 16	56 ± 16	49 ± 13	<.0001
LDL cholesterol, mg/dL	131 ± 35	130 ± 35	133 ± 38	.21
Triglycerides, mg/dL	118 (91–161)	117 (90–159)	128 (100–180)	.0001
CRP, mg/L	2.44 (1.23–4.42)	2.40 (1.20–4.34)	2.79 (1.56–6.20)	.0002
eGFR, mL/min/1.73 m2	70 ± 16	70 ± 16	67 ± 17	.0002
Medical history, n (%)
Antiplatelet medication (aspirin) (prior 2 weeks)	131 (3.3)	122 (3.4)	9 (2.7)	.52
Lipid-lowering medication (prior 2 weeks)	101 (2.6)	90 (2.5)	11 (3.3)	.36
Hypertension medication (prior 2 weeks)	1843 (46.6)	1685 (46.6)	158 (47.6)	.72
Baseline TMAO, μM	4.72 (3.20–7.51)	4.69 (3.17–7.45)	5.38 (3.62–8.85)	.0005

Continuous variables are presented as mean ± standard deviation, or median (interquartile range), as appropriate; and n (%) for categorical variables.

BMI, body mass index; CRP, C-reactive protein; eGFR, estimated glomerular filtration rate; TMAO, trimethylamine N-oxide.

The t-test or Wilcoxon rank-sum test was used for continuous variables, and the χ² test or Fisher’s exact test for categorical variables (Fisher’s exact test applied when expected cell counts were <5).

^a‘Other’ includes American Indian/Alaskan Native, Asian/Pacific Islander, and other race/ethnicities.

Table 2.

Baseline characteristics of CHS participants stratified by TMAO tertiles in the incident adverse AAA events analysis

Characteristics	All participants (N = 4442)	Tertile 1 (N = 1480) < 3.72 μM	Tertile 2 (N = 1481) 3.72–6.39 μM	Tertile 3 (N = 1481) 6.40–255 μM	P-value
Age, years	72 ± 5	72 ± 5	72 ± 5	73 ± 5	<.0001
Female sex, n (%)	2673 (60.2)	991 (67.0)	862 (58.2)	820 (55.4)	<.0001
Race, n (%)					.01
White	3635 (81.8)	1169 (79.0)	1219 (82.3)	1247 (84.2)
Black	779 (17.5)	301 (20.3)	253 (17.1)	225 (15.2)
Othera	28 (.6)	10 (.6)	9 (.6)	9 (.6)
Diabetes mellitus, n (%)	913 (20.6)	260 (17.6)	309 (20.9)	344 (23.2)	.0001
BMI, kg/m2	26.8 ± 4.7	26.5 ± 4.8	26.8 ± 4.6	27.0 ± 4.7	.003
Systolic blood pressure, mmHg	136 ± 22	136 ± 21	135 ± 21	136 ± 22	.68
Diastolic blood pressure, mmHg	71 ± 11	71 ± 11	71 ± 11	70 ± 11	.12
Smoking, n (%)					<.0001
Never smoker	2095 (47.2)	754 (50.9)	698 (47.1)	643 (43.4)
Former smoker	1856 (41.8)	550 (37.2)	635 (42.9)	671 (45.3)
Current smoker	491 (11.1)	176 (11.9)	148 (10.0)	167 (11.3)
HDL cholesterol, mg/dL	55 ± 16	57 ± 16	55 ± 15	54 ± 16	<.0001
LDL cholesterol, mg/dL	130 ± 35	131 ± 36	132 ± 34	129 ± 35	.29
Triglycerides, mg/dL	118 (91–161)	115 (88–155)	119 (92–157)	123 (94–170)	<.0001
CRP, mg/L	2.44 (1.23–4.41)	2.27 (1.11–4.10)	2.47 (1.32–4.50)	2.56 (1.29–4.43)	.002
eGFR, mL/min/1.73 m2	70 ± 16	76 ± 14	70 ± 15	64 ± 17	<.0001
Medical history, n (%)
Antiplatelet medication (aspirin) (prior 2 weeks)	146 (3.3)	39 (2.6)	49 (3.3)	58 (3.9)	.05
Lipid-lowering medication (prior 2 weeks)	108 (2.4)	44 (3.0)	33 (2.2)	31 (2.1)	.12
Hypertension medication (prior 2 weeks)	2044 (46.1)	612 (41.5)	675 (45.7)	757 (51.2)	<.0001
TMAO, μM	4.76 (3.22–7.65)	2.72 (2.18–3.22)	4.76 (4.20–5.47)	9.75 (7.65–13.96)	<.0001

Continuous variables are presented as mean ± standard deviation or median (interquartile range), as appropriate; and n (%) for categorical variables.

BMI, body mass index; CRP, C-reactive protein; eGFR, estimated glomerular filtration rate; TMAO, trimethylamine N-oxide.

Trends across TMAO tertiles were assessed using the Jonckheere–Terpstra test for continuous variables and the Cochran–Armitage trend test for binary categorical variables; for categorical variables with more than two levels, global χ² tests were used.

^a‘Other’ includes American Indian/Alaskan Native, Asian/Pacific Islander, and other race/ethnicities.

Baseline plasma TMAO and AAA risk based on ultrasound study in 1992 to 1993

Elevated fasting baseline TMAO levels were associated with larger infrarenal aortic diameter on ultrasound imaging (Figure 2, left; P = .0001). Participants with ultrasound-defined AAA had higher TMAO levels compared with those without AAA (Figure 2, right; P = .0005). When evaluating the association between baseline TMAO and AAA in 1992 to 1993, higher TMAO levels were consistently associated with increased risk of AAA, whether modelled as a continuous variable or in quantiles (e.g. Figure 3, Supplementary data online, Table S3). For instance, compared with participants in the lowest quintile of TMAO levels, participants in the highest quintile of TMAO demonstrated 62% higher odds of AAA compared with those in the lowest quintile (adjusted OR 1.62, 95% CI, 1.11–2.41), after adjustment for age, sex (male vs female), race, smoking (never, former, or current), diabetes, HDL cholesterol, LDL cholesterol, triglycerides, CRP, systolic blood pressure, diastolic blood pressure, and eGFR. This association remained similar in magnitude after additional adjustment for sociodemographic, lifestyle, adiposity, medication, and dietary factors—including study site, education level, income, self-reported health status, alcohol intake, physical activity, BMI, waist circumference, use of antiplatelet agents, lipid-lowering and antihypertensive medications, and dietary patterns (Figure 3). In sensitivity analyses, modified Poisson-derived RRs were consistent with results from standard logistic regression (OR) models, with no material differences in interpretation (see Supplementary data online, Table S4). In exploratory subgroup analyses, the association between TMAO and AAA did not differ meaningfully by traditional CVD risk factors; however, a potential interaction by race was observed, with higher effect estimates among White compared with Black participants (see Supplementary data online, Figure S2). In additional analyses, each doubling of baseline plasma TMAO was associated with an approximate .03-cm larger infrarenal aortic diameter, and when examined using restricted cubic spline analysis, a dose-dependent increase in AAA was observed with higher plasma TMAO levels (see Supplementary data online, Figure S3; P for nonlinearity = .13).

Figure 2.

Association of baseline TMAO levels with AAA screening results obtained in CHS participants. Box-and-whisker plots show baseline TMAO levels stratified by infrarenal aortic diameter and AAA status, AAA was defined as an infrarenal aortic diameter ≥3.0 cm, an infrarenal-to-suprarenal ratio ≥1.2, or based on a documented AAA repair or rupture occurring after enrolment and before the ultrasound scan in 1992 to 1993. Data are presented as box plots: middle line represents the median, bottom and top boundaries represent the 25th and 75th percentiles, and the whiskers represent the 10th and 90th percentiles. P values were calculated with the Kruskal–Wallis (K.W.) test with a Dunn post hoc analysis, Jonckheere–Terpstra test of increasing trend, and Wilcoxon rank-sum test

Figure 3.

Forest plot showing the association between baseline TMAO quintiles and the risk of ultrasound-detected AAA in CHS participants. Multivariable logistic regression model included Model 1 (blue), adjusted for age, sex (male vs female), race (White vs non-White), smoking status (never, former, or current), diabetes, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, triglycerides, C-reactive protein, systolic blood pressure, and diastolic blood pressure; Model 2 (black), adjusted model 1 additionally adjusted for renal function indices (eGFR <60 mL/min/1.73 m²); and Model 3 (red), adjusted model 2 additionally adjusted study site, education, income, self-reported health status, alcohol intake, physical activity, BMI, waist circumference, antiplatelet medication, lipid-lowering medication, anti-hypertensive medication, and alternate healthy eating index. Open circles represent odds ratios, and horizontal lines indicate 95% confidence intervals

Repeat TMAO levels and incident adverse AAA events

We next examined the association between long-term TMAO exposure, assessed from repeated TMAO measurements, and the risk of incident adverse AAA events including AAA surgical repair, AAA rupture, or AAA-related death. Among the 4442 participants with repeat TMAO levels and long-term follow-up, adverse AAA events included AAA repair (N = 53), rupture (N = 25), and AAA-related death (N = 15). Overall, 79 participants experienced at least one adverse AAA event (composite endpoint). Of the remaining participants, 3916 (88.2%) died from causes other than AAA (treated as censored events in cause-specific analyses) and 447 (10.1%) were censored at the end of follow-up. Participants with incident adverse AAA events had higher TMAO levels than those without events (Figure 4, P = .001; Supplementary data online, Figure S4). Subgroup analysis showed similar patterns among White participants and men, with higher TMAO levels observed among those experiencing AAA events, while trends were directionally consistent in Black participants and women (Figure 4).

Figure 4.

The association of time-varying TMAO levels with incident adverse AAA events (AAA surgical repair, AAA rupture, or AAA-related death) in CHS cohort. Box-and-whisker plots of TMAO levels among participants with (Yes) and without (No) incident AAA surgical repair, rupture, or AAA-related death stratified by event status, sex (male vs female), and race over a median follow-up of 12.2 years (IQR, 7.3–18.3, maximum 23.1 years) in the CHS study. Data are presented as box plots: middle line represents the median, bottom and top boundaries represent the 25th and 75th percentiles, and the whiskers represent the 10th and 90th percentiles. P-values were calculated with Wilcoxon rank-sum test

Kaplan–Meier analyses of the cohort stratified by tertiles of baseline TMAO levels and freedom from incident AAA surgical repair, AAA rupture, or AAA-related death showed that participants with higher TMAO were more likely to experience adverse AAA events (Figure 5A; Supplementary data online, Figures S5, S6). In multivariable Cox proportional hazards models, higher time-varying TMAO levels were associated with an increased risk of adverse AAA events (composite of AAA repair, rupture, or AAA-related death) after adjusting for traditional CVD and AAA risk factors, as well as CRP [Model 1 (primary model); Tertile 3 vs Tertile 1 adjusted hazard ratio (aHR) 2.46, 95% CI, 1.32–4.59]. Results were consistent in competing risk analyses using Fine–Gray models [Model 1; subdistribution hazard ratio (SHR) 2.37, 95% CI 1.27–4.42] (see Supplementary data online, Table S5). This association remained similar in magnitude after additional adjustment for eGFR, which may act as a confounder or mediator (Model 2; T3 vs T1 aHR 2.09, 95% CI, 1.09–4.01), and in a fully adjusted model including sociodemographic, lifestyle, adiposity, medication, and dietary factors (Model 3; T3 vs T1 aHR 2.06, 95% CI, 1.06–4.01) (Figure 5B). Given lower events-per-parameter ratio was in this model, results were directionally consistent, with modest attenuation of the effect estimate, in sensitivity analyses using ridge-penalized Cox regression (T3 vs T1 aHR 1.63, 95% CI 1.12–2.36). Notably, similar associations were observed when TMAO was modelled as a continuous variable. For example, for each doubling of TMAO concentration, the risk of AAA events increased by 28% in the adjusted model (Model 1, aHR 1.28; 95% CI, 1.07–1.54) (see Supplementary data online, Table S6). Using a clinically established cut-off of ≥6.2 μmol/L for elevated TMAO, individuals above this threshold had a 93% greater risk of AAA-related events (HR 1.93; 95% CI 1.25–2.99) (see Supplementary data online, Table S7).^9,11

Figure 5.

TMAO levels and likelihood of AAA surgical repair, AAA rupture, or AAA-related death in CHS cohort. (A) Kaplan–Meier estimates and the risk of incident AAA cases ranked by tertile of TMAO levels. (B) Forest plots indicate the risks of incident AAA surgical repair, rupture, or AAA-related death according to the tertiles of baseline TMAO levels. The multivariable Cox model included Model 1(blue), adjusted for baseline age, sex (male vs female), race (White vs non-White), time-varying smoking status (never, former, or current), systolic and diastolic blood pressure, diabetes, dyslipidaemia (high-density lipoprotein cholesterol ≤40 mg/dL, or low-density lipoprotein cholesterol ≥100 mg/dL or triglycerides ≥150 mg/dL), and C-reactive protein (CRP); Model 2 (black), adjusted model 1 additionally adjusted for time-varying renal function indices (eGFR <60 mL/min/1.73 m²); and Model 3 (red), adjusted model 2 additionally adjusted for study site, education, income, time-varying self-reported health status, alcohol intake, physical activity, BMI, waist circumference, antiplatelet medication, lipid-lowering medication, antihypertensive medication, and alternate healthy eating index. Open circles represent hazard ratios, and horizontal lines indicate 95% confidence intervals

AAA-related death is influenced by a broad set of pre- and post-surgical factors, including but not limited to rupture physiology, operative approach, perioperative complications, and competing cardiovascular risk.^45,46,47 To further explore adverse outcomes more directly tied to AAA progression, we also examined the composite of AAA repair or AAA rupture (see Supplementary data online, Figure S5), as well as repair alone (see Supplementary data online, Figure S6). Similar results were observed with each analysis, wherein higher TMAO levels were significantly associated with an increased risk for these AAA progression-specific events. The association between higher TMAO and increased AAA events risk was generally consistent across demographic clinical subgroups (see Supplementary data online, Figure S7). Restricted cubic spline analysis revealed a dose-dependent association between time-varying TMAO levels and risk of incident adverse AAA events, including surgical AAA repair, AAA rupture, or AAA-related death (Figure 6). In further analyses, adding TMAO to a time-to-event model including traditional CVD and AAA risk factors significantly improved time-dependent discrimination for adverse AAA events across yearly horizons (1–23 years). The event-weighted, time-integrated area under the curve increased from .79 (95% CI .77–.82) without TMAO to .82 (95% CI .80–.84) with TMAO. Addition of TMAO to the adjusted model also significantly increased both the continuous time-dependent net reclassification improvement (.31, 95% CI .21–.40) and the event-weighted integrated discrimination improvement (.004, 95% CI .002–.006) (see Supplementary data online, Table S8).

Figure 6.

Relationship between repeated time-varying plasma TMAO levels and incident risk for adverse AAA events in CHS cohort. Restricted cubic spline curves indicating a dose-dependent relationship between plasma time-varying TMAO levels, with upper-tail TMAO values truncated at the 98th percentile to enhance model stability, and the risk of incident adverse AAA events over a median follow-up of 12.2 years (IQR, 7.3–18.3, maximum 23.1 years) in a community-based cohort of 4442 participants in the Cardiovascular Health Study. Spline curves were constructed with knots at the 10th, 50th, and 90th percentiles of the TMAO distribution. P for linearity was the P value for TMAO in a model with TMAO analysed as a linear term. P for nonlinearity was obtained by a likelihood ratio test comparing the model with linear TMAO vs the model with all spline terms for TMAO. Adjusted hazard ratio (red line) and 95% confidence interval (dashed blue line) are shown, the reference value (black dot) corresponds to 10% TMAO level (2.42 μM). Adjusted for baseline age, sex (male vs female), race (White vs non-White), time-varying smoking status (never, former, or current), systolic blood pressure, diastolic blood pressure, diabetes, dyslipidaemia (high-density lipoprotein cholesterol ≤ 40 mg/dL, or low-density lipoprotein cholesterol ≥100 mg/dL or triglycerides ≥150 mg/dL) and C-reactive protein (CRP)

Discussion

In this study, we examined repeated plasma TMAO levels in a large, community-based cohort of older adults followed for a median of 12.2 years. Elevated TMAO levels were associated with both a higher future risk of AAA development and incident adverse AAA events, including AAA surgical repair, AAA rupture, or AAA-related death. The association between TMAO and AAA outcomes remained significant independent of diabetes status, an established inverse association with AAA risk,^47,48,49 as well as other cardiometabolic risk factors. Importantly, these associations remained significant in all models including adjustments for traditional CVD and AAA risk factors and remained robust after additional adjustment, including renal function, socioeconomic, lifestyle, adiposity, medication, and dietary factors (Structured Graphical Abstract).

Screening of high-risk individuals for AAA using imaging is the current strategy for early detection and surgical intervention. However, numerous participants at AAA risk are often missed until a catastrophic adverse AAA event.^4,50 Nonetheless, population-wide screening is associated with high financial costs and may yet still fail to identify individuals who do not meet current high-risk criteria but remain vulnerable to potentially catastrophic outcomes. There is thus a critical unmet clinical need to identify pathways involved in AAA pathogenesis and to develop tools that improve risk stratification. Moreover, current therapeutic strategies to slow AAA progression and prevent rupture are largely limited to lifestyle modification and risk factor management.^6,7 Most AAAs in CHS were identified by protocolized ultrasound screening performed approximately 3 years after cohort enrolment, rather than through routine clinical detection. This window of capture resembles clinical practice patterns. It is recognized that any clinical decision making (initiation of medical therapy or surgery) responds to AAA that develops over many years—sometimes decades. The high prevalence observed in the CHS cohort reflects a combination of both screen-detected AAA (predominant) and cumulative AAA events in an elderly cohort. These numbers reflect real-world expected prevalence in a community-based elderly cohort, are significantly higher to annual incidence estimates from younger, unscreened general populations, and underscore the importance of AAA as a major contributor to mortality risk in the elderly.

In the CHS, a longitudinal, community-based prospective cohort study designed to identify risk factors for development and progression of CVD in older adults, AAAs were associated with an increased risk of incident CVD events.^21,22 Long-term TMAO exposure testing in CHS has recently been reported and shown to be associated with incident risk for CVD development,³⁵ mortality,³⁶ stroke,³⁷ and additional CVD and kidney disease-related outcomes and phenotypes.^38,39 In recent studies, the gut microbiome-derived metabolite TMAO is not only clinically associated with prevalent AAA but also mechanistically linked to AAA expansion and rupture risk in animal models.¹² Furthermore, pharmacological inhibition of gut microbial TMAO production was shown to significantly reduce AAA expansion and rupture in multiple animal models of AAA.¹² In further studies, among high-risk individuals undergoing serial imaging surveillance for aortic dilation or ectasia in both European and US cohorts, elevated levels of the gut microbial metabolite TMAO were also recently reported to be associated with AAA risk, rapid AAA growth, and risk for surgical intervention.¹³ The present results extend the observation linking the gut microbial TMAO pathway to AAA risk to a community-based cohort. CHS, like many community-based studies, benefited from high participant engagement and long-term retention. The design thus reduces selection bias and allows for more comprehensive assessment of social determinants of health. Consequently, we were able to adjust not only for cardiovascular and AAA-related risk factors but also to integrate into the analysis socioeconomic, lifestyle, and environmental factors measured over time (e.g. diet and physical activity). Given that the TMAO pathway can be manipulated through both dietary and therapeutic interventions, specifically by selectively blocking gut microbial trimethylamine production, by extension, TMAO formation,^16,17 this pathway may represent a promising therapeutic target for reducing AAA risk. When combined with the present findings, these observations raise exciting possibilities of leveraging TMAO as a novel biomarker and interventional strategy for the prevention or treatment of AAA.

Currently, surgery or percutaneous interventions with an aortic stent are the only recommended treatments for AAA, typically performed when the aortic diameter reaches a critical threshold that markedly increases the risk of rupture. The present study raises the important question of whether therapeutic, dietary, lifestyle, or pharmacological modulation of the TMAO pathway could improve outcomes in individuals with AAA. We observed a dose-dependent relationship of TMAO and both risk for AAA and incident risks for adverse AAA outcomes, suggesting that even modest reductions in TMAO levels could be beneficial. Prior studies have shown that dietary interventions, such as reducing the amount of red meat consumed¹⁷ or combining a healthier diet with other lifestyle improvements⁵¹ may significantly lower TMAO levels. In addition, some existing cardiovascular therapeutics have also been reported to modestly reduce TMAO levels. For example, we previously observed that TMAO was reduced by ~20% following aspirin treatment.⁵² Rosuvastatin has been proposed to reduce the ability of gut microbes to metabolize dietary precursors of TMAO.^53,54 Beyond this, pharmacological targeting of the gut microbiome with nonlethal small molecule intervention has proven to be very effective in animal models, drastically reducing TMAO levels (e.g. >95% reduction) and significantly blocking TMAO-dependent cardiovascular phenotypes, including AAA,¹² atherosclerosis,^55,56 thrombosis,⁵⁷ and chronic kidney disease.⁵⁸ While such pharmacological targeting of the TMAO pathway in humans is yet to be explored, the current study further supports the rationale for performing human intervention studies to lower TMAO as a potential treatment of AAA.

Potential limitations in the present studies merit mentioning. Although family history is considered a risk factor for AAA, it is generally not considered a strong risk factor as it is for patients with thoracic aortic aneurysms; moreover, data on family history of AAA were not collected in this cohort. Another limitation is that protocolized AAA imaging studies were only performed in all subjects as part of the scheduled in-person 1992–1993 CHS follow-up visit. The absence of serial aortic imaging precluded assessment of aneurysm growth over time, thereby limiting evaluation of TMAO as a predictor of aneurysm expansion rate. However, we note that a recent study exploring high-risk individuals undergoing serial imaging surveillance showed an association between elevated baseline TMAO levels and both aneurysmal expansion rate and surgical risk.¹³ Another potential limitation is while the CHS cohort was community-based, the findings may not be generalizable beyond older White adults. The majority of CHS participants were White, and the limited racial and ethnic diversity may limit the applicability of our results to other populations. However, the present study represents a large, well-characterized cohort of community-dwelling older adults. It is interesting to note that the majority of CHS participants were women, a historically underrepresented group. While TMAO levels in the blood predicted incident adverse events in all subjects, sex-specific analyses showed a larger effect estimate between TMAO levels and adverse AAA risk in men compared to women. While the occurrence of AAA in women is less frequent than in men, it is recognized that women with AAA expand less predictably, often faster, and often rupture at a smaller diameter than in men.^59,60 For this reason, the updated ACC/AHA guidelines for AAA surveillance differ in frequency for men and women depending on the size of the aneurysm.⁶ We note, however, that the associations between TMAO and adverse AAA events were directionally consistent in both men and women for all AAA related outcomes monitored. Although the estimates in women were less precise, likely reflecting the smaller number of events, formal tests for interaction by sex were not significant [e.g. sex interaction P = .94 for risk of AAA upon ultrasound screening at the CHS Year 5 visit (Supplementary data online, Figure S2), and sex interaction P = .5 for risk of incident AAA events (Supplementary data online, Figure S7)]. Thus, the present findings suggest no strong evidence of differential effects between men and women with respect to elevated TMAO levels and AAA or incident AAA-related adverse outcomes. As an observational study, our findings are subject to potential unmeasured confounding variables. In addition, residual confounding may persist due to measurement error in self-reported covariates, including smoking status and alcohol intake. In addition, HRs may be subject to selection effects over time due to depletion of susceptible individuals, which can complicate the causal interpretation of long-term associations.⁶¹ The consistent association between elevated TMAO levels and multiple adverse AAA phenotypes across subgroup analyses observed underscores the robustness of the present findings.

In conclusion, this study demonstrates that elevated serial levels of the gut microbe-derived metabolite TMAO were independently associated with increased risk for AAA development, as well as higher incident adverse AAA events including AAA surgical repair, AAA rupture, or AAA-related mortality. The present results support the potential utility of TMAO as a biomarker for AAA risk stratification and highlight the need for further clinical intervention studies to clarify its role as a therapeutic target in individuals at risk for AAA progression, AAA repair and AAA rupture.

Supplementary Material

Supplementary data are available at European Heart Journal online.