Plasma acylcarnitines and risk of incident heart failure and atrial fibrillation: the Prevención con Dieta Mediterránea study

Abstract

Introduction and objectives:

Fatty acid metabolic dysregulation in mitochondria is a common mechanism involved in the development of heart failure (HF) and atrial fibrillation (AF). We evaluated the association between plasma levels of acylcarnitines and incidence of HF or AF, and whether the Mediterranean diet (MedDiet) may attenuate the association between acylcarnitines and HF or AF risk.

Methods:

Two case-control studies nested within the PREvención con DIeta MEDiterránea (PREDIMED) trial. High cardiovascular risk participants recruited in Spain: 326 incident HF and 509 AF cases individually matched with 1 to 3 controls. Plasma acylcarnitines were measured with high-throughput liquid chromatography–tandem mass spectrometry. Conditional logistic regression models were fitted to estimate multivariable OR and 95%CI. Additive and multiplicative interactions were assessed by intervention group, obesity (body mass index ≥ 30) and type-2 diabetes.

Results:

Elevated levels of medium-chain and long-chain acylcarnitines were associated with increased HF risk (adjusted OR_{per SD}, 1.28; 95%CI, 1.09–1.51 and adjusted OR_{per SD}, 1.21; 95%CI, 1.04–1.42, respectively). A significant association was observed for AF risk with long-chain acylcarnitines: 1.20 (1.06–1.36). Additive interaction of the association between long-chain acylcarnitines and AF by the MediDiet supplemented with extra-virgin olive oil (P for additive interaction = .036) and by obesity (P = .022) was observed in an inverse and direct manner, respectively.

Conclusions:

Among individuals at high cardiovascular risk, elevated long-chain acylcarnitines were associated with higher risk of incident HF and AF. An intervention with MedDiet+ extra-virgin olive oil may reduce AF risk associated with long-chain acylcarnitines.

INTRODUCTION

Heart failure (HF) and atrial fibrillation (AF) have emerged as important cardiovascular (CV) diseases contributing to the increasing burden of chronic diseases and health care costs.¹ Both HF and AF result from a cumulative exposure of shared CV risk factors that promote common and concurrent biological pathways such as the release of inflammation mediators which are involved in cardiac structural and electrophysiological remodeling.²

By studying HF and AF simultaneously, we can acquire a better understanding of their common underlying mechanisms as well as the key pathophysiological differences between them.³ Fatty acid metabolic dysregulation in mitochondria has been proposed as a common mechanism involved in the development of HF and AF. Moreover, the accumulation of long-chain acylcarnitines (LCAC) in plasma and tissue play an active role in inflammation and insulin resistance.⁴

In the PREvención con DIeta MEDiterránea (PREDIMED) study, the Mediterranean diet (MedDiet) intervention mitigated the association of high AC concentrations with the risk of major CV events (ie, myocardial infarction, stroke, or CV death) as compared to the control group.⁵ However, similar evidence is not currently available for AF and HF. We designed 2 case-control studies nested within the PREDIMED trial to evaluate the associations between ACs and incidence of HF or AF, and whether the MedDiet could mitigate the harmful effects associated with increased baseline levels of ACs. As a secondary aim, since both cardiovascular diseases share common risk factors⁶, we assessed the association between ACs as a composite outcome including participants with either HF or AF.

METHODS

Population and design

We designed 2 case-control studies nested within the PREDIMED trial.⁷ We selected 326 incident cases of HF and 509 AF cases after excluding prevalent cases, and incident cases without available samples. Incidence density sampling with replacement was used as the control sampling method.⁸ Thus, controls were randomly selected from all participants at risk at the time of the incidence case occurrence, and selected controls could be selected again as a control for another index case, and they could become later a case. Results from a simulation study considered incidence density sampling with replacement as the least biased efficient method for control sampling in nested case-control studies.⁹ Controls were matched by recruitment center, year of birth (± 5 years), and sex. We selected 1 to 3 matched controls per case. Figure 1 of the supplementary data shows the flow-chart of the participant selection process.

The protocol was approved by the research ethics committees at all study locations, and all participants provided written informed consent.

Outcomes

HF and AF were a priori defined as secondary endpoints in the PREDIMED trial protocol. In this analysis, all HF and AF incident cases diagnosed from 2003 until December 2017 were included. One center stopped follow-up in December 2014 and all participants from this center were censored at this date to be selected as controls.

Information on these outcomes was collected from continuous contact with participants and primary health care physicians, annual follow-up visits, yearly ad-hoc reviews of medical charts and annual consultation of the National Death Index. This information was collected by physicians who were blinded to the intervention groups and metabolomics measurements. If a clinical diagnosis of HF or AF was found, all relevant documentation, including clinical records of hospital discharge, outpatient clinics and family physicians’ records were obtained. Medical charts were sent anonymously to the Clinical End-Point Committee. This documentation was evaluated by two cardiologists separately and if they did not agree on the classification of the event, a third cardiologist (the committee’s chair) intervened. In some cases, more information was requested to complete the adjudication. The End-Point Committee adjudicated the events according to pre-specified criteria in a “Manual of operations” (supplementary methods of the supplementary data).

Sample collection and metabolomic analysis

During the baseline visit, ie, years before the development of AF or HF, participants provided blood samples after at least an 8-hour fast. All samples were processed at each recruiting center no later than 2 hours after collection and stored at −80°C until their analysis. Samples from matched case-control pairs were shipped and assayed in the same analytical run, varied at random to reduce bias and inter-assay variability. The metabolomics analysis is described at the Supplementary Methods. Information about the mass to charge ratio and retention time is shown in table 1 of the supplementary data.

Covariates

Baseline questionnaires were used to collect sociodemographic, lifestyle variables, prevalent and family history of diseases, and medication use. Leisure-time physical activity was measured with the validated version of the Minnesota Leisure-Time Physical Activity questionnaire.¹⁰ Incident coronary heart disease (any diagnosis of angina, myocardial infarction or coronary revascularization procedures) and stroke (any diagnosis of either ischemic or hemorrhagic stroke, and also transient ischemic attacks) were collected, blindly to the metabolomics information, according to the diagnosis criteria applied by the Clinical End-Point Committee.

Statistical analysis

Individual metabolite values were normalized and scaled to multiples of 1 SD using the rank-based inverse normal transformation. Mean ± standard deviation (SD) was used to describe quantitative traits and absolute number and percentage to describe categorical variables.

We calculated the Pearson’s correlation coefficient for free carnitine, ACs and branched-chain amino acids. We also ran linear regression models using AC scores as dependent variables and CV risk factors, as independent variables and adjusting for age and sex.

We fitted crude and multivariable conditional logistic regression models to account for the matching between cases and controls. We calculated matched odds ratio (OR) and their 95% confidence interval (95%CI) for HF or AF in the comparisons of upper quartiles of the ACs versus the lowest quartile, and for each SD of ACs using them as continuous variables. Quartiles cut-off points were calculated according to the distribution of ACs among controls (ie, participants without HF or AF). Multivariable model was adjusted for intervention group (MedDiet+extra-virgin olive oil [EVOO], MedDiet+nuts or control), smoking status (never/current/former), body mass index (BMI) (kg/m²), leisure-time physical activity (metabolic equivalent task [MET]-min/day), prevalent chronic diseases (hypertension, type 2 diabetes [T2D] and dyslipidaemia), and medication use, including ACE inhibitors, diuretics, other antihypertensive treatments, statins and other lipid-lowering agents, insulin, oral hypoglycemic agents and antiplatelet therapy. As an ancillary analysis, we additionally adjusted for incident coronary heart disease (angina, myocardial infarction, coronary revascularization procedures) and stroke (either ischemic or hemorrhagic, including transient ischemic attacks) diagnosed during the follow-up but before the diagnosis of HF or AF. Adjustment for multiple comparisons was based on the false discovery rate (FDR) procedure as proposed by Simes.¹¹ Additional statistical analyses are described in the supplementary methods of the supplementary data.

Statistical analyses were performed using Stata/SE version 15.1 (Stata Corp.).

RESULTS

We analyzed data from 2 case-control studies nested within the PREDIMED trial. The number of HF incident cases was 326 and the number of AF incident cases was 509. In total, there were 727 cases since 108 participants developed both AF and HF during the follow-up (figure 1 of the supplementary data). Participants’ characteristics in both case-control studies nested within the PREDIMED trial are shown in table 1.

Table 1.

Baseline participant characteristics of HF and AF cases and controls

	Case-control sets for HF		Case-control sets for AF
	Controls*	HF cases	Controls*	AF cases
n	426	326	617	509
Age, years	70.4 (5.9)	70.3 (5.8)	68.5 (6.1)	68.3 (6.1)
Women sex, %	54.2	58.3	49.3	49.7
Intervention group, %
MedDiet+EVOO	37.6	30.1	36.5	31.4
MedDiet+nuts	26.5	32.5	28.5	31.4
Control	35.9	36.4	35.0	37.1
Smoking, %
Never	61.3	59.8	58.0	58.7
Former	27.5	25.8	28.7	726.9
Current	11.2	14.4	13.3	14.4
Physical activity, METs-min/d	217 (220)	216 (204)	229 (220)	228 (216)
Education, %
Elementary or lower	81.7	85.0	78.8	76.0
Secondary or higher	18.3	15.0	20.2	24.0
Total energy intake, kcal/d	2279 (637)	2217 (632)	2342 (603)	2285 (600)
Score for adherence to Mediterranean diet**	8.6 (2.0)	8.5 (2.0)	8.8 (1.9)	8.7 (2.0)
Alcohol consumption, g/d	8.4 (13)	8.1 (15)	9.9 (15)	8.9 (13)
Waist circumference, cm
Women	98.1 (10.2)	101.89 (10.2)	98.0 (10.3)	100.7 (10.2)
Men	101.2 (8.8)	106.4 (9.4)	102.9 (8.9)	106.0 (8.8)
Waist circumference > 88 cm (women) or > 102 cm (men), %	64.1	80.7	68.6	76.8
Body mass index, kg/m2	29.4 (3.6)	31.1 (3.8)	29.8 (3.8)	30.7 (3.8)
Family history of premature CHD, %	19.3	19.3	20.1	19.1
Hypertension, %	82.2	87.4	82.8	88.4
Dyslipidemia, %	69.0	64.1	68.4	65.2
Type 2 diabetes, %	52.1	59.5	49.9	47.9
AF, %	0	4.3	-	-
HF, %	-	-	0.2	0.4

AF, atrial fibrillation; CHD, coronary heart disease; EVOO, extra-virgin olive oil; HF, heart failure; MET, metabolic equivalent.

^*Includes HF or AF cases selected as controls

^**Ranged from 0 to 14 points

Values are mean (SD) or percentage.

Short, medium and long-chain ACs were highly correlated (figure 2 of the supplementary data). Table 2 of the supplementary data displays the β coefficients of AC scores according to several cardiovascular risk factors. Glucose was associated with short-chain ACs, and TG with medium-chain ACs.

Acylcarnitine and heart failure risk

In multivariable model, C14, C16, C18, and C18:2 were associated with a higher risk of HF after correcting for multiple comparisons (figure 1A and table 3 of the supplementary data). Significant results remained the same for C16, C18 and C18:2 when we additionally adjusted for branched-chain amino acids (figure 3 of the supplementary data).

Figure 1.

Odds ratios (95% confidence intervals) between baseline levels of ACs and incident HF or AF in nested case-control studies (cases and controls matched by sex, age, and recruitment center). AC, acylcarnitine; AF, atrial fibrillation; HF, heart failure; MV, multivariable model adjusted for intervention group, body mass index, smoking, leisure-time physical activity, prevalent chronic diseases, and medication use; SD, standard deviation

During the follow-up, 50 HF cases developed coronary heart disease or cerebrovascular disease before the diagnosis of HF and 54 controls also developed at least one of these major CV disease events. No changes were observed in the association between individual acylarnitines and HF when we additionally adjusted for previous incident coronary heart disease or cerebrovascular disease (data not shown).

The total score with all ACs was associated with a higher risk of HF (table 4 of the supplementary data). Higher HF risk was observed for medium and long-chain AC scores in the comparison between extreme quartiles (figure 2A). Both medium and long-chain AC scores were significantly associated with HF as continuous variables (figure 2A and table 4 of the supplementary data). No significant differences for interaction tests were observed when we analyzed the potential effect modification by sex or age group (≤70 vs > 70) in the association between AC scores and HF risk (table 5 of the supplementary data).

Figure 2.

Association between baseline combined scores (weighted sum of normalized values for each metabolite [using the leave one method to avoid overfitting]) of plasma acylcarnitines and incident HF or AF in nested case-control studies (cases and controls matched by sex, age, and recruitment center) of the PREDIMED trial. AF, atrial fibrillation; HF, heart failure; MV, Multivariable model adjusted for intervention group, body mass index, smoking, leisure-time physical activity, prevalent chronic diseases, and medication use; Ref, reference; SD, standard deviation

Acylcarnitine and atrial fibrillation risk

Regarding AF risk, C16, C18, and C18:2 ACs were associated with a higher AF risk after correction for multiple comparisons (figure 1B and table 6 of the supplementary data). These associations did not change when we additionally adjusted for branched-chain amino acids (figure 3 of the supplementary data). A total of 62 AF cases and 73 controls developed coronary heart disease or cerebrovascular disease before the diagnosis of AF. No changes were observed when we additionally adjusted for these previous incident cardiovascular diseases (data not shown).

We observed that only LCACs were significantly associated with AF risk, with an OR (95%CI, p for trend) of 1.20 for each SD (1.06–1.36, P for trend = .005) (table 4 of the supplementary data). No significant effect modification by sex or age group (≤ 70 vs > 70) was found (table 5 of the supplementary data).

Acylcarnitine and a composite outcome heart failure/atrial fibrillation

In the composite outcome of HF or AF cases, a significant association was observed for both medium and LCAC scores (table 4 of the supplementary data). When we mutually adjusted for all AC scores, incident HF was only significantly associated with medium-chain ACs and incident AF only with LCACs (figure 4 of the supplementary data).

In the multinomial logistic regression models (table 2), the relative odds of developing only HF in comparison with participants without incident HF and AF, was 27% (95%CI, 8%–49%) higher per SD of the medium-chain AC score, not significant for those participants who developed only AF and 40% (95%CI, 9%–78%) higher for those who developed HF and AF per SD of the medium-chain AC score. A similar higher risk was observed for only HF or AF for each SD increase of LCAC score and a significant association was found between the short-chain AC score and only HF risk in comparison with participants without HF and AF.

Table 2.

Odds ratios and 95% confidence intervals from multinomial logistic regression models showing the association between baseline scores of plasma acylcarnitines levels and a composite outcome

	Multivariable1 adjusted OR (95%CI)
	per 1 SD increment2	P
No HF or AF (N = 726)
Short-chain-AC	1.00 (ref)	-
Medium-chain-AC	1.00 (ref)	-
Long-chain-AC	1.00 (ref)	-
Total AC	1.00 (ref)	-
Only HF (N = 218)
Short-chain-AC	1.23 (1.02–1.48)	.035
Medium-chain-AC	1.27 (1.08–1.49)	.004
Long-chain-AC	1.22 (1.09–1.38)	.001
Total AC	1.38 (1.14–1.66)	.001
Only AF (N = 401)
Short-chain-AC	1.08 (0.97–1.19)	.158
Medium-chain-AC	1.07 (0.97–1.18)	.154
Long-chain-AC	1.25 (1.09–1.43)	.001
Total AC	1.17 (1.05–1.30)	.004
AF and HF (N = 108)
Short-chain-AC	0.99 (0.81–1.21)	.909
Medium-chain-AC	1.40 (1.09–1.78)	.007
Long-chain-AC	1.29 (0.98–1.69)	.072
Total AC	1.31 (0.94–1.84)	.111

95%CI, 95% confidence interval; AC, acylcarnitine; AF, atrial fibrillation; EVOO, extra-virgin olive oil; HF, heart failure; MedDiet, Mediterranean diet; OR, odds ratio; ref., reference; SD, standard deviation

¹Adjusted for age, sex, recruitment center, intervention group (MedDiet+EVOO. MedDiet+nuts), body mass index (kg/m²), smoking (never current former), leisure-time physical activity (metabolic equivalent tasks in minutes/day), prevalent chronic diseases (dyslipidemia, hypertension, and diabetes), and medication use (angiotensin-converting enzyme inhibitors, diuretics, other antihypertensive treatments, statins and other lipid-lowering agents, insulin, oral hypoglycemic agents and antiplatelet therapy).

²Values were normalized and scaled to multiples of 1 SD using the rank-based inverse normal transformation and a weighted sum of ACs was then calculated.

Risk modification by diet and obesity

Figure 3 shows the joint analysis for quartiles of AC scores (Q1–Q2 vs Q3–Q4) and the dietary intervention (MedDiet+EVOO vs control group) and the association with HF or AF risk. The MedDiet+EVOO group modified the effect of high levels of LCACs (Q3–Q4) on AF risk compared to the control group (P for additive interaction = .036). No other additive or multiplicative interactions were found according to the intervention groups.

Figure 3.

Odds ratio¹ (95%CI) of the joint analysis of AC scores²/MedDiet+EVOO and HF or AF risk in nested case-control studies of the PREDIMED trial. A: RERI (95%CI; P value): Short-AC, −0.05 (−1.15 to 1.04; P = .924), Medium-AC, 0.53 (−0.29 to 1.35, P = .209), Long-AC, 0.23 (−0.41 to 0.87; P = .478). B: RERI (95%CI; P value): Short-AC, 0.17 (−0.37 to 0.71; P = .534), Medium-AC, −0.21 (−0.94 to 0.51, P = .562), Long-AC, 0.77 (0.05 to 1.48; P = .036). The MedDiet+nuts group was not included in this analysis. ¹Odds Ratio (95% confidence interval) adjusted for intervention group, body mass index, smoking, leisure-time physical activity, prevalent chronic diseases, and medication use. ²Weighted sum of normalized values for each metabolite. 95%CI, 95% confidence interval; AC, acylcarnitine; AF, atrial fibrillation; EVOO, extra-virgin olive oil; HF, heart failure; MedDiet, Mediterranean diet; RERI, relative excess of risk due to interaction.

We observed greater AF risk associated with high levels of LCACs (Q3–Q4) among obese participants compared with high levels among non-obese participants (P for additive interaction = .022) (figure 4). No significant interactions were observed except for the association between short-chain ACs and HF in the stratified analysis by T2D.

Figure 4.

Odds ratio¹ (95%CI) of the joint analysis of AC scores²/obesity and HF or AF risk in nested case-control studies of the PREDIMED trial. A: RERI (95%CI; P value): Short-AC, −0.65 (−2.15 to 0.84; P = .392), Medium-AC, −0.65 (−2.08 to 0.79, P = .377), Long-AC, 0.34 (−0.83 to 1.51; P = .568). B: RERI (95%CI; P value): Short-AC, 0.18 (−0.55 to 0.91; P = .634), Medium-AC, 0.56 (−0.13 to 1.25, P = .112), Long-AC, 0.92 (0.14 to 1.70; P = .022).¹ Odds Ratio (95% confidence interval) adjusted for intervention group, body mass index, smoking, leisure-time physical activity, prevalent chronic diseases, and medication use.² Weighted sum of normalized values for each metabolite. 95%CI, 95% confidence interval; AC, acylcarnitine; AF, atrial fibrillation; EVOO, extra-virgin olive oil; HF, heart failure; MedDiet, Mediterranean diet; RERI, relative excess of risk due to interaction.

DISCUSSION

In 2 matched case-control studies nested within the PREDIMED trial, the risk of incident HF and AF was higher among individuals with elevated levels of LCACs at baseline whereas some short or medium-chain ACs were more clearly positively associated with HF. Our results also suggest a potential modification of the effect of high levels of LCACs on AF risk by obesity defined as BMI≥ 30 (increasing the risk), and by a MedDiet intervention supplemented with EVOO (reducing the risk).

Our findings were consistent with some results from previous HF case-control studies. One study identified higher levels of linoleylcarnitine and hydroxybutyrylcarnitine in stage C HF patients compared with controls.¹² In the CATHGENE study, higher concentrations of 6 LCACs were found in patients with HR with reduced ejection fraction (HFrEF) than HR with reduced preserved fraction (HFpEF); and in both types of HF higher concentrations than in the controls.¹³ In a smaller case-control study, LCACs concentrations were higher in HF patients compared with controls.¹⁴ Finally, results from the Mälmo study have shown an association between medium and long-chain ACs.¹⁵ Our study found a stronger association between medium-chain ACs, instead of LCACs, and HF. One reason to explain this discrepancy is that we analyzed the association between baseline ACs and incident HF instead of measuring ACs levels in patients already diagnosed with HF.

In the MURDOCK cohort study, medium-chain ACs, short-chain dycarboxyl ACs and LCACs were associated with incident AF in patients referred for coronary angiography.¹⁶ We observed the strongest association between LCACs and AF, whereas the association with short or medium-chain ACs was no longer significant after adjusting for different confounders. No detailed information was provided in the MURDOCK study about the specific ACs associated with AF but they classified palmitoyl-L-carnitine as a medium-chain AC instead of as LCAC, as we did in our study. Another difference is that the MURDOCK study included patients who developed AF after coronary artery bypass grafting surgery¹⁶ whereas in the PREDIMED postoperative incident AF cases were excluded.¹⁷ The role of LCACs is supported by its known effect increasing intracellular calcium and inducing electrophysiological alterations in experimental systems where calcium efflux was increased, in a concentration-dependent manner, by palmitoyl-L-carnitine and stearoyl-L-carnitine but not by short-chain esters.¹⁸

The analysis and interpretation of plasma ACs levels is complex because increased concentrations have been associated with several chronic diseases. Our participants were at high CV risk, with an average age of 70 years, mean BMI was 30 kg/m² and > 50% had T2D. All these factors are related to increased plasma levels of ACs. No significant differences were observed in the association between ACs and HF or AF risk when we stratified by sex or age group (≤ 70 vs > 70). These results are in line with a study which found similar total average levels of ACs among men and women.¹⁹ However, sex-specific effects of smoking and BMI on AC metabolism was observed in another study²⁰ and further research is warranted to explore the potential effects of (or effect modification by) BMI or smoking. We observed associations between the levels of ACs and some cardiovascular risk factors. In a previous study within the PREDIMED trial we observed that short-chain and LCAC increased the risk of T2D,²¹ whereas short and medium-chain plasma ACs were associated with CV disease⁵ (myocardial infarction, stroke, and CV death). Other studies have shown that increased levels of hydroxybutyrylcarnitine, medium-chain ACs, and oleoylcarnitine were associated with prediabetic state, T2D and cardiometabolic risk factors.²² A small study found significant positive correlations between ACs and overweight/obesity measures (BMI and waist circumference), and diastolic blood pressure and negative correlation with HDL-cholesterol²³. Further studies with larger sample sizes are needed to explore these associations and to identify potential biomarkers associated with HF or AF risk and the potential modulatory effect of cardiovascular risk factors such as high glycaemia or obesity^24,25.

Propionylcarnitine and isovalerylcarnitine are byproducts of BCAA metabolism, and higher levels of plasma BCAAs and C3/C5 ACs are found in people with overweight and metabolic syndrome²⁶. We observed an association between fasting glycaemia and short-chain ACs and a stronger association between short and medium-chain ACs and HF or AF in obese participants, although the interaction was not significant. Moreover, we have previously reported the association between BCAAs with CV disease²⁷ and T2D²⁸.

Medium- and LCACs are byproducts of mitochondrial energy production through the oxidative catabolism of fatty acids. LCACs are mainly involved in muscle and cardiac metabolism and it has been suggested that heart can be the major contributor for LCACs in plasma.²⁹ The accumulation of LCACs in the diabetic myocardium reflects alterations in the fatty acids as a substrate for energy.³⁰ During the development of HF, the capacity of the heart to use fatty acids as the main source of energy is impaired and ketone bodies are used as alternative fuel sources.³¹ This mitochondrial dysfunction has been proposed as a common molecular mechanism between both HF and AF based on animal models.³²

The associations between LCAC and HF or AF were stronger among obese participants and among those assigned to the control group. The systemic chronic inflammation related to obesity and the anti-inflammatory effect of the MedDiet and EVOO may explain these associations. The accumulation of LCACs is involved in the activation of proinflammatory signaling pathways and systemic inflammation is a common mechanistic process for HF and AF.³³ This inflammatory process is linked to the expansion of the epicardial adipose tissue which is involved in the pathogenesis of AF³⁴ and it is prominent in patients with HFpEF and AF.³ However, more research is needed to better understand if the potential protective effect of the MedDiet+EVOO is through a modulation of the effect of fatty acid oxidation on AF or through a mitigation of the inflammatory effect of LCACs. Moreover, a trial conducted in Spain found an association between a MedDiet intervention and changes in circulating metabolites, including acylcarnitines, and these changes were associated with decreases in glucose, insulin and HOMA-IR³⁵. In addition, this study suggested an interplay between diet, circulating metabolites and gut microbiota. A major strength in our study was the prospective design of two matched case-control studies nested within the PREDIMED trial which allowed the preservation of the adequate temporal sequence because we measured plasma ACs years earlier than the development of incident HF and AF. Another strength was the ability to control for a wide number of potential confounders. In addition, because of the randomized design, we were able to explore whether the observed associations were modified by the Mediterranean diet interventions.

Limitations

Our study has also several limitations. First, most participants who developed HF could not be defined as HFrEF or HFpEF patients. Other studies have observed differences in the association between ACs levels and HFrEF or HFpEF.¹³ Thus, future research is needed to explore if the overlapping between HF and AF depends on the type of HF as well as the association with the MedDiet enriched with EVOO. Second, the number of cases was relatively low, especially for the number of HF cases and the overlapping between cases between HF and AF. Similarly, the study of potential interactions mediated by obesity or diet would need studies with larger sample size. However, our study has a unique design to explore the effect modification of the MedDiet and reporting both additive and multiplicative estimates were encouraged. Third, we identified a limited number of ACs and further studies are needed to examine if other ACs are associated with HF and AF. However, our results have confirmed previous associations between some individual ACs that we measured and the risk of AF and HF. Fourth, we did not quantify absolute concentrations of ACs and the practical clinical implications will probably need further assessments in the clinical setting. However, our results provide new insights into the common and differential associations between ACs and HF or AF and the potential mechanisms associated with the MedDiet. Finally, we cannot extrapolate our results to other populations with different age or prevalence of cardiovascular risk factors.

CONCLUSIONS

Among a population of individuals at high risk of cardiovascular disease, baseline LCACs were significantly associated with an increased risk of incident HF and AF, and some short or medium-chain ACs exhibited stronger positive associations with HF. The highest risk in association with increased ACs was observed for developing both AF an HF. An intervention with MedDiet+EVOO might attenuate the detrimental effect of LCACs on AF risk.

WHAT IS KNOWN ABOUT THE TOPIC?

Fatty acid metabolic dysregulation in mitochondria is a common mechanism involved in the development of heart failure (HF) and atrial fibrillation (AF), and elevated acylcarnitines are associated with higher risk of both diseases

WHAT DOES THIS STUDY ADD?

Elevated plasma long-chain ACs (LCACs) in high cardiovascular risk patients is associated with higher risk of incident HF and AF. An additive interaction of the association between LCACs and AF risk by the MedDiet supplemented with extra-virgin olive oil and by obesity was observed in an inverse and direct manner, respectively.

Abbreviations

AF

atrial fibrillation

CV

cardiovascular

HF

heart failure

MedDiet

Mediterranean diet

T2D

type 2 diabetes

Abreviaturas

CV

cardiovascular

DietMed

dieta mediterránea

DT2

diabetes tipo 2

FA

fibrilación auricular

IC

insuficiencia cardiaca