Trans-fatty acid blood levels of industrial but not natural origin are associated with cardiovascular risk factors in patients with HFpEF: a secondary analysis of the Aldo-DHF trial

This study is a post hoc analysis of the Aldo-DHF trial (ISRCTN 94726526). Baseline blood levels of TFA were analyzed. 18 whole blood aliquots were not available due to loss during storage/transfer or missing blood sampling at baseline of a total of 422 patients enrolled in the Aldo-DHF trial as described in our previous analysis of n-3 fatty acids in this cohort [21].

The Aldo-DHF trial is a multicenter, prospective, randomized, double-blind, placebo-controlled trial evaluating the effect of a 12-month aldosterone receptor blockade on diastolic function (E/e′) on echocardiography and maximal exercise capacity (VO_2peak) in HFpEF patients. Men and women aged 50 years or older were eligible to participate in the study if they had current HF symptoms consistent with New York Heart Association (NYHA) class II or III, left ventricular ejection fraction (LVEF) of 50% or greater, echocardiographic evidence of diastolic dysfunction (grade I) or atrial fibrillation at presentation and maximal exercise capacity (VO2peak) of 25 mL/kg/min or less [24]. 422 ambulatory patients (mean age 67 [SD, 8] years; 52% female) with evidence of diastolic dysfunction were included. Data acquisition was conducted between March 2007 and April 2012 at 10 sites in Germany and Austria [24].

The Aldo-DHF Trial complies with the Declaration of Helsinki and principles of good clinical practice. All responsible ethics committees approved the study protocol. All participants gave written informed consent.

Baseline characteristics are shown in Table 1. Tables 2 and 3 and Fig. 1 depict the associations of individual TFA [i.e., the naturally occurring C16:1n-7t (trans-palmitoleic acid) and the industrially processed TFA C18:1n9t, C18:2n6tt, C18:2n6ct and C18:2n6tc] with patient characteristics at baseline and 12mFU, respectively. Statistically significant associations with clinical relevance are reported in the text, and all associations are depicted in Tables 2 and 3.

Sensitivity analyses with medication group as additional independent variable within the linear regressions showed no significance regarding blood lipid profile, markers for (central) adiposity, aerobic capacity or liver enzymes. Therefore, group allocation (spironolactone^+/−) had no effect on the association between these variables and TFA blood levels at 12mFU. As expected regarding the pharmacological effects of spironolactone, sensitivity analyses with medication group as covariate showed significant effects of group allocation for the association between TFA and the 12mFU outcomes: systolic/diastolic blood pressure (p < 0.001 for all), heart rate (only C16:1n-7t and C18:1n9t), E/e′ and HbA1c. Furthermore, all correlations were adjusted for BMI, markers for truncal adiposity (waist circumference and waist-to-height ratio), HbA1c as well as systolic/diastolic blood pressure at baseline and 12mFU (Supplemental Tables 1 and 2). These adjustments did not alter our results at 12mFU. At baseline, after adjustment for the above variables, correlations between submaximal aerobic capacity and the two trans linoleic (C18:2n6) isomers (C18:2n6tt and C18:2n6ct); alanine transaminase and C18:1n9t (r = − 0.09, p = 0.095); and γ-glutamyltransferase and C16:1n-7t (r = − 0.1, p = 0.06) were not significant anymore. However, it is of importance that the overall pattern described, in particular the association of blood trans-fatty acid status and lipid profile, was not altered by adjusting for BMI, truncal adiposity, HbA1c and blood pressure.

We reduced dimensions by using a principal component analysis (PCA) to account for the number variables in relation to the present sample size. The created PCAs show a relatively low cumulative variance for the first two PCs. Therefore, these PCs were not used in further analyses. The first and second PCs are shown as bi-plot in Fig. 4 to visualize the similarities and differences between the used predictors.

We investigated the association of TFA blood levels with cardiometabolic phenotype, aerobic capacity and cardiac function in patients with HFpEF. Using Spearman’s correlation coefficient and multiple linear regression analyses, we broadly identified two groups of TFA that had opposing associations with risk factors in HFpEF patients as depicted in the Graphical Abstract: the naturally occurring TFA C16:1n-7t was associated with a lower risk cardiometabolic phenotype while the IP-TFA trans-fatty acid C18:1n9t and the three IP-trans-18:2 isomers C18:2n6tt, C18:2n6ct and C18:2n6tc were broadly associated with cardiometabolic risk factors/HFpEF comorbidities and lower functional capacity. Further substantiating these findings, the highest versus the lowest tertile group of the composite of the three trans linoleic (C18:2n6) isomer blood levels had higher LDL-C, higher non-HDL-C and lower submaximal aerobic capacity at baseline. These associations, in particular the association of blood trans-fatty acid status and lipid profile, were not altered by adjusting for BMI, markers for truncal adiposity (waist circumference and waist-to-height ratio), HbA1c and blood pressure.

Higher blood levels of trans oleic acid C18:1n9t and the trans linoleic acid isomer C18:2n6tc were associated with higher non-HDL-C and LDL-C at baseline, which are both established cardiovascular risk factors, the former being a better predictor for cardiovascular endpoints in individuals with MetS and T2D [33]. In addition, higher trans oleic acid C18:1n9t was associated with metabolic markers of a higher risk lipid and plaque phenotype such as triglycerides-to-HDL-C ratio [34‐36] and triglycerides [37] at baseline.

The trans linoleic acid isomers C18:2n6tt and C18:2n6ct were predominantly associated with higher HbA1c at baseline. T2D is a common comorbidity with prognostic impact in patients living with HFpEF and necessitates comprehensive management [1, 38]. This may, among other reasons, provide one biologically plausible explanation for the observed benefit of the oral antidiabetic drug SLGT2i Empagliflozin in patients with HFpEF [6]. Sensitivity analyses with medication group as covariate showed significant effects of group allocation (spironolactone^+/−) for the association between TFA and the 12mFU outcome HbA1c, possibly indicating that the study medication confounded the direct association between IP-TFA status and HbA1c at 12mFU.

IP-TFA foster a systemic milieu characterized by adipose tissue, vascular and systemic inflammation [8, 11] which is an intermediate risk factor for adverse outcomes in HFpEF [1]. Mozaffarian and colleagues [19] showed that in 86 ambulatory patients with established heart failure, RBC TFAs were strongly associated with NT-proBNP and biomarkers of systemic inflammation (i.e., IL 1β, IL-1 receptor antagonist, IL-6, IL-10, TNF, TNF receptor 1 and 2, monocyte chemoattractant protein 1). Regarding markers of inflammation, in our cohort, leukocytes were available as a broad and unspecific marker of inflammatory status. While leukocytes showed no association with IP-TFA, interestingly, higher blood levels of the naturally occurring C16:1n-7t were consistently associated with lower leukocyte count at baseline and prognostic for lower leukocyte count after 12mFU. One hypothesis that may offer some biological plausibility for this observation is that blood levels of naturally occurring trans-fatty acids are a biomarker, e.g., intake of dairy fat. Dairy and its components influences diverse pathways related to cardiometabolic health (reviewed in [31]), including oxidative and inflammatory pathways, intestinal integrity and endotoxemia related to broad inflammation [31].

As opposed to the findings of Mozaffarian et al. [19] we observed a significant positive association with neurohumoral activation for only one of the three IP-TFA trans linoleic (C18:2n6) isomers (C18:2n6tc). This association persisted after 12mFU. However, the strength of the association was very weak and such an association was not observed for all other IP-TFA. Reasons for this inconsistency may be the different methodology used (red blood cell TFA in the analysis of Mozaffarian et al. vs whole blood TFA in our analysis) and the different study populations (HF patients in the analysis of Mozaffarian et al. vs HFpEF patients in our analysis). No association between blood TFAs and echocardiographic markers of left ventricular diastolic dysfunction were observed in the Aldo-DHF cohort.

Higher blood levels of the trans linoleic acid isomers C18:2n6tt and C18:2n6ct were very consistently associated with lower submaximal aerobic capacity at baseline and at 12mFU. Functional capacity is an important prognostic marker in HF/HFpEF [3]. Therefore, its inverse association with trans linoleic acid isomers C18:2n6tt and C18:2n6ct may be a clinically relevant finding that (a) warrants future studies to substantiate these findings and (b) may justify discouraging HFpEF patients from consuming foods that contain trans fats. According to our literature search, no data are available on the association of IP-TFA and functional capacity in HFpEF patients.

Partially hydrogenated vegetable oils are attractive ingredients for the food industry due to their long shelf life, their stability during deep-frying and their semisolid texture, which is used to enhance the palatability of baked goods and sweets [8]. Due to uncontroversial evidence on their adverse health effects, in the USA, regulatory agencies pressured dietary oil producers to remove IP-TFA from their products. In the USA, this has resulted in declining IP-TFA levels in red blood cells since 1999 [39] and between 2009 and 2016 has collinearly led to the desired effect of declining rates of fatal ischemic heart disease (r = 0.9552, p < 0.0001) [40]. Similarly, in Europe, from 2008 to 2015, levels of IP-TFA and ruminant-derived C16:1n-7t, a marker for dairy and meat intake, have been decreasing with few individuals having levels of IP-TFA above a safe range, while many had low levels of C16:1n-7t [41]. In line, Kleber et al. [16] suggested that TFA levels up to 1.04% of total fatty acids of erythrocytes membranes might be regarded as safe and are not associated with adverse health outcomes. This overall suggests a general decline in TFA exposure due to policy changes [18]. However, it is worth noting that, even within the low levels of IP-TFA blood levels observed in the Aldo-DHF cohort, there was a consistent positive association with high-risk metabolic traits in HFpEF patients (whole blood TFA acids in our cohort, as compared to erythrocyte TFA acid levels in the analysis by Kleber et al. [16], were lower by a potency of 10 but similar to whole blood TFA concentrations in another German cohort with HF [29]). Therefore, despite the limitations of observational data in addressing causation questions, our findings support the notion that further efforts to removing IP-TFA from the food supply and/or stronger public health efforts to discourage consumption of food sources substantially contributing to trans fat consumption (i.e., fried foods, margarine, processed meats, bakery products and biscuits) [42] may ameliorate risk factor control in patients living with HFpEF, a condition where prognosis is largely determined by managing comorbidities [1]. Our findings reflect the differential association of individual TFA sources (industrial vs. natural ruminant) on HFpEF phenotype. In line with the literature, this may speak to encouraging increased consumption of whole foods such as full fat dairy that contain naturally occurring TFA or potentially supplementing the naturally occurring TFA C16:1n-7t, for improving cardiometabolic health also in HFpEF patients [28, 43]. Furthermore, these findings support the notion that trans-fatty acids, like saturated fatty acids [44] and omega-3 fatty acids [21], are a heterogeneous group with respect to health interactions. This overall speaks to considering individual fatty acids instead of referring to predefined groups of fatty acids based on their physicochemical properties.

A strength of this analysis is the large, well phenotyped cohort comprising 404 patients with HFpEF. Second, 53% of the patients included in Aldo-DHF were female, resulting in a representative gender distribution for the condition HFpEF, which is slightly more common in females [45]. Third, blood fatty acid levels are an objective biomarker of TFA status unbiased by recall-based dietary intake assessment methods (e.g., food frequency questionnaires) used in nutritional epidemiology which may not accurately reflect an individual’s consumption of nutrients due to limitations such as measurement error, recall bias, selective reporting and incompleteness of food composition databases [22].

Our analysis is not without limitations. First, our analysis does not offer information on the prognostic impact of blood TFA levels in patients with HFpEF. We analyzed associations of TFA blood levels and surrogate markers for HFpEF prognosis. Future research is warranted to determine the prognostic relevance of our findings. Second, this is an observational study that cannot address causation questions. Recognizing this limitation, our findings may support our hypothesis that efforts to remove IP-TFA from the food supply in Europe may be beneficial for patients with HFpEF. Third, little is known about the efflux of TFA blood levels, i.e., the relative contribution and determinants of elimination of these fatty acids on actual blood levels. Fourth, since 1 April 2021, the European Parliament has forbidden the marketing of foods that contain trans fat from industrial origin (excluding naturally occurring in fat of animal origin) of more than 2 g per 100 g of fat (2%) (https://​ec.​europa.​eu/​food/​safety/​labelling-and-nutrition/​trans-fat-food_​en). This may have led to declining levels of blood IP-TFA since data acquisition for the Aldo-DHF study (March 2007–April 2012). Current blood IP-TFA in HFpEF patients are not known. Finally, in the Aldo-DHF study, no data were obtained on dietary intake, and therefore, we cannot provide data on diet or related variables such as macronutrient distribution. In this regard, it is, however, worth noting that blood levels of fatty acids are determined not only by intake, but also by bioavailability, distribution volume, catabolism and other factors (e.g., de novo lipogenesis of saturated fatty acids from carbohydrates in the context of insulin resistance/non-alcoholic fatty liver disease). Therefore, circulating levels of fatty acids represent the status of an individual in these fatty acids—and as has been repeatedly demonstrated, the status of an individual of any given fatty acid correlates more closely with clinical events or other biological parameters than intake. Therefore, from the perspective of scientific rigor, reporting levels of fatty acids as an objective scientific metric but not dietary intake (as assessed by subjective memory-based methods) may alternatively be seen as a strength of our study.