Serum Metabolite Profiles in Adults With Abdominal Obesity in Response to Consuming 1 Avocado Daily for 6 Months: An Exploratory Analysis

Abstract

Background

Avocado consumption has been associated with improvements in diet quality and cardiometabolic risk factors, but effects on serum metabolite profiles remain underexplored.

Methods

Secondary analysis was conducted using untargeted metabolomics to assess fasting serum metabolite profiles at baseline (preintervention) and 6 months (postintervention) in a subset of participants with abdominal obesity from the HAT (Habitual Diet and Avocado Trial), who were randomized to the avocado group (n = 120; 70% women; 49 ± 13 years of age; body mass index 33.1 ± 5.7 kg/m²). Pre–post intervention changes in metabolites, cardiometabolic risk factors (blood pressure, lipid profile, glucose, insulin, high‐sensitivity C‐reactive protein), visceral adipose tissue volume, and hepatic fat fraction were evaluated using Wilcoxon tests. Multilevel partial‐least squares discriminant analysis, accounting for within‐subject correlation was used to examine metabolite changes associated with avocado intake, and multiple regression assessed metabolite‐cardiometabolic risk factor associations, adjusting for covariates (age, sex, body mass index, smoking status, physical activity, energy intake) and multiple testing (false discovery rate <0.1).

Results

Plasma low‐density lipoprotein cholesterol (−6%), systolic (−3%), and diastolic (−2%) blood pressure decreased, whereas visceral adipose tissue volume increased (3%) postintervention (all P < 0.05). We identified 30 primary (sugar acids/alcohols, amino/carboxylic/hydroxy acids, indoles, xenobiotics) and 45 lipid‐related metabolites (fatty acids, cholesteryl esters, glycerolipids, glycerophospholipids, sphingolipids) as key drivers of separation between pre–post intervention time points (variable importance in projection >1). Significant but weak to modest associations (multiple‐R = 0.21–0.52) were observed between 96 predominantly lipid‐related metabolites and visceral adipose tissue volume, plasma triglycerides, and total cholesterol concentrations.

Conclusions

Avocado intake was associated with subtle shifts in serum metabolites related to lipid, carbohydrate, and amino acid metabolism, with weak effects on visceral adipose tissue volume, plasma triglycerides, and total cholesterol concentrations.

Registration

URL: https://www.clinicaltrials.gov; Unique identifier: NCT03528031.

Nonstandard Abbreviations and Acronyms

CMRF

cardiometabolic risk factor

HAT

Habitual Diet and Avocado Trial

MUFA

monounsaturated fatty acids

VAT

visceral adiposity tissue volume

Clinical Perspective.

What Is New?

• In adults with abdominal obesity, daily avocado consumption for 6 months led to subtle but significant shifts in 75 serum metabolites involved in lipid, amino acid, carbohydrate, and xenobiotic pathways.

• These metabolite changes were weakly associated with changes in visceral adiposity volume and circulating lipids (total cholesterol and triglycerides), likely reflecting both direct avocado effects and its integration into overall dietary patterns.

What Are the Clinical Implications?

• Incorporating an avocado daily into habitual diets modestly supports cardiometabolic health through diverse metabolic effects, but larger mechanistic‐focused studies are needed to confirm findings.

Current dietary guidance emphasizes a nutrient‐dense dietary pattern rich in fruits, vegetables, whole grains, legumes, nuts and seeds, lean proteins, and low‐fat dairy, while low in saturated fat, added sugars, and sodium, to promote overall health and lower cardiovascular disease risk. ¹ Diet quality, which assesses adherence to dietary recommendations using such indices as the Healthy Eating Index, ² is an important predictor of cardiometabolic health outcomes. ³ , ⁴ , ⁵ A majority of the observational studies and some dietary intervention trials have reported that higher fruit and vegetable intake is associated with higher diet quality, improvements in cardiometabolic risk factors (CMRFs), as well as a lower risk of cardiovascular disease, certain cancers, and all‐cause mortality. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹²

Avocados are a nutrient‐dense fruit, rich in monounsaturated fatty acids (MUFAs), fiber, essential nutrients, and bioactive compounds. ¹³ , ¹⁴ Analysis of National Health and Nutrition Examination Survey data concluded that regular avocado consumers, compared with low avocado consumers, had better diet quality, which was associated with lower body mass index and waist circumference. ¹⁵ Systematic reviews and meta‐analyses have documented that avocado consumption is associated with a lower prevalence of metabolic syndrome and cardiovascular disease risk. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² In the HAT (Habitual Diet and Avocado Trial), a large multicenter study involving free‐living participants with abdominal obesity, ²³ we reported that incorporating 1 avocado daily into habitual diets significantly improved diet quality, as measured by the Healthy Eating Index–2015. ²⁴ Specifically, supplementing diets with 1 avocado per day resulted in a higher intake of unsaturated fats, fruits, and vegetables, and lower intake of saturated fats and empty calories, without a significant change in body weight. ²⁴ However, these improvements in diet quality did not predict changes in CMRFs. In the HAT cohort, at the end of the 6‐month intervention period, there were no significant changes in visceral adipose tissue volume (VAT), the primary study outcome, or secondary outcomes (blood pressure, glucose, insulin, and high‐sensitivity C‐reactive protein concentrations), with the exception of modest reductions in total and low‐density lipoprotein cholesterol (LDL‐C) concentrations. ²³

Metabolomics, the comprehensive analysis of small molecules such as amino acids, lipids, and organic acids, offers a powerful tool to identify specific metabolites that reflect responses to individual foods/overall dietary patterns, ²⁵ as well as how diet quality influences diverse metabolic pathways. ²⁶ , ²⁷ , ²⁸ , ²⁹ Additionally, this methodology can detect subtle changes that may not be reflected in conventional CMRF measures, thereby offering insights into intermediate metabolic steps that connect dietary intake to chronic disease progression. Little data are available on the serum metabolite profiles associated with longer‐term avocado consumption. To address this gap, we conducted an ancillary study using an untargeted metabolomics approach to measure fasting serum metabolite profiles in a subset of HAT participants from 1 study site who were randomized to the avocado group. The aims of these secondary analyses were to (1) identify pre–post intervention metabolite changes following avocado intake and (2) determine their association with CMRFs and visceral adiposity measures. Based on the diet‐quality data, we hypothesized that regular avocado intake for 6 months would result in distinct shifts in metabolite profiles that would be associated with a favorable CMRF profile.

METHODS

Data Sharing Plan

Data described in the article, code book, and analytic code will be made available upon request pending approval by the HAT steering committee.

Study Participants, Design, and Diet Intervention

HAT (NCT03528031) was a multisite, randomized, controlled, parallel‐arm study designed to assess the effects of incorporating 1 avocado per day into habitual diets, compared with minimal avocado intake (<2 avocados per month), for 6 months on measures of VAT and CMRFs. ²³ The study was conducted between June 2018 and October 2020 across 4 clinical sites in the United States: Loma Linda University (LLU), Loma Linda, California; University of California at Los Angeles, Los Angeles, California; Tufts University (Jean Mayer United States Department of Agriculture Human Nutrition Research Center on Aging), Boston, Massachusetts; and Pennsylvania State University, University Park and Hershey, Pennsylvania.

Detailed information about the HAT study design, randomization, and primary and secondary end points have been published. ²³ , ³⁰ , ³¹ Briefly, the eligibility criteria required participants be ≥25 years of age with abdominal obesity, defined as a waist circumference ≥89 cm for women and ≥102 cm for men, and typically consuming no more than 2 avocados per month. Using permuted block randomization with varying block sizes of 4 and 8, participants were randomly assigned to either the avocado‐supplemented diet group or the habitual diet group. The avocado‐supplemented diet group received 1 Hass avocado per day, whereas the habitual diet group maintained their usual diet with minimal avocado intake. No additional nutritional or dietary counseling was provided, and participants were encouraged to maintain their usual dietary and physical activity patterns. All participants provided written informed consent before randomization. A central institutional review board at the coordinating center (Wake Forest University) provided study oversight.

The present ancillary study was designed as an exploratory within‐group analysis and included 120 HAT participants from the LLU site assigned to the avocado‐supplemented diet group, with both an archived baseline (preintervention) and end of study (6 months postintervention) serum sample for metabolomic analysis (Figure S1). Our decision to use participants from a single site, rather than selecting 30 participants from each of the 4 study sites, was based on the probability that statistical power would be maximized by minimizing variability in recruitment, participant characteristics, and data collection, resulting in higher internal validity and data consistency.

Metabolomic Profiling

Untargeted metabolomic profiling of paired serum samples (pre–post intervention) was conducted by the West Coast Metabolomics Center at University of California, Davis, California. The analysis used both gas chromatography/time‐of‐flight mass spectrometry and ultra‐high‐pressure liquid chromatography/quadrupole time‐of‐flight tandem mass spectrometry. Detailed methodologies for analyzing primary metabolites, complex lipids, and biogenic amines have been previously described. ³² , ³³ For primary metabolites, data processing was performed using ChromaTOF software (LECO, St. Joseph, MI) in conjunction with BinBase databases for compound identification. Lipidomic data were analyzed using Agilent Technologies software, including Mass Hunter Qualitative Analysis, Mass Profiler Professional, and Mass Hunter Quantitative Analysis (Santa Clara, CA), to align peaks and detect metabolites, which were subsequently identified using LipidBlast tandem mass spectrometry spectral libraries. ³⁴ Biogenic amines were processed using mzMine 2.0 software for peak identification and annotated using NIST14, Metlin, and MassBank online libraries. Quantitative results were generated using Mass Hunter Quantitative Analysis (Santa Clara, CA). ³⁵

Dietary Assessment

Dietary intake was assessed using four 24‐hour dietary recalls conducted via phone: 1 at baseline (before randomization) and 3 during the intervention period. The in‐study recalls at 2, 4, and 6 months were intended to capture intake on 2 weekdays and 1 weekend day, to obtain a representative estimate of usual dietary intake during the intervention period, as previously described. ²³ Diet assessments were administered by Tufts and LLU centers, with data analyzed using the Nutrition Data System for Research (software versions 2017 and 2018). For this study, preintervention energy and nutrient intake were derived from the baseline recall, whereas postintervention values were calculated as the average of the 2‐, 4‐, and 6‐month recalls. Compliance was based on avocado intake derived from the 24‐hour recalls conducted during the study. As expected, due to recruitment criteria, no participant reported avocado intake at baseline (preintervention). During the intervention, averaged across the 3 recalls, 96.6% reported consuming at least 1 whole avocado daily, and 3.4% consumed some avocado each day.

Visceral Adiposity Measures and CMRFs

The present analyses used the baseline (preintervention) and 6‐month (postintervention) demographic, anthropometric, magnetic resonance imaging, and biochemical data for the 120 participants included in this study. Age, sex, race, ethnicity, education, physical activity, and smoking status (nicotine exposure) were collected using health and demographic questionnaires, as previously described. ³⁰ Briefly, anthropometric measurements included height, body weight, waist circumference, and body mass index, calculated as body weight divided by height squared. Systolic blood pressure and diastolic blood pressure were measured using automated devices with participants resting in the seated position. VAT and hepatic fat content were assessed by magnetic resonance imaging. Fasting blood samples were analyzed by a central laboratory (Tufts University) for high‐sensitivity C‐reactive protein, glucose and insulin concentrations, and a lipid profile that included total cholesterol (TC), high‐density lipoprotein cholesterol, and triglycerides. LDL‐C was calculated using the Friedewald equation, ³⁶ or measured directly when triglyceride levels exceeded 300 mg/dL, following standardized protocols, as previously reported. ²³

Statistical Analysis

Descriptive analyses of the baseline characteristics of the study participants included in this ancillary study (N = 120) are summarized using mean±SD for continuous variables and percentage with number for categorical variables. The pre‐ and postintervention dietary composition data are presented as mean±SD, and the CMRFs (blood pressure [systolic blood pressure], lipid profile [TC, LDL‐C, high‐density lipoprotein cholesterol, triglycerides], glucose, insulin, high‐sensitivity C‐reactive protein) and adiposity measures (VAT, hepatic fat fraction) are presented as medians with the minimum and maximum values. Pre‐ and postintervention differences in diet composition were assessed using paired t tests, whereas changes in CMRFs and visceral adiposity measures were evaluated using the Wilcoxon test, with P values adjusted for multiple testing using the Benjamini and Hochberg procedure to control the false discovery rate at <0.1.

Metabolite data, limited to annotated (identified) metabolites, were first log‐transformed and then autoscaled. Peak values for metabolites with 0 or missing peaks were imputed at half of the minimum value of that particular metabolite. The absolute metabolite change (∆‐value) for each participant was determined by subtracting their postintervention from their preintervention values. Percent change was calculated as the absolute change divided by the preintervention value multiplied by 100. The median of the individual percent change along with minimum and maximum values were then calculated. Given the exploratory nature of the study, we first conducted univariate analyses to screen for metabolites potentially associated with avocado intake using pre–post metabolite change >10% and an unadjusted P < 0.10. Multilevel partial least squares discriminant analysis (multilevel) models were constructed separately for primary and lipid‐related metabolites to account for the paired pre–post study design by subject‐specific centering using subject identification, which removes between‐subject variability and isolates within‐subject correlated changes over time. A weighted sum of squares variable influence on projection >1.0 cutoff was used to identify significant metabolites that changed over the intervention period. The classification performance of the partial least squares discriminant analysis models was evaluated using 5‐fold cross‐validation with 50 repeats. Associations of absolute change in metabolites with absolute change in CMRFs were estimated using multiple regression, adjusting for covariates (age, sex, body mass index, smoking status, physical activity, and energy intake) and multiple testing using false discovery rate at <0.1. Regression coefficients (β) and corresponding 95% CIs were obtained directly from the multiple linear regression models and calculated using symmetric Wald intervals (β ± 1.96 × SE). All analyses were conducted using R software version 4.3.3.

RESULTS

Participant Characteristics

The average age of the included participants was 49 years, of whom 70% were women (Table 1). Approximately 35% of the participants self‐identified as Hispanic or Latino. The racial distribution was 79% White, 10% Black, 3% Asian, and 8% Other (includes participants who either reported mixed‐race or did not answer). Approximately 50% of the subgroup reported attending a 4‐year college, graduate, or professional school, with <7% reporting no formal education or chose not to answer. The majority of the subgroup (81%) were nonsmokers, and approximately 69% reported engaging in regular physical activity. The baseline characteristics of the LLU subset were similar to the overall HAT cohort participants randomized to the avocado intervention (Table 1), except for a higher percentage of participants self‐identifying as Hispanic, which reflects the demographic distribution of the population at the LLU site.

Table 1.

Baseline Characteristics of Participants in the LLU Subset and the Avocado‐ Supplemented Group of the HAT Cohort^*

Variables	Avocado‐supplemented diet group
	LLU cohort (n = 120)	HAT cohort (n = 505)
Age, y	49±13	50±14
Sex, n (%)
Women	84 (70)	356 (71)
Men	36 (30)	149 (29)
Ethnicity, n (%)
Hispanic	41 (34)	106 (21)
Not Hispanic	79 (66)	399 (79)
Race, n (%)
Asian	4 (3)	26 (5)
Black	12 (10)	68 (14)
White	95 (79)	361 (71)
Other†	9 (8)	50 (10)
Education, n (%)
Community college	43 (36)	111 (21)
Four‐year college	31 (26)	167 (33)
Graduate/professional school	31 (26)	163 (32)
High/vocational school	6 (5)	33 (7)
None	9 (7)	30 (6)
Smoking history, n (%)
No	97 (81)	394 (78)
Yes	23 (19)	111 (22)
Regular physical activity, n (%)
No	37 (31)	151 (30)
Yes	83 (69)	351 (70)

HAT indicates Habitual Diet and Avocado Trial; and LLU, Loma Linda University.

^*Continuous variables are presented as the mean±SD and categorical variables are presented as number (percent).

^† The "other" category includes individuals who reported mixed race (without specifying the races) or who did not provide an answer.

Diet Composition

Incorporating 1 avocado into the diet resulted in higher daily intake of energy, total fat, predominantly MUFA, dietary fiber (both soluble and insoluble), as well as vegetable protein, postintervention compared with preintervention. No significant shifts were observed in dietary carbohydrate, total protein including animal protein, and saturated fatty acid intake. Slightly higher intake of some vitamins and micronutrients were also noted. The sodium and calcium intake did not change significantly during the 6‐month intervention period (Table S1).

Cardiometabolic Risk Factors

Avocado consumption did not significantly alter body composition measures (body mass index, waist circumference, hepatic fat fraction), or triglyceride, high‐density lipoprotein cholesterol, glucose, insulin, and high‐sensitivity C‐reactive protein concentrations (Table 2). Modest, but significant (all P < 0.05) decreases in TC (−5%), LDL‐C (−6%), systolic blood pressure (−3%), and diastolic blood pressure (−2%), as well as an increase in VAT volume (3%), were observed.

Table 2.

Cardiometabolic Risk Factors at Baseline (Preintervention) and at the End of the 6‐Month Avocado Intervention (Postintervention)

Cardiometabolic risk factors	Preintervention* (n = 120)	Postintervention* (n = 120)	% change†	P value‡
Body composition
BMI (kg/m2)	32.5 [20.7 to 51.6]	32.5 [20.8 to 52.7]	0.57 [−9.0 to 9.8]	0.14
Waist circumference (cm)	110 [89 to 151]	111 [84 to 154]	0.65 [−8.3 to 12.8]	0.21
Women	107 [89 to 151]	109 [84 to 154]	0.80 [−8.3 to 12.8]	0.25
Men	117 [102 to 138]	117 [103 to 141]	0.57 [−6.6 to 3.6]	0.57
Visceral adipose tissue (L)	3.19 [1.2 to 7.6]	3.26 [1.2 to 7.7]	3.31 [−17.8 to 104.4]	<0.01
Hepatic fat fraction (%)	7.80 [1.0 to 49]	7.70 [4.0 to 49]	11.0 [−99.7 to 528.8]	0.08
Cardiometabolic risk factors
Systolic blood pressure (mm Hg)	122 [93 to 176]	118 [90 to 164]	−2.53 [−25.2 to 29.8]	<0.01
Diastolic blood pressure (mm Hg)	80 [55 to 103]	76 [55 to 108]	−2.47 [−21.4 to 15.6]	<0.01
Total cholesterol (mg/dL)	184 [101 to 269]	173 [109 to 261]	−4.75 [−39.5 to 37.1]	<0.01
LDL‐C (mg/dL)	106 [30 to 192]	99 [29 to 194]	−6.30 [−53.7 to 42.7]	<0.01
HDL‐C (mg/dL)	46 [28 to 86]	45 [24 to 77]	−0.52 [−21.0. 43.7]	0.45
Women	48 [30 to 86]	47 [31 to 77]	−0.70 [−21.0 to 43.7]	0.34
Men	40 [28 to 58]	41 [24 to 58]	−0.45 [−18.3 to 20.2]	0.88
Triglycerides (mg/dL)	114 [36 to 995]	108 [42 to 809]	−1.34 [−48.9 to 130.8]	0.82
Insulin (μIU/mL)	15.6 [2.4 to 175]	16.5 [2.8 to 286]	8.12 [−72.2 to 197.9]	0.08
Glucose (mg/dL)	102 [83 to 282]	102 [79 to 236]	0.35 [−50.0 to 22.0]	0.48
hsCRP (mg/L)	3.8 [0.2 to 31.3]	4.2 [0.3 to 28.1]	0.01 [−53.0 to 128.0]	0.90

BMI indicates body mass index; HDL‐C, high‐density lipoprotein cholesterol; hsCRP, high‐sensitivity C‐reactive protein; and LDL‐C, low‐density lipoprotein cholesterol.

^*Values are median [minimum to maximum].

^† Percent change [median with minimum and maximum values] was calculated as postintervention−preintervention/preintervention value*100.

^‡ P values for pre–post differences were determined using the nonparametric Wilcoxon test, adjusted for multiple testing using false discovery rate ≤0.1.

Serum Metabolite Profiles

Among the 2153 metabolites detected in both pre‐ and postintervention serum samples, approximately 268 (12%) were annotated metabolites. These included 99 primary metabolites and 169 lipid‐related metabolites. The primary metabolites were grouped into 11 classes: amines, amino acids and derivatives, carboxylic acids and derivatives, keto acids, benzenes and derivatives, carbohydrates and derivatives (monosaccharides, disaccharides, sugar acids, sugar alcohols, sugar phosphates), hydroxy acids and derivatives, indoles and derivatives (gut‐derived), nucleosides and related, xenobiotics, and others (Figure 1A). The lipid‐related metabolites were grouped into 6 major classes: fatty acids, fat soluble vitamins, glycerolipids (monoglycerides, diglycerides, triglycerides), glycerophospholipids (glycerol‐α‐phosphate, lysophosphatidylcholine and phosphatidylcholine), phosphatidylethanolamine, phosphatidylinositol, sphingolipids (sphingomyelins, ceramides), and sterols (cholesteryl esters, cholesterol, squalene) (Figure 1B).

Figure 1. Summary of annotated serum metabolites.

Serum metabolites are separated into (A) primary metabolites and (B) lipid‐related metabolites. The pie charts depict metabolites by major class with number of metabolites within each class.

In exploratory univariate analyses, we observed pre–post intervention differences (percent change ≥10%, P ≤ 0.10) in 31 known metabolites including:1 sugar acid (threonic); 2 sugar alcohols (1‐5‐anhydroglucitol, xylitol): 2 carboxylic acids (succinic, malic); 3 amino acids (cysteine, taurine, glutamic acid); 1 gut‐derived metabolite (indole‐3‐acetate), and 22 lipid species (monoglycerides, triglycerides, phosphatidylcholine, and ceramides). These differences did not remain significant after false discovery rate adjustment. The median pre‐ and postintervention values and percent change for primary and lipid‐related metabolites are presented in Table S2A and Table S2B, respectively.

In multilevel partial least squares discriminant analysis of primary metabolites, there was overlap between the pre‐ and postintervention time point estimates (Figure 2A). Component 1 (x axis) explained 5% of the total variance and showed greater separation between time points, suggesting it captured subtle metabolite changes that are specifically linked to avocado consumption. In contrast, component 2 (y axis) explained a larger portion of the total variance (27%) but contributed less to separation between time points, likely reflecting individual metabolic variability unrelated to the intervention. A total of 30 metabolites with variable influence on projection scores >1 were identified as the key drivers of the pre‐ and postintervention separation (Figure 2B). These metabolites spanned multiple classes, with the majority (71%) showing an increase postintervention compared with preintervention, as depicted in the Figure 2B binary heatmap. These included amino acids (aspartic, cystine, glutaric), carbohydrates (sugar acids [threonic, deoxythreonic, glucuronic] and sugar alcohols [lyxitol, threitol]), carboxylic acids (succinic, isocitric, fumaric, gluconic), hydroxy acid (hydroxybutanoic), gut‐derived indoles (methoxy‐tryptamine, indole‐3‐acetate, indole‐3‐propionic acid), and xenobiotics (levoglucosan). There were 8 metabolites (carboxylic acids [maleic, pipecolinic, gluconic], amino acids and derivatives [hypotaurine, amino‐isobutyric, creatinine], hydroxy acid [malic], and indoles [indoxyl sulphate]) that decreased postintervention.

Figure 2. Primary metabolites.

Multilevel PLS‐DA plots depicting paired pre‐ and postintervention differences for primary metabolites (A) with corresponding VIP scores and binary heatmap of the metabolite median percent change (B). Each point represents an individual participant's metabolomic profile, with preintervention samples shown in orange and postintervention samples in blue. Ellipses represent 95% CIs for each intervention time point. Component 1 and component 2 explained 5% and 27% of the total variance, respectively. VIP scores were calculated to identify metabolites driving the separation between groups. Key metabolites (VIP scores >1) contributing to the observed difference are shown in the loadings plot. Percent change was calculated as postintervention−preintervention)/preintervention value*100, with red indicating an increase and green indicating a decrease in metabolite concentration. Expl.var indicates explained variance; PLS‐DA, partial least squares discriminant analysis; and VIP, variable importance in projection.

For lipid‐related metabolites (Figure 3A), the multilevel partial least squares discriminant analysis plot also revealed overlap in estimates, with component 1 and component 2 explaining 26% and 18% of the variation, respectively. Forty‐five lipids with variable influence on projection scores ≥1 were identified (Figure 3B). The majority were glycerolipids, including triglyceride species (73%) with long carbon chains (C46 to C58 with 1–9 double bonds), and glycerophospholipids (18%) including phosphatidylcholine (37:3, 38:6 isomer B, 40:8, 34:0 or phosphatidylcholine O 34:1), phosphatidylethanolamine (36:2, P 38:4 or phosphatidylethanolamine O 38:5), and sphingomyelin (d32:1). Most of these lipids were decreased postintervention, with the exception of 5 metabolites (fatty acid 16:1, monoglyceride 18:0, diglyceride 36:2, diglyceride 37:7, and triglyceride 54:3) (Figure 3B binary heatmap).

Figure 3. Lipid metabolites.

Multilevel PLS‐DA plots depicting paired pre‐ and postintervention differences for lipid‐related metabolites (A) with corresponding VIP scores and binary heatmap of the metabolite median percent change (B). Each point represents an individual participant's metabolomic profile, with preintervention samples shown in orange and postintervention samples in blue. Ellipses represent 95% CIs for each group. Component 1 and component 2 explained 26% and 18% of the total variance, respectively. VIP scores were calculated to identify metabolites driving the separation between groups. Key metabolites (VIP scores >1) contributing to the observed difference are shown in the loadings plot. Percent change was calculated as postintervention−preintervention)/preintervention value*100, with red indicating an increase, and green indicating a decrease in metabolite concentration. DG indicates diglyceride; expl.var, explained variance; FA, fatty acid; MG, monoglyceride; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PLS‐DA, partial least squares discriminant analysis; SM, sphingomyelin; TG, triglyceride; and VIP, variable importance in projection.

Association Between Metabolites and Cardiometabolic Risk Factors

Several metabolite‐CMRF associations were identified after covariate adjustment (P ≤ 0.05).

(Figure S2 and Data Table S3). After false discovery rate adjustment, significant associations remained between 96 metabolites and VAT, plasma triglycerides, and plasma TC concentrations, as summarized below. Of note, all associations were in the weak (multiple‐R values 0.10–0.20) to modest (multiple‐R values 0.21–0.50) range. ³⁷

Lipid Profiles

A total of 5 metabolites belonging to the glycerophospholipid (phosphatidylcholine 30:0) and glycerolipid (triglyceride 44:0, triglyceride 46:0, triglyceride 46:2, triglyceride 48:1) classes were significantly and positively associated with changes in plasma TC (Figure 4A). There were 36 triglycerides (C42 to C58 with 0–7 double bonds); and 2 glycerophospholipids (phosphatidylcholine 34:1 and phosphatidylcholine 38:3), that showed significant positive associations with changes in plasma triglyceride concentrations. Five metabolites, phthalic acid, fumaric acid, 3‐hydroxybutyric acid, lysophosphatidylcholine 18:2, and sphingomyelin d40:0 exhibited inverse associations with plasma triglyceride concentrations (Figure 4B).

Figure 4. Associations of absolute change in metabolites with absolute change in plasma total cholesterol, triglyceride and visceral adiposity tissue volume.

The absolute change (∆‐values) in plasma total cholesterol (A), plasma triglyceride (B), and visceral adiposity tissue volume (C) was calculated by subtracting the postintervention from the preintervention values. β‐coefficients with 95% CIs were estimated using multiple regression, adjusting for covariates (age, sex, body mass index, smoking status, physical activity and energy intake) and multiple testing using Benjamini and Hochberg procedure to control the FDR at <0.1. CE indicates cholesteryl ester; Cer, ceramide; DG indicates diglyceride; FDR, false discovery rate; LPC, lysophosphatidylcholine; PC, phosphatidylcholine; PE, phosphatidylethanolamine; SM, sphingomyelin; TC, total cholesterol; TG, triglyceride; and VAT, visceral adiposity tissue volume.

Visceral Adiposity

Among the 48 metabolites significantly associated with VAT (Figure 4C), the majority (95%) were lipids, including 24 glycerophospholipids (phosphatidylcholine 31 to 42, phosphatidylethanolamine 36 to 38, and lysophosphatidylcholine 18:1 and 18:2), 10 glycerolipids (diglyceride 36:2, triglyceride C54 to C58 with 1–6 double bonds), 10 sphingolipids (sphingomyelin d33 to 42 and ceramide d40:1), and 2 cholesteryl esters (18:1 and 18:2). Most of these associations were negative. In contrast, 2 primary metabolites, aspartic acid and glutaric acid, were positively associated with VAT.

DISCUSSION

This secondary analysis explored whether incorporating 1 avocado per day in habitual diets for 6 months altered serum metabolomic profiles and was associated with visceral adiposity and CMRF measures in a subset of HAT participants with abdominal obesity. Results indicate subtle but significant shifts in metabolite profiles, with 30 primary and 45 lipid‐related metabolites contributing to the separation between pre‐ and postintervention time points. These metabolites represented a wide range of classes, including amino acids, carbohydrates, lipids, microbially derived, and xenobiotics. Several of the metabolites were found in both component 1 and component 2, suggesting that avocado consumption influenced multiple metabolic pathways, some of which were directly responsible for the observed separation, whereas others most likely reflect underlying individual variability. Additionally, several metabolites were associated with changes in VAT and plasma triglycerides and TC concentrations. The change in lipid‐related metabolites (glycerolipids, glycerophospholipids, sphingolipids) were generally positively associated with change in plasma triglycerides and TC, but negatively associated with change in VAT, whereas change in primary metabolites (amino acids, carboxylic acids, hydroxy acids) were positively associated with change in VAT and negatively associated with plasma triglyceride changes. This potentially suggests that alterations in lipid metabolites may reflect systemic lipid handling and circulating lipoprotein changes, whereas primary metabolites may relate to tissue‐level energy metabolism in the context of VAT remodeling. Of note, these associations, albeit statistically significant, were mostly weak to modest, highlighting the need for further investigation in larger cohorts to confirm these findings.

In our subset of HAT participants, avocado consumption led to favorable modifications in the composition of the participants diets and was associated with a modest decrease in total and LDL‐C concentrations and blood pressure. These improvements in diet quality and lipid profile are comparable with those reported for the entire HAT cohort, ²³ and consistent with findings from other avocado intervention studies. ¹⁴ , ¹⁶ , ³⁸ Interestingly, the significant but small decreases in systolic blood pressure and diastolic blood pressure, and the increase in VAT were not observed in the main cohort, ²³ or another HAT ancillary study that assessed vascular function and blood pressure. ³⁹ This likely reflects variations in study design (between‐group versus pre–post) and differences in the demographic composition of the cohorts. However, some other studies have suggested that avocado consumption may improve vascular function, which could have favorable effects on blood pressure. ⁴⁰ , ⁴¹ Additionally, a recent study reported that higher avocado intake (5+ servings per week) compared with non‐ or low‐ avocado consumers was associated with a lower incidence of hypertension in Mexican women, ⁴² potentially attributed to its high potassium, magnesium, fiber, and MUFA content.

In regard to the effect of avocado intake on visceral adiposity, results are limited and inconsistent, likely due to the intensive nature of the measurement ²² , ²³ , ⁴³ , ⁴⁴ and differences in measurement methodology, study cohort demographic characteristics, dietary patterns, baseline adiposity, and metabolic health status. The HAT study specifically recruited individuals with elevated waist circumference, a proxy for visceral fat accumulation. Thus, the observed increase in VAT may reflect the participants' underlying metabolic trajectory over the 6‐month intervention, though intervention effects and the possibility of random variation cannot be ruled out.

The present analysis is the first to document alterations in several primary and lipid‐related metabolites in serum following avocado consumption. A few dietary biomarker discovery studies have reported higher concentrations of perseitol and mannoheptulose (sugar alcohols), ⁴⁵ , ⁴⁶ as well as glucoheptose (a xenobiotic) with avocado intake, ⁴⁷ suggesting these may be potential urinary biomarkers of avocado intake, but these results have not been corroborated. Only 1 cross‐sectional study to date has investigated serum metabolomic profiles associated with avocado intake in a multiethnic cohort of older adults. ⁴⁸ Avocado intake (consumers versus nonconsumers) was estimated using a food frequency questionnaire and correlated with baseline metabolomic data. Three highly correlated spectral features were identified that were associated with avocado intake at metabolome‐wide significance levels (P < 5.3 × 10⁻⁷). Because these metabolite annotations were not confirmed, it precludes comparison with our study results.

Avocados have a high lipid content, with neutral lipids comprising the majority (95%–97% of total lipids), primarily as triglycerides (~90%), followed by diglycerides (~4%) and monoglycerides (~2%). ⁴⁹ , ⁵⁰ Specifically, triolein (triglyceride 54:3) is the most abundant triglyceride species, with oleic acid (18:1n‐9) being the predominant MUFA in triglycerides. ⁵¹ Saturated fatty acids account for 15% to 21% of total fatty acids, with palmitic acid (16:0) being the major saturated fatty acid. Polar lipids (~5% of total lipids) include glycolipids, with phosphatidylcholine and phosphatidylethanolamine as the major phospholipid classes. Sphingolipids have also been identified in smaller proportions in avocados. ⁵² The majority of serum metabolites that were altered postintervention in our study participants included several triglyceride, phosphatidylcholine, and sphingomyelin species. Most were lower postintervention. In contrast, fatty acid 16:1 and monoglyceride 18:0 were higher, which aligns with our findings in red blood cell fatty acid MUFA profiles after avocado intake in the overall HAT cohort. ³¹ The majority of the lipid‐related changes observed most likely reflect broader effects of avocado intake within the context of habitual diets on gut microbiota composition and host metabolism, rather than direct avocado‐derived metabolites. However, our metabolomic profiling does not allow us to distinguish microbial contributions from endogenous host metabolic responses.

The observed increase in primary metabolites (sugar acids, sugar alcohols, and carboxylic acids) in serum after avocado intake may result from the fermentation of fiber by gut microbiota. ⁴⁴ , ⁵³ , ⁵⁴ The increase in dietary fiber contributed by avocados is consistent with the dietary data in our subset, as well as that reported for the entire HAT cohort, ⁵⁵ along with the finding that avocados displaced low‐fiber‐containing foods such as added sugars and refined grains. ²⁴ Similarly, the increase in methoxy‐tryptamine, indole‐3‐acetate, and indole‐3‐propionic acid are plausibly driven by gut microbial metabolism of tryptophan, consistent with the view that fiber‐rich diets favor microbial pathways toward indole production. ⁵⁶ , ⁵⁷ Additionally, prior metabolomics work has linked threonic acid (a sugar acid derived from ascorbate metabolism) with higher intake of fruits and vegetables, supporting its role as a putative biomarker of plant food consumption. ⁴⁷ , ⁵⁸ In our study, the increase in threonic acid likely reflects avocado's contribution to overall plant food exposure.

Conversely the decrease in lipid‐related metabolites (glycerolipids and glycerophospholipids) may reflect a metabolic shift in lipid metabolism toward lipid sequestration following avocado intake. This could involve modulation of lipid transport and turnover through the extracellular actions of lipoprotein lipase and lecithin–cholesterol acyltransferase, which hydrolyze and remodel triglyceride‐rich lipoproteins in circulation. Such remodeling can lower circulating lipid intermediates while promoting their uptake and deposition into tissues. ⁵⁹ The high MUFA content of avocados may modulate these enzyme‐mediated pathways, ⁶⁰ , ⁶¹ consistent with the observed inverse associations between lipid‐related metabolites and VAT.

In contrast, primary metabolites (amino acids, carboxylic acids, and hydroxy acids) showed the opposite pattern, being positively associated with VAT. As previously stated, the HAT cohort specifically focused on individuals with abdominal obesity. Higher VAT has been shown to be associated with an altered metabolite profile, especially related to amino acid and carbohydrate metabolism. ⁶² , ⁶³ In particular, aspartic acid, a key intermediate in the urea cycle and the malate–aspartate shuttle, pathways essential for maintaining energy homeostasis under metabolic stress, has previously been linked to obesity‐related phenotypes, supporting a potential role in lipid accumulation. ⁶³ , ⁶⁴ Similarly, glutaric acid has been shown to be associated with higher TC and LDL‐C levels in metabolomic studies, suggesting a possible link to altered lipid metabolism. ⁶⁵ , ⁶⁶ Because serum metabolite profiles capture both diet and metabolism, the latter driven by health status, our results most likely reflect the underlying enhanced metabolic VAT activity in our study participants.

The positive association between specific lipid species (triglycerides 44:0, 46:0, 48:1, 46:2, and phosphatidylcholine 30:0) and plasma TC concentrations might reflect alterations in lipoprotein composition or intermediates in the metabolic pathways, which were associated with a modest reduction in TC and LDL‐C. Interestingly, plasma triglyceride concentrations did not change significantly with avocado intake, although there was an overall trend toward a slight decrease, with marked variability among participants (−95% to 57%). Plasma triglyceride concentrations represent the sum of individual triglyceride species, so opposing changes (i.e., decreases in some and increases in others) may have resulted in the nonsignificant net change. Our metabolite analysis revealed alterations in several individual triglyceride species (C42 to C58) and certain glycerophospholipids, which were positively associated with plasma triglyceride concentrations, suggesting preferential incorporation of these triglyceride species into very‐low‐density lipoproteins, with avocado intake. Conversely, the inverse associations between fumaric acid, 3‐hydroxybutyric acid, lysophosphatidylcholine 18:2, and sphingomyelin d40:0 with plasma triglycerides suggest potential alterations in lipid oxidation pathways. This interpretation is supported by the fact that elevated 3‐hydroxybutyric acid, a ketone body, typically indicates increased fatty acid oxidation. Fumaric acid, an intermediate in the tricarboxylic acid and urea cycles, plays a role in energy metabolism. Lysophosphatidylcholines, including lysophosphatidylcholine 18:2, are involved in lipid signaling, transport, and improved lipid clearance. ⁶⁷ Sphingomyelins, components of cell membranes and lipoproteins, influence lipid metabolism by regulating lipoprotein processing and clearance. ⁶⁸ Increased sphingomyelin d40:0 levels might be linked to enhanced triglyceride‐rich lipoprotein turnover via increased lipolysis or improved hepatic clearance. Currently, there are no comparable data available on the role of these primary or lipid‐related metabolites, specifically in the context of avocado consumption. Further research is needed to clarify their functional significance and potential mechanisms in lipid regulation.

This ancillary study is the first to comprehensively assess metabolite profiles associated with long‐term avocado consumption, offering novel insights into the role of whole‐food dietary interventions in cardiometabolic health. The integration of lipid‐related and primary metabolites provides a broader understanding of metabolic interactions and their link to CMRFs. The study's focus on individuals with abdominal obesity, a group at elevated risk for cardiometabolic disorders, provides additional insight into this high‐risk population. However, this study was designed to be an exploratory analysis, with the analysis restricted to the avocado group, precluding causal inferences relative to a control group. The use of a convenience sample may increase the risk of selection bias. Because women were overrepresented in our data set, residual confounding related to sex cannot be ruled out even after covariate adjustment. This may limit the extent to which the results can be generalized across sexes. Only annotated metabolites were included in the reported data set, and metabolomic profiling was not conducted in duplicate, so technical variability (coefficient of variation) could not be directly assessed. Although key covariates were included in the statistical analysis, other potential confounders may have influenced the results. Also, because our participants were free‐living, dietary intake was variable. The limited metabolomic variation observed likely reflects the relatively small sample size, the free‐living dietary context, and the addition of a single food item, rather than a lack of biological effect or methodological limitations, because standardized protocols, internal standards, and pooled quality control samples were used throughout the analysis. Thus, the metabolite shifts observed likely represent both avocado intake and its integration into overall dietary patterns, which could have also weakened associations with CMRFs. In addition, we cannot distinguish microbial contributions from endogenous host metabolic responses, which restricts our ability to determine whether observed metabolite changes reflect direct host processes or microbiome‐mediated effects. Although this limits the strength of inferences, it provides preliminary data that can inform sample‐size planning and the selection of metabolites and pathways for more comprehensive future analyses.

In conclusion, daily avocado consumption for 6 months resulted in subtle shifts in serum primary and lipid‐related metabolite profiles. Changes in several of these metabolites were weakly associated with changes in VAT, TC, and triglyceride concentrations. Replication in a larger cohort is needed to confirm these trends.

Sources of Funding

This work was supported by the Avocado Nutrition Center and the US Department of Agriculture (agreement number 58‐1950‐4‐401). Any opinions, findings, conclusions, or recommendations expressed in this publication are those of authors, and do not necessarily reflect the view of the US Department of Agriculture.

Disclosures

None.

Supporting information

Tables S1–S3

Figures S1–S2

STROBE Checklist