Lipidomic Markers of Processed Meat and Unprocessed Red Meat Intake and Risk of Diabetes in American Indians

Abstract

OBJECTIVE

To identify lipidomic markers of habitual unprocessed red meat and processed meat intake and evaluate their associations with diabetes risk in American Indians.

RESEARCH DESIGN AND METHODS

We studied 1,816 participants from the Strong Heart Family Study. Using untargeted liquid chromatography-mass spectrometry, we quantified 1,542 lipids (518 known) in fasting plasma at baseline and follow-up (∼5 years apart). Meat intake was assessed via Food Frequency Questionnaires. Mixed-effects linear regression was used to identify lipids associated with meat consumption. Mixed-effects logistic regression was used to examine whether these lipids were associated with incident diabetes, independent of conventional risk factors, or with longitudinal glucose/insulin metrics.

RESULTS

Diabetes developed in 66 of 1,076 participants with normal baseline glucose. After multiple testing correction, 15 known lipids, primarily plasmalogens, were associated with unprocessed red meat intake. Three plasmalogens were linked to incident diabetes (odds ratio [OR] 1.32 [95% CI 1.02–1.70] to 1.39 [1.08–1.78] per SD increase in baseline levels) and higher red meat intake. Eight lipids, mainly sphingomyelins, were associated with processed meat intake. Two sphingomyelins were linked to incident diabetes (OR 1.33 [95% CI 1.02–1.75] and 1.36 [1.04–1.80]) and higher processed meat intake. Of 23 meat-related lipids, 20 were associated with altered glucose/insulin metrics, and 11 mediated positive associations between red or processed meat intake and fasting glucose.

CONCLUSIONS

We identified lipidomic markers of unprocessed red and processed meat consumption. Several lipids were independently associated with increased diabetes risk, potentially by mediating the association between meat intake and glucose metabolism.

Graphical Abstract

Introduction

Consumptions of unprocessed red meat and processed meat have been consistently associated with an increased risk of type 2 diabetes (T2D) in various populations, including American Indians (1,2). As such, limiting dietary intake of red or processed meat is a key component in several commonly used healthy dietary patterns (3–5). Previous studies have proposed several explanations for the link between meat consumption and diabetes risk, including the high saturated fat, heme iron, and trimethylamine N-oxide metabolism associated with red meat, as well as the high sodium, nitrates, and advanced glycation end products associated with processed meat (1,6,7). However, the mechanisms behind the observed associations between dietary intake of red meat and processed meat and risk for diabetes remain incompletely understood. American Indians suffer from the highest prevalence of T2D among all U.S. racial/ethnic groups (8). In addition, the American Indian community consumes traditional Native American foods and practices unique lifestyles. Examining the association of dietary intake, such as meat consumption, with diabetes may help us understand the high burden of T2D in this minority population.

Lipidomics is an emerging high-throughput technology that can accurately identify and quantify hundreds to thousands of lipids in a biological sample. These lipid metabolites represent the end products of cellular metabolism in response to environmental stimuli, such as meat consumption, and may thus serve as markers for food intake and be used to examine the role of dietary intake in disease risk. Previous epidemiological studies have explored markers of habitual red meat and/or processed meat intake using a lipidomics/metabolomics approach (9–16). However, existing studies have been largely cross-sectional and limited by low coverage of the blood lipidome. In addition, no research on lipidomic markers of meat consumption has specifically focused on populations with an increased risk of diabetes, such as American Indians. Despite the close link between meat intake and diabetes, only one study of a European population (10) examined the association of meat intake-associated metabolite score with diabetes risk. To date, a comprehensive assessment of lipidomic markers for habitual meat intake (i.e., unprocessed red meat and processed meat) in relation to the risk of diabetes is still missing, especially in a longitudinal setting.

We previously reported an association between high processed meat consumption, but not unprocessed red meat, with diabetes risk in American Indians (2). To elucidate the mechanisms behind these observations, the current study examined 1) the associations of 1,542 fasting plasma lipids with habitual intake of unprocessed red and processed meat; and 2) whether the identified lipids are independently associated with risk of T2D, beyond traditional risk factors. These findings may provide mechanistic insights into the meat consumption-T2D relationship and inform tailored dietary recommendations for American Indians, a minority group suffering from a disproportionately high risk of T2D.

Research Design and Methods

Study Population

The Strong Heart Family Study (SHFS, 2001–ongoing) is a family-based prospective study of cardiometabolic diseases in American Indians. Briefly, 2,786 Tribal members (aged ≥14 years) residing in Arizona, North Dakota, South Dakota, and Oklahoma were examined in 2001–2003 (baseline) and reexamined in 2006–2009 (after ∼5-year follow-up). Informed consent was obtained from all participants. The SHFS protocols were approved by the Institutional Review Boards of the participating institutions and tribes.

The flowchart of participant inclusion and exclusion is shown in Supplementary Fig. 1. We included 1,816 participants in the analysis of lipidomic markers for meat intake who had available dietary and lipidomics data at baseline. More details about the study population and inclusion criteria are provided in the Supplementary Material.

Assessment of Red Meat and Processed Meat Intake

An interviewer-administered Block 119-item Food Frequency Questionnaire with a supplemental American Indian food questionnaire was administered to SHFS participants at baseline to assess usual dietary intake over the past year. We used these dietary data to quantify habitual intake of processed meat (i.e., breakfast sausage, SPAM, hot dogs, lunch meat, and bacon) and unprocessed red meat (i.e., pork chops, pork roast, veal, lamb, deer, ribs, hamburger, cheeseburger, roast beef, steak). Consistent with previous studies in the same cohort (2), we considered 100 g (3.5 oz) and 50 g (1.8 oz) as one serving of unprocessed red meat and processed meat, respectively. Dietary data were not available at the 5-year follow-up. More details about dietary data collection in the SHFS are provided in the Supplementary Material.

Ascertainment of Incident T2D

Our primary outcome of interest was incident T2D. Incident T2D was identified in individuals with normal fasting glucose at baseline (2001–2003) who developed T2D by the end of the 5-year follow-up (2006–2009). T2D was defined as fasting plasma glucose ≥126 mg/dL or receiving hypoglycemic medications based on the American Diabetes Association criteria.

Assessment of Glucose and Insulin Metrics

Glucose and insulin metrics included fasting glucose, insulin sensitivity, and insulin resistance. Fasting blood samples were collected after overnight fasting. Fasting glucose was measured by standard laboratory methods (17). Insulin sensitivity was estimated by calculating the quantitative insulin sensitivity check index (QUICKI): QUICKI = 1/(log insulin [mU/L] + log baseline glucose [mg/dL]). Insulin resistance was assessed by HOMA-insulin resistance (IR), which was calculated as: HOMA-IR = fasting glucose (mg/dL) × insulin (μU/mL)/405. We measured the three metrics in the participants at both study visits.

Assessments of Covariates

Sociodemographic information, lifestyle factors (cigarette smoking and alcohol use), medical history, and use of prescription medications (e.g., use of antihypertensive, glucose-lowering, and/or lipid-lowering drugs) were collected using structured questionnaires at both examinations (17,18). Physical activity levels were quantified using mean pedometer readings over 7 days. Detailed laboratory and measurement methods are provided in the Supplementary Material.

Lipidomic Data Acquisition, Processing, and Normalization

Detailed methods for lipidomic data acquisition, processing, and normalization in the SHFS are described in the Supplementary Material. Briefly, the relative abundance of individual lipid species in fasting plasma samples was quantified by untargeted liquid chromatography-mass spectrometry (LC-MS). After preprocessing and quality control, we obtained 1,542 lipids (518 known and 1,024 unknown) in 1,957 participants at baseline and 1,948 participants at the 5-year follow-up.

Statistical Analysis

Supplementary Fig. 1 illustrates the procedures for participant selection and statistical analyses. Statistical analyses were performed using R 4.1.1 software (R Foundation for Statistical Computing, Vienna, Austria). All continuous variables, including lipid measurements, were standardized to zero mean and unit variance prior to analyses, unless otherwise stated.

Associations of Lipidomics With Meat Intake

To identify individual lipid species associated with meat intake at baseline, we constructed mixed-effects linear regression models. In these models, the relative abundance of each lipid (continuous) was the independent variable, and meat intake (red or processed meat intake separately) in servings/day was the dependent variable. Covariates were determined based on previous literature, including age (years), sex (male/female), study center (Arizona, Oklahoma, or North Dakota/South Dakota), education (years), BMI (kg/m²), fasting glucose (mg/dL), total energy intake (kcal/day), smoking (never/former/current), alcohol drinking (never/former/current), physical activity (steps/day), lipid-lowering medication use (yes/no), and hypertension (yes/no). The models included random effects to account for relatedness between family members. Multiple testing was controlled by false discovery rate using the Storey q-value method (19), wherein statistical significance was set at a q < 0.05 level. We also conducted pathway enrichment analysis using lipids with P < 0.05. Detailed methods are described in the Supplementary Material.

Associations of Meat Intake-Related Lipids With Incident T2D

We assessed the associations of the identified lipidomic markers with the 5-year risk of incident T2D. For these analyses, mixed-effects logistic regression models were used. Incident T2D (yes/no) was the dependent variable, and baseline level of each lipid was the independent variable. The same set of covariates described above was used.

Associations of Meat Intake-Related Lipids With Glucose/Insulin Homeostasis Metrics

The glucose/insulin homeostasis metrics included fasting glucose, insulin sensitivity, and insulin resistance. We examined the longitudinal associations between meat intake-related lipids (independent variables) and glucose/insulin metrics (dependent variables) using mixed-effects linear regression models. In these analyses, baseline and follow-up data were used as repeated measurements for an individual. We accounted for the family relatedness among the participants and correlation within the same individual between two time points by including random intercept terms in these models. The models were adjusted for all the covariates mentioned above, except fasting glucose.

Furthermore, we investigated whether the identified lipidomic markers mediate the association between meat intake and glucose/insulin metrics using the “mediation” R package (20). We used the baseline data for the mediation analyses as meat intake data were only available at baseline. Mediation percentage of lipids was calculated as the average causal mediation effects relative to the total effect. Based on 1,000 Monte Carlo draws for quasi-Bayesian approximation, we calculated the statistical significance and the CI of the estimated mediation effects of lipids.

Sensitivity Analyses

To examine the robustness of our results, we performed a series of sensitivity analyses, including sex-stratified analyses. More details of these analyses are described in the Supplementary Material.

Data and Resource Availability

The phenotype data used in this study can be requested through the Strong Heart Study (https://strongheartstudy.org/). The lipidomics data can be obtained from the corresponding author upon a reasonable request.

Results

Characteristics of Study Participants

Table 1 shows the baseline characteristics of the study participants. Among the 1,816 individuals included in the analyses, 63% were women, and the mean age was 40 years (range 18–75 years) at baseline. The mean processed and red meat intake was 0.8 (SD 0.9) and 0.7 (SD 0.8) servings/day, respectively. Participants with higher processed meat intake were more likely to be younger, male, less educated, current smokers, consume alcohol, have higher total energy intake, higher fasting glucose, and a higher prevalence of diabetes at baseline compared with those with lower intake.

Table 1.

Baseline characteristics of the participants by categories of processed meat intake

		Processed meat intake categories, servings/day
Characteristics	Total (N = 1,816)	<1/4 (n = 469)	1/4 to 1/2 (n = 423)	1/2 to 1 (n = 451)	≥1 (n = 473)
Processed meat intake, servings/day*	0.8 (0.9)	0.1 (0.1)	0.4 (0.1)	0.7 (0.1)	2.0 (1.2)
Red meat intake, servings/day*	0.7 (0.8)	0.3 (0.4)	0.5 (0.4)	0.7 (0.6)	1.2 (1.2)
Age, years	40.3 (13.9)	42.9 (15.4)	41.2 (13.9)	40.0 (13.7)	37.2 (11.9)
Female sex, n (%)	1,139 (62.7)	375 (80.0)	287 (67.8)	251 (55.7)	226 (47.8)
Education, years	12.5 (2.1)	13.0 (2.2)	12.8 (2.2)	12.5 (2.1)	11.9 (1.8)
BMI, kg/m2	31.8 (7.5)	31.2 (7.1)	31.5 (6.8)	31.9 (7.2)	32.7 (8.4)
Smoking status, n (%)
Never	668 (36.8)	201 (42.9)	158 (37.4)	162 (35.9)	147 (31.1)
Former	426 (23.5)	113 (24.1)	108 (25.5)	101 (22.4)	104 (22.0)
Current	722 (39.8)	155 (33.0)	157 (37.1)	188 (41.7)	222 (46.9)
Drinking status, n (%)
Never	147 (8.1)	57 (12.2)	38 (9.0)	31 (6.9)	21 (4.4)
Former	537 (29.6)	164 (35.0)	136 (32.2)	119 (26.4)	118 (24.9)
Current	1,132 (62.3)	248 (52.9)	249 (58.9)	301 (66.7)	334 (70.6)
Physical activity, steps/day	5,785.3 (3,810.4)	5,656.5 (3,946.4)	5,585.3 (3,371.7)	5,851.8 (4,043.2)	6,071.3 (3,807.4)
Total energy intake, kcal/day	2,671.1 (2,024.7)	1,521.3 (843.7)	2,035.2 (980.8)	2,559.3 (1,081.0)	4,486.6 (2,844.8)
Systolic blood pressure, mmHg	122.5 (15.4)	122.4 (16.2)	122.3 (15.2)	122.8 (15.0)	122.6 (15.1)
Fasting glucose, mg/dL	109.6 (46.8)	102.8 (34.5)	106.0 (38.8)	114.2 (52.7)	115.1 (56.0)
Lipid-lowering medicine use, n (%)	63 (3.5)	22 (4.7)	12 (2.8)	17 (3.8)	12 (2.5)
Hypertension, n (%)	531 (29.2)	156 (33.3)	117 (27.7)	127 (28.2)	131 (27.7)
Diabetes, n (%)	335 (18.4)	80 (17.1)	65 (15.4)	92 (20.4)	98 (20.7)

Continuous variables are expressed as mean (SD) and categorical variables as n (%).

*One serving of red meat and processed meat is 100 g (3.5 oz) and 50 g (1.8 oz), respectively.

Associations of Lipidomics With Meat Intake

Among the 1,542 lipids, 202 lipid species (89 known) were associated with red meat intake at P < 0.05 (Supplementary Table 1 and Supplementary Fig. 2). Of these, 29 lipids (15 known) remained significant at q < 0.05. Higher levels of all of the 15 known lipids were associated with higher red meat intake, including 13 glycerophospholipids (i.e., 10 phosphatidylcholine [PC] plasmalogens [PCp] and 3 phosphatidylethanolamine [PE] plasmalogens [PEp]), cholesteryl ester [CE][18:0], and sphingomyelin [SM][d41:1] A [isomer] (Fig. 1 and Supplementary Table 2).

Figure 1.

A: Lipid species associated with red meat intake (q < 0.05) and diabetes risk. B: Lipid species associated with processed meat intake (q < 0.05) and diabetes risk (n = 1,816 for left panel, n = 1,076 for right panel). The β-coefficients of lipids associated with meat intake were obtained by mixed-effects linear regression models, adjusting for age, sex, study center, education, BMI, fasting glucose, total energy intake, smoking, alcohol use, physical activity levels, lipid-lowering medication use, and hypertension status. ORs of baseline lipids associated with incident T2D were obtained by mixed-effects logistic models, adjusting for all covariates above (baseline level). Different colored dots represent different lipid classes.

For processed meat intake, 158 lipid species (58 known) were identified as significant at P < 0.05 (Supplementary Table 1), including 14 (8 known) significant at q < 0.05. These processed meat-related lipids mostly comprised SMs (six of eight). Specifically, three lipids (i.e., CE[22:5] A, SM[d42:2], and SM[d42:2] A) were associated with higher processed meat intake, while five lipids (i.e., PC[p-14:0/16:0]/PC[o-14:0/16:0], SM[d30:1] A, SM[d30:1] B [isomer], SM[d32:2] A, and SM[d32:2] B) were associated with lower intake (Fig. 1 and Supplementary Table 2). A total of 29 lipids overlapped for unprocessed red and processed meat at P < 0.05, but none at q < 0.05 (Supplementary Table 3).

Pathway enrichment analyses demonstrated that the most significantly enriched pathways for red meat-related lipids include the sphingolipid signaling pathway, necroptosis, and insulin resistance. For processed meat-related lipids, the most significantly enriched pathway is glycerophospholipid metabolism (Supplementary Figs. 3 and 4).

Associations of Meat Intake-Related Lipids With Incident T2D

Among the 1,076 participants with normal fasting glucose at baseline, T2D developed in 66 by the end of the ∼5-year follow-up. Among the 15 red meat-related lipids (known lipids), baseline levels of 3 lipids (i.e., PC[p-36:3]/PC[o-36:4], PE[p-38:5]/PE[o-38:6] A, and PE[p-38:5]/PE[o-38:6] B) were associated with a higher risk of T2D and higher red meat intake (Fig. 1; Supplementary Table 2). Each SD increase in the baseline concentrations of these lipids was associated with a 32–39% increase in risk of incident T2D (odds ratio [OR] ranged from 1.32 [95% CI 1.02, 1.70] to 1.39 [1.08, 1.78]). Meanwhile, two of the eight processed meat-related lipids (known lipids) were associated with higher processed meat intake and a higher risk of T2D (ORs were 1.33 [95% CI 1.02, 1.75] and 1.36 [1.04, 1.80] for SM[d42:2] and SM[d42:2] A, respectively).

Associations of Meat Intake-Related Lipids With Glucose/Insulin Homeostasis Metrics

Among the 15 lipids linked to higher red meat intake, 7 were significantly associated with increased fasting glucose, 8 were associated with increased insulin resistance, and 7 were associated with decreased insulin sensitivity (Fig. 2 and Supplementary Table 4). Notably, these significant lipids also included the three lipids linked to a higher T2D risk (i.e., PC[p-36:3]/PC[o-36:4], PE[p-38:5]/PE[o-38:6] A, and PE[p-38:5]/PE[o-38:6] B).

Figure 2.

A: Associations between red meat-related lipids and glucose/insulin homeostasis metrics (n = 1,492). B: Associations between processed meat-related lipids and glucose/insulin homeostasis metrics (n = 1,492). Regression coefficients were obtained from mixed-effects linear regression models, adjusting for age, sex, study center, education, BMI, total energy intake, smoking, alcohol drinking, physical activity levels, lipid-lowering medication use, and hypertension at the study visit. Red color represents positive associations and blue represents negative associations. *P < 0.05, **P < 0.01, ***P < 0.001.

Out of the eight processed meat-related lipids, all five lipids that correlated with lower processed meat intake (i.e., PC[p-14:0/16:0]/PC[o-14:0/16:0], SM[d30:1] A, SM[d30:1] B, SM[d32:2] A and SM[d32:2] B) showed negative associations with fasting glucose. Notably, SM(d42:2) and SM(d42:2) A, which were positively associated with processed meat intake, showed negative associations with fasting glucose and/or insulin resistance and positive associations with insulin sensitivity.

We further investigated the potential mediation role of the identified lipids in the associations between meat intake and glucose/insulin metrics. Among this cohort of American Indian adults, consumption of both unprocessed red and processed meat was independently associated with higher fasting glucose (Supplementary Table 5). On average, each one serving/day increase of red meat and processed meat intake was associated with an increase in fasting glucose of 3.98 mg/dL (95% CI 0.09, 7.88) and 3.53 mg/dL (95% CI 0.17, 6.88), respectively. In addition, processed meat intake was associated with lower insulin sensitivity.

We observed significant mediation effects of lipids on the positive associations of red meat and processed meat intake with fasting glucose. Figure 3 shows the top three lipid mediators with the largest proportion mediated (full results of the mediation analyses are shown in Supplementary Table 6). Six red meat-related lipids (i.e., PC[p-36:3]/PC[o-36:4], PC[p-38:4]/PC[o-38:5] B, PC[p-40:6]/PC[o-40:7] B, PE[p-38:4]/PE[o-38:5], PE[p-38:5]/PE[o-38:6] A, and PE[p-38:5]/PE[o-38:6] B) were identified as significant mediators for the positive association of red meat intake with fasting glucose, with the percentage mediated ranging from 9.7 to 20.1%. Of note, these six lipids included the three lipids that were associated with increased risk of T2D. Meanwhile, five processed meat-related lipids (i.e., PC[p-14:0/16:0]/PC[o-14:0/16:0], SM[d30:1] A, SM[d30:1] B, SM[d32:2] A, and SM[d32:2] B) significantly mediated the positive association between processed meat intake and fasting glucose. The percentage mediated ranged from 10.5 to 22.7%. No mediation effect for insulin resistance, insulin sensitivity, or incident T2D was observed.

Figure 3.

A: Lipids mediating the associations between red meat intake and fasting glucose (n = 1,492). FPG, fasting plasma glucose. B: Lipids mediating the associations between processed meat intake and fasting glucose (n = 1,492). The top three lipids with the largest proportion mediated are shown. Mediation analyses were conducted using baseline data, adjusting for age, sex, study center, education, BMI, total energy intake, smoking, alcohol drinking, physical activity levels, lipid-lowering medication use, and hypertension at baseline. Mediation percentage of lipids was calculated as the average causal mediation effects relative to the total effect.

Our main findings remain largely unchanged in the sensitivity analyses (Supplementary Tables 7–12 and Supplementary Figs. 5–8). Detailed results are described in the Supplementary Material.

Conclusions

In this large-scale lipidomic study of American Indian adults, we have several significant findings. First, we identified multiple lipidomic markers associated with habitual dietary intake of unprocessed red meat and processed meat. Specifically, consumption of unprocessed red meat was mostly associated with PCp and PEp, whereas processed meat consumption was mostly associated with SMs. These findings suggest that different metabolic pathways are likely to be involved in the metabolism of these two types of meat consumption.

Second, lipidomic markers of both types of meat consumption showed robust associations with diabetes risk: 1) a 5-year increased risk of T2D was associated with specific plasmalogens linked to red meat intake, as well as specific SMs linked to processed meat intake, independently of traditional risk factors; 2) most of the identified lipids were longitudinally associated with altered glucose/insulin homeostasis metrics; and 3) many meat intake-related lipids mediated the positive associations between unprocessed red or processed meat intake and fasting glucose levels. These findings deepen our understanding of the relationship between meat intake and diabetes risk and may help explain the high diabetes burden in American Indian communities. If validated, the identified lipidomic markers of meat consumption may help inform dietary recommendations tailored to American Indians.

Previously, we reported that consumption of processed meat, but not unprocessed red meat, was significantly associated with incident T2D among American Indians in the Strong Heart Family Study (2). In the current analysis, we found that both red meat- and processed meat-related lipids were largely associated with altered glucose/insulin homeostasis metrics, with certain lipids independently associated with a higher future risk of T2D. Notably, consumption of both red meat and processed meat were associated with higher fasting glucose levels, while specific PCp and PEp lipids may mediate the role of red meat, and PCp and SM lipids mediate the role of processed meat. In addition, insulin resistance was among the enriched pathways for red meat-related lipidomic markers. The observed associations of red and processed meat intake with higher fasting glucose in our study were supported by previous evidence (21). Interestingly, in the Multi-Ethnic Study of Atherosclerosis, red meat intake was not associated with inflammation markers; but glutamine, a metabolite linked to red meat intake, was associated with C-reactive protein levels (an inflammation marker) (9). Given the significant role of inflammation in diabetes development and progression, together, these findings offer valuable lipidomic insights into the links between red meat intake and diabetes risk. Additionally, they suggest that incorporating lipid or metabolite markers could be instrumental in assessing health risks associated with red meat consumption.

We identified plasmalogens, including PCp and PEp, as primary lipid markers of red meat intake. This is consistent with results from the Nurses’ Health Study (NHS) and the Health Professionals Follow-up Study (HPFS), which included mostly White participants. In their analyses, PCp and PEp demonstrated the strongest positive correlation with unprocessed red meat consumption among all lipid classes (14). In another NHS report, plasmalogens were also associated with higher total red and processed meat consumption as well as low diet quality (22). Particularly, PEp lipids were also found to be specifically associated with red meat intake in a Chinese cohort (23) and in a U.S. cohort of postmenopausal women (24). One notable study of a European population identified 139 metabolites associated with red meat consumption (10). Their identified lipids also mostly consist of glycerophospholipids (i.e., glycerophosphocholine, glycerophosphoethanolamine, and plasmalogens). Furthermore, the derived red meat metabolite score using these metabolites was associated with T2D incidence.

Our findings align with these studies. We additionally demonstrated that some of the red meat-related plasmalogens mediate glucose levels associated with red meat consumption. Plasmalogens are one unique subclass of glycerophospholipids that play key roles in membrane structure, cellular signaling, and antioxidant defense (25). Altered plasmalogen levels were implicated in cardiometabolic conditions, such as hypertension (26), coronary artery disease (27), prediabetes, and T2D (28). Specific PC and PE plasmalogens were associated with diabetes risk in a previous SHFS report (29), particularly PC(p-38:5)/PC(o-38:6) B and PE(p-38:5/PE(o-38:6) B, which were identified as red meat-related lipids in the current study. These associations may be explained by the role of plasmalogens in regulating oxidative stress and inflammatory responses (30), given that the ether-bond in plasmalogens is solely produced in peroxisomes, the main organelle responsive to oxidative stress. Because diet is a well-established factor influencing oxidative stress and inflammation, it is notable that glycerophospholipids, including plasmalogens, can be directly obtained and affected by dietary intakes, particularly meat (31,32). Taken together, plasmalogen metabolism and peroxisome activation may represent a pathway link between red meat consumption and metabolic health.

We identified SMs as a dominating lipid class associated with processed meat intake, and most of the SMs were further associated with incident T2D or glucose/insulin homeostasis metrics. However, research on the associations between SMs and processed meat intake is limited. SMs, a type of sphingolipid present in cell membranes, play key roles in cell signaling and lipid raft formation. SM metabolism is known to be influenced by oxidative stress and inflammation (33). Compounds in processed meats, such as advanced glycation end products and nitrosamines formed from nitrates/nitrites content, can induce oxidative stress and inflammation (34,35), thus affecting SM metabolism. Previous studies in the SHFS have showed that SMs are a major lipid class associated with temporal changes in fasting glucose (29). Notably, we found that SM(d30:1) A, SM(d30:1) B, SM(d32:2) A, and SM(d32:2) B not only correlated with lower processed meat intake and lower glucose levels but also showed significant mediation effects on the positive association between processed meat intake and glucose levels. Further investigation is required to determine which compounds in processed meat intake may explain these effects.

Some individual lipidomic markers of meat intake were supported by previous research. For instance, CE(18:0), a red meat-related lipid in our analysis, was also correlated with total red meat (including both processed and unprocessed) and processed red meat intake in another study (14). Interestingly, it was also linked to a lower Mediterranean Diet Adherence Screener score (36) and a lower Alternate Healthy Eating Index (22), both of which reflect adherence to healthy dietary patterns. In addition, our results indicate that CE(22:5) A is associated with higher processed meat intake. This finding supports our previous work, which showed a relationship between CE(22:5) and low diet quality as well as higher total red and processed meat intake among SHFS participants (37). However, this finding contrasts with results from two studies conducted in non-Hispanic White populations in the U.S. and older adults in Spain, which reported associations of CE(22:5) with olive oil intake (36) and higher diet quality (22,36). These conflicting results may be due to the underlying differences in the study populations (e.g., variations in dietary patterns, food environments across different racial/ethnic groups and geographic regions, or differences in underlying health status) or in dietary assessment methods. For instance, the clinical trial in Spain (36) assessed diet quality and olive oil intake using a brief 14-item diet screener with dichotomous questions on select dietary variables, whereas our study used a 119-item Food Frequency Questionnaire to quantify usual dietary intake over the past year. Taken together, these results suggest that CE(18:0) and CE(22:5) are important lipid metabolites closely linked to diet. Meanwhile, our study identified many lipids that have not been previously reported. Several other studies have also explored metabolites or metabolomic signatures associated with meat consumption (9,12,13,15). While direct comparisons of metabolites across studies remain challenging due to differences in analytical platforms, the collective evidence consistently suggests that red and processed meat intake is associated with altered lipid metabolism in various populations.

Previous studies measuring traditional lipid panels have associated elevated triacylglycerol (TAG) level with red meat or processed meat intake (38,39). Interestingly, several TAGs in our study showed a nominal association with red meat intake but not processed meat; however, none remained significant after multiple testing corrections. This observation aligns with another lipidomics-based study, which also found no significant association between TAG species and red or processed meat intake (23). Further research is needed to explore the specific functions and roles of small-molecule lipids in dietary risk, as they are likely to differ from those measured in clinical lipid panels.

A significant strength of our study lies in the comprehensive coverage of the lipidome. We quantified >1,500 distinct lipid species across 14 lipid classes at two time points, enabling a hypothesis-free discovery of lipids associated with red meat and processed meat intake. Furthermore, our study examined multiple outcomes related to diabetes risk, including incident T2D and glucose/insulin homeostasis metrics. In addition, we conducted a thorough evaluation of sociodemographic, lifestyle, and clinical factors, allowing for rigorous adjustment of potential confounders in our analyses.

However, some limitations should be noted. First, some of the detected lipids are unknown compounds, requiring further experiments (e.g., tandem MS/MS) to characterize their chemical structure and biological function in future research. Second, meat intake data were only available at baseline, limiting our ability to assess changes in meat consumption over time. Nonetheless, SHFS participants are known to have relatively consistent dietary intake over time (40). Third, our follow-up period is relatively short, resulting in a smaller number of incident T2D cases. Fourth, lipidomic markers were identified through cross-sectional analyses, and the findings should be interpreted with caution due to potential unknown confounders. Lastly, as our study is observational in nature, it cannot establish causal relationships.

In summary, we identified multiple lipid species as potential markers of unprocessed red meat and processed meat intake among American Indians. Meat intake-related lipids were independently associated with diabetes risk beyond traditional risk factors, potentially through mediating the positive associations of red meat and processed meat intake with fasting glucose. Future studies in diverse populations are warranted to validate our findings.

This article contains supplementary material online at https://doi.org/10.2337/figshare.28696799.

Funding Statement

This study was supported by the National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases grant R01DK107532 (J.Z.). The Strong Heart Study has been funded in whole or in part with federal funds from the National Heart, Lung, and Blood Institute, National Institutes of Health, Department of Health and Human Services, under contract numbers 75N92019D00027, 75N92019D00028, 75N92019D00029, and 75N92019D00030. The study was previously supported by National Heart, Lung, and Blood Institute research grants R01HL109315, R01HL109301, R01HL109284, R01HL109282, and R01HL109319 and by cooperative agreements U01HL41642, U01HL41652, U01HL41654, U01HL65520, and U01HL65521.