Cross-sectional study of conjugated linoleic acid in adipose tissue and risk of diabetes

Abstract

Background: Some experimental studies on conjugated linoleic acid (CLA) and insulin regulation suggested that CLA could be associated with risk of diabetes, but epidemiologic studies are lacking.

Objective: The aim of the study was to test whether the amount of CLA in adipose tissue is associated with risk of diabetes.

Design: A cross-sectional design was used to test the study hypothesis in 232 adults with diabetes and 1512 adults without diabetes who lived in Costa Rica. The cis-9, trans-11 and trans-10, cis-12 CLA isomers in adipose tissue and 48 other fatty acids were assessed by using gas chromatography. Prevalence ratios (PRs) and 95% CIs were estimated by using Poisson regression adjusted for potential confounders.

Results: The mean (±SD) percentage of total fatty acids of CLA for the cis-9, trans-11 isomer in adipose tissue was 0.57 ± 0.18% in adults without diabetes and 0.53 ± 0.17% in adults with diabetes (P = 0.0078). The trans-10, cis-12 CLA isomer was not detected in adipose tissue. The cis-9, trans-11 CLA isomer was associated with a lower risk of diabetes. In comparison with the first quintile, the PR (95% CI) for the fifth quintile was 0.48 (0.31, 0.76) (P-trend = 0.0005) in the basic and 0.46 (0.29, 0.72) (P-trend = 0.0002) in the multivariable model. Additional adjustment for other fatty acids in adipose tissue including trans-9 16:1, which is a fatty acid that was previously associated with diabetes, did not modify the results.

Conclusion: The observed inverse association between the cis-9, trans-11 CLA in adipose tissue and diabetes risk is consistent with the hypothesis that CLA may be involved in insulin regulation.

INTRODUCTION

Conjugated linoleic acid (CLA) represents a group of positional and geometric isomers of linoleic acid (1). CLA is produced in the rumen of ruminant animals by fermentative bacteria (2) or by synthesis via Δ 9-desaturase (3). CLA is in dairy products such as milk and cheese as well as ruminant meats such as beef and lamb (3). The most common CLA isomers in ruminant meats and milk are cis-9, trans-11 and trans-10, cis-12 (2). These isomers represent 80% and 10%, respectively, of all CLA isomers in the human diet (4).

Animal models suggested that CLA could have beneficial effects on diabetes (5, 6), but these effects could be isomer specific (7). Results from intervention studies in humans have yielded mixed results (8, 9), with adverse rather than beneficial effects on insulin resistance shown in some studies (10, 11). For example, a double-blinded, randomized trial in 25 obese men that measured insulin sensitivity directly by using a euglycemic clamp showed that CLA decreased insulin sensitivity by 15% (12). Another randomized, placebo-controlled trial in 32 subjects with type 2 diabetes showed that CLA increased fasting glucose and reduced postprandial insulin sensitivity (8). The interpretation of these studies is difficult because of the variety of studied subjects, lengths, doses, and types of isomers provided. For example, the typical dose of ∼3 g CLA commonly used in intervention studies is 10 times higher than the average reported dietary intake (range: ∼0.11–0.29 g/d) (13–15). Furthermore, it is possible that cis-9, trans-11 and trans-10, cis-12 have divergent health effects (16).

To our knowledge, there are no published observational studies that evaluated the association between CLA and diabetes because CLA is not generally present in nutrient databases, and most studies that assessed the fatty acid profile in human biospecimens have not included CLA. The goal of this study was to test whether CLA measured in adipose tissue is associated with risk of diabetes. Because the amount of CLA in humans is derived from ruminant fat, the amount of CLA in adipose tissue can provide an objective marker of CLA intake. A linear significant correlation has been reported between adipose tissue CLA, dietary CLA, and dairy intake, but no association has been detected for beef intake (13). Studies showed that the turnover of fatty acids in adipose tissue is >2 y; therefore, adipose tissue fatty acid concentrations represent a long-term integrated measure of exposure (17, 18). The association between CLA and plasma triglyceride and fasting glucose was also evaluated in control subjects.

SUBJECTS AND METHODS

Study population

The study population consisted of subjects who participated as control subjects in a population-based study of diet and heart disease in Costa Rica. In the 2119 control subjects in the heart-disease study, we selected 1744 subjects (82%) who had complete values for cis-9, trans-11 CLA in adipose tissue and information on diabetes and potential cofounders (19). A total of 232 subjects were classified as adults with diabetes according to the WHO diagnostic criteria of having a fasting whole blood glucose concentration ≥6.1 mmol/L and/or taking medication to control diabetic symptoms (20), and the other 1512 subjects were classified as persons without diabetes. The catchment area for this study consisted of 34 counties in the Central Valley of Costa Rica. Control subjects were randomly recruited from the free-living population to match cases of myocardial infarction by age (±5 y), sex, and area of residence. Control subjects were not recruited if they were unable to answer the questionnaire or if they had suffered a myocardial infarction in the past. Participation was 88% in control subjects. Participants signed a consent form approved by the institutional review boards at the Harvard School of Public Health and the University of Costa Rica.

Data collection

For the current study, we used already-collected data and laboratory results from the Costa Rica Heart Study (19). Data on demographic and social economic characteristics, smoking, health status, and physical activity, as well as anthropometric measurements and blood pressure, were collected at the homes of subjects. A 12–14-h fasting blood sample and an adipose tissue aspirate were collected in the morning. Adipose tissue samples were collected from the buttock with a 1.5 × 16-inch gauge needle and a 10-cc syringe by using the method of Beynen and Katan (21). The sample was left in the syringe, stored in a sealed plastic bag, and immediately immersed into ice and transported within 3 h to the field station. Samples were diluted in 1.5 cc hexane:isopropanol (3:2), sealed, and stored at –80°C until they were analyzed for fatty acid content at the Harvard School of Public Health. No changes in the fatty acid composition of adipose tissue samples taken from the buttock have been observed in samples stored for 1.5 y (21).

Fatty acids were extracted from adipose with hexane:isopropanol (3:2) and esterified as previously described by Lillington et al (22). The fatty acid methyl esters were separated on a 100 m Supelco 2330 column and Hewlett Packard 6980 gas chromatographer. The temperature program ranged from 90°C to 250°C. We identified, through qualitative analysis of the retention peak times of pure standards (NuCheck Prep), 50 fatty acids including cis-9, trans-11 CLA (23). Fatty acids were expressed as a percentage of the total fatty acid identified. The between-sample CV was 5.2% for cis-9, trans-11 CLA. The trans-10, cis-12 isomer of CLA was not detected in adipose tissue, and consequently, analyses included only the cis-9, trans-11 CLA isomer.

Plasma triglycerides were measured by using enzymatic reagents (Boehringer-Mannheim Diagnostics), and capillary whole blood glucose was measured by using an Accu-Check II Blood Glucose Monitor with Chemstrip bG Test Strips (Boehringer-Mannheim Diagnostics) immediately after sample collection.

All statistical analyses were performed with Statistical Analysis Systems software (version 9.1; SAS Institute Inc). Differences in distributions of descriptive characteristics between subjects with and without diabetes were tested by using the t test if normally distributed or Wilcoxon's signed-rank test if not normally distributed for continuous variables, and the chi-square test was used for categorical variables. Poisson regression with robust variance was used to estimate prevalence ratios (PRs) and 95% CIs of diabetes according to quintiles of fatty acids in adipose tissue (24). Tests for trends were performed by using quintiles as a continuous variable in linear regression models. Potential confounders of age (continuous), sex, area of residence, physical activity (measured in metabolic equivalent tasks), BMI, waist circumference, income, and current smoking status (yes or no) were determined by evaluating their distribution in subjects without diabetes. The following types of fatty acids were measured in adipose tissue and used as potential confounders categorized into quintiles in prevalence models: trans-9 16:1, total trans (18:1 trans, 18:2 trans), even-numbered saturated (14:0 + 16:0 + 18:0), odd-numbered saturated (15:0 + 17:0), n−3 polyunsaturated (18:3n−3, EPA, DHA, and docosapentaenoic acid), n−6 polyunsaturated (18:2n−6, 20:2n−6, and 20:4n−6), and monounsaturated (16:1n−7 and 18:1n−9).

PRs and 95% CIs between diabetes and quintiles of ruminant trans-9:16:1 fatty acid and SFAs (15:0 + 17:0) were also calculated because of their association with diabetes and ruminant fat in previous studies (25). These models were adjusted for all of the previously mentioned confounders. The associations between fasting glucose, plasma triglyceride concentrations, and CLA in adipose tissue were also examined by using Spearman's partial correlation coefficients adjusted for age, sex, BMI, physical activity, and smoking.

RESULTS

General characteristics of the study population are shown in Table 1. The prevalence of diabetes in the study population was 13%. Compared with adults without diabetes, adults with diabetes were less physically active and had a higher BMI and larger waist circumferences. Also, there were fewer smokers in adults with diabetes than in adults without diabetes. Adults with diabetes presented higher plasma triglyceride concentrations and lower LDL cholesterol. Adipose tissue amounts of cis-9, trans-11 CLA were significantly lower in adults with diabetes than in adults without diabetes. Our analyses included only the cis-9, trans-11 CLA isomer because the trans-10, cis-12 isomer was not detected in adipose tissue.

TABLE 1.

General characteristic of adults (n = 1744) with and without diabetes from the Central Valley of Costa Rica¹

Variables	Without diabetes (n = 1512)	With diabetes (n = 232)
Age (y)	57 ± 112	63 ± 9
Women (%)	24	33
Area of residence (percentage rural)	24	27
Monthly household income ($)	587 ± 432	549 ± 397
Current smokers (%)	22	15
BMI (kg/m2)	26.2 ± 4.1	280.1 ± 4.5
Waist circumference (cm)	90.4 ± 9.9	95.2 ± 9.4
Physical activity (METs)	36 ± 15	33 ± 12
Blood glucose (mmol/L)	4.1 ± 0.7	7.9 ± 2.9
Total plasma triglycerides (mmol/L)	2.4 ± 1.4	3.0 ± 1.9
HDL cholesterol (mmol/L)	1.1 ± 0.2	1.1 ± 0.3
LDL cholesterol (mmol/L)	3.2 ± 1.0	2.9 ± 1.2
Adipose tissue CLA (percentage of total fatty acids)	0.57 ± 0.18	0.53 ± 0.17

¹t Test and Wilcoxon's signed rank test for continuous variables and the chi-square test for categorical variables were used to determine the significance of differences. All variables, except area of residence, income, and HDL cholesterol, were significantly different between adults with and without diabetes (P < 0.05). CLA, conjugated linoleic acid; METs, metabolic equivalent tasks.

²Mean ± SD (all such values).

The distribution of potential confounders by quintiles of cis-9, trans–11 CLA in adipose tissue is shown in Table 2. The median for the lowest quintile of cis-9, trans-11 CLA was 0.34% of total fatty acids, whereas the highest quintile was 0.83% of total fatty acids. Quintiles of cis-9, trans-11 CLA were positively correlated with age, female sex, and trans fatty acid in adipose tissue and inversely correlated with rural area of residence, physical activity, smoking, and SFAs, MUFAs, and n−3 and n−6 PUFAs in adipose tissue. A significant inverse trend was shown for fasting glucose and plasma triglyceride concentrations.

TABLE 2.

General characteristics and potential confounders by quintiles of cis-9, trans-11 CLA in adipose tissue in adults without diabetes (n = 1512)¹

	Quintiles of CLA in adipose tissue
General characteristics	1	2	3	4	5	P-trend2
CLA (percentage of total adipose fatty acids)3	0.34	0.47	0.55	0.64	0.83
Age (y)	55	56	57	58	59	<0.0001
Women (%)	12	18	27	32	37	<0.0001
Area of residence (percentage rural)	28	28	25	23	18	<0.0001
Monthly household income ($)	537	609	635	553	604	0.3
BMI (kg/m2)	25.6	26.4	26.1	26.3	26.4	0.1
Waist circumference (cm)	89.8	91.5	90.1	90.2	90.4	0.9
Physical activity (METs)	39	37	36	34	33	<0.0001
Current smokers (%)	30	23	19	20	21	0.02
Blood glucose (mmol/L)	4.2	4.1	4.2	4.1	4.1	0.0004
Total triglycerides (mmol/L)	2.5	2.5	2.5	2.2	2.2	0.0002
HDL cholesterol (mmol/L)	1.0	1.0	1.1	1.0	1.1	0.07
LDL cholesterol (mmol/L)	3.1	3.2	3.1	3.4	3.2	0.4
Adipose tissue fatty acids (percentage of total fatty acids)
trans Fatty acids	2.8	2.9	3.2	3.3	3.6	<0.0001
trans-9 16:1	0.13	0.14	0.15	0.14	0.15	<0.0001
18:1 trans	1.79	1.85	2.00	2.08	2.28	<0.0001
18:2 trans	1.00	1.05	1.15	1.20	1.37	<0.0001
SFAs (total)	27.3	25.9	25.2	24.6	23.5	<0.0001
Even chain (14:0 + 16:0 + 18:0)	27.3	25.9	25.2	24.5	23.5	<0.0001
Odd chain (15:0 + 17:0)	0.40	0.40	0.41	0.41	0.43	<0.0001
n−3 PUFAs	1.05	1.04	1.02	1.00	1.00	0.04
n−6 PUFAs	17.4	16.8	16.6	15.8	15.2	<0.0001
MUFAs	46.6	48.1	48.6	49.7	50.4	<0.0001

¹All values are medians unless otherwise indicated. CLA, conjugated linoleic acid; METs, metabolic equivalent tasks.

²Tests for trends were performed by using quintiles as a continuous variable in linear regression models.

³All values are means.

PRs and 95% CIs for the association between quintiles of cis-9, trans-11 CLA in adipose tissue and the prevalence of diabetes are shown in Table 3. Higher amounts of cis-9, trans-11 CLA in adipose tissue were significantly associated with a 52% lower prevalence of diabetes in the basic model [PR (95% CI) = 0.48 (0.31, 0.76) for fifth compared with first quintile]. Adjustment for physical activity, smoking status, or individual fatty acids in adipose tissue did not materially change the main result. We also tested whether other fatty acid markers of ruminant products trans-9 16:1 fatty acid and SFAs (15:0 + 17:0) were associated with diabetes risk. We showed no association for the trans-9 16:1 fatty acid and a suggestive, although not significant, association between SFAs (15:0 + 17:0) and higher diabetes risk.

TABLE 3.

Risk of diabetes by quintiles of fatty acids in adipose tissue (n = 1744)

	Quintiles of adipose tissue
	1	2	3	4	5	P-trend1
cis-9, trans-11 CLA2
Percentage of total fatty acids3	0.34 (≤0.42)	0.47 (>0.42 to ≤0.51)	0.55 (>0.51 to ≤0.59)	0.64 (>0.59 to ≤0.70)	0.83 (>0.70 to ≤1.24)	—
Basic model4	1 (reference)5	0.93 (0.64, 1.34)	0.74 (0.50, 1.10)	0.67 (0.45, 1.01)	0.48 (0.31, 0.76)	0.0005
Model 16	1 (reference)	0.86 (0.59, 1.25)	0.72 (0.48, 1.07)	0.62 (0.42, 0.94)	0.46 (0.29, 0.72)	0.0002
Model 278	1 (reference)	0.85 (0.58, 1.25)	0.69 (0.46, 1.03)	0.59 (0.39, 0.90)	0.42 (0.26, 0.68)	0.0001
trans-9 16:1
Percentage of total fatty acids3	0.071 (≤0.094)	0.11 (>0.094 to ≤0.13)	0.15 (>0.13 to ≤0.16)	0.17 (>0.16 to ≤0.19)	0.21 (>0.19 to ≤0.45)	—
Basic model4	1 (reference)	0.99 (0.63, 1.56)	1.10 (0.70, 1.71)	1.17 (0.77, 1.79)	1.24 (0.82, 1.88)	0.2
Model 16	1 (reference)	0.94 (0.60, 1.48)	1.02 (0.65, 1.59)	1.09 (0.71, 1.67)	1.15 (0.76, 1.75)	0.4
Model 289	1 (reference)	0.84 (0.53, 1.34)	0.88 (0.56, 1.39)	0.95 (0.61, 1.48)	1.05 (0.66, 1.66)	0.3
15:0 + 17:0
Percentage of total fatty acids3	0.30 (≤0.33)	0.36 (>0.33 to ≤0.38)	0.40 (>0.38 to ≤0.42)	0.45 (>0.42 to ≤0.49)	0.54 (>0.49 to ≤1.23)	—
Basic model4	1 (reference)	0.78 (0.49, 1.24)	0.86 (0.55, 1.35)	1.24 (0.83, 1.85)	1.21 (0.81, 1.81)	0.07
Model 16	1 (reference)	0.80 (0.51, 1.28)	0.88 (0.58, 1.39)	1.35 (0.90, 2.03)	1.33 (0.88, 2.01)	0.02
Model 2810	1 (reference)	0.82 (0.51, 1.31)	0.92 (0.58, 1.46)	1.30 (0.85, 2.01)	1.33 (0.84, 2.11)	0.06

¹Tests for trends were performed by using quintiles as a continuous variable in linear regression models.

²CLA, conjugated linoleic acid.

³All values are medians; ranges in parentheses.

⁴Adjusted for age, sex, and area of residence.

⁵Prevalence; 95% CI in parentheses (all such values). Values were estimated by using Poisson regressions with robust variances.

⁶Adjusted as for basic model and for current smokers, BMI, and physical activity.

⁷Adjusted as for model 1 and for trans-9 16:1 and 15:0 + 17:0 fatty acids in adipose tissue.

⁸The addition of waist circumference, income, MUFAs, trans fatty acids, n−3 and n−6 PUFAs, and even-number SFAs did not modify results.

⁹Adjusted as for model 1 and for cis-9, trans-11 CLA and 15:0 + 17:0 fatty acids in adipose tissue.

¹⁰Adjusted as for model 1 and for trans-9 16:1 and cis-9, trans-11 fatty acids in adipose tissue.

A significant although weak inverse correlation was shown between concentrations of fasting glucose and adipose tissue cis-9, trans-11 CLA in adults without diabetes (r = −0.072, P = 0.006; n = 1512). In the same group, the concentration of plasma triglycerides was also inversely associated with higher adipose tissue cis-9, trans-11 CLA (r = −0.083, P = 0.001; n = 1512). These correlations were adjusted for age, sex, BMI, smoking, and physical activity. Our results were robust to outlier analysis.

DISCUSSION

We assessed amounts of cis-9, trans-11 CLA and trans-10, cis-12 CLA in adipose tissue in adults with and without diabetes. A significant inverse association was shown between the cis-9, trans-11 CLA isomer measured in adipose tissue and the prevalence of diabetes (a 42% lower prevalence of diabetes). The trans-10, cis-12 CLA isomer was not detected in adipose tissue in this population. This isomer was not detected in plasma or whole blood in the same studied population (26).

Short-term clinical trials on the role of CLA in insulin-related traits have yielded inconclusive results, and to our knowledge, there are no data from epidemiologic studies. Some studies showed detrimental effects of either cis-9, trans-11 CLA or trans-10, cis-12 CLA (12, 27) on insulin sensitivity. One study showed no effect of cis-9, trans-11 CLA on insulin sensitivity in healthy, middle-aged men after 6 wk (28). Furthermore, interventions that provided supplements that contained a 50%:50% combination of both CLA isomers (trans-10, cis-12 and cis-9, trans-11 CLA) have, in general, produced mixed results on insulin sensitivity (8, 29, 30). It is difficult to reconcile our findings with those of these intervention studies given the major differences in the way CLA intake was assessed, doses and types of isomers evaluated, and study designs. Our study was designed to evaluate an integrated long-term exposure (∼2 y) to dietary CLA by measuring its amount in adipose tissue (17, 18). Intervention studies showed that cis-9, trans-11 CLA was not incorporated into adipose tissue after 3 mo (31, 32), even with CLA supplementation that was 12-fold higher (3.9 g/d) than the dietary intake previously reported for the Costa Rican population (0.29 g/d) (13). CLA did not affect insulin sensitivity in the study in which this measure was evaluated (31, 32). These data suggested that longer exposure periods may be needed to fully assess the potential effects of CLA. Epidemiologic studies that would evaluate dietary sources of CLA are also needed in other populations.

We showed that trans-9 16:1 fatty acid and SFAs (15:0 + 17:0) fatty acids in adipose tissue, which are 2 known biomarkers of dairy intake (23, 33), were not associated with the prevalence of diabetes. Mozaffarian et al (25) recently reported that the trans-9 16:1 measured in plasma phospholipids was both positively correlated with whole-fat dairy intake and inversely correlated with risk of diabetes in the prospective Cardiovascular Health Study. Whether the results of Mozaffarian et al (25) are confounded by cis-9, trans-11 CLA is uncertain because the study did not assess CLA. Nevertheless, our results and those of Mozaffarian et al (25) indicated that fatty acids shown in dairy products may have beneficial effects on risk of diabetes and supported previous studies that showed an inverse association between dairy intake and diabetes risk (34–36).

Available evidence suggested some potential mechanisms for the observed association between adipose tissue cis-9, trans-11 CLA and diabetes. A recent experimental study suggested that CLA could increase glucose-stimulated insulin secretion by activating the free fatty acid receptor 1, which is an emerging therapeutic target to treat type 2 diabetes (37). The same study also proposed a role for CLA as an activator of insulin production and activity. Other studies have shown that CLA is able to increase the expression of the PPARG gene, which is a key regulator of adipogenesis and insulin metabolism (38), in prediabetic rats (39) and human preadipocytes (40). However, other studies have shown no effect of cis-9, trans-11 CLA on PPARG gene expression in obese mice (41), 3T3-L1 murine preadipocytes (42), or human adipose tissue (43). Other potential mechanisms may involve antiinflammatory effects (5, 44) and changes in the phospholipid membrane of skeletal muscle and adipocytes (31, 42). We showed a significant inverse correlation between adipose cis-9, trans-11 CLA and whole blood fasting glucose and plasma triglycerides, which suggested that metabolic mechanisms that regulate glucose and triglyceride homeostasis may be, at least in part, related to the content of cis-9, trans-11 CLA in adipose tissue. Whether the observed association between CLA and a reduced risk of diabetes is mediated through changes in weight is unclear because intervention studies have failed to show a consistent effect of CLA supplementation on weight and BMI reduction (45, 46). We did not find an association between adipose tissue CLA and BMI, and the estimated PRs were not modified when they were adjusted for BMI.

The major limitation of the current study was its cross-sectional design, and therefore, no causal relationships or temporality could be established. Because adults with and without diabetes were derived from the control group from the original myocardial infarction study, it remains to be assessed whether our studies can be generalized to other populations. Also, residual confounders for different factors could not be completely ruled out, and the use of self-reported diabetes and fasting glucose in whole blood to identify diabetic cases may have produced some degree of unidentified diabetic subjects. However, we used the cutoff of a fasting whole blood glucose concentration of 6.1 mmol/L for a diagnosis of diabetes as proposed by the WHO for whole blood in epidemiologic studies (20), and the use of capillary whole blood glucose is a reliable and cost-effective tool for use in epidemiologic studies (47). To our knowledge, this is the first observational study to report an association between adipose tissue cis-9, trans-11 CLA and the prevalence of diabetes; thus, our findings need to be replicated in independent studies.

A major strength of the current study was the availability of unique measurements of fatty acids in adipose tissue that reflected long-term intakes. One study showed that adipose tissue fatty acids had higher correlations with their corresponding dietary fatty acids than with whole blood or plasma fatty acids (26). Furthermore, the availability of measurements of other fatty acids in adipose tissue allowed us to control for potential confounding because of those different fatty acids. For example, in our population, the content of cis-9, trans-11 CLA was inversely correlated with even-numbered saturated fat and positively correlated with trans fat and odd-numbered saturated fat in adipose tissue. The failure to adjust for other fatty acids could lead to potential confounding by their effects.

In conclusion, higher amounts of the cis-9, trans-11 CLA isomer in adipose tissue were inversely correlated to the prevalence of diabetes, fasting plasma triglycerides, and fasting blood glucose concentrations. This study supports the hypothesis that CLA may play a role in the development of diabetes. Future studies are needed to confirm our findings as well as to elucidate potential biologic mechanisms that mediate the observed associations.