Erythrocyte Saturated Fatty Acids and Incident Type 2 Diabetes in Chinese Men and Women: A Prospective Cohort Study

The association between circulating saturated fatty acids (SFAs) and incident type 2 diabetes (T2D) is reported in Western populations with inconsistent results, while evidence from Asian populations is scarce. We aimed to examine the associations between erythrocyte SFAs and incident T2D in a Chinese population. Between 2008 and 2013, a total of 2683 participants, aged 40–75 years, free of diabetes were included in the present analyses. Incident T2D cases were ascertained during follow-up visits. Gas chromatography was used to measure erythrocyte fatty acids at baseline. The Cox proportional hazards model was used to estimate the hazard ratios (HRs) and 95% confidence intervals (CIs). During 13,508 person years of follow-up, 216 T2D cases were identified. Compared with the first quartile, multivariable-adjusted HRs (95% CIs) of the fourth quartile were 1.20 (0.82–1.76; p = 0.242) for myristic acid (14-carbon tail, zero double bonds; 14:0), 0.69 (0.48–0.99; p = 0.080) for palmitic acid (16:0), 1.49 (1.02–2.19; p = 0.047) for stearic acid (18:0), 1.46 (1.00–2.12; p = 0.035) for arachidic acid (20:0), 1.48 (0.99–2.22; p = 0.061) for behenic acid (22:0), and 1.08 (0.74–1.56; p = 0.913) for lignoceric acid (24:0). Our findings indicate that individual erythrocyte SFAs are associated with T2D in different directions, with 18:0 and 20:0 SFAs positively associated with the risk, whereas no convincing inverse association for 16:0 SFAs.


Introduction
Diabetes was a global health problem in the past decade, with an estimated 451 million adults with the disease in 2017 globally [1]. Diet is one of the most important modifiable lifestyle factors for the prevention of type 2 diabetes (T2D), the predominant type of diabetes. The role of dietary saturated fatty acids (SFAs) in the prevention of T2D is still not clear and under debate [2,3]. Circulating SFAs reflect both dietary intake and endogenous synthesis [4], and investigation of their associations with T2D may provide new insight into the role of SFAs in diabetes etiology.
Of note, all the above studies were exclusively conducted in Western populations, while there was one small nested case-control study among a Japanese population (n = 1014), which reported a null association of serum 14:0, 16:0, 18:0, or 20:0 SFAs with incident T2D [15].
The aim of the present study was to examine the associations of individual circulating SFAs with incident T2D in a community-based prospective cohort in southern China. As a secondary objective, we aimed to examine the associations of different groups of circulating SFAs with incident T2D, given the strong evidence from the EPIC-InterAct study about the diverse associations of different SFA groups [5].

Study Population
This study was based on the Guangzhou Nutrition and Health Study (GNHS), a community-based prospective cohort study designed to investigate the nutritional determinants of chronic diseases such as cardiovascular diseases, diabetes, and bone health. Between 2008 and 2013, participants were recruited via advertisements, health talks, and referrals in urban Guangzhou, China. Volunteers who lived in Guangzhou for more than five years, aged 40-75 years, were eligible for inclusion. GNHS included two batches of participant recruitment following the same criteria between 2008 and 2010 (n = 3169), and between 2012 and 2013 (n = 879), with a total of 4048 participants. All participants were followed up approximately every three years, and up to 31 May 2017, two follow-up visits were conducted for participants recruited between 2008 and 2010, and one follow-up visit for participants recruited between 2012 and 2013.
Finally, 2683 participants were included in the present analyses, with a median 5.6 years of follow-up. The flowchart is presented in Figure 1. The study protocol of GNHS was registered (NCT03179657, ClinicalTrials.gov) and approved by the Ethics Committee of the School of Public Health at Sun Yat-sen University, and all participants provided written informed consent.

Data Collection
Face-to-face interviews were conducted in each survey by trained investigators. Information on socio-demographic characteristics (e.g., age, sex, household income, and education level), lifestyle (e.g., smoking, alcohol, and tea drinking), physical activity [16], and history of chronic diseases and medications was collected. A validated food frequency questionnaire was used to estimate the habitual dietary intakes over the past year [17]. Weight and height were measured with the participants wearing light clothing but shoes-off. Participants also provided venous blood after an overnight fasting. Serum, plasma, and erythrocytes were separated and stored at −80 °C until used for analysis.

Ascertainment of Incident T2D
Incident cases of T2D were identified as fasting glucose ≥7.0 mmol/L or glycated hemoglobin ≥6.5% (n = 157), or as self-reported diabetic medications (n = 59) during the follow-up visits according to the American Diabetes Association criteria for diagnosis and classification of diabetes [18], and incident T2D was ascertained up until 31 May 2017.

Measurement of Biochemical Parameters
Serum triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and fasting glucose levels were determined with colorimetric methods using a Roche cobas 8000 c702 automated analyzer (Roche Diagnostics GmbH, Shanghai, China). The intra-

Data Collection
Face-to-face interviews were conducted in each survey by trained investigators. Information on socio-demographic characteristics (e.g., age, sex, household income, and education level), lifestyle (e.g., smoking, alcohol, and tea drinking), physical activity [16], and history of chronic diseases and medications was collected. A validated food frequency questionnaire was used to estimate the habitual dietary intakes over the past year [17]. Weight and height were measured with the participants wearing light clothing but shoes-off. Participants also provided venous blood after an overnight fasting. Serum, plasma, and erythrocytes were separated and stored at −80 • C until used for analysis.

Ascertainment of Incident T2D
Incident cases of T2D were identified as fasting glucose ≥7.0 mmol/L or glycated hemoglobin ≥6.5% (n = 157), or as self-reported diabetic medications (n = 59) during the follow-up visits according to the American Diabetes Association criteria for diagnosis and classification of diabetes [18], and incident T2D was ascertained up until 31 May 2017.

Measurement of Biochemical Parameters
Serum triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and fasting glucose levels were determined with colorimetric methods using a Roche cobas 8000 c702 automated analyzer (Roche Diagnostics GmbH, Shanghai, China). The intra-assay coefficients of variation were 5.8% for TG, 4.3% for HDL-C, 3.1% for LDL-C, and 2.5% for fasting glucose levels. Glycated hemoglobin was measured by high-performance liquid chromatography using the Bole D-10 Hemoglobin A1c Program on a Bole D-10 Hemoglobin Testing System, and the intra-assay coefficient of variation was 0.75% for glycated hemoglobin.

Statistical Analysis
We examined the difference in baseline characteristics by future T2D cases and non-cases using the t-test or Mann-Whitney test for continuous variables, and the chi-square test for categorical variables. Spearman's rank correlation coefficients were calculated between individual erythrocyte SFAs, and between food groups and erythrocyte SFAs.
The Cox proportional hazards model was used to estimate the hazard ratios (HRs) and 95% confidence intervals (CIs) of incident T2D across the second through fourth quartiles (compared with the first quartile) of erythrocyte SFAs under three statistical models to minimized the possibility of reverse causality. Model 1, adjusted for baseline age, sex, body mass index (BMI), and ratio of waist to hip circumference; Model 2, model 1 plus smoking status, alcohol drinking, tea drinking, education level, household income, physical activity, family history of diabetes, and total energy intake; Model 3, model 2 plus blood lipid and glycemic markers (TG, HDL-C, LDL-C, and fasting glucose).
We also examined the interaction of erythrocyte SFAs with several pre-defined variables: age (continuous), sex, and BMI (continuous), by including interaction terms in the above final model (i.e., model 3), and we presented the stratified analyses if a significant interaction (p < 0.05) was found.
All statistical analyses were conducted by using IBM SPSS software version 23.0 (SPSS, IBM, New York, NY, USA). A two-tailed p-value <0.05 was considered as statistically significant. Table 1 presents the baseline characteristic by future T2D cases and non-cases. At baseline, future cases aged 59.0 (5.7) years were older than non-cases aged 57.9 (5.7) years. Future cases had higher baseline BMI, ratio of waist to hip circumference, serum fasting glucose, and TG levels, and lower household income, dairy intake, and HDL-C levels, and were more likely to have a family history of diabetes compared with non-cases. Table 2 suggested that erythrocyte SFAs were significantly correlated with each other (p < 0.05), and 14:0, 16:0, and 18:0 SFAs were inversely correlated with 22:0 and 24:0 SFAs. The Spearman's rank correlation coefficients between 16:0 and 22:0 SFAs was −0.298. The correlations of erythrocyte SFAs with dietary intakes are presented in Table S2. During 13,508 person years of follow-up, 216 cases of incident T2D were identified. There were no associations among erythrocyte 14:0 and 24:0 SFAs and T2D incidence across the three models. After adjustment for baseline sociodemographic and lifestyle factors (model 1 and model 2), no association of 16:0 SFAs with T2D was observed. While after further adjustment for circulating lipid and glycemic markers in model 3, an inverse association for 16:0 SFAs was found with a HR of 0.69 (95% CIs: 0.48-0.99; p = 0.080) at the fourth quartile compared with the first quartile. In the multivariable-adjusted model (i.e., model 3), higher 18:0 and 20:0 SFA levels were associated with a higher risk of T2D (p < 0.05  (Table 3). There was no substantial difference in various sensitivity analyses (Table S4). Table 2. Spearman's rank correlation of erythrocyte saturated fatty acids (n = 2683) a .  We did not find any significant association of groupings of erythrocyte total SFAs, even-chain SFAs and VLCSFAs with incident T2D (Figure 2, Table S3).

Discussion
In this prospective study of Chinese men and women, we investigated the associations of six individual and several groupings of erythrocyte SFAs with T2D incidence. We found that higher Circulating 16:0 SFAs are the most abundant SFAs among total FA composition [5,20]. In the past decade, a positive association of circulating 16:0 SFAs with T2D was reported in several prospective cohort studies, including the EPIC-InterAct Study [5], the Atherosclerosis Risk in Communities Study [7], and the Cardiovascular Health Study [21]. Null association of circulating 16:0 SFAs with T2D was reported in several other prospective cohorts [10,12,15]. In contrast, results from 1346 Finnish men suggested a positive association of erythrocyte 16:0 SFAs with insulin sensitive index, but no association with T2D [11]. The inconsistent results across these above studies may be because of the diverse backgrounds of ethnicities and dietary habits in these study participants. Indeed, there are only two prospective studies among Asian populations reporting the 16:0-T2D association, including our present one and a Japanese cohort [15], and neither found a positive association. The discrepancy between the prior evidence and our present results may be because of the difference in dietary habits and dietary fat compositions. The majority of prior studies came from Western populations who consumed higher levels of high-fat products characterized by a high SFA content, and where the dominant SFAs in the diet are 16:0 SFAs [20,22]. In addition, the diverse intake of other dietary components, such as alcohol or carbohydrates, among Western and Asian populations may contribute to the inconsistent results, as these dietary components are also know to affect blood 16:0 SFA levels [23,24].
On the other hand, the inconsistent results may be attributed to FAs measured in different blood components. Positive associations of circulating 16:0 SFAs with T2D were more observed in plasma phospholipids than in erythrocytes (Table S1). Researchers, using the EPIC-Norfolk study, estimated the prospective association of different measures of FAs with incident T2D, and they observed greater magnitude and strength of association for FAs measured in plasma phospholipids than in erythrocytes [25], which may also partially explain our overall null results.
So far, the prospective association between VLCSFAs and T2D is less well investigated, with a few studies showing an inverse association of circulating VLCSFAs with T2D in Western populations [5,12,14], while another study reported a positive association of erythrocyte 24:0 SFAs with T2D risk [9] (Table S1). Although we did not find a positive association of erythrocyte 24:0 SFAs with T2D, we did observe a positive association of erythrocyte 20:0 SFAs with T2D in our present study. VLCSFAs have cholesterol-raising potential, and plasma phospholipid VLCSFAs were positively associated with total cholesterol [27], which was an independent risk factor for T2D [28]. In vivo experiments suggested that VLCSFA accumulation in macrophages might enhance inflammatory and oxidative responses and play a role in the pathogenesis of inflammatory diseases [29].
VLCSFAs are primarily derived from limited foods, such as macadamia nuts and peanuts [20]. Findings from the Nurses' Health Study suggested that higher nut and peanut consumption was inversely associated with T2D incidence [30], and it was reported that Western populations consumed 4-5 times the number of nuts and peanuts compared with that of Chinese populations [31], resulting in the inconsistent associations between our findings and other reports. More investigations are needed to explore the reason for the inconsistency and the related mechanisms.
We found an interaction of 14:0 SFAs with sex, and 14:0 SFAs were positively associated with T2D risk in male participants, but not in female participants. It was reported that free 14:0 SFAs would reduce the amount of albumin-bound testosterone, the main source of bioavailable testosterone [44]. A meta-analysis indicated that lower testosterone levels are associated with higher risk of T2D in men, but with lower risk in women [45]. It is possible that 14:0 SFAs might mediate bioavailable testosterone levels in relation to the risk of T2D, which might explain the opposite associations of 14:0 SFAs with T2D risk between women and men; however, more investigations are needed.
This study has several strengths. Firstly, we used an objective measurement of erythrocyte FAs. Secondly, our present study had a large sample size, and, to the best of our knowledge, provided the first evidence on this topic among Chinese populations. Thirdly, we included a variety of diet, lifestyle, and socio-demographic factors, as well as circulating biomarkers, in our statistical models, and we performed several sensitivity analyses, all of which supported our findings.
Several limitations should be considered. Firstly, we did not measure erythrocyte odd-chain SFAs, including pentadecanoic acid (15:0) and heptadecanoic acid (17:0), used as biomarkers of dairy intake [46]. Our results show that future cases had a lower dairy intake compared with non-cases (Table 1), and a previous study suggested that dairy intake was associated with a lower risk of T2D [2]. Although we adjusted for dairy intake in sensitive analyses and observed robust results, it may still be a limitation that we could not adjust for circulating 15:0 and 17:0 SFAs.
Secondly, erythrocyte FAs were measured at baseline only and erythrocyte FA compositions may change over time, which may cause potential exposure misclassification. Thirdly, we did not conduct an oral glucose tolerance test to identify diabetes, which might lead to undiagnosed diabetes cases. Finally, although we adjusted for multiple potential confounders, residual confounding due to unmeasured or imprecisely measured confounders could not be excluded.

Conclusions
In conclusion, our findings indicated that individual erythrocyte SFAs were associated with T2D in different directions, and we found that higher levels of 18:0 and 20:0 SFAs were associated with a higher risk of incident T2D, whereas no convincing inverse association was found for 16:0 SFAs. These results suggested that individual SFAs have differential associations with T2D among Chinese populations. However, these results should be interpreted with caution because we did not adjust for dairy biomarkers, and these results were based on a single measurement at baseline. More investigations are warranted to replicate our findings.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6643/10/10/ 1393/s1. Table S1: Review of literature examining association of circulating SFAs with incident type 2 diabetes. Table S2: Spearman's rank correlation of erythrocyte saturated fatty acids and dietary intakes (n = 2683). Table S3: Association of groupings of erythrocyte SFAs with incident type 2 diabetes. Table S4: Sensitivity analyses for association of erythrocyte SFAs with incident type 2 diabetes. Figure S1: Associations of erythrocyte SFAs with incident type 2 diabetes, stratified by categories of BMI and sex.