Early-Phase Changes in Serum Free Fatty Acid Levels After Glucose Intake Are Associated With Type 2 Diabetes Incidence: The Hiroshima Study on Glucose Metabolism and Cardiovascular Diseases

Several studies have reported an increase in serum FFA levels in individuals with prediabetes and type 2 diabetes (1,2). Some studies have shown that increased serum FFA levels have a negative effect on β-cell function (3) and contribute to the development of insulin resistance (4). These lines of evidence suggest the involvement of high serum FFA levels in the development of type 2 diabetes. However, previous epidemiological studies have reported inconclusive results regarding the relationship between fasting serum FFA levels and type 2 diabetes risk (5–10). Insulin regulates serum FFA levels by inhibiting lipolysis, which is the synthesis of FFA and glycerol from triglycerides in adipose tissues (11). Hence, serum FFA levels decline considerably with an increase in serum insulin levels after glucose intake (1). According to a recent study, a reduction in the decline in serum FFA levels after glucose intake rapidly weakens the suppression of hepatic glucose production (12), which may be a major defect that results in impaired glucose metabolism (13). However, information about the relationship between serum FFA levels after glucose intake and type 2 diabetes risk is limited (6), and to the best of our knowledge, no studies have been conducted to investigate whether changes in serum FFA levels after glucose intake contribute to the development of type 2 diabetes.

This retrospective cohort study aimed to investigate the association between serum FFA levels along with their changes after glucose intake and type 2 diabetes risk in a large-scale population that underwent 75-g oral glucose tolerance tests (OGTTs). To determine whether the impaired changes in serum FFA levels after glucose intake is a defect that precedes the onset and increase in plasma glucose (PG) levels, we investigated the relationship between the serum FFA-related parameters and type 2 diabetes risk in both individuals with normal fasting glucose level (NFG)/normal glucose tolerance (NGT) and those with prediabetes. In addition, a stratified analysis using indices of β-cell function and insulin resistance was performed to assess the interrelationship between β-cell dysfunction, insulin resistance, and changes in serum FFA levels after glucose intake.

This study used data collected during the Hiroshima Study on Glucose Metabolism and Cardiovascular Diseases (Hiroshima GMCVD), which was a cross-sectional and longitudinal study that examined the interrelationship between impaired glucose metabolism, hypertension, and cardiovascular disease (CVD) (14,15). Information on the Hiroshima GMCVD participants can be found in the Supplementary Material (Hiroshima Study on Glucose Metabolism and Cardiovascular Diseases population section). Among the 9,961 participants who underwent the 75-g OGTT at the Health Management and Promotion Center of Hiroshima Atomic Bomb Casualty Council between April 2003 and March 2015, 1,124 were excluded because they had type 2 diabetes, which was defined as a fasting PG (FPG) level of ≥7.0 mmol/L and/or 2-h postload PG (2-h PG) level of ≥11.1 mmol/L in the 75-g OGTT (16) at baseline. Of the remaining 8,837 participants, 1,982 were excluded because of the lack of follow-up medical checkup data, which was obtained between 4 years after the baseline examination and the follow-up end point in March 2019. Furthermore, 55 subjects were excluded because of missing baseline data, which included serum FFA levels, uric acid levels, lipid profile, and renal function (Supplementary Fig. 1). During the health checkups, participants were asked about their regular medications and medical histories, including hypertension, dyslipidemia, coronary heart disease, and stroke treatment, as well as their drinking and smoking habits. Written informed consent was obtained from all participants. This study was approved by the Hiroshima Atomic Bomb Casualty Council Committee on the Ethics of Human Research and registered under the University hospital Medical Information Network (UMIN) protocol registration system (ID: UMIN000036648).

The 75-g OGTT was conducted in the morning after overnight fasting. The samples were drawn before and 30, 60, and 120 min after glucose intake to determine serum FFA, PG, and serum immunoreactive insulin (IRI) levels. Serum FFA levels were measured using enzymatic colorimetric methods (Eiken Chemical Co., Ltd., Tokyo, Japan). PG levels were measured using the hexokinase/glucose-6-phosphate dehydrogenase method, and serum IRI levels were measured using a chemiluminescent immunoassay and the Beckman Coulter UniCell DxI at the study site. To estimate the overall serum FFA levels after glucose intake, we calculated the total area under the FFA curve (AUC_FFA) at 0–120 min of the 75-g OGTT using the trapezoidal method (17). The changes in serum FFA levels after glucose intake were assessed on the basis of changes in FFA levels at 0–30, 30–60, and 60–120 min. The percentage of changes in serum FFA levels from 0 to 30 min was calculated as 100 × (fasting FFA level − 30 min FFA level)/fasting FFA levels. Similarly, the percentage of changes in serum FFA levels from 30 to 60 min and 60 to 120 min was calculated. The adipose insulin resistance index (Adipo-IR) was calculated as the product of fasting insulin and fasting FFA levels (18). Whole-body insulin resistance was assessed using the Matsuda index, which was calculated as 10,000 divided by the square root of the levels of FPG × fasting IRI × 2-h PG × 2-h postload IRI (19). β-Cell function was assessed using the ratio of the total area under the insulin curve (AUC_ins) to the total area under the glucose curve (AUC_glu) at 0–120 min of the OGTT (AUC_ins-to-AUC_glu ratio) (20).

We categorized the participants into the NFG/NGT group (n = 3,230), which was defined as the group having an FPG level of <5.6 mmol/L and a 2-h PG level of <7.8 mmol/L, and the prediabetes group (n = 3,570), which was defined as the group having a FPG level of 5.6–6.9 mmol/L and a 2-h PG level of <7.8 mmol/L, or an FPG level of <7.0 mmol/L and 2-h PG of 7.8–11.0 mmol/L at the 75-g OGTT (16). Based on the optimal cutoff values obtained through a receiver operating characteristic (ROC) curve analysis (Supplementary Fig. 2 and Supplementary Table 1), we divided the participants into the following four groups: 1) the high Matsuda index and high AUC_ins/AUC_glu group, defined as the group having a Matsuda index of ≥14.31 and an AUC_ins-to-AUC_glu ratio of ≥24.40 (n = 2571); 2) the low Matsuda index–only group, defined as the group having a Matsuda index of <14.31 and an AUC_ins-to-AUC_glu ratio of ≥24.40 (n = 2,300); 3); the low AUC_ins/AUC_glu–only group, defined as the group having a Matsuda index of ≥14.31 and an AUC_ins-to-AUC_glu ratio of <24.40 (n = 1,730); and 4) the low Matsuda index and low AUC_ins/AUC_glu group, defined as the group having a Matsuda index of <14.31 and an AUC_ins-to-AUC_glu ratio of <24.40 (n = 199).

The incidence of type 2 diabetes was defined as the use of antidiabetes medication and/or HbA_1c ≥6.5% (48 mmol/mol) (16) at the follow-up medical checkup. Hypertension was defined as the use of antihypertensive medication, and/or a systolic blood pressure of ≥140 mmHg, and/or a diastolic blood pressure of ≥90 mmHg (21). Hyperlipidemia was defined as the use of antihyperlipidemic medications and/or a serum total cholesterol level of ≥5.70 mmol/L (22). Hyperuricemia was defined as the use of antihyperuricemic medication and/or a serum uric acid level of >0.42 mmol/L (23). Renal dysfunction was defined as an estimated glomerular filtration rate of <60 mL/min/1.73 m² (24). CVD was defined as coronary heart disease or stroke. Current smoking status was defined on the basis of a current smoking habit, regardless of the number of cigarettes smoked per day, and habitual drinking status was defined as alcohol consumption of ≥20 g/day.

Continuous variables are expressed as mean ± SD, and the normality of continuous variables was analyzed using the Kolmogorov-Smirnov test. Differences between the incident and nonincident diabetes groups were analyzed using the Wilcoxon rank sum test. Categorical variables are presented as percentages and were analyzed using the χ² test. We defined high fasting FFA levels; high FFA levels at 30, 60, and 120 min; large AUC_FFA; high Adipo-IR; low decreases in FFA levels from 0 to 30, 30 to 60, and 60 to 120 min; low rates of changes in FFA levels from 0 to 30, 30 to 60, and 60 to 120 min; low Matsuda index; and low AUC_ins-to-AUC_glu ratio based on the optimal cutoff values obtained via a ROC curve analysis to predict type 2 diabetes. The unadjusted and adjusted odds ratios (aORs) of the incidence of type 2 diabetes based on each serum FFA level–related parameter were determined using univariable and multivariable logistic regression analyses using the following three models: 1) model 1—this included age, sex, BMI, smoking habit (never, former, or current), drinking habit (yes or no), family history of diabetes (yes or no), follow-up period (years), and comorbidities, including hypertension (yes or no), hyperlipidemia (yes or no), and hyperuricemia (yes or no); 2) model 2—this included the FPG and 2-h PG levels in model 1; 3) model 3—this included the Matsuda index and AUC_ins-to-AUC_glu ratio in model 2. Next, model 3 was used to perform a stratified analysis on the basis of glycemic status (NFG/NGT vs. prediabetes). Furthermore, stratified analyses using indices of insulin resistance and β-cell function were also performed using model 3. In addition, the sensitivity analysis listed in the Supplemental Materials was conducted. A P value of <0.05 was considered statistically significant. All statistical analyses were performed using the JMP 14.2 statistical software (SAS Institute, Cary, NC).

A total of 6,800 participants (3,349 men and 3,451 women; mean age, 69.2 ± 5.7 years; mean BMI, 23.1 ± 3.0 kg/m²) were enrolled in the present analysis (Supplementary Table 2). The scatter plot of the association between serum FFA and PG levels during the 75-g OGTT at baseline is depicted in Supplementary Fig. 3. Of these, 485 participants developed type 2 diabetes over a mean 5.3-year follow-up period. Table 1 provides the clinical characteristics and OGTT results of the participants with and without the incidence of type 2 diabetes.

No significant difference was noted in the changes in serum FFA levels from 0 to 120 min after glucose intake between the incident and nonincident diabetes groups (Supplementary Table 3). When the OGTT period was divided into three periods, the decline in serum FFA levels from 0 to 30 min was significantly smaller in the incident diabetes group than in the nonincident diabetes group (0.135 ± 0.142 vs. 0.181 ± 0.158 mmol/L, respectively; P < 0.001). In contrast, from 60 to 120 min, the decline in serum FFA levels was significantly larger in the incident diabetes group than in the nonincident diabetes group. The percentage of changes in serum FFA levels was significantly smaller in the incident diabetes group than in the nonincident diabetes group from 0 to 30 min (25.4 ± 22.3 vs. 34.8 ± 23.4, respectively; P < 0.001) and 30 to 60 min (44.9 ± 16.5 vs. 47.9 ± 17.3, respectively; P < 0.001).

Serum FFA levels at 30 and 120 min and AUC_FFA were significantly associated with type 2 diabetes risk after adjusting for age, sex, BMI, and comorbidities (model 1 in Table 2); however, the significance of this association was lost after adjusting for PG levels (model 2). Conversely, serum FFA levels at 60 min (aOR 1.36; 95% CI 1.09–1.68), reduction in the decline in serum FFA levels from 0 to 30 min (aOR 1.95; 95% CI 1.55–2.46), and low rates of changes in FFA levels from 0 to 30 min (aOR 1.91; 95% CI 1.54–2.37) and 30 to 60 min (aOR 1.48; 95% CI 1.15–1.90) were significantly associated with the incidence of type 2 diabetes after adjusting for PG levels and indices of insulin resistance and β-cell function (model 3). Fasting serum FFA levels (aOR 0.59; 95% CI 0.45–0.76) were inversely associated with type 2 diabetes risk. The sensitivity analysis yielded similar results (Supplementary Tables 4–11).

Supplementary Table 12 reports the incidence rate of type 2 diabetes in the NFG/NGT and prediabetes groups. In the NFG/NGT group, serum FFA levels at 60 min (aOR 2.01; 95% CI 1.18–3.43) and low rate of changes in FFA levels from 30 to 60 min (aOR 1.97; 95% CI 1.05–3.69) were significantly associated with the incidence of type 2 diabetes (Fig. 1A). In the prediabetes group, serum FFA levels at 30 min (aOR 1.34; 95% CI 1.02–1.75) and 60 min (aOR 1.27; 95% CI 1.00–1.60), reduction in the decline in serum FFA levels from 0 to 30 min (aOR, 2.17; 95% CI, 1.68–2.81), and low rates of changes in FFA levels from 0 to 30 min (aOR 2.06; 95% CI 1.62–2.62) and 30 to 60 min (aOR 1.41; 95% CI 1.07–1.87) were significantly associated with the incidence of type 2 diabetes (Fig. 1B).

Supplementary Table 13 summarizes the incidence rate of type 2 diabetes in groups divided based on insulin resistance and β-cell function. Figure 2A shows that the low rate of change in the FFA level from 30 to 60 min (aOR 1.95; 95% CI 1.00–3.80) was significantly associated with the incidence of type 2 diabetes in the high Matsuda index and high AUC_ins/AUC_glu group. In the low Matsuda index–only group, serum FFA levels at 30 min, reduction in the decline in serum FFA levels from 0 to 30 min, and low rate of changes in FFA levels from 0 to 30 min were significantly associated with the incidence of type 2 diabetes (Fig. 2B). In the low AUC_ins/AUC_glu–only group, reduction in the decline in serum FFA levels from 0 to 30 min and low rate of changes in FFA levels from 0 to 30 min were significantly associated with the incidence of type 2 diabetes (Fig. 2C). In the low Matsuda index and low AUC_ins/AUC_glu group, serum FFA levels at 60 min, reduction in the decline in serum FFA levels from 0 to 30 min, and low rates of changes in FFA levels from 0 to 30 min and 30 to 60 min were significantly associated with the incidence of type 2 diabetes (Fig. 2D).

The link between FFA and the pathophysiology of type 2 diabetes has been well studied (25). Numerous experimental studies have shown that a decline in lipolysis suppression causes the release of an excess amount of FFA from adipose tissues and that the increased serum FFA level induces insulin resistance in the muscle and liver, stimulates excessive hepatic glucose production, and impairs insulin secretion from β-cells (26). Conversely, most epidemiological studies have investigated the relationship between fasting serum FFA levels and type 2 diabetes risk (5–10). This study focuses on the lipolysis suppression after glucose intake in a large population and adds the following important findings to this field. First, after adjusting for PG levels and indices of insulin resistance and β-cell function, serum FFA levels at 60 min, attenuation of the decrease in serum FFA levels from 0 to 30 min, and low rates of change in FFA levels at 0–30 min and 30–60 min after glucose intake were associated with new-onset type 2 diabetes. Second, even in individuals with NFG/NGT, a low rate of change in FFA levels at 30–60 min and serum FFA levels at 60 min were associated with type 2 diabetes risk. Finally, although the associations between the serum FFA–related parameters and type 2 diabetes risk were similar in the low Matsuda index–only and low AUC_ins/AUC_glu–only groups, this relationship was prominent in the low Matsuda index and low AUC_ins/AUC_glu group.

Several previous studies have reported that high fasting serum FFA levels predicted the incidence of type 2 diabetes (5–7), whereas some other studies have not (8,9). A detailed investigation conducted by Il’yasova et al. (10) showed that fasting serum FFA levels are inversely associated with type 2 diabetes risk after adjusting for confounding factors, including 2-h PG levels, BMI, and insulin sensitivity. They suggested that fasting serum FFA levels depend on the intensity of fat oxidation (27,28), which may play a protective role in the pathophysiology of type 2 diabetes. In addition, metabolic flexibility (an adaptation to changes in metabolic demand and energy substrate availability) refers to the transition from predominant lipid oxidation during fasting to glucose oxidation after diet (29). Because serum FFA levels after glucose intake decrease to near zero, the low fasting FFA levels mean a low gradient in FFA levels between before and after glucose intake, probably resulting in low metabolic flexibility, which is linked to insulin resistance and impaired glucose metabolism (30). After adjustment for PG levels and indices of insulin resistance and β-cell function, the current study found that fasting serum FFA levels were significantly and negatively associated with type 2 diabetes risk.

In contrast to the fasting FFA levels, after adjusting for age, sex, BMI, and comorbidities, we found serum FFA levels at each time point after glucose intake and the AUC_FFA during OGTT were positively associated with the incidence of type 2 diabetes. The delayed decrease in plasma FFA levels after glucose intake may be due to a reduced first-phase insulin secretion (31). These results were partially consistent with those of a previous study that reported the significant association between serum FFA levels at 2 h after glucose intake and type 2 diabetes risk in individuals with impaired glucose tolerance (6). The low rates of change in FFA levels from 0 to 30 min and from 30 to 60 min, as well as serum FFA levels at 60 min after glucose intake, were clearly associated with type 2 diabetes risk even after adjusting for PG levels and indices of insulin resistance and β-cell function at baseline, implying that the change in serum FFA levels within the 1st h after glucose intake is a primary predictor of type 2 diabetes development. The precise underlying mechanism is unknown, but there are some possibilities. First, some studies have found that a decrease in insulin-induced lipolysis inhibition causes a rapid decay of the decrease in hepatic acetyl-CoA levels, resulting in a reduction in hepatic glucose production (12,32,33). Second, it has been well established that increased FFA levels contribute substantially to the development of insulin resistance in skeletal muscles and the liver (4,34,35). Some studies have revealed that higher FFA levels had a negative impact on β-cell function (3,36). Therefore, an excess of FFA after glucose intake may contribute to the progression of β-cell dysfunction and insulin resistance during follow-up.

Obesity has a well-established association with serum FFA levels (34,35). In a recent cross-sectional study, we also found that obesity was associated with a change in serum FFA levels after glucose intake (37). Nonetheless, after adjusting for BMI, waist circumference, and the presence of metabolic syndrome, multivariate analysis in this cohort study revealed that serum FFA levels and inhibition of the decrease in serum FFA levels after glucose intake were associated with new-onset type 2 diabetes. These findings suggest that serum FFA levels influence the risk of type 2 diabetes independent of obesity.

When the participants were divided into NFG/NGT and prediabetes groups, even in the NFG/NGT group, the low rate of changes in FFA levels from 30 to 60 min and serum FFA levels at 60 min after glucose intake were associated with an increased risk of type 2 diabetes. Similarly, in the stratified analysis, the low rate of change in FFA levels from 30 to 60 min was associated with type 2 diabetes risk in the high Matsuda index and high AUC_ins/AUC_glu group. These results suggest that the impaired changes in serum FFA levels within the 1st h after glucose intake represent an early presentation associated with type 2 diabetes risk before the onset of increase in PG levels. Moreover, as the stage progresses (i.e., the high Matsuda index and high AUC_ins/AUC_glu group, then the low Matsuda index–only and low AUC_ins/AUC_glu–only group, followed by the low Matsuda index and low AUC_ins/AUC_glu group), the serum FFA levels were associated with type 2 diabetes risk increase. Therefore, there might be a vicious cycle between the reduction of changes in serum FFA levels and progression of impaired glucose metabolism, which begins at the NFG/NGT stage.

This study has some limitations. First, because of the retrospective cohort study design, the findings of this study should be confirmed by future prospective cohort studies. Second, during the mean follow-up of 5.3 years, ∼20% of the participants were lost. This may have caused a selection bias in this study. Third, the HbA_1c levels were not considered in the baseline glycemic status diagnosis in this study because the HbA_1c data at baseline for some participants were not available. Fourth, we did not consider diet and physical activity, which may have influenced the development of type 2 diabetes. Fifth, there are ethnic differences in pathophysiology in patients with type 2 diabetes, including insulin resistance, β-cell function, and the prevalence of obesity (38,39). Therefore, we determined a different Matsuda index threshold than that used in previous studies based on the optimal cutoff value from the ROC curve analysis. This may limit the applicability of our findings to patients with other ethnicities. Finally, there was no information on the classes of antihypertensive medications in this study. Therefore, we were unable to consider the effects of some antihypertensive medications that reduce the risk of type 2 diabetes (40).

In conclusion, the finding of this study showed that serum FFA levels and changes in these levels after glucose intake were predictors for the incidence of type 2 diabetes. In particular, the changes in serum FFA levels in the first hour after glucose intake may be a primary predictor of type 2 diabetes development. Even in participants with NFG/NGT, the low rate of changes in serum FFA levels from 30 to 60 min was associated with type 2 diabetes risk, suggesting that this presentation precedes the onset of increase in PG levels and is an early sign of type 2 diabetes development.

Acknowledgments. The authors thank Naomi Yuzono from the Hiroshima Atomic Bomb Casualty Council for technical and secretarial assistance.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. N.S. conducted data analysis, interpreted the data, and wrote the manuscript. R.M. contributed to data acquisition and interpretation. R.O. and K.Y. contributed to data interpretation. Y.N. reviewed and edited the manuscript. Y.H. contributed to data interpretation and reviewed and edited the manuscript. All authors had access to and verified the study data. All authors read and approved the final manuscript. N.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.