Study sample

We invited patients with T2D who attended the outpatient clinics of a tertiary hospital in southern Taiwan from November 2016 to June 2020 to participate in this observational study. The study has been described in detail previously.^12-14 In brief, a diagnosis of T2D was based on a previous history of T2D, blood glucose values defined by the American Diabetes Association, or the use of antidiabetic drugs. All patients were enrolled in an education program on diabetes at our hospital, and they followed the guidelines for a diabetic diet.

Clinical measurements

Clinical and sociodemographic data, including age, sex, cigarette smoking, alcohol use, medications, and comorbidities were obtained from medical records and interviews with the patients. Hypertension was defined as a history of hypertension, taking antihypertensive drugs, or BP equal to or above 140/90 mm Hg. Heart disease was defined as a history of myocardial infarction, ischemic heart disease, or congestive HF. Information on the use of medications, including antidiabetic agents, angiotensin‑converting enzyme inhibitors, angiotensin II receptor blockers, β-blockers, and calcium channel blockers before and after enrollment was collected from medical records. BP was measured in the sitting position after 5 minutes of rest, using a single calibrated device (TERUMO ES‑W110ZJ, Lunder Medical Instrument, Kaohsiung, Taiwan). BP was recorded as a mean of 3 consecutive measurements with a 5‑minute interval between each measurement. Body mass index (BMI) was calculated as body weight divided by body height squared (kg/m²). Information on the usual diet of the patients was obtained using a simple questionnaire. Fasting blood samples were collected for biochemical analyses.

Measurement of circulating short‑chain fatty acids

A detailed method for measurement of circulating SCFA levels was described in previous studies.^13,14 In brief, serum levels of the following 9 SCFAs were measured using liquid chromatography–mass spectrometry (LC‑MS): formate, acetate, propionate, butyrate, isobutyrate, methylbutyrate, valerate, isovalerate, and methylvalerate. Human serum (50 μl) was derivatized using 3‑nitrophenylhydrazine and N-(3‑dimethylaminopropyl)-N‑ethylcarbodiimide in methanol, incubated at 40 °C for 30 minutes, and diluted to 210 μl with 10% aqueous methanol. Then, 75-μl aliquots were mixed with 25 μl of an internal standard, and 10 μl of the mixture was analyzed with LC‑MS/MS. The analysis was performed using the Waters ACQUITY UPLC system (Waters Corporation, Milford, Massachusetts, United States) coupled with the Finnigan TSQ Quantum Ultra triple‑quadrupole MS (Thermo Electron, San Jose, California, United States) with Xcalibur software (Thermo Finnigan, Bellefonte, Pennsylvania, United States), using an electrospray ionization source in a positive mode. Separation was performed on a BEH C18 column (2.1 mm × 100 mm, 1.7 µm; Waters Corporation) at 40 °C, with a 300 μl/min flow rate.

Evaluation of cardiac structure and function

A detailed evaluation of cardiac structure and function was described in a previous study.¹² In brief, echocardiographic examinations were performed by experienced cardiologists using a VIVID 7 system (General Electric Medical Systems, Horten, Norway) with the patients breathing quietly in the left decubitus position. Two‑dimensional and M‑mode images were recorded in standardized views. Echocardiographic measurements, including left atrial diameter (LAD), aortic root diameter, left ventricular posterior wall thickness in diastole and systole (LVPWTd and LVPWTs), left ventricular internal diameter in diastole and systole (LVIDd and LVIDs), interventricular septal wall thickness in diastole and systole, peak early transmitral filling wave velocity (E), and peak late transmitral filling wave velocity (A) were obtained. The E/A ratio below 1 was defined as diastolic dysfunction. The E (early mitral inflow velocity)/E’ (mitral annular early diastolic velocity) ratio was also used to assess diastolic dysfunction, and the data were obtained from 5 beats and then averaged for analysis.¹⁵ Left ventricular systolic function was assessed according to left ventricular fraction shortening (LVFS) and left ventricular ejection fraction (LVEF). In addition, left ventricular mass (LVM) was calculated using the modified Devereux method,¹⁶ while left ventricular mass index (LVMI) was calculated by dividing the LVM by body surface area. The definition of left ventricular hypertrophy (LVH) was based on the 2007 European Society of Hypertension / European Society of Cardiology guidelines.¹⁷ Left ventricular relative wall thickness (LVRWT) was calculated as 2 × LVPWTd/LVIDd. Concentric LVH was defined as LVMI equal to or above 125 g/m² in men and equal to or above 110 g/m² in women, with LVRWT equal to or above 0.45; eccentric LVH was defined as LVMI equal to or above 125 g/m² in men and equal to or above 110 g/m² in women, with LVRWT below 0.45. Stroke volume was calculated using measurements of ventricular volumes and subtracting the volume of the blood in the ventricle at the end of a beat (end‑systolic volume) from the volume of blood just prior to the beat (end‑diastolic volume).

Assessment of brachial‑ankle pulse wave velocity and ankle‑brachial index and definitions of abnormally low and high ankle‑brachial index

Ankle‑brachial index (ABI) was measured using an ABI‑form device (VP1000; Colin Co. Ltd, Komaki, Japan) at enrollment, which automatically and simultaneously measured BP at the 4 extremities through an oscillometric method.¹⁸ ABI was calculated as the ratio of the ankle systolic BP (SBP) divided by the arm SBP. The ABI measurement was performed once for each patient. An abnormally low ABI was defined as below 0.9 and an abnormally high ABI as equal to or above 1.3 in either leg.

As a parameter of arterial stiffness, brachial‑ankle pulse wave velocity (baPWV) was measured using the same ABI‑form device that automatically and simultaneously measured BP on both arms and ankles using an oscillometric method.^19,20 Pulse waves obtained from the brachial and tibial arteries were recorded simultaneously, and the transmission time (ΔTba), which was defined as the time interval between the initial increase in brachial and ankle waveforms, was determined. The transmission distance from the brachium to ankle was calculated according to body height, and the path length from the suprasternal notch to the brachium (Lb) was obtained using the following equation: Lb = 0.2195 × height of the patient (in cm) – 2.0734. The path length from the suprasternal notch to the ankle (La) was obtained using the following equation: La = 0.8129 × height of the patient (in cm) + 12.328. Finally, the following equation was used to obtain baPWV: baPWV = (La – Lb)/ΔTba. The highest baPWV value was used to represent arterial stiffness for each patient.

Statistical analysis

Continuous variables were expressed as mean (SD) or median (interquartile range [IQR]), as appropriate. If the continuous variables were not normally distributed, their values were log10‑transformed before further analysis including univariable and multivariable regression models. Categorical variables were expressed as percentages. The Bonferroni post hoc analysis was also used after analysis of variance or the Kruskal–Wallis test. Differences in categorical variables between groups were tested using the χ² test. Linear or logistic regression was used to investigate the associations between SCFA levels and the cardiac echocardiographic parameters, ABI, and baPWV. Based on the non‑normal distribution of SCFA levels, 9 SCFA variables were introduced as continuous log‑transformed variables; moreover, we stratified the 9 SCFAs, butyrate / isobutyrate ratio, and valerate / isovalerate ratio into tertiles to evaluate the association. We applied the linear regression models to evaluate the associations between SCFAs and LVMI, LAD, E/E’, LVEF, LVFS, stroke volume (SV), and baPWV. The logistic regression models were used for LVH, LAD value above the median, E/A ratio below 1, E/E’ value above the median, baPWV value above the median, and abnormal ABI (<⁠0.9 or ≥1.3). The multivariable linear and logistic models were adjusted for traditional CVD risk factors, such as age, sex, lifestyle factors (smoking history and BMI), clinical diseases (hypertension and heart disease), and laboratory data (serum cholesterol and glycated hemoglobin [HbA_1c] levels).²¹ Statistical analyses were conducted using SPSS 22.0 package for Windows (IBM Inc., Armonk, New York, United States). Significance was set at a 2‑sided P value below 0.05.

Ethics

The study protocol was approved by the Institutional Review Board of Kaohsiung Medical University Hospital (KMUHIRB‑G(II)-20160021, KMUHIRB‑G(I)-20160036, KMUHIRB‑G(II)-20190036). All patients provided their written informed consent to participate in the study, and all clinical investigations were conducted according to the principles expressed in the Declaration of Helsinki. The participants did not give the written consent for their data to be shared publicly, so due to the sensitive nature of the research supporting data are not available.

Characteristics of the study cohort

The clinical characteristics of the study cohort are shown in Table 1. A total of 326 T2D patients were enrolled, at a mean (SD) age of 63.5 (10.4) years, with 56.4% being men. Of the 326 patients, 64.4% had hypertension, 78.5% hyperlipidemia, and 35.3% heart disease. The mean (SD) T2D duration, BMI, and serum cholesterol level were 10.6 (8.5) years, 26.6 (4.4) kg/m², and 167.8 (36.9) mg/dl, respectively, and the median (IQR) HbA_1c level was 7% (6.5%–7.9%). The median (IQR) serum levels of the 9 SCFAs were as follows: formate, 143.3 (95.8–228) μM; acetate, 95.4 (73.3–131) μM; propionate, 14.9 (11.3–21.4) μM; butyrate, 7.9 (5.2–9.7) μM; isobutyrate, 8.6 (5.7–13) μM; methylbutyrate, 7.9 (4.5–14.5) μM; valerate, 2.6 (1.6–5.6) μM; isovalerate, 15.5 (3.2–23.9) μM; and methylvalerate, 1.4 (0.6–3.6) μM (Table 1).

Relationships between short‑chain fatty acids and clinical characteristics

Univariable linear regression or nonparametric method was used to analyze the association between 9 circulating SCFAs and demographic, clinical, and laboratory variables (Supplementary material, Tables S1 and S2). BMI was inversely associated with isobutyrate and methylbutyrate, but positively with isovalerate levels; higher serum triglycerides were inversely related to formate, acetate, methylbutyrate, and valerate levels, but positively associated with propionate level (Supplementary material, Table S1). Current smoking was positively linked to acetate levels (Supplementary material, Table S2).

Relationships between short‑chain fatty acids and cardiac structure

The distribution of echocardiographic parameters is shown in Table 2. Overall, 17.2% of the patients had LVH, including 6.5% of the concentric type and 10.7% of the eccentric type. Regarding LV systolic function, mean (SD) LVEF and LVFS were 70.6% (9.5%) and 40.8% (7.7%), respectively. In addition, 71.1% of the patients had diastolic dysfunction (E/A <⁠1). Mean (SD) ABI was 1.1 (0.1), and 10.5% of the patients had abnormal ABI. We further analyzed the associations among the 9 SCFAs, butyrate / isobutyrate ratio, and valerate / isovalerate ratio with cardiac structure and systolic and diastolic function. We further stratified the SCFAs by tertiles to explore the relationship between the SCFAs and cardiac structure or function (Supplementary material, Table S3). The unadjusted unstandardized β of LVMI was significant for the tertile 3 level of formate, isobutyrate, and valerate, as compared with tertile 1 level of formate (β = 0.04; 95% CI, 0–0.08; P = 0.04), isobutyrate (β = 0.05; 95% CI, 0.01–0.09; P = 0.02, and valerate (β = 0.05; 95% CI, 0.01–0.09; P = 0.03; Supplementary material, Table S4). After adjusting for traditional risk factors, including age, sex, smoking, BMI, hypertension, heart disease, serum cholesterol level, and HbA_1c level of 7%, the patients with tertile 3 level of formate (β = 0.05; 95% CI, 0.01–0.1; P = 0.01), isobutyrate (β = 0.05; 95% CI, 0.01–0.09; P = 0.01), and valerate (β = 0.06; 95% CI, 0.01–0.1; P = 0.01) had higher LVMI than those with tertile 1 level of formate, isobutyrate, and valerate (Table 3). The adjusted odds ratio (OR) of LVH was 2.39 (95% CI, 1.06–5.41; P = 0.03) for tertile 3 level of formate, as compared with tertile 1 level of formate, and 2.6 (95% CI, 1.12–5.99; P = 0.02) for tertile 3 level of isobutyrate compared with tertile 1 level of isobutyrate. In addition, there were no significant associations between any of the SCFAs and LAD in the adjusted analysis (Supplementary material, Table S5).

Relationships between short‑chain fatty acids and cardiac diastolic or systolic function

We defined the E/E’ value greater than the median (8.8) or E/A ratio above 1 as indicating diastolic dysfunction. After adjusting for well‑known risk factors, the patients with the highest tertile of butyrate / isobutyrate ratio had a lower risk of E/E’ above the median, as compared with those with the lowest tertile of butyrate / isobutyrate ratio (OR = 0.05; 95% CI, 0.27–0.93; P = 0.03; Table 4). On the other hand, significant associations between E/A ratio below 1 and butyrate / isobutyrate ratio and acetate were found in the unadjusted model, but the results were not consistent in the adjusted model (Table 4 and Supplementary material, Table S6). In addition, none of the SCFAs were associated with SV or LV systolic function, including LVEF and LVFS (Supplementary material, Table S7).

Relationships between short‑chain fatty acids and arterial stiffness and abnormal ankle‑brachial index

We then analyzed the relationships between circulating SCFAs and the severity of arterial stiffness, defined as a baPWV value greater than the median (1719 cm/s; Supplementary material, Table S8). Among the 9 SCFAs, the patients with the highest tertile level of valerate had a lower risk of arterial stiffness than those with the lowest tertile level (adjusted OR = 0.47; 95% CI, 0.23–0.95; P = 0.04; Table 5). In addition, the patients with the highest tertile level of butyrate less often had abnormal ABI (<⁠0.9 or ≥1.3) than those with the lowest tertile level (adjusted OR = 0.26; 95% CI, 0.08–0.8; P = 0.02; Table 6).

Relationships between short‑chain fatty acids and subclinical cardiovascular disease in the subgroup analysis

We performed the subgroup analysis of the patients stratified by age (<⁠65 vs ≥65 y), sex, HbA_1c level (≤7% vs >7%), and BMI (<⁠25 kg/m² vs ≥25 kg/m²) to further examine the synergistic interactions between circulating SCFAs and these covariates with the parameters of subclinical CVD (Figure 1). A significant positive relationship between LVH and serum formate across tertile 1 to 3 was found in the patients younger than 65 years (Figure 1A), whereas a positive relationship between LVH and serum isobutyrate was found in the patients aged 65 years or older (Figure 1B). The patients younger than 65 years with the highest tertile level of butyrate / isobutyrate ratio had a lower risk of E/E’ above the median, as compared with those with tertile 1 level of butyrate / isobutyrate ratio (Figure 1C). In addition, the men with the highest tertile levels of serum valerate (Figure 1D) and butyrate (Figure 1E) had lower risks of baPWV above the median and abnormal ABI than those with the lowest tertile levels of serum valerate and butyrate, respectively. Moreover, an inverse association between LVH and serum isobutyrate was found in the patients with BMI below 25 kg/m² and those with poor glycemic control (HbA_1c >7%; Figure 1B).

Limitations

Several limitations of this study should be acknowledged. First, SCFAs and echocardiographic parameters were measured only once at enrollment, and therefore the effect of variations in SCFA levels on cardiac structure and function over time may be underestimated. In addition, the cross‑sectional study design limits the ability to establish causal relationships between SCFAs levels and cardiac structure and function. Second, the relatively small sample size may have reduced the statistical power of the results. In addition, the participants were enrolled from a single hospital, and the sample may be insufficient to represent the broader population or different ethnicities. Third, we did not have detailed information on diet composition (such as total energy, carbohydrate, protein, and fat intake) and lifestyle, which may influence the serum concentrations of SCFAs. Fourth, the proportion of patients on sodium‑glucose transporter type 2 inhibitors was low in this study, which may be not in line with current common use of this class of drugs in most countries. In addition, we did not analyze stool levels of the SCFAs to compare with their serum levels. Further research is needed to explore the interactions among SCFAs in the serum, feces, and microbiota with subclinical CVDs in T2D patients. Finally, the biological mechanisms of SCFA action in CVD remain unknown, and future in vitro and in vivo studies are necessary to investigate the pathophysiology of SCFAs in CVD and other novel biomarkers in T2D patients.

Conclusions

This study demonstrated various relationships between different SCFAs and subclinical CVD parameters in T2D patients. Our findings suggest that formate, butyrate, isobutyrate, butyrate / isobutyrate ratio, and valerate are potential biomarkers of cardiovascular remodeling in T2D patients. Future studies are warranted to investigate the pathogenic role of SCFAs in cardiovascular structure and function.