Association of Lower Cardiovagal Tone and Baroreflex Sensitivity With Higher Liver Fat Content Early in Type 2 Diabetes

Abstract

Context

Cardiovascular autonomic neuropathy (CAN) diagnosed by diminished heart rate variability (HRV) is prevalent and carries an increased risk of mortality in patients with diabetes and chronic liver diseases.

Objective

To determine whether lower HRV is associated with increased liver fat content in recent-onset diabetes.

Design

Cross-sectional study.

Setting

German Diabetes Study (GDS), Düsseldorf, Germany.

Participants

Individuals with type 1 diabetes (n = 97) or type 2 diabetes (n = 109) with known diabetes duration ≤1 year and two age- and sex-matched glucose-tolerant control groups from the GDS baseline cohort.

Main Outcome Measures

Four time and frequency domain HRV indices each were measured over 3 hours during a hyperinsulinemic-euglycemic clamp, whereas spontaneous cross-correlation baroreflex sensitivity (xBRS) was computed over 5 minutes. Hepatic fat content was determined by ¹H magnetic resonance spectroscopy, and values >5.56% were defined as hepatic steatosis.

Results

Hepatic steatosis was observed in 52% and 5% of patients with type 2 and type 1 diabetes, respectively. After adjustment for sex, age, body mass index, smoking, diabetes duration, hemoglobin A1c, M-value, and triglycerides, all four vagus-mediated time domain HRV indices, three of four frequency domain indices, and xBRS were inversely associated with liver fat content in participants with type 2 diabetes (all P < 0.05) but not in the group with type 1 diabetes.

Conclusions

Both lower cardiovagal tone and baroreflex sensitivity are strongly associated with prevalent hepatic steatosis in patients with recent-onset type 2 as opposed to type 1 diabetes, suggesting a role for hepatic steatosis in the early development of parasympathetic CAN in type 2 diabetes.

Nonalcoholic fatty liver disease (NAFLD) has become the most common chronic liver disease in high-income countries, affecting up to one-third of the general adult population, and NAFLD is reaching epidemic proportions in patients with type 2 diabetes. Patients with NAFLD are at increased risk of progression to nonalcoholic steatohepatitis (NASH) and of death from cirrhosis and hepatocellular carcinoma, and NAFLD is rapidly becoming the leading indication for liver transplantation (1, 2). The mechanisms underlying NAFLD in type 2 diabetes are not entirely understood but involve hepatic fat accumulation, alterations of energy metabolism, and inflammatory signals (2). Moreover, there is accumulating evidence suggesting a link between NAFLD and cardiovascular disease (CVD) in people with and without diabetes (3). In addition, a relationship between NAFLD and cardiac arrhythmias, such as atrial fibrillation, QTc prolongation, and ventricular arrhythmias, has been reported. This is of clinical interest because it could explain, at least in part, the increased risk of death for CVD in patients with NAFLD (4). Likewise, alterations in the cardiac autonomic nervous system regulation contribute to an increased mortality, sudden death, and cardiac arrhythmias in individuals with and without diabetes.

Cardiovascular autonomic neuropathy (CAN) affects ∼20% of people with diabetes and up to 11% of those with prediabetes and also predicts an increased risk of major cardiac events and mortality (5). However, CAN is encountered as a comorbidity and prognostic marker not only in patients with diabetes but also in those with chronic liver disease, particularly cirrhosis of various etiologies (6, 7). The fluctuations in autonomic inputs to the heart are measured by spectral analysis of heart rate variability (HRV) indicating that efferent vagal activity is a major contributor to the high-frequency (HF) component (8). Vagus nerve signaling also plays a paramount role in the regulation of metabolic homeostasis aimed at preserving energy balance and preventing fluctuations in body weight and metabolism (9). The hepatic branch of the vagus nerve is involved in mediating the suppressive effect of intestinal lipid on feeding behavior (10).

Although the invasive liver biopsy is the current clinical gold standard for assessment of hepatic steatosis, it is a suboptimal tool for research studies (11). Noninvasive methods such as ultrasonography and computed tomography can be used to assess liver fat but have limited sensitivity and accuracy in quantification. Magnetic resonance spectroscopy (MRS) can separately quantify liver fat and water signal components and therefore assess liver fat more directly than ultrasonography or computed tomography (11) and is currently considered the most accurate and standard noninvasive method with high sensitivity for liver fat assessment (12). Intracellular accumulation of triglycerides within hepatocytes is the earliest and signature histologic feature of NAFLD (13). Previous studies focusing on the relationship between NAFLD assessed by ultrasonography and diabetic sensorimotor polyneuropathy (DSPN) have shown variable results. In patients with type 1 diabetes, DSPN was positively associated with NAFLD (14), whereas in patients with type 2 diabetes, DSPN was either inversely associated (15) or not associated with NAFLD (16). However, to date, neither studies using MRS to measure liver fat content nor those assessing the relationship between the latter and autonomic function have been published. Because parasympathetic CAN is prevalent and is a prognostic marker in patients with diabetes and chronic liver diseases, we hypothesized that an early suppression of cardiovagal tone may be associated with hepatic steatosis in patients with recent-onset diabetes.

Material and Methods

Participants

Participants aged 18 to 69 years were recruited consecutively from the baseline cohort of the ongoing German Diabetes Study, a prospective observational study investigating the natural course of metabolic abnormalities and the development of diabetes-associated chronic complications in individuals recently diagnosed with diabetes and known diabetes duration ≤1 year. The present cross-sectional analysis included 97 participants with type 1 diabetes, 109 individuals with type 2 diabetes, and two corresponding age- and sex-matched glucose-tolerant control groups (control 1, n = 46; control 2, n = 56). Type 1 and type 2 diabetes were defined according to the criteria of the American Diabetes Association as previously reported (17). Informed written consent was obtained from all volunteers prior to inclusion into the trial (ClinicalTrial.gov registration no.: NCT01055093), which was approved by the ethics board of Heinrich Heine University, Düsseldorf, Germany. All participants were screened, including medical history, self-reported alcohol consumption over the past 12 months using a food frequency questionnaire, physical examination, and routine blood tests. The study design and cohort profile of the German Diabetes Study are described in detail elsewhere (17).

¹H MRS

Magnetic resonance imaging and MRS acquisition

All magnetic resonance imaging and MRS measurements were conducted on a 3.0-T MR scanner with a body coil (Achieva X-series; Philips Health care, Best, Netherlands). To acquire ¹H MRS of the liver, the location of liver was first identified using a fast localizer magnetic resonance imaging. After confirming the location of the liver, a volume of interest for ¹H MRS was selected. The size of a volume of interest was 25 × 25 × 25 mm³, which was carefully placed within the posterior region of the liver, avoiding major vessels and gallbladder. ¹H MRS measurement was acquired using stimulated echo acquisition mode sequence (repetition time/echo time/mixing time = 4000/10/16 ms) after shimming. Both water-suppressed and nonsuppressed ¹H MRS of the liver were taken in the identical voxel, and the water signal at 4.7 ppm in non-water-suppressed MRS served as an internal reference for fat quantification, as previously published (18, 19).

Processing of MRS and quantification of fat content

All liver spectra were processed using jMRUI software (20). The absolute percentage of fat content was quantified as the ratio of intensities of the methylene [-(CH₂)n-] peak in triglycerides at 1.3 ppm of ¹H liver spectra to the combined signal intensities of the water and methylene fat peaks [fat content (%) = 100 × fat/(water + fat)]. The difference in transverse relaxation times (T2) of water and fat peaks was corrected based on a previous publication (21). Hepatic fat content >5.56% (or 55.6 mg/g liver tissue) was defined as hepatic steatosis as previously suggested, which corresponds to the 95th percentile of individuals with no risk factors for hepatic steatosis (22).

HRV

R-R intervals were measured in the supine position on the second day of the study during an intravenous glucose tolerance test for 1 hour followed by a hyperinsulinemic-euglycemic clamp for at least 2 hours using a digital Spider View Holter recorder with seven electrodes to record three-channel electrocardiograms (Sorin Group, Munich, Germany). HRV was analyzed hourly as a mean of 3 hours from the Holter monitor recordings with the SyneScope version 3.00 analysis system (Sorin Group). The sampling rate of the electrocardiogram signal was 200 Hz (5-ms resolution). The system automatically filters all artifacts and ectopic beats and generates a regular signal by linear interpolation of the heart rate tachogram. Time domain measures included the standard deviation of differences between adjacent normal-to-normal (NN) intervals (SD), standard deviation of all NN intervals, the number of pairs of adjacent NN intervals differing by >50 ms in the entire recording divided by the total number of NN intervals, and the root mean square of successive differences. Frequency domain indices included the very-low-frequency (VLF) band (0.003 to 0.04 Hz), low-frequency (LF) band (0.04 to 0.15 Hz), HF band (0.15 to 0.4 Hz), and LF/HF ratio as previously suggested (8).

Cardiovascular autonomic function tests were used to diagnose CAN, including seven HRV indices measured during spontaneous breathing over 5 minutes (coefficient of R-R interval variation, VLF power, and LF power), at deep breathing (expiration-to-inspiration ratio), after standing up (maximum/minimum 30:15 ratio), and in response to a Valsalva maneuver (Valsalva ratio) using VariaCardio TF5 (MIE Medical Research Ltd., Leeds, UK), as previously described (23). Age-dependent lower limits of normal were defined at the 5th percentile of healthy individuals. The systolic blood pressure response to standing up was measured over 3 minutes using −27 mm Hg as an age-independent lower limit of normal. Borderline CAN was assumed if two of seven indices were abnormal, whereas definite CAN was diagnosed if three or more of seven indices were abnormal (23).

Baroreflex sensitivity

Cross-correlation baroreflex sensitivity (xBRS) was computed over 5 minutes in the supine position during spontaneous breathing as previously described (24). In brief, cross-correlation and regression between systolic blood pressure and R-R interval were computed over 10-second sliding windows, a time span sufficient to fully accommodate a 10-second variability in rhythm, or several cycles at ventilatory frequencies. For each window, the correlation coefficient was computed six times. The cross-correlation with the greatest value was selected, and the corresponding regression slope was taken as a determination of xBRS, provided it was positive and its probability of being a random regression was <1% (P < 0.01) (24).

Hyperinsulinemic-euglycemic clamp

All participants underwent an intravenous glucose tolerance test for 1 hour followed by a modified Botnia clamp test with [6,6-²H₂]glucose for 2 hours to measure whole-body and tissue-specific insulin sensitivity as previously described and validated (17). Whole-body insulin sensitivity [M-value; mg glucose × (body weight in kg)⁻¹ × min⁻¹] was calculated from the difference between mean glucose infusion rates during steady state in the last 30 minutes of the clamp with glucose space correction (17).

Bioelectrical impedance

Fat-free mass and fat mass were measured using Nutriguard-S (Data Input, Darmstadt, Germany) by determining resistance and reactance. Participants were examined in the morning after an overnight fast in the supine position with arms and legs abducted from their body. BIANOSTIC-AT double-size electrodes (Data Input, Pöcking, Germany) were fixed on the dorsum of the hand and foot of the dominant side of the body. Resistance and reactance were integrated into an equation to calculate fat-free mass and fat mass (25).

Laboratory analyses

Hemoglobin A1c (HbA1c), high-density lipoprotein (HDL) cholesterol, low-density lipoprotein cholesterol, triglycerides, liver enzymes, and creatinine were measured as previously described (24).

Statistical analysis

Data are presented as median (first and third quartiles) or percentages. Variables with skewed distribution were ln-transformed before analyses. The LF/HF ratio was calculated using the ln-transformed LF and HF values. To determine differences between groups or correlations between variables, the Mann-Whitney U test and Spearman rank correlation were applied for nonparametric and Student t-test and Pearson correlation analyses for parametric data. Stepwise multiple linear regression analyses with adjustments for sex, age, body mass index (BMI), smoking, diabetes duration, HbA1c, M-value, and triglycerides were performed as indicated. All statistical tests were two-sided, and the level of significance was set at α = 0.05. Where appropriate, P values obtained from univariate correlation analyses were adjusted for multiple comparisons for each diabetes group separately using the Bonferroni correction: P = 0.05/8(8 HRV indices) = 0.00625. All analyses were performed using SPSS version 22.0 software (SPSS, Inc., Chicago, IL).

Results

The demographic and clinical data of the four groups studied are listed in Table 1. Compared with control 1, the group with type 1 diabetes had higher HbA1c and M-value, whereas no differences between the groups were noted for the remaining variables. After adjustment, only the difference in M-value remained statistically significant (P < 0.05). Compared with control 2, the group with type 2 diabetes had higher BMI, alanine aminotransferase, fat mass, liver fat content, and HbA1c as well as lower HDL cholesterol and M-value. After adjustment, only the differences for HDL cholesterol, alanine aminotransferase, fat mass, and M-value remained statistically significant (P < 0.05). Furthermore, compared with control 2, patients with type 2 diabetes had higher rates of hepatic steatosis, microalbuminuria/macroalbuminuria, and CAN and used antihypertensive drugs more frequently (P < 0.05). No differences between these two groups were found for the remaining variables.

Table 1.

Demographic and Clinical Characteristics of the Four Groups Studied

Variable	Control 1	Type 1 Diabetes	Control 2	Type 2 Diabetes
n (% male)	46 (70)	97 (54)	56 (70)	109 (69)
Diabetes duration (d)	—	165 (121, 271)	—	163 (97, 258)
Age (y)	35 (26, 44)	33 (27, 43)	54 (44, 61)	52 (45, 60)
BMI (kg/m2)	25.4 (22.7, 28.4)	24.1 (22.1, 27.9)	27.3 (25.6, 30.5)	31.4 (27.4, 34.5)a
Current smoking status (% yes)	30	28	32	22
Heart rate (bpm)	65 (61, 73)	67 (60, 75)	69 (62, 74)	70 (64, 80)
Systolic blood pressure (mm Hg)	123 (112, 136)	120 (110, 129)	128 (114, 144)	126 (116, 140)
Diastolic blood pressure (mm Hg)	71 (61, 79)	68 (62, 74)	72 (65, 79)	69 (66, 78)
Triglycerides (mmol/L)	0.91 (0.61, 1.36)	0.89 (0.65, 1.23)	1.23 (0.86, 1.90)	1.41 (1.06, 2.16)
Creatinine (µmol/L)	79.2 (72.3, 87.4)	75.2 (62.8, 82.3)	78.8 (73.5, 87.8)	76.1 (66.4, 88.5)
HDL cholesterol (mmol/L)b	1.58 (1.21, 1.89)	1.57 (1.26, 1.87)	1.46 (1.15, 1.84)	1.16 (0.96, 1.32)a
LDL cholesterol (mmol/L)	2.562 (2.23, 3.42)	2.77 (2.30, 3.38)	3.62 (2.86, 4.10)	3.34 (2.66, 3.95)
Aspartate aminotransferase (U/L)	23.0 (20.0, 28.5)	20.0 (16.5, 24.0)	22.0 (20.0, 26.0)	23.0 (19.0, 31.0)
Alanine aminotransferase (U/L)b	23.5 (16.8, 30.5)	20.0 (15.0, 25.0)	25.0 (19.0, 33.0)	32.0 (23.0, 46.0)a
γ-Glutamyl transferase (U/L)	17.0 (12.5, 26.5)	15.0 (11.0, 24.0)	22.0 (16.0, 40.0)	33.0 (22.0, 49.5)
Alkaline phosphatase (U/L)	65.0 (50.0, 74.0)	63.0 (52.5, 72.5)	68.0 (57.0, 81.0)	65.0 (54.0, 80.5)
BIA fat mass (%)c	25.0 (19.2, 30.5)	24.4 (18.2, 31.5)	29.7 (25.6, 36.5)	33.7 (28.7, 39.3)a
Liver fat content (%)	0.51 (0, 1.95)	0.30 (0, 1.21)	1.89 (0.47, 5.80)	5.80 (2.70, 12.47)a
Hepatic steatosis (%)	13.0	5.2	26.8	52.3a
HbA1c (%)	5.2 (5.0, 5.3)	6.2 (5.7, 7.1)d	5.3 (5.2, 5.5)	6.2 (5.9, 6.7)a
HbA1c (mol/mmol)	33.3 (31.1, 34.4)	44.3 (38.8, 54.1)d	34.4 (33.3, 36.6)	44.3 (41.0, 49.7)a
M-value (mg × kg−1 × min−1)b	11.8 (9.0, 13.8)	8.3 (6.2, 10.1)d	9.9 (8.2, 11.9)	5.5 (4.3, 7.3)a
Diabetes treatment
Diet only (%)	—	4	—	28
Insulin only (%)	—	86	—	3
Glucose-lowering drugs without insulin (%)	—	2	—	62
Glucose-lowering drugs and insulin (%)	—	8	—	7
Antihypertensive drugs (%)	9	9	21	46a
Microalbuminuria/macroalbuminuria (%)	5	6	2	12a
CAN (incipient/definite) (%)	7 (5/2)	13 (7/6)	4 (2/2)	19 (9/10)a

Data are % or median (first, third quartile). Boldface indicates P < 0.05 after adjustment.

Abbreviations: BIA, bioelectrical impedance analysis; LDL, low-density lipoprotein.

^a

P < 0.05 vs control 2.

^b

Group comparisons adjusted for sex, age, BMI, smoking status, and HbA1c.

^c

Group comparisons adjusted for sex, age, smoking status, and HbA1c.

^d

P < 0.05 vs control 1.

The indices of HRV and baroreflex sensitivity are shown in Table 2. Standard deviation of all NN intervals, SD, and VLF power were lower in the group with type 2 diabetes than in control 2 after adjustment for sex, age, BMI, smoking status, and HbA1c. No differences between these groups were noted for the remaining parameters. Moreover, there were no differences between the group with type 1 diabetes and control 1 in any of the parameters studied.

Table 2.

Indices of HRV and Baroreflex Sensitivity in the Four Groups Studied

Variable	Control 1	Type 1 Diabetes	Control 2	Type 2 Diabetes
Time domain indices
pNN50 (%)	16.8 (5.1, 34.9)	13.9 (4.8, 27.7)	5.6 (1.8, 16.4)	3.2 (0.9, 9.4)
RMSSD (ms)	42.4 (27.2, 65.9)	37.9 (26.4, 58.3)	27.7 (20.7, 42.9)	23.6 (18.1, 34.5)
SDNN (ms)	76.1 (53.9, 91.9)	71.0 (55.4, 88.1)	60.3 (44.5, 75.3)	48.0 (39.2, 60.6)a
SD (ms)	90 (69, 121)	87 (73, 105)	78 (60, 91)	63 (51, 80)a
Frequency domain indices
VLF power (ms2)	3018 (1671, 5347)	2592 (1753, 3814)	2014 (1254, 2771)	1390 (881, 2099)a
LF power (ms2)	1296 (625, 1990)	1252 (695, 1897)	654 (406, 1373)	565 (315, 958)
HF power (ms2)	440 (146, 767)	378 (192, 888)	137 (83, 400)	124 (68, 281)
LF/HF ratio	3.21 (1.94, 4.39)	3.15 (2.11, 4.64)	3.88 (2.77, 6.02)	4.19 (2.30, 7.07)
Baroreflex sensitivity xBRS (ms/mm Hg)	12.5 (8.2, 19.5)	11.8 (8.1, 18.2)	6.9 (4.9, 9.9)	6.5 (4.5, 8.9)

Data are median (first, third quartile). Mann-Whitney U test was used to determine differences between groups followed by stepwise multiple linear regression analysis. Boldface indicates significance after adjustment for sex, age, BMI, smoking status, and HbA1c.

Abbreviations: pNN50, number of pairs of adjacent NN intervals differing by >50 ms in the entire recording divided by the total number of NN intervals; RMSSD, root mean square of successive differences; SDNN, standard deviation of all NN intervals.

^a

P < 0.05 vs control 2.

The relationships between the HRV and xBRS indices and liver fat content are shown in Table 3. In the group with type 2 diabetes after Bonferroni correction and adjustment for sex, age, BMI, smoking status, diabetes duration, HbA1c, M-value, and triglycerides, each HRV measure (except for HF power and LF/HF ratio) was strongly inversely associated with liver fat content (P ≤ 0.007), whereas the association between xBRS and liver fat content was somewhat weaker (P = 0.017). Figure 1 exemplifies the inverse associations of LF and VLF power spectrum, SD of differences between adjacent NN intervals, and xBRS with liver fat in participants with type 2 diabetes. The final models of the stepwise multiple linear regression analyses showed that liver fat was the only factor contributing to most of the associations shown in Table 3. Other independent contributors were only M-value (β = 0.286/P = 0.006) and BMI (β = 0.224/P = 0.028) in the final model for total number of NN intervals as well as age (β = −0.245/P = 0.007) and sex (β = −0.186/P = 0.040) in the final model for LF power. In contrast to the group with type 2 diabetes, no associations of the HRV and xBRS indices with liver fat content were found in the group with type 1 diabetes (Table 3).

Table 3.

Associations of HRV Indices and xBRS With Liver Fat Content

Characteristic	Liver Fat
	r	P Value	β	P Value
Time domain indices
pNN50
Type 1 diabetes	−0.157	0.125	−0.089	0.343
Type 2 diabetes	−0.331	0.0005	−0.272	0.007
RMSSD
Type 1 diabetes	−0.144	0.159	−0.099	0.292
Type 2 diabetes	−0.274	0.004	−0.280	0.003
SDNN
Type 1 diabetes	−0.191	0.061	−0.147	0.111
Type 2 diabetes	−0.330	0.0005	−0.354	0.0002
SD
Type 1 diabetes	−0.227	0.025	−0.108	0.303
Type 2 diabetes	−0.360	0.0001	−0.364	0.0001
Frequency domain indices
VLF power
Type 1 diabetes	−0.206	0.043	−0.171	0.052
Type 2 diabetes	−0.321	0.001	−0.322	0.001
LF power
Type 1 diabetes	−0.194	0.057	−0.040	0.689
Type 2 diabetes	−0.263	0.006	−0.312	0.001
HF power
Type 1 diabetes	−0.129	0.207	−0.071	0.434
Type 2 diabetes	−0.219	0.022	−0.261	0.006
LF/HF ratio
Type 1 diabetes	0.025	0.810	−0.076	0.593
Type 2 diabetes	0.065	0.502	0.114	0.333
Baroreflex sensitivity xBRS
Type 1 diabetes	−0.223	0.031	−0.165	0.088
Type 2 diabetes	−0.307	0.002	−0.248	0.017

Boldface indicates P < 0.00625 after Bonferroni correction. All data were ln-transformed before adjustment for sex, age, BMI, smoking status, diabetes duration, HbA1c, M-value, and triglycerides.

Abbreviations: pNN50, number of pairs of adjacent NN intervals differing by >50 ms in the entire recording divided by the total number of NN intervals; RMSSD, root mean square of successive differences; SDNN, standard deviation of all NN intervals.

Discussion

In this study, we demonstrate strong associations of both lower cardiac parasympathetic activity and baroreflex sensitivity (BRS) with increased hepatic fat content, independent of measures contributing or pointing to insulin resistance such as M-value, BMI, or serum triglycerides in patients with recent-onset type 2 but not type 1 diabetes. In contrast, sympathetic predominance characterized by increased LF/HF power was not linked to hepatic steatosis. Because approximately half of the patients recently diagnosed with type 2 but only 5% of those with type 1 diabetes had evidence of NAFLD, these data suggest that the latter may be relevant in the early development of parasympathetic CAN in type 2 as opposed to type 1 diabetes. Indeed, the prevalence of incipient/definite CAN was higher in the former than in the latter group.

There are no published studies available with which our findings could be directly compared. Previous studies have focused on possible associations of ultrasound-diagnosed NAFLD with DSPN (14, 15, 26) or with retinopathy and nephropathy (27–29) and have shown variable results. In patients with type 1 diabetes, prevalent DSPN was associated with NAFLD (14), whereas DSPN in patients with type 2 diabetes was associated with a lower prevalence of NAFLD (15) or no association of DSPN with NAFLD was found (26). Likewise, in some studies, prevalent or incident chronic kidney disease and prevalent retinopathy were associated with NAFLD (27–29), whereas in others, NAFLD was associated with lower prevalence of both nephropathy and retinopathy in patients with type 2 diabetes (15, 16). Thus, the available evidence from ultrasound-based studies suggests that NAFLD in type 1 diabetes is linked to prevalent or even incident microvascular complications, whereas in type 2 diabetes, for unknown reasons, it relates to an even lower prevalence of microvascular complications and DSPN.

Although CAN has previously been recognized as a feature of chronic liver diseases (6, 7), there are no studies assessing the relationship between NAFLD and cardiovascular autonomic function in patients with diabetes. In a small group of nondiabetic individuals with histologically proven, noncirrhotic NAFLD, autonomic symptoms such as orthostatic hypotension were relatively frequent, but autonomic function was not assessed (30). In another small study in patients with NAFLD, among whom one-third had type 2 diabetes, heart rate recovery after exercise, an indicator of vagal reactivation, was inversely associated with body composition and body fat distribution (31). Both NAFLD and CAN are predictors of CVD, increased mortality, sudden death, and cardiac arrhythmias (4). Thus, autonomic nervous system alterations could represent an integral bridging factor contributing to the heightened cardiovascular risk shared by NAFLD and CAN.

This study shows consistent inverse associations between liver fat content and several vagus-modulated HRV indices, VLF power, and LF power in patients with type 2 diabetes; therefore, the lack of correlation of the LF/HF ratio with hepatic steatosis in the opposite direction may not be surprising but rather is compatible with the critique that LF power, with or without adjustment for HF power, does not provide an index of cardiac sympathetic tone, because it is unrelated to cardiac sympathetic tone during supine rest and also does not relate to sympathetic nervous responses to acute manipulations (32). Several lines of evidence fit with the concept that LF power is of central origin and particularly support an association between LF power and baroreflex modulation of autonomic outflows (32). In line with this notion, we also found an inverse association between lower BRS and increased liver fat content in type 2 diabetes. The baroreceptor–heart rate reflex is a key mechanism contributing to the neural regulation of the cardiovascular system. The quantitative estimation of BRS is regarded as a synthetic index of neural regulation at the sinus atrial node and has been shown to provide clinical and prognostic information in a variety of CVDs (33). Thus, we suggest that NAFLD in recent-onset type 2 diabetes is accompanied by lower vagal activity rather than by sympathetic predominance as derived from spontaneous HRV and BRS recordings. Compatible with this concept is also the strong association we found between VLF power and liver fat content in type 2 diabetes. VLF power can almost be completely attenuated by parasympathetic blockade using atropine and modestly blunted by angiotensin-converting enzyme blockade (34). Although VLF heart period rhythms are influenced by the renin-angiotensin-aldosterone system, as LF and HF R-R interval rhythms, they depend primarily on the presence of parasympathetic outflow (34).

The liver is innervated by both the sympathetic and the parasympathetic nervous systems. These nerves are derived from the splanchnic and vagal nerves that surround the portal vein, hepatic artery, and bile duct. The afferent fiber delivers information regarding osmolality, glucose level, and lipid level in the portal vein to the central nervous system. In contrast, the efferent fiber is crucial in the regulation of metabolism, blood flow, and bile secretion (35). The hepatic branch of the vagus nerve is critically involved in mediating the suppressive effect of intestinal lipid on feeding behavior. Hepatic portal infusions of lipids increase hepatic vagal afferent activity, and vagal denervation markedly raises plasma triglycerides (10). The hepatoportal vagal sensing of lipids may therefore not only be important for the “reflexive” regulation of feeding behavior but also may play a role in the pathophysiology of metabolic abnormalities such as hepatic insulin resistance. Elevated levels of free fatty acids (FFAs) in the portal vein have been suggested to induce insulin resistance, because they might directly reduce insulin clearance by the liver. The overloading of the liver by FFAs may finally cause an increased synthesis of triglycerides and an excess secretion of very-low-density lipoprotein (10). On the other hand, it has also been argued that energy storage as triglycerides in lipid droplets in hepatocytes may be considered an attempt to protect the liver from FFA-induced lipotoxicity and from hyperinsulinemia-driven de novo lipogenesis, and those who display a better ability to clear fat from the liver would probably be those at lower risk of developing long-term hepatic complications (2). Thus, in line with our results, preserved vagal activity could be protective in the context of hepatic fat accumulation in type 2 diabetes.

Furthermore, the vagus nerve has an important role in the regulation of metabolic homeostasis, and efferent vagus nerve–mediated cholinergic signaling controls immune function and proinflammatory responses via the inflammatory reflex. Dysregulation of metabolism and immune function in obesity are associated with chronic low-grade inflammation, a critical step in the pathogenesis of insulin resistance and type 2 diabetes. Cholinergic mechanisms within the inflammatory reflex have been implicated in attenuating obesity-related inflammation and metabolic complications (9). Steatosis in NAFLD only rarely is an “isolated” finding and is frequently accompanied by a chronic mononuclear cell inflammatory infiltrate located in the acini and composed of lymphocytes, rare plasma cells, and monocytes (36). In fact, experimental studies suggest that inflammation is a driver for the development and progression of NAFLD (36). In line with these pathogenetic concepts, the current study, together with our recent data, indicates that a complex interplay between subclinical inflammation (37), insulin resistance (38), and hepatic steatosis, the latter possibly driven by incipient inflammatory changes, makes a substantial contribution to the early development of parasympathetic CAN in patients recently diagnosed with type 2 diabetes.

The results of this study also open up therapeutic perspectives supporting the concept of possible synergistic drug effects aimed at concomitantly improving NASH via attenuating insulin resistance and augmenting BRS by insulin sensitizers such as pioglitazone or via reducing oxidative stress to improve steatosis and inflammation in NASH by antioxidants such as vitamin E (1). Similarly, antioxidant treatment was shown to improve CAN in patients with type 2 diabetes (39).

The strengths of this study are the relatively large sample sizes of individuals with well-controlled diabetes or normal glucose tolerance who underwent comprehensive metabolic characterization assessed by state-of-the-art methodology, including the gold standards for the assessment of both insulin sensitivity and hepatic fat content. Furthermore, HRV indices were measured over 3 hours during a hyperinsulinemic-euglycemic clamp, avoiding the impact of confounding factors such as blood glucose fluctuations. A study limitation is the cross-sectional design, which does not provide insight into the temporal sequence of the observed abnormalities. Moreover, because genetic and epigenetic factors act as modifiers of NAFLD development and progression (40), their possible influence cannot be excluded.

In conclusion, we demonstrated that both lower cardiac parasympathetic activity and BRS are strongly and independently associated with increased hepatic fat content quantified by MRS in patients with recent-onset type 2 but not type 1 diabetes, suggesting that NAFLD may be relevant in the early development of parasympathetic CAN in type 2 diabetes. Large-scale controlled studies should ultimately determine whether favorable modulation of autonomic tone toward augmenting vagal activity accompanied by attenuated hepatic steatosis and insulin resistance can be translated into a reduction of cardiovascular end points in people with diabetes.

Figure 1.

Inverse associations of LF and VLF power spectrum, SD of differences between adjacent NN intervals, and xBRS with liver fat (%) in participants with type 2 diabetes.

Abbreviations:

 
• BMI

  body mass index

 
• BRS

  baroreflex sensitivity

 
• CAN

  cardiovascular autonomic neuropathy

 
• CVD

  cardiovascular disease

 
• DSPN

  diabetic sensorimotor polyneuropathy

 
• FFA

  free fatty acid

 
• HbA1c

  hemoglobin A1c

 
• HDL

  high-density lipoprotein

 
• HF

  high frequency

 
• HRV

  heart rate variability

 
• LF

  low frequency

 
• MRS

  magnetic resonance spectroscopy

 
• NAFLD

  nonalcoholic fatty liver disease

 
• NASH

  nonalcoholic steatohepatitis

 
• NN

  normal to normal

 
• SD

  standard deviation of differences between adjacent normal-to-normal intervals

 
• VLF

  very low frequency

 
• xBRS

  cross-correlation baroreflex sensitivity.

Acknowledgments

The German Diabetes Study (GDS) Group consists of A. E. Buyken, J. Eckel, G. Geerling, H. Al-Hasani, C. Herder, A. Icks, J. Kotzka, O. Kuss, E. Lammert, D. Markgraf, K. Müssig, W. Rathmann, J. Szendroedi, D. Ziegler, and M. Roden (speaker). We appreciate the voluntary contribution of all study participants. We also thank F. Battiato, M. Schroers-Teuber, J. Schubert, and the staff of the Clinical Research Center and Technical Laboratory of the German Diabetes Center for excellent technical assistance and taking care of the patients. We also thank the radiographers (N. Achterath and A. Nagel) for acquiring magnetic resonance imaging and MRS data.

Financial Support: This work was supported by the Ministry of Science and Research of the State of North Rhine–Westphalia and the German Federal Ministry of Health. This study was supported in part by a grant of the Federal Ministry for Research to the German Center for Diabetes Research and in part by a grant of the German Center for Diabetes Research. The funding sources had no influence on the design and conduct of this study; collection, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.

Clinical Trial Information: ClinicalTrial.gov registration no. NCT01055093 (registered 22 January 2010).

Author Contributions: D.Z. researched the data and wrote the article. A.S., Y.K., A.B, G.J.B., K.B., K.M., J.S., J.-H.H., and D.Z. researched data, contributed to the discussion, and reviewed and edited the article. D.F.M. performed laboratory analyses. M.R. contributed to the discussion and reviewed and edited the article. D.Z. designed the study. D.Z. and M.R. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Disclosure Summary: The authors have nothing to disclose.