Importance of cardiometabolic risk factors in the association between nonalcoholic fatty liver disease and arterial stiffness in adolescents

Authors

  • Rae-Chi Huang,

    Corresponding author
    1. School of Paediatrics and Child Health, UWA, Nedlands, Australia
    • School of Medicine and Pharmacology, University of Western Australia (UWA), Nedlands, Australia
    Search for more papers by this author
  • Lawrence J. Beilin,

    1. School of Medicine and Pharmacology, University of Western Australia (UWA), Nedlands, Australia
    Search for more papers by this author
  • Oyekoya Ayonrinde,

    1. School of Medicine and Pharmacology, University of Western Australia (UWA), Nedlands, Australia
    2. Department of Gastroenterology, Fremantle Hospital, Fremantle, Australia
    3. Curtin Health Innovation Research Institute, Curtin University, Bentley, Australia
    Search for more papers by this author
  • Trevor A. Mori,

    1. School of Medicine and Pharmacology, University of Western Australia (UWA), Nedlands, Australia
    Search for more papers by this author
  • John K. Olynyk,

    1. School of Medicine and Pharmacology, University of Western Australia (UWA), Nedlands, Australia
    2. Department of Gastroenterology, Fremantle Hospital, Fremantle, Australia
    3. Curtin Health Innovation Research Institute, Curtin University, Bentley, Australia
    4. Institute for Immunology and Infectious Diseases, Murdoch University, Murdoch, Australia
    Search for more papers by this author
  • Sally Burrows,

    1. School of Medicine and Pharmacology, University of Western Australia (UWA), Nedlands, Australia
    Search for more papers by this author
  • Beth Hands,

    1. University of Notre Dame, Fremantle, Australia
    Search for more papers by this author
  • Leon A. Adams

    1. Department of Gastroenterology and Hepatology, Sir Charles Gairdner Hospital, Nedlands, Australia
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.

  • Supported by the National Health and Medical Research Council (NH&MRC) program grants (353514 and 634445) and project grant (403981), Gastroenterology Society of Australia Astra Zeneca Career Development Award (to L. Adams), Fremantle Hospital Medical Research Foundation Medical Research Grant, NH&MRC Medical Postgraduate Scholarship (Ayonrinde, 404166). J.K. Olynyk is the recipient of an NH&MRC Practitioner Fellowship (513761).

Address reprint requests to: Dr. Rae-Chi Huang, Medical Research Building, Level 4 Rear 50 Murray Street, GPO Box X2213, Perth WA 6847, Australia. E-mail: rae-chi.huang@uwa.edu.au; fax: +61 8 9224 0246.

Abstract

Nonalcoholic fatty liver disease (NAFLD) is the most common liver disease worldwide and is regarded as the hepatic manifestation of the metabolic syndrome. In adults, NAFLD is a determinant of arterial stiffness and cardiovascular risk, independent of the metabolic syndrome. Our aim was to ascertain if NAFLD is associated with arterial stiffness, independent of cardiometabolic factors in a population-based cohort of adolescents. The 17-year-olds (n = 964) from an Australian birth cohort had measures of anthropometry, blood pressure, fasting insulin, glucose, lipids, and NAFLD by ultrasound. Two-step cluster analysis identified youth at high metabolic risk. Measures of arterial stiffness (pulse wave velocity [PWV] and augmentation index corrected for heart rate [AI@75]) were obtained using applanation tonometry. The overall prevalence of NAFLD was 13.3%. The “high risk” metabolic cluster at age 17 years included 16% males and 19% females. Compared to “low risk,” the “high risk” cluster participants had greater waist circumference, triglycerides, insulin, systolic blood pressure, and lower high-density lipoprotein (HDL) cholesterol (all P < 0.0001). Those who had NAFLD but were not in the “high risk” metabolic cluster did not have increased PWV or AI@75. However, males and females who had NAFLD in the presence of the metabolic cluster had greater PWV (b = 0.20, 95% confidence interval [CI] 0.01 to 0.38, P = 0.037). Males who had NAFLD in the presence of the metabolic cluster had greater AI@75 (b = 6.3, 95% CI 1.9 to 10.7, P = 0.005). Conclusion: NAFLD is only associated with increased arterial stiffness in the presence of the “high risk” metabolic cluster. This suggests that arterial stiffness related to the presence of NAFLD is predicated on the presence of an adverse metabolic profile in adolescents. (Hepatology 2013;58:1306–1314)

Abbreviations
AI@75

augmentation index corrected for heart rate

AIx

aortic augmentation index

CVD

cardiovascular disease

HOMA-IR

homeostasis model for insulin resistance

hsCRP

high sensitivity C-reactive protein

NAFLD

nonalcoholic fatty liver disease

PWC170

physical work capacity at a heart rate of 170 beats per minute

PWV

pulse wave velocity

Nonalcoholic fatty liver disease (NAFLD) is the most prevalent liver condition worldwide, affecting between 15%-30% of adults and 13%-18% of adolescents.[1, 2] The development of NAFLD is intimately related to key components of the metabolic syndrome, namely, central obesity, disturbed lipid metabolism, and insulin resistance. Thus, NAFLD has been considered the hepatic component of the metabolic syndrome.[3] As the metabolic syndrome is a cluster of factors that predict cardiovascular disease (CVD), it is not surprising that CVD is the leading cause of death in patients with NAFLD, accounting for ∼30% of all deaths.[4, 5] In contrast, only a small proportion of subjects with NAFLD develop cirrhosis or die from complications of their liver disease.

Although CVD is the leading cause of death in subjects with NAFLD, it is less clear whether NAFLD increases CVD-related morbidity or mortality independently of traditional cardiovascular risk factors.[6, 7] Cross-sectional studies consistently demonstrate the magnitude of association between NAFLD and CVD is attenuated after adjustment for metabolic factors,[8] although a limited number of cohort studies suggest NAFLD remains an independent risk factor for incident CVD disease.[7, 9] As the prevalence of NAFLD in children and adolescents is approaching alarming levels, it is important to establish whether NAFLD independently increases the risk of future CVD in these individuals. It is established that atherosclerosis may begin in childhood,[10] and thus pediatric patients with NAFLD are potentially at life-long cardiovascular risk from an early age with greatest potential for early death related to CVD. Cardiovascular risk can be objectively and noninvasively quantified using applanation tonometry (SphygmoCor), which measures arterial stiffness. Arterial stiffness reflects atherosclerotic changes in the arterial system and predicts cardiovascular events and mortality independent of other traditional risk factors.[11] Therefore, we aimed to examine the association between NAFLD, the metabolic syndrome, and arterial stiffness in a well-defined community-based cohort of adolescents.

Participants and Methods

Study Population

The Western Australian Pregnancy Cohort (Raine) Study is a longitudinal pregnancy cohort study of 2,868 children born between 1989 and 1992 in Perth, Western Australia. Recruitment and follow-up of the cohort has been described in detail.[1, 12] Since birth, the cohort has undergone serial cross-sectional assessments with the current study performed when the cohort had reached the age of 17 years. At the time of this cohort review, the Raine cohort was representative of the Western Australia adolescent population.[13]

At age 17 years, 1,754 participants underwent assessment including (1) a detailed questionnaire regarding nutrition, physical activity, and drug and alcohol behavior; (2) anthropometric assessment including height, weight, body mass index (BMI), and skinfold thickness; (3) abdominal ultrasonography; (4) fasting serum biochemistry; (5) assessment of arterial stiffness using SphygmoCor[14]; and (6) assessment of aerobic fitness. Aerobic fitness was measured using a cycle ergometer to assess physical work capacity at a heart rate of 170 beats per minute (PWC170). The PWC170 test has been described and validated in adolescents elsewhere.[15, 16] Cigarette smoking was assessed by self-reported, confidential questionnaire administered to the 17-year-olds and was coded as having never smoked, smoked 10 cigarettes or less, or greater than 10 cigarettes over their lifetime. The study was conducted with appropriate institutional (King Edward Memorial Hospital and Princess Margaret Hospital for Children) ethics approval. Written informed consent was obtained from the primary caregiver and assent obtained from adolescents.

Arterial Stiffness Assessment and Blood Pressure Assessment

Blood pressure was measured by oscillometric sphygmomanometer (DINAMAP vital signs monitor 8100, DINAMAP XL vital signs monitor or DINAMAP ProCare 100) after resting 5 minutes and using the appropriate cuff size. Six readings were taken while the subject was supine, every 2 minutes for 10 minutes. The average value was calculated after excluding the first reading.

The Pulse Wave Analysis System (SCOR-Px) and SphygmoCor Pulse Wave System (SCOR-Vx) were used.[14] After 5 minutes rest, six blood pressure measurements were recorded in the supine position and three electrocardiogram (ECG) leads attached to left leg, right arm, and left arm. The tonometers were applied to two sites (the carotid artery and distal dorsalis pedis). Distance measurement was taken in millimeters between the manubrium sternum and the two sampling sites. Pulse wave analysis was collected from the supported radial artery with wrist facing upwards. Data were entered after the waveform was maintained for 10 seconds and the test repeated until at least two captures recorded with a quality index of more than 80. Pulse wave velocity (PWV) was calculated by dividing the distance between tonometers by the transit time of the arterial pulse wave. PWV increases as vessels become less compliant with early atherosclerosis and arterial stiffness. An Aortic Augmentation Index (AIx) was defined as the difference in the second and first systolic pressure peaks as a percentage of pulse pressure. AIx measures the central arterial reflected pressure wave occurring after distension of peripheral vessels with systole. When the peripheral arterial vessels are less elastic, the reflected pressure wave occurs earlier after cardiac ejection and augments systolic pressure.[11]

Assessment of NAFLD

The diagnosis of NAFLD required the presence of ultrasound-detected hepatic steatosis and daily alcohol consumption less than 10 grams for females and less than 20 grams for males.[1] Ultrasound was performed by trained sonographers using a Siemens Antares ultrasound machine with a CH 6-2 curved array probe (Sequoia, Siemens Medical Solutions, Mountain View, CA) according to a standardized protocol.[1] A single radiologist then interpreted standard images and determined the presence or absence of hepatic steatosis based on scoring of echotexture, deep attenuation, and vessel blurring according to the protocol described by Hamaguchi et al.,[17] which provides 92% sensitivity and 100% specificity for the histological diagnosis of fatty liver. Scores of 0-3, 0-2, and 0-1 were determined for liver echotexture, deep attenuation, and vessel blurring, respectively. The diagnosis of fatty liver required a total score of at least 2, including echotexture score of at least 1. Hepatic fatty infiltration (steatosis) severity was classified by total fatty liver score as 0-1 (no fatty liver), 2-3 (mild fatty liver), and 4-6 (moderate-severe fatty liver). The intraobserver reliability (Κ statistic) for hepatic steatosis was excellent at 0.78 (95% confidence interval [CI] 0.73-0.88).[18]

Biochemical Analysis

Venous blood samples were taken after an overnight fast. Serum insulin, glucose, triglycerides, total cholesterol, HDL-cholesterol, LDL-cholesterol, and high sensitivity C-reactive protein (hsCRP) were measured in the PathWest Laboratory at Royal Perth Hospital as described.[19] The homeostasis model for insulin resistance (HOMA-IR) was calculated by insulin [μU/mL] × glucose [mmol/L]/22.5).[20]

Metabolic Syndrome

This was defined using age-specific definitions from the International Diabetes Federation for those who were 16 years or older.[21]

Cardiometabolic Cluster

Two-step cluster analysis was undertaken to define those with high and low cardiometabolic risk without recourse to arbitrary cutoffs.[22] To avoid arbitrary assignment of adult derived cutoffs we have previously used cluster analysis to identify individuals with features akin to the metabolic syndrome at the 8-year-old,[19] 14-year-old,[23] and 17-year-old[22] follow-ups. In these previous studies in children we used cluster analysis to generate “natural groupings” of composite metabolic parameters that predict inflammation and disease risk. In this way, individuals in the same group (or cluster) share metabolic parameters that differ from those in the alternative group (or cluster). This technique uses a scalable cluster analysis algorithm designed to handle large datasets.[19, 23, 24] This algorithm allows the identification of subclusters within the population of 17-year-olds, such that both within-group similarities and between-group differences are maximized using log-likelihood distance. Cluster analysis was applied separately to males and females with the continuous measures of systolic blood pressure (SBP), logarithmically transformed HOMA-IR, triglycerides, and BMI at age 17 years. The two clusters identified were designated as “high” or “low” cardiometabolic risk. Using this method we have previously defined “high-risk metabolic clusters” that not only predict metabolic and cardiovascular risk factors, but also inflammatory markers at various ages in children.[23]

Statistical Analysis

We aimed to determine the effect of NAFLD on arterial stiffness, independent of traditional cardiometabolic risk. The metabolic and biochemical features of 17-year-old subjects were initially assessed according to the presence of NAFLD or cardiometabolic risk using one-way analysis of variance (ANOVA). Using multiple linear regression models with NAFLD and individual metabolic risk factors as covariates to explore this would be flawed by inevitable collinearity[25, 26]; as NAFLD and each individual traditional cardiometabolic risk factor are highly correlated potentially giving rise to incorrect estimates of association. The use of the metabolic cluster avoids the collinearity between traditional cardiovascular risk factors. We checked for the severity of the multicollinearity between the metabolic cluster and NAFLD in the models using the variance inflation factor, which was in all cases <2. As multicollinearity was not a significant issue between NAFLD and metabolic clusters, these covariates were included concurrently in regression models, with the outcomes being arterial stiffness measures (PWV and AI adjusted for HR = 75 [AI@75]). As sex, age, and smoking are related to arterial stiffness, they were considered for inclusion in these models. Four groups were also determined with combinations of either the presence or absence of NAFLD and high or low cardiometabolic risk to facilitate communication of the results. The reference group consisted of individuals with low cardiometabolic status and absent NAFLD. IBM SPSS v. 19 (Chicago, IL) was used for statistical analysis and significance was set at P < 0.05.

Results

Population Characteristics

Of the 1,754 subjects who attended the 17-year assessment, 1,053 had both blood pressure and fasting blood samples, 991 also had completed liver assessments performed, and 964 had complete data available for analysis. Of those participants (n = 964) with both NAFLD diagnosis and valid cluster analysis, the mean age was 16.5 ± 0.5 years with 53% male and 47% female. The overall prevalence of NAFLD was 13.3% and the overall proportion of subjects in the high metabolic risk cluster was 17.5% (16% males, 19% females). Table 1 shows the features of those with or without NAFLD and those at high or low cardiometabolic risk. As has been previously shown, compared to those within the low-risk cardiometabolic cluster, the high-risk cardiometabolic cluster was associated with greater SBP, BMI, waist, fasting triglycerides, and lower HDL for both males and females (all P < 0.001).[22] Similarly, those youths with NAFLD also had greater SBP, BMI, waist circumference, triglycerides, and lower HDL compared to those without NAFLD.[1] There were greater differences in lipids, insulin, and HOMA when categorized by metabolic cluster as opposed to categorization by diagnosis of NAFLD (Table 1).

Table 1. Metabolic and Biochemical Features of 17-Year-Old Subjects According to Presence of NAFLD, Cardiometabolic Risk
CovariateMalesFemales
NAFLDP ValueCardiometabolic ClusterP ValueNAFLDP ValueCardiometabolic ClusterP Value
++++
  1. Mean (SD) are shown where relevant.

N45455 43079 37877 37877 
Age (years)16.5 (0.5)16.6 (0.6)0.4016.5 (0.5)16.5 (0.5)0.4416.6 (0.5)16.5 (0.5)0.3416.5 (0.5)16.5 (0.6)0.91
BMI (kg/m2)21.9 (3.1)28.8 (5.9)<0.00121.5 (2.7)28.7 (4.9)<0.00122.2 (3.3)26.7 (6.1)<0.00121.6 (2.6)29.5 (4.9)<0.001
Waist Circumference (cm)78.3 (7.7)96.3 (15.9)<0.00177.5 (6.9)95.7 (13.8)<0.00175.7 (9.3)85.9 (15.0)<0.00174.4 (7.8)93.1 (12.8)<0.001
Height (cm)178.1 (7.2)178.9 (6.8)0.44178.3 (7.2)177.8 (7.0)0.60165.9 (6.7)166.9 (6.1)0.23166.1(6.7)165.9 (6.1)0.78
SBP (mm Hg)117.5 (9.3)122.3 (9.4)<0.001116.5 (8.7)125.8(9.4)<0.001107.9 (9.4)110.4 (8.8)0.038107.0 (8.2)114.7 (11.5)<0.001
DBP (mm Hg)58.0 (6.5)59.6 (6.8)0.1057.8 (6.5)60.3 (6.6)0.00259.1 (6.5)59.8 (6.2)0.4459.0 (6.2)60.5 (7.3)0.059
HR (beats/min)63.1 (9.3)65.9 (11.6)0.03862.3 (9.4)66.0(10.5)0.00967.4(8.7)68.7 (9.8)0.2166.9 (8.9)71.1 (8.0)<0.001
Fasting insulin8.1(6.0)13.3 (10.0)<0.0017.1 (4.5)17.0 (10.1)<0.0019.3 (9.2)15.0 (26.7)0.0017.8 (4.3)22.4 (29.6)<0.001
HOMA1.8 (1.4)3.0 (2.2)<0.0011.5 (1.1)3.8 (2.3)<0.0012.0 (2.2)3.3 (6.8)0.0021.6 (0.9)5.1 (7.5)<0.001
Fasting glucose4.8 (0.7)5.0 (0.5)0.0674.8 (0.7)5.0 (0.5)0.0354.6 (0.4)4.7 (0.4)0.804.6 (0.4)4.8 (0.4)<0.001
Triglycerides1.0 (0.6)1.2 (0.6)0.0110.9 (0.3)1.8 (0.9)<0.0011.0 (0.5)1.1 (0.6)0.0200.9 (0.4)1.4 (0.7)<0.001
LDL2.2 (0.6)2.3 (0.8)0.392.2 (0.6)2.5 (0.8)0.122.4 (0.6)2.5 (0.7)0.332.4 (0.6)2.5 (0.7)<0.001
HDL1.22 (0.24)1.08 (0.19)<0.0011.22 (0.24)1.09 (0.21)<0.0011.43 (0.31)1.31 (0.30)0.0021.44 (0.30)1.23 (0.31)<0.001
Measures of arterial stiffness
Pulse wave velocity (m/sec)6.6 (0.7)6.8 (0.9)0.116.6 (0.7)6.7 (0.8)0.176.3 (0.7)6.3 (0.6)0.336.2 (0.7)6.4 (0.6)0.027
Augmentation index97.9 (12.1)101.7 (11.3)0.03197.6 (11.9)102.2 (11.9)0.00299.6 (10.6)98.6 (10.0)0.4499.5 (10.7)99.2 (9.3)0.82
Augmentation index corrected for HR = 75−10.4 (10.7)−5.6 (11.4)0.003−10.8 (10.6)−4.8 (10.8)<0.001−6.7 (9.7)−7.2 (9.0)0.70−7.0 (9.6)−4.8 (9.3)0.12
% with NAFLD0100 10.6%56.4% 0100 11.6%44.2% 

There was, however, overlap between the NAFLD and metabolic cluster groups. Of males with NAFLD, 56% were classified with “high risk” metabolic risk cluster and 44% at “low risk” (chi-square P < 0.001). Of females with NAFLD, 44% were classified with the “high risk” cardiometabolic cluster and 56% at “low risk” of cardiometabolic cluster (chi-square P < 0.001).

As shown in Table 2, 454 (80%) of males and 378 (73%) of females had neither NAFLD nor cardiometabolic risk. Eight percent (5% males and 10% females) had NAFLD without cardiometabolic risk and 8% (9% males and 8% females) had NAFLD with cardiometabolic risk.

Table 2. Characteristics of the Four Exclusive Groups (With and Without NAFLD and High and Low Metabolic Cluster Risk)
 MalesFemales
NAFLD++++
Metabolic Risk++++
  1. Mean (SD) are shown separately by males and females.

  2. a

    P < 0.05 on one-way ANOVA.

  3. b

    P < 0.001 on one-way ANOVA.

N406244831334434434
BMI (kg/m2)21.3 (2.6)23.8 (3.5)26.2 (3.4)32.6 (4.3)b21.5 (2.5)22.5 (2.7)27.5 (3.9)32.1 (4.8)b
Waist (cm)77 (6)84 (9)89 (9)107 (13)b74.0 (7.7)76.7 (7.8)88.5 (11.1)99.5(12.4)b
Suprailiac skin folds (cm)10.4 (5.8)19.3 (9.5)20.0 (7.9)32.7 (6.3)b16.2 (6.1)20.7 (8.3)24.4 (8.1)32.3 (7.5)b
Height (cm)178.1 (7.3)180.3 (6.1)177.8 (6.9)177.8 (7.2)166.1 (6.8)166.7 (5.4)164.9 (5.3)167.2 (6.9)
Insulin (mmol/L)7.0 (4.5)7.1 (4.6)16.1 (9.9)18.2 (10.3)b7.4 (3.8)8.9 (4.4)19.2 (16.0)22.6 (38.9)b
HOMA1.5 (1.1)1.5 (1.1)3.6 (2.3)4.1 (2.3)b1.6 (0.9)1.8 (1.0)5.0 (5.1)5.1 (9.9)b
SBP (mmHg)116 (9)118 (7)126 (9)126 (10)b107 (8)107 (6)115 (12)114 (10)b
DBP (mmHg)58 (6)58 (7)60 (6)61 (7)a59 (6)60 (6)61 (8)60 (7)a
Triglycerides(mmol/L)0.9 (0.3)0.8 (0.3)1.9 (1.0)1.6 (0.6)b0.9 (0.4)1.0 (0.4)1.5 (0.8)1.4 (0.7)b
HDL(mmol/L)1.2 (0.2)1.2 (0.2)1.1 (0.2)1.0 (0.2)b1.4 (0.3)1.4 (0.3)1.3 (0.3)1.1 (0.3)b
hsCRP (mmol/L)1.9 (6.9)2.9 (7.0)3.3 (3.6)5.1 (6.7)a1.1 (3.1)2.3 (7.5)1.6 (3.5)2.9 (3.0)a
Pulse wave velocity (m/sec)6.6 (0.7)6.7 (0.7)6.7 (0.6)6.9 (1.0)6.2 (0.7)6.3 (0.7)6.5 (0.7)6.4 (0.6)
Augmentation index97.4 (11.9)99.9 (11.7)101.7 (12.6)103.2 (10.8)a99.4 (10.7)100.0 (10.6)101.1 (9.2)96.8 (9.1)
Augmentation index corrected for HR = 75−11.0 (10.6)−8.9 (11.5)−6.4 (10.5)−2.9 (10.7)b−7.1 (9.7)−7.5 (9.1)−3.9 (9.2)−6.8 (9.6)
% smoked in last 12 months4154466251416653

Based on IDF criteria, only 2.7% of individuals had the metabolic syndrome (5 females and 12 males with concurrent NAFLD and 7 females and 7 males without NAFLD). As the numbers were small in these groups, further modeling was not undertaken.

Effect of NAFLD and Cardiometabolic Risk on Arterial Stiffness

Effect of NAFLD and Cardiometabolic Risk on PWV

Males had greater PWV than females (6.7 ± 0.7 versus 6.3 ± 0.7 m/s, P < 0.001) and thus the association between PWV, NAFLD and the metabolic cluster was adjusted for sex (Fig. 1). PWV was not different in subjects with or without NAFLD or between females in the high- or low-risk metabolic cluster (Table 1).

Figure 1.

The effect of NAFLD and cardiometabolic risk on PWV. Predicted fitted values with 95% CI shown for the sex-adjusted model. Closed diamonds represent low-risk cluster and open diamonds represent high-risk metabolic cluster. The P values indicate the difference in the two groups in the sex-adjusted model.

Those with both NAFLD and high metabolic risk had increased PWV (Table 3, adjusted coefficient = 0.20, 95% CI 0.01 to 0.38, P = 0.037), compared to those with neither NAFLD nor high-risk metabolic cluster. Subjects with NAFLD and the low-risk cluster, and those without NAFLD but with the high-risk metabolic cluster did not have increased PWV (P = 0.15 and 0.60, respectively).

Table 3. Adjusted Linear Regression Models Showing the b-Coefficient, 95% Confidence Interval, and P Value Indicative of Significance
 Adjusted Model 1Adjusted Model 2
Arterial Stiffness OutcomeCovariatebSig95% CICovariatebSig95% CI
  1. Smoking refers to self-reported history of cigarette smoking asked confidentially of 17-year-olds. It is categorized as never smoked, 10 cigarettes or less, or greater than 10 cigarettes.

Pulse wave velocityMetabolic cluster0.110.114−0.02 to 0.25NAFLD-Cluster   
 NAFLD0.070.60−0.11 to 0.19NAFLD(+) Metabolic(-)0.120.15−0.04 to 0.28
 Sex (male)0.39<0.0010.30 to 0.48NAFLD(-) Metabolic(+)0.050.60−0.13 to 0.23
 PWC170−0.0020.004−0.003 to −0.001NAFLD(+) Metabolic(+)0.200.0370.01 to 0.38
     Sex (male)0.50<0.0010.38 to 0.62
     PWC170−0.0020.004−0.003 to −0.001
Augmentation index adjusted for HR = 75Metabolic cluster3.40.0011.4 to 5.4NAFLD-Cluster   
 NAFLD0.10.91−2.0 to 2.2NAFLD (-) Metabolic(-)
 Sex (male)−0.90.31−2.7 to0.8NAFLD(+) Metabolic(-)2.50.17−1.0 to 6.0
 Smoking   NAFLD(-) Metabolic(+)−0.20.90−3.6 to 3.2
 NeverNAFLD(+) Metabolic(+)0.0450.98−3.8 to 3.9
 <10 cigarettes0.40.63−1.2 to 2.0Sex (male)−1.70.076−3.7 to −0.18
 >10 cigarettes2.20.0160.4 to 4.0Smoking   
 PWC170−0.04<0.001−0.06 to-0.02Never
     <10 cigarettes0.30.710.5 to 4.0
     >10 cigarettes2.30.010.5 to 4.1
     Sex* NAFLD Met interaction   
     Sex* NAFLD(-) Metabolic(-)
     Sex* NAFLD(+) Metabolic(-) 0.31 
     Sex* NAFLD(-) Metabolic(+) 0.44 
     Sex* NAFLD(+) Metabolic(+) 0.033 
     PWC170−0.04<0.001−0.06 to −0.02

Effect of NAFLD and Cardiometabolic risk on Central Augmentation Index (AI@75)

AI@75 was higher in males with NAFLD and the high-risk metabolic cluster (P = 0.003, P = <0.001 respectively), but not in females (P = 0.12 and P = 0.70, respectively) (Table 1, Fig. 2). There was a strong positive effect of the metabolic cluster on AI@75 (adjusted coefficient = 3.4, 95% 1.4 to 5.4, P = 0.001). When the composite four groups (made up of high or low metabolic risk, and with or without NAFLD) were used in the model, a sex interaction was seen for those with both NAFLD and high metabolic risk compared to those who were disease/risk-free (P = 0.033). The estimated coefficient in males was 6.3 (95% CI 1.9 to 10.7, P = 0.005) compared to the estimated female coefficient of −0.1 (95% CI −3.6 to 3.6, P = 0.99). On univariate exploration, smoking and PWC170 were significant covariates, and when included in the model (Table 3) the associations of NAFLD/metabolic cluster with augmentation were unchanged.

Figure 2.

The effect of NAFLD and cardiometabolic risk on augmentation index standardized for heart rate = 75 bpm (AI@75). The predicted means and 95% CI of AI@75 in the adjusted models are shown.

Discussion

Our study demonstrated in a population-based cohort of nearly 1,000 17-year-olds that NAFLD was only a risk factor for arterial stiffness in the presence of cardiometabolic risk factors. Those with NAFLD, but with concurrent low cardiometabolic risk, did not have increased PWV or AI@75, suggesting that in young adults the increased arterial stiffness seen in association with NAFLD is mediated primarily through cardiometabolic dysfunction. Gender-specific relationships between NAFLD and augmentation index (AI@75) were observed. Only males had additive effects of NAFLD upon AI@75.

Studies of older adult subjects have generally shown additive effects of NAFLD above metabolic dysfunction for arterial stiffness.[27-29] In our study of younger participants, NAFLD was associated with increased PWV only in subjects with underlying metabolic risk factors and only had an additive effect in males with underlying metabolic risk factors for AI@75. These results are consistent with previous reports from the Western Australian Busselton Population Study in which serum alanine aminotransferase levels (as a surrogate marker of NAFLD) were not independently predictive of adverse cardiovascular outcomes in adults after correction for metabolic syndrome components.[8] Therefore, aggressive attention and management of cardiovascular risk factors may be important only in NAFLD patients who have high-risk metabolic features. Differences between our observations and those of others may be due to age-related increasing visceral adiposity[30] and atherosclerosis, or progression of NAFLD to nonalcoholic steatohepatitis (NASH) and/or cirrhosis.[31, 32]

Furthermore, in this age group our data show that arterial stiffness coexists with the high-risk cardiometabolic cluster. These individuals often had mean values of SBP, fasting glucose, and some lipids below designated “abnormal” cutoffs and some could be described as having subclinical evidence of cardiometabolic dysfunction. This has also been reported in other studies of older, albeit not elderly, adults[33] and healthy adolescents and young adults,[34] confirming that insulin resistance and the metabolic syndrome are independent predictors of arterial stiffness. The central and causative role of insulin resistance in development of arterial stiffness is supported by a randomized placebo-controlled trial administering metformin (an insulin sensitizer) to NAFLD patients over a 4-month period, resulting in lower PWV and AI in the metformin treatment group, as well improvement in glucose metabolism.[35]

It is not unexpected that different relationships were observed between NAFLD and AI and PWV, as the latter are measuring different aspects of vessel elasticity and hemodynamics.[36] PWV is calculated as the distance : transit time ratio and is expressed as meters per second. Greater velocity of blood passage is presumed to be due to reduced elasticity in large arteries. Young and old arterial walls differ in their viscoelastic properties; at any given pressure, young arteries show increasing wall stiffness towards the periphery compared to old arteries.[37] Thus, in young individuals it is not correct to assume that any reduction in elasticity is due to underlying atherosclerosis, as seen in elderly populations. AI (the difference between early and late pressure peaks divided by pulse pressure) is calculated by a computer algorithm derived from invasive pressure and flow data, and expressed as a percentage. The late pressure peak is due to the reflected wave from the periphery. Therefore, while arterial stiffness may play a role, other physiological factors such as cardiac load and peripheral artery stiffness will also affect this measurement.

We observed several differences between males and females. Similar to other studies, males had higher levels of PWV[38] and lower readings of augmentation index, as previously reported in adults[39] and children.[40] Tomiyama et al.[38] demonstrated that males had higher PWV than females until the age of 60 years. We also observed a sex difference in the relationship between NAFLD and AI@75. Males with NAFLD had a greater AI@75 compared to males without NAFLD. Females with NAFLD did not have evidence of increased AI@75. This may relate to the inappropriate higher vascular loading conditions that already exist in females. Females have larger reflected waves than males, partly because they are generally shorter, with closer distances between heart and reflecting sites,[41, 42] and have greater tapering of arteries between the aorta and periphery.[43] We have also previously found that males compared with females with NAFLD have greater visceral adiposity, higher aminotransaminase levels, and a more severe metabolic phenotype with higher systolic blood pressure, blood glucose levels, but lower adiponectin and HDL cholesterol.[1] Thus, the increased AI@75 levels seen in males with NAFLD and the high-risk metabolic cluster may be related to greater visceral adipose and adipocytokine disturbance which have been implicated in cardiovascular risk.

Some limitations of the current work include the ability to adjust for other potential confounders, such as diet, which are limited by the assumptions of regression modeling techniques. These factors are clearly on the causal pathway towards both cardiometabolic risk, atherosclerosis, and NAFLD and are frequently highly correlated. However, as we tested for these variables using univariate regression and few were apparently associated, most were omitted from final modeling, and significant issues of collinearity were avoided. We also were not able to control for family history of cardiovascular disease. Another limitation of this study is that arterial stiffness and NAFLD measurements are taken at the same time, hence cause and effect cannot be inferred.

Given that this is a large-scale study on healthy participants, ethical considerations restrict us from obtaining gold standard liver histology, so a noninvasive method of liver ultrasound was used.[17] However, we acknowledge that this a limitation, as ultrasound can result in false-negative results, particularly when less than a third of the liver parenchyma is infiltrated by fat.[44, 45] Thus, this limits the ability to correlate with the severity of liver injury and could potentially introduce false negatives.

In conclusion, NAFLD is only associated with increased arterial stiffness in the presence of the “high-risk” metabolic cluster. This suggests that arterial stiffness related to the presence of NAFLD is predicated on the presence of an adverse metabolic profile in adolescents, and that NAFLD per se is not an independent risk factor for arterial stiffness.

Ancillary