Association of Nonalcoholic Fatty Liver Disease With Lower Brain Volume in Healthy Middle-aged Adults in the Framingham Study

Abstract

Importance

Nonalcoholic fatty liver disease (NAFLD) is a common condition that is most often asymptomatic. It is associated with metabolic syndrome, incident diabetes, carotid atherosclerosis, and endothelial dysfunction, conditions that in turn are strongly linked with brain damage and cognitive impairment. However, it is not known whether NAFLD is associated with structural brain measures in humans.

Objective

To assess the association between prevalent NAFLD and brain magnetic resonance imaging (MRI) measures.

Design, Setting, and Participants

The cross-sectional association between NAFLD and brain MRI measures was assessed from November 6, 2002, to March 16, 2011, in 766 individuals from the Offspring cohort of the Framingham Study. Participants were included if they did not have excessive alcohol intake and were free of stroke and dementia. Data analysis was conducted from December 30, 2015, to June 15, 2016.

Exposures

Nonalcoholic fatty liver disease was assessed by multidetector computed tomographic scans of the abdomen.

Main Outcomes and Measures

Linear or logistic regression models were used to evaluate the cross-sectional association between NAFLD and brain MRI measures, adjusting for age, sex, alcohol consumption, visceral adipose tissue, body mass index, menopausal status, systolic blood pressure, hypertension, current smoking, high-density lipoprotein and low-density lipoprotein cholesterol levels, lipid treatment, type 2 diabetes, cardiovascular disease, physical activity, insulin resistance, C-reactive protein levels, and plasma homocysteine values. Brain MRI measures included total cerebral brain volume, hippocampal and white matter hyperintensity volumes, and presence or absence of covert brain infarcts.

Results

Of the 766 individuals in the study sample (410 women and 356 men; mean [SD] age at the time of brain MRI, 67 [9] years), 137 (17.9%) had NAFLD. Nonalcoholic fatty liver disease was significantly associated with smaller total cerebral brain volume even after adjustment for all the covariates included in the study (β [SE], –0.26 [0.11]; P = .02). Differences in total cerebral brain volume between those with and without NAFLD corresponded to 4.2 years of brain aging in the general sample and to 7.3 years in individuals younger than 60 years of age. No statistically significant associations were observed between NAFLD and hippocampal or white matter hyperintensity volumes or covert brain infarcts.

Conclusions and Relevance

Nonalcoholic fatty liver disease is associated with a smaller total cerebral brain volume, independent of visceral adipose tissue and cardiometabolic risk factors, pointing to a possible link between hepatic steatosis and brain aging.

This cross-sectional analysis among participants from the Offspring cohort of the Framingham Study assesses the association between prevalent nonalcoholic fatty liver disease and brain magnetic resonance imaging measures.

Key Points

Question

Is nonalcoholic fatty liver disease associated with structural brain measures?

Findings

In this cross-sectional analysis among participants from the Offspring cohort of the Framingham Study, prevalent nonalcoholic fatty liver disease was linked with smaller total cerebral brain volumes. This finding remained statistically significant after controlling for health and lifestyle parameters, including visceral adipose tissue.

Meaning

Hepatic steatosis may have an independent association with brain aging; future studies are warranted to validate these findings and to test whether remission of nonalcoholic fatty liver disease has benefits in preserving brain health.

Introduction

Nonalcoholic fatty liver disease (NAFLD) is emerging globally as the major cause of chronic liver disease, with a prevalence of approximately 30% in the general population and 80% among morbidly obese individuals. Lifestyle factors, including unhealthy diet and lack of physical activity, play a key role in the onset of NAFLD and their modification is the standard treatment. Insulin resistance, type 2 diabetes, and obesity are the strongest risk factors for NAFLD. However, a reciprocal association is also evident, in which NAFLD may lead to type 2 diabetes and metabolic syndrome independently from body weight or even visceral adiposity.

In recent years, there has been a growing understanding that not only advanced forms of chronic liver disease but even precirrhotic stages may be linked with risk of cognitive impairment and dementia. Specifically, a link between NAFLD and brain health has been suggested. Liver steatosis may be linked with brain structure and function, at least partly, through shared risk factors including diabetes, obesity, and physical inactivity, which are established risk factors for dementia as well as for brain aging. However, other mechanisms may include insulin resistance, inflammation, hormonal alterations, endothelial dysfunction, and change in levels of secreted hepatokines. Accumulating evidence also suggests that NAFLD contributes directly to an increased risk of clinical and subclinical cardiovascular diseases, which in turn are strongly correlated with brain structure and function; however, distinct mechanisms may also be involved. Despite the abundance of literature suggesting biologically plausible links, the association of NAFLD with markers of brain aging has been rarely studied, to our knowledge. In animal models, induction of NAFLD was shown to be associated with brain damage. Moreover, emerging data in humans demonstrate associations between NAFLD and poor cognitive function as well as with decreased brain activity.

Structural magnetic resonance imaging (MRI) allows identification of anatomical abnormalities of brain morphologic and vascular changes. These measures correlate well with cognitive performance and can estimate future risk of dementia, stroke, and mortality. In the current study, our objective was to examine the cross-sectional associations between NAFLD and MRI measures of total cerebral brain volume (TCBV), hippocampal volume, white matter hyperintensity volume, and covert brain infarcts in middle-aged community-dwelling participants.

Methods

Study Sample

The study sample is based on participants from the Offspring cohort of the Framingham Heart Study. Between November 16, 2002, and April 29, 2005, a total of 1418 participants underwent multidetector computed tomographic (CT) scanning of the abdomen as part of an ancillary study. Inclusion criteria for that study were weighted toward individuals who still resided in the greater New England area. Minimum age cutoffs were 35 years in men and 40 years in women, and pregnant women and individuals weighing more than 160 kg were excluded. A total of 1409 Offspring cohort participants had CT scans that were interpretable for fatty liver. Data were obtained under a protocol approved by the Boston University Medical Center institutional review board and written informed consent was obtained from all participants.

Covariates including laboratory measurements were performed at the seventh examination cycle (September 14, 1998-October 4, 2001). We excluded individuals if they did not attend the seventh examination cycle when covariates were measured (n = 23), reported consuming more than 14 alcoholic drinks per week for men and more than 7 alcoholic drinks per week for women (n = 149), or had missing information on alcohol consumption (n = 3). Of the 1234 remaining eligible participants, 811 also underwent brain MRI, which was conducted between August 27, 2004, and March 16, 2011, subsequent to the abdominal CT scan. We additionally excluded participants with prevalent dementia (n = 15), prevalent stroke (n = 16), or other neurologic conditions that might interfere with interpretation of the MRI findings (n = 14), resulting in a final sample size of 766 Offspring participants.

Measurement of Fatty Liver

Multidetector CT scans were performed using 8-slice multidetector CT technology (LightSpeed Ultra, General Electric). A calibration phantom (Image Analysis) with a water-equivalent compound (CT-Water, Light Speed Ultra, General Electric) and calcium hydroxyapatite at 0, 75, and 150 mg/cm³ was placed under each participant. Three areas from the liver and 1 from an external phantom were measured, and the mean of the liver measures was then calculated and used to create liver to phantom ratios. Both intrareader and interreader reproducibility of the liver to phantom ratios were 0.99. Nonalcoholic fatty liver disease was defined as a liver to phantom ratio of 0.33 or less, consistent with prior Framingham Heart Study publications. Additional details on multidetector CT scan protocol and measurement of fatty liver can be found elsewhere.

Methods for Brain MRI Evaluation

Methods for segmentation and quantification of total cerebral brain, hippocampal, and white matter hyperintensity have been described previously. Briefly, participants were evaluated with a 1.5-T Siemens Avanto scanner (Siemens). Three-dimensional T1 and double echo proton density and T2 coronal images were acquired in 4-mm contiguous slices. All images were transferred to the centralized reading center at the University of California–Davis Medical Center and analyses were performed on QUANTA, version 6.2 (University of California–Davis Medical Center), a custom-designed image analysis package. Images were read centrally, masked to the participant’s identity, risk factors, and abdominal CT scan findings. Brain volume calculation was performed using semiautomated analysis of pixel distributions based on mathematical modeling of MRI pixel intensity histograms for cerebrospinal fluid and brain matter (white matter and gray matter) to determine the optimal threshold of pixel intensity to best distinguish cerebrospinal fluid from brain matter. Hippocampal volume was defined by the operator using automated traced boundaries. Total, hippocampal, and white matter hyperintensity volumes were expressed as a proportion of total cranial volume to correct for head size. Covert brain infarcts were defined as an area of abnormal signal intensity in a vascular distribution, based on a size of 3 mm or more, as well as location and imaging characteristics of the lesion.

Covariate Assessment

Educational achievement was defined as a 4-class variable (no high school degree, high school degree only, some college, and at least a college degree). Levels of fasting blood glucose, insulin, lipid markers, and serum alanine aminotransferase (ALT) and aspartate aminotransferase were measured in fasting morning blood samples. Type 2 diabetes was defined as a fasting blood glucose level of 126 mg/dL or more (to convert to millimoles per liter, multiply by 0.0555) or use of an antidiabetic therapy. History of cardiovascular disease (myocardial infarction, angina, coronary insufficiency, stroke, transient ischemic attack, intermittent claudication, or congestive heart failure) was assessed by a review panel. The physical activity index was calculated as a composite score based on information collected from a structured questionnaire. Body mass index was defined as weight in kilograms divided by height in meters squared. Women were considered menopausal if their menstrual periods had stopped for at least 1 year. When controlling for this variable, men were coded as nonmenopausal. Alcohol consumption was defined as number of drinks per week through a series of physician-administered questions. Participants were considered current smokers if they had smoked at least 1 cigarette per day in the year preceding the Framingham Heart Study examination. Volume of visceral adipose tissue was assessed using an 8-slice supine multidetector CT scan as previously described. Homeostatic model assessment of insulin resistance was defined as levels of fasting insulin (μIU/mL) × fasting glucose (mmol/L), and serum C-reactive protein (CRP) was measured using high-sensitivity assay. Values for total plasma homocysteine were measured using high-performance liquid chromatography with fluorometric detection.

Statistical Analysis

Statistical analysis was conducted from December 30, 2015, to June 15, 2016. Descriptive statistics were calculated for all variables, stratified by NAFLD status. Skewed variables were natural log transformed as necessary. Age and age-squared adjusted residuals were calculated for the MRI variables, which then were standardized to a mean of zero and an SD of 1 (z scores) to facilitate comparisons across the measures.

Linear regression models were constructed to calculate β coefficients and SEs for the association between NAFLD status and each continuous MRI measure. A logistic regression model was constructed to calculate odds ratios and 95% CIs for the association between NAFLD status and presence of covert brain infarcts (≥1 covert brain infarct vs none). Owing to numerous potential confounders, we created 3 adjustment models. The basic model included age, age-squared, sex, alcohol consumption, years between CT scan and covariate assessment, and years between CT scan and MRI measurement. Model 2 further adjusted for visceral adipose tissue, body mass index, and menopausal status. Model 3 additionally included the following vascular risk factors: hypertension, systolic blood pressure, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, lipid treatment, current smoking, history of type 2 diabetes, history of cardiovascular disease, physical activity index, homeostatic model assessment of insulin resistance, C-reactive protein, and homocysteine levels. We additionally examined whether the associations of NAFLD with brain MRI measures were modified by sex, type 2 diabetes, and impaired fasting glucose and ALT levels by including interaction terms in the regression model, adjusting for the covariates included in model 1.

We estimated the number of years necessary for individuals with NAFLD to achieve the same loss of TCBV (years of brain aging) as those without NAFLD. Calculation of years of brain aging was based on regressing TCBV onto age at MRI, where the slope is the estimate of the change in TCBV per year of age (β = –0.26 cm³/y of age). Mean TCBV specific to age groups (<60, 60-74, and ≥75 years) was calculated for those with and without NAFLD. For each age group, the difference between the mean TCBV among those with and without NAFLD divided by the β coefficient yields the years of brain aging due to NAFLD. Because the association between age and the MRI measures were nonlinear, the estimations of years of brain aging were performed by age groups, and model 1 included a term for age-squared.

Statistical analyses were performed using SAS, version 9.4 (SAS Institute Inc) and all statistical tests were 2-sided. P < .05 was used to indicate statistical significance, with the exception of interaction assessments, in which case P < .10 was considered statistically significant, owing to the low power of the test.

Results

The mean (SD) age of the participants at the brain MRI examination was 67 (9) years, 410 were women (53.5%), and 137 participants (17.9%) had NAFLD. Table 1 presents the characteristics of participants with and without NAFLD. Those with NAFLD were more likely to be men (78 [56.9%] vs 278 [44.2%]) and to have prevalent diabetes (19 [13.9%] vs 39 [6.2%]). Compared with healthy individuals, participants with NAFLD had higher mean [SD] body mass index (30.6 [5.5] vs 27.3 [4.7]), increased mean (SD) visceral adipose tissue volume (2818 [1009] vs 1842 [983] cm³), higher median serum triglyceride levels (149 vs 104 mg/dL [to convert to millimoles per liter, multiply by 0.0113), higher median C-reactive protein levels (3.05 vs 1.70 mg/L [to convert to nanomoles per liter, multiply by 9.524), higher median homocysteine levels (1.09 vs 1.00 mg/L [to convert to micromoles per liter, multiply by 7.397]), and higher median ALT levels (22.0 vs 18.0 U/L [to convert to microkatals per liter, multiply by 0.0167]), as well as increased insulin resistance (median homeostatic model assessment of insulin resistance, 3.82 vs 2.31 mmol/L x μU/mL) and lower mean (SD) levels of high-density lipoprotein cholesterol (47 [14] vs 53 [16] mg/dL [to convert to millimoles per liter, multiply by 0.0259]). The mean (SD) time between the abdominal CT scan and brain MRI examinations was slightly longer for those with NAFLD compared with those without NAFLD (3.2 [1.3] vs 3.0 [1.2] years).

Table 1. Study Sample Characteristics.

Characteristic	Participants, No. (%)a		P Value
	No NAFLD (n = 629)	NAFLD (n = 137)
Women	351 (55.8)	59 (43.1)	.007
Age at covariate measurement, mean (SD), y	59.5 (9.0)	59.5 (8.5)	.93
Age at CT scan, mean (SD) y	63.8 (8.9)	63.6 (8.6)	.88
Age at MRI, mean (SD) y	66.7 (8.9)	66.9 (8.4)	.88
Time between CT scan and covariate measurement, mean (SD), y	−4.2 (1.0)	−4.2 (1.0)	.56
Time between CT scan and MRI, mean (SD), y	3.0 (1.2)	3.2 (1.3)	.03
Educational level
<High school degree	12 (1.9)	4 (2.9)	.23
High school degree	147 (23.4)	33 (24.1)
Some college	186 (29.6)	50 (36.5)
≥College degree	284 (45.2)	50 (36.5)
Postmenopausalb	291 (82.9)	49 (83.1)	.98
Current smoker	47 (7.5)	12 (8.8)	.61
Type 2 diabetes	39 (6.2)	19 (13.9)	.002
History of cardiovascular disease	55 (8.7)	14 (10.2)	.58
Body mass index, mean (SD)c	27.3 (4.7)	30.6 (5.5)	<.001
Systolic blood pressure, mean (SD), mm Hg	124 (18)	126 (15)	.12
Hypertension	204 (32.4)	65 (47.4)	.001
Lipid treatment	109 (17.3)	20 (14.6)	.41
Total cholesterol, mean (SD), mg/dL	200 (36)	198 (34)	.63
LDL cholesterol, mean (SD), mg/dL	122 (32)	120 (31)	.55
HDL cholesterol, mean (SD), mg/dL	53 (16)	47 (14)	<.001
Triglycerides, median (25th-75th percentile), mg/dL	104 (73-152)	149 (101-203)	<.001
Alcohol, median (25th-75th percentile), No. drinks/wk	1.7 (0.0-3.3)	1.7 (0.0-5.0)	.23
Physical activity index score, mean (SD)	37.6 (6.2)	36.8 (5.7)	.13
Visceral adipose tissue, mean (SD), cm3	1841 (983)	2818 (1009)	<.001
HOMA-IR, median (25th-75th percentile), mmol/L × μU/mL	2.31 (1.71-3.36)	3.82 (2.65-5.61)	<.001
C-reactive protein, median (25th-75th percentile), mg/L	1.70 (0.81-3.72)	3.05 (1.38-5.90)	<.001
Homocysteine, median (25th-75th percentile), mg/L	1.00 (0.85-1.22)	1.09 (0.91-1.28)	.02
Alanine aminotransferase, median (25th-75th percentile), U/L	18.0 (15.0-25.0)	22.0 (16.0-34.0)	<.001
Aspartate aminotransferase, median (25th-75th percentile), U/L	21.0 (18.0-24.0)	22.0 (18.0-27.0)	.06
MRI brain measures
Total cerebral brain volume, mean (SD), %d	79.4 (3.7)	78.3 (3.6)	.002
Hippocampal volume, median (25th-75th percentile), %d	0.547 (0.051)	0.552 (0.056)	.20
White matter hyperintensities, median (25th-75th percentile), %d	0.09 (0.05-0.18)	0.10 (0.04-0.17)	.47
Covert brain infarct	85 (13.5)	20 (14.6)	.74

Abbreviations: CT, computed tomographic; HDL, high-density lipoprotein; HOMA-IR, homeostatic model assessment of insulin resistance; LDL, low-density lipoprotein; MRI, magnetic resonance imaging; NAFLD, nonalcoholic fatty liver disease.

SI conversion factors: To convert total cholesterol, HDL cholesterol, and LDL cholesterol to millimoles per liter, multiply by 0.0259; to convert triglycerides to millimoles per liter, multiply by 0.0113; to convert C-reactive protein to nanomoles per liter, multiply by 9.524; to convert homocysteine to micromoles per liter, multiply by 7.397; and to convert alanine aminotransferase and aspartate aminotransferase to microkatals per liter, multiply by 0.0167.

^aData are presented as number (percentage) of participants unless otherwise indicated.

^bDenominator is total number of women. Men were coded as nonmenopausal.

^cCalculated as weight in kilograms divided by height in meters squared.

^dExpressed as percentage of total cranial volume.

After adjustment for age, sex, alcohol consumption, and time between measurements, NAFLD was associated with smaller TCBV (β [SE], –0.34 [0.094]; P < .001) (Table 2). This association remained statistically significant even after additional adjustment for all potential confounders available in the study (β [SE], –0.26 [0.11]; P = .02). No statistically significant associations of NAFLD with hippocampal volume, white matter hyperintensity volumes, or covert brain infarcts were demonstrated. There were no statistically significant interactions between NAFLD and sex, diabetes, impaired fasting glucose levels, or serum ALT levels.

Table 2. Regression Results for the Association Between NAFLD and MRI Brain Measures.

MRI Brain Measurea	Modelb	No.	β (SE)	P Value
Total cerebral brain volumec	1	766	−0.34 (0.094)	<.001
	2	736	−0.28 (0.10)	.007
	3	688	−0.26 (0.11)	.02
Hippocampal volumec	1	762	0.15 (0.095)	.12
	2	732	0.21 (0.11)	.05
	3	684	0.21 (0.11)	.06
White matter hyperintensitiesc	1	766	−0.093 (0.095)	.33
	2	736	−0.051 (0.10)	.63
	3	688	−0.016 (0.11)	.89
Covert brain infarct (yes vs no)d	1	766	1.13 (0.65-1.94)e	.67
	2	736	1.64 (0.90-2.97)e	.10
	3	688	1.64 (0.85-3.16)e	.14

Abbreviations: MRI, magnetic resonance imaging; NAFLD, nonalcoholic fatty liver disease.

^aExpressed as SD increment of age and age-squared adjusted residuals.

^bModel 1 is adjusted for age at MRI, age-squared at MRI, sex, alcohol (drinks per week), years between computed tomographic scan and covariate assessment, and years between computed tomographic scan and MRI. Model 2 is adjusted for model 1 covariates plus visceral adipose tissue, body mass index, and menopausal status. Model 3 is adjusted for model 2 covariates plus systolic blood pressure, hypertension, levels for high-density lipoprotein cholesterol and low-density lipoprotein cholesterol, lipid treatment, current smoking, diabetes, history of cardiovascular disease, physical activity index, homeostatic model assessment of insulin resistance, and levels for C-reactive protein and homocysteine.

^cExpressed as percentage of total cranial volume.

^dCovert brain infarct models are not adjusted for menopausal status, owing to the lack of covert brain infarcts occurring in nonmenopausal women.

^eOdds ratio (95% CI).

Table 3 describes the estimated years of brain aging associated with NAFLD. In individuals younger than 60 years of age, the difference in TCBV between those with and without NAFLD corresponds to 7.3 years of brain aging. In older participants, the estimated number of years of brain aging was smaller, at 4.2 years for those 60 to 74 years of age and 1.5 years for those 75 years of age or older.

Table 3. Differences in TCBV and Correspondent Years of Brain Aging by Age Groups.

Age Group, y	No.	Mean TCBV, Adjusted for Age Group, % of Total Cranial Volume		Years of Brain Aging Due to NAFLDa
		No NAFLD	NAFLD
<60	174	82.3	80.4	7.3
60-74	429	79.7	78.6	4.2
≥75	163	75.6	75.2	1.5
All participants	766	79.4	78.3	4.2

Abbreviations: NAFLD, nonalcoholic fatty liver disease; TCBV, total cerebral brain volume.

^aCalculation of years of brain aging is based on regressing TCBV onto age at magnetic resonance imaging, where the slope is the estimate of the change in TCBV per year of age (β, –0.26 cm³/y of age). Mean TCBV specific to age groups (<60, 60-74, and ≥75 y) was calculated for those with and without NAFLD. For each age group, the difference between the mean TCBV among those with and without NAFLD divided by the β coefficient yields the years of brain aging due to NAFLD.

Discussion

Our study suggests that NAFLD may be associated with brain aging. Participants with NAFLD had smaller TCBV compared with those with a healthy liver, regardless of multiple cardiometabolic risk factors and visceral fat. These differences in TCBV corresponded to an estimated 4.2 years of brain aging in the overall sample and 7.3 years of brain aging in participants younger than 60 years of age.

Obesity and overweight have been previously associated with lower total and regional brain volumes. However, it is evident today that specific fat depots, rather than global adiposity, may have a direct role in various metabolic and vascular pathologic conditions. In terms of the effect of obesity on the brain, findings suggest that abdominal adiposity may be a more sensitive indicator compared with global obesity. Furthermore, increased visceral fat on CT assessment was linked to lower TCBV independently of body mass index and insulin resistance in a previous study among the Framingham Offspring cohort. Evidence shows that visceral and liver fat are correlated; however, these correlations have been estimated as not more than modest (ρ = –0.34) in the Framingham study sample, suggesting that the 2 fat components are distinct. Our study now expands previous findings by highlighting fat accumulation in the liver as a potential independent risk factor for brain aging.

The mechanisms underlying the association of NAFLD with brain atrophy are hard to elucidate, as NAFLD is considered a single component in a complex multisystem disease. The liver is a central regulator of systemic energy homeostasis and has a pivotal role in glucose and lipid metabolism. Apart from the association with TCBV suggested in the current study, the involvement of liver steatosis per se in various metabolic and vascular conditions is rapidly gaining acknowledgment. This involvement includes dyslipidemia and dysglycemia, type 2 diabetes and the metabolic syndrome, and cardiovascular disease and atherosclerosis. These health conditions are interrelated and have been linked with decreased TCBV. Although many of these cardiometabolic factors were controlled for in the current analyses, unknown factors may have affected the results.

Nonalcoholic fatty liver disease may be linked with cerebral neurodegeneration through microvascular structural alterations. In a recent study among more than 2200 participants, the association of NAFLD with the peripheral arterial tonometry ratio, a measure of microvascular dysfunction, was significant even after adjustment for vascular risk factors as well as of visceral adiposity. In another study, NAFLD was associated with distal artery calcification. Such injuries to brain microvasculature, although not directly examined in the context of NAFLD, are thought to reduce brain blood flow, subsequently leading to microvascular ischemia, quantifiable brain tissue damage, and atrophy.

Another explanation for the link between fat accumulation in the liver and global brain atrophy is mediation through subclinical inflammation. Liver adiposity has been shown to induce systemic subclinical inflammation and secretion of multiple cytokines, which in turn have been linked with decreased TCBV. In animal models, induction of NAFLD resulted in the development of significant liver inflammation, which coincided with increased numbers of activated microglial cells in the brain, an increased inflammatory cytokine profile, and neurodegeneration. Nevertheless, this mechanism is not restricted to NAFLD, as other fat depots such as visceral and subcutaneous adipose tissues also contribute to systemic inflammation.

Last, there is also a possible involvement of hepatokines in the association between liver fat and brain atrophy. Similar to adipokines (eg, leptin, resistin, and ghrelin), which are proteins secreted by adipose tissues, hepatokines are hormones derived exclusively or predominantly by the liver. Data show that these proteins contribute to the pathogenesis of cardiovascular diseases and stroke; however, future research is warranted to assess their possible contribution to brain structure and function.

Strengths and Limitations

The strength of our study is the large, well-characterized cohort of individuals with a wide variety of metabolic and lifestyle covariates measured in a standardized and validated manner. However, we acknowledge that our study has several limitations. First, this is an observational study with a cross-sectional design that does not permit the determination of a temporal sequence between hepatosteatosis and brain MRI measures. Hence, cause-and-effect inferences cannot be made. Second, covariates were measured at times separate from the CT scans and the latter were measured at a separate time from the MRI, with a slight difference between the NAFLD and healthy liver groups. Although these time gaps were controlled for in our models, we cannot exclude the possibility that cardiometabolic risk factors or end-stage liver disease emerged with higher frequencies in the interim period among patients with NAFLD. Third, we were unable to perform liver biopsies, considered the criterion standard for the diagnosis of NAFLD, in this cohort of healthy volunteers. However, such possible misclassification bias is nondifferential; therefore, it may result in underestimation of the observed association. The lack of liver biopsy as well as of noninvasive fibrosis markers limits any inference regarding the association between brain damage and liver inflammation or fibrosis. In using elevated levels of ALT as a marker for hepatocyte injury secondary to inflammation, we did not notice a significant interaction between ALT and NAFLD with regard to brain atrophy or other measures. Last, the study population was limited to individuals of European ancestry from 1 geographic area and of a relatively high socioeconomic status, which is reflected by a lower than average estimation of NAFLD prevalence; therefore, the external validity for other ethnicities and varying populations is uncertain.

Conclusions

This study shows that the diagnosis of NAFLD is associated with decreased TCBV, independently of visceral obesity and other cardiovascular risk factors. Furthermore, the fact that estimates for years of brain aging associated with NAFLD were larger in younger compared with older individuals may stress the importance of NAFLD in this age group, for which prevalence of NAFLD is substantially increasing owing to unhealthy lifestyles. These findings, together with previous literature, strongly suggest that NAFLD is not merely a liver disease with implications for hepatic functions but is also a key factor in the evolution of a complex metabolic process with significant extrahepatic implications, including for the brain. Exploration of the association between liver disease of other possible causes and brain MRI measures was beyond the scope of this study; therefore, it is uncertain whether our findings are specific to NAFLD. Furthermore, the study was underpowered at this time to assess whether NAFLD is also associated with an increased risk of clinical dementia, Alzheimer disease, and stroke, events that have been associated with total brain atrophy.

Nonalcoholic fatty liver disease is a common condition, has a relatively early age of onset, and most important, is modifiable. As such, it might be a risk marker or a modifiable risk factor for accelerated brain aging and its consequences. Encouraging data for reversible metabolic damage was recently provided by a prospective study indicating prevention of diabetes following remission of NAFLD. Thus, our results, if confirmed in prospective studies and clinical trials, may indicate that the prevention and treatment of NAFLD can bear an even greater extrahepatic benefit including perhaps in preserving brain function.