H.H.H. is an investigator for the Howard Hughes Medical Institute.
Liver Failure and Liver Disease
Prevalence of hepatic steatosis in an urban population in the United States: Impact of ethnicity
Article first published online: 24 NOV 2004
Copyright © 2004 American Association for the Study of Liver Diseases
Volume 40, Issue 6, pages 1387–1395, December 2004
How to Cite
Browning, J. D., Szczepaniak, L. S., Dobbins, R., Nuremberg, P., Horton, J. D., Cohen, J. C., Grundy, S. M. and Hobbs, H. H. (2004), Prevalence of hepatic steatosis in an urban population in the United States: Impact of ethnicity. Hepatology, 40: 1387–1395. doi: 10.1002/hep.20466
- Issue published online: 24 NOV 2004
- Article first published online: 24 NOV 2004
- Manuscript Accepted: 31 AUG 2004
- Manuscript Received: 29 JUN 2004
- Donald W. Reynolds Cardiovascular Clinical Research Center at Dallas
- National Institutes of Health. Grant Number: HL-20948
- postdoctoral fellowship award from the American Liver Foundation and the National Institutes of Health training grant. Grant Number: T32-DK-07745
Despite the increasing prevalence of nonalcoholic fatty liver disease (NAFLD), its pathogenesis and clinical significance remain poorly defined. In this study, we examined and compared the distribution of hepatic triglyceride content (HTGC) in 2,287 subjects from a multiethnic, population-based sample (32.1% white, 48.3% black, and 17.5% Hispanic) using proton magnetic resonance spectroscopy. HTGC varied over a wide range (0.0%-41.7%; median, 3.6%) in the population. Almost one third of the population had hepatic steatosis, and most subjects with hepatic steatosis had normal levels of serum alanine aminotransferase (79%). The frequency of hepatic steatosis varied significantly with ethnicity (45% in Hispanics; 33% in whites; 24% in blacks) and sex (42% in white men; 24% in white women). The higher prevalence of hepatic steatosis in Hispanics was due to the higher prevalence of obesity and insulin resistance in this ethnic group. However, the lower frequency of hepatic steatosis in blacks was not explained by ethnic differences in body mass index, insulin resistance, ethanol ingestion, or medication use. The prevalence of hepatic steatosis was greater in men than women among whites, but not in blacks or Hispanics. The ethnic differences in the frequency of hepatic steatosis in this study mirror those observed previously for NAFLD-related cirrhosis (Hispanics > whites > blacks). In conclusion, the significant ethnic and sex differences in the prevalence of hepatic steatosis documented in this study may have a profound impact on susceptibility to steatosis-related liver disease. (HEPATOLOGY 2004;40:1387–1395.)
Nonalcoholic fatty liver disease (NAFLD) is the most common cause of abnormal liver function tests among adults in the United States.1 The spectrum of NAFLD ranges from simple steatosis to steatohepatitis, which can progress to end-stage liver disease. Hepatic steatosis develops for a variety of reasons2 but is most commonly associated with obesity, insulin resistance, and hyperlipidemia,3–7 all of which are components of the metabolic syndrome.8 However, not all subjects with the metabolic syndrome develop hepatic steatosis, nor do all subjects with hepatic steatosis develop nonalcoholic steatohepatitis or cirrhosis.2, 7 Why only a subset of at-risk individuals develop steatosis and a smaller fraction ultimately develop steatohepatitis remains unclear. Day et al.9 have proposed a “two-hit” hypothesis, in which the development of hepatic steatosis constitutes the first hit and cellular events leading to hepatic inflammation constitute the second hit.
Previous studies of NAFLD in the general population have relied on qualitative or semiquantitative techniques to assess hepatic fat content. The histological grading system commonly used to estimate hepatic triglyceride content (HTGC)10 is subject to overestimation (due to processing artifacts) and underestimation (due to microvesicular steatosis).11, 12 This technique requires liver biopsy and is not typically used in healthy individuals. Radiographic techniques (e.g., ultrasonography, computed tomography, magnetic resonance imaging) have the advantage of being noninvasive, but none of these techniques can quantify the triglyceride content of liver. Furthermore, only computed tomography has been directly correlated with direct biochemical measurements of HTGC.13, 14 Most studies examining the prevalence of NAFLD have been small and/or have been performed in highly selected populations.3, 15–17 Thus, the true prevalence and distribution of NAFLD in the general population remains unknown.
Two recent retrospective studies suggest that the prevalence of cirrhosis attributable to NAFLD may differ among ethnic groups.18, 19 To determine if the ethnic differences in NAFLD-related cirrhosis are associated with underlying differences in susceptibility to hepatic triglyceride accumulation, we compared the distribution of HTGC and the frequency of hepatic steatosis in the three major ethnic groups in the United States: whites, blacks, and Hispanics. We used proton nuclear magnetic resonance spectroscopy (1H-MRS),20 a sensitive and quantitative noninvasive technique, to measure HTGC in a large, ethnically diverse, probability-based population sample from Dallas, Texas.21
Patients and Methods
The study population included all participants in the Dallas Heart Study (DHS) in whom HTGC was measured with 1H-MRS. The DHS is a multiethnic, population-based probability sample of Dallas County, Texas, weighted to include 50% black and 50% nonblack subjects.21 A total of 2,971 subjects completed a clinic visit and, of these subjects, 2,349 underwent 1H-MRS of the liver to quantify HTGC. Some participants failed to complete a 1H-MRS study for the following reasons: claustrophobia (n = 191), medical contraindications (n = 49), equipment failure (n = 19), refusal (n = 74), and scheduling conflicts (n = 289). Because of the weight limitations of the Gyroscan Intera table (Philips Medical Systems, Eindhoven, the Netherlands), individuals with extreme obesity (>320 lbs.) were excluded (n = 58); consequently, the prevalence of obesity was slightly lower in the study group (43%) than in the DHS clinic cohort (47%).
Sampling weights reflecting the different probabilities of selection and sample attrition at each step of the study were constructed to generate unbiased estimates of population frequencies. Weighted population estimates for measured variables agreed closely with those of the United States Census and were relatively stable from the first interview sample to the final clinic-based sample.21
Each participant completed a 60-minute structured questionnaire that was designed to obtain detailed data regarding the demographics, medical history, comorbid conditions, and ethanol intake of each subject. Alcohol consumption (g/d) was determined from responses to previously validated questions.22
The ethnicity of each subject was self-assigned according to categories used in the National Health and Nutrition Examination Survey.23 Blood pressure was determined from five measurements using an automatic oscillometric device while subjects were in the seated position (Welch Allyn, Series #52,000, Arden, NC).21 Body weight was measured with a portable scale; height, waist circumference, and hip circumference were determined by trained personnel using standard tape measures.
The study was approved by the institutional review board of University of Texas Southwestern Medical Center, and all subjects provided written informed consent prior to participation. Because of ethical concerns raised by the institutional review board, subjects were not screened for any infectious diseases, including viral hepatitis.
All study participants underwent phlebotomy after an overnight fast. A total of 40 mL of blood was collected in tubes containing a serum separator or citrate–ethylenediaminetetraacetic acid and maintained at 4°C for less than 4 hours prior to processing. The tubes were centrifuged (1,000g for 15 minutes at 4°C) and plasma was isolated. Serum chemistries were performed within 24 hours. Fasting serum insulin levels were measured by a Radio Immuno-Assay (RIA) assessed with an ICN Micromedic gamma counter from Linco Research Incorporated (St. Charles, MO). The definition of elevated alanine aminotransferase (ALT) levels published by the National Center for Health Statistics (>40 U/L in men and >31 U/L in women) was used.1, 23 Homeostasis model assessment (IRHOMA) was calculated from fasting values of insulin and glucose.24 Insulin resistance was defined as the top quartile of IRHOMA in nondiabetic, normoglycemic DHS subjects.25
1H-NMR Spectroscopy of Liver Triglyceride Content.
Localized 1H-NMR spectra of the liver were acquired with subjects in the supine position using a 1.5T Gyroscan Intera MR system (Philips Medical Systems, The Netherlands).20 Sagittal, coronal, and axial slices through the right lobe of the liver were acquired, and a 27-cm3 volume of interest was positioned, avoiding major blood vessels, intrahepatic bile ducts, and the lateral margins of the liver. After the system was tuned and shimmed, spectra were collected using a Q-body coil for radio frequency transmission and signal reception and a double-echo point-resolved spectroscopy (PRESS) sequence with an interpulse delay of Tr = 3 seconds and an echo time of Te = 25 milliseconds. All data were processed using automated procedures implemented on a Philips console to integrate water and triglyceride signals over the intervals of 3.0 to 5.5 ppm and 0.5 to 3.0 ppm, respectively (Fig. 1A). HTGC was calculated as a ratio of methylene and combined methylene and water signals corrected for spin–spin relaxation and is expressed as weight percent (g triglyceride per 100 g wet liver tissue) via methods previously validated in humans and animals.14, 26 Of the 2,349 1H-MRS measures obtained, 2,287 subjects had spectra of sufficient quality (i.e., without significant motion artifact) to determine HTGC.
Statistical analyses were performed using SigmaStat 3.0 software (SPSS, Inc., Chicago, IL). Differences between two groups were evaluated using unpaired t tests (means) or Mann-Whitney rank sum tests (medians). One-way ANOVA or ANOVA based on ranks, followed by multiple pairwise comparisons, was used for multigroup comparisons. Chi-square analysis was used to compare proportions between groups, and the Yates' correction was used in the comparison of fewer than three groups. Correlations between variables were determined using the Spearman rank test. A P value of less than .05 was considered statistically significant.
HTGC was measured in 2,287 participants of the DHS, including 734 whites, 1,105 blacks, 401 Hispanics, and 47 individuals of other ethnicities (Table 1). The Hispanics in the sample were significantly younger than the whites and blacks, reflecting the demographics of Dallas County (Census 2000) (see Table 1). Hepatic steatosis was previously shown to be associated with the metabolic syndrome.8 Therefore, we first compared the prevalence of the various components of the metabolic syndrome in the three ethnic groups. Mean body mass index (BMI) was lower in white women than in black and Hispanic women. Blacks had significantly higher systolic and diastolic blood pressures, lower plasma levels of triglyceride, and higher high-density lipoprotein cholesterol levels than the other ethnic groups. The prevalence of obesity, diabetes/impaired fasting glucose, and insulin resistance (IRHOMA) was similar in blacks and Hispanics and was significantly lower in whites. The prevalence of elevated serum ALT levels was significantly higher in Hispanics than in whites or blacks. Self-reported ethanol intake was generally low, with a higher proportion of blacks and Hispanics abstaining from ethanol use compared with whites.
|All (n = 2,240)||Men (n = 1,046)||Women (n = 1,194)|
|White (n = 734)||Black (n = 1105)||Hispanic (n = 401)||White (n = 375)||Black (n = 499)||Hispanic (n = 172)||White (n = 359)||Black (n = 606)||Hispanic (n = 229)|
|Age (yr)||46 ± 9||46 ± 10||41 ± 9†||45 ± 9||46 ± 10||41 ± 9†||47 ± 10||46 ± 10||41 ± 9†|
|Body mass index (kg/m2)||29 ± 6†||31 ± 8||30 ± 7||29 ± 5||29 ± 7||29 ± 5||28 ± 7†||33 ± 8||31 ± 8|
|Waist circumference (cm)||96 ± 7||100 ± 17†||96 ± 15||102 ± 14||100 ± 16||99 ± 11||91 ± 16||100 ± 17†||94 ± 17|
|Systolic blood pressure (mm Hg)||124 ± 14†||131 ± 19†||121 ± 15†||128 ± 13†||132 ± 18†||124 ± 13†||121 ± 14†||130 ± 19†||118 ± 16†|
|Diastolic blood pressure (mm Hg)||78 ± 9†||80 ± 10†||75 ± 9†||79 ± 9||79 ± 10||76 ± 9†||76 ± 8||80 ± 10†||74 ± 9|
|Fasting glucose (mg/dL)||95 ± 24||104 ± 47||106 ± 42†||96 ± 23||108 ± 55||105 ± 35†||94 ± 25||102 ± 38||107 ± 47†|
|Triglyceride (mg/dL)||139 ± 109||107 ± 98†||152 ± 142||157 ± 125||118 ± 123†||168 ± 131||119 ± 86||98 ± 71†||140 ± 149|
|HDL cholesterol (mg/dL)||49 ± 15||53 ± 16†||46 ± 11||42 ± 10||50 ± 16†||42 ± 10||55 ± 16||55 ± 15||49 ± 12†|
|AST (U/L)||24 ± 17||25 ± 19||27 ± 42||25 ± 11||30 ± 24||38 ± 62||23 ± 22||21 ± 11||22 ± 11|
|ALT (U/L)||24 ± 16||23 ± 16||29 ± 28†||29 ± 18||29 ± 18||39 ± 38†||19 ± 11||18 ± 11||21 ± 14†|
|Obesity (BMI > 30 kg/m2)||34†||48||48†||34†||38†||46†||33†||57||50†|
|DM and/or IFG (glucose > 110 mg/dL)||9†||20||20†||11†||21†||22†||7†||19||18†|
|Insulin resistance (IRHOMA > 4.04)||23†||38||37†||28†||36†||37†||20†||38||39†|
|Elevated ALT (M > 40 U/L; W > 31 U/L)||12†||12||23†||14†||15†||26†||7†||7||15†|
|Moderate (M < 30 g/d; W < 20 g/d)||76†||57||60†||76†||61†||72†||76†||53||50†|
|Excessive (M > 30 g/d; W > 20 g/d)||6†||8†||3†||6†||13†||3†||6†||3||2†|
Proton magnetic resonance spectra representative of 3 subjects are shown in Fig. 1A. The distribution of HTGC was positively skewed, with a median value of 3.6% (interquartile range, 2.1%-6.6%). The median HTGC for Hispanics (4.6%) was significantly higher than for whites (3.6%) or blacks (3.2%) (P < .001) (Fig. 1B, Table 2). Among men, Hispanics and whites had a significantly higher median HTGC (4.6% and 4.4%, respectively) than blacks (3.2%). Hispanic women had a significantly higher median HTGC (4.6%) than either black (3.3%) or white (3.0%) women.
|Ethnicity||n||Hepatic Triglyceride Content, %||Hepatic Steatosis, %|
The upper limit of normal for HTGC was determined in this population by examining the distribution of HTGC in 345 study participants who did not have secondary causes of hepatic steatosis, including increased adiposity (BMI > 25), glucose intolerance, or excessive alcohol use, and had normal liver function tests.20 Hepatic steatosis was defined as a HTGC greater than 5.5%, which corresponds to the 95th percentile in this low-risk subgroup.20, 27 Participants were stratified based upon the presence or absence of hepatic steatosis (HTGC > 5.5% and < 5.5%, respectively). Among the 2,287 DHS subjects, 708 (31%) had hepatic steatosis (see Table 2). Correction of this value for population sampling indicated an overall prevalence of hepatic steatosis in Dallas County of 34%.21
Striking differences in the prevalence of hepatic steatosis were present among the three major ethnic groups. Compared with whites, the prevalence of hepatic steatosis was significantly higher in Hispanics and significantly lower in blacks (see Table 2). The prevalence of hepatic steatosis was similar between the sexes in blacks and Hispanics, whereas in whites, men had an approximately 2-fold higher prevalence of hepatic steatosis than women. The demographic and clinical characteristics of subjects with and without hepatic steatosis are presented in Table 3. As has been reported previously, individuals with hepatic steatosis were significantly more likely to be obese, diabetic, or hypertriglyceridemic; have low high-density lipoprotein cholesterol levels; and meet criteria for the metabolic syndrome.8
|Characteristic||Normal Hepatic Triglyceride Content (<5.5%)(n = 1,579)||Elevated Hepatic Triglyceride Content (>5.5%)(n = 708)||P Value|
|Age (yr)||45 ± 9||46 ± 10||.003|
|Sex (M/F ratio)||0.8:1||1.1:1||.003|
|Obesity (%) BMI > 30 kg/m2||33||67||<.001|
|DM and/or IFG (%) Glucose > 110 mg/dL||11||18||<.001|
|Insulin resistance IRHOMA ≥ 4.04 (%)||23||58||<.001|
|Lipid abnormalities (%)TG > 150 mg/dL; HDL: M < 40 mg/dL, F < 50||40||64||<.001|
|Metabolic syndrome* (%)||8||30||<.001|
|Elevated ALT (%) M > 40 U/L; W > 31 U/L||9||21||<.001|
|Ethanol use (g/d)||6.3 ± 15.6||6.6 ± 15.3||NS|
HTGC correlated significantly with components of the metabolic syndrome, including BMI and waist circumference (Table 4). The correlation coefficient for BMI and HTGC was highest in Hispanics and lowest in blacks (Fig. 2A; see Table 4). A strong correlation was also present between IRHOMA and HTGC in all subgroups. Serum levels of ALT, but not aspartate aminotransferase, also correlated significantly with HTGC in all subgroups, but the correlation coefficient was highest in Hispanics (Fig. 2B; see Table 4). No significant positive correlation was found between daily ethanol intake and HTGC.
|Parameter||Men (n = 1,046)||Women (n = 1,194)|
|White (n = 375)||Black (n = 499)||Hispanic (n = 172)||White (n = 359)||Black (n = 606)||Hispanic (n = 229)|
|Body mass index||0.550†||0.436†||0.567†||0.519†||0.402†||0.573†|
|Systolic blood pressure||0.330†||0.161†||0.191*||0.286†||0.197†||0.343†|
|Diastolic blood pressure||0.386†||0.174†||0.258†||0.254†||0.165†||0.340†|
To determine if the observed ethnic and sex differences in the prevalence of hepatic steatosis were due to differences in the frequency of risk factors for NAFLD, we performed subgroup analyses. First, we examined the prevalence of hepatic steatosis in those subjects who were nonobese (BMI < 30 kg/m2) and in those subjects who were insulin-sensitive (IRHOMA < 4.04). The prevalence of hepatic steatosis was not significantly different in nonobese whites (n = 570) and nonobese Hispanics (n = 207) (20% vs. 26%; P = .118). Similarly, the proportion of Hispanics who did not have insulin resistance (IRHOMA > 4.04) and had hepatic steatosis (62 of 243; 26%) was similar to that seen in the insulin-sensitive whites (126 of 543; 23%; P = .653). In contrast, the prevalence of hepatic steatosis in nonobese blacks (65 of 570; 11%) or in insulin-sensitive blacks (87 of 655; 13%) was significantly lower than in their white and Hispanic counterparts (P < .001). The sex difference in prevalence of hepatic steatosis in whites (i.e., men > women) was also not attributable to differences in body weight or insulin sensitivity. Nonobese white men had a higher prevalence of hepatic steatosis (66 of 246; 27%) than nonobese white women (31 of 241; 13%; P < .001). Likewise, among white subjects who were insulin-sensitive, men were more likely (79 of 268; 29%) than women (48 of 283; 17%) to have hepatic steatosis (P < .001).
The ethnic differences in prevalence of hepatic steatosis were not attributable to differences in ethanol intake; the ethnic-specific frequency of hepatic steatosis in those subjects who did not consume alcohol was similar to that of the entire sample (whites, n = 132, 36%; blacks, n = 386, 25%; Hispanics, n = 152, 54%) (see Table 2).
Although no significant differences were found in the overall prevalence of hepatic steatosis in subjects who abstained from ethanol (221 of 694; 32%) versus those who reported moderate ethanol intake (434 of 1439; 30%) (P = .459), alcohol use was associated with sex differences in hepatic triglyceride accumulation. Men who reported a moderate intake of ethanol had a higher prevalence of hepatic steatosis than those who did not drink (36% vs. 29%; P = .047). Conversely, in women, moderate ethanol use was associated with a lower prevalence of hepatic steatosis (24% vs. 34%; P < .001). These differences were significant only in the white subgroup; white men who reported moderate ethanol intake (n = 284) had a significantly higher prevalence of hepatic steatosis (42%) than their female counterparts (n = 271; 20%) (P = .030).
Subjects with hepatic steatosis had a significantly higher prevalence of elevated ALT levels when compared with subjects who had normal HTGC (21% vs. 9%, respectively; P < .001), even after restricting the analysis to the nonobese (22% vs. 8%; P < .001) or insulin-sensitive subjects (16% vs. 8%; P < .001). However, the vast majority of subjects with elevated HTGC had normal serum ALT levels (79%) (see Table 3). Although the prevalence of an elevated ALT level was higher in Hispanics than in the general population (see Table 1), there were no ethnic differences in the prevalence of elevated ALT levels in subjects without hepatic steatosis (P = .705) or among the subjects with hepatic steatosis who were not obese (P = .740) or were insulin-sensitive (P = .057).
The ethnic and sex differences in the prevalence of hepatic steatosis persisted after controlling for medications known to cause hepatic steatosis (data not shown).28
In this study, HTGC was measured in a large, ethnically diverse population (n = 2,287) to determine and compare the prevalence of hepatic steatosis in three major ethnic groups: whites, blacks, and Hispanics. This is the largest study to date to use 1H-MRS, a highly sensitive and quantitative technique, to determine the prevalence of hepatic steatosis in a population-based sample. A major finding of this study was that hepatic steatosis was present in nearly one third (31%) of the sample, corresponding to a 34% prevalence of hepatic steatosis in Dallas County, Texas. Furthermore, the prevalence of hepatic steatosis varied significantly among ethnic groups: Hispanics had a significantly higher prevalence than whites, while blacks a significantly lower prevalence than whites. The higher prevalence of hepatic steatosis in Hispanics was positively correlated with obesity and insulin resistance, whereas the lower prevalence of hepatic steatosis in blacks was not attributable to these factors, or to ethnic differences in ethanol intake. In whites, but not in Hispanics or blacks, the frequency of hepatic steatosis was approximately 2-fold higher in men than in women. This sex difference was not due to differences in the frequency of obesity or insulin resistance; rather, it was related to ethanol intake. Moderate alcohol intake was associated with an increase in prevalence of hepatic steatosis in men and a decrease in women. Finally, although hepatic steatosis was associated with a higher prevalence of elevated serum ALT levels, the vast majority of subjects with excess hepatic triglyceride had normal serum aminotransferase levels.
The current standard used to assess HTGC is based on histomorphmetric analysis of liver biopsy specimens.10 This method has several disadvantages, including the clinical risks associated with liver biopsy, sampling error,29 and possible underestimation of the true HTGC.14 Noninvasive imaging modalities such as ultrasonography, computed tomography, and magnetic resonance imaging can be used to detect hepatic triglyceride, but only if there is moderate to severe fatty infiltration (>33% of hepatocytes affected).13 Furthermore, these noninvasive methods are not quantitative. In contrast, 1H-MRS provides a safe, quantitative assessment of hepatic triglyceride by directly measuring protons in acyl groups of liver tissue triglycerides; it also samples a much larger liver volume than can be obtained through routine liver biopsy (≈27 g vs. ≈75 mg). The values obtained using 1H-MRS correlate well with histomorphmetric analysis.14, 26, 29, 30 To date, 1H-MRS has been used in small numbers of selected subjects.31, 32
Most prior studies assessing the prevalence of hepatic steatosis have been performed in either clinic- or hospital-based populations, thus introducing selection and ascertainment bias.33 Ultrasonography has been used in two studies to assess the prevalence of hepatic steatosis in the general population. The largest of these studies (n = 2,574 subjects) was performed 16 years ago in Japan.34 The prevalence of hepatic steatosis in this study was 14%. A second study using ultrasonography was performed in Northern Italy (The Dionysus Study).17 In this study, the prevalence of hepatic steatosis in the “control” population (nonobese nondrinkers) was 16.4%.17 The low prevalence of hepatic steatosis found in these two studies likely reflects the low frequency or absence of obese or diabetic subjects. The prevalence of hepatic steatosis in these two samples is similar to the 16.7% prevalence found in the nonobese DHS participants who consumed no alcohol (n = 341).
The higher prevalence of hepatic steatosis in Hispanics and the low prevalence in blacks are due to different factors. The higher prevalence of obesity and insulin resistance in Hispanics accounted for the higher prevalence of hepatic steatosis in this population. In contrast, the lower prevalence of hepatic steatosis in blacks was not due to a lower frequency of associated risk factors. Blacks and Hispanics had a similar prevalence of both obesity and insulin resistance, yet the prevalence of hepatic steatosis was dramatically lower in blacks (24% vs. 45%). The prevalence of hypertriglyceridemia and low plasma levels of high-density lipoprotein cholesterol was also significantly lower among blacks (see Table 2), suggesting that the difference in prevalence of hepatic steatosis in blacks likely reflects more fundamental ethnic differences in lipid homeostasis.
Whites were the only major ethnic group in which there was a significant difference in the prevalence of hepatic steatosis in men and women (42% vs. 24%, respectively). This sex difference in hepatic steatosis prevalence was only present among those subjects who consumed moderate amounts of alcohol. Ethanol ingestion was associated with a significant increase in hepatic steatosis in men and a decrease in women. Thus, moderate ethanol use may protect against excess hepatic triglyceride accumulation in women. This finding is consistent with previous data from a predominantly female population with severe obesity showing that light to moderate ethanol consumption reduced the risk of hepatic steatosis.35
The reduced prevalence of hepatic steatosis in women who drink alcohol may result from an ethanol-associated increase in insulin sensitivity.36, 37 Ethanol intake was associated with increased insulin sensitivity in both men and women in this study (data not shown). Although hepatic metabolism of ethanol is known to be sexually dimorphic,38 we do not know why the prevalence of hepatic steatosis was lower only in women with moderate ethanol intake and not men. Presumably, hepatic triglyceride accumulation associated with ethanol use39 outweighs any beneficial effects of alcohol on insulin sensitivity in men.
The present study provides further evidence that a normal serum ALT level provides little diagnostic or prognostic value when assessing patients for NAFLD, because almost four fifths (79%) of the subjects with hepatic steatosis had normal serum ALT levels. Even when more stringent criteria were used (ALT >30 U/L in men and >19 U/L in women),27 most subjects with hepatic steatosis (54%) had normal serum ALT levels. Therefore, serum ALT levels appear to be an insensitive marker for both hepatic steatosis and nonalcoholic steatohepatitis.40
Interpretation of the present study is subject to certain limitations. The DHS was a multistep study that introduced the possibility of dropout at every stage of the study. It is possible that heavy ethanol users were underrepresented in the study. Recent estimates of frequency of heavy ethanol intake from the 2000 National Household Survey on Drug Abuse (whites, 6.6%; blacks, 4.5%; Hispanics, 4.7%)41 are similar to those seen in the DHS (see Table 1). However, others have estimated the frequency of heavy ethanol intake in the United States to be as high as 18%.41 Ethanol intake may also be underestimated because we relied on self-reporting. Only 148 subjects reported excess ethanol use and only 12 subjects reported the use of ethanol in quantities associated with alcoholic liver disease (>80 g/d).42
A second limitation of the study is that because we could not screen for infectious diseases, we were unable to evaluate the potential impact of hepatitis C virus (HCV) on hepatic steatosis. Infection with HCV, perhaps especially genotype 3, is an independent risk factor for the development of hepatic steatosis, regardless of BMI or insulin resistance.43 Based on available United States prevalence data, the DHS population would be expected to have 86 subjects with HCV, of which 7 (8%) would be expected to have genotype 3.44–46 Thus, it is unlikely that HCV infection contributed significantly to the ethnic and sex differences in the prevalence of hepatic steatosis in this study. Moreover, HCV infection is associated with low plasma levels of apoB-containing lipoproteins, possibly because of inhibition of very low-density lipoprotein cholesterol secretion.47, 48 Only 12 subjects with hepatic steatosis had a very low-density lipoprotein cholesterol level of less than 10 mg/dL (0.11 mmol/L), and 17 subjects had a low-density lipoprotein cholesterol level of less than 50 mg/dL (0.56 mmol/L); most of these subjects were black (8 of 12 and 8 of 17, respectively), which was the group with the lowest prevalence of hepatic steatosis.
The prevalence of NAFLD and its associated complications (nonalcoholic steatohepatitis and cirrhosis) is increasing worldwide and is expected to more than double by the year 2025.49 Two studies have reported significant ethnic disparities in the prevalence of NAFLD-related cirrhosis.18, 19 One study, also performed in Dallas, found that Hispanics had a disproportionately high prevalence of NAFLD-related cirrhosis, while that of blacks was disproportionately low.19 The results of this study suggest that ethnic differences in the prevalence of hepatic steatosis may contribute to the observed differences in NAFLD-related cirrhosis between the three featured racial groups. These data indicate that the burden of steatosis-related liver disease is likely to disproportionately affect Hispanics, the fastest growing ethnic group in the United States. Understanding the mechanisms responsible for the ethnic differences in the prevalence of hepatic steatosis and steatosis-related liver injury may provide clues to the development of new therapeutic approaches for the prevention and treatment of this disorder.
We wish to thank Dwain Thiele and Gregory Fitz for helpful discussions. We also thank the DHS investigators, especially Mujeeb Basit and DuWayne Willett, for data organization and management.
- 1The prevalence and etiology of elevated aminotransferase levels in the United States. Am J Gastroenterol 2003; 98: 960–967., , .Direct Link:
- 10Nonalcoholic steatohepatitis: a proposal for grading and staging the histological lesions. Am J Gastroenterol 1999; 94: 2467–2474., , , , .Direct Link:
- 18Is NASH underdiagnosed among African Americans? Am J Gastroenterol 2002; 97: 1496–1500., , , .Direct Link:
- 19Ethnic differences in the prevalence of cryptogenic cirrhosis. Am J Gastroenterol 2004; 99: 292–298., , , .Direct Link:
- 22Institute for Survey Research. Technical Report for the 1995 National Alcohol Survey. Philadelphia, PA: Temple University, 1996.
- 23Anonymous. Plan and Operation of the Third National Health and Nutrition Examination Survey. Washington, DC: National Center for Health Statistics, 1994.
- 28Nonalcoholic fatty liver disease. In: FeldmanM, FriedmanLS, Sleisenger, MH, eds. Gastrointestinal and Liver Disease: Pathophysiology/Diagnosis/Management. Volume 2. 7th ed. Philadelphia: Saunders, 2002: 1393–1400., .