We performed a cross-sectional study of newly diagnosed cases of nonalcoholic fatty liver disease (NAFLD) identified between December 1998 and December 2000 in the Chronic Liver Disease Surveillance Study. We compared the demographic and clinical features of NAFLD in a racially diverse representative U.S. population (Alameda County, CA). Diagnostic criteria for probable NAFLD were persistent unexplained elevation of serum aminotransferase levels, radiology (ultrasound or computed tomography scan) consistent with fatty liver, and/or two or more of the following: (i) body mass index of 28 kg/m2 or more, (ii) type 2 diabetes, or (iii) hyperlipidemia, in the absence of significant alcohol use. Definite NAFLD cases required histological confirmation. Of the 742 persons with newly diagnosed chronic liver disease, 159 (21.4%) had definite or probable NAFLD. The majority were nonwhite: Hispanics (28%), Asians (18%), African Americans (3%), and other race(s) (6%). African Americans with NAFLD were significantly older than other racial or ethnic groups (P < .001), and in Asians, NAFLD was 3.5 times more common in males than in females (P = .016). Clinical correlates of NAFLD (obesity, hyperlipidemia, diabetes) were similar among racial and ethnic groups, except that body mass index was lower in Asians compared with other groups (P < .001). Compared with the base population (Kaiser Permanente members), Hispanics with NAFLD were overrepresented (28% vs. 10%) and whites were underrepresented (45% vs. 59%). In conclusion, these racial and gender variations may reflect differences in genetic susceptibility to visceral adiposity, including hepatic involvement, and may have implications for the evaluation of persons with the metabolic syndrome. Clinicians need to be aware of the variable presentations of NAFLD in different racial and ethnic groups. (HEPATOLOGY 2005;41:372–379.)
Nonalcoholic fatty liver disease (NAFLD) is increasingly recognized to be among the most common causes of chronic serum aminotransferase elevation in the United States1 as well as in several other countries, including Japan, Australia, Europe, and the Middle East.2—6 Additionally, the histological lesion associated with more advanced disease, nonalcoholic steatohepatitis, has emerged as an important cause of chronic liver disease, cirrhosis, and hepatocellular carcinoma.3, 7, 8 Clinical conditions commonly associated with a fatty liver include central obesity, type 2 diabetes mellitus, and dyslipidemia.1, 9 As the prevalence and severity of these conditions continues to rise in the United States, the prevalence and disease burden of NAFLD are predicted to increase substantially.10 The prevalence rates for an overweight state, defined as a body mass index (BMI) of more than 25 kg/m2, have risen to 64.5%11 in the United States and the age-adjusted prevalence of obesity, defined as a BMI of 30 kg/m2 or more, is more than 30%. Based on current U.S. census figures, it is estimated that more than 30 million obese adults may have steatosis and that 8.6 million may have steatohepatitis.1 Prevalence rates for type 2 diabetes have similarly risen over the past decade. Currently, 7.8% of the U.S. population has diabetes, representing a 49% increase from the period 1990 through 2000.12
NAFLD has been described in persons of all ages in the United States, but the populations studied have been predominantly white.2, 4, 13—16 In this study, we focused on the clinical presentation of NAFLD in different racial and ethnic groups in a geographically representative and racially diverse segment of the U.S. population.
NAFLD, nonalcoholic fatty liver disease; BMI, body mass index; KP, Kaiser Permanente.
Patients and Methods
The institutional review boards of each of the participating institutions approved the study.
The study population was composed of adult members of the Kaiser Permanente Medical Care Program who participated in the Alameda County Chronic Liver Disease Surveillance Study. This site is one of three in the United States performing surveillance for chronic liver disease. In this cooperative study supported by the Centers for Disease Control and Prevention, new cases of chronic liver disease are identified and detailed demographic and clinical data are collected to establish the etiology, risk factors, and comorbidities associated with liver disease. The Kaiser Permanente (KP) membership is generally representative of the Alameda County population, except at the extremes of income.17 The proportion of KP members with an annual household income of less than $15,000 or more than $100,000 is lower than in the general Northern California population (6% vs. 10% with income of less than $15,000, and 18% vs. 24% with income of more than $100,000). Alameda County residents followed at one of four KP medical facilities (Pleasanton, Fremont, Hayward, and Oakland) serving Alameda county residents were eligible to participate. Additionally, study participants were required to be 18 years of age or older and to have continuous health plan membership for 18 months before initial identification for inclusion in the study.
A case of newly identified chronic liver disease was defined by specific laboratory, histological, or radiological criteria. A computer-based search was used to identify cases using predefined criteria. Laboratory criteria were persistently elevated liver blood tests (serum alanine aminotransferase and serum aspartate aminotransferase levels, alkaline phosphatase, or bilirubin), with nonliver causes excluded. The same liver test had to be elevated on at least two occasions at least 6 months apart, but not greater than 18 months apart, with the second test abnormality identified between December 1998 and December 2000. Radiological criteria required evidence of cirrhosis (nodular or shrunken liver) or portal hypertension (splenomegaly, varices, recanalized umbilical vein, or ascites). Histological criteria required a liver biopsy result consistent with chronic liver disease.18
Participants in the Chronic Liver Disease Surveillance Study were assigned the diagnosis of definite or probable NAFLD based on the following criteria. Definite NAFLD cases required histological confirmation. Probable NAFLD cases were defined as persistent unexplained elevation of serum aminotransferases with ultrasound or computed tomography scan results consistent with fatty liver and/or two or more of the following clinical conditions: (i) a BMI of 28 kg/m2 or more, (ii) type 2 diabetes, or (iii) hyperlipidemia; and an absence of significant alcohol use, defined as a current daily alcohol consumption of 40 g/d or more in males and 20 g/d or more in females based on lifetime drinking history. Patients with serological or histological evidence of viral hepatitis B or C, autoimmune, metabolic, or other identifiable liver diseases were excluded based on available laboratory data. Serological tests for viral hepatitis (anti-hepatitis C virus, total anti–hepatitis B core antigen, and hepatitis B surface antigen) were performed on all study participants. Hepatitis C virus infection was confirmed by testing for hepatitis C virus RNA using qualitative and quantitative assays in those persons with positive anti-hepatitis C virus test results. Autoimmune markers (serum antinuclear antibody ± anti smooth muscle antibody) were available in 48% of participants, and metabolic markers (ferritin, total iron binding capacity, percent iron saturation, serum ceruloplasmin, and α1-antitrypsin level) were available in 25%.
All participants had the following performed: (i) review of computerized medical records; (ii) in-person interview including self-reporting of race and ethnicity, comorbidities, and alcohol intake19; (iii) phlebotomy with testing of blood samples for viral hepatitis; (iv) height, weight, and BMI at the time of interview; and (v) standardized review of liver pathology, if available.
Review of Computerized Records.
Review of computerized medical records was performed on all participants to confirm eligibility. Additionally, clinical data, pharmacy records, and physician-assigned diagnoses between the month in which the patient met the case definition (index month) and the 12 months of follow-up were reviewed. Laboratory variables included serum bilirubin, serum aspartate aminotransferase, serum alanine aminotransferase, alkaline phosphatase, serum albumin, prothrombin time, blood glucose, glycosylated hemoglobin, and triglycerides. Pharmacy record review focused on use of corticosteroids, lipid-lowering agents, oral hypoglycemic agents, and insulin.
All patients were interviewed using a standardized questionnaire that included (i) demographic information such as age, sex, race, and ethnicity; (ii) a detailed history of all medications, including current use of corticosteroids or lipid-lowering agents as well as potential hepatotoxins; (iii) quantitation of current and lifetime alcohol intake; and (iv) comorbidities, including obesity, type 2 diabetes, dyslipidemia, ulcerative colitis, Crohn's disease, and asthma within a 5-year period before interview. Race and ethnicity were determined by self-report.
Liver biopsies obtained between July 1997 and December 2001 were retrieved and reviewed by a single pathologist. Liver biopsy results were evaluated for steatosis, ballooning of hepatocytes, inflammation, and fibrosis using the Brunt criteria.18
Reference Population Used for Comparison of NAFLD Distribution by Race and Ethnicity.
To gain an understanding of differences in the racial and ethnic distribution of NAFLD in the study population versus the Kaiser Permanente base population, we used membership race and ethnicity information from a survey conducted in 1999. The survey, developed by the Division of Research at Northern California Kaiser Permanente, was mailed to a stratified random sample of 5,080 adult health plan members followed up at the Pleasanton, Fremont, Hayward, and Oakland KP facilities in Alameda County. Only Kaiser Permanente Medical Care Program members continuously enrolled within 3 months before administration of the survey were eligible. Respondent data were weighted so that the final sample used to create the profiles of sociodemographic characteristics reflected the actual age and sex distribution of the four medical center service populations in 1999 rather than that of the respondent sample. The overall survey response was 53%. Lower response rates were seen for younger adults (20-44 yr) compared with older age groups and for African Americans compared with Hispanics, Asians, and whites (NP Gordon. Characteristics of Adult Health Plan Members in the Northern California Region Membership, as Estimated from the 1999 Member Health Survey. Personal communication, December 31, 2003.).
Median and mean, and range and standard deviation were used for descriptive statistics, as appropriate. Categorical data were analyzed using the Fisher's exact test and chi-square test. Continuous variables were tested with Student t test or a 1-way ANOVA. A P value of less than .05 was considered statistically significant.
Race and Ethnicity Categories Used for Analysis.
There were six race and ethnicity categories used for analysis: (i) white, (ii) African American, (iii) Asian, (iv) Hispanic, (v) other (which included Pacific Islander, Native American, and other races), and (vi) more than one race. Participants who indicated “yes” for Hispanic ethnicity were classified as Hispanic, regardless of race category. Persons who specified Latino or Hispanic for race were also included in this category.
Of the 742 newly identified cases of chronic liver disease, 333 (39.1%) had definite or probable NAFLD (Fig. 1). A total of 159 (47.7%) of eligible patients with a diagnosis of NAFLD participated, which was similar to the proportion of eligible patients with other diagnoses who participated in the Chronic Liver Disease Surveillance Study. The mean age of the NAFLD study participants was 50 years (range, 20-74 years) and 52.2% were male, which was similar to nonparticipants (mean age, 51 years; 60% male). Among participants, the average BMI was 33.1 kg/m2 (range, 22.9-78.1 kg/m2), and 44.7% and 49.7% had a diagnosis of non-insulin-dependent diabetes mellitus and hyperlipidemia, respectively. The mean serum alanine aminotransferase level was 72 IU/L (range, 15-147 IU/L) and the mean serum aspartate aminotransferase level was 47 IU/L (range, 17-177 IU/L) for the total group.
Most participants in this study had probable NAFLD, because liver biopsies were uncommon in persons with a diagnosis of NAFLD. Only 19 participants (12% of the total NAFLD group) underwent liver biopsy. Among those participants with available histological results, 18 (94.7%) had features consistent with NAFLD (Fig. 1). Six had simple macrosteatosis, 12 had steatohepatitis, and the single remaining participant had nondiagnostic liver histological results. A total of 112 (70.0%) of the participants with probable and definite NAFLD had either abdominal ultrasound or computed tomography study results. Radiological findings consistent with fatty liver were present in 88% of these patients. Ultrasound was performed more frequently than computed tomography. In 90 of 103 (87.4%) participants who had undergone abdominal ultrasound and in 19 of 29 (65.5%) participants who had undergone a computed tomography scan of the abdomen, findings were consistent with fatty infiltration.
Factors Associated With Fatty Liver: Alcohol and Medication Use
Based on prescription records, only 20 (12.4%) participants had a history of being prescribed medications such as prednisone, amiodarone, methotrexate, or hepatotoxic substances associated with fatty liver. Forty-four (27.7%) participants were lifelong nondrinkers; 54 (34.0%) were current drinkers. The remainder (n = 61; 38.4%) were former drinkers who had not consumed alcohol within the 1 year before the interview. The median period of abstinence for former drinkers was 14 years. Among current drinkers, the average amount of alcohol consumed was 8.6 g/d (range, 0.9-31.3 g/d) for a mean of 13.6 years (range, 1-45 years). Current alcohol consumption was more common in males than in females; females were likely to be lifelong nondrinkers (Fig. 2). Of females who were current drinkers, the lifetime mean number of drinks consumed per day was 0.36, equivalent to 4.87 g/d of alcohol (range, 0.9-18.1 g/d), and the mean number of years of alcohol consumption was 15.3 years (range, 1-37 years). Currently drinking males consumed a lifetime average of 0.79 drinks per day or 10.57 g/d of alcohol (range, 2.27-31.3 g/d) over a mean period of 12.9 years (range, 1-45 years).
Race and Ethnicity and NAFLD
Of the 19 participants with definite NAFLD, 8 (44.4%) were white, 5 (27.8%) were Hispanic, 3 (16.7%) were Asian, 1 (5.7%) was American Indian, and 1 (5.7%) was of more than one race. In the combined definite and probable NAFLD population, 44.7% were white, 28.3% were Hispanic, 17.6% were Asian, 3.1% were African American, and 6.3% other or more than one race or ethnicity. Figure 3 depicts the distribution of race and ethnicity in the combined definite and probable NAFLD groups compared with the KP membership. The proportion of Hispanics in the NAFLD study population was higher than the KP membership, and the proportion of whites in the NAFLD study population was lower than the KP membership. Whites, who accounted for 59% of the KP membership, made up 45% of the NAFLD study population, and Hispanics, who accounted for 10% of the KP membership, made up 28% of the NAFLD study population. The proportion of Asians in the KP membership and the NAFLD study population was similar. African Americans, who comprised 9% of the KP members, accounted for only 3% of the NAFLD study population, but the number of African-American participants with NAFLD was small.
The clinical and biochemical characteristics of participants with NAFLD by racial and ethnic group in shown in Table 1. The mean age of NAFLD participants was 50 years, with most NAFLD participants (61.5%) between the ages of 45 and 64 years. There were differences in the mean age of participants by racial and ethnic group (P < .001; Table 1). African Americans were 9 to 15 years older, on average, than whites, Hispanics, or Asians. NAFLD was as frequent in males as females (52.2% vs. 47.8%) overall, but sex differences in the distribution of NAFLD were seen among Asians and African Americans (Fig. 4).
Table 1. Characteristics of NAFLD Participants by Race and Ethnicity*
Further evaluation of participants of Asian race with NAFLD revealed that males were significantly younger than females (median age, 44.0 vs. 55.0 yr; P = .0085). There were no significant differences in the proportion of Asian males (vs. females) with BMI of more than 28 kg/m2 (P = .91), diabetes mellitus (P = .11), dyslipidemia (P = .14), or current alcohol use (P = .93). However, 68% of Asian males were previous drinkers, compared with 17% of Asian females (P < .02).
NAFLD is rapidly emerging as one of the most important chronic liver diseases of the twenty-first century. Epidemiological studies in patients with other types of chronic liver disease have shown that race and ethnicity can be predictive of disease complications and response to treatment. For example, among patients with chronic hepatitis C, African Americans have a lower response to antiviral treatment than whites.20 Asians and Alaskan Natives with chronic hepatitis B virus infection experience higher rates of hepatocellular carcinoma compared to whites.21–23 Thus, identification of racial-ethnic correlates in persons with NAFLD may lead to an improved understanding of the natural history and treatment of this disease. Additionally, sensitivity to differences in disease characteristics in different racial-ethnic groups will aid in the development of appropriate educational programs aimed at the early recognition and prevention of NAFLD within specific communities.
This study determines the distribution of NAFLD in a racially and ethnically defined population in the United States. Our study confirms the emerging importance of NAFLD as a cause of chronic liver disease, with one in every five cases of newly diagnosed chronic liver disease being definite or probable NAFLD. The above figures are in keeping with previous population estimates of NAFLD and of the metabolic syndrome in North America.1
Among cases of newly diagnosed NAFLD, nonwhites comprised two thirds of the study population, with Hispanics making up the largest group (28%), followed by Asians (18%) and African Americans (3%). Pacific Islanders, Native Americans, and Native Hawaiians were also represented in the NAFLD study population, but at a low frequency. Because all NAFLD patients were enrolled in the KP healthcare plan, this study is not biased in terms of healthcare access. However, because the diagnosis of NAFLD was dependent on patients being seen by a physician and certain test results being obtained, it is possible that differences in healthcare-seeking behavior among persons of different racial and ethnic groups may affect rates of newly identified NAFLD in specific groups. Persons at the extremes of income (less than $15,000 or more than $100,000 per year) are underrepresented in the KP membership. This may influence the generalizability of our findings to other populations, but does affect the differential representation of NAFLD in the various racial- ethnic groups. The racial-ethnic distribution of our study population is not necessarily representative of that of all NAFLD patients in the Unites States and reflects, at least in part, the underlying racial-ethnic distribution of our base population (i.e., KP members in Alameda County).
Compared with the estimated racial-ethnic distribution of the base population (KP members) from Alameda County, the distribution of participants in the NAFLD study group showed a higher proportion of Hispanics (28% vs. 10% in the membership survey). The higher proportion of Hispanics in the NAFLD group suggests this racial-ethnic group is at higher risk for this condition. Differential rates of participation in the NAFLD study versus the KP membership survey would be unlikely to explain this substantial difference in proportions. Additionally, there are other lines of evidence suggesting Hispanics may be an ethnic group at higher risk for NAFLD. Other studies have found Hispanics have a higher BMI compared with several other ethnic groups,24–26 and in women with similar BMI and socioeconomic status, Hispanic women have a greater amount of adiposity compared with white women.27 Although the overall representation of African Americans in this study population is too small to draw any firm conclusions, our results are consistent with those of Caldwell et al.,28 who similarly reported a low prevalence of NAFLD in African Americans. In contrast, in the Third National Health and Nutritional Evaluation Survey (NHANES III) study of possible NAFLD cases (based on abnormal liver test results only), an increased prevalence of NAFLD among African-American males was found.29, 30 Although differences in study methodology may account for the lack of consistency across studies, several studies suggest genetic or intrinsic physiological factors may underlie the differences in the prevalence of NAFLD in African Americans versus other racial groups.31–37
We found that NAFLD patients who were of Asian race had a significantly lower BMI than all other racial groups. Although the World Health Organization currently defines overweight and obesity as a BMI of 25 kg/m2 or more and 30 kg/m2 or more, respectively, these definitions may vary by racial-ethnic group. Ko et al.38 recently proposed a lower cutoff value for BMI in persons of Asian race (BMI ≥ 23 kg/m2 for overweight and BMI ≥ 27 kg/m2 for obesity) based on the observation that Hong Kong Chinese have a higher percentage of body fat than whites for a given BMI. Our results support the recommendation for a lower cutoff for overweight and obesity in Asians. In another study, Asians were shown to have a higher amount of visceral adiposity39 and more subcutaneous fat than whites,40 highlighting the limitations of BMI as a single surrogate marker for obesity and obesity-related comorbidities in non-Caucasian racial-ethnic groups. Racial-ethnic differences in the relationship between BMI, percent body fat, and body-fat distribution among whites, Hispanics, and African Americans also have been reported and may contribute to the sex and ethnic variations in NAFLD prevalence.41, 42 Although waist circumference was not measured in our study, persons with an increased waist circumference have been shown to be more likely to exhibit type 2 diabetes, dyslipidemia, and other components of metabolic X syndrome than persons with a normal waist circumference independent of race, BMI, age, income, socioeconomic class, or physical activity.43 Although waist-to-hip ratio generally is thought to correlate with visceral adiposity, Conway et al.44 have shown that waist-to-hip ratio correlates with visceral adiposity in white women but not in African-American women, again highlighting the existence of racial variability in these anthropometric parameters.
One of the limitations of this study was the lack of liver biopsies in most patients. Additionally, although tests to exclude viral hepatitis were available for all participants, testing for autoimmune and metabolic liver diseases was incomplete in many patients. Thus, it is possible that a few participants with incomplete laboratory testing to exclude these less common chronic liver diseases (autoimmune hepatitis, hemochromatosis, Wilson's disease, and α1-antitrypsin deficiency) were included in the study cohort. However, given the low prevalence of these conditions in the population, the effects of this bias are not likely of sufficient magnitude to change the results of our study.
This U.S. study finds sex differences in NAFLD among Asians. NAFLD was strikingly more common in males than females. It is possible that our findings may reflect differences in study participation or healthcare-seeking behavior by sex among Asian KP members. However, in support of sex differences in NAFLD among Asians, a recent study of 3,432 Japanese adults seen at a single hospital in Nagasaki in which ultrasound was used for the diagnosis of fatty liver, NAFLD was twofold more prevalent in men (25.1%) than in women (12.2%).45 Gender-related differences in BMI among Asians did not reach statistical significance in our study. However, sex differences in fat distribution that vary by race and ethnicity may exist. Accumulation of visceral fat was previously shown to be more common in males regardless of total body fat, and deep subcutaneous adipose tissue has been associated with insulin resistance in men, but not in women.45, 46 Whether sex differences in fat distribution and insulin resistance are more or less pronounced in individuals of different races or ethnicities remains unclear, but the current study highlights the variability in risk factors by sex as well as race and ethnicity.
The exact amount of alcohol ingestion required to cause fatty liver is not known. Previous studies in patients with NAFLD have used various cutoffs for alcohol consumption ranging from 40 to 180 g/d as criteria to exclude alcohol-related liver disease. This study uses a validated instrument for quantification of lifetime and current alcohol consumption to ascertain alcohol use accurately. In this study, two thirds of patients were either lifelong nondrinkers or had been abstinent from alcohol for at least 12 months. In those who were current consumers of alcohol, the mean daily alcohol consumption was 8.6 g/d, which is below the maximum recommended daily alcohol intake for men (40 g/d) and women (20 g/d) and well below the level traditionally associated with alcohol-related liver injury.47, 48
In summary, we have shown that NAFLD affects individuals of diverse race and ethnicity within Alameda County, Northern California. Our findings highlight the need for increased awareness of NAFLD within minority racial-ethnic groups, the need for culturally sensitive educational programs aimed at the early recognition and prevention of NAFLD, as well as the need for ensuring adequate representation of different racial-ethnic groups in future clinical studies.
The authors thank Dr. Arthur Reingold for his ongoing support in the Chronic Liver Disease Surveillance Study and Dr. Raphael Merriman for his helpful comments during study planning.