Prevalence of Nonalcoholic Fatty Liver Disease in the United States: The Third National Health and Nutrition Examination Survey, 1988–1994

Study population

The NHANES III was conducted between 1988 and 1994 and is a cross-sectional national examination study conducted in the United States by the National Center for Health Statistics (Hyattsville, Maryland). The NHANES III was based on a complex multistage, stratified, clustered probability-sample design and comprised a representative sample of the US population living in households. The overall participation rate in the interview and examination was 78%. The complex NHANES III design and sampling can be used to generate estimates of health and disease in the general US population, incorporating adjustments for nonresponse. The NHANES III was approved by the institutional review board of the National Center for Health Statistics.

Data collection

Detailed descriptions of the NHANES III data collection are available elsewhere (19). Briefly, individuals received standardized interview questionnaires to provide sociodemographic data. Race/ethnicity was categorized as non-Hispanic white, non-Hispanic black, and Mexican American. Individuals who did not identify themselves as belonging to 1 of these categories were not analyzed separately but were included in total estimates. Additional information included urbanization of the area and US geographic region of residence, smoking, alcohol use (a series of questions about the frequency per week and amount per day on a drinking day over the past 12 months), medical history (e.g., medication use, diagnosis of diabetes or hypertension, history of cardiovascular events), and physical activity. Adults were defined as sedentary if they answered “no” to all questions regarding engaging in any of the following activities over the last month: jogging/running, bicycling, swimming, aerobics, other dancing, calisthenics, garden/yard work, weight lifting, or other sports. Average alcohol consumption was estimated by multiplying the number of drinking days by the average number of drinks per day on a drinking day. “Never drinkers” were individuals who replied “no” to the following question: In your entire life, have you had at least 12 drinks of any kind of alcoholic beverage?

Body mass index was calculated as weight (kg)/height (m)². Gallbladder ultrasonography was performed by using a Toshiba Sonolayer V SSA-90A (Toshiba America Medical Systems, Inc., Tustin, California) with 3.75- and 5.0-MHz transducers (20).

Plasma glucose, serum insulin, serum total cholesterol, serum triglycerides, and serum high-density lipoprotein cholesterol levels were measured by using standardized methods. Low-density lipoprotein cholesterol levels were calculated from measured values of total cholesterol, triglycerides, and high-density lipoprotein cholesterol. Serum biochemistry analysis was performed by using the Hitachi 737 automated multichannel chemistry analyzer (Boehringer Mannheim Diagnostics, Inc., Indianapolis, Indiana). The following liver tests were assayed: asparatate aminotransferase, alanine aminotranferase, γ-glutamyl transferase, alkaline phosphatase, and total bilirubin. Presence of antibodies to hepatitis C was tested by using a second-generation enzyme immunoassay test (Abbott Laboratories, Ltd., Chicago, Illinois) and confirmed with the MATRIX assay (Abbott Laboratories, Ltd.). Antibodies to the hepatitis B core antigen were measured by using a solid-phase competitive immunoassay (Abbott Laboratories, Ltd.). Iron, total iron binding capacity, and ferritin were also measured. Serum transferrin saturation was calculated as serum iron (μg/dL) / serum total iron binding capacity (μg/dL) × 100 (21).

We defined diabetes on the basis of self-reported physician diagnosis, medication use, fasting plasma glucose ≥126 mg/dL, or 2-hour glucose tolerance test ≥200 mg/dL. Hypertension was defined on the basis of self-reported physician diagnosis, medication use, systolic blood pressure ≥140 mm Hg, or diastolic blood pressure ≥90 mm Hg. Finally, we defined hypercholesterolemia on the basis of self-reported physician diagnosis, medication use, or total cholesterol >240 mg/dL.

Among people without diabetes, we calculated insulin resistance by using the following homeostasis model assessment insulin resistance (HOMA-IR) formula: fasting serum insulin (μU/mL) × fasting plasma glucose (mmol/L) / 22.5. We used the adjusted insulin value for examinees. Elevated HOMA-IR was defined as >2.86, representing a level above the fourth quartile of HOMA-IR among people without diabetes (22).

Hepatic steatosis ultrasonography image assessment in the NHANES III

Between 2009 and 2010, hepatic steatosis examination was conducted to assess the presence of fat within the hepatic parenchyma. We retrieved and reviewed 13,856 gallbladder ultrasonography videotapes (96.6% of the original sample) that were originally obtained in the NHANES III between 1988 and 1994.

A more detailed description of the protocol can be found elsewhere (23). Briefly, the following information was recorded: 1) the presence of liver-to-kidney contrast (yes, no, or not assessed); 2) the degree of brightness of the liver parenchyma (none, intermediate, moderate, or severe); 3) the presence of posterior deep beam attenuation (yes, no, or not assessed); 4) the presence of echogenic walls in the small intrahepatic vessels (yes, no, or not assessed); and 5) the definition of the gallbladder walls (clear, intermediate, obliterated, or not assessed). By using a standardized algorithm, we made an overall primary finding of hepatic steatosis on the basis of the presence or absence of each of the parameters listed above, and we graded the condition of the liver as normal, mild, moderate, or severe steatosis. Steatosis was also treated as a dichotomous variable (i.e., present (moderate or severe) or absent (normal-mild)). The rates of intra- and interrater reliability of the presence of hepatic steatosis were 0.77 (95% confidence interval (CI): 0.73, 0.82) and 0.70 (95% CI: 0.64, 0.76), respectively (23). The raters were unaware of the participants’ health characteristics when grading the ultrasonography videotapes.

Definition of hepatic steatosis and NAFLD

We defined hepatic steatosis as the presence of moderate or severe hepatic steatosis by ultrasonography regardless of the presence of other liver disease markers (e.g., hepatitis C infection). Because there is no consensus definition in the literature, we defined NAFLD as the presence of hepatic steatosis in the absence of elevated alcohol consumption (≥1 drink/day for women and ≥2 drinks/day for men) as previously defined (24–26), or the use of zydovudine or didanosine, which are medications associated with the presence of hepatic steatosis (27).

Elevated levels of alanine aminotransferase or aspartate aminotransferase were defined on the basis of the upper limit of normal of the NHANES III laboratory values (alanine aminotransferase of >40 U/L for men and >31 U/L for women; aspartate aminotransferase of >37 U/L for men and >31 U/L for women) (21).

Statistical methods

Analyses included 12,454 NHANES III participants for whom we had complete information on all of the following: hepatic steatosis, racial/ethnic group, sex, age, body mass index, medication use, self-reported diabetes, alcohol consumption, viral hepatitis (B and C), transferrin levels, and levels of alanine aminotransferase and aspartate aminotransferase. No significant differences were observed between persons included in the analyses compared with those excluded. Thus, the analytical sample comprised 89.9% of all participants with hepatic steatosis data and 87.1% of all participants who underwent gallbladder ultrasonography during the 1988–1994 examinations. All analyses were performed by incorporating the sampling weights to obtain unbiased estimates from the complex NHANES III sampling design. The standard errors for all estimates were obtained by using the Taylor series (linearization), as recommended in the NHANES III (28). These methods allow findings to be generalized and to be considered nationally representative of the noninstitutionalized US population aged 20–74 years.

Age-adjusted prevalence estimates of NAFLD were computed by using direct standardization with age-based group proportions based on 2000 US Census data. Estimates of the affected population were obtained by multiplying the sex-race–specific prevalence estimates by the sex-race population totals (28). Given the high prevalence of NAFLD, we used the robust Poisson method to estimate the prevalence ratios of NAFLD by key metabolic characteristics while adjusting for confounders (29).

Although the main focus of the current analyses was NAFLD, we examined the prevalence and epidemiology of any hepatic steatosis and hepatic steatosis other than NAFLD, and the results are presented in detail in the Web tables available at http://aje.oxfordjournals.org/.

Analyses were conducted by using Stata, versions 10 and 11, software (StataCorp LP, College Station, Texas) and SAS software (SAS Institute, Inc., Cary, North Carolina). P ≤ 0.05 was considered statistically significant. No corrections were made for multiple comparisons.

Prevalence of NAFLD

The age-adjusted prevalence of hepatic steatosis was 21.4% (95% CI: 19.7, 23.1), which corresponds to 32.5 (95% CI: 29.9, 35.0) million people in the United States. Ninety percent of these were considered to have NAFLD. The age-adjusted prevalence of NAFLD was 19% (95% CI: 17.5, 20.6), which corresponds to 28.8 (95% CI: 26.6, 31.2) million people in the United States.

As shown in Table 1, the prevalence of NAFLD was substantially higher among Mexican Americans compared with non-Hispanic whites and non-Hispanic blacks and among men compared with women. Across race and sex categories, the prevalence of NAFLD was lower among people aged 20–39 years compared with those aged ≥40 years. For every age category and for both men and women, Mexican Americans had a higher prevalence of NAFLD. The highest observed prevalence was 33% among Mexican-American men aged 40–59 years, followed by 29% among Mexican-American women of the same age group. As shown in Table 2, we also found a strong association between body mass index categories and the prevalence of NAFLD across racial/ethnic groups. A similar association was observed among people with high waist circumference (32.4% of whom had NAFLD compared with 9.7% of those with normal waist circumference).

To take into account differences in the prevalence of these risk factors across racial/ethnic groups, we calculated the adjusted prevalence of NAFLD for each racial/ethnic group. After adjustment for age, sex, education, income, diabetes, and body mass index, the 24.1% (95% CI: 20.8, 27.5) prevalence of NAFLD remained significantly higher for Mexican Americans (24.1%, 95% CI: 20.8%, 27.5%) compared with the 17.8% (95% CI: 16.1, 19.5) and 13.5% (95% CI: 11.3, 15.7) prevalence values for non-Hispanic whites and non-Hispanic blacks, respectively. Similarly, the estimated prevalence of NAFLD by sex adjusted for age, race, obesity status, diabetes, education, and income showed a significantly higher 20.2% (95% CI: 18.0, 22.5) prevalence among men compared with 15.8% (95% CI: 14.3, 17.2) prevalence among women (P < 0.001).

Association between metabolic risk factors and NAFLD

The adjusted prevalence ratios of NAFLD according to selected risk factors are shown in Table 3. In the age-, sex-, and race/ethnicity–adjusted analyses, those in the overweight and obese categories were significantly more likely to have NAFLD compared with individuals of normal weight. The prevalence ratios remained significant even after further adjustment for other metabolic abnormalities. Similarly, there was a strong and independent association between insulin resistance, with or without the presence of diabetes, and NAFLD. Additionally, sedentary individuals had a significantly higher prevalence of NAFLD independent of other risk factors. No significant interactions were found between age and sex, age and race/ethnicity, race/ethnicity and sex, or diabetes and body mass index.

Association between liver enzymes and NAFLD

The prevalence of elevated alanine aminotransferase was 6.0% (95% CI: 5.0, 7.2), and among these subjects, 41% (95% CI: 35.6, 45.9) had NAFLD. The prevalence of elevated aspartate aminotransferase was 5.6% (95% CI: 4.9, 6.4), and 33.8% (95% CI: 28.4, 39.7) of these subjects had NAFLD.