Study Importance What is already known? ► Individuals with obesity or diabetes mellitus (DM) with elevated alanine aminotransferase or steatosis are believed to be at a high risk for steatohepatitis with advanced liver fibrosis (nonalcoholic steatohepatitis fibrosis) and cirrhosis. ► However, this knowledge derives largely from selected populations and has not yet been demonstrated in a large cohort in the general US population, in which the interaction and/or added risk of having both overweight or obesity and DM for the development of fibrosis remains unclear. What does this study add? ► In this large US cohort, steatosis was primarily driven by obesity and was only modestly impacted by DM. ► In contrast, the presence of DM added a significant risk of advanced fibrosis and about a twofold risk of cirrhosis to individuals with overweight or obesity. ► Owing to the fact that advanced fibrosis is an established risk factor for cirrhosis and increased mortality, screening is imperative in adults with DM. How might these results change the direction of research or the focus of clinical practice? ► This study highlights the additive impact of DM on the development of steatosis and fibrosis for individuals with overweight or obesity, as well as the need for improved diagnostic approaches in the future. ► Based on the current findings, the American Diabetes Association guidelines that recommend screening for nonalcoholic fatty liver disease in individuals with elevated alanine aminotransferase or steatosis are a first step in the right direction. INTRODUCTION

Nonalcoholic fatty liver disease (NAFLD) is the most common cause of chronic liver disease in the United States. Individuals with obesity or diabetes mellitus (DM) have the highest risk of developing its more severe form, nonalcoholic steatohepatitis (NASH), with inflammation, hepatocyte injury, and severe fibrosis. There are ~18.2 million people in the United States with DM and NAFLD, with about one-third having NASH ((1-3)). The American Diabetes Association (ADA) recommended screening for advanced fibrosis in all patients with prediabetes or DM with elevated plasma alanine aminotransferase (ALT) and/or hepatic steatosis ((4)). NAFLD not only predisposes individuals to advanced liver diseases (cirrhosis and hepatocellular carcinoma), but it also is often associated with worse insulin resistance ((5)), dyslipidemia ((6)), DM ((7)), and cardiovascular disease (CVD) ((8)). Over the next 20 years, NASH in patients with DM will be responsible for ~812,000 liver-related deaths and ~1.37 million cardiovascular-related deaths, at an estimated cost of $55.8 billion ((1, 3)).

The diagnosis of NAFLD is usually based on history and serum diagnostic panels that combine clinical parameters with routine laboratory tests, followed by imaging and liver biopsy ((2, 9, 10)). The Fatty Liver Index (FLI) ((11)) and United States FLI (USFLI) ((12)) are the most suitable validated diagnostic panels ((12)). A recent meta-analysis based largely on liver ultrasonography studies reported a worldwide prevalence of NAFLD in ~55% of patients with DM ((13)). However, ultrasonography may underestimate steatosis ((14)). Studies using controlled attenuation parameter ((15-19)) or magnetic resonance-based techniques (the gold standard) ((20-22)) have suggested an even higher prevalence of steatosis in patients with DM, but access to these imaging techniques is more costly and often limited. However, the true target of screening is fibrosis, not steatosis per se, because fibrosis (not steatosis) is associated with an increased risk of mortality from end-stage liver disease and CVD ((8-10, 23)). Individuals with DM and elevated ALT or steatosis ((4)) are especially at risk for steatohepatitis with advanced liver fibrosis, defined as having either fibrosis stage F3 (advanced fibrosis) or F4 (cirrhosis) ((9)). An earlier diagnosis would facilitate treatment according to current guidelines with lifestyle management, vitamin E, or pioglitazone ((9, 10, 24, 25)). Additionally, many new drugs are under development ((26)). However, the true prevalence of advanced fibrosis associated with NASH in the general US population, or in those with DM, remains unclear. The most widely accepted blood diagnostic panels are the fibrosis-4 index (FIB-4) and NAFLD fibrosis score (NFS) ((9, 10, 27-30)), as they can predict liver-related mortality in patients with NASH ((31)). However, their use in patients with DM has suffered from small sample size, study design heterogeneity, and populations in which patients with and without DM were analyzed together ((15, 16, 20-22, 30, 32)). Additionally, few analyses have been performed in the United States ((21, 30)) or have examined the contributing role of obesity to fibrosis in patients with DM.

We aimed to address the prevalence of steatosis and of NASH-associated advanced liver fibrosis in a US population, taking advantage of the large, mixed ethnicity population–based data from the National Health and Nutrition Examination Survey (NHANES) 2015-2016, which was conducted by the US National Center for Health Statistics (NCHS) ((33-35)).

METHODS Data source

Data from the NHANES 2015-2016 cycle collected by the NCHS were obtained for this study ((33, 34)). The survey combined interviews with physical examinations designed to provide a cross-sectional view of the health status of adults in the United States. The NHANES interviews collected data on demographic, socioeconomic, and medical conditions, whereas the physical examinations consisted of physiological measurements and laboratory testing.

Participants

Participants were selected if they were ≥18 years old and had undergone a medical examination (Figure 1). Of the 9,971 participants surveyed during the NHANES 2015-2016 cycle, a total of 3,841 were ≥18 years old, had available medical examination and laboratory results, no evidence of significant alcohol consumption (≤2 drinks/d for men and ≤1 drink/d for women), and a BMI ≥ 25 kg/m2. There were 1,696 participants with overweight (BMI ≥ 25-<30) and 2,145 participants with obesity (BMI ≥ 30) included in the analysis. The population with overweight was composed of 263 participants with DM, defined as having a self-reported physician diagnosis of DM, receipt of DM medications, or glycated hemoglobin (HbA1c) ≥ 6.5%, and there were 1,433 participants with no evidence of DM. The population with obesity was composed of 571 participants with evidence of DM and 1,574 participants “without DM.” Hypertension was defined from the physical exam data as systolic blood pressure ≥ 140 mm Hg and/or diastolic blood pressure ≥ 90 mm Hg. CVD was a patient-reported diagnosis of coronary heart disease, angina, heart attack, or stroke.

Flow diagram of NHANES participants and study subgroups. NHANES = National Health and Nutrition Examination Survey

Study measures The presence of NAFLD was predicted using two indices: the FLI, comprising triglycerides, BMI, gamma-glutamyl transferase (GGT), and waist circumference ((11)), and the more recently developed USFLI ((12)), comprising age, ethnicity, GGT, waist circumference, fasting glucose, and fasting insulin ((4)). FLI was calculated as follows:

An FLI cutoff value of ≥ 60 was used to denote NAFLD ((11)). For FLI values ≥ 60, the reported sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) are 44%, 90%, 67%, and 76%, respectively ((36)).

Owing to the fact that NHANES is a US database, we also established USFLI scores, believed to be more precise in assessing steatosis ((12)), at least as validated in US populations and considering relevant variables such as insulin resistance (i.e., fasting plasma glucose and insulin) and ethnicity. USFLI ((12)) was calculated using the following formula, where Non-Hispanic Black and Mexican American have a value of 1 if the participant is of that ethnicity and 0 if not of that ethnicity:

A USFLI cutoff value of ≥30 was used to diagnose fatty liver, with a reported sensitivity, specificity, likelihood ratio positive, and likelihood ratio negative values of 62%, 88%, 5.2, and 0.43, respectively ((12)).

The USFLI could be used only in a subset of NHANES participants, as serum fasting insulin was unavailable in 58% of study participants. Therefore, we focused on FLI for the overall population analysis.

The FIB-4, NFS, and aspartate aminotransferase (AST)-to-platelet ratio index (APRI) were chosen as the diagnostic panels to identify participants with advanced fibrosis, as these are well validated, supported by the literature ((9, 10, 27-29, 31, 37, 38)), and most widely used in clinical practice. NFS, FIB-4, and APRI were calculated as follows: NFS = −1.675 + 0.037 × Age (years) + 0.094 × BMI (kg/m2) + 1.13 × (Impaired Fasting Glucose or DM) + 0.99 × (AST/ALT) − 0.013 × Platelet (× 109/L) − 0.66 × Albumin (g/dL), in which Impaired Fasting Glucose or DM had a value of 1 if the participants had impaired fasting glucose and 0 if they did not

We chose the traditionally accepted cutoffs to define the relative risk of having clinically significant fibrosis for each panel. Participants with FIB-4 < 1.3 were considered as being at the lowest risk of advanced liver fibrosis, followed by those with FIB-4 ≥ 1.3 and <1.67. The cutoff of ≥1.67 was chosen based on a prior cohort we studied that indicated the cutoff of ≥1.67 as a good discriminatory value to identify those with clinically relevant fibrosis among patients with DM ((29)). Participants with FIB-4 ≥ 1.67 to < 2.67 were classified as having a moderate risk, whereas those with FIB-4 ≥ 2.67 were classified as having a high risk of advanced liver fibrosis (F3 or higher; ≥F3) based on the literature ((9, 10, 27-29, 31)). For NFS, we also used accepted cutoffs for the risk of liver fibrosis ((9, 10, 27-29, 31)). An NFS value of <−1.455 was considered as low risk, −1.455 to 0.676 as intermediate risk, and >0.676 as high risk for advanced liver fibrosis (≥F3). The reported areas under the receiver operating characteristic curve for the diagnosis of advanced fibrosis have been reported to be 0.78 to 0.80 for FIB-4 and 0.72 to 0.75 for NFS ((28-31)). For instance, based on a recent analysis by Anstee et al. ((27)), for a threshold of <1.3 for FIB-4, the sensitivity, specificity, PPV, and NPV were 82%, 57%, 83%, and 56%, respectively. In contrast, FIB-4 with a cutoff value of ≥2.67 has a sensitivity of 36%, specificity of 93%, PPV of 93%, and NPV of 37%. A cutoff of <−1.455 for NFS is able to detect advanced fibrosis with a sensitivity of 89%, specificity of 37%, PPV of 85%, and NPV of 46%. NFS with a threshold value > 0.676 had a sensitivity, specificity, PPV, and NPV of 88%, 89%, 93%, and 26%, respectively.

At upper and lower cutoff values (<0.50, >1.50), APRI has been reported to have a sensitivity of 31%, specificity of 99%, PPV of 67%, and NPV of 94% ((29)).

Statistical analyses

The data were analyzed using descriptive statistics. Health indices were used to approximate the prevalence of NAFLD and NASH with fibrosis among populations of participants with overweight and participants with obesity. Group comparisons were made using the χ2 test for categorical measures and the Student t test for continuous measures. Comparisons involving steatosis and fibrosis indices were adjusted for multiple comparisons using the Holm–Bonferroni method. Categorical measures were reported as percentages, and continuous measures were reported as mean (SD) values. Logistic regression models examining the impact of DM on fibrosis while controlling for age were performed. Corresponding odds ratios (ORs) and 95% confidence intervals (CIs) are reported.

Analyses were conducted using the survey weights provided with the NHANES data. The weights were used to account for the complex survey design, including nonresponse and oversampling of certain age and ethnic groups. Weighting of the NHANES data produced results that were representative of the US noninstitutionalized civilian resident population. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, North Carolina).

RESULTS

Data were collected from 3,841 participants ≥ 18 years old without evidence of significant alcohol consumption and BMI ≥ 25. Of these, a total of 2,145 participants (56%) had a diagnosis of obesity, and 1,696 (44%) had overweight. Among included participants, 834 (21.7%) had a self-reported physician diagnosis of DM, were taking DM medications, or had HbA1c ≥ 6.5%, a prevalence consistent with previous Centers for Disease Control and Prevention reports in patients with higher BMI and older age (https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf).

Table 1 shows the baseline characteristics and risk factors for the development of NAFLD, categorized into participants with either overweight or obesity with or without DM. CVD and hypertension were significantly more prevalent in participants with overweight or obesity with DM versus without DM (p < 0.05-0.001). Patients with DM were older versus those without DM. In the group of individuals with overweight, there were fewer female participants with DM versus without DM (p < 0.05). There were no significant differences in ethnicity between the groups; this is representative of a typical US population. Metabolic parameters (HbA1c, glucose, and lipids) were worse, as expected, in those with DM versus without DM.

TABLE 1. Baseline characteristics of the NHANES populations with overweight and obesity Population with overweight† Population with obesity‡ Without DM (n = 1,433) With DM§ (n = 263) Without DM (n = 1,574) With DM§ (n = 571) Age (y), mean (SD) 48 (9.2) 63 (4.4)*** 46 (9.0) 57 (5.9)*** Sex, female (%) 47 32.2** 55.3 55.3 Ethnicity (%) Non-Hispanic White 64.7 58.2 61.6 58.2 Non-Hispanic Black 10.1 11.3 12.5 16.6 Hispanic¶ 16.4 21.3 19.3 17.8 Other¶¶ 8.8 9.2 6.6 7.4 BMI (kg/m2), mean (SD) 27.4 (0.7) 27.7 (0.2) 35.9 (5.0) 37.0 (3.0)* History of CVD¶¶¶ (%) 6.3 21.5** 7.2 23.1*** Hypertension¶¶¶¶ (%) 25.4 42.8** 29.5 43.7** DM (%) NA 100.0 NA 100.0 Glucose (mg/dL), mean (SD) 93 (7.3) 144 (12.0)*** 96 (7.4) 149 (20.5)*** HbA1c (%), mean (SD) 5.4 (0.2) 7.3 (0.4)*** 5.5 (0.2) 7.3 (0.6)*** Total cholesterol (mg/dL), mean (SD) 199 (21) 178 (7)** 195 (19) 185 (12)** HDL (U/L), mean (SD) 56 (9) 50 (3)** 50 (7) 46 (4)** LDL (U/L), mean (SD) 119 (11) 96 (4)*** 118 (11) 106 (6)*** Triglycerides (mg/dL), mean (SD) 158 (83) 185 (22) 171 (58) 208 (43)* ALT (U/L), mean (SD) 25 (6) 28 (3)* 29 (10) 29 (5) AST (U/L), mean (SD) 25 (5) 27 (2) 26 (5) 29 (8)* Platelet count (109/L), mean (SD) 233 (26) 225 (10) 247 (28) 239 (18)* Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; CVD, cardiovascular disease; DM, diabetes mellitus; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; LDL, low-density lipoprotein; NA, not applicable; NHANES, National Health and Nutrition Examination Survey. † BMI ≥ 25 to <30 kg/m2. ‡ BMI ≥ 30 kg/m2. § Self-reported based on physician diagnosis and/or taking DM medications. ¶ Hispanic includes “Mexican American” and “other Hispanic.” ¶¶ Other includes “non-Hispanic Asian” and “other race, including multi-racial” categories. ¶¶¶ CVD was a patient-reported diagnosis of coronary heart disease, angina, heart attack, or stroke. ¶¶¶¶ Hypertension was defined from the physical exam data as systolic blood pressure ≥ 140 mm Hg and/or diastolic blood pressure ≥90 mm Hg. * p < 0.05 vs. without DM ** p < 0.01 vs. without DM *** p < 0.001 vs. without DM. The without DM group was used as the reference group within the population with overweight and obesity. Prevalence of hepatic steatosis (NAFLD)

Table 2 summarizes the NAFLD and liver fibrosis indices across the different cohorts. A diagnosis of hepatic steatosis by FLI ≥ 60 was observed more frequently among individuals with obesity with DM versus without DM (94.4% vs. 90.1%; p = 0.396). This was higher than that in the participants with overweight with DM (52.4%) or without DM (32.5%; p = 0.074 vs. with DM).

TABLE 2. Hepatic steatosis and fibrosis indices among participants Overweight group† Obesity group‡ Overall Without DM (n = 1,433) With DM§ (n = 263) Without DM (n = 1,574) With DM§ (n = 571) Without DM (n = 3,007) With DM§ (n = 834) Steatosis indices FLI > 60 (%) 32.5 52.4 90.1 94.4 62.6 82.6*** USFLI¶ ≥ 30 (%) 17.4 48.3** 57.6 79.9** 38.4 70.8*** Advanced fibrosis risk stratification FIB-4 < 1.3 (%; lowest risk) 56.1 38.3 66.7 49.7* 61.5 46.4* FIB-4 ≥ 1.3 to < 1.67 (%¶¶; low risk) 19.5 27.8 17.3 19.3 18.4 21.8 FIB-4 ≥ 1.67 to < 2.67 (%)¶¶; moderate risk) 19.0 24.1 13.7 24.9* 16.3 24.7 FIB-4 ≥ 2.67 (%; high risk of advanced fibrosis) 5.4 9.7 2.4 6.1 3.8 7.1 FIB-4 ≥ 1.67 (%; moderate-to-high risk) 24.3 33.9 16.1 31.0** 20.1 31.8* NFS < −1.455 (%) 70.8 12.6*** 53.7 9.7*** 61.9 10.5*** NFS −1.455 to 0.676 (%) 27.0 68.4*** 39.8 56.2** 33.7 59.7*** NFS > 0.676 (%) 2.2 18.9** 6.4 34.1*** 4.4 29.8*** Abbreviations: DM, diabetes mellitus; FIB-4, fibrosis-4 index; NFS, nonalcoholic fatty liver disease fibrosis score; USFLI, United States Fatty Liver Index. † BMI ≥ 25 to <30 kg/m2. ‡ BMI ≥ 30 kg/m2. § Self-reported based on physician diagnosis and/or taking DM medications. ¶ USFLI measurements required fasting insulin data, which were not available for all participants of the survey. ¶¶ Cutoff based on Bril et al. ((30)). * p < 0.05 vs. without DM ** p < 0.01 vs. without DM *** p < 0.001 vs. without DM; p values adjusted for multiple comparisons using the Holm–Bonferroni method.

When hepatic steatosis was calculated using USFLI ≥ 30, a similar trend was observed but with an o