Prevalence of Nonalcoholic Steatohepatitis and Associated Fibrosis Stages Among US Adults Using Imaging-Based vs Biomarker-Based Noninvasive Tests

Introduction: Nonalcoholic fatty liver disease (NAFLD) is believed to be the most common chronic liver disease worldwide. Therapies are under development for nonalcoholic steatohepatitis (NASH), the progressive form of NAFLD, such that the prevalence of NASH with liver fibrosis, which is likely to require treatment, may be of interest to healthcare decision makers. Noninvasive tests are used in initial screening for NASH, as well as in observational studies of NASH prevalence. However, existing evidence does not address how estimated prevalence varies with different noninvasive tests. This analysis estimated the prevalence of NASH among US adults and assessed variation with different noninvasive tests. Methods: A cross-sectional analysis was conducted using the 2017–March 2020 National Health and Nutrition Examination Survey cycle. Participants with presumed NAFLD (steatosis and without alternative causes of liver disease) were identified, among whom NASH was predicted based on FAST score, Fibrosis-4 (FIB-4), and AST-to-Platelet Ratio Index (APRI) cutoffs across 11 scenarios. Among NASH participants, fibrosis stages were explored based on distribution across the spectrum of liver-stiffness measurements. Results: Among participants with complete data for the analysis (N=6969), prevalence of presumed NAFLD was 25.6%. Within presumed NAFLD, prediction of NASH using imaging-based NIT cutoffs yielded estimated prevalence of 1.3%-4.8% (3.3 million-12.2 million) based on FAST score cutoffs from 0.35-0.67. Using biomarker-based NIT cutoffs yielded estimated prevalence of 0.4%-12.3% (1.0 million-14.5 million) based on FIB-4 cutoffs from 0.90-2.67, and 0.1%-1.9% (0.2-5.0 million) based on APRI cutoffs from 0.50-1.50. Conclusion: Prevalence of NASH among US adults was estimated to range from 1.3% to 4.8% when predicted using imaging-based noninvasive test values for participants with presumed NAFLD, generally aligning with estimates in the literature of prevalence of biopsy-confirmed NASH. Use of biomarker-based noninvasive test values for prediction of NASH yielded a wider range of estimates with FIB-4, and a considerably lower range of estimates with APRI.


Journal of Health Economics and Outcomes Research
Fishman J, et al S6  In the NHANES 2017-March 2020 cycle, MRI-PDFF and liver biopsy were not conducted.Consequently, the following screening steps were modeled:

Journal of Health Economics and Outcomes Research
Step 1: Evidence of ≥3 risk factors of significant fibrosis, including: • AST >20 U/L or AST/ALT ≥1 • Diabetes (self-report of being told by a healthcare professional that one had diabetes or "borderline" diabetes) • Dyslipidemia (total cholesterol ≥200 mg/dL, triglycerides ≥150 mg/dL, LDL-C ≥130 mg/dL, or low HDL-C defined as <50 mg/dL for women and <40 mg/dL for men) • Hypertension (systolic blood pressure ≥130 mmHg or diastolic blood pressure ≥85 mmHg) • Metabolic syndrome (≥3 of: HbA1c ≥5.7% and/or treatment for high blood glucose, waist circumference >35 inches for women and >40 inches for men, hypertension and/or treatment for high blood pressure, triglycerides >150 mg/dL or treatment for high cholesterol, low HDL-C) Step 2: Steatosis reflected by controlled attenuation parameter (CAP) ≥280 dB/m, from VCTE Step 3: Liver stiffness measure (LSM) ≥8.5 kPa, from VCTE In addition to the steps above, three scenarios were modeled varying access to care, reflective of individuals who might be diagnosed with NASH in practice.These included: •

Journal of Health Economics and Outcomes Research
Fishman J, et al

AACE (2022) Screening Algorithm
The scenario based on the AACE (2022) "Cirrhosis Prevention in NAFLD" screening algorithm was informed by the process described in Cusi et al (2022), 10 as replicated below.
In the NHANES 2017-March 2020 cycle, the ELF blood test was not conducted.Consequently, the following screening steps were modeled: Step 1: Evidence of ≥1 of: • Prediabetes or diabetes (self-report of being told by a healthcare professional that one had diabetes, "borderline" diabetes, or prediabetes) • Obesity (BMI >30 kg/m2) and/or ≥2 cardiometabolic risk factors (HbA1c ≥5.7% and/or treatment for high blood glucose, waist circumference >35 inches for women and >40 inches for men, hypertension and/or treatment for high blood pressure, triglycerides >150 mg/ dL or treatment for high cholesterol, HDL-C <50 mg/dL for women and <40 mg/dL for men) • Steatosis on imaging (CAP ≥288 dB/m) and/or elevated aminotransferases (ALT >30 U/L or AST >30 U/L) Step 2: FIB-4 ≥1.30 Step 3: FIB-4 >2.67 or LSM ≥8.0 kPa

Study design
4 Specify the study design in the methods section with a commonly used term (eg, cross-sectional or longitudinal).

5a
Describe the questionnaire (eg, number of sections, number of questions, number and names of instruments used).

Data collection methods 5b
Describe all questionnaire instruments that were used in the survey to measure particular concepts.Report target population, reported validity and reliability information, scoring/classification procedure, and reference links (if any).

NHANES documentation 5c
Provide information on pretesting of the questionnaire, if performed (in the article or in an online supplement).Report the method of pretesting, number of times questionnaire was pre-tested, number and demographics of participants used for pretesting, and the level of similarity of demographics between pre-testing participants and sample population.

NHANES documentation 5d
Questionnaire if possible, should be fully provided (in the article, or as appendices or as an online supplement).

Sample characteristics 6a
Describe the study population (ie, background, locations, eligibility criteria for participant inclusion in survey, exclusion criteria).

NHANES documentation 6b
Describe the sampling techniques used (eg, single stage or multistage sampling, simple random sampling, stratified sampling, cluster sampling, convenience sampling).Specify the locations of sample participants whenever clustered sampling was applied.

NHANES documentation 6c
Provide information on sample size, along with details of sample size calculation.
35, 36, and Figure 1 6d Describe how representative the sample is of the study population (or target population if possible), particularly for population-based surveys.

35, 37
Survey administration 7a Provide information on modes of questionnaire administration, including the type and number of contacts, the location where the survey was conducted (eg, outpatient room or by use of online tools, such as SurveyMonkey).

NHANES documentation 7b
Provide information of survey's time frame, such as periods of recruitment, exposure, and follow-up days.

NHANES documentation 7c
Provide information on the entry process: ->For non-web-based surveys, provide approaches to minimize human error in data entry.
->For web-based surveys, provide approaches to prevent "multiple participation" of participants.

Fishman J,
et al S7 Source: Replicated from Loomba et al (2022). 9SCENARIO ANALYSIS DEFINITIONS MAESTRO-NASH eligibility criteria Scenarios based on the MAESTRO-NASH eligibility criteria were informed by the algorithm depicted in Loomba et al. (2022), 9 as replicated below.
Scenario A: No restriction on access to care • Scenario B: Initial restriction to individuals with ≥1 healthcare visits in last year (NHANES variable HUQ051) • Scenario C: Initial restriction to individuals with ≥1 healthcare visits in last year (NHANES variable HUQ051) and no evidence of other liver disease (excessive alcohol consumption, hepatitis B, or hepatitis C)

Source:
Replicated from Cusi et al (2022). 10Introduction Background 2 Provide a background about the rationale of study, what has been previously done, and why this survey is needed.

Table S1
Definitions of Variables in the Analysis DIQ010Journal of Health Economics and Outcomes Research

Table S2
Distributions of NASH Prevalence vs LSM

Table S2
Distributions of NASH Prevalence vs LSM

Table S1 10c
Report details about how missing data was handled.Include rate of missing items, missing data mechanism (ie, missing completely at random [MCAR], missing at random[MAR]or missing not at random[MNAR]) and methods used to deal with missing data (eg, multiple imputation)., provide information on the model building process, model fit statistics, and model assumptions (as appropriate).anysensitivity analysis performed.If there are considerable amount of missing data, report sensitivity analyses comparing the results of complete cases with that of the imputed dataset (if possible).35,37, 40, and Table2