Development and Validation of the Nonalcoholic Fatty Liver Disease Familial Risk Score to Detect Advanced Fibrosis:A Prospective, Multicenter Study

BACKGROUND & AIMS: Nonalcoholic fatty liver disease (NAFLD)-related fibrosis is heritable, but it is unclear how family history may be used to identify first-degree relatives with advanced fibrosis. We aimed to develop and validate a simple risk score to identify first-degree relatives of probands who have undergone assessment of liver fibrosis who are at higher risk of NAFLD with advanced fibrosis. METHODS: This prospective, cross-sectional, familial study consisted of a derivation cohort from San Diego, California, and a validation cohort from Helsinki, Finland. This study included consecutive adult probands (n = 242) with NAFLD and advanced fibrosis, NAFLD without advanced fibrosis, and non-NAFLD, with at least 1 of their first-degree relatives. All included probands and first-degree relatives underwent evaluation of liver fibrosis, the majority by magnetic resonance elastography. RESULTS: A total of 396 first-degree relatives (64% male) were included. The median age and body mass index were 47 years (interquartile range, 32–62 y) and 27.6 kg/m2 (interquartile range, 24.1–32.5 kg/m2), respectively. Age (1 point), type 2 diabetes (1 point), obesity (2 points), and proband with NAFLD and advanced fibrosis (2 points) were predictors of advanced fibrosis among first-degree relatives in the derivation cohort (n = 220) and formed the NAFLD Familial Risk Score. The area under the receiver operator characteristic curve of the NAFLD Familial Risk Score for detecting advanced fibrosis was 0.94 in the validation cohort (n = 176). The NAFLD Familial Risk Score outperformed the Fibrosis-4 index in the validation cohort (area under the receiver operator characteristic curve, 0.94 vs 0.70; P = .02). CONCLUSIONS: The NAFLD Familial Risk Score is a simple and accurate clinical tool to identify advanced fibrosis in first-degree relatives. These data may have implications for surveillance in NAFLD.

NAFLD with advanced fibrosis (n = 66) and probands without advanced fibrosis (n = 90) (comprising probands without NAFLD [n = 73], and probands with NAFLD but without advanced fibrosis [n = 17]) were enrolled in the study along with their first-degree relatives (Supplementary Figure 1).All subjects were recruited from December 2011 to July 2021.In the Helsinki (validation) cohort, consecutive probands who had NAFLD with advanced fibrosis (n = 21), probands without advanced fibrosis (n = 65) (comprising probands without NAFLD [n = 46], and probands with NAFLD but without advanced fibrosis [n = 19]), and their first-degree relatives were enrolled from November 2017 to March 2021.The baseline characteristics of probands are shown in Supplementary Table 1.All participants completed written informed consent.This study was performed per the ethical principles of the Declaration of Helsinki, and applicable regulatory requirements (University of California San Diego Institutional Review Board 14088).

Inclusion and Exclusion Criteria
All probands with at least 1 first-degree relative enrolled were included in the study.Probands and their first-degree relatives were included if they were adults age 18 years and older in the UCSD (derivation) cohort and aged between 18 and 74 years in the Helsinki (validation) cohort.All probands and their first-degree relatives in both cohorts underwent a standardized medical history, anthropometric measurements, physical examination, biochemical testing, as well as assessment of liver fibrosis and steatosis.All probands and first-degree relatives in both cohorts were assessed for other liver diseases (eg, alcohol-associated liver disease, viral hepatitis, autoimmune hepatitis, and primary biliary cholangitis), and participants with chronic liver disease other than NAFLD were excluded.Alcohol consumption was assessed using the Alcohol Use Disorders Identifications Test and the Skinner questionnaire.
Exclusion criteria (for both probands and relatives) in both cohorts included any of the following: (1) significant alcohol consumption (defined as ≥14 drinks/wk for men or ≥7 drinks/wk for women) within the past 2 years; (2) underlying liver disease including hepatitis B, hepatitis C, hemochromatosis, Wilson's disease, α-1 antitrypsin deficiency, glycogen storage disease, autoimmune hepatitis, and cholestatic or vascular liver disease; (3) evidence of secondary causes or chronic conditions associated with hepatic steatosis including nutritional disorders, and human immunodeficiency virus infection based on laboratory data and clinical history; (4) use of steatogenic drugs, major systemic illnesses; and (5) pregnancy or breastfeeding.

Study Definitions
The presence of NAFLD was defined by either proton density fat fraction of 5.0% or greater 13 or proton magnetic resonance spectroscopy of 5.56% or greater 14 in the majority (89%) of participants, and by controlled attenuation parameter of 288 dB/m or greater 15 in the remainder of participants.
In probands, advanced fibrosis was defined by MRE in the majority of participants, with advanced fibrosis defined by liver stiffness by MRE of 3.63 kPa or greater. 16The minority that were unable to undergo MRE underwent assessment by VCTE (with advanced fibrosis defined as ≥10 kPa 17 ) or histologic assessment (advanced fibrosis defined as fibrosis stages 3-4 by the Nonalcoholic Steatohepatitis Clinical Research Network Histologic Scoring System). 18Based on these criteria, probands were classified as having NAFLD with advanced fibrosis or without advanced fibrosis (comprising probands who either had NAFLD without advanced fibrosis or did not have NAFLD).None of the probands without NAFLD had advanced fibrosis.
In first-degree relatives, NAFLD with advanced fibrosis was defined by previously validated criteria using MRE of 3.63 kPa or greater 16 in the majority (84%) of participants.VCTE or acoustic radiation force impulse was used for all other participants (16%) who were not able to undergo MRE (advanced fibrosis was defined as VCTE ≥10 kPa 17 or acoustic radiation force impulse ≥2.07 m/s 19 ).

Imaging Assessments
Advanced magnetic resonance examinations including MRE and proton density fat fraction using a 3-T research scanner (GE Signa EXCITE HDxt; GE Healthcare, Waukesha, WI) at the UCSD Liver Imaging Group or MRE and proton magnetic resonance spectroscopy using a 1.5-T research scanner (GE Signa HDxt; GE Healthcare) at SYNLAB Kamppi, Helsinki, Finland, were used to assess liver fibrosis and steatosis.

Primary Objective
The primary objective was to create and validate a simple risk score using readily available clinical parameters to detect NAFLD with advanced fibrosis among first-degree relatives of probands (with no other liver disease apart from NAFLD) who had undergone an assessment of liver fibrosis.

Statistical Analysis
Descriptive statistics of participant characteristics were presented at baseline and dichotomized by proband status (presence of NAFLD with advanced fibrosis vs no).Baseline categoric variables were compared with the chi-square test, and continuous variables were compared using a t test or the Wilcoxon 2-sample test where appropriate.Models were derived using the UCSD (derivation) cohort to detect the presence of NAFLD with advanced fibrosis.Univariable and multivariable logistic regression analyses were performed in the UCSD (derivation) cohort for factors associated with NAFLD and advanced fibrosis.Age, sex, race/ethnicity, body mass index (BMI), and T2DM are known risk factors for NAFLD with advanced fibrosis and are assessed easily in clinical practice, hence these factors were selected a priori for multivariable-adjusted logistic regression analysis.Two logistic regression models were constructed, with the first model using all a priori selected factors (model 1), while the second model included only significant predictors (P < .01)from the univariable analysis (model 2).Calibration and discrimination of the models were assessed by the Hosmer-Lemeshow test and the area under the receiver operating characteristic curve (AUC), respectively.The model with the higher AUC and lower Akaike information criterion value in the derivation cohort was selected as the final model.
To develop a simple risk score based on a points system, regression coefficients from the logistic regression model were transformed into scores by rounding to an integer.The projected risk of advanced fibrosis based on a given score was estimated by the following equation: 1/ (1 + e −[−5.613+ 0.917*score] .The agreement between risk estimates based on the points system and the multivariable model was evaluated using a weighted kappa.The score then was validated externally in the Helsinki (validation) cohort.We estimated that the odds ratio for the score would be 2.5 (per unit increase), and 190 first-degree relatives would provide a power of 0.8 with an α value of .05,hence there was adequate power.
Statistical significance was defined as P < .05.All statistical analyses were performed using a graphic user interface for R (The R Foundation for Statistical Computing, Vienna, Austria) and SPSS version 28.0 software (SPSS, Inc, Chicago, IL).

Characteristics of First-Degree Relatives
A total of 396 first-degree relatives (64% female) were included in this study.The median age and BMI were 47 years (interquartile range [IQR], 32-62 y) and 27.6 kg/m 2 (IQR, 24.1-32.5 kg/m 2 ), respectively.The UCSD (derivation) cohort comprised 220 first-degree relatives, of whom 92 were relatives of probands without advanced fibrosis and 128 were relatives of probands who had NAFLD and advanced fibrosis (Table 1).Among first-degree relatives in the UCSD cohort, 48% were parents or offspring of the proband and 52% were siblings of the proband.The Helsinki (validation) cohort comprised 176 first-degree relatives, of whom 119 were relatives of probands without advanced fibrosis, and 57 were relatives of probands with NAFLD and advanced fibrosis.
There was a higher prevalence of NAFLD with advanced fibrosis among first-degree relatives of probands with NAFLD and advanced fibrosis vs first-degree relatives of probands without advanced fibrosis in both the derivation and validation cohorts (Table 1).

Factors Associated With Advanced Fibrosis in First-Degree Relatives in the Derivation Cohort
We examined the factors associated with advanced fibrosis in first-degree relatives in the UCSD (derivation) cohort (Table 2).In unadjusted analysis, a proband with NAFLD and advanced fibrosis (odds ratio [OR], 13.7; 95% CI, 1.8-104.0;P = .01)was a significant predictor of advanced fibrosis in first-degree relatives.Age 50 years and older, obesity (BMI, ≥30 kg/m 2 ), and T2DM also were significant risk factors for advanced fibrosis in the first-degree relatives.The age threshold of 50 years was selected based on Youden's index and the median age of the cohort.
Model 1 included all prespecified factors for multivariable analysis.A proband with NAFLD and advanced fibrosis remained a statistically significant and independent predictor of advanced fibrosis in first-degree relatives (adjusted OR, 6.7; 95% CI, 1.3-35.7;P = .03),after adjusting for confounders (Table 2).
Model 2 included only significant predictors on univariable analysis for NAFLD with advanced fibrosis: age 50 years and older, obesity, T2DM, and family history of a proband with NAFLD and advanced fibrosis (adjusted OR, 5.0; 95% CI, 1.1-23.6;P = .04)(Table 2).Both models calibrated well with the observed risk (P values for Hosmer-Lemeshow goodness-of-fit test: model 1, 0.19; model 2, 0.19).Model 2 was selected as the final model based on a higher area under the receiving operator characteristic curve (AUROC) (0.85 vs 0.81) and lower Akaike information criterion (118.1 vs 125.1) compared with model 1 (Supplementary Table 2).The variance inflation factor of all factors in the final model was less than 2.

Derivation of the Nonalcoholic Fatty Liver Disease Familial Risk Score
Supplementary Table 3 provides the regression coefficients for the 4 variables identified in model 2. The regression coefficients were simplified to an integer scoring system, the NAFLD Familial Risk Score, to simplify the computation of the risk of NAFLD with advanced fibrosis in first-degree relatives and facilitate its use in a clinical setting without the need for a calculator.The NAFLD Familial Risk Score ranges from 0 to 6, and the projected risk of advanced fibrosis corresponding to this score is shown in Supplementary Table 4.A score of 4 points or higher corresponds to a 13% or higher risk of NAFLD with advanced fibrosis.The AUROC of the NAFLD Familial Risk Score in the UCSD (derivation) cohort was 0.85 (95% CI, 0.76-0.92)(Figure 1A).The estimated risk by the multivariable model and the NAFLD Familial Risk Score correlated well (weighted κ, 0.83).
The optimal cut-off value based on Youden's index that maximized both sensitivity and specificity of the score was 4 points, with a corresponding sensitivity, specificity, positive predictive value, and negative predictive value (NPV) of 90.9%, 67.7%, 23.8%, and 98.5%, respectively (Table 3).

Sensitivity Analyses
We analyzed proband status as 3 categories (no liver disease, NAFLD without advanced fibrosis, and NAFLD with advanced fibrosis), and determined that a proband status of NAFLD without advanced fibrosis was not a significant predictor of advanced fibrosis (P = .29)in the derivation cohort.In a sensitivity analysis excluding first-degree relatives without liver disease, the NAFLD Familial Risk Score remained robust in detecting advanced fibrosis in the validation cohort (AUC, 0.89; 95% CI, 0.78-0.99).In a sensitivity analysis of the validation cohort that excluded all participants aged 35 years or younger, the AUC of the NAFLD Familial Risk score remained robust (AUC, 0.91) and higher than that of FIB-4 (AUC, 0.79).Proband age was not associated significantly with advanced fibrosis in first-degree relatives (OR, 1.0; 95% CI, 0.93-1.1;P = .93).

Discussion
Using 2 unique, prospective, well-phenotyped familial cohorts, we developed and validated a clinically applicable risk score to identify advanced fibrosis in first-degree relatives of probands who have undergone an assessment of liver fibrosis.The NAFLD Familial Risk Score comprises age, obesity, T2DM, and family history of NAFLD with advanced fibrosis using a derivation cohort from Southern California.The NAFLD Familial Risk Score performed robustly in a geographically and ethnically distinct external validation cohort from Helsinki, with an AUC, sensitivity, specificity, and NPV of 0.94, 90%, 87%, and 99%, respectively.The NAFLD Familial Risk Score outperformed FIB-4 (AUROC, 0.94 vs 0.70) in first-degree relatives.
The NAFLD Familial Risk Score potentially can be used by family members who are aware of the diagnosis of advanced fibrosis in the proband.Information on how to calculate and interpret the score can be conveyed to first-degree relatives by the proband, or by medical staff to first-degree relatives who accompany the proband to medical appointments.Firstdegree relatives with a score of 4 points or more (corresponding to ≥13% risk of NAFLD with advanced fibrosis) may consider undergoing an imaging-based fibrosis assessment (Figure 2).The NAFLD Familial Risk Score is simple, does not require a calculator, and relies on information that can be derived from clinical history taking.It may be a helpful alternative to FIB-4 for identifying NAFLD with advanced fibrosis among first-degree relatives in clinical practice because it does not require laboratory tests.Several genome-wide association studies have shown an association between fibrosis and single-nucleotide polymorphisms including PNPLA3, TM6SF2, and MBOAT7. 20,21These single-nucleotide polymorphisms are associated with the accumulation of fat in the liver and influence the development of fibrosis, emphasizing the potential for familial clustering of NAFLD with advanced fibrosis. 22,23A recent study described a genetic risk score comprising 11 single-nucleotide polymorphisms and determined that individuals in the top quartile of the score had approximately 3 times the risk of NAFLD with cirrhosis. 24here was a higher prevalence of Hispanics among first-degree relatives of probands with advanced fibrosis in the derivation cohort, consistent with current literature that suggests that Hispanics may have a higher risk of nonalcoholic steatohepatitis. 25However, the score performed well in an ethnically distinct cohort.A substantial number of first-degree relatives in the derivation cohort were female, consistent with the increasing burden of NAFLD among females. 26A recent prospective study determined the prevalence of advanced fibrosis among first-degree relatives of probands with advanced fibrosis and determined predictors of advanced fibrosis in first-degree relatives in the combined cohort (UCSD and Helsinki). 12wever, it remained unclear how family history can be used in a clinical setting to identify first-degree relatives at higher risk of advanced fibrosis.The present study focuses on the development and validation of a simple, clinically applicable score (NAFLD Familial Risk Score) to predict advanced fibrosis in family members.First, we developed the logistic regression model using only the derivation cohort (UCSD), which resulted in a separate set of predictors being identified.We then developed a simple risk score and externally validated this risk score in a geographically distinct validation cohort (Helsinki).The novelty of this study is to provide an actionable score for the practicing clinician.
The strengths of this study included its prospective nature; detailed clinical phenotyping; well-characterized cohort of participants, all of whom underwent assessment of liver fibrosis and with more than 84% using advanced magnetic resonance techniques; external validation of the score in a geographically and ethnically distinct cohort; and unique familial design.However, it was not without limitations.There was a high prevalence of advanced fibrosis among probands, which may affect generalizability in the general population.The score does not include genetic or environmental data, which may provide more granularity, but may increase its complexity.The low number of probands with NAFLD, but without advanced fibrosis, may limit the generalizability of the score in the primary care setting, in which the pretest probability of advanced fibrosis is likely to be lower.Regardless, this was a large study with an independent validation cohort and is unlikely to be replicated.The utility of FIB-4 may be impacted in populations enriched for advanced fibrosis, hence the comparison between the risk score and FIB-4 should be interpreted with caution.
In summary, the NAFLD Familial Risk Score accurately identifies NAFLD with advanced fibrosis in first-degree relatives of probands who have undergone an assessment of liver fibrosis.It is simple, does not require a calculator or extensive laboratory investigations, and may be a helpful alternative to FIB-4 for screening first-degree relatives.These data may have implications for surveillance in NAFLD.

Background
It is unclear how family history may be used to identify first-degree relatives who are at high risk of nonalcoholic fatty liver disease (NAFLD) with advanced fibrosis in the clinical setting.

Findings
Using 2 unique familial cohorts, we developed and validated a clinically applicable risk score to identify advanced fibrosis in first-degree relatives of probands who have undergone an assessment of liver fibrosis.The NAFLD Familial Risk Score comprised age, obesity, diabetes mellitus, and family history of NAFLD with advanced fibrosis, and performed robustly in a geographically and ethnically distinct external validation cohort.

Implications for patient care
The NAFLD Familial Risk Score is simple, does not require a calculator or laboratory tests, and relies on information that can be derived from clinical history taking.It may be a helpful tool for identifying NAFLD with advanced fibrosis among first-degree relatives in clinical practice.Clin Gastroenterol Hepatol.Author manuscript; available in PMC 2024 June 07.

Figure 1 .
Figure 1.(A) Area under the receiver operating characteristic curve (AUROC) of the nonalcoholic fatty liver disease (NAFLD) Familial Risk Score for identifying advanced fibrosis among first-degree relatives in the derivation cohort.(B) AUROC of the NAFLD Familial Risk Score for identifying advanced fibrosis among first-degree relatives in the validation cohort.(C) AUROC of the NAFLD Familial Risk Score vs the Fibrosis-4 index (FIB-4) for identifying advanced fibrosis in the validation cohort.

Figure 2 .
Figure 2. Proposed clinical algorithm to identify first-degree relatives at risk of advanced fibrosis using the nonalcoholic fatty liver disease (NAFLD) Familial Risk Score.T2DM, type 2 diabetes mellitus.*Family history of NAFLD with advance fibrosis Refer to Web version on PubMed Central for supplementary material.27.Shah AG, Lydecker A, Murray K, et al.Comparison of noninvasive markers of fibrosis in patients with nonalcoholic fatty liver disease.Clin Gastroenterol Hepatol 2009;7:1104-1112.[PubMed: 19523535]

Table 1 .
Characteristics of First-Degree Relatives in the Derivation and Validation Cohorts Clin Gastroenterol Hepatol.Author manuscript; available in PMC 2024 June 07.

Table 2 .
Factors Associated With Advanced Fibrosis in First-Degree Relatives in the Derivation (UCSD) Cohort a Model 2 used significant predictors from the univariable analysis.Clin Gastroenterol Hepatol.Author manuscript; available in PMC 2024 June 07.

Table 3 .
Performance Characteristics of the NAFLD Familial Risk Score and the FIB-4 Index for Identifying Advanced Fibrosis fatty liver disease; AUROC, area under the receiving operator characteristic curve; FIB-4, Fibrosis-4; PPV, positive predictive value; NPV, negative predictive value.
a Optimal cut-off point determined by Youden's index in the derivation cohort.