Original Research

CLivD score modifies FIB-4 performance in liver fibrosis detection in the US general population

Abstract

Background and aims Steatotic liver disease (SLD) is a growing global concern. The Chronic Liver Disease (CLivD) risk score predicts liver-related outcomes in the general population using easily accessible variables with or without laboratory tests (CLivDlab and CLivDnon-lab). We assessed CLivD’s associations with liver steatosis, fibrosis and its combined performance with fibrosis-4 (FIB-4) for advanced fibrosis detection.

Methods Using the National Health and Nutrition Examination Survey data (2017–2020), 3603 participants aged 40–70 years with valid liver stiffness measurements (LSMs) were included. Advanced fibrosis was defined as LSM ≥12 kPa, and SLD as controlled attenuation parameter ≥288 dB/m.

Results Significant associations were found between CLivD and SLD and advanced fibrosis. CLivDlab had an area under the curve (AUC) for advanced fibrosis of 0.72 (95% CI 0.68 to 0.77), while CLivDnon-lab had an AUC of 0.68 (95% CI 0.64 to 0.72), both slightly higher than FIB-4 (AUC 0.66, 95% CI 0.60 to 0.72). Among participants without obesity, AUC of CLivDlab was 0.82 (95% CI 0.76 to 0.88) and AUC of CLivDnon-lab was 0.72 (95% CI 0.65 to 0.79). The CLivD score improved FIB-4’s AUC for advanced fibrosis detection from <0.5 at minimal CLivD scores to >0.8 at high CLivD scores. A sequential CLivD→FIB-4 strategy outperformed universal FIB-4 testing, enhancing specificity from 72% to 83%, with sensitivity at 51%–53%. This strategy identified a subgroup with a 55% prevalence of advanced fibrosis, while 47% had minimal-risk CLivD scores, eliminating the need for FIB-4 testing.

Conclusions The CLivD score, designed for predicting liver-related outcomes, effectively identifies liver steatosis and advanced fibrosis in the general population. Combining CLivD with FIB-4 enhances advanced fibrosis detection accuracy. The CLivD score could enhance population-based liver fibrosis screening, optimising resource allocation.

What is already known on this topic

  • Steatotic liver disease and associated liver fibrosis is on the rise globally, emphasising the need for accurate early detection methods.

  • The user-friendly Chronic Liver Disease (CLivD) score has shown potential in predicting liver-related outcomes.

What this study adds

  • This study demonstrates that the CLivD score can be used to identify liver steatosis and advanced fibrosis.

  • When combined with fibrosis-4, it significantly enhances advanced fibrosis detection accuracy.

How this study might affect research, practice or policy

  • These findings offer a promising approach for population-based liver fibrosis screening programmes, enabling efficient resource allocation, cost reduction and improved patient outcomes.

  • Further research and cost-effectiveness analyses are warranted to explore the CLivD score’s applicability across diverse populations and age groups.

Introduction

Morbidity and mortality related to chronic liver disease is rising in many countries.1 Among the chronic liver diseases, steatotic liver disease (SLD) stands out as the most prevalent. There is an urgent need to develop user-friendly, non-invasive methods to identify evolving chronic liver disease during its asymptomatic phase, prior to the manifestation of end-stage complications.

The recently devised Chronic Liver Disease (CLivD) score addresses this need by predicting future liver-related clinical outcomes in the general population.2 Based on widely accessible risk-factor information such as age, sex, diabetes, smoking, alcohol consumption and waist-to-hip ratio, with or without gamma-glutamyltransferase (GGT), the CLivD score has potential in identifying high-risk individuals who may warrant subsequent liver fibrosis assessment using non-invasive tests such as fibrosis-4 (FIB-4).2 3

However, crucial aspects remain unexplored. The associations between the CLivD score and liver steatosis or fibrosis have yet to be investigated in a cross-sectional fashion. Additionally, data on the effectiveness of combining the CLivD score and FIB-4 for advanced liver fibrosis detection are scarce. Specifically, the focus lies on the potential use of the CLivD score to identify individuals at risk who require fibrosis testing through FIB-4.

In this study based on a US general population sample, we analyse the associations between the CLivD score and liver steatosis and fibrosis assessed using transient elastography. Furthermore, we assess the potential of the combined CLivD score and FIB-4 approach for the detection of advanced liver fibrosis. The outcomes hold substantial implications for the formulation of future strategies for population-based liver fibrosis screening.

Methods

Source population

This study used data from the National Health and Nutrition Examination Survey (NHANES) conducted between 2017 and 2020. NHANES is a cross-sectional, nationally representative survey conducted by the National Center for Health Statistics (NCHS), which collects data on various health and nutritional aspects of the US population.4 The survey employs a complex, multistage sampling design to ensure its representativeness. Due to COVID-19, NHANES suspended field operations in March 2020, truncating the 2019–2020 cycle. Data from 2019 to March 2020 were merged with the 2017–2018 cycle for a nationally representative 2017−March 2020 prepandemic dataset.

Demographic, anthropometric and clinical data were collected during the NHANES examinations through questionnaires, physical examinations and laboratory tests. Liver stiffness measurements (LSMs) and the controlled attenuation parameter (CAP) using the FibroScan model 502 V2 Touch were measured on all participants aged 12 years and older, except those who were unable to lie down on the exam table, pregnant, had an implanted electronic medical device or wearing a bandage or had lesions on the right side of their abdomen by the ribs.

Participants

We included individuals aged between 40 and 70 years who participated in NHANES 2017–2020 and had available complete LSMs, that is, individuals had to comply with a fasting time of at least 3 hours and have ≥10 stiffness measurements with an IQR of <30% from the median.

We excluded participants with incomplete LSMs, missing data to calculate the CLivD score or FIB-4, or positive hepatitis B s-antigen (n=30) or hepatitis C RNA (n=72). Among individuals within the age range 40–70 years (n=4905), 1302 were excluded by these criteria.

Covariates

Alcohol intake was evaluated based on a quantity-frequency questionnaire, and we calculated the average number of weekly drinks for each participant. One standard drink in the USA is considered to contain 14 g of pure ethanol. Data on current smoking were based on self-reports or, if self-report data were missing, an elevated serum cotinine level (>3.08 ng/mL).5 Body mass index (BMI), waist and hip circumferences were measured at baseline. Diabetes was defined by either a fasting blood glucose ≥7.0 mmol/L, glycated haemoglobin >6.5% or a previous known diabetes diagnosis. We also collected data on GGT, alanine aminotransferase (ALT), aspartate aminotransferase (AST) and platelet count measurements.

Risk scores

The non-laboratory CLivDnon-lab score was calculated based on age, sex, smoking status (current vs previous/never), weekly alcohol use, waist-to-hip ratio and diabetes (yes vs no) as previously described.2 The laboratory version, CLivDlab, also included GGT (U/L). Both CLivD scores were categorised into subgroups (minimal, low, intermediate and high risk) based on previously reported thresholds.2 FIB-4 was calculated based on age, ALT, AST and platelet count as previously described.6 FIB-4 was categorised into low, intermediate and high subgroups using the cutoffs 1.3 and 2.67 suggested by both European Association for the Study of the Liver (EASL) and American Association for the Study of Liver Diseases (AASLD).7 8

Liver steatosis and fibrosis

We considered an LSM ≥12 kPa as the primary outcome measure defining advanced liver fibrosis. As secondary outcomes, we also considered other established LSM cutoffs: 8, 10, 15 and 20 kPa. Liver steatosis was evaluated using the CAP in dB/m with median and IQR calculated for each participant. SLD was defined by a CAP value ≥288 dB/m.8

Statistical analyses

To compare the groups, we used the χ2 test or t-test, as appropriate. Correlations were assessed using the Spearman’s method. Associations between the CLivD score and risk of steatosis or fibrosis were assessed using locally estimated scatter plot smoothing. Performance measures included area under the curve (AUC), sensitivity, specificity and positive and negative predictive values. To evaluate possible effect modification of FIB-4 performance by the CLivD score, we calculated the covariate-specific AUC of FIB-4 for presence of fibrosis using the npROCRegression package in R. The performance of a sequential combined use of CLivD and FIB-4 scores was evaluated in terms of sensitivity and specificity in a strategy where individuals with a CLivD score above a defined threshold are referred to subsequent FIB-4 testing. Here, a CLivD score above the threshold for low risk or intermediate risk and FIB-4 >1.3 was considered test positive. This sequential strategy was compared with a strategy of performing FIB-4 testing in the entire study population. Statistical significance was defined as a two-tailed p value of <0.05. Data were analysed using R software, V.4.3.1.

Results

The study comprised 3603 participants with a mean age of 55 years and with 52% women (table 1). Some 23% had diabetes, average alcohol use was 3.8 units per week, mean BMI was 30.3 kg/m2 and 25% were current smokers. The mean LSM was 5.9 kPa, and 124 (3.4%) had an LSM of ≥12 kPa. There were statistically significant differences in almost all demographic variables between participants with or without an LSM ≥12 kPa (table 1). The proportion of participants belonging to the intermediate-risk or high-risk CLivDlab group was 3.3% among those with an LSM <12 kPa, compared with 17% among those with an LSM ≥12 kPa. Findings were similar regarding CLivDnon-lab. Interestingly, 47% of participants with an LSM ≥12 kPa belonged to the low-risk FIB-4 group, while only 13%–15% of participants with an LSM ≥12 kPa belonged to the minimal-risk CLivD groups.

Table 1
|
Demographics of the study population

Of participants, 402 (11.2%) had an LSM ≥8 kPa, 207 (5.7%) ≥10 kPa, 75 (2.1%) ≥15 kPa and 45 (1.2%) ≥20 kPa. SLD was observed in 1504 (41.7%).

Detection of SLD

The correlation coefficient between CLivDlab and CAP was 0.26 (p<0.001), and between CLivDnon-lab and CAP, 0.25 (p<0.001). The risk of SLD increased along with the CLivD score from 32% in the minimal-risk CLivDlab group to 60% in the high-risk CLivDlab group (figure 1). The AUC value in detection of SLD for CLivDlab was 0.625 (95% CI 0.607 to 0.643), and for CLivDnon-lab, 0.614 (95% CI 0.596 to 0.632).

Figure 1
Figure 1

Locally estimated scatter plot smoothing estimated risk of steatosis (panels A and B) and advanced fibrosis (panels C and D) according to the CLivDlab and CLivDnon-lab scores. The shaded area represents 95% CIs. The dashed vertical lines show the cutoffs of the CLivD score separating minimal-risk, low-risk, intermediate-risk and high-risk groups. CLivD, Chronic Liver Disease score; LSM, liver stiffness measurement.

Detection of advanced fibrosis

The correlation coefficient between CLivDlab and LSM was 0.23 (p<0.001), and between CLivDnon-lab and LSM, 0.18 (p<0.001), whereas it was 0.05 between FIB-4 and LSM (p=0.002) (online supplemental figure 1). The risk of LSM ≥12 kPa increased almost linearly along with CLivDlab (figure 1), while there was large uncertainty at the higher end of CLivDnon-lab scores, as evidenced by the wide CI.

The prevalence of LSM ≥12 kPa increased from 1.1% in the minimal-risk CLivDlab group to 17.9% in the high-risk CLivDlab group (table 2). Similar trends were seen using alternative LSM cutoffs (online supplemental figure 2). The trends were similar, although less consistent between CLivDnon-lab groups and LSM (table 2 and online supplemental figure 3).

Table 2
|
Prevalence of liver fibrosis within CLivD risk categories using different cutoffs of LSM

The AUC value in detection of LSM ≥12 kPa for CLivDlab was 0.723 (95% CI 0.681 to 0.765), and for CLivDnon-lab, 0.679 (95% CI 0.638 to 0.720), slightly higher than for FIB-4, 0.661 (95% CI 0.604 to 0.719). Sensitivities and specificities for the CLivD scores are shown in figure 2. The sensitivity at low/intermediate/high CLivDlab scores was 85%, while the specificity at high CLivDlab scores was 98% (figure 2).

Figure 2
Figure 2

Sensitivity and specificity of the CLivDlab (upper panel) and CLivDnon-lab (lower panel) scores in detection of advanced liver fibrosis (liver stiffness measurements ≥12 kPa). The dashed vertical lines represent the cutoffs of the CLivD score separating minimal-risk, low-risk, intermediate-risk and high-risk groups, and the results for sensitivity/specificity at those cutoffs. CLivD, Chronic Liver Disease.

Among 2003 participants without obesity (BMI <30 kg/m2), AUCs in detection of LSM ≥12 kPa for CLivDlab was 0.821 (95% CI 0.759 to 0.883), and for CLivDnon-lab, 0.722 (95% CI 0.649 to 0.795).

Combined use of CLivD and FIB-4

The AUC value of FIB-4 in detection of LSM ≥12 kPa improved along with higher CLivD scores, suggesting effect modification by CLivD of FIB-4’s discriminatory performance (figure 3). Similarly, within the minimal-risk CLivDlab group, prevalences of LSM ≥12 kPa were <1.5% across all FIB-4 risk strata (table 3). In contrast, in the higher-risk CLivDlab groups, the prevalence of LSM ≥12 kPa varied from 2% to 55% depending on FIB-4 (table 3).

Figure 3
Figure 3

The area under the curve (AUC) of fibrosis-4 in detection of advanced liver fibrosis (liver stiffness measurements ≥12 kPa) across the spectrum of CLivDlab (upper panel) and CLivDnon-lab (lower panel) scores. The dashed lines represent 95% CIs to the AUC estimate. CLivD, Chronic Liver Disease.

Table 3
|
Prevalence of advanced liver fibrosis (LSM ≥12 kPa) according to CLivD and FIB-4 risk categories

Compared with a strategy of screening the entire study population using FIB-4 for an LSM ≥12 kPa, targeting FIB-4 only to those with a CLivDlab or CLivDnon-lab score above the threshold for low risk (sequential strategy) improved overall specificity from 72% to 83% while maintaining sensitivity at 51%–53% (table 4). These improved measures were observed under the sequential strategy despite that 36%–47% of the population were classified as being at low risk based on the CLivD score alone without a need for subsequent FIB-4 testing.

Table 4
|
Performance measures for detection of advanced liver fibrosis, according to two liver stiffness cutoffs (8 and 12 kPa), of FIB-4 for all and a sequential combined strategy where FIB-4 is performed to those with a CLivD score above a defined threshold

Among 2003 participants without obesity (BMI <30 kg/m2), the sensitivity of the sequential CLivD→FIB-4 strategy was 72% (the same for CLivDlab and CLivDnon-lab), and the specificity was 78%–81%.

Discussion

While the CLivD score was originally developed to predict future liver-related outcomes in the general population, the present study shows that it also holds promise in identifying both liver steatosis and advanced liver fibrosis. Furthermore, the integration of the CLivD score substantially improved the discrimination performance of FIB-4 in the detection of advanced liver fibrosis. By sequentially applying the CLivD and FIB-4 scores, we were able identify a subgroup of individuals in the US general population with a prevalence of advanced liver fibrosis as high as 55%. Conversely, approximately 47% of the population exhibited a CLivDlab score categorised as minimal risk, resulting in an exceptionally low prevalence of advanced fibrosis (<1.5%) regardless of the FIB-4 score and thus rendering FIB-4 testing unnecessary for this particular group. These findings underscore the pivotal role of the CLivD score as the initial step in early liver disease identification and emphasise its synergistic potential when combined with FIB-4.

Study strengths include the use of the nationally representative NHANES dataset, enhancing the generalisability of our results. The comprehensive data collection methods, including LSMs, laboratory tests and clinical assessments, ensured robust analyses. The use of established cutoffs for LSMs and risk scores further facilitated clear risk stratification.

However, we acknowledge certain limitations inherent to our study. The cross-sectional design restricts our ability to establish causal relationships, and the reliance on self-reported data, such as alcohol consumption and smoking, may introduce recall bias. Additionally, the use of surrogate markers for liver fibrosis rather than histological confirmation may introduce measurement error. Furthermore, our study focused on a specific age range (40–70 years), which may limit its generalisability to younger or older populations. This age range was selected because it corresponds to the demographic for which the CLivD score was originally designed.

The performance of CLivD for detecting advanced fibrosis was improved among participants without obesity. We speculate that this phenomenon may be attributed to the diminished accuracy of LSMs in individuals with obesity, thereby increasing the likelihood of a false positive result when LSM ≥12 kPa is observed in this subgroup. Obesity-related false positive LSM values are probably especially common when using lower LSM cutoffs such as 8 kPa. Higher LSM cutoffs (12–15 kPa) have recently been advocated for ruling in advanced liver disease.7

Previous research has highlighted the suboptimal discriminatory performance of the FIB-4 test in detecting advanced liver fibrosis and predicting liver-related outcomes in the unselected general population.9 10 These studies found that the sensitivity of FIB-4 was poor in unselected individuals but improved when applied to those with specific risk factors for liver disease. Our study revealed analogous results, demonstrating that FIB-4 performance improved along with increasing CLivD scores. However, these earlier studies primarily assessed the performance of the FIB-4 test and its variations based on pretest probability determined by various risk factors. They did not comprehensively evaluate the entire screening process, which includes how the target risk group is defined—an aspect essential to optimising screening effectiveness. There are both false positive and false negative rates associated with defining the target risk group.11

In our investigation, we discovered that when compared with a strategy of applying FIB-4 to all individuals, targeting FIB-4 testing specifically to those with CLivD scores falling into the low-risk/intermediate-risk/high-risk category led to an improvement in specificity while maintaining consistent sensitivity levels. This refined approach enhances the overall accuracy of liver fibrosis detection and highlights the significance of carefully defining the target risk group within the screening process.

Our findings align with current guidelines that discourage the use of non-invasive tests like FIB-4 for liver fibrosis screening in the unselected general population, particularly in individuals without risk factors.7 8 Instead, the recommendation is to target screening towards individuals with specific risk factors for liver disease to increase pretest probability.7 8 However, the challenge lies in identifying the ideal target group for screening. We previously found that defining the target group by the CLivD score was associated with substantially better sensitivity than defining the target group by traditional risk factors considered in isolation.11

One distinctive feature of the CLivD score is its reliance on widely available variables, with the CLivDnon-lab score not even requiring a single blood test. This accessibility means that individuals can easily calculate their CLivD score using a web calculator, even without healthcare contact. Implementing this sequential approach (CLivD→FIB-4) in population-based screening programmes for liver fibrosis holds the potential to optimise resource allocation, thereby reducing healthcare costs while effectively identifying those at risk. Since the CLivD score is a continuous variable, it allows choosing the cut-off according to the desired sensitivity and specificity level.

To comprehensively evaluate the financial implications of incorporating CLivD-based risk assessment into routine healthcare screening programmes, conducting cost-effectiveness analyses is imperative. While the CLivD score was originally derived in a Finnish population and externally validated in various populations,2 12 further investigation in diverse populations and different age groups is warranted to ensure its applicability.

In conclusion, our study sheds light on the potential of the CLivD score, extending its original prognostic scope to include the identification of liver steatosis and advanced fibrosis. When combined with FIB-4, it significantly enhances the accuracy of advanced fibrosis detection. These findings offer a pathway towards more efficient and targeted strategies in population-based liver fibrosis screening, ultimately leading to early intervention and improved patient outcomes.