FIB-4 Score: Exploring Current Limitations & Improvement Opportunities

The FIB-4 index is a commonly used calculation for guiding the diagnostic strategy and management of liver disease. It helps to rule in—and rule out—the need for further diagnostic interventions to assess liver fibrosis. By integrating age, liver enzyme tests, and platelet counts, it offers a glimpse into the fibrosis stage of a patient, a crucial predictor of liver-related morbidity and mortality.

As the digital age progresses in medicine and advanced data analytics offer new predictive opportunities, we can take a closer look at tools such as FIB-4 and see if alternatives might drive better patient outcomes. This is particularly important because of growing liver disease prevalence: we absolutely would benefit from more targeted methods to find and diagnose liver disease early and accurately.

 

Understanding the Limitations of FIB-4 Score

EHRs Are Not Always Configured to Compute FIB-4 Score

Despite their advanced capabilities, EHRs are not universally configured to automatically compute scores such as FIB-4. The FIB-4 calculation requires inputs of patient age, aspartate aminotransferase (AST), alanine aminotransferase (ALT), and platelet count. While these data are typically available within EHR systems, a provider may need to customize its EHR for automatic FIB-4 computation.

Dependency on Comprehensive Datasets

The FIB-4 score’s effectiveness, as underscored by research from ScienceDirect, points to a nuanced performance that heavily relies on the quality and breadth of patient data available. [1] This study highlights the significant discrepancy in the FIB-4 score’s performance, with a FIB-4 ≥1.3 resulting in false positives in 35% of patients (resulting in unnecessary referrals). Also, the score has varied accuracy across different liver diseases, which emphasizes the need for more inclusive datasets.

The research proposes a refined strategy, suggesting that using the ELF test alone or in conjunction with FIB-4 in indeterminate cases could reduce unneeded referrals. This indicates the crucial role comprehensive datasets play in enhancing diagnostic accuracy and reducing unnecessary medical interventions.

Suboptimal Screening in At-Risk Populations

Research published in Cureus and the International Journal of Molecular Sciences further highlights limitations of FIB-4 in accurately screening at-risk populations. [1] The Cureus study reveals that both FIB-4 and FIB-5 scores exhibit modest sensitivity in predicting advanced fibrosis, with area under receiver operator characteristic (AUROC) curves of 0.712 and 0.655, respectively. Despite their negative predictive values (NPV) greater than 90%, their sensitivity ranges suggest limitations in effectively screening for advanced fibrosis in MASLD (formerly known as NAFLD) patients, especially in resource-limited settings.

Learn why NAFLD is now MASLD, in our recent blog post.

Similarly, the research from the International Journal of Molecular Sciences indicates a notable percentage of false-negative rates among patients with FIB-4 scores < 1.3 who still had significant liver stiffness measurements (LSM ≥ 8 kPa). This discrepancy was more pronounced in patients over 60 years with diabetes mellitus, arterial hypertension, or a body mass index (BMI) ≥ 27 kg/m^2. [2] These findings underscore the challenges of using FIB-4 as a universal screening tool, particularly for at-risk populations, and highlight the necessity for advancements that would accommodate a broader spectrum of patients.

 

Going Beyond The FIB-4 Score

Integration of AI and Machine Learning

AI and ML technologies have the capability to process and analyze vast amounts of data with speed and accuracy. By leveraging common data elements from EHRs, such as laboratory results, diagnoses, vitals, and patient demographics, AI models can identify subtle patterns and markers indicative of liver disease that might be overlooked by traditional diagnostic methods.

Overcoming Current Limitations

One of the significant advantages of integrating AI and ML into liver disease detection and diagnostics is their ability to overcome certain limitations of tools like the FIB-4 score. For instance, while FIB-4 relies on specific laboratory values that may not always be available, an AI model could utilize a broader range of data inputs to provide a more reliable assessment even if certain data are uncharted. Additionally, AI can help mitigate the impact of confounding factors such as age or comorbidities, which can limit the accuracy of traditional scoring systems.

Example 1: Detection Assistance

For screening of other diseases, AI models have the remarkable capability to analyze existing laboratory data in conjunction with other routinely collected patient information within EHRs, such as demographics, vitals, and medical history. For example, an AI model called LGI-Flag, created by Medial EarlySign, uses patient age, sex, and CBC component scores to flag patients with higher apparent risk for lower GI disorders such as colorectal cancer (CRC) and its precursors. By integrating these data points, LGI-Flag identifies patterns and correlations that may not be immediately apparent to human observers. We incorporate LGI-Flag in our Reveal for Lower GI solution. For provider organizations facing long backlogs in their endoscopy suites, Reveal for Lower GI can help to risk stratify patients who are overdue for CRC screening and warrant extra outreach and encouragement to obtain a screening colonoscopy.

A similar approach could support predictive models that highlight patients who exhibit a higher apparent risk for liver diseases. These models would flag patients for closer monitoring or further diagnostic testing, facilitating early intervention. Early detection is crucial as it can significantly alter the disease trajectory, leading to better management and improved outcomes. Moreover, by proactively identifying at-risk individuals, healthcare systems can allocate resources more effectively, ensuring that high-risk patients receive timely and appropriate care.

Example 2: Diagnostic Assistance

Developers are increasingly creating AI models to recognize specific imaging patterns that are indicative of liver conditions such as fibrosis or steatosis. [4] By leveraging advanced machine learning techniques, these algorithms can analyze imaging data from modalities like ultrasound, CT, or MRI with a level of detail and consistency that surpasses manual interpretation. This capability should strengthen non-invasive diagnostic tests to assess liver health, reducing the need for more invasive procedures like biopsies. AI in imaging can lead doctors to make earlier and more accurate diagnoses by detecting subtle changes in liver tissue that might be missed otherwise. Furthermore, AI can process large volumes of imaging data quickly, providing rapid diagnostic support to clinicians and care teams.

These AI-enabled approaches allow for earlier detection and diagnosis of liver disease, even before a patient may experience symptoms, enabling timely intervention and management.

FIB-4 remains a valuable diagnostic measure for liver fibrosis. And by augmenting FIB-4 with new tools powered by clinical AI models, we can further improve the accuracy and cost-effectiveness of diagnosing liver disease.