FIB-4 Sweep: Primary Care Case-Finder for Silent MASLD

Concept

A batch analytics module that runs nightly against a health system's lab feed. It filters adults showing any of the metabolic features named in the MASLD definition (prediabetes or type 2 diabetes, abdominal overweight/obesity, hypertension, elevated triglycerides, low HDL), plus the risk factors the review highlights (age ≥50, male sex, with a male:female ratio of ~2). For each patient it computes the Fibrosis-4 index — age, AST, ALT, and platelet count, all routine chemistry/CBC values — and produces a clinician-reviewed worklist of patients with elevated FIB-4 and no recorded fibrosis assessment, each with a one-click order for vibration-controlled transient elastography (FibroScan) and a cardiovascular risk review, since CVD remains the leading cause of death in MASLD. Two design changes reflect the newest evidence: the sweep is now explicitly longitudinal — it re-scores on every new chemistry/CBC and tracks each patient's FIB-4 trajectory, because MASLD progression is highly variable and a single baseline value is not a risk stratification — and FIB-4 is positioned strictly as a first-tier filter, never a diagnosis.

Why now

The case-finding rationale keeps strengthening. MASLD is the most prevalent chronic liver disease worldwide, accounts for up to 38% of hepatocellular carcinoma cases, and is projected to drive as many as 135,000 US HCC cases between 2015 and 2030. Advanced fibrosis is the recognized surrogate for the liver-related outcomes that matter (cirrhosis and HCC), and the surveillance literature explicitly flags the absence of validated methods to stratify risk in routine practice — the exact gap a lab-feed case-finder fills. The commercial trigger stands: resmetirom and subcutaneous semaglutide are now conditionally FDA-approved for adults with MASH and moderate-to-advanced fibrosis, so a staged patient is now a treatable one. Older corpus evidence reinforces that the obesity-driven metabolic syndrome components the sweep filters on are the recognized precursors of this population.

Honest limits — and how the design adapts

• FIB-4's false-positive burden is real. In a prospective screening study of 3,378 participants (reference standard: transient elastography ≥8 kPa), FIB-4 at the ≥1.3 cutoff produced 35% false positives, versus 11% for the ELF test. A FIB-4-triggered worklist will over-flag, full stop. The design adapts rather than ignores this: (a) FIB-4 is a filter only, and every flagged patient goes to VCTE confirmation before any referral or treatment decision; (b) where the health system has ELF access, an optional reflex second tier cuts false-positive VCTE volume; (c) age-adjusted cutoffs (higher thresholds for patients over 65) trim over-referral; (d) the worklist is sized and staffed for the known false-positive rate, not an idealized one.
• Most MASLD does not progress. The same surveillance literature notes MASLD carries a lower risk of progression to cirrhosis than other liver diseases — which is precisely why the sweep concentrates on the advanced-fibrosis minority, and why the false-positive discipline above matters: over-flagging a largely indolent condition is the product's main failure mode.
• NIT accuracy varies by phenotype. A multicenter study of 2,156 biopsy-proven MASLD patients found that biochemical pattern (hepatocellular vs. mixed vs. cholestatic, defined by the ALT/ALP ratio) tracks distinct histology and prognosis — cholestatic-pattern patients had nearly double the cirrhosis rate of hepatocellular-pattern patients (10.9% vs. 5.8%) — and that noninvasive test performance differs across these patterns. The sweep already holds ALT and ALP, so it computes the pattern and flags cholestatic-pattern patients for clinician judgment rather than letting FIB-4 be the sole arbiter.
• FIB-4-based composites are not the best MASH detectors. In 108 biopsy-proven patients, MEFIB (MRE + FIB-4) underperformed FAST and MAST for identifying at-risk MASH (AUC 0.671 vs. 0.803/0.799). The claim is therefore bounded: this tool finds candidates for fibrosis staging, not at-risk MASH; activity and drug-eligibility assessment happens downstream with VCTE ± MRE-based scores.

Remaining open questions

• Will payers reimburse VCTE for screening based on an automated FIB-4 trigger without a prior ultrasound finding?
• Does alert fatigue from automated EHR worklists cause clinicians to ignore high-risk flags? (Mitigation: the worklist is clinician-reviewed and volume-capped, not a passive interruptive alert.)
• Does FIB-4 accuracy vary across ethnic groups in ways that would bias the sweep? (The phenotype finding above confirms accuracy is not uniform across subgroups; ethnicity-specific validation is still needed.)
• Does the cost of implementation plus staging outweigh value-based savings from preventing liver failure and transplant?
• What software-as-a-medical-device regulatory requirements attach when an algorithm triggers a diagnostic order?