Article — Maddrey Discriminant Function Calculator
Maddrey Discriminant Function: Severity in Alcoholic Hepatitis
This article is educational only. It does not provide medical advice and is not a substitute for diagnosis or treatment by a licensed physician. Maddrey discriminant function interpretation, corticosteroid decisions, and transplant referrals in alcoholic hepatitis require a complete clinical assessment by a hepatology team. Seek immediate medical care for jaundice, confusion, GI bleeding, or signs of liver failure.
The Maddrey discriminant function (DF) grades severity in acute alcoholic hepatitis. The formula is DF = 4.6 × (PTpatient − PTcontrol) + bilirubin (mg/dL). A value at or above 32 identifies severe disease and is the conventional threshold for corticosteroid therapy per AASLD and EASL guidance.
William Maddrey introduced the score in 1978 from a randomized trial at Johns Hopkins. Patients with DF ≥ 32 had 28-day mortality near 35 percent untreated and around 25 percent on prednisolone. Forty years later, the score is still the bedside standard for the steroid decision.
What is Maddrey's discriminant function?
The Maddrey DF is a two-input score: prothrombin time and total bilirubin. PT reflects hepatic synthetic function (the liver's ability to make clotting factors II, VII, IX, X), and bilirubin reflects excretory function. Both rise as alcoholic hepatitis worsens, so the combined score climbs in step with disease severity.
The score is part of the AAH workup, not the diagnosis itself. Diagnosis still requires history of heavy alcohol use, jaundice, an AST that is 2 to 6 times upper limit and exceeds ALT (AST/ALT > 1.5), exclusion of other liver disease, and sometimes liver biopsy in unclear cases.
The Maddrey DF formula in plain math
DF = 4.6 × (PT_pt − PT_ctrl) + bili mg/dLbili (mg/dL) = bili (μmol/L) ÷ 17.1 unit conversionDF < 32 mild / moderateDF ≥ 32 severe — consider steroidsThe coefficient 4.6 came from a regression in the 1978 trial — it gave the best mortality discrimination at the time and has not been re-tuned since. Bilirubin contributes mg/dL directly with no multiplier, so a 1 mg/dL change in bilirubin moves DF by 1.0; a 1-second change in ΔPT moves it by 4.6.
The Maddrey DF ≥ 32 threshold
The 32 cutoff is the most consequential number in the score. Below it, mortality is under 10 percent at 28 days and steroids show no meaningful benefit. At or above 32, untreated mortality jumps to 35 to 45 percent, and corticosteroid therapy reduces it to roughly 25 to 35 percent. The threshold preserved its place in current AASLD and EASL guidance through the STOPAH trial and subsequent meta-analyses.
Maddrey's 1978 paper enrolled only 55 patients but produced one of the most enduring scores in hepatology. Multiple competing scores have been proposed (ABIC, Glasgow alcoholic hepatitis score, MELD), but DF ≥ 32 remains the steroid-decision standard because it needs only two cheap labs and discriminates the high-mortality group reliably.
Corticosteroids and the Maddrey DF cutoff
The standard regimen for severe alcoholic hepatitis is prednisolone 40 mg orally daily for 28 days, followed by a 2-week taper. The STOPAH trial (2015) confirmed a short-term mortality reduction at 28 days but no benefit at 1 year — partly because non-treatment-related complications (sepsis, hepatorenal syndrome, ongoing alcohol use) dominate later outcomes.
Active uncontrolled infection (especially spontaneous bacterial peritonitis), active gastrointestinal bleeding, acute pancreatitis, hepatorenal syndrome, and uncontrolled diabetes are AASLD-listed contraindications. Screen for infection at presentation and at any clinical change. Pentoxifylline was previously used as an alternative but offers limited mortality benefit per STOPAH.
Maddrey DF vs MELD: when to use each
The Model for End-Stage Liver Disease (MELD) was developed for cirrhosis severity and transplant priority. It uses creatinine, bilirubin, and INR, capturing renal dysfunction that DF misses. Both scores apply in alcoholic hepatitis, and many centers use them together: DF for the steroid decision, MELD for transplant priority and overall prognosis.
MELD has higher accuracy at predicting 90-day mortality. DF is faster to compute, requires fewer labs, and aligns historically with the steroid-trial evidence base. A MELD of 21 or higher in a patient with DF ≥ 32 suggests very severe disease and triggers a hepatology-led discussion about transplant evaluation.
Maddrey DF and the Lille score for steroid response
The Lille model is computed at day 7 of corticosteroid therapy. It combines baseline age, albumin, PT, bilirubin, and creatinine with day-7 bilirubin to predict 6-month survival. A Lille score ≥ 0.45 identifies non-responders. AASLD recommends discontinuing steroids in non-responders and considering transplant referral.
Maddrey DF and Lille play complementary roles: DF decides whether to start steroids, Lille decides whether to continue them. Track both on a patient's chart through the 28-day course.
Limits of the Maddrey discriminant function
The DF was tuned with PT in seconds, not INR. Different lab thromboplastin reagents produce different PT controls, so using a non-local control can throw the score off by several points. Many centers now standardize on INR and use (INR − 1) × 10 as a proxy for ΔPT — accurate enough for most patients but not exactly the historical formula.
DF also misses renal dysfunction. Hepatorenal syndrome raises mortality sharply but does not affect PT or bilirubin in the short term. A patient with DF = 30 and a rising creatinine may be at higher risk than the score suggests; cross-check with MELD or daily kidney function.
A third limit is the static snapshot. DF reflects severity at admission but does not track change. Falling bilirubin during the first week of treatment is a favorable sign; the Lille score formalizes that early-response observation. Two patients can present with identical DF and follow very different trajectories — protein synthesis, encephalopathy grade, and infection burden all matter.
Common Maddrey DF mistakes
- Using INR directly in the formula instead of PT seconds — the coefficient 4.6 does not apply
- Using a generic 12-second control instead of the local lab reference
- Forgetting μmol/L → mg/dL conversion (divide by 17.1) when reading SI units
- Starting steroids without infection screen — sepsis is a hard contraindication
- Stopping at the DF — also compute MELD and check renal function
- Treating the DF as diagnostic — it grades severity once AAH is already established