About the MELD Excluding INR (MELD-XI)
Anticoagulation raises the INR for reasons that have nothing to do with liver function, which inflates a standard MELD score and can wrongly advance an anticoagulated patient's transplant priority. MELD-XI solves that by dropping INR entirely and rescaling the remaining two terms: 5.11 × ln(bilirubin) + 11.76 × ln(creatinine) + 9.44, with both values floored at 1.0 mg/dL, giving a minimum possible score of 9.44. Despite losing a variable, it tracks MELD closely — correlations of 0.90 to 0.95 across the derivation and validation cohorts, with c-statistics only marginally lower. Unlike MELD it has no single universal cut-off; the threshold depends on the population, and much of its modern use is outside hepatology altogether, in heart failure and ventricular assist device assessment.
Formula
MELD-XI = 5.11 × ln(bilirubin mg/dL) + 11.76 × ln(creatinine mg/dL) + 9.44- bilirubin
- Total bilirubin in mg/dL, floored at 1.0.
- creatinine
- Serum creatinine in mg/dL, floored at 1.0.
- Both inputs are floored at 1.0 mg/dL, so the minimum possible score is 9.44 — a normal patient does not score zero, and a score near 9.4 means 'no derangement detected', not 'low risk of nothing'.
- Creatinine's coefficient (11.76) is more than double bilirubin's (5.11), so MELD-XI is driven substantially more by renal than by hepatic dysfunction.
- The constant 9.44 and the coefficients come from normalising the two-variable model onto the same scale as MELD, so scores are broadly comparable between the two indices.
- There is no upper cap in the formula itself, unlike the conventional MELD ceiling of 40 applied in allocation policy.
Interpreting the result
MELD-XI is a continuous index and, unlike MELD, has no single agreed threshold — this is the most important thing to know about interpreting it. Higher scores mean worse combined hepato-renal function and worse short-term survival, but the cut-off that matters depends entirely on the population: transplant assessment, advanced heart failure, ventricular assist device candidacy and adult congenital heart disease cohorts each have their own thresholds in the literature, and a value of, say, 17 carries a different meaning in each. Because comparable MELD and MELD-XI scores were associated with comparable prognosis in the derivation work, a MELD-XI can reasonably be read against MELD-based expectations in a cirrhotic transplant population. In cardiac cohorts it is better treated as a stratifying variable against the specific threshold used in the relevant study — and one that can improve, which itself carries prognostic weight.
| Score | Band | What it means | Action |
|---|---|---|---|
| 9.44 (minimum) | No detectable derangement | Both bilirubin and creatinine at or below 1.0 mg/dL, so both are floored — the lowest score the formula can produce | Read as absence of measured hepato-renal derangement, not as a reassuring risk estimate in its own right |
| Rising score | Increasing hepato-renal dysfunction | Worse short-term survival, driven more by creatinine than bilirubin given the coefficient weighting | Interpret against the threshold published for your specific population rather than a universal cut-off |
What the MELD-XI needs (2 inputs)
- Total bilirubin (mg/dL)
- Values below 1.0 mg/dL are set to 1.0 before the log is taken.
- Creatinine (mg/dL)
- Also floored at 1.0 mg/dL. It carries more than twice bilirubin's coefficient, so renal function dominates the score.
Units. MELD-XI is an mg/dL formula. Convert from SI units before use: bilirubin µmol/L ÷ 17.1 = mg/dL; creatinine µmol/L ÷ 88.4 = mg/dL. Note this is the opposite convention from UKELD, which requires µmol/L.
What it returns
- MELD-XI score
- A continuous index on approximately the same scale as MELD, with a floor of 9.44 when both inputs are at or below 1.0 mg/dL.
How it is calculated
Heuman and colleagues analysed survival in 554 cirrhotic veterans referred for transplant consideration and used logistic regression to derive a 90-day pre-transplant mortality model from bilirubin and creatinine alone, deliberately omitting INR. They then applied a linear regression to normalise the result onto MELD's scale, so that a MELD-XI of 20 means roughly what a MELD of 20 means. Accuracy was tested in a 278-patient holdout group and in an independent set of 7,203 US adults listed for transplant. The design goal was substitutability rather than improvement: the point was to produce a number a clinician could use in place of MELD without recalibrating their intuition about what it signifies.
Facts & figures
| Cohort | Correlation with MELD | c-statistic MELD vs MELD-XI |
|---|---|---|
| Training (n = 554) | r = 0.930 | — |
| Holdout (n = 278) | r = 0.954 | 30 d: 0.939 vs 0.906; 90 d: 0.842 vs 0.829; 180 d: 0.795 vs 0.797 |
| Independent validation (n = 7,203) | r = 0.902 | 90 d survival: 0.857 vs 0.843 (non-cholestatic); 0.905 vs 0.894 (cholestatic) |
At 180 days MELD-XI marginally outperformed MELD in the holdout cohort — the accuracy cost of dropping INR is small and not uniformly in one direction.
Evidence
Derivation — Heuman
2006 · n = 554554 cirrhotic veterans referred for consideration of liver transplantation before 1 December 2003, used as a training group to derive a 90-day pre-transplant mortality model from bilirubin and creatinine while omitting INR, then normalised onto the MELD scale by linear regression.
Yielded MELD-XI = 5.11 ln(bilirubin) + 11.76 ln(creatinine) + 9.44. MELD-XI and MELD correlated at r = 0.930 in training, and comparable scores on the two indices were associated with comparable prognosis.
Holdout and independent validation — Heuman
2006 · n = 7,203A holdout group of 278 cirrhotic veterans referred after 1 December 2003, plus an independent dataset of 7,203 cirrhotic adults listed for liver transplantation in the United States between May and October 2001.
Correlations with MELD were 0.954 and 0.902. In the holdout cohort c-statistics for MELD versus MELD-XI were 0.939 vs 0.906 at 30 days and 0.842 vs 0.829 at 90 days, with MELD-XI marginally ahead at 180 days (0.795 vs 0.797). For 90-day survival in the independent set: 0.857 vs 0.843 in non-cholestatic and 0.905 vs 0.894 in cholestatic disease.
Ventricular assist device cohort — Yang
2012 · n = 255255 patients with advanced heart failure who received a long-term ventricular assist device between 2000 and 2010 (147 pulsatile, 108 continuous-flow), with hepatic dysfunction quantified by both MELD and MELD-XI and grouped around a score of 17.
MELD and MELD-XI correlated at R ≥ 0.901. Liver dysfunction predicted worse survival, but where MELD-XI improved during device support, post-transplant survival matched that of patients without prior liver dysfunction — establishing MELD-XI as a viable alternative in anticoagulated heart failure patients and as a dynamic prognostic marker.
How it compares
MELD-XI vs MELD-Na
Use MELD-Na as the default in a non-anticoagulated cirrhotic patient, and MELD-XI only when anticoagulation makes the INR uninterpretable — MELD-XI trades a small amount of discrimination for freedom from a corrupted input.
MELD-Na keeps INR and adds sodium, so it carries more information about hepatic synthetic function and circulatory dysfunction than MELD-XI's two terms. That advantage evaporates when the INR is pharmacologically driven. In the derivation work, dropping INR cost only a few points of c-statistic against standard MELD — 0.842 versus 0.829 at 90 days — and at 180 days MELD-XI was marginally ahead, so the substitution is defensible when indicated and unnecessary when not.
MELD-XI vs UKELD
MELD-XI and UKELD both depart from standard MELD but in opposite directions — MELD-XI removes a variable to survive a corrupted input, while UKELD adds sodium and recalibrates to a specific national waiting list.
UKELD retains INR and is therefore vulnerable to exactly the anticoagulation problem MELD-XI was built to solve, so an anticoagulated UK patient poses a genuine difficulty that neither score fully resolves. The two also differ in units — UKELD requires µmol/L, MELD-XI mg/dL — and in purpose: UKELD sets a policy threshold for listing eligibility, whereas MELD-XI is a continuous research and risk-stratification index with no fixed cut-off.
Pearls & pitfalls
- The floor of 9.44 catches people out. A patient with entirely normal bilirubin and creatinine scores 9.44, not 0, because both inputs are floored at 1.0 mg/dL.
- There is no universal cut-off. Any threshold you apply must come from literature in the same population — transplant, heart failure, VAD or congenital heart disease thresholds are not interchangeable.
- Creatinine dominates: its coefficient is more than twice bilirubin's, so MELD-XI is more a renal index than a hepatic one, and acute kidney injury moves it sharply.
- Use it only when INR is genuinely unreliable. In a non-anticoagulated patient, standard MELD (or MELD 3.0) retains information MELD-XI discards.
- It corrects for anticoagulation but not for the other things that distort MELD — dialysis, haemolysis, Gilbert's syndrome and sarcopenia-related low creatinine still mislead.
- A falling MELD-XI is meaningful in its own right, particularly after ventricular assist device implantation, where improvement was associated with post-transplant survival matching patients who never had liver dysfunction.
Critical actions
- Establish why the INR is raised before choosing between MELD and MELD-XI — the substitution is justified by anticoagulation, not by convenience.
- State which threshold and which source population you are interpreting the score against, since MELD-XI carries no standalone cut-off.
- Check creatinine trajectory alongside the score; because renal function drives it, a single value in evolving acute kidney injury can misrepresent the baseline.
- In cardiac and VAD populations, repeat the score over time rather than relying on a single pre-operative value — the change carries prognostic information.
- Do not report a MELD-XI as though it were a MELD in allocation documentation; label it explicitly, since the two are scaled to be comparable and are therefore easy to confuse.
Why this score exists
The authors framed this explicitly as a fairness problem in 'sickest first' allocation rather than a modelling exercise. MELD had made US allocation objective, but objectivity fails where an input is corrupted: a patient anticoagulated for a cardiac indication appears sicker than they are, and one whose INR is genuinely deranged by liver failure appears equivalent. Rather than propose an INR correction factor — which would require knowing how much of the prolongation was iatrogenic — they removed the variable and asked what the remaining two could do. The decision to normalise onto MELD's existing scale was equally deliberate: it meant the new score could enter practice without clinicians having to learn a second set of reference points.
About the creator
First author, 2006 derivation study
Derived an INR-free MELD so that patients anticoagulated for cardiac indications would not be assigned falsely high priority.
Limitations
- It has no universally accepted threshold, which makes it harder to act on than MELD and means published cut-offs cannot be moved between populations.
- Dropping INR discards genuine information about hepatic synthetic function, so it is slightly less discriminating than MELD in patients whose INR is not pharmacologically altered.
- Because creatinine carries the dominant coefficient, it behaves substantially as a renal index and can be driven by kidney injury unrelated to liver disease.
- It remains vulnerable to the other well-known MELD distortions — dialysis, haemolysis, Gilbert's syndrome, and low creatinine from sarcopenia or low muscle mass.
- It was derived in a predominantly male US veteran cirrhotic cohort, and much of its subsequent use is in cardiac populations quite unlike that derivation setting.
- It carries no cap in the formula itself, so extreme values produce scores beyond the range where MELD-based allocation intuitions apply.
If you are the patient
MELD-XI is a version of the MELD liver score designed for people taking blood-thinning medication. The standard MELD score includes a clotting blood test called the INR, but blood thinners such as warfarin deliberately raise the INR, which makes the ordinary score look worse than the liver really is. MELD-XI leaves the INR out and uses just two blood results instead — bilirubin, which reflects jaundice, and creatinine, which reflects kidney function. The lowest possible score is about 9.4, so a normal result is not zero. Higher scores mean more strain on the liver and kidneys. There is no single number that means 'good' or 'bad' for everyone, because the score is used in several different situations — including in heart failure and before heart-pump surgery — and each has its own reference points, so your team will interpret it against what is relevant to you.
Frequently asked questions
What is MELD-XI used for?#
It estimates severity of combined liver and kidney dysfunction when the INR cannot be trusted as a marker of liver function — most often because the patient is anticoagulated. It is widely used in liver transplant assessment and in advanced heart failure, ventricular assist device and adult congenital heart disease cohorts.
What is the MELD-XI formula?#
MELD-XI = 5.11 × ln(bilirubin in mg/dL) + 11.76 × ln(creatinine in mg/dL) + 9.44, with both bilirubin and creatinine floored at 1.0 mg/dL.
Why is the lowest MELD-XI score 9.44?#
Because both inputs are floored at 1.0 mg/dL, and the natural log of 1 is zero, leaving only the constant. A patient with entirely normal values therefore scores 9.44 rather than 0.
What is a high MELD-XI score?#
There is no single universal cut-off — that is a genuine feature of the score, not an omission. Thresholds differ by population, so a MELD-XI must be interpreted against the threshold published for the setting in question, whether transplant assessment, heart failure or VAD candidacy.
How accurate is MELD-XI compared with MELD?#
Very close. The two correlated at 0.90–0.95 across derivation and validation cohorts, and c-statistics differed only slightly — 0.842 versus 0.829 at 90 days in the holdout group, with MELD-XI marginally ahead at 180 days.
Should MELD-XI be used in patients who are not anticoagulated?#
Generally no. If the INR genuinely reflects liver function, standard MELD or MELD 3.0 uses that information and discriminates slightly better. MELD-XI's advantage is specifically that it is immune to a pharmacologically raised INR.