MELD 3.0: what changed from MELD-Na and why it matters

Calculator for this article: our MELD score calculator (MELD 3.0).

The MELD score estimates short-term mortality in end-stage liver disease, and in the United States it sets the order of the liver transplant waiting list. Most clinicians outside hepatology meet it in a different role: as a quick read on how sick a patient with cirrhosis is. Both uses changed with MELD 3.0, and the changes are specific enough to explain in a few paragraphs.

From MELD to MELD-Na

The original MELD combined bilirubin, INR and creatinine. In its validation, the c-statistic for 3-month mortality was 0.87 in hospitalized patients and 0.80 in non-cholestatic ambulatory patients.^[1]

MELD-Na added serum sodium. In the 2008 analysis of the waiting list, both MELD and sodium independently predicted death, with hazard ratios of 1.21 per MELD point and 1.05 per 1-unit decrease in sodium for values between 125 and 140 mmol/L, and the effect of sodium was greater when the MELD score was low.^[2] In other words, hyponatremia flags a sicker patient than bilirubin, INR and creatinine alone suggest, most of all at the lower end of the scale.

What MELD 3.0 changes

MELD 3.0 was developed on the national waiting-list registry and published in 2021.^[3] The changes from MELD-Na:

• Female sex is a new variable (+1.33 points). The authors' reasoning: serum creatinine overestimates GFR in women, and so underestimates their risk of death compared with men at the same creatinine.^[3]
• Albumin is a new variable, bounded between 1.5 and 3.5 g/dL.^[3]
• Two interaction terms: bilirubin with sodium, and albumin with creatinine.^[3]
• A lower creatinine ceiling: 3.0 mg/dL instead of 4.0 mg/dL.^[3]

In the development cohort, MELD 3.0 discriminated slightly better than MELD-Na (C-statistic 0.869 vs 0.862) and correctly reclassified a net 8.8% of patients who died to a higher MELD tier. In the allocation simulation it was associated with fewer waitlist deaths than MELD-Na (7,788 vs 7,850).^[3]

Those are modest numbers, and the paper presents them that way. The change is incremental, not a new model.

The bounds that matter when you calculate it

These rules come from the OPTN allocation policy and are what our calculator applies:^[4]

• Bilirubin, INR and creatinine values under 1.0 are set to 1.0.
• Creatinine over 3.0 mg/dL is set to 3.0, and so is creatinine in a candidate who had two or more dialysis treatments, or 24 hours of continuous veno-venous hemodialysis, in the prior 7 days.
• Sodium is bounded between 125 and 137 mmol/L.
• Albumin is bounded between 1.5 and 3.5 g/dL.
• The minimum score is 6 and the maximum is 40.

Candidates aged 12 to 17 use a variant without the sex term. Our calculator does not implement that variant, and it says so.

A practical consequence of these bounds: a patient on dialysis gets the creatinine cap regardless of the measured value, so recording dialysis status correctly matters as much as the lab value.

A worked example

Take a patient with bilirubin 3.0 mg/dL, INR 1.8, creatinine 1.4 mg/dL, sodium 130 mmol/L and albumin 2.8 g/dL, not on dialysis. Our MELD 3.0 calculator, applying the bounds above, returns 26 for a woman and 25 for a man with identical labs. That one-point difference is the female-sex term at work.

The paper's own scenarios make the same point at the intermediate range of the scale: severe hypoalbuminemia raised MELD 3.0 by one point, corresponding to a 2.4 percentage-point increase in predicted 90-day mortality, and female sex added another point, associated with another 3 percentage-point increase.^[3] At the low end, the paper shows that mild hyponatremia in a patient with a MELD of 12 raised MELD-Na by 6 points, while MELD 3.0 added fewer, because of the bilirubin-sodium interaction.^[3]

The lesson for anyone reading a score in a chart: one or two points can matter, and they depend on inputs that are easy to leave out (sex, albumin, dialysis).

Did it fix the sex disparity?

OPTN implemented MELD 3.0 for adult allocation in July 2023.^[5] An analysis of candidates before and after found that women's transplant odds, lower than men's before the change (OR 0.86), rose immediately after implementation (OR 1.36) and were 1.18 at three months, but by the end of the study period the estimate had returned toward the earlier disparity (OR 0.81).^[5] The same study examined body size as a second, persisting source of disparity. The score narrowed a known bias; it did not remove the structural reasons for it.

What it means outside the transplant center

• Use MELD 3.0 as the current version. It is what allocation uses, and our calculator keeps the original MELD only as a superseded reference.
• Albumin is now an input. If it is not in today's labs, the score is incomplete.
• The score answers a narrow question. It estimates short-term mortality without transplant. It does not grade encephalopathy, varices or frailty, and it does not replace a referral decision. A rising MELD in a patient who has not yet been referred is a reason to talk to hepatology, not a substitute for that conversation.

Where decision support helps

The equation is easy to compute and easy to compute wrong: a missed dialysis flag, an unbounded sodium, an old version. Pogosh offers MELD 3.0 when the note describes cirrhosis or decompensation, shows the inputs it needs, and applies the published bounds. It does not run the calculator on its own, and it does not decide transplant candidacy. The clinician confirms the values and makes the referral call.