Why are ALT and AST such poor surrogates for what is happening in MASH?

Accepted answer

All three of your claims are true, and they fail for different reasons. Taking them in order.

1. Normal ALT does not exclude disease

This is the most firmly established. A large fraction of people with biopsy-proven steatohepatitis, including some with advanced fibrosis, have ALT within the laboratory reference range [1]. Part of this is that the reference ranges themselves were derived from populations that included undiagnosed fatty liver, so "normal" is calibrated against a partly diseased distribution. Proposals to lower the upper limit to roughly 30 U/L for men and 19 to 20 U/L for women exist for this reason and are not universally adopted.

2. High ALT does not indicate severity

ALT reflects hepatocyte membrane injury and turnover, which is an activity signal, not a scar signal. Fibrosis stage - the thing that predicts outcomes - is not what ALT measures. Worse, ALT characteristically falls as disease advances towards cirrhosis, because there is less viable hepatocyte mass to leak enzyme. So the relationship between ALT and severity is not merely weak, it is non-monotonic: low in early disease, high in active steatohepatitis, low again in burnt-out cirrhosis. Any linear reading of ALT as a severity gauge will misclassify the sickest patients as the healthiest.

The AST/ALT ratio partially recovers information here. In uncomplicated fatty liver ALT typically exceeds AST; as fibrosis advances the ratio rises towards and above 1. That directional shift is why AST rather than ALT appears in most fibrosis scores.

3. Changes in ALT do not reliably track changes in histology

This is the one that defeats your proposed coherent position, and it is the reason enzyme falls in trials are not accepted as a surrogate. The correlation between enzyme change and histological change at the individual level is weak. Participants with large ALT falls frequently show no histological improvement, and some with unchanged enzymes show resolution. Enzyme change also responds rapidly to weight loss and to reduced hepatic fat, which can happen within weeks and is largely a steatosis effect, while fibrosis moves over many months if at all.

So a drug can produce a fast, large, real ALT fall by clearing hepatic fat and reducing hepatocyte stress, and still do nothing to scar. That is precisely the scenario the paired histological endpoints described elsewhere in this tag were designed to catch.

Why trials report them anyway

Three legitimate reasons. They are supportive evidence that the drug is doing something hepatic. They are a safety endpoint in their own right, because a drug that raises transaminases is signalling injury - this is exactly how the small-molecule programmes in this space have run into trouble. And they are cheap, continuous and measured at every visit, so they provide a dose-response and time-course picture that biopsy cannot.

What to look at instead

Non-invasive tests, arranged from cheapest to most informative:

• FIB-4. Calculated as age in years multiplied by AST, divided by the product of platelet count in 10^9/L and the square root of ALT. A worked example: age 52, AST 46, platelets 180, ALT 62. Numerator = 52 x 46 = 2392. Square root of 62 = 7.87. Denominator = 180 x 7.87 = 1417. FIB-4 = 2392 / 1417 = 1.69. Conventional cutoffs put below 1.3 as ruling out advanced fibrosis with reasonable negative predictive value, above 2.67 as ruling it in, and the space between as indeterminate - which is where the worked example lands, and where a large fraction of real patients land [2]. In people over about 65 the lower cutoff is usually raised to 2.0 because age is in the numerator.
• ELF score. A proprietary panel combining hyaluronic acid, PIIINP and TIMP-1 - direct products and regulators of extracellular matrix turnover rather than indirect surrogates. Better performance than FIB-4 for advanced fibrosis, at higher cost, with a commonly used threshold around 9.8.
• Vibration-controlled transient elastography, reported in kPa. Measures liver stiffness. Below about 8 kPa argues against advanced fibrosis, above about 12 kPa argues for it. The controlled attenuation parameter from the same device estimates steatosis separately.

The right way to use these is sequentially: FIB-4 first to triage the large low-risk majority out, then ELF or elastography on the indeterminate and high groups. Using them in parallel and picking whichever agrees with your expectation is how people get into trouble.

Any of this applied to your own results needs a clinician. A FIB-4 you calculated yourself is a screening arithmetic exercise, not a diagnosis.

edited 16 Oct 2024 by per_haugen — clarified the distinction between purity and content