How many draws do I need before a lab trend means anything? My ALT went 24 to 41

Baseline ALT was 24 U/L. Fourteen weeks into semaglutide it came back at 41 U/L, still inside the lab's quoted range of 0 to 50 but a 71% rise from where I started. I have read that a rise of that size within the reference range is meaningless, and I have also read that a 71% change in anything ought to be taken seriously. Those cannot both be right.

What I would like to understand is the general machinery, not just the answer for ALT. There must be a way to calculate how big a change has to be, for a given analyte, before it is more likely to be real than to be noise — something combining how reproducible the assay is with how much the analyte bounces around within one person from week to week. If that number exists I would rather compute it for each of the tests I am tracking than ask a forum every time a value moves.

Related: is the answer different if I have three points instead of two? Intuitively a trend across three draws should be more convincing than a single jump, but I cannot articulate why in a way that survives contact with the fact that each individual draw is equally noisy.

One statistical phenomenon deserves its own answer here, because it is the reason repeat draws so reliably look better than the one that alarmed you: regression to the mean.

Your true ALT is not a number, it is a distribution. Say your personal set point is 26 U/L with a within-person CV of 25%, so a standard deviation of about 6.5 U/L. Draws from that distribution mostly land between 13 and 39. A value of 41 is a high draw from that distribution — plausible, if uncommon.

Now: you did not select the 41 at random. You noticed it because it was high. Conditional on having selected a value from the upper tail, the expected value of the next independent draw is closer to the mean, purely as a property of sampling. No biological recovery is required. This is why:

• Repeat tests after an alarming result usually improve.
• People attribute that improvement to whatever they changed in between — dropped the dose, cut alcohol, started milk thistle — when the arithmetic predicted it regardless.
• Uncontrolled before-and-after series in people selected for an abnormal baseline consistently show apparent benefit. Selecting on a high baseline guarantees a fall.

The practical defences are unglamorous. Decide the retest interval and the criterion for action before you see the retest, so you are not adjusting the goalposts to fit the number. Change one thing at a time between draws, or you will not be able to attribute anything. And be sceptical of your own narrative: if you dropped alcohol, cut the dose and started a supplement in the same fortnight, a fall in ALT tells you nothing about any of the three.

The same logic runs in reverse and catches people out more subtly. If you selected a starting point that happened to be a low draw, the next value will tend to rise, and you will read a benign return-to-baseline as a deterioration. This is the mechanism behind a good fraction of "my labs got worse on the drug" posts where the baseline was a single flattering value taken during a fasting week.

On the specific ALT question, three confounders that are worth eliminating before you conclude anything hepatic, because all three are common in exactly this population and all three are free to rule out.

Skeletal muscle. ALT is not liver-specific. It is present in muscle, and eccentric or unfamiliar resistance training raises ALT and AST for several days, sometimes substantially. The tell is the AST:ALT relationship together with creatine kinase: a muscle source typically raises AST proportionally more than ALT and comes with a CK well above its own reference limit. If your CK was not measured, that is the cheapest test to add to draw three. People starting resistance training to protect lean mass while losing weight are precisely the people who will see this.

Weight loss itself. This runs in the opposite direction to your result and is worth knowing. In hepatic steatosis, sustained weight loss lowers ALT, and it does so early and reliably enough that the trials in steatohepatitis use it as a supporting endpoint. A rise in ALT during effective weight loss is therefore the less expected direction, which is a modest argument for taking your value seriously rather than dismissing it. It is not an argument for panic: transaminases can also move transiently during rapid mobilisation of hepatic fat.

Everything else that is not the drug. Alcohol in the preceding week, paracetamol at the upper end of its dose range, over-the-counter non-steroidal anti-inflammatories, most bodybuilding and "liver support" supplements, a recent viral illness, and any new prescription. Anabolic androgenic steroid use is a much larger effect than anything discussed here and is common enough in the peptide-buying population that it deserves explicit mention rather than politeness.

Note also that the reference interval you were given is doing you no favours. An upper limit of 50 U/L for ALT descends from reference populations that included people with undiagnosed fatty liver, and lower thresholds — roughly 30 U/L for men and 20 for women — have been argued for on the basis of outcome data rather than population distribution. Under that framing your 41 is not "within range"; it is mildly elevated. This is the reference-interval-versus-decision-threshold distinction in miniature, and it is why "the lab did not flag it" is a weak reassurance.

None of the above is a diagnosis, and a persistent ALT elevation with any risk factor for liver disease is a genuine reason to see a clinician rather than to keep drawing bloods and reasoning about them.