My TSH keeps drifting and I take levothyroxine — is it the weight loss or the gastric emptying?

Hypothyroid on 125 µg levothyroxine for six years, stable TSH around 1.4 mIU/L that whole time. Since starting tirzepatide nine months ago and losing 21 kg, my TSH readings have been 0.9, then 2.6, then 0.4, then 1.9. Same lab each time. My clinician has not changed my dose and says the values are all within range, which they are, but the scatter is unlike anything I saw in the previous six years and I would like to know what is generating it.

Candidate explanations I can construct:

• Delayed gastric emptying is changing how much levothyroxine I absorb, so my effective dose is varying.
• Losing 21 kg means I now need less levothyroxine, so I am mildly over-replaced and the readings are drifting down — except two of the four went up.
• The scatter is just what TSH does and I never noticed before because I was only tested annually.

I would also like to know what the right retest interval is after any dose change, and whether there is a way to estimate what dose a 21 kg lighter version of me should be on rather than adjusting blindly.

Most of your scatter is TSH being TSH, and you are now seeing it because you are measuring four times a year instead of once. The weight loss is a real second-order effect and the gastric emptying is probably the smallest of the three. All three are worth quantifying because the relative sizes are counterintuitive.

How much scatter is normal

Within-person biological variation for TSH is around 19 to 20%, which is high, and analytical variation adds about 5%. The reference change value:

RCV = 2.77 × sqrt(25 + 361) = 2.77 × 19.6 = 54%

So two TSH values must differ by more than about 54% before anything changed. Check your series, taking 1.4 as reference:

• 0.9 versus 1.4: a 36% fall. Inside the noise.
• 2.6 versus 1.4: an 86% rise. Clears the threshold.
• 0.4 versus 1.4: a 71% fall. Clears it.
• 1.9 versus 1.4: a 36% rise. Inside the noise.

Two of four excursions are formally significant, which is more than noise alone would give but not by a large margin over four draws. TSH is also log-distributed, so the RCV framework fits it imperfectly and these figures are indicative. The honest reading: genuine variability, but 0.9, 2.6, 0.4, 1.9 is scatter, not drift. A drift would be monotonic.

Add the diurnal component. TSH follows a circadian rhythm with a nocturnal peak; values in the small hours can be 50% or more above the mid-afternoon nadir, and even within clinic hours an 08:00 draw runs meaningfully higher than a 15:00 one. Four draws at four different times can generate most of your spread by themselves. This is the cheapest thing to fix.

Levothyroxine requirement and weight

The weight-based rule of thumb is about 1.6 µg/kg/day for full replacement. Applied naively:

• Before, if you were 118 kg: 118 × 1.6 = 189 µg.
• After, at 97 kg: 97 × 1.6 = 155 µg.
• Implied reduction: 34 µg, or about 18%.

Notice that both exceed your actual stable dose of 125 µg, which is the first clue that the weight-based rule is a starting-dose heuristic rather than a description of an individual's requirement. It also suggests you retain some endogenous thyroid function, which makes requirement less sensitive to anything.

The more important correction: requirement tracks lean body mass considerably better than total body weight, which is why weight-based dosing performs poorly in obesity. If your 21 kg loss was roughly 75% fat and 25% lean, you lost about 5 kg of the relevant compartment, not 21. So the honest estimate of your requirement change is 5 to 8%, not 18%.

Applied to 125 µg, that is 6 to 10 µg per day — smaller than the increment between tablet strengths and well inside the noise band of a single TSH measurement. Which explains your clinician's position. The situation where it matters more is full replacement after thyroidectomy or radioiodine, with no endogenous reserve, where the buffer is absent.

Gastric emptying and absorption

Levothyroxine is absorbed principally in the jejunum and upper ileum, and its dissolution is favoured by an acidic gastric environment — which is why proton pump inhibitors, calcium, iron, coffee and soy all reduce absorption, and why the fasting-with-water instruction exists.

Delayed gastric emptying, which this class produces reliably, principally shifts time to peak concentration rather than total quantity absorbed. With a plasma half-life of about seven days and daily dosing, plasma concentration is a heavily smoothed average of many doses, so a shift in absorption timing is almost invisible at trough. There is no strong evidence of a clinically important bioavailability reduction from GLP-1 receptor agonists specifically, and no routine dose adjustment is recommended in their labelling.

Where absorption genuinely changes: during vomiting, where a dose may be lost outright, and if the drug changed your morning routine — coffee earlier, a calcium or iron supplement with breakfast because you now eat differently, or a new proton pump inhibitor for reflux. Those are far more plausible sources of a 50% TSH excursion than transit time. Audit the routine rather than the pharmacology.

The retest interval, with the arithmetic

Six weeks, and the number comes from the half-life. Levothyroxine's plasma half-life is approximately seven days. Reaching a new steady state after a dose change requires roughly four to five half-lives:

• 5 × 7 days = 35 days, so five weeks to be within a few per cent of steady state.
• Add time for the pituitary-thyroid axis to re-equilibrate to the new free T4 level, which lags further, and the conventional 6 to 8 weeks falls out.

Testing earlier gives you a value on a rising or falling curve, which is uninterpretable and reliably leads to over-adjustment. This is the single most common error in levothyroxine management and it produces exactly the oscillating series you are describing, so be glad your dose was not changed.

What to actually do

Standardise before adjusting anything: same lab, same time of day, ideally early morning before the dose, fasting, and no biotin supplements beforehand since biotin interferes with many immunoassays and can produce a spuriously low TSH. Take two or three draws under those conditions across three to six months and read the central tendency, not the individual values.

Ask for free T4 alongside at least one draw. TSH alone in a person with a changing body and changing intake is one number describing a two-variable system.

edited 3 Jul 2026 by Dr_Idris_Coulibaly — added the placebo-arm figures

Small mechanical addition on the dosing logistics, since it is the part that most often explains an unstable TSH series and it is entirely fixable.

Levothyroxine has a narrow therapeutic index by the standards of most oral drugs, and its absorption is more fragile than its reputation suggests. Roughly 70 to 80% of an oral dose is absorbed under ideal conditions, and that figure degrades with:

• Food. Taking it with breakfast rather than fasting reduces absorption appreciably. The conventional instruction is 30 to 60 minutes before food, or at bedtime at least three hours after the last meal, which some trials suggest performs as well or better.
• Coffee. Espresso reduces levothyroxine absorption measurably. If your morning routine changed because your appetite changed, this is a plausible and invisible cause of a TSH shift.
• Calcium and iron — including in multivitamins and in fortified foods — bind levothyroxine in the gut. A four-hour separation is the usual advice. Anyone who started iron supplementation because their intake dropped has introduced this without noticing.
• Proton pump inhibitors. Reflux is common during rapid weight loss and PPIs are commonly started for it. They reduce levothyroxine absorption by raising gastric pH.
• Bile acid sequestrants and some fibre supplements. Same binding mechanism.

Given that list, the audit worth doing before touching the dose is: has anything about the timing, the accompanying drink, the supplement stack or the acid-suppression status changed in the last nine months? In someone whose eating pattern has been rebuilt from scratch, the answer is very often yes, and at least one of those changes is usually large enough to account for a 50% TSH excursion by itself.

Two further practical points. Tablet brand and formulation matter more than they should — switching manufacturer occasionally shifts TSH, and a mid-series switch is worth knowing about. And on vomiting: if a dose is vomited within an hour or so, essentially none of it was absorbed, and a week of that during a dose escalation is a real interruption in therapy rather than a rounding error. Whether and how to replace missed doses is a question for the prescriber, and the answer depends on how many were missed.