Levothyroxine is unusually fussy about the stomach it lands in. It is taken fasting, held apart from calcium and iron, and titrated in increments of 12.5 or 25 micrograms against a blood test. A drug whose whole mechanism includes holding the stomach shut for longer is therefore a plausible disturbance, and it is one of the few disturbances in this class that a manufacturer has measured and printed. The wider picture of swallowed drugs taken alongside these injections is mostly reassuring; thyroid replacement is the named exception.
The one dedicated trial was run on the tablet
Forty-five healthy subjects took a single 600-microgram dose of levothyroxine alone, then with the absorption enhancer SNAC, then alongside oral semaglutide 14 mg at steady state. Total thyroxine exposure was 33% higher with oral semaglutide than with levothyroxine alone, while baseline-corrected peak concentration was unaffected. SNAC by itself did not raise total T4 exposure.[1]
Two things about that design carry into practice. The subjects were healthy volunteers taking one large dose, not hypothyroid patients on a daily replacement dose. And the co-formulated tablet is a different delivery problem from a weekly injection, a distinction covered in the comparison of the two routes. A 2025 review of pharmacokinetic interactions across the whole class reached the same short list: metabolic and transporter interactions are essentially absent, and the exposure changes worth monitoring were the oral contraceptive with tirzepatide and levothyroxine with oral semaglutide.[2]
The same trial measured the reverse direction, and that half is rarely quoted. When oral semaglutide was swallowed alongside five placebo tablets, semaglutide exposure fell 34%, with peak concentration down as well.[1] The tablet needs an almost-empty stomach and a small sip of water to be absorbed at all, and so does levothyroxine. Two drugs competing for the same fasting window is not a theoretical conflict; it is the reason the tablet carries dosing instructions that a weekly injection does not.
What the labels carry, and what they leave out
The semaglutide label states the finding directly. Its drug interactions section reports that in a trial with the semaglutide tablet, levothyroxine exposure was increased 33% (90% CI, 1.25 to 1.42), and instructs prescribers to consider increased clinical or laboratory monitoring for medications with a narrow therapeutic index.[3] The clinical pharmacology section repeats that maximum exposure was unchanged.
The tirzepatide label does not name a thyroid drug at all. It warns generally that gastric emptying is delayed, asks for caution with concomitant oral medications, and gives warfarin as its worked example of a narrow therapeutic index.[4] So a reader on thyroid replacement who switches molecules moves from a label naming their medicine to one that does not, without anything having been shown about tirzepatide and thyroxine either way.
Weight loss moves the dose requirement in the same direction
Absorption is only half the mechanism. Levothyroxine is dosed roughly by body weight, so a person who loses a fifth of their mass on a stable prescription is being replaced against a body that no longer exists. Both halves push the same way: more hormone absorbed, and less hormone needed.
A 2024 case report describes the endpoint of that arithmetic — a 62-year-old man who developed thyrotoxicosis from excess levothyroxine exposure during the rapid weight loss produced by tirzepatide.[5] A single case establishes that the event happens, not how often, and the report is published as a teaching example rather than as an incidence estimate.
The one outcome study points against the class reputation
A 2025 target trial emulation used a 15% nationally representative sample of United States Medicare beneficiaries over 65 who had been on a stable levothyroxine dose for at least six months. After propensity matching, 2,384 patients who started a GLP-1 receptor agonist were compared with 2,384 who started an SGLT2 inhibitor, over a median 1.05 years.[6]
Starting the GLP-1 was associated with a higher rate of atrial fibrillation or flutter: hazard ratio 1.46 (95% CI, 1.28 to 1.67). Stroke, including ischemic stroke and transient ischemic attack, did not differ significantly, at HR 1.17 (95% CI, 0.98 to 1.39). The result held when outcomes were restricted to inpatient visits, under an intention-to-treat analysis, under a prevalent new-user design, and with DPP-4 inhibitors substituted as the comparator.[6]
That is the half a summary tends to drop. This drug class is best known for the cardiovascular results in the outcome trials, and in a population defined by an interacting prescription the arrhythmia signal runs the other way. The proposed mechanism is not the drug acting on the heart but thyroid hormone over-replacement arriving through the two routes above. The authors call for further research into the roles of weight loss, TSH fluctuation and levothyroxine dosing, which is the correct weight to put on one retrospective emulation in adults over 65 — a group whose evidence base is thin in general, as the age-specific literature shows.
The monitoring did not follow the mechanism
If over-replacement is the pathway, the countermeasure is a TSH test. A 2026 emulation of the same Medicare data asked whether one was being ordered. Among 5,370 matched adults aged 65 and over with type 2 diabetes on a stable levothyroxine dose — mean age 73.2 years, 71.6% women — approximately 83.0% in both the GLP-1 and SGLT2 arms had a TSH test within one year of follow-up, and the mean time to that test was about 130.5 days in both groups.[7]
No difference at all, in either the rate or the timing, despite the GLP-1 group being the one more likely to need a dose change. The authors describe this as a missed opportunity for weight-responsive thyroid monitoring. A test that would have been ordered anyway is not a response to a new prescription.
What is genuinely unmeasured
A 2026 systematic review searched for studies of thyroid function tests during GLP-1 treatment published between January 2021 and December 2025. Six studies met the criteria, covering 900,225 participants — and only one of the six directly assessed thyroid function tests, reporting a mild decrease in free thyroxine and slight nodule progression at 12 months, with no significant causal effect on Mendelian randomization.[8] The other five measured malignancy outcomes or weight, not thyroid hormone.
That review also reported no increased risk of thyroid tumors across the large cohorts it included.[8] That is a separate question from this page, and the rodent C-cell mechanism, the boxed warning and the conflicting human epidemiology behind it are set out in the article on thyroid cancer risk. Nothing on this page concerns the gland; it concerns the tablet.
Treated hypothyroidism does not appear to blunt the weight loss
The question patients ask first is usually whether the drug will work at all with an underactive thyroid. A 2025 retrospective case-control study compared 53 adults with hypothyroidism on stable levothyroxine against 145 matched controls without thyroid disease, all treated with liraglutide. After a mean 10 months, weight change was −10.8% against −8.9% (P = 0.940), the proportion reaching at least 5% weight loss was 79.2% against 71.0% (P = 0.248), and the share of loss coming from fat-free mass was 29.9% against 33.3% (P = 0.729).[9]
Baseline TSH within the normal range did not correlate with weight outcomes. That is a small single-center study of an older molecule, and it answers the efficacy question rather than the safety one.
Why the signal would be invisible without a test
Mild over-replacement does not announce itself. Its symptoms — a faster resting pulse, poor sleep, heat intolerance, anxiety, a lighter appetite — are the same symptoms a person starting a weight-loss drug is braced for and inclined to attribute to the injection. Palpitations in particular read as a side effect rather than as a thyroid result, which is how a rising free T4 goes unexamined for months.
The 2026 evidence review is explicit that this has not been characterized: five of its six included studies did not systematically evaluate thyroid function parameters at all.[8] What that means in practice is that the population-level frequency of over-replacement on this drug class is unknown, and the only instrument that would detect it in one person is a blood test somebody has to order.
What this changes at a cash-pay intake
Nothing here argues for or against starting treatment. It argues that thyroid replacement is a reason for a blood test on a schedule, and that the schedule does not currently appear in the data. A prescriber who knows the levothyroxine dose can order a TSH before the first injection and again as weight comes off; a form that never asked about thyroid medication cannot, which is one of the gaps described in the telehealth intake article.
One more thing is specific to this market. The compounded semaglutide and tirzepatide sold by these services are not FDA-approved drugs, and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed. The interaction described above is a property of the molecule rather than of the manufacturer, so it applies regardless — but the labeled monitoring language that carries it belongs to the branded products, and a compounded vial arrives without one. How claims on this site are established is set out in the methodology.