The Switch a GLP-1 Flips Inside the Beta Cell
The Switch a GLP-1 Flips Inside the Beta Cell
Salk researchers found one phosphate tag on one protein that turns on more than a thousand genes after a long exposure to a GLP-1 drug. It’s rat cells and mice, not people. Here’s what it shows, and what it doesn’t.
Give a beta cell a GLP-1 drug for one hour and 184 genes change. Leave the drug on for 16 hours and 1,065 genes change. That is the observation at the center of a study from the Salk Institute, published in PNAS in March 2026, and the team went looking for what makes the long exposure so different from the short one.
What they found is a switch. A protein called Med14, part of a large machine called Mediator that helps a cell read its genes, picks up a phosphate tag at one specific spot when the GLP-1 receptor is active. Tag on, and the beta cell reprograms itself: more activity at the insulin gene, faster fuel burning, a gene profile the authors compare to what they see in calorie-restricted mice. Tag blocked, and most of the long-exposure response disappears.
One thing before the details. Everything here was done in a rat beta-cell line and in mice. No human cells, no human islets, no patients. The study’s first author, Sam Van de Velde, told The Scientist it was “just one cell type.” Keep that in your pocket while you read.
Here is the chain, one step at a time.
What the study shows
Worth naming up front: the work was funded by the National Institutes of Health and by philanthropy, including Breakthrough T1D (the organization formerly called JDRF) and the Helmsley Charitable Trust. No drug company is listed as a funder of the study. The senior author is Marc Montminy, whose lab has spent decades on CREB, the well-known first responder in this pathway.
On the evidence ladder this is the bottom rung: mechanistic laboratory and animal research. That is not a knock. It is the kind of study that explains how something happens. It is not the kind that tells you whether a drug helps a person. Tap through where the evidence sits.
One wrinkle is worth pulling out of the evidence box because it changes how to picture the switch. Cells that could not tag Med14 did not simply go quiet. At rest, many of the genes a GLP-1 would normally turn on were already running higher than normal in the mutant cells, and the authors report that feedback signals which normally rein in cAMP were repressed. Their read is that the same phosphate site both turns the program on when the hormone is present and helps keep it in check when it isn’t. A switch, not just a gas pedal.
What nobody knows yet
The paper is candid about its edges. Two of them matter most for anyone reading this from outside a lab.
There is also a smaller, stranger limit. The team suspects the phosphate tag may change how Mediator gathers into little droplets inside the nucleus, which could explain why it favors cell-specific genes. They present that as a hypothesis, and it should be read as one.
One phosphate on one protein turned a short hormone signal into a long genomic program, in rat cells and in mice. The human chapter hasn’t been written.
Where this standsWhy this matters if you use insulin
The fair question from this audience is obvious: what does beta-cell resilience mean when type 1 has already taken most of your beta cells? Three things are worth knowing. First, a type 1 organization helped pay for this. Breakthrough T1D is a listed funder, and keeping beta cells alive under stress is a live question in type 1: at diagnosis, when some cells remain, and in cell-replacement therapies, where transplanted islets have to survive in a hostile environment. That connection is an editorial read, not the paper’s claim. The paper never mentions type 1.
Second, plenty of people with type 1 use a GLP-1 off-label, and this study does not move that decision in either direction. No GLP-1 is approved for type 1 diabetes as of this writing; phase 3 trials adding semaglutide to insulin therapy are registered and running on ClinicalTrials.gov. This work says nothing about safety, DKA risk, insulin needs, or dosing in a person. It says something about a protein in a dish.
Third, and this is the part that lingers: the whole finding is a timing story. Your own GLP-1 shows up at a meal and is gone in minutes, long before Med14 would ever get tagged. The drugs stick around for days, and in these experiments that staying power is what flipped the switch. The molecule was never the surprise. The duration was. If any of this bears on your care, bring it to a doctor who knows type 1 and knows you.
The study. Van de Velde S, Yu J, Evensen KG, Pakhlevanyan E, Williams AE, Shaw RJ, Montminy M. Med14 phosphorylation shapes genomic response to GLP-1 agonists. Proceedings of the National Academy of Sciences 123(10): e2536772123, published March 4, 2026 (PubMed PMC12974444). Funding listed: National Institutes of Health, Breakthrough T1D, Paul F. Glenn Foundation, Clayton Foundation, Leona M. and Harry B. Helmsley Charitable Trust. No industry funding listed. Evidence tier: mechanistic cell-line and mouse research.
Institutional release. Salk Institute, How do GLP-1 agonists affect gene expression?, March 5, 2026. Used for the funding statement and author affiliations.
The coverage. Roberts R. Prolonged GLP-1 Exposure Remodels Gene Expression in the Pancreas by Flipping a Molecular Switch, The Scientist, 2026. Used for the Van de Velde interview quotes.
Type 1 trial status. ClinicalTrials.gov phase 3 listings for semaglutide added to insulin in type 1 diabetes, for example NCT06894784 and NCT06082063, checked September 2026.
How to read this evidence. This is basic science: a rat insulinoma cell line (INS-1) and genetically modified mice, with the drug applied directly to cells or cultured islets. It shows a mechanism. It cannot show whether a GLP-1 helps, harms, or does nothing for any person, and it did not test human tissue. Semaglutide (Ozempic, Wegovy) and tirzepatide (Mounjaro, Zepbound) are FDA-approved for type 2 diabetes and weight management; none is approved for type 1 diabetes.
This post is for educational purposes only and is not medical advice. Nothing here is a recommendation to start, stop, or change any medication. Talk with your doctor before making any changes to your care.
Listen to the Juicebox Podcast
Conversations about diabetes, five days a week, since 2015. Search “GLP-1” in the Juicebox FAQ for the episodes on off-label use in type 1.
Read the full disclaimer
A Sensor That Watches Your Ketones
A Sensor That Watches Your Ketones
The FDA has cleared the first wearable that reads glucose and ketones at the same time. Here is what the agency reviewed, and what nobody has shown yet.
The FDA authorized Abbott’s Libre Duo 10 Day on August 25, according to the agency’s announcement. It is a wearable sensor that reads glucose and ketones at the same time, for people age 2 and up, and the FDA cleared it through the De Novo pathway — the route used when a device has no existing category to fit into. The agency built the category as part of the authorization.
The ketone half is the new part. Ketones have always been a snapshot: a strip, a stick, a meter you hope is not expired, checked at the moment you already suspect something has gone wrong. This turns them into a line that moves.
Abbott says the U.S. launch comes later this year. Before that, it is worth seeing how the category arrived, because the guardrails were written before the device existed.
What the FDA reviewed, and what it did not
According to the FDA’s announcement, the review covered six clinical studies enrolling more than 600 participants age 2 and older. The agency reported that the sensor tracked clinically meaningful differences in ketone levels across the full wear period, including picking up elevated ketones before ketoacidosis set in. Michelle Tarver, who directs the FDA’s device center, said having that information in real time “can be the difference between early intervention and a life-threatening emergency.”
Read that again, because it describes accuracy. The question the agency answered is whether the sensor measures ketones well enough to be relied on. The question it did not answer is whether wearing one lowers anybody’s odds of ending up in a hospital bed. Those are different studies, and the second kind has not been reported. Abbott designed the device, ran the studies, and wrote the release; the release also quotes Jennifer Sherr of Yale, a co-author of the Lancet panel that set the thresholds, and Aaron Kowalski of Breakthrough T1D, which led that same paper. It is a small field. None of that makes the work wrong. It is just worth knowing who is standing where.
The part nobody has shown yet
Finding out sooner only matters if the extra time changes what happens next. That is the claim underneath this entire category, and it is the claim that has not been put to a trial.
There are two reasonable ways to think about it. Both are still arguments.
The most honest sentence available here belongs to the panel itself. Its members wrote that they met “in the absence of substantial evidence” able to identify the right ketone thresholds for this technology. The thresholds now sitting inside an authorized device were set by expert judgment, because there was nothing sturdier to set them with.
What this changes for you
Nothing this week. Abbott has not announced a price or any insurance coverage, and the product is not on shelves. According to reporting in MedTech Dive, compatibility with Beta Bionics and Sequel Med Tech pumps is expected near the end of the year, with Insulet, Tandem, and MiniMed following the year after. For now this is a thing that exists, not a thing you can wear.
When it does arrive, the dull advice is the durable advice: keep the blood ketone meter. The expert panel recommends exactly that, for the times when how you feel does not match what the sensor says, and Abbott’s own published safety language allows that the urgent high ketones alarm may not detect every instance of ketones above 3.0 mmol/L. The company’s wear-duration study also reported that roughly 16 percent of adult sensors and 31 percent of pediatric sensors did not last the full ten days. None of that is disqualifying. It is the difference between a tool and a promise. The sick day plan is still what does the work — the sensor can move the moment you find out, but it cannot tell you what to do about it. Build that plan with your doctor before you need it, not during.
The authorization. U.S. Food and Drug Administration. FDA Authorizes First Wearable Device That Continuously Monitors Both Ketone Levels and Blood Sugar. FDA news release, August 25, 2026. Device details, quotes, alarm limitations and sensor wear-duration figures also come from Abbott’s press release of the same date. Abbott makes the device and funded the studies supporting it.
Ketone thresholds and how to use them. Dhatariya K, Bergenstal RM, Sherr J, et al. Continuous ketone monitoring for people with diabetes: international expert recommendations on the application of a new technology. The Lancet Diabetes & Endocrinology, 2026;14(1):82–92. Expert consensus, published as a Personal View and endorsed by ISPAD. Led by Breakthrough T1D.
DKA hospitalization trend. Benoit SR, Zhang Y, Geiss LS, Gregg EW, Albright A. Trends in Diabetic Ketoacidosis Hospitalizations and In-Hospital Mortality — United States, 2000–2014. MMWR Morb Mortal Wkly Rep, 2018;67:362–365. National surveillance data.
Continuous ketone sensing against blood. Haas NL, Korley FK, Schneider RM, et al. Continuous Ketone Monitoring in Diabetic Ketoacidosis: Prospective Multicenter Method-Comparison and Feasibility Study. Diabetes Technology & Therapeutics, 2026. A 34-participant feasibility study of a different manufacturer’s sensor, not Abbott’s.
Euglycemic DKA and SGLT inhibitors. Danne T, Garg S, Peters AL, et al. International Consensus on Risk Management of Diabetic Ketoacidosis in Patients With Type 1 Diabetes Treated With Sodium–Glucose Cotransporter (SGLT) Inhibitors. Diabetes Care, 2019;42(6):1147–1154.
Launch and pump-integration timing. MedTech Dive, August 2026 trade reporting.
How to read this evidence. The FDA review is a regulatory accuracy assessment, not an outcomes trial. The Lancet recommendations are expert consensus, and the panel states plainly that trial evidence for its thresholds does not yet exist. The emergency department comparison is a 34-participant feasibility study of a different manufacturer’s sensor, so it supports the idea that continuous ketone sensing can be accurate — not any claim about this particular product. The CDC hospitalization figures are surveillance data. They describe how often something happened, not why, and the most recent published national trend ends in 2014. Libre Duo 10 Day is authorized but not yet for sale in the United States, and no price or insurance coverage has been announced. Nothing here describes what any individual should do.
Disclosure. The Juicebox Podcast is sponsored by, among others, Dexcom, Omnipod, Tandem, MiniMed, twiist and Eversense — several of which are named above. Abbott is not a Juicebox sponsor. This post was not solicited, reviewed, or paid for by any company.
This post is for educational purposes only and is not medical advice. Nothing here is a recommendation to start, stop, or change any medication. Talk with your doctor before making any changes to your care.
Listen to the Juicebox Podcast
Conversations about diabetes, five days a week, since 2015.
Read the full disclaimer
Way more than a FAQ
Ask the podcast anything
How the Episode FAQ and site search turn more than a decade of conversations about Type 1 diabetes into answers you can find in seconds.
More than a decade of making this podcast has produced a strange problem: the answer to almost any Type 1 diabetes question you have is probably already in an episode. The trouble is finding it. There are more than 1,800 episodes. Nobody is going to scrub through them looking for the four minutes where Jenny explains why your blood sugar spikes at 3 a.m.
So we built a way to ask.
The Episode FAQ
Go to juiceboxpodcast.com/jbfaq and type your question the way you'd actually say it. Not keywords. Not "hypoglycemia nocturnal management." Type "I'm afraid of going low at night." Type "how do you bolus for pizza." Type "my teenager won't manage their diabetes." The search is built to understand real language — the way you'd describe your problem to a friend, not the way a textbook would file it.
Behind that search box are 834 real questions and struggles pulled from 296 episodes, each one written from what was actually said on the show. Every answer tells you what you'll learn, gives you a quote from the episode, and links you straight to it — listen on the website, Spotify, or Apple Podcasts, or jump to the whole series it came from.
A few things worth knowing about how it thinks:
It understands struggles, not just questions
A lot of what people bring to diabetes isn't a tidy question — it's "I feel like a bad diabetic" or "I'm exhausted and nothing is working." The search recognizes over ninety of the most common struggles people describe, in more than a thousand different phrasings, and routes you to the episodes that speak to them. You don't have to translate your problem into search language. That's the search's job, not yours.
It forgives typos
Spell "exercize" however you want. It'll figure it out.
It meets you where you are
Filter by experience — newly diagnosed, experienced, or parent of a child with T1D — and the results reorder themselves. Someone three weeks into this doesn't need the same episode as someone three years in trying to fine-tune settings. You can also browse by series (Bold Beginnings, Pro Tip, Defining Diabetes, Small Sips, and the rest) or by category, from insulin and dosing to burnout to thyroid.
Every question has its own link
Open any answer and tap "Copy link." Send it to a spouse, a newly diagnosed friend, a grandparent who watches your kid on weekends. The link opens directly to that question. When you follow a related question and want to return, there's a back button that takes you right to where you were — your search, your filters, your place on the page.
If the curated set doesn't have it, the archive does
When your search comes up thin, the page automatically checks the full 1,800+ episode archive and surfaces episodes the FAQ hasn't cataloged yet. The curated answers come first; the deep archive is the safety net.
And because plenty of you are searching from bed at 2 a.m. with a CGM alarm still ringing in your ears — there's a dark mode, and the whole thing is built to work one-handed on a phone.
The search on the main page
The search bar on the homepage is connected to the same brain. Type a question there and it pulls live answers from the FAQ, along with matching series, guides, and tools — all in one place. You don't have to know where something lives on the site. Ask, and it routes you.
The estimators
Some questions aren't really questions — they're math. For those, the FAQ sits alongside our interactive tools. Timing and amount — that's the whole game, and the tools exist to help you think about both.
Bolus Estimator
Think through meal insulin — carbs, timing, and strategy.
Settings Simulator
Understand basal, ICR, and ISF — and how they depend on each other.
Fat & Protein Estimator
For the meals that carb counts alone can't explain.
A1c Estimator
Translate between A1c and average glucose in either direction.
Why this exists
Most people leave their diagnosis appointment with a prescription and almost no working model of the disease. That's not their fault, and it's usually not even their doctor's fault — there's no time. The podcast has been filling that gap every week since January 2015, and the FAQ is the fastest door into it. You bring the question you're actually living with; it brings you to the conversation where we worked through it.
Nothing here is medical advice — I'm not a doctor, and the FAQ will remind you of that too. It's the lived experience of this community, organized so you can find it the moment you need it.
Talk to your doctor about your care. But walk in there knowing what to ask.
Bring the question you're living with
834 answers, searchable the way you actually talk.
Search the Episode FAQ →Read the full disclaimer
Are People With Type 1 Diabetes Immunocompromised?
Are People With Type 1 Diabetes Immunocompromised?
The short answer is no. The longer answer is the useful one — and it changes how you think about a sick day.
Type 1 diabetes is an autoimmune disease. It is not an immunodeficiency. Those two words get used as if they point at the same problem, and they are closer to opposites.
An immunodeficiency means the immune system is missing something it needs. Type 1 is the other failure. CDC describes it as an autoimmune disorder — the immune system mistakenly attacking part of the body. The system worked, aggressively and precisely, and aimed itself at the beta cells. Nothing in that process leaves a person short on defenses against a head cold.
So why does the question keep coming back? Because the honest answer has a “however” attached to it, and that however is worth understanding. It decides how you handle a sick day, what you say to a school nurse, and which box you land in on a vaccine form.
What high blood sugar does to immune cells
Here is the part that gets flattened into “diabetics are immunocompromised.” Sustained high blood sugar does blunt immune function. But it does it as a state, not as a trait.
Zhou and Lansang, reviewing the topic for Endotext, lay out the specific machinery. Neutrophils are recruited less well and produce less hydrogen peroxide, which is part of how they kill bacteria. Glycosylation of complement protein C3 impairs its ability to attach to pathogen surfaces. In vitro, advanced glycation end products block neutrophils from migrating out of the bloodstream toward an infection. A separate finding on macrophages — impaired receptor pathways for engulfing pathogens — was observed in cells from people with type 2 diabetes, so it does not carry over to Type 1 without more evidence.
The word that matters there is “state.” These are defects that move with glucose, not permanent properties of the diagnosis. But don't push that too far. The review's evidence that lowering glucose helps comes from hospital settings — a cardiac surgery protocol where tighter post-operative glucose tracked with fewer deep wound infections. It also flags the counterexample: in the NICE-SUGAR trial, driving glucose down hard with IV insulin in intensive care increased deaths, through hypoglycemia. The association is well described. A dose of glucose-lowering that reliably repairs it is not.
It also explains why illness is dangerous in a way that has very little to do with catching more colds.
What the research actually shows
The strongest evidence here comes from cohort studies, which describe association rather than cause. Carey and colleagues, writing in Diabetes Care in 2018, matched 5,863 people with Type 1 against 11,696 people without diabetes in English primary care records and counted infections from 2008 to 2015. One limit to hold onto: the cohort was restricted to people aged 40 to 89 whose diabetes was diagnosed by 2008, so it is not a picture of children or of newly diagnosed adults.
What they found is a gradient, and the shape of it is the real answer to the question. Infections needing a prescription came in modestly higher in the Type 1 group — an incidence rate ratio of 1.66. Infection-related hospitalization came in at 3.71. Infection-related death came in at 7.72, but that rests on 80 deaths, and the confidence interval runs from 4.47 to 13.33 — a threefold spread, which is the study telling you how much it does not know.
So it is not that people with Type 1 catch far more things. It is that the serious end of the range is heavier. The authors name an alternative reading worth sitting with: some of the gap could reflect a greater tendency to diagnose, prescribe, admit, or code a death as infection-related once a diabetes diagnosis is on file.
The part nobody has separated
Every number above comes from observational data, and observational data is not built to tell you why. Two explanations fit the same findings, and the studies were not designed to referee between them. They are not mutually exclusive — most likely both contribute, and nobody has measured how much of the gap belongs to each.
What the research can show is that the association exists and that, for infection-related hospitalization, it is graded — higher HbA1c, higher risk, band by band. What no cohort study can show is how much of that gap would close if glucose alone changed. That question is still open.
The immune system in Type 1 is not weak. It can be slowed down — and what slows it down is the same thing that makes everything else harder.
What this means when you are the one who is sick
The label matters in a few specific places. “Moderately or severely immunocompromised” is a defined clinical category with a specific list behind it — transplant recipients, people in active cancer treatment, primary immunodeficiencies, certain immunosuppressive drugs. CDC does not include diabetes of either type on that list, though it does say the list is “not limited to” those conditions. And on the question people ask most, CDC is unusually direct: people with diabetes might not be more likely to catch COVID or flu, but research shows they are at increased risk of severe illness from them.
One honest wrinkle: a different CDC page says that because Type 1 is autoimmune, people with it “may have weaker immune systems.” That sits awkwardly beside the immunocompromised list that leaves diabetes off, and CDC does not reconcile the two. If a clinician quotes one page at you and you have read the other, that is why.
That distinction is worth carrying into the conversation with a school nurse, an HR department, or the appointment where someone waves a hand and says immunocompromised. You are not describing a broken immune system. You are describing a condition where getting sick is more likely to go sideways, and where a large part of what decides how sideways is glucose — the part that is within reach. None of that is a reason to change anything about your care on your own. A sick-day plan is a conversation to have with your doctor before you need it, not during.
Infection risk in Type 1. Carey IM, Critchley JA, DeWilde S, Harris T, Hosking FJ, Cook DG. Risk of Infection in Type 1 and Type 2 Diabetes Compared With the General Population: A Matched Cohort Study. Diabetes Care, 2018;41(3):513–521. Retrospective cohort: 5,863 people with Type 1 vs 11,696 age-, sex- and practice-matched controls, English primary care (CPRD), 2008–2015, ages 40–89 with diabetes diagnosed by 2008. Unfunded; authors reported no conflicts.
HbA1c and infection risk within Type 1. Critchley JA, Carey IM, Harris T, DeWilde S, Hosking FJ, Cook DG. Glycemic Control and Risk of Infections Among People With Type 1 or Type 2 Diabetes in a Large Primary Care Cohort Study. Diabetes Care, 2018;41(10):2127. 4,496 people with Type 1, follow-up 2010–2015. Described by the authors as an unfunded study. This is the source for the graded HbA1c relationship — not the COVID data below.
How high glucose affects immune cells. Zhou K, Lansang MC. Diabetes Mellitus and Infection. Endotext, NCBI Bookshelf, last updated June 2024. Narrative review of mechanism. Note that its macrophage findings are drawn from cells of people with type 2 diabetes, its advanced-glycation finding is in vitro, and its evidence that lowering glucose improves outcomes comes from perioperative and inpatient settings rather than everyday outpatient management. The chapter does not describe people with diabetes as immunocompromised; it describes specific, partial immune defects.
Illness outcomes and HbA1c. Holman N, Knighton P, Kar P, et al. Risk factors for COVID-19-related mortality in people with type 1 and type 2 diabetes in England: a population-based cohort study. The Lancet Diabetes & Endocrinology, 2020;8(10):823–833. 264,390 people with Type 1, 464 COVID-19-related deaths, 16 Feb to 11 May 2020. Hazard ratio 2.23 (95% CI 1.50–3.30) for HbA1c at or above 86 mmol/mol (10.0%) versus 48–53 mmol/mol (6.5–7.0%). Important: in Type 1 the elevation reached significance only in that highest band; the stepwise gradient across all bands was found in type 2 diabetes.
Infection as a trigger for ketoacidosis. Umpierrez GE, Davis GM, ElSayed NA, et al. Hyperglycaemic crises in adults with diabetes: a consensus report. Diabetologia, 2024;67:1455–1479. Source for the finding that infection is the most common precipitant of DKA worldwide, occurring in 14–58% of cases, and for insulin omission dominating in US cohorts. The counterregulatory hormone mechanism is described in Dhatariya K, Mustafa OG, Stathi C. Hyperglycemic Crises. Endotext, NCBI Bookshelf, 2025.
Co-occurring autoimmune conditions. Nederstigt C, Uitbeijerse BS, Janssen LGM, Corssmit EPM, de Koning EJP, Dekkers OM. Associated auto-immune disease in type 1 diabetes patients: a systematic review and meta-analysis. European Journal of Endocrinology, 2019;180(2):135–144. 180 articles, 293,889 people with Type 1.
What CDC says. COVID-19 Vaccination Guidance for People Who Are Immunocompromised (last updated November 2025) and Your Immune System and Diabetes (last reviewed December 2023, last updated May 2024), US Centers for Disease Control and Prevention.
How to read this evidence. The infection and mortality figures come from cohort studies. A cohort study shows that two things occur together in a population; it does not establish that one causes the other, and it can only adjust for the factors the researchers measured. The mechanism descriptions come from narrative reviews, which summarize a field rather than test a hypothesis. The autoimmune prevalence figures come from a systematic review and meta-analysis, which is the strongest tier used here. No randomized trial in Type 1 diabetes has been designed to test whether lowering blood sugar reduces infection risk; that link rests on observational data. The closest randomized evidence comes from intensive-care insulin trials in mixed, mostly non-Type 1 populations, and those results conflict — one found fewer bloodstream infections, a larger one found no difference and more deaths.
This post is for educational purposes only and is not medical advice. Nothing here is a recommendation to start, stop, or change any medication. Talk with your doctor before making any changes to your care.
Listen to the Juicebox Podcast
Conversations about diabetes, five days a week, since 2015.
Read the full disclaimer
Your Thyroid Dose Was Set for a Different Body
Your Thyroid Dose Was Set for a Different Body
Levothyroxine is dosed by body weight. Lose a lot of it on a GLP-1 and the same pill can become too much — and the early warning signs look like half a dozen other things.
A 62-year-old man with type 1 diabetes, autoimmune hypothyroidism, and obesity arrived at an emergency department with palpitations, heavy sweating, confusion, fever, and shaking hands. His ECG showed atrial fibrillation. His TSH came back at 0.001 mIU/L. Six months earlier that number had been 1.9 and he had weighed 132 kg. In between, he started tirzepatide for obesity, titrated up to 10 mg weekly as instructed, missed his follow-up visit because he lives seasonally in different states, and lost more than 36 kg. His levothyroxine stayed at 200 micrograms a day the entire time. His clinicians traced the atrial fibrillation to thyrotoxicosis in the setting of that rapid weight loss.
That is one patient, written up in JAMA Internal Medicine in 2024 as a Teachable Moment by clinicians at the Barbara Davis Center. A single case report sits at the bottom of the evidence ladder and settles nothing on its own. It is here because of who he was. Autoimmune thyroid disease is the most frequent autoimmune condition traveling with type 1 diabetes, and the ADA’s 2026 Standards of Care put it in 17 to 30 percent of people with type 1.
That overlap is about to get more crowded. The 2026 Standards added Recommendation 8.29, which puts GLP-1-based therapy on the table as an obesity treatment for adults with type 1 diabetes and a BMI of 30 or above. More people on thyroid replacement are going to start losing weight quickly. What happened to that man is not exotic. It is arithmetic.
What the research shows
Only one study has looked at this at population scale, and it is not a trial. It is a target trial emulation published in Cardiovascular Diabetology in December 2025, built from a 15 percent sample of U.S. Medicare claims. Researchers matched 2,384 adults over 65 with type 2 diabetes who had been on a stable levothyroxine dose for at least six months and then started a GLP-1, against 2,384 matched patients who started an SGLT2 inhibitor instead. Over a median of about a year, starting a GLP-1 was associated with a higher rate of atrial fibrillation or flutter: hazard ratio 1.46, with a 95 percent confidence interval of 1.28 to 1.67. Stroke showed no significant difference.
Read the limits, because the authors do not bury them. This was type 2 diabetes, not type 1, in people over 65. There were only 13 crude atrial fibrillation events in each arm, and the authors write that the small event count makes the estimate unstable and that the results should be interpreted cautiously. Medicare claims carry no weights and no TSH values, so nobody can confirm that over-replacement was the mechanism. One exploratory finding is still hard to look away from: within 18 months of starting the drug, close to 60 percent of the atrial fibrillation events in the GLP-1 group occurred in people who had not yet had a single TSH test. Another quarter happened after a TSH was drawn but before anyone changed the dose. Three separate mechanisms could be pushing thyroid levels around, and they are not equally well studied.
The part nobody can tell you in advance
Here is what makes this hard rather than merely under-monitored: the requirement does not reliably move in one direction. In that 93-patient bariatric cohort, 47 people needed less levothyroxine after losing weight, 34 needed no change at all, and 12 needed more. All 12 who needed more had autoimmune thyroiditis. Their own residual thyroid function was still declining in the background, and that decline outran the effect of the weight loss.
For this audience, that detail is the whole ballgame. Hashimoto’s is usually the reason a person with type 1 is on levothyroxine in the first place, which puts a lot of readers squarely in the group where the arrow could point either way.
No prospective study has tracked weight, thyroid labs, and levothyroxine dosing together through a course of GLP-1 therapy. The Medicare authors call for exactly that, and say the mechanisms linking the drug, the weight loss, and the cardiovascular outcomes have not been disentangled. Until somebody runs it, the honest answer to “will my dose need to change” is that it might, in either direction, and the only way to find out is to measure.
The prescription never changed. The body it was calculated for did. That gap is where the trouble lives.
Editorial emphasis · Juicebox Podcast- Levothyroxine is dosed by body weight — roughly 1.6 micrograms per kilogram per day per the ATA’s 2014 guideline, then tuned by TSH. Your dose was fitted to a body of a particular size.
- Lose a lot of that weight quickly and the fit changes. In a 93-patient bariatric cohort, the size of the levothyroxine reduction tracked the loss of lean body mass.
- It does not always move down. In that same cohort, 47 people needed less, 34 needed no change, and 12 needed more — and all 12 who needed more had autoimmune thyroiditis, the usual reason someone with type 1 is on levothyroxine at all.
- Absorption may shift too. Oral semaglutide raised total T4 exposure by 33 percent in a 45-person pharmacokinetic trial. Measured with the tablet only; no equivalent number exists for the injectables.
- The one population-scale study — Medicare claims, type 2 diabetes, over 65 — found starting a GLP-1 was associated with more atrial fibrillation than starting an SGLT2 inhibitor (HR 1.46). An association, not a demonstrated cause, and only 13 crude events per arm.
- In that study, close to 60 percent of those atrial fibrillation events happened in people who had not yet had a single TSH test.
- The symptoms hide well. Palpitations, sweating, tremor, anxiety, and weight loss all have other plausible explanations in someone with type 1 diabetes who is losing weight on purpose.
- This is the overlap: the ADA puts autoimmune thyroid disease in 17 to 30 percent of people with type 1 diabetes, and its 2026 Standards now list GLP-1-based therapy as an obesity option for adults with type 1 and a BMI of 30 or above.
- None of this is a reason to avoid a GLP-1, and none of it is something to act on alone. It is a reason to ask how often your thyroid gets checked while your weight is moving — and to make sure whoever manages your thyroid knows about the other prescription.
Why this slips past people
The symptoms have nowhere distinctive to land. Levothyroxine prescribing information lists the effects of too much hormone: palpitations, rapid heart rate, tremor, excessive sweating, heat intolerance, anxiety, irritability, insomnia, muscle weakness, diarrhea, weight loss. Now set that list beside someone who is deliberately losing weight on a drug whose common side effects are gastrointestinal, who has type 1 diabetes, and who already knows exactly what a racing heart and a sweaty, shaky, anxious twenty minutes feels like. Every symptom on that list has three plausible explanations that are not the thyroid.
There is a second moving part worth naming. Levothyroxine labeling warns that thyroid therapy in people with diabetes can worsen glycemic control and change insulin requirements, and GLP-1 medications change insulin needs on their own. If thyroid status drifts during a stretch of weight loss, insulin dosing is drifting at the same time, for a reason that will not be obvious from a CGM graph.
None of this is an argument against taking a GLP-1, and none of it is something to act on alone. Levothyroxine has a narrow therapeutic index; too much and too little each carry real risk, and changing it without labs is not a thing to do. What this is, is a specific question worth putting on the table before or shortly after starting: how often will my thyroid get checked while my weight is moving, and what would make us check sooner? GLP-1 medications are not FDA-approved for type 1 diabetes, though the ADA’s 2026 Standards now list them as an option for treating obesity in adults with type 1. Bring the question to whoever manages your thyroid and whoever manages your diabetes, and make sure each one knows about the other prescription.
The index case (single case report · tier 5). Karakus KE, Shah VN, Akturk HK. Tirzepatide-Induced Rapid Weight Loss–Related Thyrotoxicosis. JAMA Internal Medicine 2024;184(10):1246–1247. Barbara Davis Center for Diabetes, University of Colorado. One patient. It illustrates a mechanism; it does not establish a rate.
Population-scale signal (target trial emulation of claims data · tier 3). Du F, Singh Ospina NM, Chen Y, et al. Glucagon-like peptide-1 receptor agonists and risk for cardiovascular events in older adults treated with levothyroxine: a target trial emulation. Cardiovascular Diabetology 2025;24:459. Medicare, adults over 65, type 2 diabetes, 2,384 matched pairs. The authors state that the small crude event count (13 per arm) makes the estimate unstable and that results should be read cautiously.
Weight-based dosing (clinical practice guideline). Jonklaas J, Bianco AC, Bauer AJ, et al. Guidelines for the treatment of hypothyroidism: prepared by the American Thyroid Association task force on thyroid hormone replacement. Thyroid 2014;24(12):1670–1751.
Weight loss and levothyroxine requirement (meta-analyses · tier 2; cohorts · tier 3). Azran C, Hanhan-Shamshoum N, Irshied T, et al. Surgery for Obesity and Related Diseases 2021;17(6):1206–1217 (28 studies, 1,284 patients). Hassan MM, Ahmad Z, Hassan MB, et al. Cureus 2025;17(7):e88354 (17 cohort studies, 922 patients; reduction not statistically significant, I² = 84%). Fierabracci P, Martinelli S, Tamberi A, et al. Thyroid 2016;26(4):499–503 (93 patients; 47 reduced, 34 unchanged, 12 increased). Barzin M, Molavizadeh D, Mahdavi M, et al. Thyroid 2024;34(9):1105–1116 (Tehran Obesity Treatment Study; three-year dose-change percentages).
Absorption (pharmacokinetic trial in healthy volunteers · tier 1 design, narrow question). Hauge C, Breitschaft A, Hartoft-Nielsen ML, Jensen S, Bækdal TA. Expert Opinion on Drug Metabolism & Toxicology 2021;17(9):1139–1148. n = 45. Worth naming plainly: four of the five authors were Novo Nordisk employees or shareholders, and the contract research organization that ran the trial was funded by Novo Nordisk, which makes semaglutide. The 33% figure also appears in the FDA-approved Rybelsus label.
Over-replacement symptoms and risk. FDA prescribing information, Synthroid (levothyroxine sodium), adverse reactions and warnings sections. Flynn RW, Bonellie SR, Jung RT, et al. Journal of Clinical Endocrinology & Metabolism 2010;95(1):186–193 — a 17,684-person Scottish cohort (tier 3) in which suppressed TSH was associated with higher rates of cardiovascular disease, dysrhythmia, and fracture, while a low-but-not-suppressed TSH was not.
Type 1 diabetes context (clinical practice guideline). American Diabetes Association Professional Practice Committee. Standards of Care in Diabetes—2026. Diabetes Care 2026;49(Suppl 1) — autoimmune thyroid disease in 17–30% of people with type 1 diabetes; screening recommended soon after diagnosis. Recommendation 8.29, added in the 2026 revision, lists GLP-1 RA–based therapy (evidence level B) and metabolic surgery (level C) as obesity management options for adults with type 1 diabetes and a BMI of 30 or above.
How to read this evidence. The strongest thing here is an association found in claims data, not a demonstrated cause, in a population that is older and has type 2 diabetes rather than type 1. The dose-change numbers come from bariatric surgery research, which is not the same intervention as a GLP-1. The 33% absorption figure was measured with oral semaglutide only. No randomized trial has tested thyroid monitoring during GLP-1 therapy, and the researchers involved are the ones asking for one.
What I could not confirm. Several sources describe ADA guidance to monitor thyroid levels when starting or uptitrating a GLP-1. I could not read that passage of the Standards directly, so it is attributed here to the researchers who cite it rather than stated as guideline text.
This post is for educational purposes only and is not medical advice. Nothing here is a recommendation to start, stop, or change any medication. Talk with your doctor before making any changes to your care.
Listen to the Juicebox Podcast
Conversations about diabetes, five days a week, since 2015.
Read the full disclaimer