| Clinical Takeaway | |||||||||||||||||||||
|
He was 52 years old, and the first real conversation of his metabolic life began on the day it was already too late to have it.
That is not entirely fair. Things had been said over the years. His weight had been noted, in the way weight gets noted. Somebody had once circled a fasting glucose of 108 mg/dL and written "recheck at next visit," and the next visit had been about a shoulder. At some point the word prediabetes had appeared in his chart, and nobody in the room, including him, had treated it as a word that meant something was happening.
Then an A1c came back at 7.4%. The diagnosis arrived through a patient portal message. He was started on metformin 500 mg twice daily, handed a printout about carbohydrates, and told to come back in 3 months.
Here is what nobody told him at that visit. By the time his glucose crossed the threshold we use to name this disease, roughly half of his beta cell function was already gone. The years in which it was most recoverable were mostly behind him. And the therapy he had just been given was chosen, in large part, because it is what we have always given first.
The myth that should die is that type 2 diabetes begins when the glucose rises. It does not. Glucose is the smoke. We have built a treatment paradigm around waiting for smoke, confirming the smoke twice, and then bringing a garden hose to a house that has been burning for a decade.
I want to make a specific argument here, and I want to be precise about where the evidence supports me and where it runs out. Type 2 diabetes is the clearest example in American medicine of a disease we understand well, can detect early, and treat late and weakly anyway.
Hyperglycemia is the last thing to break
The Whitehall II cohort followed 6,538 British civil servants without diabetes and, over a median of 9.7 years, watched 505 of them develop it. Because the investigators had repeated measurements, they could run the film backward from the moment of diagnosis.
People who went on to develop diabetes were already metabolically distinguishable 13 years before anyone gave them the label. Their fasting glucose ran higher across the entire window and climbed on a steeper slope. Then, close to diagnosis, the curves stopped being gradual. Fasting glucose bent sharply upward over the final 3 years. Insulin sensitivity fell steeply across the last 5. Beta cell function, by HOMA modeling, actually rose between years 4 and 3 before diagnosis, from 85.0% to 92.6%, and then collapsed to 62.4% at the moment of diagnosis.
Read that sequence again, because it is the whole story. The pancreas compensates. It works harder, and for a while the compensation is good enough that every screening test we routinely order comes back acceptable. The glucose only rises when compensation fails. Hyperglycemia is not the beginning of the disease. It is the announcement that a decade of silent compensation has run out.
The UKPDS put a number on where that leaves people at diagnosis. Beta cell function, again by HOMA modeling, sat at roughly 50% of normal at the time of diagnosis and continued falling to about 28% over the following 6 years, on treatment. The paper carrying that finding was subtitled, with real understatement, "a progressive disease."
| UKPDS · Beta cell function by HOMA modeling | |||||||||||||||||||||||
| Half of it is gone before we name the disease | |||||||||||||||||||||||
| Percentage of normal beta cell function, modeled from fasting glucose and insulin. | |||||||||||||||||||||||
| |||||||||||||||||||||||
| HOMA-B is a modeled estimate of function, not a direct measure of beta cell mass. Autopsy series suggest mass is reduced by roughly 60% in established disease. The two numbers describe related but distinct things. | |||||||||||||||||||||||
| The core claimType 2 diabetes is endocrine pancreatic failure. We diagnose it by its last symptom, treat it first with our least potent tools, and reserve intensity for the years when the tissue we were trying to protect is already gone. |
The window is real, and it is short
If this were simply a story about decline, it would be depressing but not actionable. It is actionable because the failure is partly reversible, and reversibility has an expiration date.
Roy Taylor's twin cycle hypothesis proposed that type 2 diabetes results from fat accumulating where it does not belong. First in the liver, driving hepatic insulin resistance and VLDL export. Then in the pancreas, where it suppresses beta cell function. If that model is right, then removing the fat should restore the function. DiRECT was built to test exactly that.
DiRECT randomized 306 people with type 2 diabetes of up to 6 years duration, by general practice cluster, to a total diet replacement program of roughly 825 to 853 kcal per day for 12 to 20 weeks followed by structured food reintroduction and maintenance support, or to usual care. All antidiabetic medication was withdrawn at the start.
| DiRECT · Remission off all medication, intervention vs usual care | ||||||||
|
Then the mechanistic substudy did something more important than the headline. In 64 intervention participants who had detailed metabolic phenotyping, liver fat fell from 16.0% to 3.1%, and pancreatic fat fell in essentially everyone who lost weight, whether or not their glucose normalized. What separated the people who achieved remission from those who did not was one thing: recovery of first-phase insulin response. The liver responds to weight loss in nearly everybody. The beta cell responds only if it still can.
And whether it still can is largely a function of how long it has been failing. Responders had a shorter diabetes duration than non-responders, 2.7 years versus 3.8 years. Across the wider literature the gradient is steeper still. Compared with a duration of 0 to 2 years, the chance of achieving remission drops below 50% at 2 to 7 years and to roughly 20% past 7 years. Bariatric surgery cohorts show the same shape at 10 years: about 60% remission for those operated on within 1 year of diagnosis, just over 20% at 1 to 3 years, around 15% beyond 4 years.
The same pattern appears in the insulin literature. Among patients given short-course intensive insulin at diagnosis, a duration under 2 years independently predicted whether remission persisted, displacing both baseline A1c and baseline beta cell function in the regression.
Three different interventions. Diet, surgery, insulin. One consistent finding. The clock started before we were paying attention, and it is still running.
What we actually do at diagnosis
Here is the part of this argument I expected to be easy and was not. I assumed the guidelines were the problem. They are not, at least not anymore.
The 2026 ADA Standards of Care support considering a GLP-1 receptor agonist or SGLT2 inhibitor from the point of diagnosis rather than after metformin has been given a chance to fail. The 2022 ADA and EASD consensus report had already removed metformin as the sole first-line agent. Agents with demonstrated cardiovascular benefit are recommended independent of A1c, with or without metformin. Arguing against the guidelines as though they still mandate a metformin waiting period is fighting a battle that was won on paper several years ago.
The problem is that almost nothing changed in the exam room.
A cross-sectional analysis of 5,783 patients with newly diagnosed type 2 diabetes across 60 primary care clinics found that 19.9% received either an SGLT2 inhibitor or a GLP-1 receptor agonist within 3 months of diagnosis. The subgroups where the indication is strongest did not do better. Patients with chronic ischemic heart disease: 20.0%. Patients with impaired kidney function: 19.3%.
| Newly diagnosed type 2 diabetes · 60 primary care clinics, n=5,783 | ||||||||||||||||||||||||
| Having the indication changed nothing | ||||||||||||||||||||||||
| Received an SGLT2 inhibitor or GLP-1 receptor agonist within 3 months of diagnosis. | ||||||||||||||||||||||||
| ||||||||||||||||||||||||
| Shaded remainder is the proportion who received neither agent. Cross-sectional EMR review, April 2021 to January 2023, West Michigan. | ||||||||||||||||||||||||
Higher BMI, hyperlipidemia, ischemic heart disease and higher A1c all raised the odds of receiving one of these drugs, which tells you clinicians were reasoning rather than choosing at random. They were reasoning inside a framework that treats these agents as an escalation instead of a starting point.
Cost and coverage are part of this and I am not going to pretend otherwise. But cost does not explain why the cardiac and renal subgroups, where prior authorization is most defensible and the guideline language is strongest, performed no better than the cohort overall. Something upstream of the formulary is deciding that a new diagnosis is a reason to start gently.
The insulin question, stated precisely
The reflex to reach for insulin in type 2 diabetes deserves a more careful answer than it usually gets, including from me.
The case against is real. Insulin promotes weight gain. In a patient whose disease is driven by ectopic fat and insulin resistance, gaining weight while receiving a hormone whose job is storage is a strategy that treats the number and worsens the thing producing it. If the goal is to unload the liver and the pancreas, chronic basal insulin as a maintenance strategy in an insulin-resistant patient works directly against that goal.
But the blanket version of that claim does not survive contact with the trial data, and I want to be honest about where it breaks.
Weng and colleagues randomized 382 patients with newly diagnosed type 2 diabetes to a short course of intensive insulin, delivered either by pump or by multiple daily injections, or to oral agents. Treatment was withdrawn once normoglycemia was achieved. At 1 year, remission rates were roughly 45% to 51% in the insulin arms against 26.7% in the oral arm. Acute insulin response improved more in the insulin groups. The authors concluded that early intensive insulin favorably affected both recovery and maintenance of beta cell function.
That is not an argument for chronic insulin. It is an argument for the thing this entire article is about. A short, intense intervention at the moment of diagnosis, aimed at breaking glucotoxicity and giving the beta cell room to recover, outperforms a gentle one. The insulin was not the active ingredient. Intensity and timing were.
So the precise version of the claim is this. Transient intensive insulin at diagnosis is a legitimate, evidence-supported strategy for beta cell rescue, and it is badly underused. Chronic basal insulin as a first-line maintenance therapy in an insulin-resistant patient with obesity is a different intervention with a different physiology, and it is overused. Conflating the two has let us do the wrong one while skipping the right one.
What maximal therapy at diagnosis would look like
If you accept that the diagnosis marks a window closing rather than a disease opening, the treatment plan reorganizes itself. Here is what I think the first visit should contain.
Name the disease accurately. Tell the patient this is a condition of the pancreas and the liver, that it has been developing for years, and that the next 24 months matter more than any 24 months that will follow. Nobody escalates for a condition they were told to watch.
Lead with the agents that change the trajectory. A high-potency incretin, an SGLT2 inhibitor, or both, chosen on comorbidity rather than on sequence. The guidelines already permit this and the 2026 Standards specifically support it from diagnosis. Metformin remains reasonable, cheap and well tolerated as a companion. It is a poor sole answer to a pancreas shedding close to 4 percentage points of modeled function a year, which is what the UKPDS curve works out to.
Offer a real dietary intervention, not a printout. Total diet replacement has the strongest remission evidence and should be offered explicitly, with the DiRECT numbers attached, to anyone within a few years of diagnosis. Therapeutic carbohydrate restriction is a legitimate alternative for patients who will not do formula replacement. In the LoBAG30 work, a weight-neutral diet of 30% carbohydrate, 30% protein and 40% fat lowered glycohemoglobin from 10.8% to 9.1% over 5 weeks in untreated patients, with weight loss doing none of that work. A separate 12-week randomized comparison reported a final A1c of 6.6% on LoBAG30 against 7.2% on control, though both arms lost weight in that one, so it is not a clean test of composition alone. I flag the evidence quality below, because it matters.
Set the target at remission, and say so out loud. Not an A1c below 7%. Remission, defined as an A1c below 6.5% off medication, stated with honest odds and an honest acknowledgment that those odds fall every year.
Book the follow-up in weeks. Not 3 months. The most consequential variable in this disease is how fast you move, and a 3-month recheck interval is a decision to move slowly.
| Free reference | |
| If you are choosing an incretin at diagnosis, the trial data is scattered across a dozen publications. I put the efficacy, safety and dosing figures for 12 GLP-1 and dual-agonist trials on a single page. | |
|
And why wait for the diagnosis at all
Everything above accepts the diagnosis as the starting gun. It should not be.
Prediabetes is treated in most American primary care as a warning rather than a condition. The outcome data does not support that reading. In a meta-analysis of prediabetes and cardiovascular outcomes, prediabetes carried a relative risk of roughly 1.15 for composite cardiovascular disease and 1.16 for coronary heart disease against normoglycemia, with measurable excess all-cause mortality. Those are modest relative risks attached to an enormous population, which is exactly how you generate a very large number of events out of a condition nobody treats.
We also know the intervention works. The Diabetes Prevention Program showed that intensive lifestyle intervention reduced progression to diabetes by 58% and metformin by 31% against placebo. That result is nearly 25 years old. It has not meaningfully changed what happens to a patient whose A1c comes back at 6.0%.
And prediabetes is itself a late marker. By the time the A1c moves, insulin sensitivity has been falling for years. The earlier signals are ones we already collect and then ignore: a rising waist circumference, a triglyceride to HDL ratio drifting upward, hepatic steatosis noted incidentally on imaging, the whole constellation we label metabolic syndrome and then do nothing about. Central adiposity is not a risk factor sitting next to this disease. On the twin cycle model, it is the mechanism.
Which leaves the question I cannot get past. If the disease is a decade of ectopic fat accumulation ending in pancreatic failure, and we can watch the fat accumulating the entire time, why is the first real intervention scheduled for the year the pancreas fails?
The honest ledger
| What would make me wrong | ||||||
|
None of that makes the underlying claim fake. It makes it an argument from convergent evidence rather than from a single trial. The duration gradient appears independently in diet trials, surgical cohorts and insulin studies, which is the kind of consistency that usually means something real. But a clinician who wants a randomized answer to whether starting hard at diagnosis preserves beta cells is entitled to one, and does not have one.
| Speculation · extrapolation beyond the data |
| I suspect that within 10 years we will look back on the A1c threshold of 6.5% the way we now look at treating hypertension only after the stroke. The threshold is a statistical artifact of retinopathy risk, not a biological event. Nothing happens to the pancreas at 6.5% that was not already happening at 6.2%. If that is right, the next useful diagnostic category is not a glucose number at all. It is a measure of ectopic fat and beta cell reserve: hepatic fat fraction, a first-phase insulin response, or some proxy nobody has built yet. A patient with 12% liver fat, a blunted first-phase response and a perfectly normal A1c is, on the twin cycle model, already sick. We currently tell that person their labs look fine. Nobody has run the trial that would justify treating them. Until someone does, this is a hypothesis about where the field is heading and nothing more. |
The bottom line
By the time we diagnose type 2 diabetes, beta cell function is around half of normal, the disease has been running for a decade, and the odds of remission are already falling by the year. Everything we know about reversibility says the first 24 months carry more leverage than the following 20 years. And in the largest look at what actually happens in American primary care, 4 out of 5 newly diagnosed patients walked out without the agents most likely to change that trajectory. The guidelines are no longer the obstacle. The habit is. We treat the diagnosis as a beginning when it is much closer to a deadline.
| The single most useful thing a clinician can do at a new diagnosis is compress the timeline. Start the agents that alter trajectory, set remission as the stated target, and book the follow-up in weeks rather than months. At Vineyard, our clinicians treat obesity and its metabolic consequences as the chronic, time-sensitive disease they are. See how Vineyard approaches obesity care → |
Following the latest GLP-1 clinical trials?
| Explore the trial tracker → |
The patient described at the opening is a composite drawn from many clinical encounters and does not represent any individual. Disclosure: the author is Chief Medical Officer of Vineyard, a telehealth obesity medicine practice. This article is educational and is not individualized medical advice. Talk with your own clinician before making changes to your care.
REFERENCES
1. Tabák, A. G., Jokela, M., Akbaraly, T. N., et al. (2009). Trajectories of glycaemia, insulin sensitivity, and insulin secretion before diagnosis of type 2 diabetes: An analysis from the Whitehall II study. The Lancet, 373(9682), 2215–2221. https://doi.org/10.1016/S0140-6736(09)60619-X
2. UK Prospective Diabetes Study Group. (1995). U.K. Prospective Diabetes Study 16: Overview of 6 years' therapy of type II diabetes: A progressive disease. Diabetes, 44(11), 1249–1258. https://doi.org/10.2337/diab.44.11.1249
3. Lean, M. E. J., Leslie, W. S., Barnes, A. C., et al. (2018). Primary care-led weight management for remission of type 2 diabetes (DiRECT): An open-label, cluster-randomised trial. The Lancet, 391(10120), 541–551. https://doi.org/10.1016/S0140-6736(17)33102-1
4. Lean, M. E. J., Leslie, W. S., Barnes, A. C., et al. (2019). Durability of a primary care-led weight-management intervention for remission of type 2 diabetes: 2-year results of the DiRECT open-label, cluster-randomised trial. The Lancet Diabetes & Endocrinology, 7(5), 344–355. https://doi.org/10.1016/S2213-8587(19)30068-3
5. Taylor, R., Al-Mrabeh, A., Zhyzhneuskaya, S., et al. (2018). Remission of human type 2 diabetes requires decrease in liver and pancreas fat content but is dependent upon capacity for β cell recovery. Cell Metabolism, 28(4), 547–556.e3. https://doi.org/10.1016/j.cmet.2018.07.003
6. Taylor, R. (2021). Type 2 diabetes and remission: Practical management guided by pathophysiology. Journal of Internal Medicine, 289(6), 754–770. https://doi.org/10.1111/joim.13214
7. Weng, J., Li, Y., Xu, W., et al. (2008). Effect of intensive insulin therapy on beta-cell function and glycaemic control in patients with newly diagnosed type 2 diabetes: A multicentre randomised parallel-group trial. The Lancet, 371(9626), 1753–1760. https://doi.org/10.1016/S0140-6736(08)60762-X
8. Kramer, C. K., Zinman, B., Choi, H., & Retnakaran, R. (2016). Predictors of sustained drug-free diabetes remission over 48 weeks following short-term intensive insulin therapy in early type 2 diabetes. BMJ Open Diabetes Research & Care, 4(1), e000270. https://doi.org/10.1136/bmjdrc-2016-000270
9. Cai, X., Zhang, Y., Li, M., et al. (2020). Association between prediabetes and risk of all cause mortality and cardiovascular disease: Updated meta-analysis. BMJ, 370, m2297. https://doi.org/10.1136/bmj.m2297
10. Nuttall, F. Q., Schweim, K., Hoover, H., & Gannon, M. C. (2008). Effect of the LoBAG30 diet on blood glucose control in people with type 2 diabetes. British Journal of Nutrition, 99(3), 511–519. https://doi.org/10.1017/S0007114507819155
11. Anderson, J. D., Hanson, M. A., Lysne, L. F., et al. (2023). 49-LB: A low biologically available glucose diet reduces hemoglobin A1c in type 2 diabetes mellitus: A randomized controlled trial. Diabetes, 72(Supplement 1), 49-LB. https://doi.org/10.2337/db23-49-LB
12. Müller, F., Bouthillier, M. J., Alshaarawy, O., Azhary, H., & Holman, H. T. (2025). SGLT2 inhibitor and GLP-1 receptor agonist prescriptions in newly diagnosed type 2 diabetes patients with cardiorenal risks: A cross-sectional study. Journal of Diabetes Research, 2025, 6656982. https://doi.org/10.1155/jdr/6656982
13. American Diabetes Association Professional Practice Committee for Diabetes. (2026). 9. Pharmacologic approaches to glycemic treatment: Standards of Care in Diabetes—2026. Diabetes Care, 49(Supplement 1), S183–S215. https://doi.org/10.2337/dc26-S009
14. Knowler, W. C., Barrett-Connor, E., Fowler, S. E., et al. (2002). Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. New England Journal of Medicine, 346(6), 393–403. https://doi.org/10.1056/NEJMoa012512
Evidence-based obesity medicine, twice a week. No hype, no telehealth grifts.