In routine care, remission of type 2 diabetes is uncommon and usually brief

FOR DECADES type 2 diabetes was taught as a disease that only gets worse. In it the body loses control of blood sugar, usually after years of carrying too much weight, and patients typically move from diet advice to one pill, then two, and often to insulin. A 2011 paper from Newcastle University summed up the textbook view as one of "amelioration but not cure", adding that "a steady rise in plasma glucose occurs irrespective of the degree of control or type of treatment". Doctors track that rise with a blood test that reflects a person's average blood sugar over roughly the previous three months (HbA1c). A reading of 6.5% or above is the usual threshold for diagnosing diabetes.

Since then a series of studies has shown that the disease can retreat, at least for a while, when people lose a large amount of weight. The question patients now put to their doctors is whether they might come off their diabetes medicines, and it comes up more often as the newer weight-loss injections have spread. A new study of American health records by Felipe Lobelo of Emory University and colleagues, posted on medRxiv on October 16th, gives a sober answer for routine care. Of 96,817 adults who had been treated with diabetes drugs, 4,188 (4.33%) had a recorded remission, and for half of them it had ended within about seven and a half months.

What counts as remission matters to that answer. In 2021 a panel convened by the American Diabetes Association, with members from the European Association for the Study of Diabetes, Diabetes UK, the Endocrine Society and the Diabetes Surgery Summit, published an agreed definition in Diabetes Care. A person is in remission when their HbA1c is below 6.5% when measured at least three months after they have stopped all glucose-lowering drugs. The panel considered the words "resolution", "reversal" and "cure", which had all been used, and chose remission because it implies "that a notable improvement may not be permanent". On cure specifically, the group "agreed that the term should be avoided in the context of T2D", because blood sugar can rise again and because the underlying faults in how the body makes and responds to insulin are "rarely completely normalized" by treatment.

Fat in the wrong places

The 2011 paper, by Ee Lin Lim, Roy Taylor and colleagues, published in Diabetologia, tested the textbook view directly. Eleven people with type 2 diabetes ate 600 calories a day for eight weeks. After a single week their fasting blood sugar had fallen from an average of 9.2 to 5.9 mmol/l, which is within the normal range. Over the eight weeks the fat stored inside their livers, measured by magnetic resonance imaging, fell from 12.8% of the organ to 2.9%, and fat inside the pancreas fell from 8.0% to 6.2%.

The pancreas is where insulin, the hormone that moves sugar out of the blood and into the body's tissues, is made, by clusters of specialized cells (beta cells). In healthy people these cells release a quick burst of insulin within minutes of sugar rising in the blood. At the start of the study this early response in Dr Lim's volunteers was about a third of the level measured in a matched group of people without diabetes. By eight weeks it had more than doubled and, in the paper's words, "approached control values". The Newcastle group's account of the disease follows from this. Excess fat builds up in the liver, which then releases too much sugar into the blood, and in the pancreas, where it impairs the beta cells, and losing enough weight to clear that fat can restore both organs to something close to normal working. The study was small and had no comparison group on an ordinary diet, so it showed that the change was possible rather than how often it would happen.

A trial in family practice

Answering that needed a randomized trial, and the same researchers, working with colleagues at the University of Glasgow, ran one. The Diabetes Remission Clinical Trial, known as DiRECT, was reported in the Lancet in 2018 (the authors' accepted version is here). It recruited 306 adults who had been diagnosed with type 2 diabetes within the previous six years and were not taking insulin, through 49 family-doctor practices in Scotland and the Tyneside area of England. The practices, rather than individual patients, were assigned at random either to offer the program or to give usual care following national guidelines.

People in the program stopped their diabetes and blood-pressure drugs on the first day. For the next three to five months they replaced all their food with formula shakes and soups providing 825 to 853 calories a day (total diet replacement). Ordinary food was then brought back in steps over two to eight weeks, and a nurse or dietitian at each practice supported them in keeping the weight off for the rest of the two-year trial.

After 12 months, 68 of the 149 people in the program group (46%) were in remission, against 6 of the 149 in the usual-care group (4%). Remission in DiRECT meant an HbA1c below 6.5% after at least two months off all diabetes drugs, a slightly shorter interval than the three months the 2021 definition later required. The results tracked weight loss closely. None of the 76 participants who gained weight reached remission, but 31 of the 36 who lost 15kg or more (86%) did. Losing that much was hard, and only 36 of the 149 people in the program group (24%) managed it. Nine serious adverse events were reported by 7 of the 157 people in the program group (4%), compared with two events in two people in the usual-care group, and two of the program group's events, in a single participant, were judged possibly related to the treatment.

Five years on

Whether remission lasts is the question that matters most to patients, and the trial team followed a subset for longer. Their five-year results, published in the Lancet Diabetes & Endocrinology in 2024, show how much is lost over time. Of the 149 people originally in the program group, 101 were still in the trial at two years and 95 agreed to three further years of lighter support. At five years, 11 of the 85 people in that group with data (13%) were in remission. Of the 95, 48 had been in remission at two years, and 43 of those were followed to five years, of whom 26% were still in remission. In the original usual-care group, 5 of 93 people with data (5%) were in remission at five years.

The longer study also counted harms in patient-years, a measure that adds up the time each person spent in the study so that groups followed for different lengths can be compared. Serious adverse events ran at 4.8 per 100 patient-years in the original program group, against 10.2 in the usual-care group, less than half the rate. Lim's study, DiRECT and its five-year follow-up all come from the same Newcastle and Glasgow research program, so they are not independent confirmations of one another.

Outside the trials

Trials enroll volunteers, give them intensive support and measure them on a schedule. The new preprint asks what happens in ordinary practice. Dr Lobelo and three co-authors, from Emory, the University of Washington and Meta, ran a retrospective cohort study, one that looks back through records already collected rather than assigning anyone a treatment. They used Protégé, a database linking the electronic records and insurance claims of 4.95 million American patients, and found 136,781 adults with type 2 diabetes confirmed by laboratory tests. They counted a remission when a first HbA1c below 6.5% was recorded at least 90 days after all glucose-lowering treatment had stopped.

Among the 96,817 adults who had already been treated with diabetes drugs when they entered the study, 4,188 (4.33%) had a remission on record. It was more common in the 39,964 adults who had not been treated at entry, of whom 6,223 (15.57%) did. The first remission in previously treated adults lasted a median of 226 days. As in DiRECT, it was likelier when the disease was less advanced. Over five years, the estimated share of previously treated adults reaching remission was 6.89% for those whose HbA1c at entry was 6.5% to 6.9%, but 1.47% for those whose HbA1c was 9.0% or higher.

The study's most unusual finding concerns how remission ends. Among the 4,188 previously treated adults who reached remission, the first event to end it within five years was a restart of diabetes treatment in 47.83%, a rise in HbA1c back to the diabetic range in 31.94%, and death in 4.65%. In other words, a recorded restart of treatment ended remission about one and a half times as often as a recorded return of high blood sugar did. The records show when treatment was restarted but cannot show what prompted each decision.

The study has several serious limits. It is a preprint, which means it has not been peer reviewed, the checking of a paper by independent experts before a journal publishes it. The authors state that the database "is not a probability sample" of American patients, so its rates may not hold for the country as a whole. Body weight, the factor DiRECT tied most closely to remission, could not be tested properly because body-mass index was missing for 50,308 of the 96,817 previously treated adults (52.0%). Medical records also miss things, and in 1,264 of the 4,188 remissions (30.2%) the end of treatment was inferred from a planned end date or the end of a prescription supply rather than a recorded stop. The work was funded by Meta Superintelligence Labs, which also supplied the computing environment and staff support. All four authors were "contracted and compensated by Meta Superintelligence Labs" to carry out the analyses. The paper also states that an internal Meta AI model "was used to draft the analyses, study cohorts, statistical plan, protocols, and during manuscript preparation", with its output checked by the lead author.

What remission is worth

Even a remission that ends brings a period without daily drugs and with blood sugar below the diagnostic threshold, and in DiRECT the program group had fewer serious adverse events over five years than the usual-care group. The ADA panel was careful not to treat remission as the end of the disease. It noted that "the classic complications of diabetes", including damage to the eyes, kidneys and nerves and a higher risk of heart disease, "can still occur" after a remission. It said that people in remission should keep having regular eye screening, kidney tests, foot checks and blood-pressure measurements, because there is "no long-term evidence" that any of those checks can safely be dropped. Once the preprint has been reviewed, and if its findings are repeated in other databases, doctors will have a clearer picture of how often remission happens and lasts in patients who never joined a trial.