Condition: Type 2 Diabetes

68 of 149 (46%) in the intervention group were in remission at 12 months, against 6 of 149 (4%) on usual care (adjusted OR 19.7, 95% CI 7.8 to 49.8). Mean weight change was -22.0 lb (-10.0 kg, SD 8.0) against -2.2 lb (-1.0 kg, SD 3.7). Measured in: 306 adults aged 20 to 65 with type 2 diabetes diagnosed within the previous 6 years and BMI 27 to 45, recruited through 49 primary care practices in Scotland and Tyneside. Not taking insulin.. Randomization was by practice, not by person, so the effective sample is smaller than 306 and practice-level differences in support cannot be separated from the intervention. It was open-label, which it had to be, and remission was assessed after a protocol-mandated withdrawal of glucose-lowering medication that would itself be unsafe without supervision. Nobody taking insulin was enrolled.

Remission at 12 months by weight lost, across both trial arms: 0 of 76 (0%) among those who gained weight, 6 of 89 (7%) losing 0 to 11 lb (0 to 5 kg), 19 of 56 (34%) losing 11 to 22 lb (5 to 10 kg), 16 of 28 (57%) losing 22 to 33 lb (10 to 15 kg), and 31 of 36 (86%) losing 33 lb (15 kg) or more. Measured in: The 306 DiRECT participants pooled across intervention and control arms, grouped after the fact by weight change. This is a within-trial analysis by achieved weight loss, not by randomized assignment, so the groups are self-selected by their own success. Whatever made someone able to lose 33 lb (15 kg) may also independently favor remission, and the gradient would look the same either way.

53 of 149 (36%) were in remission at 24 months against 5 of 149 (3%) on usual care (adjusted OR 25.82). 17 of 149 (11%) had kept off 33 lb (15 kg) or more, down from 24% at one year. Among those sustaining a 22 lb (10 kg) loss, 29 of 45 (64%) were in remission. Measured in: The same DiRECT cohort followed to 24 months, with continued structured weight-maintenance support. The remission rate fell as maintained weight loss fell, which is the same relationship as at one year, not a separate finding. Two years is still short for a condition measured in decades, and the maintenance support was more than most people are offered.

Partial or complete remission reached 11.5% at year 1 and 7.3% at year 4 in the intensive lifestyle arm, against 2.0% in the control arm at both points. Remission sustained across all four years occurred in 3.5% against 0.5%. Measured in: 4,503 adults with type 2 diabetes and overweight or obesity in the Look AHEAD trial, randomized to intensive lifestyle intervention or diabetes support and education. Participants had longer-established diabetes than the DiRECT cohort, which is the most likely reason the numbers are a quarter of DiRECT's, and the definition used here includes partial remission, not the stricter 2021 consensus threshold.

Structured exercise lowered HbA1c by 0.67 percentage points against control. By mode: aerobic 0.73, resistance 0.57, combined 0.51. By volume: more than 150 minutes a week gave 0.89, 150 minutes or less gave 0.36. Physical activity advice alone changed HbA1c only when combined with dietary advice (0.58). Measured in: 47 randomized controlled trials, 8,538 people with type 2 diabetes. Trial programs were supervised, which is not what most people do at home, and the mode comparison is between trials, not within them, so the aerobic-versus-resistance ordering is not a head-to-head result. Trials with shorter follow-up dominate.

Incidence was 11.0 cases per 100 person-years on placebo, 7.8 on metformin (31% reduction) and 4.8 with lifestyle intervention (58% reduction), over an average 2.8 years. The lifestyle arm targeted 7% weight loss and 150 minutes a week of physical activity. Measured in: 3,234 people with impaired fasting and post-load glucose, 68% women, mean age 51. The lifestyle arm was delivered with 16 individual teaching sessions and ongoing case-manager contact, which is more support than routine care provides. The endpoint is a diagnostic threshold crossing, not a clinical event.

In a head-to-head trial of people with type 2 diabetes already on metformin, the dual-hormone drug tirzepatide brought HbA1c down by up to 2.3 percentage points and took off several kilograms more than semaglutide. Both are weekly injections, and both commonly cause nausea and other gut effects early on. This trial ran 40 weeks and measured glucose and weight, so it speaks to control of the condition, not to long-term heart or kidney events.

11 of 85 (13%) of intervention participants in the extension were in remission at 5 years, against 5 of 93 (5%) of controls. Of those in remission at year 2, 26% were still in remission at year 5. Mean weight loss from baseline was 13.4 lb (6.1 kg) in the extension group and 10.1 lb (4.6 kg) in controls. Measured in: DiRECT participants who continued into a 3-year extension offering low-intensity dietary support, plus the original control participants followed alongside. What could explain it instead: Self-selection into the extension. The participants who agreed to three more years of dietary support are plausibly the ones who were doing better already, which would inflate the extension group's remission rate independently of the support itself.. The extension was not randomized. Participants chose whether to continue, so the comparison at 5 years is between people who opted into further support and people who did not. The absolute numbers are small: 13% is 11 people.

HbA1c of 6.0% or below, with or without medication, at 5 years: 14 of 49 (29%) after gastric bypass, 11 of 47 (23%) after sleeve gastrectomy, 2 of 38 (5%) on intensive medical therapy alone. Mean HbA1c reduction from baseline was 2.1 percentage points after surgery against 0.3 on medical therapy. Measured in: 134 people completing 5-year follow-up in STAMPEDE, mean age 49 (SD 8), mean baseline HbA1c 9.2%, mean BMI 37, BMI range 27 to 43. The endpoint allows medication to be in use, so it is glycemic control, not remission under the 2021 consensus definition. This is a single-center trial and the arms are small: 29% is 14 people. Operative risk, reoperation and lifelong nutritional consequences sit outside the endpoint entirely.

At 10 years, remission by intention to treat was 9 of 18 (50.0%) after biliopancreatic diversion, 5 of 20 (25.0%) after gastric bypass and 1 of 18 (5.5%) on medical therapy (p=0.0082). Of the 34 participants in remission at 2 years, 20 (58.8%) had relapsed by 10 years. Diabetes complications were far less frequent in the surgical arms (relative risk 0.07). Measured in: 60 people with type 2 diabetes randomized at a single center in Italy, with 95% follow-up at 10 years. Sixty people across three arms is a very small trial, so each percentage point rests on a fraction of a person. Biliopancreatic diversion, the arm with the highest remission rate, is also the most malabsorptive and is rarely performed now.

Over 9 months against a non-exercise control, combined aerobic and resistance training lowered HbA1c by 0.34 percentage points (p=0.03). Aerobic alone (0.24, p=0.14) and resistance alone (0.16, p=0.32) did not reach significance. Baseline HbA1c in the control group was 7.7%. Measured in: 262 adults with type 2 diabetes, 63% women, 47.3% non-white, mean age 55.8, in the HART-D trial. The single-mode arms were not powered to detect effects this small, so a non-significant result there is not the same as no effect. Baseline HbA1c was close to target at 7.7%, which leaves less room to improve than in trials recruiting worse-controlled participants.

Walking 10 minutes after each main meal lowered the 3-hour incremental glucose area under the curve to 0.88 of the same total walking done as one 30-minute daily bout (95% CI 0.78 to 0.99), a 12% reduction. After the evening meal the ratio was 0.78 (0.67 to 0.91), a 22% reduction. Measured in: 41 adults with type 2 diabetes, mean age 60 (SD 9.9), each completing both 2-week conditions in randomized order with continuous glucose monitoring. This measures postprandial glucose over two weeks, not HbA1c and not any clinical outcome. The evening-meal advantage is confounded with meal composition, since that was the meal carrying the most carbohydrate in this population.

Against continuous sitting, light-intensity walking breaks lowered postprandial glucose (SMD -0.72, 95% CI -1.03 to -0.41) and insulin (SMD -0.83, -1.18 to -0.48). Standing breaks lowered glucose less (SMD -0.31, -0.60 to -0.03) and did not lower insulin. Neither changed systolic blood pressure. Measured in: 7 randomized crossover trials in mixed-sex adults, predominantly with overweight or obesity. These are acute laboratory studies measuring a single day, so nothing here shows the effect persists or changes HbA1c. The authors specifically note the free-living feasibility has not been tested.

Across 10 dietary approaches, ketogenic, low-carbohydrate and low-fat patterns significantly reduced HbA1c, and moderate-carbohydrate, low glycemic index, Mediterranean, high-protein and low-fat patterns significantly reduced fasting glucose. The ketogenic pattern ranked highest for HbA1c. Measured in: 42 randomized controlled trials, 4,809 people with type 2 diabetes. Network meta-analysis infers most of these comparisons indirectly, and diet trials are short and open-label, so ranking reflects short-term adherence as much as physiology. The ketogenic ranking rests on the fewest and shortest trials in the network.

At 6 months, low and very low-carbohydrate diets produced higher rates of remission (defined as HbA1c under 6.5%) than control diets, alongside clinically meaningful weight loss and HbA1c reduction. At 12 months the effects ranged from minimal to trivial. Adverse events did not differ between groups at either point. Measured in: 23 randomized trials, 1,357 people with type 2 diabetes, including unpublished trial data. The six-month remission definition allowed continued glucose-lowering medication in some trials, which is not remission under the 2021 consensus. Remission rates fell substantially when insulin-treated participants were included, and certainty of evidence was rated moderate to low.

Over 4 years, 44% of the Mediterranean-style group started antihyperglycemic drug therapy against 70% of the low-fat group, an absolute difference of 26 percentage points. The Mediterranean group also had greater improvements in glycemic control, coronary risk factors and weight. Measured in: 215 people with newly diagnosed type 2 diabetes and overweight, randomized to a Mediterranean-style diet with under 50% of calories from carbohydrate (n=108) or a low-fat diet with under 30% of calories from fat (n=107). Single-center, unblinded, and the decision to start medication was made against an HbA1c threshold of 7%, so the endpoint depends on the glycemic effect, not being independent of it. The low-fat comparator was a specific prescription, not usual eating.

Real-time CGM lowered HbA1c by 0.19 percentage points and intermittently scanned CGM by 0.31. Real-time CGM reduced user satisfaction; intermittently scanned CGM improved it. Both increased adverse event risk (real-time CGM RR 1.22, intermittently scanned somewhat higher). Neither changed body composition, blood pressure or lipids. Measured in: 26 randomized controlled trials (17 real-time CGM, 9 intermittently scanned), 2,783 people with type 2 diabetes. Sensor technology changed substantially across the span of the included trials, so the pooled figure averages devices that are no longer comparable. Most trials were short, and the participants were mostly on insulin or intensive regimens, not diet alone.

Over 12 months, diabetes developed in 18.18% of the Tianqi group against 29.32% on placebo (p=0.01), a 32.1% relative risk reduction. Normal glucose tolerance was restored in 63.13% against 46.60% (p=0.001). Body weight and BMI did not differ between groups. No severe adverse events occurred. Measured in: 420 people with impaired glucose tolerance randomized in a double-blind multicenter trial in China; 389 completed (198 Tianqi, 191 placebo). This is one trial of one ten-herb proprietary capsule, conducted entirely in China, and it has not been independently replicated outside that setting. It tested prevention in prediabetes, so it says nothing about treating established diabetes. The endpoint is conversion to diabetes defined by an oral glucose tolerance test, a diagnostic threshold, not a hard clinical outcome. That is still unusual rigor for this literature, and it is not the same as preventing an event.

Across 69 different herbal preparations, several showed significant glucose lowering against placebo, seven showed better metabolic control than prescription hypoglycemic drugs, and 15 combinations showed additional benefit over conventional medication alone. The reviewers judged none recommendable on this evidence. Measured in: 66 randomized trials, 8,302 participants with type 2 diabetes, searched to April 2004. Low methodological quality, small samples and probable publication bias run through the whole set, and the review is now two decades old. The claim that seven preparations beat prescription drugs is exactly the kind of result that low trial quality generates and that has not been replicated in independent settings.

Berberine combined with lifestyle intervention lowered fasting glucose, post-load glucose and HbA1c compared with lifestyle intervention alone. Combined with oral hypoglycemic drugs it added further reduction. For lipids it lowered triglycerides and raised HDL cholesterol. No serious adverse reactions were reported across the 27 trials. Measured in: 27 randomized controlled trials, 2,569 people, across type 2 diabetes, hyperlipidemia and hypertension. The reviewers rated the included trial quality as limited, and almost all trials were conducted in China where the adulteration of herbal antidiabetic products with sulfonylureas is documented, which is a specific reason to treat glucose-lowering findings in this literature with care. Berberine also inhibits CYP enzymes and P-glycoprotein, so interactions with prescription drugs are plausible and largely unmeasured.

403 primary events in the intervention group against 418 in the control group, HR 0.95 (95% CI 0.83 to 1.09, p=0.51), over a median 9.6 years. Weight loss was 8.6% against 0.7% at year 1 and 6.0% against 3.5% at the end. HbA1c, fitness and most risk markers improved. Measured in: 5,145 adults with type 2 diabetes and overweight or obesity across 16 US centers, randomized to intensive lifestyle intervention or diabetes support and education. The trial was stopped early for futility, so it cannot exclude a benefit emerging later. Control-group participants also received standard diabetes care including increasing use of statins and glucose-lowering drugs, which narrows the gap the intervention had to beat.

Compared with conventional diet-first treatment, metformin gave a 32% risk reduction for any diabetes-related endpoint (95% CI 13 to 47, p=0.002), 42% for diabetes-related death (9 to 63, p=0.017) and 36% for all-cause mortality (9 to 55, p=0.011), over a median 10.7 years. Metformin also caused less weight gain and fewer hypoglycemic episodes than insulin or sulfonylureas. Measured in: 1,704 overweight people with newly diagnosed type 2 diabetes in UKPDS, of whom 753 formed the primary randomized comparison, mean age 53. The comparator was a diet-first policy from the 1980s, not modern care, so the size of the benefit against today's alternatives is not what this trial measured. A secondary UKPDS comparison in which metformin was added to sulfonylurea showed increased diabetes-related death, a result the investigators could not explain and that has not been replicated. Metformin monotherapy typically lowers HbA1c by around 1 percentage point. Set that beside the exercise figure on this page: at 0.67 points, structured exercise reaches a good fraction of what metformin does.

Major adverse cardiovascular events fell 14% (HR 0.86, 95% CI 0.80 to 0.93, p<0.0001), all-cause mortality 12%, hospital admission for heart failure 11%, and a composite kidney outcome 21%. No increase in severe hypoglycemia, retinopathy, pancreatitis or pancreatic cancer. Measured in: 8 cardiovascular outcome trials, 60,080 people with type 2 diabetes, most at high cardiovascular risk. The trials enrolled people at elevated cardiovascular risk, so the absolute benefit in someone newly diagnosed and otherwise well is smaller than these relative figures imply. Most of the kidney composite is driven by albuminuria change, not by dialysis or death.

Primary composite of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke occurred in 10.5% on empagliflozin against 12.1% on placebo (HR 0.86, 95.02% CI 0.74 to 0.99, p=0.04). Cardiovascular death fell 38%, hospitalization for heart failure 35% and death from any cause 32%. Genital infection was more common. Measured in: 7,020 people with type 2 diabetes at high cardiovascular risk, median observation 3.1 years. Everyone enrolled had established cardiovascular disease, so this does not describe someone recently diagnosed with no vascular history. Empagliflozin did not reduce non-fatal myocardial infarction or stroke individually; the composite was driven by the cardiovascular death component.

In drug-naive people at low cardiovascular risk, no glucose-lowering treatment differed from placebo for vascular outcomes. In people at increased cardiovascular risk already on metformin, oral semaglutide, empagliflozin, liraglutide, extended-release exenatide and dapagliflozin reduced all-cause mortality, and SGLT2 inhibitors reduced heart failure hospitalization and end-stage kidney disease. Measured in: 453 randomized trials of glucose-lowering drugs in adults with type 2 diabetes. Network meta-analysis compares treatments that were often never tested head to head, so the comparisons rest on the assumption that the trial populations were similar enough to pool. The low-risk group also had fewer events, which limits what could be detected there.

In the largest tirzepatide trial so far, people with type 2 diabetes and existing cardiovascular disease had a similar rate of heart attack, stroke and cardiovascular death on tirzepatide as on dulaglutide, another injectable already shown to protect the heart. Tirzepatide did not beat dulaglutide on that endpoint, and it delivered larger improvements in glucose, weight and kidney measures. Everyone here already had heart disease, so the day-to-day event rate for someone recently diagnosed is lower.

The primary composite of end-stage kidney disease, doubling of serum creatinine, or renal or cardiovascular death occurred at 43.2 per 1,000 patient-years on canagliflozin against 61.2 on placebo (HR 0.70, 95% CI 0.59 to 0.82, p=0.00001). End-stage kidney disease alone fell 32% (HR 0.68, 0.54 to 0.86). Measured in: 4,401 people with type 2 diabetes and albuminuric chronic kidney disease (eGFR 30 to under 90, urinary albumin-to-creatinine ratio above 300), all on renin-angiotensin blockade, median follow-up 2.62 years. This was a trial in people who already had significant albuminuric kidney disease, and everyone was already on an ACE inhibitor or ARB, so the benefit is additive to that, not instead of it. It was stopped early on efficacy, which tends to overestimate effect size.

Weight loss at 12 months was -11.7 lb (-5.3 kg) on healthy low-fat against -13.2 lb (-6.0 kg) on healthy low-carbohydrate, a between-group difference of 1.5 lb (0.7 kg), 95% CI -0.2 to 1.6. No significant interaction with genotype pattern (p=0.20) or with baseline insulin secretion (p=0.47). Measured in: 609 adults aged 18 to 50 with overweight or obesity, 57% women, mean BMI 33; 481 completed. These participants did not have type 2 diabetes, so the transfer is partial: the diet comparison here is about weight and the predictors, not about glycemic control in diabetes. Both arms received intensive dietitian support that most people do not get.

Among normoglycemic people classified as having the most variable glucose pattern, about a quarter of the sample, monitored time reached up to 15% in the prediabetic range and 2% in the diabetic range. Measured in: 57 adults wearing continuous glucose monitors, characterized alongside standard glucose tolerance and insulin testing. What could explain it instead: Sensor bias against blood glucose. The device systematically reads higher than capillary blood, so time spent in the prediabetic and diabetic ranges is partly a property of the measurement, not of the person.. This describes the most variable quarter, not normoglycemic people generally, and 15% is an upper bound, not an average. Read as a general figure it would make continuous monitoring look more alarming than the paper supports.

CGM-estimated fasting and postprandial glucose ran 16 ± 11 and 16 ± 9 mg/dL (0.9 ± 0.6 and 0.9 ± 0.5 mmol/L) above capillary estimates (both p<0.001). The size of the bias varied by test food and by individual. CGM overestimated time above 140 mg/dL (7.8 mmol/L) roughly fourfold, falling to roughly twofold after adjusting for the baseline difference. Measured in: 15 healthy adults, each completing seven laboratory visits with randomized carbohydrate challenges including glucose, whole fruit, blended fruit and commercial smoothies, sampled every 15 minutes for 120 minutes. Fifteen people, one sensor type, one laboratory. The finding is about a specific device against capillary sampling, and it should not be read as a general property of every monitor on the market.