The main finding
In one sentence: Five years after randomization, people who had bariatric surgery had lost more weight (23% after gastric bypass, 19% after sleeve gastrectomy) than people on intensive medical therapy alone (5%), but the main blood-sugar comparison was not statistically significant in the analysis that accounted for missing data.
Group averages, not an individual prediction. A trial result describes what happened to a defined group over a defined time.
On this page
Why this study matters
Long-term randomized trials comparing surgery with medical treatment for type 2 diabetes are rare. This report follows the 150 people in the STAMPEDE trial out to five years. It is important for two reasons. First, it shows what happened to weight and blood sugar over a longer time. Second, it shows how the way researchers handle missing data can change whether a result counts as "significant". Canadian readers living with obesity and type 2 diabetes may find both points useful.
Who was in the study
The trial randomized 150 adults with type 2 diabetes and a body mass index (BMI) of 27 to 43. BMI is weight in kilograms divided by height in metres squared. One person died during the five-year follow-up, and 134 of the remaining 149 people (90%) completed five years. Among those 134, the average age at the start was 49 years (standard deviation 8), 66% were women, the average HbA1c was 9.2% (standard deviation 1.5), and the average BMI was 37 (standard deviation 3.5). HbA1c is a blood test that reflects average blood sugar over about three months.
What the researchers did
This is an extension report of a randomized trial, meaning the same people were followed for longer. At the start, a computer assigned each person by chance to intensive medical therapy alone, medical therapy plus Roux-en-Y gastric bypass, or medical therapy plus sleeve gastrectomy. The trial was unblinded, since surgery cannot be hidden, and ran at a single centre. All groups received intensive medical therapy. The main outcome was an HbA1c of 6.0% or less, with or without diabetes medicines.
The researchers reported the main outcome in two ways. The primary analysis used only completers, the people who came back at five years. They also ran intention-to-treat (ITT) analyses, which include everyone who was randomized and fill in missing values using statistical methods (called multiple imputation). ITT is generally the stricter approach.
What they found
Among completers, the HbA1c target was met by 2 of 38 people (5%) on medical therapy alone, 14 of 49 (29%) after gastric bypass, and 11 of 47 (23%) after sleeve gastrectomy. For gastric bypass versus medical therapy, the P values were 0.01 (unadjusted), 0.03 (adjusted), and 0.08 in the ITT analysis. For sleeve gastrectomy versus medical therapy, they were 0.03, 0.07, and 0.17. A P value below 0.05 is usually treated as statistically significant. In plain terms: the completer results favoured surgery, but the ITT results, which account for missing people, were not statistically significant (R32-C103).
Weight and other outcomes at five years were reported as observed changes from the start:
| Outcome at 5 years | Gastric bypass | Sleeve gastrectomy | Medical therapy alone |
|---|---|---|---|
| Body weight change | loss of 23% | loss of 19% | loss of 5% |
| Triglyceride change | loss of 40% | loss of 29% | loss of 8% |
| HDL cholesterol change | gain of 32% | gain of 30% | gain of 7% |
| Insulin use change | loss of 35% | loss of 34% | loss of 13% |
| General health score change (RAND 36, scale 0 to 100) | gain of 17 | gain of 16 | gain of 0.3 |
The summary reports P less than 0.05 for all of these comparisons (R32-C102). The percentage reduction in HbA1c from the start was 2.1% in the surgical groups versus 0.3% with medical therapy (P=0.003).
Side effects and people who stopped
The summary reports no major late surgical complications except one reoperation. One person in the medical therapy group died during follow-up. The extracted data show that 8 medical therapy participants and 1 sleeve gastrectomy participant withdrew early, and 6 more people were lost to follow-up later. The summary does not report side effects in detail.
What this study does not tell you
- The main blood-sugar result depends on the analysis method. The multiple-imputation ITT comparisons were not statistically significant, so the five-year blood-sugar benefit is less certain than the completer numbers suggest.
- The trial was small (150 people), unblinded, and run at one centre.
- Each report on the same trial adds more comparisons, which raises the chance that some results look significant by chance.
- The medical therapy predates the newer diabetes and weight-management medicines used today. This trial does not compare surgery with those medicines.
- The one-year and three-year STAMPEDE reports describe the same people and are not separate confirmations.
- The trial was funded in part by a maker of surgical devices.
What it means in Canada
STAMPEDE is often cited in Canadian discussions about type 2 diabetes and obesity. It does not tell you whether surgery is right for you, and it does not compare surgery with the medicines used today. For information on medicines, see semaglutide in Canada and tirzepatide in Canada. For questions about what provincial plans pay for, see public drug coverage by province.
Common questions
What were the five-year weight results in the STAMPEDE trial?
Observed mean weight change was a loss of 23% after gastric bypass, 19% after sleeve gastrectomy, and 5% with intensive medical therapy alone.
Was the main STAMPEDE result still statistically significant at five years?
It depends on the analysis. Among completers, 29% (bypass) and 23% (sleeve) versus 5% (medical) reached the HbA1c target, but in the intention-to-treat analysis the P values were 0.08 and 0.17, which are not significant.
How many people finished five years of follow-up in STAMPEDE?
One person died during follow-up, and 134 of the remaining 149 (90%) completed five years.
Related reading
Other studies in this library that bear on the same question.
STAMPEDE: bariatric surgery versus medical therapy for type 2 diabetes
STAMPEDE at three years: surgery versus medical therapy
Life expectancy after surgery: the Swedish Obese Subjects study
A 2025 Canadian cost-effectiveness model of surgery, medicines and behavioural care
SURMOUNT-2: tirzepatide for weight loss in type 2 diabetes
Source
Original paper: Philip R Schauer, STAMPEDE 5-year, The New England journal of medicine, 2017. PubMed: https://pubmed.ncbi.nlm.nih.gov/28199805/. DOI: https://doi.org/10.1056/NEJMoa1600869. Funding: The trial was funded by Ethicon, with additional support from LifeScan, the Cleveland Clinic and the National Institutes of Health. How we summarized it: full text and extracted data.
This page explains a published study. It is not medical advice and does not describe whether a medicine is right for you. Talk to your doctor, nurse practitioner or pharmacist about your own situation. Reviewed and signed by two pharmacists registered in British Columbia, 2026-09-18.
Evidence records behind this page
Each record is a row in the clinical evidence file, taken from the published paper named above. The caution column is the limit the researcher recorded for that number.
This table scrolls sideways.
| Record | What the paper reports | Type | Population | Caution | Source |
|---|---|---|---|---|---|
| R32-C102 | At 5 years in STAMPEDE, observed mean weight changes were-23% after gastric bypass,-19% after sleeve gastrectomy and-5% with intensive medical therapy alone. | direct evidence | Adults with T2D and BMI 27 to 43 | Small unblinded single-centre trial. Primary comparison sensitive to multiplicity/missing-data method; multiple-imputation ITT not significant. Medical therapy predates current incretin options. Primary 22449319;3-year 24679060. | PubMed 28199805 |
| R32-C103 | STAMPEDE's 5-year primary glucose-control comparisons were not statistically significant in the multiple-imputation intention-to-treat analysis, although observed-completer results favoured surgery. | direct evidence | Adults with T2D and BMI 27 to 43 | Small unblinded single-centre trial. Primary comparison sensitive to multiplicity/missing-data method; multiple-imputation ITT not significant. Medical therapy predates current incretin options. Primary 22449319;3-year 24679060. | PubMed 28199805 |