Issue 11 · 16 September 2026
A special issue on intermittent fasting. Eight papers from the last two years on a single methodological fact that runs underneath the whole field: measured against usual diet, fasting produces large effects with high certainty, and measured against a matched calorie restriction, most of them disappear.
Issue 11: Cochrane read twenty two trials, and the comparator decided the answer
28:52 · 16 September 2026
What fasting was measured against
Four views of the same point. Each panel carries its own outcome and its own units, and no panel should be read against another.
One review, two comparisons, two answers
Percentage of body weight lost, from the same Cochrane review of the same twenty two trials. The only thing that changes between the two rows is what fasting was compared against.
The interval on the first row runs from minus 0.92 to plus 0.26, which includes zero and includes fasting being slightly worse. The second row is the only moderate certainty estimate in the review.
Hold the calories equal and the differences go away
Three separate trials, three different designs, each comparing a fasting arm against a control that was matched on intake. Each card stands alone.
2.7
against 2.5 percent body weight a ten hour eating window against a sixteen hour one, with all food supplied and calories matched and set to hold weight steady. Duan 2026
0.46
p value for the difference in lean tissue between lifters on a sixteen eight window and lifters eating across the day, with calories and protein statistically matched. Blake 2025
384
calories a day less, reported during a four hour eating window, which is where the weight loss came from and why the hunger hormones responded as they did. Kramer 2026
The Duan trial is the only design in this issue that supplied the food. The other two matched intake by prescription and measurement.
Two studies that watched what happened after twelve weeks
Different outcomes on the same axis of elapsed time. The bars mark how long each measurement ran, not how long any effect took to appear.
The Parr margin was 0.10 percent glycated haemoglobin, written down before the data were seen. The changes in both arms were small and the authors call them not clinically meaningful at either time point.
Eight weeks, four arms, and what actually came off
Fat mass lost in kilograms, from one trial with one protocol and one set of units. The lean tissue numbers are in the note, because they move in the opposite direction and do not belong on the same axis.
Lean tissue went the other way. The window alone lost 2.3 kilograms of it, training alone gained 1.6, and the combination held it steady. The arm that lost the most body weight without training lost none of its fat.
In this issue
The first Cochrane review of intermittent fasting split the question in two. Against regular dietary advice the difference in weight was minus 0.33 percent and the interval crossed zero. Against no intervention at all it was minus 3.42 percent. Same review, same trials, two comparators.
Cochrane Database of Systematic · Garegnani et al. · the Research Department at Universidad Hospital Italiano de Buenos Aires · 16 February 2026 · PMID 41692034
Three degrees of ordinary calorie cutting and four kinds of fasting, compared at matched absolute energy restriction across 167 trials and nearly twelve thousand people. The ranking came out by how much energy each regimen removed. The authors state the conclusion as a negative.
Nutrition Reviews 84(3):487 to · Wu et al. · the Shanghai Institute of Endocrine and Metabolic Diseases at Ruijin Hospital, Shanghai Jiao Tong University School of Medicine · undefined March 2026 · PMID 40367516
Twelve weeks, food provided, calories matched between arms and set to maintain weight. Ten hours of eating against the sixteen people normally use. Weight fell 2.7 percent and 2.5 percent. Appetite hormones, inflammatory markers and eating behaviour did not separate.
Obesity 34(9):1732 to 1741 · Duan et al. · the Division of Endocrinology, Diabetes and Metabolism at Johns Hopkins University School of Medicine · 27 July 2026 · PMID 42508996
Thirty nine adults with type 2 diabetes did six weeks of twenty hours fasting and six weeks of their usual life, in random order. The weight came off through a large reported calorie deficit. Leptin fell, fasting ghrelin rose, and the gut peptides did not move.
International Journal of Obesity · Kramer et al. · the Leadership Sinai Centre for Diabetes at Mount Sinai Hospital and the University of Toronto · 11 March 2026 · PMID 41814085
Two hundred and forty seven Australian adults at raised risk of type 2 diabetes. Half got a nine hour window, half got individualised dietetic guidance, both got the same three hours of telehealth contact. The trial wrote down in advance how much worse the window was allowed to be.
Diabetologia 69(9):2458 to 2471 · Parr et al. · the Mary MacKillop Institute for Health Research at Australian Catholic University · 6 June 2026 · PMID 42251202
Four arms for eight weeks: a ten hour eating window, resistance training, both, and neither. The window alone dropped more than two kilograms of body weight, and the fat mass did not move. What came off was lean tissue. Adding the training changed the answer completely.
Journal of the International · Cui et al. · the Institute of Human Movement and Sports Engineering at Zhejiang Normal University · 19 March 2025 · PMID 40108888
Seventeen well trained lifters, eight weeks of supervised training, both arms ten percent above maintenance at 2.2 grams of protein per kilogram, and intake statistically matched. The fasting arm did fewer repetitions, rated its daily energy lower, and gained about four kilograms less on the squat.
Journal of the International · Blake et al. · the Center for Sport Performance at California State University Fullerton · 16 April 2025 · PMID 40241374
Ninety seven adults with mild to moderate Crohn's disease. Sixty five did five days a month of a fasting mimicking diet for three months. Seven in ten met the response threshold against four in ten of the control group, and faecal calprotectin fell twenty two percent while it rose in the control arm.
Nature Medicine 32(3):1023 to · Kulkarni et al. · the Division of Gastroenterology and Hepatology at Stanford University School of Medicine · 13 January 2026 · PMID 41530382
Ask what it was measured against
Opinion Our reading of what these papers add up to. Everything above this line came from the papers. This part did not.
The honest summary of this literature is short. Fasting protocols produce clinically useful weight loss and the metabolic improvements that follow from it. They do so about as well as ordinary calorie restriction, and no better. That is not a small finding and it is not a dismissal, but it does change what the remaining question is, because if the schedule is not the active ingredient then the only thing worth asking about a protocol is whether it makes eating less easier for you.
What makes the field confusing is that both of those sentences are supported by published evidence, and which one you meet depends on the comparison a given paper chose. The Cochrane review is the cleanest demonstration, because it ran both comparisons on the same twenty two trials and printed both answers. Against ordinary dietary advice, a third of one percent of body weight and an interval that includes zero. Against no intervention, three and a half percent and the highest certainty rating in the review. Nothing about the intervention changed between those two lines.
There is a disagreement on the record about what to do with this, and it is worth stating rather than smoothing. In June of this year three of the field's principal trialists published a Personal View in Lancet Diabetes and Endocrinology arguing that the evidence has matured enough for intermittent fasting to enter diabetes practice guidelines, citing reductions in glycated haemoglobin of typically 0.3 to 1.2 percent, and they are explicit that those figures are relative to usual eating patterns. A Personal View is an opinion piece rather than a systematic review. In July, a systematic review and meta-analysis in Diabetologia put the pooled figure at minus 0.30 percent, with an interval running from minus 0.63 to plus 0.02 and a p value of 0.06. Same year, same question, and the gap between the two answers is a comparator.
The other thing this field has taught, and it was taught the hard way, is what a conference abstract can do. In March 2024 an abstract reporting that an eight hour eating window carried a ninety one percent higher risk of cardiovascular death travelled around the world on a press release. It was a conference abstract, not peer reviewed, resting on four hundred and fourteen exposed participants and thirty one cardiovascular deaths. The peer reviewed paper arrived in 2025 carrying a different number, reached after the reference category had been narrowed, and reporting no association with all cause mortality. Its authors close by asking whether the finding reflects the short eating duration itself or residual confounding. That paper is not in this issue. It is worth mentioning anyway, because the abstract is still the version most people met.
Two practical things fall out of these eight papers, and neither is advice about whether to fast. The first is that the one reproducible property of a compressed window is that it suppresses intake, which is useful if the goal is to eat less and a problem if the goal is to hit a protein or energy target. The second is that when a window runs without resistance training, the eight week trial here found the weight it removed was lean tissue and the fat did not move at all. That is not an argument against the window. It is an argument that the window and the training are not separable questions.
And the issue ends where it does for a reason. The Crohn's trial is the one study here where a fasting protocol clearly beat what it was measured against, on a clinical outcome, with an objective marker moving in the same direction. It is open label, its control arm received no diet at all, and its primary outcome is a symptom score. It also has a stool marker that does not care what anyone believed. That combination is what a result looks like when it is worth building a blinded trial around, which is a different thing from a result that is ready to act on.
Substrate Digest
Newsletter
Each issue by email, when the newsletter launches. Leaving your address puts you on the list, nothing is sent yet.