You’re caught up.

Home/Issues/Issue 11/Study

Bayesian network meta-analysis PubMed records a month level publication date, March 2026, with no day. Volume 84, issue 3, pages 487 to 499 are assigned, so the record is final rather than ahead of print.

Across one hundred and sixty seven trials, the size of the energy cut predicted the weight loss and the meal timing did not

AI narration, generated on first listen
Journal
Nutrition Reviews 84(3):487 to 499
Authors
Wu X, Ding Y, Cao Q, Huang J, Xu X, Jiang Y, Xu Y, Lu J, Xu M, Wang T, Zhao Z, Wang W, Ning G, Bi Y, Li M
Institution
the Shanghai Institute of Endocrine and Metabolic Diseases at Ruijin Hospital, Shanghai Jiao Tong University School of Medicine
Published
undefined March 2026
Source
PMID 40367516 · DOI 10.1093/nutrit/nuaf056
Design
Systematic review and Bayesian random effects network meta-analysis of randomised controlled trials, covering three degrees of continuous energy restriction and four categories of intermittent fasting, with regimens compared at matched absolute energy restriction. GRADE used to rate certainty. PROSPERO registration CRD42022379621.
Sample
One hundred and sixty seven trials, 11,998 adults with overweight, obesity or metabolic abnormalities.

What network meta-analysis is

A network meta-analysis compares treatments that were never tested against each other, by chaining them through the comparisons that were actually run. If A beat B in some trials and B beat C in others, the network estimates A against C. The strength is that it uses all the evidence at once. The weakness is that the answer depends on whether the trials being chained together are similar enough to chain.

Drawn from standard methodology, not from this paper.

Why they ran it

Continuous energy restriction and intermittent fasting are both recommended for weight loss, and the authors set out to evaluate them against each other rather than each against a control. The design choice that matters is that they held the absolute size of the energy restriction constant, which turns the question from whether fasting works into whether the eating pattern adds anything on top of the deficit.

Drawn from the paper's stated objective.

One hundred and sixty seven eligible randomised trials, 11,998 participants, searched from database inception to December 2022. Bayesian random effects network meta-analysis, GRADE applied throughout, registered on PROSPERO.

Most fasting regimens induced significant weight loss that was comparable with the weight loss from ordinary calorie restriction at a similar absolute energy restriction. Certainty across those comparisons ran from low to high.

Severe continuous energy restriction was the most effective single regimen, at 11.50 kilograms, interval 10.07 to 12.93, on moderate certainty evidence.

Alternate day fasting came next at 5.07 kilograms, interval 3.44 to 6.72, and this one carries high certainty evidence. Moderate continuous energy restriction was 6.09 kilograms, interval 5.26 to 6.93, moderate certainty.

The same pattern held for body measurements, blood pressure, blood lipids and glycemic profiles.

In subgroup analysis the weight loss effects of the fasting regimens rebounded after twelve weeks. The continuous restriction regimens did not.

The numbers

Trials and participants167 trials, 11,998 participants
Severe continuous energy restriction, weight change11.50 kg, 95 percent credible interval 10.07 to 12.93, moderate certainty
Alternate day fasting, weight change5.07 kg, 95 percent credible interval 3.44 to 6.72, high certainty
Moderate continuous energy restriction, weight change6.09 kg, 95 percent credible interval 5.26 to 6.93, moderate certainty
Subgroup analysis after twelve weeksfasting regimens rebounded, continuous restriction regimens did not
Search cutoffdatabase inception to December 2022

Why this might happen

Proposed by the authors This is the explanation the authors offer in their discussion. This study did not test it.

The authors offer no physiological mechanism for fasting, because their result is that there is nothing left to explain. They rank the regimens by how much energy each removes, find that the ranking reproduces the weight loss ordering, and write the conclusion as a negative: effectiveness in weight loss mainly depends on the extent of the energy restriction, regardless of the mealtime patterns.

Drawn from the paper's stated conclusion.

What this does not show

  • It does not show that intermittent fasting fails to cause weight loss. Most of the fasting regimens produced significant weight loss. The finding is that they did not produce more of it than an equivalent energy cut delivered a different way.
  • It does not show that severe restriction is the best diet. It shows severe restriction removed the most weight in trials that were mostly short. What happens after the trial ends is a different question, and the rebound subgroup is the only hint the paper offers.
  • It does not cover the newest evidence. Searches ran to December 2022, which excludes a large part of the time restricted eating literature, including everything else in this issue.
  • The rebound finding is a subgroup analysis. It is not a prespecified primary outcome and the authors offer no mechanism for it. It should be read as a lead rather than a result.

Where this leaves us

The largest analysis in this field says the eating window is not the active ingredient. The size of the deficit is.

Adults with overweight, obesity or metabolic abnormalities, in randomised trials indexed up to December 2022.

A trial that holds the deficit constant and varies only the window, with food supplied, which is the next study in this issue.

Caveats worth holding

  • The search stops at December 2022, so the most recent trials are absent.
  • Certainty across the comparisons ranges from low to high, and only the alternate day fasting estimate reaches high.
  • Credible intervals from a Bayesian network are not the same object as frequentist confidence intervals and should not be read as one.

Newsletter

Each issue by email, when the newsletter launches. Leaving your address puts you on the list, nothing is sent yet.