You’re caught up.

Home/Issues/Issue 11/Study

Randomised non-inferiority trial PubMed records 6 June 2026 as the publication date. Volume 69, issue 9, pages 2458 to 2471.

A nine hour eating window matched a dietitian at four months, and at twelve months it no longer counted as close enough

AI narration, generated on first listen
Journal
Diabetologia 69(9):2458 to 2471
Authors
Parr EB, Charrouf R, Hutchison AT, Flint SA, Teong XT, Vincent AD, Bravo-Garcia AP, Siviour Z, Wittert GA, Brennan L, Devlin BL, Hawley JA, Heilbronn LK
Institution
the Mary MacKillop Institute for Health Research at Australian Catholic University
Published
6 June 2026
Source
PMID 42251202 · DOI 10.1007/s00125-026-06762-x
Design
Two arm parallel group multicentre randomised non-inferiority trial. Primary outcome the between group difference in glycated haemoglobin at four months, analysed by covariate linear regression with multiple imputation against a prespecified margin of 1.1 millimoles per mole. Secondary outcomes at twelve months. Registered as NCT04762251.
Sample
Two hundred and forty seven Australian adults with overweight or obesity and a raised type 2 diabetes risk score.

What a non-inferiority trial is

Most trials ask whether a new treatment is better. A non-inferiority trial asks whether it is close enough, which is the more useful question when the new treatment is cheaper or easier to deliver. The researchers state the largest acceptable gap before they see any data, and if the interval stays inside it the treatment is non-inferior. It can pass without ever being better.

Drawn from standard trial design, not from this paper.

Why they ran it

Time restricted eating is usually tested against a control that receives nothing. This team tested it against the real world alternative, a dietitian, with contact time matched in both arms, and asked whether the window is close enough to serve as a substitute where dietetic support is scarce.

Drawn from the paper's stated aim.

One hundred and twenty four participants randomised to the nine hour self selected eating window with the last eating occasion by seven in the evening, and 123 to individualised dietetic guidance. Two Australian clinical research institutes, staff conducting assessments and analyses blinded to allocation.

Both groups received five personalised telehealth consultations between zero and three months, three hours of contact in total, specific to their allocated group.

Baseline glycated haemoglobin averaged 40 millimoles per mole, which is 5.8 percent, a normal value.

The non-inferiority margin was set in advance at 1.1 millimoles per mole, which is 0.10 percent.

At four months the window was non-inferior but not superior: minus 0.22 millimoles per mole, interval minus 0.72 to 0.30, p equals 0.40.

At twelve months the upper bound of the difference exceeded the margin, so non-inferiority could no longer be concluded: 0.47 millimoles per mole, interval minus 0.19 to 1.23, p equals 0.14.

The authors state that the absolute changes were small and not clinically meaningful in either group at either time point. Adverse events were minor and did not differ between interventions.

The numbers

Randomised124 to time restricted eating, 123 to individualised dietetic guidance
Baseline glycated haemoglobin40 mmol/mol, standard deviation 3, which is 5.8 percent
Non-inferiority margin, set in advance1.1 mmol/mol, or 0.10 percent
Four monthsminus 0.22 mmol/mol, 95 percent CI minus 0.72 to 0.30, p equals 0.40, non-inferior and not superior
Twelve months0.47 mmol/mol, 95 percent CI minus 0.19 to 1.23, p equals 0.14, non-inferiority no longer concluded
Contact time in both armsfive telehealth consultations, three hours total
Eating windownine hours, self selected, last eating occasion by 19:00
Adverse eventsminor and not different between interventions, tracked twelve months

Why this might happen

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

No physiological mechanism is at issue and the authors propose none. Their framing is about service delivery: a self selected nine hour window finishing before seven in the evening can be taught in a few consultations and then run without professional input, and the question is what that costs in glycaemic control. Their answer at four months is nothing measurable, and at twelve months they say they can no longer make the claim.

Drawn from the authors' stated conclusion.

What this does not show

  • It does not show the eating window made anyone worse. The point estimate at twelve months is 0.05 percent glycated haemoglobin and the interval crosses zero in both directions. Failing to conclude non-inferiority is not the same as demonstrating inferiority.
  • There was very little room to improve. Baseline glycated haemoglobin averaged 5.8 percent, which is normal. This is a population at risk, not a population with diabetes, and a trial in a normal range has a low ceiling.
  • It does not test either arm against no intervention. Both groups received three hours of personalised telehealth support, so nothing here says whether either approach beats doing nothing.
  • It does not tell you why the twelve month result drifted. The trial measured the outcome, not the adherence mechanism behind it. Whether people stopped keeping the window, or kept it and it stopped mattering, is not separated.

Where this leaves us

The eating window held level with a dietitian for four months, and by twelve months the trial could no longer say it was close enough.

Australian adults with overweight or obesity and a raised diabetes risk score, starting from a normal glycated haemoglobin.

The same design in people who actually have elevated glycated haemoglobin, where there is room for either arm to move.

Caveats worth holding

  • Participants and dietitians were unblinded. Only the assessors and analysts were blinded.
  • The population started with normal glycated haemoglobin, which limits how much either arm could move.
  • The twelve month comparison is a secondary outcome, not the primary.

Newsletter

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