Two carbohydrate days a week, and one result that survived review#
A refeed is a short, planned stretch — usually one or two days — where you raise intake to about maintenance and put nearly all of the extra energy into carbohydrate, then drop straight back into the deficit. It is not a cheat day, which is unbudgeted and food-agnostic, and it is not a diet break, which is one to two weeks at maintenance. Exactly one randomized controlled trial has isolated the refeed itself, and the most useful thing about this topic is what happened to that trial after publication: it reported preserved fat-free mass and resting metabolic rate, an independent reanalysis found only one of those outcomes held, and the lead author published a reply agreeing.
So the defensible claim is narrow — about 0.9 kg of dry lean tissue over seven weeks, in trained lifters, on a diet that averaged a 26% deficit. What did not survive matters more for how people actually use refeeds, because the leptin rationale the practice is sold on was contradicted inside the same trial that produced the good news. Below: the study, the reanalysis, the design detail nobody quotes, and the one mechanism that might be doing the work and was never measured.
The trial, then the reanalysis, then the reply#
Twenty-seven resistance-trained men and women were randomized for seven weeks alongside four training sessions a week and 1.8 g of protein per kilogram of bodyweight in both arms. The refeed group ran a 35% energy deficit for five days and then ate predetermined maintenance calories for two consecutive days, with the added energy specified as carbohydrate only. The continuous group ran a flat 25% deficit every day1.
The paper's title claimed attenuated loss of fat-free mass. Then Jackson Peos, Andrew Brown, Colby Vorland, David Allison and Amanda Sainsbury reanalyzed the reported data both completers-only and intention-to-treat, and found the headline did not hold2.
| Outcome | Original paper | On independent reanalysis |
|---|---|---|
| Fat-free mass | −0.4 vs −1.3 kg, reported P = 0.006 | not significant (P = 0.18 completers, P = 0.17 ITT) |
| Resting metabolic rate | −38 vs −78 kcal/day | not significant (P = 0.49 completers, P = 0.45 ITT) |
| Dry fat-free mass | −0.2 vs −1.9 kg | significant, ~0.9 kg difference (P = 0.0004 completers, P = 0.0028 ITT) |
| Fat mass | −2.8 vs −2.3 kg | no group difference claimed |
The lead author's response is the part that makes this a model rather than a scandal. He conceded the framing directly — "we apologize if the manuscript unintentionally implied that we observed changes that represented a statistically significant between-group difference at an alpha level of 0.05" — while defending per-protocol analysis as normal practice in sports-nutrition work, and proposed a revised conclusion limited to dry fat-free mass3.
One disclosure belongs on the record before spending any of this: the trial was funded by the Dymatize Athletic Nutrition Institute and the lead author sits on that company's scientific advisory board, though the paper states the funder had no role in design, analysis or the decision to publish. The finding that survived is also the modest one, which is not the shape a marketing result usually takes.
Why the surviving outcome is the one worth having#
Dry fat-free mass sounds like a technicality and is not. Fat-free mass as normally measured includes the water in the tissue; dry fat-free mass strips that out and leaves the protein and mineral — the part you actually mean when you say muscle.
That distinction cuts in a specific direction here, and the reasoning is mine rather than either paper's. A carbohydrate refeed refills muscle glycogen, and glycogen is stored wet, so the refeed arm has a mechanism for carrying extra water inside its wet fat-free mass reading. If the group difference were an artifact of the intervention's own water, you would expect it to show up in the wet measure and vanish in the dry one. It did the opposite: the wet number failed and the dry number was the strongest result in the reanalysis. The finding sits in the compartment that is hardest to fake by refeeding.
The design carries a second wrinkle that runs the same way. Averaged across the week, the refeed arm ran a 26% deficit and the continuous arm 21% — the group with the pause was, in net terms, the group eating less. A 27-person completer sample drawn from 58 recruits is small and attrition-heavy, and none of this promotes one trial into a body of evidence. But the two obvious ways to explain the result away — extra water, an easier diet — both point the wrong direction.
The hormone story, contradicted by the trial that tells it#
Almost every refeed protocol on the internet is justified by leptin: eat carbohydrate, leptin rebounds, metabolism restarts. The first half of that has good support. In 21 women with obesity across four weeks of restriction and one week of refeeding, plasma leptin fell by up to 66% within the first week, tracked changes in glucose rather than changes in fat mass, and rose again during refeeding while fat mass was still falling5. Leptin is reporting today's energy intake, not this month's body fat — which is exactly why a carbohydrate day moves it. (What short-term carbohydrate overfeeding does to leptin and to expenditure in non-dieting people is measured in calorie cycling; what the hormone signals in general is leptin resistance explained.)
The second half is where it breaks, twice. Campbell's own blood subsample — eight participants — showed leptin falling from 3.9 to 1.6 ng/mL across the seven weeks, refeeds and all. Two carbohydrate days a week did not hold the hormone up over a diet.
And even if they had, the payoff people expect is not on offer. Six to eight lean men were studied through 72-hour fasts with placebo, low-dose leptin, or a replacement dose designed to hold leptin at fed-state levels. Keeping leptin up was "necessary and sufficient" to protect the reproductive axis — it fully restored luteinizing-hormone pulsatility and reversed a roughly 40% drop in testosterone — and it partially regulated TSH pulsatility and IGF-1. Fuel utilization, energy expenditure, cortisol, growth-hormone pulsatility, hunger and post-fast intake all moved independently of leptin4.
The one experiment that held leptin at fed-state levels through a fast changed the reproductive axis and left metabolic rate and hunger where they were.
That is the cleanest available test of the refeed's advertised mechanism, and it was run in the direction that would have proved it. Leptin moves when you eat carbohydrate. It is not the wire connected to your metabolic rate.
The mechanism with better odds is the one nobody measured#
The alternative explanation is mechanical: five days of restriction drains muscle glycogen, two high-carbohydrate days refill it, and a refilled lifter trains harder — more volume, more stimulus, less lean tissue given up. Campbell's group proposed something close to this and were explicit about the gap, noting that not measuring acute insulin changes during the refeeds "limited our ability to describe potential mechanisms." Training volume was not reported as an outcome either. The mechanism is a hypothesis the trial was not built to test.
It also has a complication. A review of carbohydrate availability in resistance-based exercise concludes that severe carbohydrate restriction may not impair strength adaptations over a training program, while carbohydrate before a testing session does help maximal performance, and that the effect on hypertrophy remains unsettled6. So the glycogen route is plausible for the acute quality of a hard session and weaker as a story about muscle retention over weeks — which leaves the surviving dry-lean-mass result standing without an established mechanism underneath it. That is a normal state for a single trial to be in, and a bad reason to build a rulebook.
Running one, and reading the scale afterwards#
| Refeed | Cheat day | Diet break | |
|---|---|---|---|
| Length | 1–2 days | one meal or one day | 1–2 weeks |
| Intake | maintenance, calculated | unbudgeted | maintenance, calculated |
| Composition | the extra energy is carbohydrate | anything | normal diet, protein held |
| Evidence for it | one small RCT, one outcome after reanalysis | none as a protocol | pooled trials: no fat-loss gain, smaller RMR fall |
| Best argument | protecting dry lean mass and hard sessions | none | making a long diet survivable |
Three practical notes follow from the trial as actually run. The refeed days were maintenance days, not surplus days — nothing in this evidence supports eating above maintenance. The extra energy was carbohydrate, protein was held at 1.8 g/kg in both arms, and the restricted days were correspondingly steeper so the week still averaged a deficit; a refeed you did not subtract from anywhere is just a bigger diet. And expect the scale to jump a kilogram or more within a day or two of the refeed, because refilling glycogen brings its water with it — that rebound is ordinary fluid movement, and reading it as fat is the single most common way people abandon the practice after one attempt.
What a refeed is not is insurance. The lean tissue you keep through a diet is decided mostly by protein intake and resistance training, and the amount you lose is decided by the week's energy total. A refeed is a small, contested adjustment sitting on top of both.
FAQ#
What is the difference between a refeed day and a cheat day?#
A refeed is planned, capped at calculated maintenance, and composed deliberately — in the one trial that tested it, the added energy was carbohydrate only, protein stayed at 1.8 g/kg, and the surrounding days were cut deeper so the week still averaged a 26% deficit1. A cheat day has no ceiling, no composition target and no subtraction anywhere else. They are different interventions that happen to share a calendar slot.
How many calories should a refeed day be?#
Maintenance for your current weight, not more. The trial's refeed days were "predetermined maintenance calories," with the increase over the restricted days delivered as carbohydrate; nothing in the evidence supports a surplus. Because that maintenance figure falls as you get lighter, the number needs re-deriving during a long diet rather than inherited from the start of one.
Do refeed days actually preserve muscle?#
One outcome in one small trial says maybe. After independent reanalysis of the only RCT to isolate a refeed, fat-free mass and resting metabolic rate showed no significant group difference, and dry fat-free mass — the protein-and-mineral fraction, with tissue water removed — favored the refeed group by about 0.9 kg over seven weeks2. Twenty-seven trained completers, seven weeks, no measured mechanism. Treat it as a promising single result, not a requirement.
Sources#
- Campbell BI, Aguilar D, Colenso-Semple LM, et al. Intermittent Energy Restriction Attenuates the Loss of Fat Free Mass in Resistance Trained Individuals. A Randomized Controlled Trial. J Funct Morphol Kinesiol. 2020;5(1):19.
- Peos JJ, Brown AW, Vorland CJ, Allison DB, Sainsbury A. Contrary to the Conclusions Stated in the Paper, Only Dry Fat-Free Mass Was Different between Groups upon Reanalysis. J Funct Morphol Kinesiol. 2020;5(4):85.
- Campbell BI. A Reply to Contrary to the Conclusions Stated in the Paper, Only Dry Fat-Free Mass Was Different between Groups upon Reanalysis. J Funct Morphol Kinesiol. 2020;5(4):86.
- Chan JL, Heist K, DePaoli AM, Veldhuis JD, Mantzoros CS. The role of falling leptin levels in the neuroendocrine and metabolic adaptation to short-term starvation in healthy men. J Clin Invest. 2003;111(9):1409-1421.
- Wisse BE, Campfield LA, Marliss EB, Morais JA, Tenenbaum R, Gougeon R. Effect of prolonged moderate and severe energy restriction and refeeding on plasma leptin concentrations in obese women. Am J Clin Nutr. 1999;70(3):321-330.
- Cholewa JM, Newmire DE, Zanchi NE. Carbohydrate restriction: Friend or foe of resistance-based exercise performance? Nutrition. 2019;60:136-146.
