You’re caught up.

Home/Issues/Issue 07

Issue 07 · 19 August 2026

Sixty years of carb loading, and two blinded trials

A special issue on one question: which parts of the carbohydrate loading protocol actually matter. Eight papers from 1981 to this year, including the only meta analysis of loading protocols that exists, the closest thing to a head to head comparison of depletion against a taper, and a count of how many performance trials were ever blinded.

8 studies Special issue. Evidence published between 1981 and 2026. 25 min listen Listen on Spotify

What sixty years of carb loading actually settled

Four beats, in the order the papers came. Every figure here appears on a study page in this issue, verified at its source.

The protocol got shorter, and nothing was lost

Three arms of one 1981 trial, run in the same runners with the same units, plus a later trial that removed the depletion phase altogether.

Classical3 days at 15% carbohydrate, then 3 at 70%
207
Modified3 days at 50%, then 3 at 70%
203
Taper only6 days at 50% throughout
159
One daya separate trial, 10 g/kg, no depleting bout
maximal
Day 0Race day

The first three rows are one trial and share their units, millimoles of glucosyl units per kilogram of wet tissue. The fourth is a separate trial with different methods and is placed on the day axis only, not on the same scale. Its result was maximal glycogen in type I, IIa and IIb fibres within 24 hours. Run times did not differ between the first three arms.

Six things were tested. Three predicted the overshoot

Meta regression across the cycling trials in the only quantitative meta analysis of loading protocols. Bars are R squared, the share of the variation between studies each factor explains.

Carbohydrate as a share of energyp less than 0.001
0.56
Glycogen left after the boutp less than 0.001, the emptier the bigger
0.49
Glycogen before startingp = 0.011, the same headroom story
0.18
Carbohydrate in grams per kilogramp = 0.177
0.03
Size of the depleting boutp = 0.574
0.00
The athlete's maximal oxygen uptakep = 0.949
0.00
0R² 0.60

Adding the top two factors to one model dropped heterogeneity from 92.4 to 66.1 percent, which says the trials have been disagreeing about protocol rather than about physiology. The bottom row is the one worth sitting with: how fit the athlete was explained none of the difference.

The knob that matters is the one being missed

Measured intake in the 48 hours before a race, against the loading guideline, in female triathletes at the 2024 IRONMAN World Championships.

The guidelinewhat the papers recommend
8 to 12 g/kg
What they atemean of the field
6.4 g/kg
0 g/kg14 g/kg

Six of twenty three reached the guideline. The interviews say the reason was not that they did not know it: travel, nerves, food access and how it feels to eat that much.

Mean intake was 6.4 plus or minus 2.1 grams per kilogram. Higher carbohydrate intake correlated with faster finish times, which is a correlation across twenty three athletes at one event and not a demonstration that the carbohydrate caused it.

The muscle filled. The stopwatch did not notice

A placebo controlled trial over 100 kilometres, with carbohydrate taken during the ride as a racer would.

+18%

Muscle glycogen. 572 against 485 millimoles per kilogram of dry weight after loading, a significant difference.

no change

100 km time. 147.5 against 149.1 minutes, p = 0.4. Loading was numerically faster by about one percent, which seven riders cannot resolve.

And of the fourteen loading trials with a performance outcome, two were double blind and placebo controlled. Both of those also fuelled during the test, and neither found a benefit.

Two blinded trials finding nothing is not the same as nothing being there. The glycogen literature is solid. It is the step from a fuller muscle to a faster race that has barely been tested.

Sherman 1981 . Bussau 2002 . Goforth 2003 . Solem 2025 . Fortis 2026 . Burke 2000 . Jones 2026

In this issue

Where the protocol argument actually sits

Opinion Our reading of what these eight papers add up to. Everything above this line came from the papers. This part did not.

The protocol most people picture comes from the late nineteen sixties. Several days of very low carbohydrate paired with hard training to empty the muscle, then several days of flooding it. The depletion half of that came out of a single elegant experimental design, one legged cycling, in which only the leg that had worked overshot. That is a good reason to think depletion matters. It is not the same thing as a trial that gave one group a depletion phase and another group none, and then compared them.

The worry usually raised about carbohydrate loading is that athletes do a one day version of a protocol built to run for a week, and that this is why it disappoints. It is a reasonable worry and the evidence does not support it. Sherman's runners reached the same muscle glycogen on a moderate diet as on a severe one, in 1981. Bussau's athletes hit their ceiling in twenty four hours with no depletion at all, in every fibre type. Goforth ran the closest thing to a head to head, and the depletion group did finish ahead, on day seven. On day five, which is where a real taper sits, the two protocols could not be told apart.

The meta analysis is the useful document here, because it is the only one that tests the knobs rather than the outcome. Two things predicted the size of the overshoot: how carbohydrate rich the diet was, and how empty the muscle was when loading began. Three things did not: the gram dose, the size of the depleting bout, and the athlete's maximal oxygen uptake. That last null deserves more attention than it will get. The idea that better trained athletes have less to gain from loading is widespread, and across sixty years of these studies aerobic fitness explained none of the variation in how much glycogen was stored.

The same lesson turns up in the most consequential error this literature has made. For six years the field carried the claim that women cannot supercompensate at all, on the strength of one trial. Change the denominator of the dose from a share of habitual energy intake to grams per kilogram of lean body mass, and the difference does not appear. The failure was not in the women and it was not in the physiology. It was in the prescription.

Then there is the field data, which is the least glamorous paper in this issue and possibly the most useful. At the IRONMAN World Championships, six of twenty three women reached the loading guideline, and the average intake was closer to an ordinary training day than to a loading day. The interviews say knowledge was not the missing piece. Travel, nerves, food access and how it feels to eat that much were. If the meta analysis is right that how carbohydrate rich the diet is drives the number, then the protocol failure that matters is not a skipped depletion phase. It is a dose that never arrived.

Which leaves the question everyone actually cares about, and here our confidence should drop. Burke's cyclists loaded, gained about eighteen percent more muscle glycogen, and rode a hundred kilometres in the same time. And of the fourteen loading trials with a performance outcome, twelve could not separate the diet from the belief in the diet. The two that could, both of which also fed the athletes during the test, found nothing.

It would be easy to land that harder than the evidence allows, so here is the other side. Two blinded trials is a thin basis for a null. Burke's loading condition was numerically faster by about one percent, which is a margin that decides races and a margin seven riders can never resolve. The depletion protocol did store more glycogen by day seven, and that is a real finding rather than a rounding error. And nobody has ever run a one day loading protocol with a performance test on the end of it, which is a strange gap given how long the one day protocol has been on the table.

So the honest position is narrower than either camp would like. The muscle physiology is settled and reproducible. Give a trained athlete a carbohydrate rich diet and a rest and glycogen overshoots, in men and in women, at almost any level of fitness, and faster than the classical calendar assumes. What is not settled is the last step, from a fuller muscle to a faster race. The field has spent sixty years measuring the first step very well and reporting it as though it were the second.

That is the gap worth naming. The problem is not that athletes are doing the protocol wrong. It is that the protocol was validated on a proxy.

Substrate Digest

Newsletter

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