A Real-World Keto Protocol Reported 7.7% Weight Loss — and No Scheduled Safety Labs
What Was Published
A research group at the Fondazione Policlinico Universitario Agostino Gemelli in Rome published a real-world observational study on 11 September 2026 in Nutrients, describing a sequential ketogenic protocol for patients with complicated obesity who had not responded to a calorie-restricted Mediterranean diet. The paper is open access, and the full text is at doi.org/10.3390/nu18182982.
The protocol has a name — KETOCOMP — and a two-step structure: four weeks of a food-based ketogenic phase at roughly 1,200 kcal per day and under 20 g of carbohydrate per day, followed by four weeks of controlled carbohydrate reintroduction at roughly 1,300 kcal per day and up to 100 g of carbohydrate. Both phases are supervised, delivered through ordinary food rather than meal-replacement products, and delivered in a clinical nutrition outpatient setting.
The headline result is a 7.7 percent reduction in body weight across the whole observed pathway, from a mean of 99.9 kg to 92.3 kg, with fat mass down 23.1 percent. The headline caveat is a sentence in the paper's own limitations section: biochemical safety assessments were not carried out on a predefined schedule at each study timepoint. In plain terms, the study did not systematically measure blood lipids, blood glucose or liver enzymes at fixed points, so it can report body composition but cannot speak to metabolic safety. Both of those things are in the same document, and a reader who takes only the first one away has taken away half the story.
The Design, in Numbers
This is a retrospective, observational, single-arm study. There was no control group, no randomisation and no blinded assessment. Each participant was compared against their own earlier measurements, which is the weakest of the common designs for attributing a change to an intervention.
| Item | Value |
|---|---|
| Design | Retrospective, observational, single-arm |
| Participants completing the keto phase | 25 |
| Participants with data after reintroduction | 21 |
| Sex split | 16 women / 9 men |
| Mean age | 48.3 years |
| Mean baseline BMI | 34.4 kg/m² |
| Keto phase | 4 weeks, ~1,200 kcal/day, <20 g carbohydrate/day |
| Reintroduction phase | 4 weeks, ~1,300 kcal/day, up to 100 g carbohydrate/day |
Two details in that table deserve attention before any result does. First, the sample is 25 people, shrinking to 21 by the final timepoint, because four participants were lost to follow-up. A sample that size cannot support subgroup conclusions, and the authors say so. Second, the paper notes that participation required completing the keto phase — so patients who stopped early were excluded from the cohort by definition. That makes a dropout rate impossible to calculate, which the authors also state outright. A protocol that only counts the people who finished it will always look more tolerable than it is.
The keto phase used a standardised 1,200 kcal target for everyone, regardless of body size. For a participant weighing 99.9 kg, that is a substantial energy deficit. The paper lists the absence of an isocaloric non-ketogenic control as a limitation, which means the study cannot separate the effect of carbohydrate restriction from the effect of eating roughly 1,200 calories a day.
What the Body Composition Numbers Showed
The results are reported as within-subject changes with significance tests. The keto phase produced the largest shifts, which is what the design would predict given it was also the deepest energy deficit.
| Measure | Keto phase (T1→T2, n=25) | Reintroduction (T2→T3, n=21) | Combined (T1→T3, n=21) |
|---|---|---|---|
| Body weight | −6.10 kg | −1.20 kg | −7.79 kg |
| BMI | −2.08 | −0.43 | −2.65 |
| Fat mass | −4.80 kg | not significant | −6.29 kg |
| Waist circumference | −5.62 cm | −1.14 cm | −6.40 cm |
| Fat-free mass | −1.40 kg | not significant | −1.50 kg |
Across the full pathway from baseline, body weight fell 7.7 percent, BMI fell 9.45 percent from 34.4 to 31.2, absolute fat mass fell 23.1 percent from 40.2 kg to 30.9 kg, and waist circumference fell 6.65 percent from 110.0 cm to 102.7 cm. Those are the figures the coverage will repeat, and they are accurately reported.
The row worth reading twice is fat-free mass. It fell 1.40 kg during the keto phase with a p value of 0.0051, which meets the study's threshold for significance. The paper is careful about what that does and does not mean: it states that skeletal muscle mass and function were not directly assessed, with no imaging, strength testing or physical performance measures, and concludes that the stability of two bioelectrical impedance-derived indices cannot be treated as confirmation that muscle was protected. That is an unusually direct piece of self-criticism, and it is the reason we are not writing that this protocol preserved lean mass. The study did not measure the thing that claim would require.
The Gap the Paper Names Itself
The limitations section runs through twelve items. Four of them determine how far any of the above can be generalised.
No scheduled laboratory safety data. The paper states plainly that biochemical safety assessments were not performed according to a predefined schedule at each timepoint. Blood lipids, fasting glucose, insulin, HbA1c, liver enzymes and inflammatory markers are therefore absent from the results. A ketogenic phase is exactly the intervention where a reader would want those numbers, and a study that did not collect them cannot rule out changes in either direction. Absence of reported harm is not evidence of absence of harm when nothing was measured on a schedule.
No isocaloric control. Without a comparison group eating the same calories without ketosis, the observed changes cannot be attributed to carbohydrate restriction rather than to the energy deficit. The authors list this first among the design limitations.
Adverse events were collected from routine clinical notes. The paper reports no serious adverse events among the 25 analysed participants, and lists the symptoms that were assessed — headache, nausea or dizziness, fatigue, irritability, gastrointestinal symptoms, reduced exercise tolerance and halitosis. It then states that because of the retrospective design these were not collected with standardised questionnaires or prospective grading, so frequency and severity cannot be reliably quantified. Our own electrolyte guide covers the same symptom list from the mineral side, and the fact that a trial of this size cannot quantify them is worth knowing.
No correction for multiple comparisons. The paper tested many outcomes across multiple timepoints without adjusting for the number of tests, which inflates the chance of a false positive. The authors describe the secondary endpoints as exploratory on that basis.
None of this makes the study worthless. It makes it a hypothesis-generating report from a real clinical setting, which is what the authors call it. The pathway it describes — an initial ketogenic phase followed by deliberate carbohydrate reintroduction rather than an indefinite very-low-carbohydrate state — is a design worth testing properly. This paper is not that test.
What 20 Grams of Carbohydrate Looks Like
The protocol's most concrete instruction is the easiest to misread, because 20 g per day and 100 g per day are numbers without pictures. Our own food records give the pictures. Everything below is per 100 g and measured the same way across the library.
| Food | Net carbs / 100 g | FDC ID |
|---|---|---|
| Cooked white rice | 27.77 g | 168878 |
| Peanut butter, smooth | 17.31 g | 174266 |
| Dry-roasted peanuts | 12.86 g | 173806 |
| Greek yogurt, plain | 3.98 g | 171304 |
| Broccoli, raw | 4.04 g | 170379 |
| Spinach, raw | 1.43 g | 168462 |
| Butter | 0.06 g | 173410 |
A single 100 g portion of cooked white rice carries 27.77 g of net carbs, which on its own exceeds the phase-one ceiling. That is the arithmetic that makes a 20 g daily allowance restrictive in practice: it is not one food swap, it is most of the day's budget spent on a side dish. The reintroduction ceiling of 100 g is a different regime entirely and lands closer to what a standard low-carb pattern looks like — our net carb methodology page explains why we subtract fiber rather than counting total carbohydrate, and the difference matters most for foods in the middle of that table. Peanut butter, at 17.31 g per 100 g, sits just under the keto phase ceiling on paper while a two-tablespoon serving measures 5.54 g, which is a separate calculation we worked through in the peanut butter verdict.
One further note for context: the diet composition used here is not what most people mean when they say they are eating keto. It is a supervised, calorie-capped, protein-specified protocol delivered by a clinical nutrition unit, with individualised electrolyte and micronutrient supplementation where needed. The paper positions it as a less strict alternative to very-low-calorie ketogenic diets under 800 kcal per day, not as a home protocol.
Did the KETOCOMP study measure cholesterol or blood sugar?
No, not on a schedule. The paper states that biochemical safety assessments were not performed according to a predefined schedule at each timepoint, so blood lipids, glucose, insulin, HbA1c and liver enzymes are not reported. The study reports body composition outcomes only. Anyone reading the weight-loss figures should read that sentence in the same sitting.
How many people were in the study?
Twenty-five participants completed the four-week ketogenic phase and 21 had data after the four-week carbohydrate reintroduction phase, after four people were lost to follow-up. The mean age was 48.3 years, 16 were women and 9 were men, and mean baseline BMI was 34.4 kg/m². All participants had obesity with metabolic complications and had not responded to a calorie-restricted Mediterranean diet.
Was the weight loss caused by ketosis or by the calorie deficit?
The study cannot answer that. There was no isocaloric non-ketogenic control group, and every participant followed the same standardised target of about 1,200 kcal per day during the ketogenic phase. The authors list the absence of an isocaloric control as a limitation, which means the observed changes cannot be separated between the effect of carbohydrate restriction and the effect of a large energy deficit.
Did participants lose muscle?
The study reports a 1.40 kg reduction in fat-free mass during the ketogenic phase (p = 0.0051), but it did not directly measure skeletal muscle mass or function — there was no imaging, strength testing or physical performance assessment. The paper states that two bioelectrical impedance-derived indices cannot be treated as proof that muscle was protected, so the honest answer is that the question was not tested.
Related reading
Filed under newsresearchweight-loss