Tip · Fasting

Fasting-mimicking diet (FMD, 5 days)

Q: Does the FMD really work, or is it just calorie reduction?

In type 2 diabetes, the effect is clearly measurable in a twelve-month randomized trial: HbA1c improved by 0.3 percent even though glucose-lowering medication was reduced at the same time, and weight fell by 3.6 kilograms. Whether the same could be achieved with an equivalent permanent calorie reduction has never been compared directly – a study of FMD versus the same calorie reduction does not exist.

Q: Does the FMD really have to be low in protein?

That is the most interesting open question. In the only direct comparison so far, an FMD with 30 percent protein performed better on visceral fat, heart rate variability, gut microbiome and blood lipids than one with 10 percent protein – at the same calorie count and with autophagy likewise detectable at the molecular level. Only ketosis was more pronounced under the low-protein variant.

Q: Is autophagy really activated during the FMD?

Two studies have measured it. A pilot study with 30 healthy people found a significant group difference in autophagic flux in blood cells over eight days, though not at all measurement time points, and it was funded by the manufacturer. An independent seven-day study also reports that both FMD variants triggered autophagy at the molecular level. In each case the measurement is in blood cells, not in muscle, liver or brain.

Q: Can I put together the FMD myself?

The composition in the studies was low in calories, low in sugar and protein and rich in unsaturated fats, at 850 calories per day in the independent comparison study. Whether a homemade variant has the same effects has not been studied – all published studies used a standardized product or a diet assembled by the study team.

Q: How many people stick with it?

In the twelve-month diabetes study, 61 percent completed the program, 31 percent dropped out because of problems with the diet itself. Taste, quantity and frequency were cited. Then there is the price: in the first year, health care costs were significantly higher in the cost-effectiveness analysis.

Q: Who is the FMD not for?

For underweight people, people with a history of eating disorders, pregnant and breastfeeding women, children and adolescents. Anyone taking insulin or sulfonylureas will find no study data on this – the large diabetes study explicitly included only people taking metformin alone or no glucose-lowering medication at all.

Low in protein, high in fat, ~800 kcal. Randomized trials show effects on surrogate markers – endpoint data are missing.

Eating very few calories five days a month instead of nothing at all – the fasting-mimicking diet is the best-studied fasting protocol of all. It is here because it is the only form of fasting that can point to randomized trials with clinical endpoints, and because its best-known design principle – as little protein as possible – performed worse than the opposite in the only direct comparison study.

In short

The FMD works, and beyond surrogate markers. In a twelve-month randomized trial in 100 people with type 2 diabetes, the medication effect score fell by 0.3 (p < 0.001) and HbA1c by 3.2 mmol/mol or 0.3 % (p = 0.04) – both primary endpoints met, weight fell by 3.6 kg. In mild to moderate Crohn’s disease, 69.2 % of the FMD group achieved a clinical response compared with 43.8 % of the control group (p = 0.03). Hard endpoints such as mortality are still missing, some of the studies come from the developer’s circle, and the low-protein design has not been shown to be necessary. The most important number in practice: 61 % completed the one-year program, 31 % dropped out because of the diet itself.

What is behind it

The FMD is designed to push the body into fasting metabolism without leaving it completely without food: few calories, very little sugar, little protein, plenty of unsaturated fat. The low protein content is meant to lower IGF-1, the growth factor that maintains the anabolic state via the IGF-1, PKA and mTOR axis. In mouse experiments, lowering this axis is the precondition for hematopoietic stem cells switching into a regenerative mode.

In humans, the IGF-1 drop in this chain is established and has been measured in two independent studies: around 35 % after seven days of FMD and significantly after three cycles. The rise in ketone bodies is also established and was greater with the low-protein variant – difference in serum β-hydroxybutyrate 0.64 mmol/l (95 % CI 0.13 to 1.15; p = 0.015).

What the clinical trials found

The best study runs over twelve months in primary care. 100 people with type 2 diabetes, BMI of 27 kg/m² or more, all treated with metformin as their only glucose-lowering medication or with diet alone, received either five FMD days per month in addition to usual care or usual care alone. The actual finding: HbA1c improved even though medication was reduced at the same time. The composite category of glycemic management improved in 53 % of FMD participants compared with 8 % of controls and worsened in 23 % compared with 59 % (p < 0.001).

Two MRI findings come from the same data. Liver fat fell by 2.8 percentage points in proton density fat fraction (95 % CI −4.7 to −0.8; p < 0.01), and the marker for liver inflammation and fibrosis by 29.9 ms (95 % CI −51.8 to −8.0; p < 0.01). And abdominal fat decreased without loss of muscle: visceral abdominal fat −37.9 cm², subcutaneous −20.9 cm², abdominal muscle area −1.6 cm² (95 % CI −4.6 to 1.4) and thus no significant change.

The open protein question

The requirement that the FMD must be low in protein is the weakest point of the concept. An independent study compared two seven-day FMDs with the same 850 kcal per day but opposite distribution – 10 % protein and 45 % fat versus 30 % protein and 25 % fat – and both against an isoenergetic control diet. Both reduced weight and fat mass (interaction effects p < 0.0001), fasting glucose by about 10 % and IGF-1 by about 35 %, and according to the authors both triggered autophagy at the molecular level.

The high-protein variant was even better on several points: only it reduced visceral fat compared with control (−0.09 kg; 95 % CI −0.15 to −0.03; p = 0.006), and only it improved heart rate variability (p < 0.0001), microbiome diversity (p = 0.003), triglycerides (p = 0.009) and saturated fatty acids (p = 0.008). The low-protein variant was superior only for ketosis.

What is well supported

Two randomized trials with clinically relevant primary endpoints carry the field. In type 2 diabetes over twelve months: medication effect score −0.3 (95 % CI −0.4 to −0.2; p < 0.001), HbA1c −3.2 mmol/mol or −0.3 % (p = 0.04), body weight −3.6 kg (95 % CI −5.2 to −2.1; p < 0.001). In mild to moderate Crohn’s disease after three monthly cycles: clinical response in 45 patients (69.2 %) compared with 14 (43.8 %) (p = 0.03), clinical remission in 42 (64.6 %) compared with 12 (37.5 %) (p = 0.02), fecal calprotectin −22.0 % compared with +8.0 % (p = 0.03). In addition, the original study with 100 mostly healthy participants, 71 of whom completed three cycles: lower body weight, trunk and total body fat, lower blood pressure and lower IGF-1, no serious adverse events.

What the studies show

Twelve months of FMD in type 2 diabetes

Randomized controlled trial with blinded outcome assessment in 100 people with type 2 diabetes: 51 in the FMD group with 5 days per month in addition to usual care, 49 in the control group; 8 FMD and 10 control participants were lost to follow-up. Both co-primary endpoints were met. Limitations: open-label intervention, 18 % loss to follow-up, and the developer of the FMD is a co-author.

FMD in Crohn’s disease

Open-label randomized controlled trial, FMD on 5 consecutive days per month over three months compared with an unchanged diet. The primary endpoint was clinical response, defined as a decrease in the Crohn’s Disease Activity Index of at least 70 points or an index of no more than 150 after the third cycle. Exploratory analyses showed decreases in inflammatory lipid mediators. Limitation: open-label design with a control group that simply kept eating – the placebo component of a subjective index therefore cannot be separated out.

Low protein versus high protein, conducted independently

46 healthy men and women, randomized to three groups over 7 days: isoenergetic control (n = 16), FMD with 850 kcal and 10 % protein / 45 % fat (n = 15), FMD with 850 kcal and 30 % protein / 25 % fat (n = 15). Both FMD variants reduced weight, fat mass, fasting glucose and IGF-1 compared with control. The high-protein variant was superior for visceral fat, heart rate variability, microbiome and blood lipids. The developer of the FMD was not involved in this study.

Where the data stop

Endpoint data are missing: no FMD study has examined mortality, heart attacks or strokes. Some of the positive findings come from the developer’s circle – he is a co-author of the original study, the biomarker analysis and the diabetes study, and the autophagy pilot study was explicitly funded by the manufacturing company. The autophagy measurement itself is thin: in the pilot study with 30 healthy people over 8 days, there were significant group differences for body weight, fasting glucose, β-hydroxybutyrate, HOMA-IR and autophagic flux (each p < 0.05), but according to the authors not at all time points – measured in blood cells, not in muscle, liver or brain. The claim about biological age, a decrease in the median value of 2.5 years, also comes from an exploratory secondary analysis using a computational model. Not everything points in the favorable direction: in a randomized three-arm trial in 56 Surinamese South Asian people with type 2 diabetes, BMI and HbA1c improved only temporarily, while the dynamic perfusion border region worsened. Then there is the price: in the first year, quality-adjusted life years were not significantly lower (−0.04), but health care costs were significantly higher (+€2,241; 95 % CI +182 to +2,660).

How to do it

What the studies actually did: 5 consecutive days per month over 3 months, with a normal diet on the remaining days – that is how the original study ran and how the Crohn’s study ran. 5 days per month over 12 months in addition to usual primary care – that is the diabetes study. 6 monthly cycles in a crossover study with 102 participants with overweight or obesity, in which the proportion with impaired sense of smell fell from 38.1 % to 6.4 %. And 850 kcal per day over 7 days, once and plant-based, in the independent comparison study. On the common split of about 800 kcal on day 1 and about 500 kcal on days 2 to 5: this corresponds to the commercial implementation; the published studies describe the FMD as low in calories, low in sugar and protein and rich in unsaturated fats, without a daily split. There is no study on a frequency of once per quarter – what has been studied are monthly cycles over three, six and twelve months. And whether a self-assembled variant has the same effects has not been studied.

Safety

In the original study with 100 participants, no serious adverse events occurred, and in the periodontitis studies only minimal adverse events were reported – nausea, fatigue, weakness, dizziness – with no difference compared with control. The most important practical limitation is adherence: 61 % completed the one-year program, 31 % dropped out because of diet-related problems, 8 % for other reasons; taste, quantity and frequency were cited. The FMD is not suitable in underweight, with a history of eating disorders, during pregnancy and breastfeeding, or in childhood and adolescence. Anyone taking glucose-lowering medication should be under medical supervision – the large diabetes study explicitly included only people taking metformin as their only glucose-lowering medication or none at all; there are no study data on FMD under insulin or sulfonylureas. On refeeding risk: five days at 500 to 850 kcal are not complete food deprivation, and under the NICE criteria little or no food intake for more than 5 days counts as high risk only together with a second criterion – a person of normal weight after five FMD days does not meet this.

BK-Score Thin human evidence

Human evidence		5
Mechanism		5
Safety data		7
Hype gap		4
Track record of use		6

There are randomized trials with around a hundred participants showing changes in blood pressure, blood glucose and inflammatory markers – mostly surrogate markers over a few cycles; two studies now reach clinical endpoints in diabetes and Crohn’s disease. Hard endpoints such as mortality are missing. The original studies come from the developer’s circle; the newer papers no longer do.

The score rates the state of knowledge, not the effect. “Safety data 9” means well studied – not harmless.
Subjective assessment by Biohacking Kompakt based on published scoring rules – not a scientific rating and not a medical recommendation. Rules and all ratings (German)

Frequently asked questions about the fasting-mimicking diet (FMD, 5 days)

Does the FMD really work, or is it just calorie reduction?