What breaks a fast · 4 min read
What breaks a fast, and what the 50-calorie rule gets wrong
The 50-calorie rule has no source. It appears in no trial, no guideline and no physiology paper, only in blogs and app copy. What the evidence does show is that carbohydrate suppresses fasting metabolism at doses where fat and protein do not, so what you consume matters more than the calorie count.
Where the 50-calorie rule came from: nowhere
I went looking for the primary source. There is not one. The figure appears in no randomised trial, no clinical guideline and no physiology paper. It circulates entirely within blogs, forums and app copy, each citing the last.
It is also in direct tension with the largest studies available. The fasting-mimicking diet provides roughly 1,100 calories on day one and 720 on the following days, and in a randomised trial of 100 people it still produced reduced body weight, reduced trunk fat, lower blood pressure and lower IGF-1. The largest prolonged-fasting cohort on record, 1,422 people fasting 4 to 21 days in a residential clinic under daily medical supervision, documented the metabolic switch into ketosis on 200 to 250 calories a day. That last one was an observational study run by the clinic itself, and those fast lengths are a medical procedure rather than a template. Nothing on this site is about fasts of that length and they are not safe to attempt at home.
If a thousand calories of real food does not abolish fasting biology, thirty calories of cream almost certainly does not either. That is not proof, but the claim runs against the evidence rather than with it.
The honest gap: nobody has measured the small doses
Here is what makes this question harder than it looks. No human study has ever given someone 20 to 50 calories during a fast and measured ketones, insulin or anything else against a water-only control.
The dose-response work that exists starts at meal-sized amounts and works upward. Nobody has run the low end, so every confident claim about cream in coffee is extrapolated downward into a gap where no measurement exists.
So when someone tells you a splash of milk breaks your fast, and when someone else tells you it does not, neither is reporting a measurement. Both are guessing, with equal confidence and equal evidence.
What the evidence does support
Macronutrient appears to matter more than calorie count. In a study where people received 500 calories a day for a week at a time, carbohydrate abolished or greatly attenuated ketosis. The same 500 calories as fat or protein did not: ketones reached 4.08 mM, with glucose and insulin essentially indistinguishable from total starvation.
There is no fasting-state dose-response study for protein either. The closest data comes from a trial where 10 grams of leucine raised insulin and C-peptide while 5 grams did not, but that was given alongside a 400-calorie drink, so it describes how leucine changes a meal response rather than what a gram of protein does to a fasted person. Either way, both doses are far above anything you would add to a coffee.
Non-caloric sweeteners look inert, on evidence the analysts themselves grade as low to moderate confidence. A network meta-analysis of 36 trials and 472 participants found no effect on glucose, insulin, GLP-1, GIP, PYY, ghrelin or glucagon when taken alone. The cephalic-phase insulin response people worry about appears to need sweet taste plus actual glucose; sucralose alone produced neither an insulin nor a GLP-1 response.
| Item | Status | What is known |
|---|---|---|
| Black coffee | Fine on calories | ~2 kcal a cup, but caffeine transiently blunts insulin sensitivity |
| Non-caloric sweeteners | No measurable effect found | 36 trials, 472 people, low to moderate confidence |
| A splash of cream or milk | Unstudied | Nobody has measured 20 to 50 kcal against a control |
| Bone broth | Breaks it | Roughly 40 to 90 kcal a cup, varying a lot by recipe |
| 500 kcal of carbohydrate | Suppresses ketosis | Directly measured |
| 500 kcal of fat or protein | Does not | Same study in adults with obesity, ketones reached 4.08 mM |
Black coffee, the one real caveat
Energetically, black coffee is a non-issue at around two calories a cup. But caffeine is not metabolically neutral, and the direction is the opposite of how it is usually sold.
Caffeine acutely reduces insulin sensitivity, by roughly 15 percent in one clamp study, though that study infused caffeine intravenously rather than serving coffee. This is not "breaking a fast" in the sense people mean, but coffee is not a fasting enhancer either.
The autophagy-and-coffee claims you may have read come from mice, using a fluorescent marker in transgenic animals. There is no human autophagy data on coffee at all.
A more useful way to think about it
"Breaking a fast" is not a physiological event with a threshold. It is a question about your goal, and different goals have different answers.
If you are fasting to eat less overall, which is why it works for most people, then a splash of milk in your coffee is irrelevant and worrying about it is counterproductive. If you are fasting before a blood test, the rules are set by the test, not by us. If you are fasting for a specific metabolic effect, the honest answer is that the doses involved have not been studied.
The most useful rule is not a number. It is this: if what you are adding is small enough that you are asking whether it counts, it is almost certainly small enough not to matter for the reason you are fasting.
Sources
- Wei M et al. Fasting-mimicking diet and markers/risk factors for aging, diabetes, cancer, and cardiovascular disease. Sci Transl Med. 2017;9(377):eaai8700.
- Wilhelmi de Toledo F et al. Safety, health improvement and well-being during a 4 to 21-day fasting period in an observational study including 1422 subjects. PLoS One. 2019;14(1):e0209353.
- Vazquez JA et al. Effect of dietary fat, carbohydrate, and protein on branched-chain amino acid catabolism during caloric restriction. J Clin Invest. 1985;76(2):737-43.
- Ullrich SS et al. Intragastric administration of leucine or isoleucine lowers the blood glucose response to a mixed-nutrient drink by different mechanisms in healthy, lean volunteers. Am J Clin Nutr. 2016;104(5):1274-1284.
- Zhang R et al. Effects of non-nutritive sweeteners on glycaemic and appetite responses: a network meta-analysis. Nutrients. 2023;15(4):1050.
- Ford HE et al. Effects of oral ingestion of sucralose on gut hormone response and appetite in healthy normal-weight subjects. Eur J Clin Nutr. 2011;65(4):508-13.
- Keijzers GB et al. Caffeine can decrease insulin sensitivity in humans. Diabetes Care. 2002;25(2):364-9.
- Thong FS et al. Caffeine-induced impairment of insulin action but not insulin signaling in human skeletal muscle is reduced by exercise. Diabetes. 2002;51(3):583-90.
- Pietrocola F et al. Coffee induces autophagy in vivo. Cell Cycle. 2014;13(12):1987-94.animal or lab study
Every source here has been checked to exist and to say what this article says it says. Findings from animal or laboratory studies are labelled, because they often do not transfer to people.