My results so far: two months of three-day fasts, measured

The headline numbers

Weight: the trend is much flatter than the endpoints

If I wanted to sell something, I would show you two photographs and the two numbers at the ends of the series: 180.0 lb on 21 June, 172.2 lb on 21 August. That reads as 7.8 lb lost in nine weeks.

The daily data does not support that story. Averaging the first seven days gives 177.5 lb; averaging the most recent seven gives 177.1 lb. Across 62 readings the weight rose as high as 182.2 lb (13 August) and fell as low as 172.2 lb, and it crossed back and forth through the high 170s the entire time. Both endpoints happen to sit at the extremes of a noisy band, which is exactly how misleading before-and-after numbers get made.

Most of that swing is not tissue. A full fasting day depletes liver and muscle glycogen, and every gram of stored glycogen carries roughly three grams of water with it, so several pounds can appear and disappear inside 48 hours without a single gram of fat changing hands. This is the effect the fasting weight loss calculator separates out, and it is why smart scale body-fat percentages during a fast should be ignored almost entirely.

My reading of my own data: nine weeks is long enough to prove that the protocol is sustainable and to characterise the noise, and far too short to claim a fat-loss rate. Ask me again at week 26 with a second DEXA scan in hand.

Blood pressure: a small, cautious improvement

Average systolic pressure fell from 122.4 mmHg across the first fortnight to 119.1 mmHg across the most recent one. Diastolic did not move meaningfully (79.4 to 80.4 mmHg). The single lowest reading was 107/81 on 28 July, during a run of unusually low systolic days in the last week of July.

I am deliberately underselling this. A 3 mmHg systolic difference is small enough to be produced by cuff position, time of day, caffeine on eating days, or how long I sat still first. It is consistent with the direction the literature reports for fasting and caloric restriction, but one unblinded person with a home monitor cannot distinguish a real effect of that size from measurement drift. What the record does show is that nothing alarming happened: no hypotensive episodes on fasting days, no hypertensive readings after refeeding.

Ketones: fast length matters far more than fasting frequency

This is where the data is most interesting, because the pattern is unambiguous even with a small number of readings. Breath acetone tracked fast duration closely:

• 72-hour fast (23 June): 45 ppm — the highest reading of the experiment.
• 48-hour fast (29 June): 25 ppm.
• 24-hour fasts: wildly inconsistent — 26 ppm on 7 July, but 0 ppm on 1 July, 5 July, 9 July and 28 July.

Repeated zero readings after a full 24 hours without food surprised me at first, then stopped surprising me. A single day frequently is not enough to exhaust glycogen and drive ketosis far enough for a breath meter to detect, particularly after a carbohydrate-heavy eating day. The switch described in metabolic flexibility is a gradient, not a light switch, and my own readings show the second and third days doing most of the work.

The two paired blood and breath measurements agree with each other. On 11 August, blood beta-hydroxybutyrate was 0.3 mmol/L against a 6 ppm breath reading. On 21 August, deeper into a fast, blood BHB was 2.9 mmol/L against breath readings of 30 and 34 ppm. Two pairs is not a calibration curve — the meters measure different molecules, as explained in ketones: blood vs breath — but they are at least telling the same story.

Body composition and fitness: baselines only

The 1 June 2026 DEXA scan recorded 31.1 percent body fat and 117.8 lb of lean mass, and the 14 June laboratory VO2 max test returned 49.1 ml/kg/min. Both are baselines, taken before and at the very start of the protocol, and neither tells you anything about the effect of fasting yet.

I am stating them anyway because the comparison is the whole point of scanning early. The outcome I care most about is not the fat number — it is whether lean mass holds. Losing muscle alongside fat is the real risk of repeated multi-day fasting, which is why resistance training continues through the year and why sarcopenia and hypertrophy have their own pages here.

Five years of blood panels — the pre-experiment context

The metabolic panels run from January 2022 to May 2026, so they describe the body that walked into this experiment rather than any result of it. Two threads stand out.

Triglycerides were volatile: 86 mg/dL in January 2022, a spike to 258 mg/dL in July 2023, a long stretch at 172 mg/dL, then 88 mg/dL in the most recent panel of May 2026. Total cholesterol drifted between 190 and 220 mg/dL and sat at 181 mg/dL in May 2026, with LDL at 113 mg/dL and HDL at 50 mg/dL.

Kidney markers moved in a way worth watching rather than ignoring. Estimated GFR fell from 100 in January 2023 to 65 in July 2025, with creatinine climbing from 0.85 to 1.40 mg/dL over the same period, before recovering to 81 and 1.05 mg/dL respectively by May 2026. Creatinine is influenced by muscle mass, protein intake and hydration, so a single trajectory is not a diagnosis — but it is precisely the marker that repeated fasting and heavy training can both perturb, and it is the one I will be following most closely for the rest of the year.

Fasting glucose stayed in a narrow 85 to 95 mg/dL band throughout and was 85 mg/dL in the latest panel.

What I have actually learned in nine weeks

• Three fasting days a week is sustainable for me, socially and physically. That was the open question in June, and it is largely settled.
• Daily weighing teaches you humility about scales. The noise band is roughly 10 lb wide; anything smaller than that is not a result.
• Fast duration drives ketosis, not fast frequency. Three separate 24-hour fasts did not produce anything like one 72-hour fast.
• Blood pressure has drifted very slightly downward, and I am not yet willing to call it an effect.
• The measurements that will matter most — a second DEXA scan and a mid-protocol blood panel — have not happened yet.

How to read this responsibly

This is a sample size of one, unblinded, uncontrolled, self-recorded, on a vegetarian diet, with resistance training in the background and no attempt to hold other variables still. It cannot tell you what fasting will do to you. It can show you what an honest set of measurements looks like, what magnitude of noise to expect, and which numbers are worth collecting if you ever do something similar under medical supervision.

Every figure here is reproducible from the dashboard, which updates as I log. When something contradicts what I expected, it stays on the page.