Dry Fasting — The Record

The record · multi-day

Does pulling water drive ketosis faster?

The claim that dehydration deepens ketosis, tested against the one controlled human trial that measured it and against what a urine strip can actually detect.

PART 1: DISCONFIRMING EVIDENCE, REPORTED FIRST

D1. A randomized crossover human trial restricted fluid for 17 hours and BHB did not rise significantly. GRADE A.

This is the most direct experimental test of the hypothesis that exists, and it is negative.

Berneis K, Ninnis R, Häussinger D, Keller U. "Effects of hyper- and hypoosmolality on whole body protein and glucose kinetics in humans." Am J Physiol 1999;276(Endocrinol Metab 39):E188-E195. PMID 9886966.

Design, READ: randomized crossover, 10 healthy young men (25 +/- 1 yr, BMI 23.0 +/- 0.8), each completing hyperosmolar, hypoosmolar and isoosmolar studies in random order at least a week apart, with [1-13C]leucine and [6,6-2H2]glucose tracers.

How the dehydration was produced, READ, and this is why the study matters here: "hyperosmolality was induced by fluid restriction and intravenous infusion of hypertonic NaCl" over 17 h. Subjects were told not to drink after 8 PM, then received 2% NaCl at 1 ml/kg/h overnight and 5% NaCl at 200 ml/h from 8 AM. This is an experimental dry fast with the osmolar stress amplified by infused salt.

The hyperosmolar state was genuinely achieved, READ: plasma osmolality "increased during the hyperosmolality study from 283.4 to 296.4" mmol/kg H2O (P < 0.0001). Sodium rose from 142 to 149 mmol/l.

The ketone endpoint, READ, hyperosmolar versus isoosmolar:

AnalyteHyperosmolarIsoosmolarReported significance
beta-hydroxybutyrate86 +/- 26 umol/l66 +/- 13 umol/lno significant difference reported
acetoacetate54 +/- 7 umol/l42 +/- 6 umol/lno significant difference reported
NEFA392 +/- 31 umol/l414 +/- 17 umol/lno significant difference reported
glycerol45 umol/l43 umol/lno significant difference reported

Two things follow. First, BHB under 17 h of fluid restriction was 0.086 mmol/L. That is below the 0.5 mmol/L threshold for nutritional ketosis (see F1), so the subjects were not in ketosis at all. Second, the direction the study did find significant runs against the hypothesis, READ: "Glucose Ra during the experimental period was higher in the hyperosmolality study" (12.4 +/- 0.4 vs 11.2 +/- 0.4 umol/kg/min, P < 0.03), which the authors call "increased hepatic glucose production." More endogenous glucose is the anti-ketogenic direction, not the pro-ketogenic one.

Limitations, stated by the authors and not to be hidden: they note the rise in osmolality "may have been too modest to cause sufficient cell shrinkage," and that 17 h may have been too short. An insulin-glucose-amino acid clamp was applied, which constrains insulin and therefore isolates the tonicity effect cleanly but does not reproduce a free-living multi-day dry fast. So this trial refutes "tonicity alone accelerates ketogenesis over 17 hours." It does not refute a claim about day three of a dry fast, because nobody has run that.

D2. When tonicity did move lipolysis and ketogenesis in humans, it moved them in the opposite direction: hydration up, not down. GRADE B+.

Keller U, Szinnai G, Bilz S, Berneis K. "Effects of changes in hydration on protein, glucose and lipid metabolism in man: impact on health." Eur J Clin Nutr 2003;57(Suppl 2):S69-S74. DOI 10.1038/sj.ejcn.1601904.

Design, READ: 10 male subjects, three conditions. Hyperosmolality by fluid restriction plus IV hypertonic NaCl (2-5%) for 17 h; hypoosmolality by desmopressin, liberal drinking and hypotonic saline; isoosmolality with water ad libitum. Lipolysis measured by 2-[13C]glycerol turnover.

Exact sentence relied on, READ: glycerol turnover, "reflecting whole-body lipolysis, increased significantly during hypo-osmolar conditions."

And the conclusion, READ: the hypoosmolar response "results in protein sparing associated with increased lipolysis, ketogenesis and lipid oxidation."

Measured osmolality, READ: "Plasma osmolality increased from 285+/-1 to 296+/-1 mosm/kg" in the hyperosmolar arm and "decreased from 286+/-1 to 265+/-1 mosm/kg during hypo-osmolality" (P<0.001 both).

So the arm in which lipolysis and ketogenesis rose was the arm in which subjects were water-loaded and their cells were swollen, not dehydrated. Honest qualifier: the word "ketogenesis" in that conclusion is the authors' interpretation; the abstract reports no ketone numbers for the hypoosmolar arm, so the lipolysis finding is the measured part and the ketogenesis extension is their inference.

Companion primary paper, READ as a citation on the above record: Bilz S, Ninnis R, Keller U. "Effects of hypo-osmolality on whole-body lipolysis in man." Metabolism 1999;48:472-476. DOI 10.1016/S0026-0495(99)90106-3. PMID 10206440. Not retrieved in full; cited here as the primary record behind the hypoosmolar lipolysis result, not as a source of numbers.

Underlying framework, READ: Häussinger D, Lang F, Gerok W. "Regulation of cell function by the cellular hydration state." Am J Physiol 1994;267:E343-E355. Per the Keller abstract, hyperosmotic shrinkage "promotes protein breakdown, glycolysis and glycogenolysis." Note what is on that list and what is not: glycogenolysis is, ketogenesis is not.

D3. The most dehydrated state in clinical medicine has the least ketosis. GRADE A, with a stated confound.

ADA diagnostic criteria, READ, as reproduced in the AACE document "Diagnosis and Management of Hyperglycemic Crises":

ParameterDKAHHS
Effective serum osmolality, mOsm/kgVariable (lab table: <= 320)> 320
Serum ketones (nitroprusside)PositiveSmall
Urine ketones (nitroprusside)Positive (lab table: moderate to high)Small; "None or trace"
Estimated total water deficit, 70 kg patient~100 mL/kg, about 7 L100-200 mL/kg, about 10.5 L

Exact phrase relied on, READ, describing HHS: "Minimal ketonuria and ketonemia."

Hyperosmolar hyperglycemic state carries the larger measured water deficit, about 10.5 L in a 70 kg patient, and an effective osmolality above 320 mOsm/kg, and it is defined by having almost no ketones. Diabetic ketoacidosis has the smaller water deficit and the lower osmolality and is defined by having a lot of them.

The confound, stated plainly because it is the honest reading: the accepted explanation for the difference is residual insulin in HHS suppressing ketogenesis, not the water deficit itself. So this does not prove dehydration suppresses ketosis. What it does prove is narrower and still fatal to the hypothesis as stated: severe dehydration and hyperosmolality are not sufficient to produce ketosis, because the clinical state that maximizes both produces nearly none.

D4. The 3 g water per gram of glycogen figure, which is load-bearing for the mechanism, is contested in the primary literature. GRADE B+ for the contest.

Source, READ: "Muscle Glycogen Assessment and Relationship with Body Hydration Status: A Narrative Review." Nutrients 2023;15(1):155.

Supporting the ratio, READ: "1 g of liver glycogen binds 3 g of water" (attributed to early work cited to Zuntz 1906 and Pavy 1860); "1 g of muscle glycogen is bound to 3-4 g of water" (Olsson & Saltin 1970). Animal liver values tabulated: 1:2.4 (Puckett & Wiley, rats), 1:3.8 (MacKay & Bergman, rats), 1:1.63 (Fenn & Haege, cats), 1:2.7 (McBride, rats).

Contradicting it, READ, with the review's own DOIs:

  • MacKay & Bergman 1932, DOI 10.1016/S0021-9258(18)76276-5: relationship in liver, but not in muscle.
  • Bridge & Bridges 1931, DOI 10.1016/S0021-9258(18)76501-0, and Kaplan & Chaikoff 1936, DOI 10.1016/S0021-9258(18)74639-5: no relationship in liver.
  • Sherman 1982, DOI 10.1055/s-2008-1026056; Richter 1988, DOI 10.1152/ajpendo.1988.255.5.E621: no relationship in muscle.
  • Shiose 2018, DOI 10.1152/japplphysiol.00666.2017: no change in segmental water with glycogen depletion in humans.
  • Fernández-Elías 2015: "The ratio of stored glycogen and water in muscle changed from 1:3 to 1:17," i.e. not a fixed ratio at all.

And the decisive methodological sentence, READ: "glycogen-binding water was not directly measured" in the human studies.

So "3 g of water per gram of glycogen" is a reasonable central estimate for liver in old animal work and is not an established constant in human muscle. It can be stated as an estimate with a range of roughly 1.6 to 4, never as a fixed fact, and the human evidence that glycogen depletion moves tissue water is mixed.


PART 2: THE NUMBERED QUESTIONS

(a) Normal time course of ketosis in a water-only fast in humans

VERDICT: established, with wide between-study variance. BHB reaches roughly 0.5 mmol/L by 24 h, 1 to 2 mmol/L by 36 h, and 2 to 3 mmol/L by 48 to 72 h. GRADE A.

Primary source, READ: Ulupınar et al. "Duration-Dependent Ketone Body Responses to Fasting: A Systematic Review, Meta-Analysis, and Meta-Regression." IJMS 2026. DOI 10.3390/ijms27219570. 30 studies, 44 effect sizes; the BHB analysis covers 24 studies and 26 duration-specific estimates. Per-study n from 6 to 57.

Values as reported in that review, READ (baseline to post-fast, mmol/L unless noted):

HoursStudy as named in the review (n)BHB baselineBHB post
12Bouck 2025 (15)~0.08not reported
24Gonzalez & Cooke 2022 (25)0.1280.536
30Deru 2025 (30)0.1451.067
36Andriessen 2023 (12)0.0340.619
36Jørgensen 2015 (18)0.082.21
36Rubio-Aliaga 2011 (10)0.071.17
48Gälman 2008 (7)~0.08~3.24
58Qvigstad 2002 (6)0.011.84
60Andriessen 2023 (12)0.0342.236
72Bak 2018 (9 lean / 9 obese)0.33.0 lean; 1.9 obese
72Nuttall 2020 (7)0.0912.233
72Pedersen 2017 (9)15.8 umol/L~2.33
<= 72Lundholm 2025 (57)median 0.572.73
7 dKolnes 2025 (13)0.074.01
8 dOgłodek & Pilis 2021 (12)0.304.77
~10 dMa 2021 (13)0.185.87 experienced

Exact sentences relied on, READ:

  • "beta-Hydroxybutyrate increased from 0.091 +/- 0.210 mmol/L at baseline to 2.233 +/- 0.529 mmol/L after 72 h of fasting." (Nuttall 2020)
  • "Blood ketone concentrations increased from 0.08 +/- 0.02 to 2.21 +/- 0.23 mmol/L (mean +/- SE)." (Jørgensen 2015)
  • "Plasma beta-hydroxybutyrate concentrations increased from 0.145 +/- 0.071 to 1.067 +/- 0.567 mM (mean +/- SD)." (Deru 2025)

Meta-regression, READ: BHB rises "beta = 0.300 per day (cluster-robust SE = 0.082; t = 3.66; 95% CI: 0.130 to 0.470; p = 0.001)". Fasting duration explained about 31% of between-study variance. Pooled BHB effect Hedges' g = 2.794 (95% CI 2.281 to 3.307). Acetoacetate did not track duration significantly (beta = 0.156/day, p = 0.365).

Note on the 48 h versus 72 h values: the 48 h figure (~3.24) is higher than several 72 h figures (1.9 to 3.0). That is real between-study scatter, not a decline. Only 31% of variance is explained by duration, so any single-timepoint number should be quoted as a range, never as a landmark.

Hydration in these studies, READ, and this matters for the whole strand: "water, mineralized water, and non-caloric beverages such as unsweetened black coffee or tea were permitted." The review lists hydration among the inconsistently reported factors: "including differences in supervision, hydration practices, and experimental conditions." So hydration was permitted but not controlled or measured. There is no hydration-stratified arm anywhere in this corpus.

Supporting serial case, READ, GRADE C: Sekizuka E, Sekizuka D, Zheng T. Glycative Stress Research 2018;5(4):181-191. DOI 10.24659/gsr.5.4_181. Single subject, 7-day water-only fast with oral intake of more than 2 L of water per day. Total ketones 66 to 1,066 to 2,418 umol/L; BHB 59 to 923 to 2,030 umol/L; acetoacetate 7 to 143 to 388 umol/L. Reported as "ケトン体は 37 倍、β ヒドロキシ酪酸は 34 倍、アセト酢酸は 55 倍に上昇した" (total ketones rose 37-fold, BHB 34-fold, acetoacetate 55-fold). This is a drinking fast reaching BHB above 2 mmol/L, and it supplies the BHB-to-acetoacetate ratio used in (e).

(b) Time course under fluid restriction or dawn-to-dusk fasting

VERDICT: no usable time course exists. The one controlled experiment that actually restricted fluid and measured BHB found no ketosis. The Ramadan literature overwhelmingly does not measure ketones at all. GRADE A for the single controlled result, and the rest is an evidence gap.

1. The only controlled fluid-restriction measurement, READ, GRADE A: Berneis 1999 (D1). 17 h fluid restriction plus hypertonic saline, osmolality to 296.4 mmol/kg, BHB 0.086 mmol/L, not significantly different from the drinking arm.

2. Ramadan studies largely do not measure ketones, READ: Fenni Y, Ikiouane M, Redjal M, Remini H, Allam A, Zaabar S. Apunts Sports Medicine 2025;60:100467. DOI 10.1016/j.apunsm.2024.100467. Across the eight included Ramadan studies, none is described as measuring ketones. That is the finding, not a search failure.

3. Where Ramadan metabolomics was run, BHB was not among the reported changes, READ, GRADE B+: Frontiers in Nutrition 2022, DOI 10.3389/fnut.2022.1008730. n = 25 with overweight and obesity, baseline about one week before Ramadan and a second sample at the end of the month. The paper reports no measured change in 3-hydroxybutyrate; the discussion mentions 3-hydroxybutyrate only when citing other prolonged-fasting studies. No hydration or osmolality markers were measured.

4. One Ramadan study reports a significant BHB rise but supplies no retrievable numbers and did not measure during the fast, READ, GRADE B: Diabetology & Metabolic Syndrome 2024, DOI 10.1186/s13098-023-01234-3. n = 30 with type 2 diabetes. "Patients were subjected to time restricted fasting for 16 h (from dawn to sunset) during Ramadan." Result sentence: "The present study revealed a statistically significant increase in betaHB levels post fasting (p < 0.001)." Three disqualifiers for our purpose: samples were taken 1 week before and 3 weeks after Ramadan, not during; no numeric BHB values or units were retrievable; and the text does not state fluid restriction, so it cannot be scored as a dry-fast measurement.

So the honest position for (b): there is no BHB-versus-hours curve for a fluid-restricted fast. A claim about how fast ketosis develops without water has no time course behind it in any language searched.

(c) Does any study directly compare the two?

VERDICT: NO. No study compares a dry fast against a water fast with a ketone endpoint. The absence is the finding.

Stated plainly, as the brief requires. Across nine languages, the searches below returned no trial, cohort, case series or case report that measured ketones in the same or matched subjects under both water-restricted and water-permitted fasting.

The closest existing design, READ, and the reason it is not the missing study: Berneis 1999 does compare fluid restriction against free drinking within the same subjects, randomized, with BHB measured in both arms. But it runs 17 h, not days; it adds hypertonic saline, so the osmolar stress is not that of a plain dry fast; and it applies an insulin-glucose-amino acid clamp, so insulin is not free to fall as it would in a real fast. It is the best available proxy and it is negative. It is not the direct comparison.

What the missing study would need: same subjects, randomized order, matched fast duration, water ad libitum in one arm and withheld in the other, with blood BHB as the primary endpoint and plasma osmolality, body mass, urine specific gravity and urine flow rate recorded. Nothing of that shape exists.

Consequence for copy and for the hypothesis: "enters a deep level of ketosis very quickly" when water is withheld is, at this date, an untested claim rather than a contested one. The only adjacent controlled data point runs against it.

(d) Mechanisms that could plausibly accelerate ketosis

VERDICT: of the five candidate mechanisms, the two that are measured in humans do not accelerate ketogenesis, and one of them actively opposes it. Two are real but not specific to water withholding. The glycogen-water figure is contested. GRADE mixed, stated per mechanism.

d1. Glycogen depletion and glycogen-bound water. GRADE B+, and the ratio is contested. See D4. The ratio is roughly 1.6 to 4 g water per g glycogen in old animal liver work and 3 to 4 in Olsson & Saltin human muscle, with multiple primary studies finding no relationship in muscle and one human study finding no change in segmental water on depletion. Critically, this mechanism explains water loss and scale weight, not ketone production. Losing the water that accompanies glycogen does not itself make a ketone. Glycogen depletion does drive ketogenesis, but glycogen depletion is a consequence of not eating, and it proceeds the same way whether or not the person drinks. No source found shows that withholding water depletes glycogen faster.

d2. Insulin fall. GRADE B+, and not specific to water. The insulin fall that permits ketogenesis is driven by carbohydrate and energy absence. In Berneis 1999, READ, insulin under hyperosmolality was 37.8 +/- 3.6 versus 40.8 +/- 4.8 pmol/l isoosmolar, with no significant difference reported; only the hypoosmolar arm showed lower insulin. So in the one human experiment that restricted fluid, dehydration did not lower insulin.

d3. Counter-regulatory hormones and hepatic glucose output. GRADE A, and the direction is anti-ketogenic. Berneis 1999, READ: glucose Ra 12.4 +/- 0.4 versus 11.2 +/- 0.4 umol/kg/min, P < 0.03, and plasma glucose 5.1 +/- 0.05 versus 4.9 +/- 0.1 mmol/l, P < 0.03. The authors call this "increased hepatic glucose production." Raising endogenous glucose availability opposes the switch to ketones. Consistent with the Häussinger framework, READ, in which hyperosmotic shrinkage "promotes protein breakdown, glycolysis and glycogenolysis," a list that does not include ketogenesis.

d4. Cell volume and tonicity acting on lipolysis and ketogenesis. GRADE B+, and the measured direction is the reverse of the hypothesis. See D2. Lipolysis rose under hypoosmolality (cell swelling, water loading). Under hyperosmolality, Berneis 1999 measured NEFA, glycerol, acetoacetate and BHB and reported no significant difference from isoosmolar in any of them. So the cell-volume mechanism, as actually measured in humans, does not deliver accelerated ketogenesis under dehydration; where it moves lipolysis at all, it moves it the other way.

d5. Vasopressin as a candidate pathway. NOT GRADED, lead only. Two records were located and not retrieved, so no numbers are claimed and no grade is assigned: "Osmotic stimulation of vasopressin acutely impairs glucose regulation: a counterbalanced, crossover trial" (Am J Clin Nutr, fetch refused by the publisher), and "Vasopressin and hydration play a major role in the development of glucose intolerance and hepatic steatosis in obese rats," DOI 10.1007/s00125-015-3496-9 (not opened). Flagged as the one open mechanistic lane worth a dedicated pass. Per Rule 25 this is a named source with nothing claimed from it, which is lawful but not finished.

One genuine arithmetic point in the hypothesis's favor, flagged as arithmetic on read values, not a reported result: there is a real route by which a dry fast could produce a higher measured number without producing more ketones, which is exactly what (e) is about. That is a measurement effect, not a metabolic acceleration.

(e) THE CRITICAL CONFOUND

This splits into the urine question, which is where the founder's artifact sits, and the blood question, which was the original brief.


(e1) URINE ACETOACETATE STRIPS UNDER DEHYDRATION

Context: the artifact is a TRUEplus urine ketone strip dated 2017-11-16 reading at the top of the scale, 160 mg/dL, "LARGE," taken during a dry fast and offered as evidence of unusually deep ketosis.

VERDICT: the strip cannot support a claim of deeper ketosis. It is uninterpretable as a depth measure, for four independent reasons, two of which push the reading up under dehydration and two of which push it down in deep ketosis. GRADE B+ overall, assembled from multiple independent witness-level records.

e1.1 What the strip measures, and what it is blind to. GRADE B+, three independent sources.

Source 1, READ: Huang J, Yeung AM, Bergenstal RM, Castorino K, Cengiz E, Dhatariya K, Niu I, Sherr JL, Umpierrez GE, Klonoff DC. "Update on Measuring Ketones." J Diabetes Sci Technol 2024;18(3):714-726. DOI 10.1177/19322968231152236.

  • "Assessment of ketonemia is performed by the nitroprusside reaction, which provides a semiquantitative estimation of acetoacetate and acetone levels."
  • The assay "does not recognize the presence of beta-hydroxybutyrate, the main ketone body in ketoacidosis."
  • Comparison table, urine disadvantage: "Measures only acetoacetic acid, not beta-hydroxybutyrate."

Source 2, READ: Chinese Society of Endocrinology. 中国糖尿病血酮监测专家共识 [Expert consensus on the monitoring of diabetic blood ketone bodies in China]. Chinese Journal of Endocrinology and Metabolism 2014;30(3):177-183. DOI 10.3760/cma.j.issn.1000-6699.2014.03.001.

  • "目前使用的尿酮测定方法只能测定乙酰乙酸(灵敏度约为50 mg/L)及丙酮(灵敏度约为500 mg/L),不能测定β-羟丁酸。" Current urine ketone methods can measure only acetoacetate (sensitivity about 50 mg/L) and acetone (about 500 mg/L), and cannot measure beta-hydroxybutyrate.
  • "该法的主要不足在于无法检测酮症时浓度最高、且与病情平行的β-羟丁酸水平,且结果只能定性或半定量显示。" The main shortcoming is that it cannot detect beta-hydroxybutyrate, the ketone that is highest in ketosis and tracks severity, and results are only qualitative or semi-quantitative.

Source 3, READ: the ADA criteria document reproduced by AACE states that the nitroprusside method "measures acetone and acetoacetate, not beta-hydroxybutyrate."

Consequence: a 160 mg/dL reading is an acetoacetate concentration in urine. It is not a measure of BHB, and BHB is the dominant ketone in deep ketosis. The strip is structurally blind to the thing the claim is about.

e1.2 Dehydration raises the reading. Elevated specific gravity is a documented false-positive cause for the ketone pad, in two independent published reviews 17 years apart. GRADE B+.

Source 1, READ: Simerville JA, Maxted WC, Pahira JJ. "Urinalysis: A Comprehensive Review." Am Fam Physician 2005;71(6):1153-1162.

  • Table 2, ketone false positives: "Acidic urine, elevated specific gravity, mesna (Mesnex), phenolphthalein, some drug metabolites (e.g., levodopa)"
  • Same article, on specific gravity: "a value greater than 1.020 indicates relative dehydration"

Source 2, READ: Hitzeman N, Greer D, Carpio E. "Office-based urinalysis: a comprehensive review." Am Fam Physician 2022;106(1):27-35B.

  • Table 2, ketone false positives: "High urine specific gravity, low pH, some drug metabolites."
  • Same article: "A value exceeding 1.020 may indicate dehydration, glucosuria, or syndrome of inappropriate antidiuretic hormone." And: "Dipstick urine specific gravity indicates only current hydration status."

The chain closes inside these two sources without any inference of mine: dehydration is what high specific gravity indicates, and high specific gravity is a listed false-positive cause for the ketone pad. Both reviews also list acidic or low pH as the other false-positive cause, and concentrated urine during a fast is characteristically acidic, so a dry fast loads both listed false-positive conditions at once.

Honest gap, stated because the brief demands it: no study was found that held ketone production constant, varied hydration, and measured the strip's response. That titration experiment appears not to exist. The false-positive attribution above is a published reference-standard statement, not a dose-response measurement. This is the single most useful experiment nobody has run.

e1.3 The size of the dilution effect, and the published correction. GRADE B+ for the method.

Source, READ: MacPherson S, Arbuckle TE, Fisher M. "Adjusting urinary chemical biomarkers for hydration status during pregnancy." J Expo Sci Environ Epidemiol 2018;28:481-493. DOI 10.1038/s41370-018-0043-z.

The published specific-gravity correction (their Eq. 7), READ: adjusted concentration = observed concentration x (SG_m - 1) / (SG_i - 1), where SG_i is the sample's specific gravity and SG_m the reference median.

Their creatinine correction (Eq. 6), READ: concentration divided by creatinine, expressed per gram creatinine.

Their excretion-rate approach (Eq. 5), READ: concentration x urinary flow rate, where flow rate is void volume divided by time since last void.

Reference range quoted in that paper, READ: "normal urine specific gravity usually ranges from 1.013 to 1.029."

Arithmetic on those read values, flagged as arithmetic and not a reported ketone result: applying Eq. 7 across that very normal range, a void at SG 1.029 against a reference of 1.013 carries a factor of (0.013)/(0.029) = 0.45, meaning the raw concentration is inflated about 2.2-fold by concentration alone, at identical excretion. Dehydration beyond that range widens the factor further. Two hops: (1) the published formula, (2) substitution of the paper's own stated normal range.

Their own caution, READ: specific gravity reproducibility was poor, the largest intraclass correlation being "a weekend day, having an ICC of 0.38."

So the dilution artifact on a urine concentration is not a few percent. It is plausibly a factor of about two within the normal hydration range, which is the same order as the gap between adjacent blocks on a ketone strip's color scale.

e1.4 The reading is also biased DOWNWARD in deep ketosis, which is the well-documented discordance. GRADE B+.

Source 1, READ, and this came from the Turkish-language channel: Kuru B, Sever M, Aksay E, Dogan T, Yalcin N, Seker Eren E, Ustuner F. "Comparing Finger-stick beta-Hydroxybutyrate with Dipstick Urine Tests in the Detection of Ketone Bodies." Turk J Emerg Med 2014;14(2):47-52. DOI 10.5505/1304.7361.2014.14880. Cross-sectional prospective, 265 patients analyzed.

  • "Twenty-nine (53.7%) of the 54 patients whose capillary blood ketone levels were positive had no ketonuria." So over half of the blood-ketone-positive patients had a negative urine strip.
  • Of the four patients meeting ADA DKA criteria, with blood BHB of 3.5, 5.2, 6.3 and 6.7 mmol/L, "the urine dipsticks only identified significant ketonuria ('3+') in one of these patients." Three of four patients in frank ketoacidosis were not flagged as significantly ketonuric by the strip.
  • Also, READ, cited by them to the literature rather than their own data: "the specificity of urine dipsticks is low (<50%)."

Source 2, READ, Chinese consensus: "在血糖>13.9 mmol/L时,尿酮诊断DKA阳性率低于血酮(58%对93%, P<0.001),且尿酮诊断DKA较血酮滞后约1 h" Urine ketone is positive for DKA less often than blood ketone, 58% versus 93%, P<0.001, and lags by about an hour.

Source 3, READ, Huang 2024, the mechanism of the paradox: as DKA resolves, "beta-hydroxybutyrate is oxidized to acetoacetate, which is the predominant ketone in the urine," and "the elevated urine result is a lagging indicator of the severity of ketoacidosis." The strip can therefore read higher while the person is getting better.

Why depth itself biases the strip down, READ: Huang 2024 reports that in DKA "the ratio of beta-hydroxybutyrate to acetoacetate increase from 1:1 to as much as 10:1." Measured in an actual human 7-day water fast, READ, Sekizuka 2018: BHB 2,030 umol/L against acetoacetate 388 umol/L at the deepest point, a ratio of about 5.2 to 1 (arithmetic on their two reported values). So the deeper the ketosis, the larger the share of total ketones that the nitroprusside strip cannot see.

e1.5 Renal handling under dehydration: does reduced clearance concentrate urine ketones independently of production? Partly, and the two effects oppose each other. GRADE B+, with the net effect unmeasured.

Source 1, READ: Palmer BF, Clegg DJ. "Starvation Ketosis and the Kidney." Am J Nephrol 2021;52(6):467-478. DOI 10.1159/000517305. PMID 34350876.

  • "At low plasma concentrations, filtered ketone bodies are completely reabsorbed by the saturable Na+-coupled monocarboxylate transporters SMCT1 (SLC5A8) and SMCT2 (SLC5A12) in the proximal tubule."
  • "Ketonuria develops as plasma levels rise and the filtered load of ketoacid salt increases."
  • Urinary losses "peak after 3-4 days then slightly fall as fasting extends into the protein conservation phase."
  • "In prolonged starvation kidney reabsorption of ketone bodies spares approximately 225 kcal/day, which would otherwise be lost in the urine."
  • Stated limitation of that source, READ: it "does not directly examine how dehydration or volume depletion changes ketone clearance or plasma ketone concentration."

Source 2, READ, Chinese consensus: "需要注意的是糖尿病酮症患者肾功能严重损伤而肾阈值增高时,尿酮体可减少,甚至完全消失。" When renal function is severely impaired and the renal threshold rises, urine ketone bodies may decrease or even disappear completely.

So the two opposing arrows, both grounded in the above:

  • Dehydration reduces renal perfusion and filtered load, which pushes ketone excretion down, and can in the limit abolish ketonuria entirely.
  • Dehydration reduces urine volume, which pushes ketone concentration, the thing the strip reports, up.

Net effect on a strip reading during a dry fast: not measured by anyone found. Do not state a direction for the net. What can be stated is that the strip reports the quantity that the water loss inflates, while the quantity that reflects production, excretion per unit time, is the one not being reported.

e1.6 One further property of this specific reading: it is at the ceiling.

READ, from the scale as reported: 160 mg/dL is the top block of the TRUEplus urine ketone scale. A top-of-scale reading is right-censored: it cannot distinguish 160 from any higher value, and it cannot be ranked against another top-of-scale reading. Arithmetic point, flagged as such: a censored maximum cannot support a comparative claim of the form "deeper than," in either direction, regardless of hydration. Even a perfectly calibrated strip at the ceiling carries no depth information above the ceiling.

Also, READ, both AAFP reviews list a false negative cause: "Delay in examination of urine." Acetoacetate is unstable, so any error from sample delay or strip age runs downward, not upward.

(e1) PLAIN VERDICT, for a reader to act on

A urine ketone strip reading taken during a dry fast cannot support a claim of deeper ketosis. It is uninterpretable on its own, for reasons that are documented rather than speculative:

1. It measures acetoacetate and acetone, never BHB, and BHB is the dominant ketone in deep ketosis (three independent sources). 2. High urine specific gravity, which is the signature of dehydration, is a listed false-positive cause for that exact pad (two independent published reviews). 3. Low or acidic urine pH, also characteristic of a fasting concentrated void, is the other listed false-positive cause for that exact pad (same two reviews). 4. The published dilution correction implies a factor of roughly two within the normal specific-gravity range alone, before any dehydration. 5. The strip misses over half of blood-ketone-positive people and missed three of four frank DKA cases in a 265-patient series, so it is unreliable in the deep direction too. 6. This particular reading is at the ceiling of the scale and is therefore censored, so it carries no "deeper than" information even in principle.

The honest statement is that the 2017 strip documents that acetoacetate was present in concentrated urine on that day. It does not document how deep the ketosis was, and it cannot be compared against a strip taken while drinking.

What would have to be measured alongside it, which is the protocol to run next

On the same void, same clock time each day: 1. Blood BHB by an enzymatic meter (beta-hydroxybutyrate dehydrogenase method, as the Chinese consensus specifies for blood strips). This is the depth measure. Everything else is support. 2. Urine specific gravity by refractometer, not by dipstick. The 2022 AAFP review states dipstick specific gravity "indicates only current hydration status," and the dipstick pad has its own interferences. This is the hydration marker that makes the strip's number interpretable at all. 3. Urine creatinine on the same void, to express the result per gram creatinine (MacPherson Eq. 6). 4. Timed void volume, so concentration can be converted to an excretion rate (MacPherson Eq. 5: concentration x urine flow rate). This is the only step that separates "more ketones were made" from "there was less water." 5. Morning body mass, same scale and conditions, to quantify total water loss. 6. Plasma sodium or measured osmolality, to document the hyperosmolar state actually reached, so the condition can be compared to Berneis 1999's 296 mOsm/kg. 7. A paired control arm: the same person, same fast duration, drinking water freely, measured identically. Without this arm there is no basis for "faster" or "deeper," because there is nothing to be faster or deeper than.

Scope caution on the result of that protocol: one person across two conditions is a GRADE C result. It can settle the question for him personally and it must not be published as a general claim. Under the lane law a C does not open a lane.


(e2) IS AN OBSERVED BLOOD KETONE RISE PARTLY HEMOCONCENTRATION AND REDUCED CLEARANCE?

VERDICT: for blood BHB the artifact is real but small, on the order of a few percent, and it cannot explain a multi-fold rise. The blood rise in a fast is overwhelmingly a real increase in production. The artifact problem belongs to urine, not blood. GRADE B+.

Measured plasma volume contraction under dawn-to-dusk fluid-restricted fasting, READ: Khalfoun et al. "Ramadan Intermittent Fasting and Plasma Volume Variations in Individuals with Different Body Weights." Medicina 2024;60(7):1143. DOI 10.3390/medicina60071143. PMID 39064573 appears in the BVS/Medline record (mdl-39064573) and not on the article page, so it is reported here as a bibliographic-record identifier rather than a page-confirmed one.

  • Plasma volume variation reached about -1.9% to -4.2% in the obese and severely obese groups and about 0% in normal-weight participants.
  • Hematocrit in the obese group rose from 49.3 +/- 2.4 to 50.4 +/- 2.4% at T2, about +2.2%, READ: "Post hoc tests indicated increases in hematocrit in OB at T2 (p = 0.03; ES = 0.36)".
  • Caveat the paper itself creates, READ: the text describes plasma volume variation as increasing in obese participants while Table 4 reports negative values and the discussion refers to "decreases in plasma volume variation." The sign is internally inconsistent in that paper, so only the magnitude, a few percent, should be relied on.

A second, larger estimate under harsher conditions, READ: Trabelsi K, El Abed K, Trepanowski JF, et al. "Effects of Ramadan fasting on biochemical and anthropometric parameters in physically active men." Asian J Sports Med 2011;2(3):134-144. DOI 10.5812/asjsm.34775. As reported in the Apunts review: "Fasters' hematocrit increased by 5.3% (P<0.001) and hemoglobin increased by 6.3% (P=0.01)", serum sodium +1.9%, chloride +2.6%, body weight -1.9%, with aerobic exercise in heat.

Arithmetic on those read values, flagged as arithmetic: hemoconcentration in real fluid-restricted fasting runs about 2% to 6%. A BHB rise from roughly 0.1 to 2.0 mmol/L is about 20-fold. A 2% to 6% contraction of plasma volume cannot account for more than a few percent of a 20-fold change. Two hops: (1) the measured contraction range above, (2) the measured BHB time course in (a).

Two further points of honesty:

  • No published hemoconcentration correction for BHB was found. None was located in any language searched. So no correction can be quoted.
  • A plasma-volume correction would in any case be the wrong instrument. BHB distributes in total body water rather than being confined to plasma, so scaling by plasma volume is not the right adjustment, and the right one has not been published. This is a genuine open methodological gap, and it is small in magnitude.
  • Reduced renal clearance does raise plasma ketone levels for a given production rate, and the renal mechanism is documented (Palmer & Clegg, e1.5), but no source found quantifies that contribution during dehydration in humans.

The asymmetry that resolves the whole strand: plasma concentration is physiologically defended within a few percent, which is why hemoconcentration is a small artifact. Urine concentration is the variable that absorbs the water deficit, and it ranges widely, which is why a urine strip is the fragile measurement. The founder's artifact is a urine strip, so it lands squarely in the fragile category, while a blood BHB reading would not.

(f) Ketoacidosis risk, and what raises it under dehydration

VERDICT: the operational line is blood BHB above 3.0 mmol/L with acidosis. Nutritional ketosis starts at 0.5 mmol/L and a prolonged water fast typically peaks between 2 and 6 mmol/L, so a long fast approaches the ketoacidosis threshold on concentration alone, and dehydration removes the renal route for clearing the acid load. GRADE A for the thresholds.

Thresholds, READ, Huang 2024 (J Diabetes Sci Technol 18(3):714-726, DOI 10.1177/19322968231152236):

  • "A normal serum concentration of ketone bodies is defined as <= 0.5 mM."
  • "nutritional ketosis, which is a beta-hydroxybutyrate concentration of at least 0.5 mmol/L."
  • "hyperketonemia is defined as being in excess of 1.0 mM."
  • "ketoacidosis as ketone concentrations in excess of 3.0 mM."
  • Starvation: "starvation ketosis is defined as occurring when the blood/plasma concentration of acetoacetate is about 1.0 mmol/L," with BHB above 2 to 5 mM.
  • "plasma ketone concentrations, in particular beta-hydroxybutyrate, can rise to as high as 6 mmol/L" in prolonged starvation.
  • DKA resolution: "beta-hydroxybutyrate to 1.5 mmol/L or less has been proposed as defining the resolution."

Diagnostic criteria, READ, ADA via the AACE document: DKA requires arterial pH <= 7.30, bicarbonate <= 18 mmol/L, anion gap > 10 mmol/L and BHB >= 40 mg/dL in adults. HHS requires effective osmolality > 320 mOsm/kg.

Chinese consensus, READ: blood ketone >= 3 mmol/L or a positive urine ketone is one of the three major DKA criteria; bicarbonate therapy is considered only when pH is below 6.9; treatment targets are HCO3- above 15 mmol/L and pH above 7.3. In children, capillary blood ketone above 0.5 mmol/L "indicates metabolic decompensation."

A conflict between sources, reported rather than resolved: Palmer & Clegg 2021 describe steady-state starvation ketosis with "a beta-hydroxybutyrate concentration of 8-10 mmol/L," while Huang 2024 gives a prolonged-starvation ceiling of about 6 mmol/L and the meta-analysis in (a) reports measured 8 to 10 day values of 4.0 to 5.9 mmol/L. The measured values in (a) sit below both review figures. The 8 to 10 mmol/L figure is the outlier and should not be quoted as the expected value of a fast.

What raises the risk when the person is also dehydrated, with the mechanism named and the evidence status stated: 1. The renal escape route closes. READ, Chinese consensus: with a raised renal threshold, "尿酮体可减少,甚至完全消失" (urine ketones may decrease or even disappear completely). READ, Palmer & Clegg: reabsorption in prolonged starvation already "spares approximately 225 kcal/day" of ketones that would otherwise be excreted. Volume depletion reduces filtered load further. So the acid load is retained rather than excreted. This mechanism is documented; its magnitude during a dry fast is not measured. 2. Volume depletion itself is the treated abnormality in both hyperglycemic crises, with estimated deficits of about 7 L in DKA and 10.5 L in HHS for a 70 kg patient, READ. Dehydration is not a benign co-traveler of a fast; it is the thing emergency protocols correct first. 3. The monitoring instrument fails in the direction that matters. Per (e1.4), the urine strip missed three of four frank DKA cases, and per Huang 2024 the urine result "is a lagging indicator of the severity of ketoacidosis." A person relying on a urine strip during a dry fast is using the one instrument documented to under-call the dangerous state. 4. Euglycemic ketoacidosis is documented with prolonged fasting, so a normal glucose does not exclude it. Records located and not retrieved, listed as leads with no claims attached: "Euglycemic diabetic ketoacidosis induced by sodium-glucose cotransporter 2 inhibitor in the setting of prolonged fasting: a case report," DOI 10.1186/s13256-022-03347-1; "Euglycemic ketoacidosis induced by therapeutic fasting in a non-diabetic patient" (Nutrition, publisher blocked); "Risk of diabetic ketoacidosis during Ramadan fasting: A critical reappraisal" (Diabetes Metab, publisher blocked); and, from the Japanese channel, a case of starvation ketoacidosis diagnosed by rapid BHB measurement, Nihon Kyukyu Igakukai Zasshi 2024, DOI 10.1002/jja2.12903 (publisher blocked). Four separate records exist in this space; none was opened, so none is graded.


PART 3: RECEIPT

Questions answered: 7 / 7

#QuestionAnsweredVerdict in one lineGrade
DisconfirmingReported first, as requiredyesFour independent disconfirming lines, one of them a randomized human trialA
(a)Normal time course, water-only fastyes~0.5 mmol/L at 24 h, 1-2 at 36 h, 2-3 at 48-72 h, wide scatterA
(b)Time course under fluid restrictionyesNo usable curve exists; the one controlled measurement found no ketosisA for that one result; otherwise an evidence gap
(c)Direct comparison dry vs water fastyesNone exists. Stated plainly as the findingn/a, absence
(d)Accelerating mechanismsyesThe measured human mechanisms do not accelerate ketogenesis; one opposes it; the 3:1 glycogen-water ratio is contestedmixed, stated per mechanism
(e1)Urine strips under dehydration (added)yesUninterpretable as a depth measure; six documented reasons; protocol specifiedB+
(e2)Blood hemoconcentration and clearanceyesReal but a few percent; cannot explain a multi-fold rise; no published BHB correction existsB+
(f)Ketoacidosis riskyesLine is BHB > 3.0 mmol/L with acidosis; dehydration closes the renal escape route and the urine strip under-calls the dangerA

Languages: 9 / 9 searched, all named

LanguageQuery runReturned
Englishmultiple, across all sub-questionsThe bulk of the evidence base: Berneis 1999, Keller 2003, Huang 2024, both AAFP reviews, Palmer & Clegg 2021, the IJMS meta-analysis, MacPherson 2018, Nutrients 2023, Khalfoun 2024, ADA/AACE criteria
Russianсухое голодание кетоз бета-гидроксибутират измерения обезвоживание кетоновые тела мочаEmpty of primary data. Consumer and clinical-laboratory pages only (citilab.ru, Yandex Health, translated keto vendor content). No indexed primary measurement of ketones in dry fasting, despite dry fasting being a well-known practice in Russian-language sources
GermanTrockenfasten Ketose Beta-Hydroxybutyrat Dehydratation Ketonkörper Urin Messung StudieEmpty of primary dry-fast data. Fasting-clinic pages (Buchinger Wilhelmi), a forum thread pointing to the 1,610-subject fasting-induced ketosis study, a DRKS trial registry entry (DRKS00036696), a forensic-medicine dissertation. All water or liquid fasting, none dry
Frenchjeûne sec cétose bêta-hydroxybutyrate déshydratation corps cétoniques urine mesure étudeEmpty of primary data. Consumer and encyclopedia pages, plus a French historical review of ketone bodies. No primary measurement
Arabicصيام رمضان بيتا هيدروكسي بيوتيرات الأجسام الكيتونية قياس الجفاف الكيتوزيةEmpty of primary data. Keto blogs, a translated vendor article, a news health feature. Notable as a null given that Ramadan is the largest fluid-restricted fasting practice in the world
TurkishRamazan orucu beta-hidroksibutirat keton cisimleri ölçüm dehidratasyon ketozis kanPRODUCTIVE. Yielded Kuru et al. 2014, Turk J Emerg Med 14(2):47-52, full text retrieved: the 265-patient finger-stick BHB versus urine dipstick comparison that supplies the measured discordance in e1.4
Persianروزه خشک کتوز بتا هیدروکسی بوتیرات کم آبی اجسام کتونی اندازه گیریEmpty of primary human data. Laboratory test reference pages and veterinary bovine ketosis articles
Chinese干性断食 酮症 β-羟丁酸 脱水 尿酮体 测量 血酮PRODUCTIVE. Yielded the national expert consensus, Chin J Endocrinol Metab 2014;30(3):177-183, full text retrieved: the acetoacetate and acetone sensitivity figures, the 58% versus 93% DKA detection comparison, and the renal-threshold statement in e1.5
Japanese断水 断食 ケトーシス βヒドロキシ酪酸 脱水 尿ケトン体 測定 濃縮PRODUCTIVE. Yielded Sekizuka et al., Glycative Stress Research 2018;5(4):181-191, full text retrieved: serial BHB and acetoacetate across a 7-day water-only fast, supplying the BHB-to-acetoacetate ratio. Also surfaced a starvation-ketoacidosis case report, DOI 10.1002/jja2.12903, publisher blocked and not retrieved

Three of the eight non-English channels produced primary evidence that the English channel had not surfaced in the same form. The Turkish and Chinese results are load-bearing for sub-question (e1).

Channels blocked, abandoned on first refusal per channel law

Never retried after refusal: eutils.ncbi.nlm.nih.gov (proxy 403), www.ebi.ac.uk Europe PMC REST (proxy 403), www.ncbi.nlm.nih.gov/books (captcha), pmc.ncbi.nlm.nih.gov (captcha after one success), pubmed.ncbi.nlm.nih.gov (empty response), journals.physiology.org (client error), onlinelibrary.wiley.com and wchh.onlinelibrary.wiley.com (client error, robots disallowed), sciencedirect.com (client error), ajcn.nutrition.org (client error), eje.bioscientifica.com (robots unreachable), endocrino.org.br (resolved to a homepage).

Effect on the result: four records named in (d5) and (f) could not be opened and are reported as leads with no claims and no grades attached. Everything graded in this file was retrieved and read.

Open enumeration, stated as required before a verdict

1. No study anywhere found that titrates hydration against a urine ketone strip reading at constant ketone production. This is the experiment that would settle e1.2 directly. 2. No published hemoconcentration correction for blood BHB exists in any language searched. 3. The vasopressin pathway (d5) has located but unopened records and is the one mechanistic lane not closed. 4. Four euglycemic and starvation ketoacidosis records in (f) are located but unopened. 5. The Palmer & Clegg 8-10 mmol/L starvation figure conflicts with both Huang 2024 and the measured meta-analysis values and is unreconciled.

PLAIN VERDICT

Withholding water does not measurably accelerate ketosis in humans, and the proposition is simultaneously unproven and contradicted by the one controlled experiment that tested it.

Stated in three parts, because the three are different claims:

1. Unproven. No study compares a dry fast against a water fast with a ketone endpoint. There is no BHB-versus-time curve for a fluid-restricted fast. The claim has no supporting measurement behind it, in nine languages.

2. Contradicted as far as it has been tested. A randomized crossover trial in 10 healthy men restricted fluid and infused hypertonic saline for 17 hours, reached 296.4 mOsm/kg, and measured BHB at 0.086 mmol/L, not significantly different from the drinking arm and below the threshold for nutritional ketosis. The same trial found significantly increased hepatic glucose production, which is the anti-ketogenic direction. Where tonicity has been shown to move lipolysis in humans, it moved it upward under water loading, not water restriction. And the clinical state with the largest measured water deficit, HHS at roughly 10.5 L, is defined by having almost no ketones.

3. The apparent effect, as evidenced by a urine strip, is substantially an artifact. A urine ketone strip reports the concentration of acetoacetate, which is not the ketone that dominates deep ketosis, in a fluid whose concentration is precisely what dehydration raises. Elevated specific gravity and acidic pH, the two signatures of a dehydrated fasting void, are both listed false-positive causes for that exact reagent pad in two independent published reviews. The published dilution correction implies a factor of about two across the normal specific-gravity range alone. The founder's 2017 reading is additionally at the ceiling of the scale and therefore censored, so it carries no comparative depth information in either direction.

The distinction that matters for the brand: a blood BHB rise during a fast is real and hemoconcentration accounts for only a few percent of it, so nothing here undermines ordinary fasting ketosis. What fails is specifically the inference from a urine strip during a dry fast to deeper or faster ketosis. That inference cannot be made from that instrument, and no published evidence supports the underlying physiological claim.

If the question is worth settling, the protocol in (e1) settles it for one person, and the paired drinking-arm control is the step without which no answer is possible.


dryfasting.com has been published since 2014. This evidence record was compiled on 22 November 2026. Every row names the fasting modality the study actually ran, the system it was measured in and the grade. Where no source carries a DOI or PMID the row says so rather than supplying one. Nothing is sold on this site.