The Beverage Hydration Index is reproducible. The harder question is whether it measures what its name appears to promise.
Permanent research record: Nikitin, I. (2026). What Does the Beverage Hydration Index Actually Measure? A Construct-Validity Audit (Version 1.0). The Null Institute. https://doi.org/10.5281/zenodo.23194332 · Archived report and supporting tables. For citation and archival use, use the DOI report. It is the final narrowed Version 1.0; the broader exploratory analyses retained on this web page are preserved for transparency and are not part of the archived report.
In 2016, a randomized trial introduced the Beverage Hydration Index (BHI) as a way to compare the short-term fluid balance produced by different drinks. The experiment was straightforward: euhydrated participants consumed one litre of a beverage, urine was collected for several hours, and the result was expressed relative to water [1].
The index has since escaped the laboratory. Scientific papers describe the “hydration properties” of beverages. Media articles rank drinks by how hydrating they are. Commercial pages say that milk, electrolyte mixtures, or proprietary formulations “hydrate better than water.”
Those statements may sound like natural translations of the original experiment.
They are not necessarily equivalent.
Our question was not whether the BHI is reproducible. It was whether a higher BHI is a valid surrogate for a broader claim: that a beverage hydrates the body better in a physiologically or functionally meaningful sense.
We approached that question in three steps: a construct-validity audit of the index itself, an audit of downstream exercise outcomes, and an exploratory 30-document audit of how BHI claims change as they move from primary science into media and professional or commercial communication.
1. What does the BHI actually measure?
The original trial tested 13 commonly consumed drinks in 72 healthy men. Participants began euhydrated and consumed 1 L over 30 minutes. Urine output and net fluid balance were then followed for four hours [1].
Operationally, the BHI is built from cumulative urine output. At a given time point, it compares how much urine is produced after water with how much is produced after the test beverage.
BHI(t) ≈ cumulative urine after water / cumulative urine after test beverage
If less urine is produced after the test drink, the BHI rises.
That is a perfectly legitimate measurement. But notice what is not directly measured by the index: how rapidly the water entered the circulation, how it was distributed among body compartments, whether intracellular or extracellular hydration improved, whether cognition or symptoms changed, or whether a person subsequently performed better.
The distinction matters because urine output is only one term in the water-balance equation.
Systemic water gain ≈ absorbed water − urine − other losses
Two beverages can therefore produce similar urine output while delivering very different amounts of water into the circulation over the same interval.
2. A higher BHI can accompany slower fluid delivery
A 2024 randomized crossover experiment provides an unusually direct test of this problem. Fourteen young adults consumed water or beverages containing glycerol and sodium, with or without sucrose or the slower-absorbing carbohydrate isomaltulose [2].
Adding isomaltulose increased the BHI by about 45%: 3.43 versus 2.50 for glycerol plus sodium alone. If BHI were simply a scale of how effectively water had already been delivered into the body, that would suggest a substantial improvement.
But plasma-volume expansion moved in the opposite temporal direction. It appeared earliest with glycerol plus sodium, at about 30 minutes, later with sucrose, at about 90 minutes, and latest with isomaltulose, at about 120 minutes. The authors attributed the higher BHI with isomaltulose to slower absorption [2].
A higher BHI can therefore coexist with slower delivery of fluid into the circulation.
This is not merely an external objection to the method. A 2021 BHI study explicitly noted that the model cannot determine whether lower urine output reflects greater retention of absorbed water or slower entry of water from the gastrointestinal tract [3].
Those mechanisms can produce the same score while representing different physiological states.
3. Renal physiology is also inside the score
The BHI is also affected by how quickly the kidneys excrete a water load.
A randomized trial comparing 12 younger and 12 older adults found that older participants remained in positive fluid balance longer after beverage ingestion. The conventional two-hour BHI was still changing beyond that point for some beverages, and the authors concluded that a four-hour BHI may be more appropriate in older adults because of attenuated urine excretion rates [4].
This creates another construct problem. A physiological state that slows renal clearance can make fluid remain in positive balance longer. That does not automatically mean that a beverage has acquired a superior intrinsic “hydrating power.”
The score is therefore better understood as a property of a system:
beverage × gastrointestinal kinetics × renal handling × baseline state × population × observation window
That is substantially narrower than treating BHI as an intrinsic ranking of drinks.
4. Reliability is not validity
A 2018 randomized trial tested the reproducibility of the BHI and whether body mass or sex materially altered the results. The method performed reasonably well: the authors found no meaningful effect of sex or body mass and reported robust reproducibility analyses [5].
That is evidence in favour of the BHI, not against it.
But it answers a different question.
Reliability: does the measurement behave consistently?
Construct validity: does the measurement represent the concept we say it represents?
Criterion validity: does it predict an independent outcome that gives the construct practical meaning?
A ruler can be perfectly reproducible while measuring the wrong dimension for the question being asked.
Other BHI experiments reinforce this distinction. An isotonic beverage produced a higher BHI and more positive net fluid balance than water, yet the study did not find corresponding beverage effects on total, intracellular, or extracellular body water [6]. These measures are not perfect gold standards either, but the dissociation illustrates the problem: different operational definitions of “hydration” need not move together.
5. Does greater fluid retention predict a meaningful outcome?
To test the stronger surrogate claim, we moved away from BHI studies themselves and asked a downstream question:
When an intervention clearly increases fluid retention, does a meaningful functional outcome improve with it?
Exercise performance provides a useful natural laboratory because hyperhydration experiments often quantify retained fluid, plasma volume, body-mass loss, cardiovascular strain and a subsequent time-trial outcome in the same participants.
We froze a working audit set of 12 time-trial studies: nine contained in a 2023 systematic review of pre-exercise hyperhydration [7], plus three later studies identified after the review’s search window [8–10].
The result was:
3 positive time trials / 9 null time trials.
This ratio is not a meta-analysis. The studies differ in temperature, exercise mode, duration, hydration agent, comparator and access to fluid during exercise. A count of significant studies cannot estimate an average treatment effect.
A separate 2024 meta-analysis reached a different but compatible result. Pooling hyperhydration studies, it found a small-to-moderate improvement in time-trial performance (Hedges’ g = 0.25, 95% CI 0.002–0.51; p = 0.049) and did not identify significant moderation by ambient temperature, hyperhydration agent, exercise mode or extent of hyperhydration [12]. We therefore do not interpret 3/12 as evidence that hyperhydration has no average ergogenic effect. Our narrower test is whether the amount of fluid retained behaves as a monotonic surrogate for downstream benefit.
But it can test a weaker proposition: if retained fluid were a broadly valid monotonic surrogate, very large increases in retention should not repeatedly coexist with null functional outcomes.
A useful falsification pair
In one running experiment, sodium hyperhydration increased pre-exercise body mass by about 1 kg, reduced subsequent body-mass loss from 3.1% to 1.4%, lowered heart rate by about 5 beats/min and reduced rectal temperature by about 0.3°C. Yet the 18-km time trial was essentially unchanged: 85.6 versus 85.3 minutes, p = 0.82 [8].
A much newer half-marathon experiment produced an even cleaner retention contrast. Glycerol plus sodium increased fluid retention by 1,189 mL and plasma volume by 9.0%, but did not significantly improve half-marathon performance, heart rate or core temperature [9].
Yet a 2025 trial in female cyclists found a positive result under severe heat stress. Sodium hyperhydration increased pre-exercise fluid retention by about 509 mL and improved a 200-kJ cycling time trial, despite no significant improvement in the thermal or cardiovascular strain measures highlighted in the original performance paper [10,11]. A published erratum exists for the performance article [13]. Because the correction text was not available in the sources used for this audit, we treat this trial as a positive counterexample rather than as a decisive estimate of effect size.
The important result is not that hyperhydration “does not work.” It sometimes does.
The important result is that the magnitude of retention itself does not behave as a simple dose-like predictor of benefit.
6. The largest retention effects were not the positive ones
For six studies in our audit, we could extract a reasonably direct numerical difference in retained fluid between conditions.
| Outcome | Approximate additional retention |
|---|---|
| Positive time trials (3 studies) | ~350–577 mL |
| Null time trials (3 studies) | ~846–1,354 mL |
Every positive study in this small numerical subset had a smaller retention increment than every null study.
This is a post-hoc descriptive observation, not evidence that retaining more water harms performance. The subset is small and selected by availability of extractable data. But it is strongly incompatible with a simple linear story in which more retained fluid should mean more functional benefit. The small pooled time-trial benefit found in the 2024 meta-analysis [12] does not resolve this surrogate question: an intervention can have a positive average effect without the magnitude of an intermediate marker reliably predicting the magnitude of that effect.
We also tried to rescue the relationship with simple moderators. Severe heat appears relevant, and all of the clearest positive time trials occurred under substantial heat stress. But heat is not sufficient: several null studies were also performed near or above 30°C [7]. A simple threshold such as “benefits appear once dehydration exceeds 2% of body mass” also fails because the 18-km running study reduced the control condition’s 3.1% loss to 1.4% without improving performance [8].
The best remaining model is contextual: extra fluid matters when it changes a physiological bottleneck that actually limits the outcome. That model is plausible, but it is much less convenient than a universal hydration ranking and is difficult to infer from BHI alone.
7. We looked for criterion validation and did not find it
We then searched specifically for studies that used BHI prospectively as a predictor of an independent meaningful endpoint: cognition, symptoms, exercise performance, heat illness, clinical outcomes or another criterion external to the urine/fluid-balance construct.
We did not identify such a validation study in the corpus we found.
This is a search result, not proof of non-existence. A rare study may have escaped our search. But the BHI literature we identified mainly validates the assay against itself or nearby physiological measures: reproducibility, beverage composition, urine output, fluid balance, population differences and related biomarkers.
That supports a narrow interpretation:
BHI is a standardized measure of the short-term urine-retention / fluid-balance response to a beverage challenge.
It does not by itself validate the stronger statement:
This beverage hydrates the body better in a generally meaningful sense.
8. How a narrow measurement becomes a broad claim
We next examined the language surrounding the index.
The original abstract is cautious. It says BHI may be useful for identifying the short-term hydration potential of beverages consumed in a euhydrated state [1].
But semantic broadening begins surprisingly early. The discussion of the same paper describes BHI as an objective assessment of the effectiveness of beverages in maintaining hydration status and suggests that industry might use hydration potential in product labelling. Later scientific papers use formulations such as “hydration potential” and “relative hydration properties” [3,5].
Consumer communication often removes the remaining qualifiers. The question becomes not “which drink produced less urine during a four-hour laboratory test in already hydrated participants?” but “which drink is most hydrating?” Milk and other drinks are then described as “more hydrating than water.”
The shift is subtle because every step sounds linguistically reasonable. But the endpoints are not equivalent.
less urine → higher BHI → greater hydration potential → more hydrating → better hydration
Each arrow adds an inference.
9. A 30-document exploratory claim audit
To see whether this impression survived a simple falsification attempt, we froze an exploratory sample of 30 documents: 10 scientific sources, 10 media sources and 10 industry or professional-education sources.
Before coding the sample, we defined two simple scores.
- Claim strength, 0–4: from a purely operational fluid-balance description to categorical statements such as “more hydrating than water” or “most hydrating.”
- Qualifier preservation, 0–4: one point each for retaining the short-term time window, euhydrated/resting context, urine/retention mechanism, and an explicit methodological limitation.
| Source type | n | Mean claim strength | Mean qualifier score | Strong claim ≥3 | ≥3 qualifiers |
|---|---|---|---|---|---|
| Science | 10 | 1.8 | 2.8 | 2/10 | 6/10 |
| Media | 10 | 3.7 | 1.1 | 10/10 | 0/10 |
| Industry / professional | 10 | 3.3 | 2.1 | 8/10 | 4/10 |
The gradient is large enough to be interesting, but it must not be overinterpreted. This was a purposeful stratified web sample, not a random bibliometric sample of every paper or webpage discussing BHI. Source discovery was search-engine based and coding was not blinded. The numbers therefore do not estimate the prevalence of misleading claims in science, media or industry.
They do show that our qualitative claim-drift hypothesis did not disappear when tested against a fixed mixed sample.
The result is also less moralistic than a standard “bad media” story. Scientific language itself becomes broader after the original experiment. Professional sources are heterogeneous: some preserve the laboratory limitations very well, while promotional sources often compress the result into a product comparison.
The pattern is better described as distributed claim amplification.
10. A second problem: ranking inflation
BHI results are often displayed as league tables: beverage A ranks above beverage B, which ranks above water.
But rank order is not the same as demonstrated difference. In the original experiment, most tested beverages were not statistically distinguishable from water at the relevant time points; the clearest differences were concentrated in oral rehydration solution and milk conditions [1].
A continuous noisy measurement can always be sorted. Sorting creates a first place, second place and tenth place even when uncertainty intervals overlap and pairwise evidence is weak.
Rank ≠ established superiority.
11. What survived the audit?
Several potential criticisms did not survive.
The BHI is not meaningless. It measures a real physiological response. The method appears reasonably reproducible [5]. Beverage composition can change urine output and short-term fluid balance [1,3]. Under some conditions, increasing retained fluid can also improve subsequent exercise performance [10].
What did not survive is the stronger universal interpretation.
A higher BHI can be produced partly by slower gastrointestinal delivery [2]. It changes with population and observation window because renal handling changes [4]. Larger retention effects do not monotonically predict better time-trial outcomes [7–10]. And in our targeted search we did not find independent prospective criterion validation that would justify treating BHI as a general surrogate for meaningful hydration.
The most defensible description is therefore narrow: BHI is a reproducible short-term urine-retention / fluid-balance response to a standardized beverage challenge.
Calling one beverage “more hydrating” than another may sometimes be a useful shorthand. But scientifically, that phrase hides assumptions about absorption, renal function, time horizon, physiological state and the outcome that actually matters.
12. The broader lesson
This case is not mainly about milk, electrolyte drinks or water.
It is about a recurrent problem in measurement science.
A narrow biomarker can be useful, reliable and mechanistically interesting. Once it receives an intuitive name, however, the name can quietly become the construct. Reproducibility is then mistaken for validation, correlated mechanisms become causal interpretations, and an operational measurement becomes a property of the product being measured.
The error is difficult to notice because no single step has to be absurd.
The BHI illustrates the sequence unusually clearly:
urine output → fluid retention → hydration potential → hydration quality → consumer ranking
The first step is measured. The later steps require progressively more evidence.
Our audit suggests that the language expanded faster than that evidence did.
Data and materials
The audit workbook contains the 12-study time-trial dataset, sensitivity subsets, BHI construct-validity evidence, the qualitative claim chain, and the full 30-document exploratory claim sample with coding rules and source URLs.
The workbook is provided to make the judgments in this article inspectable. The 30-document claim sample is exploratory and non-random; it should not be treated as a population estimate.
References
- Maughan RJ, Watson P, Cordery PA, et al. A randomized trial to assess the potential of different beverages to affect hydration status: development of a beverage hydration index. American Journal of Clinical Nutrition. 2016;103(3):717–723. doi:10.3945/ajcn.115.114769.
- Otsuka J, Okamoto Y, Enoki Y, et al. Effects of ingesting beverages containing glycerol and sodium with isomaltulose or sucrose on fluid retention in young adults: a single-blind, randomized crossover trial. Applied Physiology, Nutrition, and Metabolism. 2024;49(5):667–679. doi:10.1139/apnm-2023-0483.
- Millard-Stafford M, Snow TK, Jones ML, Suh H. The Beverage Hydration Index: Influence of Electrolytes, Carbohydrate and Protein. Nutrients. 2021;13(9):2933. doi:10.3390/nu13092933.
- Clarke MM, Stanhewicz AE, Wolf ST, et al. A randomized trial to assess beverage hydration index in healthy older adults. American Journal of Clinical Nutrition. 2019;109(6):1640–1647. doi:10.1093/ajcn/nqz009.
- Sollanek KJ, Tsurumoto M, Vidyasagar S, Kenefick RW, Cheuvront SN. Neither body mass nor sex influences beverage hydration index outcomes during randomized trial when comparing 3 commercial beverages. American Journal of Clinical Nutrition. 2018;107(4):544–549. doi:10.1093/ajcn/nqy005.
- Bechke EE, Zaplatosch ME, Choi J-Y, Adams WM. Utility of an Isotonic Beverage on Hydration Status and Cardiovascular Alterations. Nutrients. 2022;14(6):1286. doi:10.3390/nu14061286.
- Jardine WT, Aisbett B, Kelly MK, et al. The Effect of Pre-Exercise Hyperhydration on Exercise Performance, Physiological Outcomes and Gastrointestinal Symptoms: A Systematic Review. Sports Medicine. 2023;53:2111–2134. doi:10.1007/s40279-023-01885-2.
- Gigou P-Y, Dion T, Asselin A, Berrigan F, Goulet EDBG. Pre-exercise hyperhydration-induced bodyweight gain does not alter prolonged treadmill running time-trial performance in warm ambient conditions. Nutrients. 2012;4(8):949–966. doi:10.3390/nu4080949.
- Jardine WT, Condo D, Aisbett B, et al. The Addition of Glycerol and Sodium Chloride to a Hyperhydration Protocol Does Not Improve Half-Marathon Time-Trial Performance in Trained Runners in Warm Conditions. International Journal of Sports Physiology and Performance. 2025;20(6):770–778. doi:10.1123/ijspp.2024-0268.
- Convit L, Orellana L, Périard JD, et al. Sodium Hyperhydration Improves Performance With No Change in Thermal and Cardiovascular Strain in Female Cyclists Exercising in the Heat Across the Menstrual Cycle. International Journal of Sport Nutrition and Exercise Metabolism. 2025;35(2):99–111. doi:10.1123/ijsnem.2024-0125.
- Convit L, Orellana L, Périard JD, et al. Hydration Responses to Pre-Exercise Sodium Hyperhydration at Rest and During Cycling in the Heat and Across Menstrual Cycle Phases. Nutrients. 2025;17(23):3672. doi:10.3390/nu17233672.
- McCubbin AJ, Irwin C. The effect of pre-exercise oral hyperhydration on endurance exercise performance, heart rate, and thermoregulation: a meta-analytical review. Applied Physiology, Nutrition, and Metabolism. 2024;49(5):569–583. doi:10.1139/apnm-2023-0384.
- International Journal of Sport Nutrition and Exercise Metabolism. Erratum: Sodium Hyperhydration Improves Performance With No Change in Thermal and Cardiovascular Strain in Female Cyclists Exercising in the Heat Across the Menstrual Cycle. International Journal of Sport Nutrition and Exercise Metabolism. 2025;35(3):269. doi:10.1123/ijsnem.2025-0013.
Article type: Research Article · Evidence Audit
Scope note: This is a falsification-oriented evidence audit, not a preregistered systematic review or meta-analysis. The exercise dataset was frozen after a systematic-review backbone plus identified post-review studies. The 30-document claim sample was purposefully stratified across science, media and industry/professional sources; it was not randomly sampled and its percentages should not be generalized to the entire literature or web. Negative searches for criterion validation should be read as “not identified in this audit,” not as proof that no such study exists.