Shifting Focus: From Hydration for Performance to Hydration for Health

Water has been described as the most essential nutrient: neglected, underappreciated, and under researched [1, 2, 3, 4]. Early work on water intake, dehydration, and survival was conducted during the 1940s when Adolph published a series of experiments demonstrating that “life and efficiency in a hot climate depends upon the degree of dehydration, which is the overriding factor” [5, 6]. However, the bulk of modern research into the effects of (de)hydration was not conducted until the 1970s through the 1990s, when a number of research groups began examining the acute effects of dehydration on physical performance. Conducted largely from the perspective of athletic and military performance and safety, this body of work demonstrated conclusively that hypovolemia (a reduction in plasma volume) and hypertonicity (an increase in plasma osmolality) due to uncompensated sweat loss impact both cardiac output and thermal strain [7, 8, 9]. In order to minimize performance and safety concerns during exercise, authorities recommend limiting body water loss to no more than 2% of body mass [7, 10, 11].

This body of work also had implications that reached beyond the realm of sports performance and safety. In establishing the Dietary Reference Values for water in 2004, the Institute of Medicine found insufficient evidence to set a water requirement as a means to treat or prevent a specific disease or condition [12]. Therefore, “hydration status, as assessed by plasma or serum osmolality, (was) the primary indicator” (pp. 144) was used to establish the Adequate Intake (AI) for water. The AI for total water was set as the median consumption reported in NHANES III, as serum osmolality was found to be unchanged across deciles of total water intake [13]. A normal hydration status, or hydration state (i.e., the maintenance of total body water content, represented by unchanged plasma osmolality) can be achieved with a broad range of total water intake volumes.

The concept of hydration as a state, representing the amount of water in the body at a given time, is highly relevant for short-term performance and safety, particularly during physical activity in hot environments or in the absence of free water access. The hydration state is also of critical concern in vulnerable populations, including the elderly, who have an increased prevalence of plasma hypertonicity [14]. Several studies suggest that dehydration is associated with increased mortality rates among hospitalized older adults [15, 16]. Moreover, dehydration may precipitate emergency hospitalization or increase the risk of repeated hospitalization [15, 17, 18], and some evidence suggests that in the elderly, dehydration may contribute to constipation, impaired cognitive function, falling, orthostatic hypotension, salivary dysfunction, poor control of hyperglycemia in diabetes, or hyperthermia; and that water supplementation may alleviate some of these conditions [18, 19, 20, 21]. In considering this evidence, it is important to remember that older adults represent a vulnerable population with specific hydration challenges, including reduced thirst, lower lean body mass (and thus lower body water), and a reduced capacity to concentrate urine. In contrast, there is little to no evidence that any significant proportion of the general, healthy population is chronically dehydrated.

Thus, if we restrict our definition of hydration to that of a state, there is little evidence for broadly generalizable health benefits. Yet, over the past decade, the hydration research landscape has shifted from studying hydration as a state to investigating the processes and mechanisms through which the hydration state is maintained. While individuals who consume low fluid volumes ad-libitum (“low drinkers”) maintain a normal plasma osmolality, other physiological markers are significantly different between low-volume and high-volume drinkers. Low drinkers have significantly lower 24-h urine volume, higher urine concentration (osmolality, specific gravity, and color) [22, 23, 24, 25, 26, 27] and, in some studies, demonstrate higher plasma vasopressin (AVP) or even higher cortisol [27, 28, 29], compared to their high-drinking counterparts. Although the values for urinary and blood markers in low drinkers remain within normal limits, these differences suggest that chronic low drinking triggers a sustained antidiuretic response to conserve body water using the normal, renal concentrating mechanism. Shifting the paradigm toward redefining hydration as a dynamic, ongoing process to maintain water balance allows for distinctions between low total water intake, and low urinary output; from high water intake, and high urinary output [30]. Recent studies linking low water intake, low urine output, or high AVP (or its surrogate, copeptin) to various health outcomes or risk factors [31, 32, 33, 34, 35, 36, 37, 38, 39] suggest that maintaining the hydration state in the face of habitual low intake has a cost. The entire process of hydration (from what goes in, to what comes out, to what happens internally to maintain body water homeostasis) may be more important than the hydration state when considering implications for kidney or cardiometabolic health (Fig. 1) [40]. Thus, the time is ripe for a fundamental shift in the research questions linked to hydration: what are the metabolic or health costs of maintaining the hydration state when drinking is low?

Multiple associations have been reported between fluid intake and kidney and metabolic health outcomes, including chronic kidney disease (CKD), recurrence of kidney stones, and some metabolic markers of hyperglycemia. In an Australian cohort, Strippoli et al. [41] found that individuals with the highest quintile of fluid intake had significantly reduced risk of CKD. In a follow-up publication, this same research group further reported that if one excludes plain water from the analysis [42], there was no longer an association with kidney function or with cardiovascular or all-cause mortality, suggesting that plain water intake may play a determining role. This is further supported by an analysis of NHANES data by Sontrop et al. [32], who found CKD prevalence to be the highest among those with the lowest total water intake. When intake was subdivided into plain water versus other beverages, the association was found to hold only for plain water, further suggesting that water, in particular, may play a protective role in kidney disease. Higher total fluid intake has also been associated with lower relative risk for kidney stones in the Nurses' Health Study II [43], and an intervention to increase water intake reduced recurrence of kidney stones in a randomized, controlled trial [39]. In terms of metabolic disease endpoints, low water intake has been associated with new-onset hyperglycemia in a prospective French cohort with a 9-year follow-up [31]. More recently, Carroll et al. [38] also reported an association between higher plain water intake and lower HbA1c in men, but not in women, participating in the UK National Diet and Health Survey. Their substitution analysis revealed that the effect of plain water was not due to substituting water for other beverages; rather, it was the addition of plain water that appeared to be the pertinent factor. Overall, a key challenge is to design and carry out intervention studies in order to demonstrate that fluid intake is causally related to kidney and metabolic health endpoints. Nevertheless, we appear to be approaching a critical mass of associative evidence for the role of fluid intake, and specifically plain water intake, in kidney and cardiometabolic health.

High urine volume (and low urine concentration) is the logical end result of a high fluid intake. It is not surprising, then, that the associations seen between fluid intake and kidney health endpoints are also reported using urinary output as a predictor. Clark et al. [33] found that maintaining a high versus low 24-h urine volume reduced the rate of estimated glomerular filtration rate (GFR) decline in a prospective cohort with a 5.7-year follow-up. In another short-term intervention study, increasing urine volume and lowering urine concentration by drinking more water was also shown to reduce the Tiselius Crystallization Risk Index in 24-h urine samples, a risk factor for recurrent kidney stones [44]. The European Urology Association also recommends that a low urine concentration be maintained in order to prevent stone recurrence [45]. For stone disease, it is likely that reducing urine concentration simply decreases the likelihood for supersaturation and thus crystal formation. For CKD, however, the explanation for why increased urine output may be protective is more complex. Clark et al. [33] and Sontrop et al. [32] offer several hypotheses: the kidney requires a minimum “obligatory” urine volume to remove waste, and it is possible that the kidneys function more efficiently with a more abundant supply of water. It is also possible that the production of more concentrated urine incurs a greater metabolic demand on the kidney [33], or that higher urine concentration may, over time, contribute to glomerular hyperfiltration and albuminuria [32]. One significant challenge in assessing the relation between urinary output and health endpoints is that in a majority of cohorts, if urine is collected at all, it is a single (random) or morning (fasting) sample, which may or may not represent 24-h urine. The existing literature, however, consistently supports a benefit of increased fluid or plain water intake. One possible solution to the challenge of collecting 24-h urine samples in a large cohort is to instead evaluate plasma copeptin as a surrogate to vasopressin, as an indicator of the activity of the urine concentrating or diluting mechanism, and as an indicator of hydration.

It is difficult to measure the levels of vasopressin, the primary hormone responsible for regulating water reabsorption in the kidney, under normal circumstances, as its concentration in the blood is quite low (25th-75th percentile in one of our research group's prior studies: 1.1-3.4 pmol·L^-1[27]) and its half-life is short (i.e., 10-20 min in plasma). Vasopressin is secreted in a 1:1 ratio along with copeptin and neurophysin II, which together form the pre-pro-vasopressin precursor. The recent development of a sensitive copeptin assay kit for clinical use [46] has popularized the measurement of copeptin in clinical studies, where it is considered to be a surrogate measure for vasopressin. Over the past several years, plasma copeptin has also been measured in some cohort studies, allowing for associations to be made with health outcomes. Copeptin has been independently associated with diabetes, metabolic syndrome, abdominal obesity, metabolic heart disease, and death [34, 35, 36, 37], as well as with GFR decline and progression toward CKD [47, 48]. One limitation is that widespread copeptin measurement is relatively new, and thus the number of large cohorts is limited. A majority of the available results were obtained from analyses performed on the Malmö Diet and Cancer Study (Sweden) and D.E.S.I.R. (France) cohorts [47, 48]. One major strong point of the analyses, however, is that copeptin remained an independent risk factor after controlling for covariates most often associated with kidney and metabolic disease. The inclusion of copeptin in future cohorts, or analysis of stored blood samples in existing cohorts, would facilitate the confirmation of these findings.

Given the evidence described above, a strong case exists for pursuing further research into the links between the global hydration process (including water intake, urinary output, and copeptin concentration) and health. Selecting the appropriate biomarkers to assess the global hydration process, and measuring them under the right conditions, is thus fundamental to advancing the research on hydration and health. This is especially important when considering that in different experimental conditions, the same biomarker may be an indicator of the hydration state and of the global hydration process to maintain total body water or both. Consider the following 2 studies on similar populations but with vastly different protocols: Munoz et al. [49] subjected young, healthy volunteers to both passive (sedentary) and active (exercise) dehydration in the heat; serum osmolality, body mass loss, and urinary hydration biomarkers were tracked throughout the progressive body water loss. In a separate study, Perrier et al. [27] measured plasma and urinary hydration biomarkers in young, healthy adults with habitually different fluid intakes (low: 0.74 L/day; high: 2.70 L/day) over 4 consecutive days, with no dehydration intervention. Figure 2 illustrates plasma (or serum) osmolality (P_Osm), urine osmolality (U_Osm), and urine specific gravity (U_SG) at baseline vs. 1% body mass loss for the dehydration study [49], and for high vs. low drinkers for the habitual fluid intake study [27]. Only during the active dehydration condition, did P_Osm significantly rise at 1% body mass loss. Conversely, a significant change in urine concentration (U_Osm and U_SG) was detected at 1% body mass loss only for the passive dehydration condition, and for high vs. low drinkers. High urine concentration may thus indicate mild, gradual dehydration, or chronic low fluid intake, depending upon the context, whereas a change in plasma osmolality is a sensitive measure of mild to moderate acute dehydration but does not appear to differ as a function of fluid intake or during mild passive body water loss.

While U_Osm and U_SG may be used in research or clinical practice as hydration markers, both measures require some degree of specialized equipment and technical expertise, and are thus less well-suited for day-to-day hydration monitoring. Individualized hydration tracking via a portable densitometer has been shown to be feasible for daily hydration monitoring and the secondary prevention of urinary tract infection [50], and maintaining a U_SG below 1.010 is recommended in the secondary prevention of kidney stone disease [45]. While patients may be motivated to track hydration via U_SG to prevent recurrence, a simpler method to increase awareness of drinking habits for the general population is to simply monitor urine color (U_Col). Using an 8-point color scale originally developed by Armstrong et al. [51, 52] for dehydration assessment, the utility of U_Col to accurately detect elevated U_Osm has been demonstrated in healthy adults [53], children [54], and in pregnant and breastfeeding women [55]. Periodically monitoring U_Col and increasing water intake in response to more concentrated urine (shade 4 or higher on the 8-point scale) is likely a sufficient level of hydration monitoring for most individuals in the general population.

Because fluid intake occurs periodically throughout the day, and not on a set schedule [56], and because there exists circadian variation in vasopressin [57] that, independently of fluid intake, influences urinary excretion [24], the hydration process is best assessed through the collection of a 24-h urine sample. Twenty-four-hour urine concentration also tracks day-to-day changes of daily fluid intake [24, 28, 29], and thus it can be used as a clinical measure during fluid intake studies as an outcome or as an indicator of subject compliance with drinking instructions. However, 24-h collections are cumbersome, often difficult for subjects, and impractical for any research situation other than short-term clinical trials. This poses a particular challenge in linking the hydration process to long-term health outcomes, as first evidence for associations is often obtained from large cohort studies, which cannot ensure 24-h collections on large sample sizes in free-living conditions.

Larger studies and cohorts such as NHANES in the United States, and the HELENA study in Europe, typically instead obtain single urine samples, which may represent either the first morning collection or a random sample obtained during the day [58]. However, first morning samples tend to overestimate 24-h urine concentration [22], even under situations of controlled fluid intake [24], since AVP levels peak overnight and sleeping effectively represents an overnight fluid restriction. Further, the difference between first morning and 24-h urine concentration may be dependent upon fluid intake volume. For low drinkers, who are essentially in an antidiuretic state all day, we previously observed that first morning U_Osm may not at all be different than the 24-h value (794 vs. 767 mOsm·kg^-1, respectively); whereas for high drinkers, the difference between the first morning and 24-h sample may be quite large (590 vs. 371 mOsm·kg^-1) [27]. Thus, even the use of a correction factor to adjust first morning urine values may prove to be complex.

Another approach involves determining a window during which a single urine sample would reasonably approximate 24-h U_Osm. To date, and to our knowledge, only one study has evaluated the equivalence of single to 24-h U_Osm in healthy, free-living young adults, with the aim of establishing a collection window that may provide guidance as to the best time for single urine collection in research and clinical practice. The results suggest that U_Osm measured on urine samples obtained in mid-afternoon to early evening (14:00-20:00) was equivalent (defined as falling within 100 mOsm·kg^-1) to the 24-h value, while samples taken earlier in the day tended to overestimate 24-h U_Osm[59]. One caveat to this conclusion, however, is that this study was performed in European subjects with Western dietary practices and meal times. It is possible that substantial deviations from this eating and drinking pattern, or night and shift work, may alter the window during which equivalence can be expected.

A key question in hydration assessment involves determining the thresholds or cut-off values that represent adequate hydration for health outcomes. A commonly referenced cut-off for the upper limit of euhydration is a U_Osm of 800 mOsm·kg^-1, which has been used in studies of hydration in both children [60, 61, 62, 63] and adults [64, 65]. The principle behind the 800 mOsm·kg^-1 cut-off has roots in early work by Katz et al. [66], who studied the urine-concentrating capacity of the kidney in hot conditions, and determined that nearly all participants could concentrate their urine up to at least 800 mOsm·kg^-1. This concept of the “minimum, maximum urine concentrating capacity” was also later explored by Manz and Wentz [67], who suggested that an upper limit for euhydration could be calculated as the mean maximum U_Osm, minus 2 SDs, as way to ensure euhydration for nearly all members of a given population. However, with the emerging distinction between the hydration state and hydration process [30], and with recent associations between low water intake, low urinary output, higher plasma copeptin, and kidney and metabolic outcomes [31, 32, 33, 34, 35, 36, 37], a lower target for 24-h urine concentration may be warranted.

Our research group has proposed a target for 24-h U_Osm of 500 mOsm·kg^-1[68], which represents a U_Osm that is consistent with daily intake sufficient to replace all water losses, ensure a sufficient urinary output to reduce the risk of kidney-related pathologies, and which may reduce circulating plasma vasopressin (copeptin). Given the growing evidence linking water intake, urine output, and copeptin to not only the kidney but also to cardiometabolic health, we argue that this value represents a reasonable target for 24-h hydration in the general population, and may contribute to reducing kidney and cardiometabolic risk over time. In order to expand the utility of this cut-off beyond clinical research, we also calculated the optimal thresholds for U_SG (≥1.013) and participant-assessed U_Col (≥4) to detect U_Osm >500 mOsm·kg^-1[53], thus opening the door for clinicians, health care practitioners, and individuals to monitor and adjust hydration behaviors accordingly.

These values represent a first, but not definitive, step toward refining and obtaining consensus regarding a target urine concentration for being well hydrated. In a recent publication, Cheuvront and Kenefick [69] acknowledge that there may be room for a “gray area between dehydration and euhydration,” which may be of interest in long-term health. Hydration, then, may be best viewed as a continuum between dehydration (fluid losses exceed fluid gains), the lower limit of euhydration (fluid losses are compensated by fluid gains; however, the kidneys conserve body water in order to avoid excess losses, resulting in low urine volume, high urine concentration, and increased circulating vasopressin), and being well hydrated (fluid losses are compensated by fluid gains, with a water reserve that allows for ample urine output, at a lower concentration, accompanied by lower circulating vasopressin). It is the latter, that is, being well hydrated, that may be the most important to long-term health, but Cheuvront and Kenefick [69] also rightly point out that the existing research is insufficient. Today, while there is some good evidence for increased water intake for renal health, the evidence linking being well hydrated to other health benefits, including cardiometabolic health, is more limited, is primarily based upon associations observed in large cohorts, and is centered around copeptin (vasopressin) as an independent predictor of health outcomes. While vasopressin is the central node in a working model linking water intake, hydration, and health (Fig. 1), more evidence is required to demonstrate a causal relationship between vasopressin and metabolism, and to demonstrate that increasing water intake can reduce circulating vasopressin and improve risk markers for disease.

After Cheuvront and Kenefick's paper was published in early 2016 [69], another study was published, which offers a first suggestion of a causal relationship between vasopressin and metabolism. Roussel et al. [70] investigated associations between baseline plasma copeptin and various glycemic and insulinemic risk markers for metabolic disease in a large French cohort with a 9-year follow-up. To address the question of causality between vasopressin and disease, they looked at several single nucleotide polymorphisms in the AVP locus and assessed the relationship between AVP gene variants, copeptin, and clinical outcomes. They found that baseline plasma copeptin concentration was associated with lower insulin sensitivity as well as increased incidence of impaired fasting glucose or Type 2 diabetes mellitus (IFG/T2DM) during follow-up. Moreover, they also reported that in men, genotypes associated with higher circulating copeptin also showed a higher cumulative incidence of IFG/T2DM. This pattern of Mendelian randomization, though not a conclusive demonstration of causality, represents the first suggestion of a causal relationship between vasopressin and IFG/T2DM.

Moreover, there also exists first evidence that normal, healthy adults can lower circulating plasma copeptin over the medium-term by increasing water intake. In a stratified sample of young French adults with habitual low, medium, and high fluid intake, we found that baseline plasma copeptin was associated with 24-h U_Osm (unpublished observations); then, in the low- and medium-drinkers, a 6-week water intake intervention resulted in a significant decrease of circulating plasma copeptin. While this was a secondary analysis, and although it contained certain experimental limitations, this study demonstrated that plasma copeptin can be modified through increased water intake, even in young, healthy individuals with normal baseline copeptin levels.

The existing literature allows a working model of fluid intake, hydration, and health to be put into place. However, the level of evidence for each link in the model is uneven, and many research questions remain insufficiently answered. With regard to plasma copeptin, fluid intake, and hydration, we must quantify the relationships between urinary hydration biomarkers and copeptin, and quantify changes in copeptin in response to increased water intake in a broader range of subjects. The proposed threshold of 500 mOsm·kg^-1 U_Osm requires more research to support, build upon, or update this target. With respect to copeptin and disease, thresholds for copeptin that are predictive of disease in men and women are needed; then, it will be necessary to demonstrate that increasing water intake to lower copeptin also impacts risk markers for disease. Finally, there is a paucity of research regarding ways that hydration and copeptin may impact specific segments of the population. In pregnant women, limited research suggests that plasma copeptin as early as the first trimester may be predictive of preeclampsia [71, 72]; in children, research regarding water intake, hydration, and vasopressin is virtually nonexistent outside of enuresis. Today, the field is wide open and, for young researchers in particular, represents a huge opportunity for novel research.

E.T.P. is a full-time employee of Danone Research.