We performed a pooled analysis of four independently conducted studies. All four studies were conducted according to the Declaration of Helsinki and were approved by the appropriate ethics committees, and all subjects provided informed consent. Together, the four studies reflected a variety of study settings, interventions (increasing daily TFI in low drinkers, decreasing daily TFI in high drinkers, or maintaining daily TFI habits), and drinking habits representative of typical North American and Western European consumption habits (Table 1). The pooled database included 238 participants, 50 % male, with a broad representation of daily TFI (5th–95th percentile: 0.37–3.22 L/d) and corresponding 24-h urine samples.
Participant lifestyle habits and relevant inclusion criteria
Lifestyle habits and relevant inclusion or non-inclusion criteria, which may impact fluid balance and homeostasis, were reviewed prior to analysis. In all studies, participants were sedentary or participated in only light-to-moderate physical activity. In the case in which some physical activity was permitted , exercise logs were kept to ensure that physical activity was within the prescribed limits (actual physical activity reported: 1 ± 1 sessions during the intervention with an average duration of 25 ± 18 min/session). Caffeine intake of ≤500 mg/day was permitted in 3 of the 4 studies. This level of daily caffeine intake has been shown, in free-living young adults, to cause no dehydration or measurable differences in hydration indices . In all studies, exclusionary criteria included metabolic or gastrointestinal disease (acute or chronic), renal, hepatic, or cardiac failure, and any drug or concomitant medication that may interfere with renal function or water balance.
Total fluid intake
In all four studies, intake was recorded throughout each study day. Daily TFI was defined as the total volume of drinking water plus other beverages. In studies A and C, fluid intake data were collected via an online food and fluid intake diary, which participants filled out daily (NutriSaas-24WQ-waters; MXS, France) and which included questions specific to fluid intake both during and between meals. Study B was conducted using an inpatient setting, meaning that food and water consumption was controlled, monitored, and recorded by the study staff. In Study D, subjects filled out daily paper food and fluid diaries, which were double-checked for completeness via interviews with trained staff each morning and were analysed using commercial nutrition software (Nutritionist Pro, Axxya Systems, Redmond, WA).
Twenty-four-hour urinary variables
In studies A, B, and C, 24-h urine volume (U
Vol) was measured from urine mass (to the nearest gram) and corrected for density using specific gravity. In study D, urine volume was measured from urine mass alone.
In studies A, B, and C, 24-h urine specific gravity (U
SG) was measured using a commercially available digital hand-held refractometer (ATAGO Pen Urine-SG, Atago Corp., Japan). In study D, U
SG was measured using a clinical light refractometer (ATAGO A300CL, Atago Corp., Japan).
All four studies evaluated 24-h U
Col using the 8-shade urine colour scale published by Armstrong et al. . In all four studies, a transparent urine collection container was placed against a plain white background in a well-lit room. The colour of the sample was compared against the colour scale, and the number corresponding to the closest shade (1 = very pale; 8 = very dark) was recorded.
Data recombination and analysis
Extractions of demographic, TFI, and urine variables were performed on each of the study databases; variable names were recoded, and the extractions were merged into a common database. For each participant, between-day change scores were calculated for TFI, U
SG, and U
Col. In studies A, B, and D, where variables were collected over consecutive days, change scores reflect the changes in TFI and urinary parameters between consecutive 24-h periods. In study C, variables were collected at baseline and upon completion of the intervention, and thus change scores reflect the change between baseline and intervention periods. Because the database included studies where participants were asked to increase, decrease, or maintain their daily fluid intake, data were then grouped by the directionality of the U
Col change score (i.e. U
Col became lighter; U
Col became darker; or no change in U
Col). These three data groups were then analysed separately. For each data group, the respective changes in TFI, U
Vol, and U
SG that were associated with a change in U
Col were calculated.
To determine the change in daily fluid intake volume needed to adjust U
Col by 2 shades (Aim 1), two separate approaches were used to evaluate the change in TFI required to modify U
Col by 2 shades. In the first approach, all data points displaying the same change in U
Col were grouped, and the mean change in TFI, U
Vol, and U
SG that was required to modify U
Col by ± 1, 2, 3, or 4 shades was calculated. In the second approach, change in U
Col was recoded as a binomial variable (0: change in U
Col < 2 shades; 1: change in U
Col ≥ 2 shades) and receiver operating characteristic (ROC) analysis was used to determine the ‘optimal’ change in TFI associated with a change of 2 or more U
To evaluate whether starting U
Col altered the relationship between the change in fluid intake and change in U
Col (Aim 2), it was also necessary to consider two cases (U
Col becoming lighter; U
Col becoming darker) separately. This separation was necessary because subjects with 24-h U
Col that was already relatively dark (shades 5 or 6 on the 8-colour scale) would be unlikely to exhibit even darker 24-h urine the following day, while subjects with very pale urine (1 or 2) would not be able to lighten their U
Col by 2 or more shades on the 8-point scale. Thus, for data points where U
Col became lighter, we compared change in TFI between urine starting at a moderate colour (shades 3 or 4), to urine starting at a darker colour (shades 5 or 6). Urine samples starting at very dark shades (7 or 8) could not be assessed since there were only four samples that fell in this colour range. For data points where U
Col became darker, we compared the change in TFI between urine starting at a very pale colour (shades 1 or 2), to urine starting at a moderate colour (shades 3 or 4).