Water intake and progression of chronic kidney disease: the CKD-REIN cohort study

Abstract

Background

Optimal daily water intake to prevent chronic kidney disease (CKD) progression is unknown. Taking the kidney’s urine-concentrating ability into account, we studied the relation of kidney outcomes in patients with CKD to total and plain water intake and urine volume.

Methods

Including 1265 CKD patients [median age 69 years; mean estimated glomerular filtration rate (eGFR) 32 mL/min/1.73 m²] from the Chronic Kidney Disease–Renal Epidemiology and Information Network cohort (2013–19), we assessed fluid intake at baseline interviews, collected 24-h urine volumes and estimated urine osmolarity (eU_osm). Using Cox and then linear mixed models, we estimated hazard ratios (HRs) and 95% confidence intervals (CIs) for kidney failure and eGFR decline associated with hydration markers, adjusting for CKD progression risk factors and eU_osm.

Results

Patients’ median daily intake was 2.0 L [interquartile range (IQR) 1.6–2.6] for total water and 1.5 L (1–1.7) for plain water, median urine volume was 1.9 L/24 h (IQR 1.6–2.4) and mean eU_osm was 374 ± 104 mosm/L. Neither total water intake nor urine volume was associated with either kidney outcome. Kidney failure risk increased significantly with decreasing eU_osm ˂292 mosm/L. Adjusted HRs (95% CIs) for kidney failure associated with plain water intake were 1.88 (1.02–3.47), 1.59 (1.06–2.38), 1.76 (0.95–3.24) and 1.55 (1.03–2.32) in patients drinking <0.5, 0.5–1.0, 1.5–2.0 and >2.0 L/day compared with those drinking 1.0–1.5  L/day. High plain water intake was also significantly associated with faster eGFR decline.

Conclusions

In patients with CKD, the relation between plain water intake and progression to kidney failure appears to be U-shaped. Both low and high intake may not be beneficial in CKD.

GRAPHICAL ABSTRACT

Open in new tabDownload slide

INTRODUCTION

Hydration is an important aspect of nutrition that should receive more attention. The 2010 European Food Safety Authority guidelines recommend a total water intake of 2.5 L/day for men and 2 L/day for women [1]. The kidneys play a central role in controlling water balance, a function that is impaired in chronic kidney disease (CKD), which affects 10–15% of the population worldwide [2]. International guidelines on CKD management, however, do not include recommendations about fluid intake [3, 4]. The National Kidney Foundation suggests a total daily water intake of 3 L for men and 2.2 for women with CKD, while French health authorities recommend 1.5 L, to be adjusted to thirst and excretion [5], but the evidence supporting these recommendations is poor.

In the general population, higher water intake has been associated with a lower prevalence of CKD and slower kidney function decline [6–8]. Studies among CKD patients have yielded conflicting results: two showed worse kidney outcomes with higher 24-h urine osmolarity (presumably lower fluid intake) [9, 10] and three with lower osmolality (presumably higher fluid intake) [11, 12]. The Modification of Diet in Renal Disease trial found high urine volume (>2.85 L/day) and low urine osmolality to be associated with faster CKD progression, suggesting that ‘pushing fluids’ may not be appropriate in patients with CKD [13]. Finally, a randomized clinical trial found that coaching patients with moderate CKD to increase water intake did not have any beneficial effect on kidney function [14]. Thus, while population-based studies tend to support the recommendation of high water intake to prevent CKD onset, findings from CKD cohorts based on urine volume or osmolality are inconclusive and no study has assessed water intake in this population.

We therefore studied the relation of water intake, evaluated from both input, including total and plain daily water intake, and output, i.e. urine volume, to CKD progression to kidney failure in nephrology patients. We also assessed the effect of decreasing urine-concentrating ability related to kidney function decline on this relation.

MATERIALS AND METHODS

Study design and population

The Chronic Kidney Disease–Renal Epidemiology and Information Network (CKD-REIN) is a prospective cohort study that enrolled 3033 outpatients with CKD from 40 nephrology clinics nationally representative geographically and for legal status (i.e. public or private). Details of the study protocol have been published elsewhere [15, 16]. In brief, during a census phase (2013–15) we identified all eligible patients, i.e. adult patients ≥18 years and with an estimated glomerular filtration rate (eGFR) of 15–60 mL/min/1.73 m² measured twice with no prior chronic dialysis or kidney transplantation. They were enrolled during their next nephrology visit a few weeks later (2013–16), regardless of whether their eGFR was still within 15–60 mL/min/1.73 m², and followed up annually thereafter. The study protocol was approved by the Institut national de la santé et de la recherche médicale institutional review board (IRB00003888) and the ethics committee (CCTIRS12.360/CPP). All patients provided written informed consent to participate.

Data collection

Clinical data, including CKD history and comorbidities, were collected by trained clinical research associates from medical records. Sociodemographic data and smoking habits were recorded either in interviews or by self-administered questionnaires, as was detailed information about medication use and dietician visits during the past year. Physical activity was assessed through the 16-item Global Physical Activity Questionnaire and was expressed in metabolic equivalents per minute. Blood pressure, height and weight were measured. All patients were prescribed a set of standard blood and urine tests as recommended by the French health authorities for routine CKD care, to be performed at their usual laboratory. Patients were considered to have hypertension if it was reported in their medical records or if they used antihypertensive medications and to have diabetes if it was reported in their records or if they used glucose-lowering medication or had haemoglobin A1c ≥6.5% or fasting glucose ≥7.0 or random glucose ≥11 mmol/L. eGFR was estimated with the Chronic Kidney Disease Epidemiology Collaboration equation [17]. Albuminuria or proteinuria was measured and classified according to the Kidney Disease: Improving Global Outcomes 2012 guideline categories: A1, normal (<3 mg/mmol); A2, moderately increased (3–30 mg/mmol); A3, severely increased (>30 mg/mmol) [16].

Assessment of water intake, urine volume and urine osmolarity

All but 3 of the 3033 cohort participants were interviewed about their intake of all types of beverages, including water, sodas, juices, hot beverages, soups, wine, beer and other alcoholic drinks. They were asked about the daily volume of plain water intake and the number of glasses or cups of each beverage they had drunk over the past week. We calculated total water intake from all beverages, which was analysed in five classes (<1.5, 1.5–2.0, 2.0–2.5, 2.5–3.0 and >3.0 L/day), as well as plain water intake, analysed in five categories (<0.5, 0.5–1.0, 1.0–1.5, 1.5–2.0 and >2.0 L/day), and other beverage intake. To take hydration needs related to weather conditions into account, we collected outdoor temperatures on the day of the interview from the Météo-France website (http://www.meteofrance.com/climat/meteo-date-passee).

Among cohort participants, 1265 had 24-h urine collected, in accordance with their nephrologist’s usual practice. Estimated urine osmolarity was calculated as the sum of the concentrations of major solutes measured in 24-h urine: eU_osm (mosm/L) = (sodium + potassium) × 2 + urea [18]. We also calculated osmolar excretion (mosm/24 h) as follows: osmolar excretion = eU_osm × urine volume.

Kidney outcomes

We studied two endpoints: eGFR slope over time (in mL/min/1.73 m²/year), based on all routine measurements collected over the study follow-up, and progression to kidney failure, defined as initiation of kidney replacement therapy, either dialysis or pre-emptive transplantation. Patients were followed up through January 2019. Only one patient was lost to follow-up; his observation was censored at the time of the last visit.

Statistical analyses

The main analysis is based on the 1265 patients with both water intake data and 24-h urine collection (Supplementary data, Figure S1). We first described their baseline characteristics overall and according to water intake categories, which were also described for all 3030 patients with water intake data. Second, after excluding 55 patients with CKD Stage 5, we described the means of urine volume, eU_osm and osmolar excretion according to plain water intake categories at baseline as well as mean water intake and urine volume according to CKD stage. We also analysed the crude association between baseline eU_osm and eGFR.

Third, we performed cause-specific Cox hazard models to estimate crude and adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for kidney failure associated with plain and total water intake categories, with 1.0–1.5 and 2.0–2.5 L/day, respectively, as the reference categories. All-cause mortality was considered a competing risk for kidney failure. We systematically adjusted for age, sex, body mass index (BMI) and albuminuria, as well as for selected covariates with P < 0.20 in crude analyses, and then adjusted for eU_osm. We stratified for nephropathy type, which did not meet the proportional hazards assumption, and used a time-dependent coefficient as well as a restricted cubic spline for eGFR, because it did not satisfy either the proportional hazards or log-linearity assumptions.

Fourth, we used Cox models with penalized splines to estimate HRs for kidney failure associated with urine volume and eU_osm, both treated as continuous variables, and adjusted for covariates selected as above.

Fifth, we used linear mixed models with a random intercept for patient identification and a random slope for time since baseline to study the crude and adjusted associations of plain and total water intake in five categories and urine volume and eU_osm in tertiles, with mean eGFR slope in mL/min/1.73 m²/year. Covariates included in the adjusted models were the same as those for the above Cox models. All models were run with the lme4 package [19] and overall P-values were calculated by an aggregated likelihood-ratio test following Meng and Rubin [20], available in the mitml package [21]. All continuous adjustment variables were standardized to facilitate model convergence.

Finally, in a sensitivity analysis we computed the linear mixed models and the Cox models to estimate eGFR decline and the HRs of kidney failure with total and plain water in the 3030 patients with available data, before and after adjusting for the same confounders as used in the above models except eU_osm.

Missing data were addressed by multiple imputations using all covariates and outcomes (n = 25 imputed datasets, with fully conditional specification, maximum 30 iterations), with the MICE package [22]. All statistical analyses were performed on each imputed dataset and all estimates and P-values were calculated by pooling across imputed datasets following Rubin’s rules [23].

All statistical analyses were performed with R version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria) [24]. The data underlying this article will be shared on reasonable request to the corresponding author.

RESULTS

Baseline characteristics

Baseline characteristics in all patients and the study population with eU_osm were broadly similar, particularly for demographics (median age 69 years, mostly men), lifestyle, median eGFR, comorbidities and fluid intake, but differed by region of residence (Supplementary data, Table S1).

Among the 1265 patients with eU_osm, the median total water intake was 2.0 L/day [interquartile range (IQR) 1.6–2.6], 1.5 L/day (IQR 1.0–1.7) for plain water and 0.6 L/day (IQR 0.4–0.9) for other beverages. Higher plain water intake was significantly associated with younger age, with men, a higher education level, living in southern France, more physical activity, more frequent dietician visits and fewer comorbidities (Table 1). As expected, lower plain water intake was associated with a higher intake of other beverages.

Cross-sectional associations between hydration markers and CKD stages

At baseline, patients had a median urine volume of 1.9 L/24 h (IQR 1.6–2.4), an average eU_osm of 374 ± 104 mosm/L and a median osmolar excretion of 691 mosm/24 h (IQR 554–861). Higher plain water intake was strongly associated with higher urine volume and osmolar excretion and with lower eU_osm (Figure 1; all P < 0.0001). Plain and total water intake and urine volume did not differ across CKD stages (Figure 2A–C; P > 0.3).

Association of estimated urine osmolarity with baseline eGFR, eGFR decline and kidney failure risk

Among the 1265 patients with eU_osm, the cross-sectional relation of eU_osm to eGFR, assessed with loess regression, was not linear; the association was positive above 369 ± 15 mosm/L and non-existent below that cut-off (Supplementary data, Figure S2). After excluding patients with CKD Stage 5 at baseline, the adjusted HR of kidney failure increased significantly as eU_osm decreased to ˂293 mosm/L (Supplementary data, Figure S3). No association was observed, however, between baseline eU_osm in tertiles and eGFR change over time, before or after adjustment (Supplementary data, Table S2).

Association of water intake with kidney failure risk and eGFR decline

During a median 3-year follow-up (IQR 2.1–3.5), 221 of the 1210 patients with CKD Stages 3 and 4 at baseline progressed to kidney failure. There was no significant association between total water intake and the risk of kidney failure (Figure 3B). In contrast, daily plain water intake >2.0 L was significantly associated with a higher crude HR of kidney failure [2.16 (95% CI 1.38–3.39)] than the reference daily intake of 1.0–1.5 L (Figure 3A). After adjustment for CKD progression risk factors and osmolar excretion reflecting osmolar load, the relation was U-shaped; further adjustment for eU_osm changed these associations only slightly, with HRs of 1.92 (95% CI 1.03–3.56), 1.69 (1.11–2.56), 1.17 (0.72–1.89) and 2.43 (1.49–3.97), for plain water intake <0.5, 0.5–1.0, 1.5–2.0 and >2 L, compared with the reference of 1.0–1.5 L, respectively.

Over a median 2.7-year follow-up (IQR 2.0–3.1), a median of 7 (IQR 5–12) eGFR values per patient were available for estimating eGFR slopes. Mean eGFR decline was 1.78 mL/min/1.73 m²/year (IQR 1.56–1.99). Linear mixed models showed statistically significant crude associations between annual eGFR change and both plain and total water intake at baseline (Tables 2 and 3). The association was no longer significant for total water intake after adjusting for the main confounders and eU_osm, but remained significant for plain water intake.

In the sensitivity analysis, based on the overall cohort population, the crude association between plain water intake and kidney failure risk was similar to that observed in the main study subgroup. However, although adjusting for confounders, except eU_osm, also resulted in a U-shaped relation, the HRs for kidney failure associated with the lowest and highest daily water intake were not statistically significant (Supplementary data, Figure S4).

Association of urine volume with kidney failure and eGFR decline

No association was observed between continuous urine volume and the HR for kidney failure (Figure 4). Nor was baseline urine volume divided in tertiles associated with eGFR change over time, either before or after adjustment for confounders (Supplementary data, Table S3).

DISCUSSION

This prospective cohort study in patients with moderate and advanced CKD showed a significant association between plain water intake and disease progression to kidney failure. This association was independent of risk factors for CKD progression and appeared non-linear, with a higher risk observed for both the lowest and highest water intakes. Low baseline eU_osm was associated with a higher risk of kidney failure but did not explain the relation observed with high water intake. To our knowledge, this is the first study to investigate the association of hydration status with kidney outcomes in patients with CKD based on direct assessment of water intake rather than from urine volume or osmolality.

Population-based studies have consistently showed that higher water intake is associated with a lower prevalence of CKD and slower eGFR decline [6–8]. In contrast, in patients with CKD, both lower and higher urine osmolality have been associated with disease progression [9–12]. The first study to suggest that high fluid intake may not slow kidney disease progression in humans was the Modification of Diet in Renal Disease (MDRD) study, which found that sustained high urine volume and low urinary osmolality were independent risk factors for faster eGFR decline [12]. Nevertheless, because kidney function decline can decrease the kidneys’ urine-concentrating ability and increase urine volume, it is difficult to ascertain whether the fluid intake assessed from these markers was a cause or a consequence of CKD progression. Our main analysis, based on CKD patients with both fluid intake assessments and 24-h urine collection, showed that the association between daily water intake and either kidney failure risk or eGFR slope was not linear, but rather was U-shaped after adjustment for multiple confounders and eU_osm.

This observation may explain previous conflicting results about the association of both low and high urinary osmolality with CKD progression. Our finding that only plain water intake, and not total water, was associated with disease progression is consistent with a previous population-based study [7] that suggested that plain water intake per se is important. In contrast with the MDRD study, however, we found no significant association between urine volume and CKD progression. This may be due to non-differential measurement error in urine volume estimates based on a single value at baseline in this study versus the average of monthly measurements in the MDRD study.

The kidneys’ water needs depend in part on their osmolar load, a factor rarely measured in previous studies. The osmolar load decreases with CKD progression, probably due to loss of appetite and diet restrictions [25]. In this study, the higher the osmolar excretion, the higher the water intake, consistent with the fact that heavy water drinkers are also heavy eaters of food generating higher osmolar excretion. Nevertheless, this excretion does not explain the association of high water intake with both kidney failure risk and eGFR decline, as shown in multivariable analyses adjusted for this potential confounder.

Hence, while high water intake is beneficial for preventing CKD onset [6, 7] and kidney stones [26, 27], the findings from our study reinforce those of the MDRD study [12] and the negative findings of the clinical trial [14] aimed at increasing water intake. All three studies show that higher water intake may not be helpful for patients with moderate or advanced CKD. Our study suggests that there may be an optimum range of daily water intake for CKD patients, ranging from 1 to 2 L/day.

The existence of a limit above which increased water intake in CKD patients may be harmful rather than beneficial can be understood better by observing the relations between eGFR and urine osmolarity. Few epidemiological studies have explored the association between eU_osm and later kidney outcomes, and they have had conflicting results. Some studies have shown that low fasting eU_osm is associated with a higher risk of either kidney failure or faster mGFR decline [11–12]. In contrast, other studies report an inverse relation between eU_osm and eGFR decline [9, 10]. Our results showed a non-linear relation between eU_osm and kidney failure and support the growing body of evidence that low osmolarity (<293 mosm/L in our study) is associated with kidney failure. Studies in normal conscious rats and a clinical investigation in healthy humans showed that the urine-concentrating mechanism (due to either low hydration or the infusion of a selective vasopressin V2 receptor agonist) induces glomerular hyperfiltration and kidney hypertrophy similar to those induced by high protein intake, and probably dependent on decreased intensity of the tubuloglomerular feedback [28, 29]. Importantly, this is observed only when the kidney produces hyperosmotic urine. It is not observed below a certain threshold of urine osmolarity [30]. Consistent with these observations, a recent epidemiologic study showed that the association between eGFR and urine osmolarity (evaluated by the specific gravity of the urine) in a large population is J-shaped [31]. Altogether, these findings suggest that the kidney is best off when excretion of urinary solutes does not require either the concentration or the dilution of urine, both of which probably impose a burden on the kidney.

Our study has several strengths. It is based on a large, prospective CKD cohort, representative of the patient population receiving nephrologist-led care in France. We collected extensive clinical, biological and environmental data that enabled adjustment for the principal potential confounders. Our detailed questionnaire about fluid intake enabled us to analyse plain and total water intake separately. Our measurements of hydration markers, including water intake, urine volume and eU_osm, resulted in consistent associations between them and reflect the internal validity of the study.

This study also has limitations. First, its observational design does not allow any causal inferences. Second, these analyses were based on a single assessment of beverage intake during the week before the patient’s interview, although this may vary over time and with outdoor temperature. Nevertheless, we adjusted for outdoor temperature. Protein intake and other food data were not available, but we were able to adjust for osmolar excretion, reflecting the load of ingested proteins and major electrolytes. Third, 24-h urine collection may not be accurate. However, the osmolarity range was within that found in other studies. Moreover, water intake estimates from patient interviews were consistent with measured urine volume. Finally, while our main analysis was based on patients with available 24-h urine collection, i.e. 42% of all CKD-REIN participants, sensitivity analyses in the 3030 with interview data showed a similar crude association between plain water intake and kidney failure risk. The adjusted analysis, however, failed to show significant HRs for the lowest and highest categories after adjustment, despite a trend towards higher HRs in both. The study protocol did not require 24-h urine collection, which was available for some but not all patients, depending on the nephrology practices in different regions. Indeed, region was the only factor that differed between the overall cohort and the main study population.

In conclusion, our study, based on a detailed assessment of water intake, suggests that the relation between plain water intake and CKD progression is not linear but U-shaped. This relation is independent of the decreased urine-concentrating ability resulting from kidney function decline. These findings may have important clinical implications for the management of CKD. Further investigations in other cohorts and clinical trials are needed to confirm the optimal level of water intake in CKD patients suggested by our results and define specific water recommendations tailored for non-dialysis CKD patients.

SUPPLEMENTARY DATA

Supplementary data are available at ndt online.

ACKNOWLEDGEMENTS

We acknowledge the CKD-REIN study coordination staff for their efforts in setting up the CKD-REIN cohort: Elodie Speyer, Reine Ketchemin and all the clinical research associates. We thank Jo Ann Cahn for editing the English version. We also thank the participating clinical sites, their investigators cited above and all the patients. Members of the CKD-REIN study group: Steering committee and coordination: Natalia Alencar de Pinho, Carole Ayav, Serge Briançon, Dorothée Cannet, Christian Combe, Denis Fouque, Luc Frimat, Yves-Edouard Herpe, Christian Jacquelinet, Maurice Laville, Ziad A Massy, Christophe Pascal, Bruce M Robinson, Roberto Pecoits-Filho, Bénédicte Stengel, Céline Lange, Karine Legrand, Sophie Liabeuf, Marie Metzger and Elodie Speyer. CKD-REIN investigators/collaborators: Thierry Hannedouche, Bruno Moulin, Sébastien Mailliez, Gaétan Lebrun, Eric Magnant, Gabriel Choukroun, Benjamin Deroure, Adeline Lacraz, Guy Lambrey, Jean Philippe Bourdenx, Marie Essig, Thierry Lobbedez, Raymond Azar, Hacène Sekhri, Mustafa Smati, Mohamed Jamali, Alexandre Klein, Michel Delahousse, Christian Combe, Séverine Martin, Isabelle Landru, Eric Thervet, Ziad A Massy, Philippe Lang, Xavier Belenfant, Pablo Urena, Carlos Vela, Luc Frimat, Dominique Chauveau, Viktor Panescu, Christian Noel, François Glowacki, Maxime Hoffmann, Maryvonne Hourmant, Dominique Besnier, Angelo Testa, François Kuentz, Philippe Zaoui, Charles Chazot, Laurent Juillard, Stéphane Burtey, Adrien Keller, Nassim Kamar, Denis Fouque and Maurice Laville.

FUNDING

The CKD-REIN study is funded by the Agence Nationale de la Recherche through the 2010 Cohortes-Investissements d’Avenir programme and by the 2010 national Programme Hospitalier de Recherche Clinique. CKD-REIN is also supported through a public–private partnership with Amgen, Fresenius Medical Care and GlaxoSmithKline since 2012; Otsuka Pharmaceutical since 2015; Baxter and Merck Sharp & Dohme-Chibret from 2012 to 2017; Sanofi-Genzyme from 2012 to 2015; Lilly France from 2013 to 2018; and Vifor Fresenius and AstraZeneca since 2018. Inserm Transfert set up and has managed this partnership since 2011. This project received a grant from the ISN-H4KH Initiative supported by an unrestricted grant from Danone Nutricia Research.

AUTHORS’ CONTRIBUTIONS

S.W., L.B., Z.M. and B.S. designed the study. T.M. performed the statistical analyses. S.W., L.B. and B.S. drafted the manuscript. All the authors were involved in the interpretation of the results, critical review of the manuscript and approved the manuscript.

CONFLICT OF INTEREST STATEMENT

S.W. received a grant from the ISN-H4KH Initiative for this project. The ISN-H4KH Initiative received an unrestricted grant from Danone Nutricia Research. Amgen, AstraZeneca, Fresenius Medical Care, Vifor Fresenius, GlaxoSmithKline and Otsuka contribute to the public–private partnership supporting the CKD-REIN. D.F. reports personal fees from Sanofi, Lilly, Fresenius Kabi and grants from Fresenius Medical Care outside the submitted work. Z.A.M. reports grants and other funding from Baxter, Outsuka, Daichi and Astellas, outside the submitted work. The remaining authors have nothing to disclose.

REFERENCES

Author notes