ABSTRACT
Fluid balance is pivotal in the management of children with chronic kidney disease (CKD) and on dialysis. Although many techniques are available to assess fluid status, there are only a few studies for children, of which none have been comparable against cardiovascular outcome measures.
We performed a longitudinal study in 30 children with CKD5-5D and 13 age-matched healthy controls (71 measurements) to determine a correlation between optimal weight by bioimpedance spectroscopy (Wt-BIS) and clinical assessment (Wt-CA). The accuracy of Wt-BIS [relative overhydration (Rel-OH)] was compared against indicators of fluid status and cardiovascular measures.
There was poor agreement between Wt-CA and Wt-BIS in children on dialysis (P = 0.01), but not in CKD5 or control subjects. We developed a modified chart to plot Rel-OH against systolic blood pressure (SBP) z-score for the appropriate representation of volume status and blood pressure (BP) in children. In total, 25% of measurements showed SBP >90th percentile but not with concurrent overhydration. Rel-OH correlated with peripheral pulse pressure (P = 0.03; R = 0.3), higher N-terminal pro-brain natriuretic peptide (P = 0.02; R = 0.33) and left ventricular end-diastolic diameter (P = 0.05; R = 0.38). Central aortic mean and pulse pressure significantly associated with the left ventricular end-diastolic diameter (P = 0.03; R = 0.47 and P = 0.01; R = 0.50, respectively), but not with Rel-OH. SBP was positively associated with pulse wave velocity z-score (P = 0.04). In total, 40% of children on haemodialysis and 30% on peritoneal dialysis had increased left ventricular mass index.
BIS provides an objective method for the assessment of hydration status in children on dialysis. We noted a marked discrepancy between BP and hydration status in children on dialysis that warrants further investigation.
INTRODUCTION
Assessing fluid balance and achieving the optimal weight in children with chronic kidney disease (CKD) and on dialysis can be challenging. Fluid overload plays a causal role in the development of left ventricular hypertrophy (LVH) and contributes to overall cardiovascular morbidity and mortality [1, 2].
On the other hand, underestimating the optimal weight carries a risk of excessive ultrafiltration on dialysis leading to accelerated loss of residual renal function [3, 4] and myocardial stunning [5]. Optimal weight has been defined as the lowest weight a patient can tolerate without developing hypotension and not in apparent fluid overload [6].
Optimal weight or ‘dry weight’ has been traditionally determined by clinical assessment, taking into consideration the presence of clinical signs of volume overload, blood pressure (BP) trends and changes in weight. Clinical assessment of optimal weight is subjective, requiring experience and careful assessment. This can be particularly challenging in children on dialysis in whom optimal weight changes with growth and BP may be affected by the underlying renal or renovascular disease per se and not be a manifestation of volume overload [7–9]. Various methods to assess fluid status have been studied including bioimpedance spectroscopy (BIS) [10], measurement of circulating natriuretic peptide [11], lung ultrasound for extravascular lung water or lung B lines [12], inferior vena cava diameter and its collapsing index [13] and heavy water dilution studies in children and adults [14, 15]. The BIS particularly has been widely used in adult dialysis patients but there are few data in children on dialysis. Furthermore, reference levels among healthy children are not available. In fact, the association between relative overhydration (Rel-OH) against BP has been based on adult normograms. Thus, it is not known if BIS provides a true reflection of volume status in children, and no ‘gold standard’ is available for comparison.
Our aim was to identify the optimal technique for assessing hydration status in children with CKD and on dialysis, and to determine its accuracy by comparing against cardiovascular measures of fluid status. In this study, we evaluated: (i) the correlation between optimal weight by clinical assessment (Wt-CA) and optimal weight by BIS (Wt-BIS); (ii) the accuracy of Rel-OH against other indicators of fluid status including peripheral BP, central BP, N-terminal pro-brain natriuretic peptide (NT-proBNP) levels and serum sodium (se Na); and (iii) the association between Rel-OH with cardiovascular outcome indicators, including pulse wave velocity (PWV) z-score adjusted for age and increased left ventricular mass index (LVMI).
MATERIALS AND METHODS
Study design and study participants
This was a prospective cross-sectional study prevalent in CKD Stage 5 and maintenance dialysis (haemodialysis, HD and peritoneal dialysis, PD) patients between 5 and 18 years of age. We recruited all children who fulfilled the inclusion criteria from our CKD and dialysis clinics at Great Ormond Street Hospital for Children, London, from February 2016 to June 2016. Age-matched healthy controls were recruited from an international school and studied for comparison. Children with arrhythmias or in situ cardiac conduction devices were excluded as these interfere with the BIS device. Children had an average of two measurements at different times at least 4 weeks apart. The HD group had the measurements taken before a mid-week HD session. The PD and CKD5 group had their measurements during clinic review. In the PD group, any indwelling dialysate, if present, was subtracted from measured weight to give the patient’s actual weight.
At each assessment period the attending paediatric nephrologist or senior dialysis nurse (who was blinded to the outcome of BIS assessment) estimated the child’s optimal weight (Wt-CA) from their measured weight and BP. We then measured the height and weight, peripheral BP, whole body BIS, applanation tonometry (to determine central BP and PWV) and collected a blood sample for NT-proBNP alongside routine clinical bloods. Throughout the study period, patients had one echocardiographic assessment timed during mid-week session of dialysis with an interdialytic interval of 48 h. The healthy controls had all the above measurements except blood sampling and echocardiography assessment. The study was approved by the local research ethics committee. Consent was obtained from parents/carers and assent from the child if deemed competent.
Whole body bioimpedance spectroscopy
BIS was performed using the BCM® (Fresenius Medical Care, GmbH, Bad Homburg, Germany) device, which operates on the basis of conductance and reactance properties of different body tissues. This multi-frequency bioimpedance analysis measures 50 frequencies between 5 and 1000 kHz. Using the Hanai model, extracellular, intracellular and total body water is calculated from resistance obtained from a Cole–Cole plot [16, 17]. This is shown to be highly reproducible in adults [18].
BIS electrodes were applied on the non-fistula arm with the child in semi-recumbent position. Parameters generated directly from the device namely absolute overhydration (OH, L) and relative overhydration (Rel-OH, %) were recorded. The OH was then subtracted from the patient’s measured weight from the scale to derive a proposed optimal weight (Wt-BIS). Rel-OH is hydration normalized to extracellular water. Normality is defined within a range of −7 to 7%, which corresponds to the 10th and 90th percentiles of a reference healthy adult population. OH is defined as Rel-OH ≥7% and severe OH as Rel-OH ≥15% [19].
In children, BP limits differ with age, gender and height. Normal BP is defined as systolic BP (SBP) and diastolic BP (DBP) that is <90th percentile for sex, age and height, corresponding to a z-score of 1.28 [20]. Although hypertension is defined as a systolic BP and/or diastolic BP that is greater than or equal to the 95th percentile for sex, age and height, it is recommended that in children with CKD, BP-lowering agent should be initiated when the BP is consistently >90th centile [21, 22]. In the original hydration reference chart utilized with the BCM®, Rel-OH is plotted against SBP in a hydration reference chart which assumes 100 mmHg and 140 mmHg as the threshold for hypo- and hypertension. This is clearly not appropriate for children. We have adapted the hydration reference chart by plotting the Rel-OH against SBP z-score. The Rel-OH measurements were also plotted against the corresponding SBP on the original hydration reference chart to highlight the inaccuracy in assessment using the adult-based chart for children (Supplementary data, Figure S1).
Percentage difference between Wt-BIS and Wt-CA. A one-way Welch ANOVA was conducted to identify if the percentage difference was different for the various groups. The percentage difference was significantly different between the groups, Welch’s F (3, 32.64) = 18.53, P = 0.0005. Games-Howell post hoc analysis revealed that the higher percentage difference in HD compared with control group was statistically significant (2.72, 95% CI: 1.48–3.97; P = 0.0005). Similar observation was noted for the PD group (3.93, 95% CI: 1.73–6.12; P = 0.0005).
BP measurement
Peripheral BP
Peripheral BP was measured by auscultation method using an aneroid sphygmomanometer. An appropriate-sized cuff was applied after the patient had rested for at least 5 min. BP was recorded prior to a mid-week haemodialysis session in all HD patients. Peripheral pulse pressure was calculated as the difference between the measured SBPs and DBPs. One-quarter of the 12 children who required antihypertensive medications had SBP >90th centile for age, height and sex.
Central aortic BP
Pulse wave analysis was performed using the SphygmoCor Arterial Waveform Analysis System (AtCor Medical, Sydney, Australia). A 0.5-mm sensor was used to detect the radial waveform, which was then used to reconstruct the aortic pulse waveform using a validated mathematical transfer function from 20 sequential radial waveforms acquired. Recorded parameters were aortic systolic pressure, aortic diastolic pressure and aortic pulse pressure. The central aortic pulse pressure reflects the LV afterload; it is dependent on the LV filling pressure and is a marker of diastolic dysfunction. Using the same device, carotid-radial PWV was determined. PWV reflects the speed of the pulse originated from the heart as it circulates through the peripheral vasculature. In a stiff vessel, the travel time is faster; hence, a higher PWV will be recorded. The PWV is then expressed as a z-score of the value after adjusted for age [23, 24].
All applanation tonometry and BIS measurements were performed by a single operator who was blinded to the clinical assessment of weight.
Echocardiography
All 20 children on HD and PD and 7 out of 10 CKD5 children had echocardiography (ECHO) assessment. ECHO was performed by the in-centre paediatric echocardiographer using a 2D M-mode echocardiography. The ECHO parameters recorded were interventricular septal thickness at end-diastolic, left ventricular end-diastolic diameter (LVEDd) and LV posterior wall thickness at diastole. These indices were then utilized to calculate LV mass indexed to height [27] using the formula by Devereux et al. [25]. LVH is defined as LV mass indexed to height [27] more than the 95th percentile for gender and chronological age referenced from a large cohort of healthy children [26].
N-terminal pro B-type natriuretic peptide
NT-proBNP levels rise with increase in stress from stretched LV walls. It is widely used as a biomarker of circulatory congestion [27, 28] and decompensated heart failure [29, 30]. Serum NT-proBNP levels were measured with the Siemens (DPC) Immulite® 2000 systems, which is based on the principle of chemiluminescence immunoassay.
Statistical methodology
Demographic parameters were expressed as z-scores wherever possible. We described data as mean (standard deviation, SD) or median (interquartile range) as appropriate and categorical variables in percentages. Tests for normality were conducted with the Shapiro–Wilk test. Comparisons between groups were performed with either one-way analysis of variance (ANOVA) or Kruskal–Wallis H Test followed by Dunn’s procedure with a Bonferroni correction for multiple comparisons post hoc whenever appropriate. Intra-rater reliability for measurements from the BCM® and the SphygmoCor were measured and a good level of consistency accepted as Cronbach’s alpha >0.7. The level of agreement (LoA) between Wt-CA and Wt-BIS and the 95% confidence interval (95% CI) for the respective upper and lower limit was calculated in Bland–Altman analysis. Association between Rel-OH and other study parameters was determined with either Pearson’s or Spearman’s correlation co-efficient, R wherever appropriate for continuous data and chi-square test for association between categorical data. Level of significance was set at P < 0.05. All statistical analysis was performed using the SPSS software, version 23.0 (SPSS Inc., Chicago, IL, USA).
RESULTS
Demographics of the study population
Forty-three children were recruited into four study groups: 10 each in the HD, PD and CKD5 groups, and 13 age-matched healthy children served as control. A total of 71 sets of measurements were recorded: 46 from dialysis patients (25 from HD and 21 from PD), 12 from the CKD5 group and 13 from the healthy controls. Demographics of the study population are described in Table 1. All patient groups had a lower weight z-score compared with healthy controls. The age, height and body mass index z-scores were comparable across groups. A total of 50% of children on HD and 40% on PD had residual renal function. None of the patients had nephrotic syndrome or nephrotic range proteinuria that could have caused oedema and fluid overload. As a unit policy, the dialysate sodium in children on haemodialysis was in the range of 136–138 mEq/L.
Clinical characteristics of study and control groups
| HD | PD | CKD5 | Control | Pa | |
|---|---|---|---|---|---|
| Total | 10 | 10 | 10 | 13 | |
| Male gender | 7 (70.0) | 5 (50.0) | 6 (60.0) | 5 (38.5) | |
| Mean age (years) | 12.7 ± 3.71 | 12.3 ± 3.96 | 11.4 ± 3.79 | 11.4 ± 4.49 | 0.81 |
| Median z-score of weight for age and sex | −1.61 (−2.26 to − 0.01) | −1.57 (−2.30 to − 0.69) | −0.14 (−0.80 to 0.11) | −0.05 (−0.72 to 0.80) | 0.03 |
| Median z-score of height for age and sex | −1.7 (−2.7 to − 0.8) | −1.1 (−1.9 to − 0.5) | −0.9 (−1.9 to − 0.5) | 0.1 (−0.7 to 1.0) | 0.21 |
| Median z-score BMI | −1.3 (−2.3 to 0.8) | −0.8 (−1.5 to 0.7) | 0.1 (−0.1 to 1.0) | 0.1 (−0.7 to 0.7) | 0.47 |
| Median duration of renal replacement therapy (months) | 27.8 (2.7 to 65.0) | 6.0 (3.3 to 18.5) | NA | NA | NA |
| Number of patients on antihypertensive treatment | 3 (30) | 5 (50) | 4 (40) | 0 | NA |
| Median number of antihypertensive agents | 0 | 1 | 0 | 0 | NA |
| Number of patients on ACEI/ARB | 2 (20) | 1 (10) | 0 | NA | NA |
| Number of patients with arteriovenous fistula | 4 (40) | 0 | 1 (10) | NA | NA |
| Percentage with residual renal function | 50 | 40 | |||
| Aetiology of underlying CKD | |||||
| Congenital anomalies of the kidneys and urinary tract | 2 | 6 | 7 | ||
| Cystic kidney disease | 1 | 1 | |||
| Primary glomerular disease | 3 | 1 | |||
| Lupus nephritis | 1 | NA | NA | ||
| Others (HIVAN, Wilm's tumour, metabolic disease, undefined genetic syndrome) | 3 | 3 | 2 |
| HD | PD | CKD5 | Control | Pa | |
|---|---|---|---|---|---|
| Total | 10 | 10 | 10 | 13 | |
| Male gender | 7 (70.0) | 5 (50.0) | 6 (60.0) | 5 (38.5) | |
| Mean age (years) | 12.7 ± 3.71 | 12.3 ± 3.96 | 11.4 ± 3.79 | 11.4 ± 4.49 | 0.81 |
| Median z-score of weight for age and sex | −1.61 (−2.26 to − 0.01) | −1.57 (−2.30 to − 0.69) | −0.14 (−0.80 to 0.11) | −0.05 (−0.72 to 0.80) | 0.03 |
| Median z-score of height for age and sex | −1.7 (−2.7 to − 0.8) | −1.1 (−1.9 to − 0.5) | −0.9 (−1.9 to − 0.5) | 0.1 (−0.7 to 1.0) | 0.21 |
| Median z-score BMI | −1.3 (−2.3 to 0.8) | −0.8 (−1.5 to 0.7) | 0.1 (−0.1 to 1.0) | 0.1 (−0.7 to 0.7) | 0.47 |
| Median duration of renal replacement therapy (months) | 27.8 (2.7 to 65.0) | 6.0 (3.3 to 18.5) | NA | NA | NA |
| Number of patients on antihypertensive treatment | 3 (30) | 5 (50) | 4 (40) | 0 | NA |
| Median number of antihypertensive agents | 0 | 1 | 0 | 0 | NA |
| Number of patients on ACEI/ARB | 2 (20) | 1 (10) | 0 | NA | NA |
| Number of patients with arteriovenous fistula | 4 (40) | 0 | 1 (10) | NA | NA |
| Percentage with residual renal function | 50 | 40 | |||
| Aetiology of underlying CKD | |||||
| Congenital anomalies of the kidneys and urinary tract | 2 | 6 | 7 | ||
| Cystic kidney disease | 1 | 1 | |||
| Primary glomerular disease | 3 | 1 | |||
| Lupus nephritis | 1 | NA | NA | ||
| Others (HIVAN, Wilm's tumour, metabolic disease, undefined genetic syndrome) | 3 | 3 | 2 |
Data expressed as n (%) or median and interquartile range.
Comparison between groups.
ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; IQR, interquartile range; BMI, body mass index; HIVAN, human immunodeficiency virus-associated nephropathy; NA, not applicable.
Clinical characteristics of study and control groups
| HD | PD | CKD5 | Control | Pa | |
|---|---|---|---|---|---|
| Total | 10 | 10 | 10 | 13 | |
| Male gender | 7 (70.0) | 5 (50.0) | 6 (60.0) | 5 (38.5) | |
| Mean age (years) | 12.7 ± 3.71 | 12.3 ± 3.96 | 11.4 ± 3.79 | 11.4 ± 4.49 | 0.81 |
| Median z-score of weight for age and sex | −1.61 (−2.26 to − 0.01) | −1.57 (−2.30 to − 0.69) | −0.14 (−0.80 to 0.11) | −0.05 (−0.72 to 0.80) | 0.03 |
| Median z-score of height for age and sex | −1.7 (−2.7 to − 0.8) | −1.1 (−1.9 to − 0.5) | −0.9 (−1.9 to − 0.5) | 0.1 (−0.7 to 1.0) | 0.21 |
| Median z-score BMI | −1.3 (−2.3 to 0.8) | −0.8 (−1.5 to 0.7) | 0.1 (−0.1 to 1.0) | 0.1 (−0.7 to 0.7) | 0.47 |
| Median duration of renal replacement therapy (months) | 27.8 (2.7 to 65.0) | 6.0 (3.3 to 18.5) | NA | NA | NA |
| Number of patients on antihypertensive treatment | 3 (30) | 5 (50) | 4 (40) | 0 | NA |
| Median number of antihypertensive agents | 0 | 1 | 0 | 0 | NA |
| Number of patients on ACEI/ARB | 2 (20) | 1 (10) | 0 | NA | NA |
| Number of patients with arteriovenous fistula | 4 (40) | 0 | 1 (10) | NA | NA |
| Percentage with residual renal function | 50 | 40 | |||
| Aetiology of underlying CKD | |||||
| Congenital anomalies of the kidneys and urinary tract | 2 | 6 | 7 | ||
| Cystic kidney disease | 1 | 1 | |||
| Primary glomerular disease | 3 | 1 | |||
| Lupus nephritis | 1 | NA | NA | ||
| Others (HIVAN, Wilm's tumour, metabolic disease, undefined genetic syndrome) | 3 | 3 | 2 |
| HD | PD | CKD5 | Control | Pa | |
|---|---|---|---|---|---|
| Total | 10 | 10 | 10 | 13 | |
| Male gender | 7 (70.0) | 5 (50.0) | 6 (60.0) | 5 (38.5) | |
| Mean age (years) | 12.7 ± 3.71 | 12.3 ± 3.96 | 11.4 ± 3.79 | 11.4 ± 4.49 | 0.81 |
| Median z-score of weight for age and sex | −1.61 (−2.26 to − 0.01) | −1.57 (−2.30 to − 0.69) | −0.14 (−0.80 to 0.11) | −0.05 (−0.72 to 0.80) | 0.03 |
| Median z-score of height for age and sex | −1.7 (−2.7 to − 0.8) | −1.1 (−1.9 to − 0.5) | −0.9 (−1.9 to − 0.5) | 0.1 (−0.7 to 1.0) | 0.21 |
| Median z-score BMI | −1.3 (−2.3 to 0.8) | −0.8 (−1.5 to 0.7) | 0.1 (−0.1 to 1.0) | 0.1 (−0.7 to 0.7) | 0.47 |
| Median duration of renal replacement therapy (months) | 27.8 (2.7 to 65.0) | 6.0 (3.3 to 18.5) | NA | NA | NA |
| Number of patients on antihypertensive treatment | 3 (30) | 5 (50) | 4 (40) | 0 | NA |
| Median number of antihypertensive agents | 0 | 1 | 0 | 0 | NA |
| Number of patients on ACEI/ARB | 2 (20) | 1 (10) | 0 | NA | NA |
| Number of patients with arteriovenous fistula | 4 (40) | 0 | 1 (10) | NA | NA |
| Percentage with residual renal function | 50 | 40 | |||
| Aetiology of underlying CKD | |||||
| Congenital anomalies of the kidneys and urinary tract | 2 | 6 | 7 | ||
| Cystic kidney disease | 1 | 1 | |||
| Primary glomerular disease | 3 | 1 | |||
| Lupus nephritis | 1 | NA | NA | ||
| Others (HIVAN, Wilm's tumour, metabolic disease, undefined genetic syndrome) | 3 | 3 | 2 |
Data expressed as n (%) or median and interquartile range.
Comparison between groups.
ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; IQR, interquartile range; BMI, body mass index; HIVAN, human immunodeficiency virus-associated nephropathy; NA, not applicable.
Comparison of ‘optimal weight’: Wt-CA against Wt-BIS
At the start of their dialysis, mean excess weight (difference between physical weight and given optimal weight assessed clinically as a percentage of body weight) for patients in the HD and PD groups were 2.3% (95% CI: 1.23–3.35) and 2.0% (95% CI: 0.59–3.30), respectively. In children on PD and HD, Wt-CA was lower than Wt-BIS (Figure 1). Significantly higher percentage differences between Wt-BIS and Wt-CA were observed in the HD and PD groups compared with the control group: 2.5% and 2.7%, respectively, P = 0.01 each. The median difference between weights in the CKD5 group was 0.4%, P = 0.32, and comparable to the control group in which there was no difference between Wt-BIS and Wt-CA.
Based on the Bland–Altman analyses, the LoA for the different groups of children is shown (Figure 2). Mean difference of weight (Wt-CA and Wt-BIS) for the HD and PD groups was −0.52 (95% CI: −0.96 to −0.05) and −0.96 (95% CI: −1.46 to −0.46), respectively. In the CKD5 group, the LoA was excellent (−0.06; 95% CI: −0.43 to 0.30) and comparable to the healthy controls (−0.04; 95% CI: −0.15 to 0.07). Forty-six measurements were performed in the 20 children. Of these, 5 measurements in 3 children showed OH, 16 measurements in 9 children showed underhydration and 25 in 14 children showed normal hydration. The repeated measurements from the BCM® had high level of consistency, Cronbach’s alpha 0.96.
Bland–Altman plots for LoA between Wt-CA and Wt-BIS in the (A) HD group, (B) PD group, (C) CKD5 group and (D) healthy control group. The average weight from both the methods were plotted on the x-axis and the differences (Wt-BIS and Wt-CA) were plotted on the y-axis. Dotted lines represent the mean while the solid black lines represent ±2 SDs, respectively.
Comparison of hydration status and BP
Twenty three of 58 (40%) measurements (8 out of 21 from PD, 8 out of 25 from HD and 7 out of 12 from CKD5 group) had normal SBP z-score and normal Rel-OH (shaded area in Figure 3). Seventeen of 58 (29%) measurements (7 out of 21 from PD, 8 out of 25 from HD and 2 out of 12 from CKD5 group) had a SBP z-score >1.28 (90th percentile). Of these, only 2 (12%) of the measurements (1 PD and 1 HD patient) showed Rel-OH >7% (Quadrant I in Figure 3). The remaining 88% (15 measurements; 6 from PD, 7 from HD and 2 from CKD5 group) had normal hydration (n = 11) or even underhydration (n = 4) (Quadrant II in Figure 3). All measurements in healthy controls were entirely normal for BP and hydration status. When measurements in Quadrant I (↑BP and ↑Rel-OH) were compared against those in Quadrant II (↑BP but normal or ↓Rel-OH), the Rel-OH was significantly higher in Quadrant I (+11.4%) than in Quadrant II (−4.5%), P = 0.001. No correlation was seen between Wt-CA and Rel-OH. Relationship between Rel-OH and BP is illustrated in Table 2.
Relationship between degree of Rel-OH and BP
| Degree of Rel-OH, % | Number of measurements with SBP ≥ 90th centile/z-score ≥1.28 (%) | Number of measurements with SBP ≥95th centile/z-score ≥1.96 (%) |
|---|---|---|
| Rel-OH < −7%, underhydration | 4 out of 58 (6.90) | 3 out of 58 (5.17) |
| Rel-OH −7% to + 7%, normohydration | 11 out of 58 (18.97) | 7 out of 58 (12.07) |
| Number of measurements with SBP <90th centile/z-score <1.28 (%) | Number of measurements with SBP ≥95th centile/z-score ≥1.96 (%) | |
| Rel-OH +7% to + 15%, moderate overhydration | 2 out of 58 (3.45) | 2 out of 58 (3.45) |
| Rel-OH > 15%, severe overhydration | 1 out of 58 (1.72)a | – |
| Degree of Rel-OH, % | Number of measurements with SBP ≥ 90th centile/z-score ≥1.28 (%) | Number of measurements with SBP ≥95th centile/z-score ≥1.96 (%) |
|---|---|---|
| Rel-OH < −7%, underhydration | 4 out of 58 (6.90) | 3 out of 58 (5.17) |
| Rel-OH −7% to + 7%, normohydration | 11 out of 58 (18.97) | 7 out of 58 (12.07) |
| Number of measurements with SBP <90th centile/z-score <1.28 (%) | Number of measurements with SBP ≥95th centile/z-score ≥1.96 (%) | |
| Rel-OH +7% to + 15%, moderate overhydration | 2 out of 58 (3.45) | 2 out of 58 (3.45) |
| Rel-OH > 15%, severe overhydration | 1 out of 58 (1.72)a | – |
The measurements belong to a child who was persistently overhydrated. The initial encounter was recorded with concurrent high BP while the subsequent two other encounters demonstrated normal BP even with worsening Rel-OH. Interestingly, this child’s ECHO showed LVH and early manifestation of right ventricular dysfunction.
Relationship between degree of Rel-OH and BP
| Degree of Rel-OH, % | Number of measurements with SBP ≥ 90th centile/z-score ≥1.28 (%) | Number of measurements with SBP ≥95th centile/z-score ≥1.96 (%) |
|---|---|---|
| Rel-OH < −7%, underhydration | 4 out of 58 (6.90) | 3 out of 58 (5.17) |
| Rel-OH −7% to + 7%, normohydration | 11 out of 58 (18.97) | 7 out of 58 (12.07) |
| Number of measurements with SBP <90th centile/z-score <1.28 (%) | Number of measurements with SBP ≥95th centile/z-score ≥1.96 (%) | |
| Rel-OH +7% to + 15%, moderate overhydration | 2 out of 58 (3.45) | 2 out of 58 (3.45) |
| Rel-OH > 15%, severe overhydration | 1 out of 58 (1.72)a | – |
| Degree of Rel-OH, % | Number of measurements with SBP ≥ 90th centile/z-score ≥1.28 (%) | Number of measurements with SBP ≥95th centile/z-score ≥1.96 (%) |
|---|---|---|
| Rel-OH < −7%, underhydration | 4 out of 58 (6.90) | 3 out of 58 (5.17) |
| Rel-OH −7% to + 7%, normohydration | 11 out of 58 (18.97) | 7 out of 58 (12.07) |
| Number of measurements with SBP <90th centile/z-score <1.28 (%) | Number of measurements with SBP ≥95th centile/z-score ≥1.96 (%) | |
| Rel-OH +7% to + 15%, moderate overhydration | 2 out of 58 (3.45) | 2 out of 58 (3.45) |
| Rel-OH > 15%, severe overhydration | 1 out of 58 (1.72)a | – |
The measurements belong to a child who was persistently overhydrated. The initial encounter was recorded with concurrent high BP while the subsequent two other encounters demonstrated normal BP even with worsening Rel-OH. Interestingly, this child’s ECHO showed LVH and early manifestation of right ventricular dysfunction.
Relative overhydration (Rel-OH) against SBP z-score. The blue plot represents normal, i.e. SBP z-score −1.56 (5th percentile) to 1.28 (90th percentile) and Rel-OH from −7 to 7%. The solid black line across the y-axis represents the 90th percentile or z-score = 1.28. The dotted black line represents z-score = 1.96 corresponding with the 95th percentile, the threshold for defining hypertension. The four respective quadrants represent the hydration status in relation to the SBP z-score for age, gender and height. N, normal.
With the currently used hydration reference chart based on adult data, only one patient would be thought to have hypertension (Supplementary data, Figure S1), highlighting that paediatric reference ranges are essential for precise interpretation of volume status and BP.
Correlation between techniques of hydration assessment
Rel-OH significantly correlated with peripheral pulse pressure (R = 0.28, P = 0.03; Figure 4A). LVEDd, which directly reflects central aortic BP, showed a correlation with Rel-OH (R = 0.36, P = 0.04). Central aortic mean pressure and the aortic pulse pressure both correlated significantly with LVEDd, R = 0.47, P = 0.03 and R = 0.5, P = 0.01, respectively (Figure 4B). A direct correlation between central aortic pulse pressure and Rel-OH was not seen (R = 0.26, P = 0.08).
Spearman’s correlation between relative overhydration (Rel-OH). (A) Peripheral pulse pressure (n = 58 measurements), (B) LVEDd (n = 27 patients) and (C) plasma NT-proBNP levels (n = 54 measurements).
Higher Rel-OH positively correlated with higher NT-proBNP levels (R = 0.33, P = 0.02; Figure 4C). A subgroup analysis showed that mean NT-proBNP was significantly elevated when bioimpedance analysis showed overhydration (P < 0.005). However, no correlation was seen between NT-proBNP levels and clinical estimation of fluid overload, P = 0.56. se Na did not correlate with Rel-OH (R = 0.321, P = 0.12).
Correlation between BIS measures and cardiovascular outcome indicators
PWV z-score adjusted for age was associated with SBP z-score >1.28 (>90th centile), X2(1) = 3.89, P = 0.04. No correlation was seen with Rel-OH. Forty percent of the children on HD and 30% on PD had an LVH. Within the group, the mean NT-proBNP was relatively higher in those with LVH (2013.70 pg/mL versus 1803.36 pg/mL in HD group, P = 0.61 and 670.7 pg/mL versus 479.5 pg/mL in PD group; P = 0.78). In addition, we did not observe an association between Rel-OH and LVH, X2(1) =0.28, P = 0.59.
DISCUSSION
In this study, we have shown that in children on dialysis the estimation of hydration status by clinical assessment alone may be misleading, whereas BIS provides an accurate assessment of the hydration status that correlates with established biomarkers like NT-proBNP and cardiovascular measures. Also, almost 25% of children with hypertension did not have overhydration, suggesting that other causes of hypertension need to be explored in order to allow appropriate treatment.
Both overestimation and underestimation of optimal weight in the dialysis population can have deleterious effects. Overestimation of optimal weight can lead to an inadequate ultrafiltration prescription, eventually resulting in chronic fluid overload and LV strain. Overhydration is an important predictor of cardiovascular morbidity and mortality [31, 32]. On the other hand, underestimation of dry weight puts patients at risk of higher ultrafiltration rates. Vigorous ultrafiltration results in higher incidence of intradialytic hypotension manifesting as abdominal cramps, nausea and vomiting, dizziness and fainting spells. Excessive ultrafiltration can also be harmful: a recent study in children on hemodialysis showed a correlation between high ultrafiltration rates with LVMI [33]. Myocardial stunning is observed with intradialytic hypotension and this increases the risk of cardiovascular mortality [34]. Indeed, euvolaemia per se might be a more important dialysis adequacy parameter than small solute clearance, since fluid status better predicts patient outcome [35–37]. Given the potential adverse outcome of over- as well as underestimation of hydration, an objective assessment of optimal weight is important, both as a single time point measure, as well as to monitor trends in hydration status.
The hydration reference chart plots measurements taken over a period of time allows the clinician to have a better idea of patients’ hydration status in relation to their BP. In children with CKD, the KDIGO (Kidney Disease: Improving Global Outcomes) guideline recommends treatment for hypertension for BP >90th centile for age, sex and height. In order to adopt the hydration reference chart for paediatric practice, we propose that Rel-OH is plotted against the SBP z-score. We have adapted the hydration chart by dividing the quadrants into normal or high BP at z-score corresponding to the 90th instead of 95th centile on the SBP axis (Figure 1 versus Supplementary data, Figure S1). We believe that expressing the SBP as a z-score will represent the paediatric cohort better as BP limits in children are closely related to gender, age and height. In our study, if the conventional hydration chart was used (Supplementary data, Figure S1) only 1 child would be recognized as hypertensive, whereas using the modified hydration chart 17 (29%) had hypertension.
We observed that in 15 measurements (∼25%), an SBP >90th percentile was not accompanied by overhydration. A similar observation has been reported in a multicentre paediatric study that evaluated 463 pre-dialysis sessions whereby approximately one-third of the events were plotted in Quadrant II (elevated BP with normal or reduced Rel-OH) [38 ]. This suggests that high SBP is not always associated with fluid overload, and may be due to volume-independent factors. In our centre we practise a strict policy of fluid management and salt restriction on all dialysis patients. Children and families have frequent and easy access to renal dietetic support and all dialysis nurses are trained to recognize and control fluid overload using measures such as isolated ultrafiltration, sodium and ultrafiltration profiling, lowering dialysate sodium and lowering dialysate temperature to optimize ultrafiltration. Hence, volume-related hypertension was not prevalent in our centre. Incorporating BIS in day-to-day practice helps dialysis care providers to differentiate between volume-dependent and volume-independent hypertension and avoid excessive ultrafiltration. Vascular stiffness, congestive heart failure, residual renal function and the underlying renal disease could lead to a discrepancy between hydration measurement and BP [39–41]. Therefore, BP alone is not an effective tool to evaluate hydration status in chronic dialysis patients. Importantly, the misconception that elevated BP will benefit from increased ultrafiltration in order to achieve a lower dry weight may be incorrect and dangerous.
The clinical assessment of optimal weight requires experience and expertise, and carries the risk of both overestimation and underestimation of hydration status. In routine clinical practice the optimal weight of a patient on dialysis is assessed based on multiple factors that include the trend in weight, as well as the correlation of weight with BP, presence of oedema and the jugular venous pressure. In this study, we have attempted to seek an association between the clinical assessment of weight and presence of Rel-OH on BIS measurement, and shown that only 6.5% of measurements showed consistent overhydration. Also, clinical assessment may be inaccurate if the observer assumes that a high BP is a sign of volume overload. Our study showed that Rel-OH correlated with the LVEDd as well as another marker of volume overload, NT-proBNP. LVEDd has been shown to be superior to other echocardiographic parameters in predicting all-cause mortality in chronic HD patients [42]. Many studies have described the usefulness of NT-proBNP biomarker in determining congestive heart failure. In an adult study, NT-proBNP was associated with volume overload, LV dysfunction and inflammation [43, 44]. NT-proBNP has the longest half-life and hence is the most stable among the natriuretic peptides. Excretion is reduced in renal insufficiency [45] and variable clearance influenced by dialyser characteristics [46] leads to its wide range of plasma levels. In children, more studies are emerging on the role of this natriuretic peptide as indicator of both fluid overload as well as cardiac dysfunction [47, 48]. The outcome of combined data showed that the upper limit of normal NT-proBNP level did not differ in any age groups regardless of gender [49]. A prospective cohort study in dialysis patients that includes serial and simultaneous assessments of fluid overload and cardiac function would be required before NT-proBNP can be considered a reliable biomarker of fluid overload in clinical practice.
Clinical assessment may miss changes in lean tissue mass, an important aspect in the assessment of any child with CKD. Variation in body composition is a dynamic process and this does not coincide with change in body weight. The child on dialysis who loses lean tissue mass following an episode of catabolic stress may be falsely deemed ‘dry’ if the weight appears to be within target but in fact, the loss in tissue mass may be replaced by excessive fluid as renal function deteriorates. Similarly, children who gain lean tissue mass with improved dialysis clearance and vigorous nutritional support, are at risk of unnecessary ultrafiltration if their optimal weight is not appropriately adjusted.
BCM® measurements are feasible in regular clinical practice. BCM® is easy to perform by the bedside, non-invasive, well tolerated even by young children, operator-independent and therefore highly reproducible, and inexpensive. A high correlation between within-patient measurements suggests that trends in hydration status can also be monitored by BCM®. Monitoring for overhydration as well as changes in lean tissue mass provide a comprehensive insight into the nutritional and hydration status. The application of BCM® has been evaluated even in younger children with different outcomes reported. Several studies have demonstrated it to have high reproducibility and good LoA even when it is compared against dilution technique [47–51]. It is technically possible to perform BCM® even in younger children with some practice and play therapy, as long as the child is cooperative and able to engage with the operator [52–54]. However, BCM® does not distinguish between fat and fluid well, and is not suggested for use in infants, at least at the present time [55].
Studies in adults suggest that patients with post-dialysis OH of the extracellular space (i.e. excess extracellular water, and referred to as subclinical fluid overload) have been identified as being hypertensive but were not thought to be fluid overloaded on clinical assessment. Gradual dry weight adjustment using a BCM® was shown to promote regression of LV mass, reduced arterial stiffness and reduced BP in these patients. The study has also shown that the ECW–TBW ratio before and after HD was greater in those with intradialytic hypertension when compared with those who developed intradialytic hypotension [56]. BIS measures were able to detect this subclinical overhydration: the difference in mean reduction in weight correlated with the difference in mean reduction in OH as measured by BIS, but target dry weights and ultrafiltration volumes, which were determined clinically, failed to identify this subclinical fluid overload.
Lung ultrasound is an emerging tool for assessment of hydration status in the dialysis population, but pulmonary permeability in the uraemic milieu may contribute to excess lung water giving rise to a potentially false-positive impression [57, 58]. This is also dependent on the cardiac status of the individual [59]. On the contrary, specific assessment of electrical properties of tissues that is dependent on tissue hydration in BIS reflects changes in body water composition better [60]. In addition, BIS detects both over- and underhydration but lung ultrasound only picks up the former.
Our study is limited by the small number of patients. The short timeline of the study did not allow for correlations with outcome measures such as LVH. A 2.5-year prospective study in adults utilizing BIS to guide volume control reported improvement in overall mortality and cardiovascular endpoints [61]. Carotid-femoral PWV would be a better measurement for central arterial stiffness compared with carotid-radial PWV but the former may be less acceptable to patients, and was not allowed by our ethics committee in healthy controls. We were not able to perform 24-h ambulatory BP measures for this study. In children on PD we performed BCM® measurements with a full abdomen, as per the manufacturer’s instructions, but note that one study in 25 adults on PD suggests that BCM® measurements with an empty abdomen better reflects OH [62]. In this study, all measurements were performed by a single trained operator, with all analysis performed in a blinded fashion, allowing excellent consistency and contributing to the strength of this study.
In conclusion, we have shown that the clinical estimation of hydration status in children on dialysis can be misleading and BIS provides an objective assessment of hydration that correlates with biomarkers and cardiovascular measures of OH. A prospective multicentre study in children on dialysis is required to assess the benefits, if any, of BIS monitoring on cardiovascular outcomes.
SUPPLEMENTARY DATA
Supplementary data are available at ndt online.
ACKNOWLEDGEMENTS
This work was supported by the National Institute for Health Research Biomedical Research Centre at Great Ormond Street Hospital for Children NHS Foundation Trust and University College London. R.C.S. holds a Career Development Fellowship with the National Institute for Health Research (NIHR).
CONFLICT OF INTEREST STATEMENT
None declared.
REFERENCES
Comments