Comparison of fluid volume estimates in chronic hemodialysis patients by bioimpedance, direct isotopic, and dilution methods

Comparison of fluid volume estimates in chronic hemodialysis patients by bioimpedance, direct isotopic, and dilution methods

clinical investigation http://www.kidney-international.org & 2013 International Society of Nephrology see commentary on page 738 Comparison of flui...

990KB Sizes 0 Downloads 0 Views

clinical investigation

http://www.kidney-international.org & 2013 International Society of Nephrology

see commentary on page 738

Comparison of fluid volume estimates in chronic hemodialysis patients by bioimpedance, direct isotopic, and dilution methods Jochen G. Raimann1, Fansan Zhu1, Jack Wang2, Stephan Thijssen1, Martin K. Kuhlmann3, Peter Kotanko1, Nathan W. Levin1 and George A. Kaysen4 1

Renal Research Institute, New York, New York, USA; 2St Luke’s-Roosevelt Hospital Center, New York, New York, USA; 3Vivantes Klinikum im Friedrichshain, Berlin, Germany and 4University of California Davis, Davis, California, USA

Bioimpedance analysis (BIA) is accepted for the assessment of total-body water (TBW), intracellular fluid (ICF) and extracellular fluid (ECF). We aimed to compare precision and accuracy of single and multi-frequency-BIA to direct estimation methods (DEMs) of TBW, ECF, and ICF in hemodialysis patients. Linear regression analysis of volume estimates in 49 patients by single- and multi-frequency-BIA correlated significantly with DEMs. Bland-Altman analysis (BAA) found systemic bias for ECF single-frequency-BIA vs. ECF-DEMs. No other systematic biases were found. Proportional errors were found by BAA of ICF and ECF assessments with single- and multi-frequency bioimpedance spectroscopy compared to the DEMs. Comparisons of indirect methods (IEMs) to DEMs showed no significant differences and proportional errors. Root mean-squared-error analysis suggested slightly better accuracy and precision of ICF single-frequency-BIA vs. DEMs over ICF multi-frequency-BIA and IEMs to DEMs, and slightly better performance for ECF multi-frequency-BIA over both respective other methods. Compared to DEMs, there is slightly better accuracy for ECF multi- over single-frequency-BIA and ICF single- over multi-frequency-BIA. However the margin of differences between direct and indirect methods suggests that none of the analyzed methods served as a true ‘‘gold standard’’, because indirect methods are almost equally precise compared to DEMs. Kidney International (2014) 85, 898–908; doi:10.1038/ki.2013.358; published online 25 September 2013 KEYWORDS: chronic hemodialysis; hemodialysis; nutrition; water–electrolyte balance

Correspondence: Jochen G. Raimann, Renal Research Institute, 315 East 62nd Street, Suite 4-05, New York, New York 10065, USA. E-mail: [email protected] Received 13 February 2013; revised 22 July 2013; accepted 25 July 2013; published online 25 September 2013 898

Estimates of extracellular and intracellular fluid (ECF, ICF) volumes in dialysis patients are useful for the estimation of nutritional status and to distinguish between physiologic and pathologic expansion of ECF. Currently available methods include deuterium oxide (D2O) for the measurement of total-body water (TBW), bromide dilution (Br) for the measurement of ECF and total-body potassium (TBK) for the determination of ICF. Volume estimation using bioimpedance with either single-frequency bioimpedance analysis (mostly 50 kHz) or multi-frequency bioimpedance spectroscopy (SF-BIA, MF-BIS) is simple to perform. Estimates of both techniques have been proposed as useful measures of body composition (BC) and fluid overload in dialysis patients.1,2 The advantages of bioimpedance compared with other techniques are its low cost and easy portability allowing bed-side measurements without the need for blood draws. However, biases between bioimpedancederived estimations and ‘gold-standard’ methods have been reported.3,4 Other studies indicate that isotope dilution methods themselves may not be without error or bias.5 Inter-individual differences in BC have been shown to be associated with the lack of agreement between the methods in healthy subjects4 and dialysis patients.6 Differences in BC contribute to sex-specific differences between extra- and intracellular resistivities. It is interesting to note that the difference is more accentuated for extracellular resistivity, which appears to be due to the higher fat mass (and a subsequent ECF over fat-free mass ratio) found in women. Intracellular resistivity is basically determined by muscles and organs; thus, the presence of a higher number of cells containing only a smaller cytoplasmic volume (which in essence represents the ICF) will alter the results and the resistivity of the body segment of interest.7 To evaluate the inherent error(s) intrinsic to the three direct estimation methods (DEMs) compared with bioimpedance technology in our data, we estimated the fluid compartments directly and compared those estimates with the ones calculated by the other two dilution methods. This approach was based on the assumption that if each estimate were accurate the results using either the direct method or Kidney International (2014) 85, 898–908

JG Raimann et al.: Comparison of methods of fluid volume estimation

clinical investigation

the sum (or difference) of the other two (indirect estimation method (IEM)) will provide the same value. We also assessed fluid volume in the study cohort using SF-BIA and MF-BIS and compared those with DEM results.

a Bland–Altman analysis (BAA) did not show significant regression (Figure 2a and b; all P40.05). In addition, no proportional error in BAA (Figure 2c) was found for TBW DEM versus TBW IEM. The root mean squared error (RMSE) as a marker of accuracy and precision combined was slightly lower for TBW SF-BIA (RMSE of 5.2) than for TBW MF-BIS (RMSE of 6.1) when the estimates were compared with DEM. Both RMSE values were lower as compared with the RMSE from the comparison of DEM versus IEM (Table 2). Intracellular fluid. Estimates of ICF were not significantly different when measured with MF-BIS or SF-BIA (Table 2) and compared with ICF DEM. The difference between ICF DEM and IEM (  1.8 (95% CI  3.7 to 0.1)) was also not significant. The regression plot showing ICF estimations by all techniques versus DEM is shown in Figure 3 and the

Fifty-four subjects were studied from March 2002 to March 2011. Two subjects were excluded because of fit errors in the MF-BIS measurements and three had missing DEM values. See the demographics of the final 49 subjects in Table 1. Measurements

Volume estimates are shown in Table 2. DEM results were normally distributed, and so parametric tests were employed. Total-body water. Estimates of TBW by SF-BIA did not significantly differ in comparison with TBW DEM (  2.1 (95% confidence interval (CI)  5.3 to 1.1)) and MF-BIS (1.9 (95% CI  1.3 to 5.1)) (Table 2). Comparison between TBW DEM and TBW IEM also showed no significant systematic bias (1.8 (95% CI  1.4 to 5.0; Table 2)). The regression plot showing TBW estimations by all techniques versus DEM is shown in Figure 1 and the regression function and coefficient are provided in Supplementary Table S1 online. On the regression plot the slopes of the fit lines appeared similar between SF-BIA (m ¼ 0.86; Po0.001) and MF-BIS (m ¼ 0.76; Po0.001). However, the systematic bias was slightly larger in SF-BIA than in MF-BIS, which is reflected by a parallel shift on the regression plot (Figure 1). Comparison of TBW SF-BIA and MF-BIS with TBW DEM in

70 TBW (SF and MF and (Br+TBK)) (l)

RESULTS Patient selection

TBW (SF vs. D2O) TBW (MF vs. D2O) TBW ((Br+TBK) vs. D2O)

60

50

40

30

20 20

Table 1 | Demographics and comparison of male and female subjects Parameter (mean±s.d.)

All subjects Male subjects Female subjects (n ¼ 49) (n ¼ 29) (n ¼ 20)

Age (years) 56.1±11.2 Race (non-black/black) 14/35 Height (cm) 166.7±9.3 Pre-hemodialysis 75.7±16.2 weight (kg) Body mass index (kg/m2) 27.2±5.5

55.8±10.1 10/19 171.9±7.1a 79.2±12.6

56.4±13.0 4/16 159.3±6.8a 70.8±19.5

26.8±3.9

27.8±7.2

a

Po0.05 (male vs. female).

30

40

50

60

70

TBW (D2O) (l)

Figure 1 | Linear regression plot: total-body water (TBW) (single-frequency (SF) and multi-frequency (MF) and (bromide (Br) þ total-body potassium (TBK)) vs. deuterium oxide (D2O)). Linear regression of TBW as estimated by a direct ‘gold-standard approach’ with D2O dilution (TBW D2O) versus TBW as assessed by whole-body MF and SF bioimpedance analysis, and an indirect ‘goldstandard approach’ using the sum of intracellular fluid volume (as assessed per TBK (ICF TBK)) and extracellular fluid volume ((as per bromide dilution (ECF Br)). Solid symbols (m, K, and ’, respectively) indicate those excluded from the sensitivity analysis.

Table 2 | Comparison of measures of distribution of fluid volume by direct or indirect estimation methods (DEM, IEM) or single or multi-frequency bioimpedance

TBW (l) ICF (l) ECF (l)

DEM

IEM

MF-BIS

SF-BIA

DEM–MF-BIS

DEM–SF-BIA

DEM–IEM

RMSE (DEM–MF-BIS)

RMSE (DEM–SF-BIA)

RMSE (DEM–IEM)

40.0±7.8 20.6±4.3 17.6±5.0

38.2±8.2 22.4±5.2 19.4±5.3

38.1±8.1 20.4±5.4 17.7±3.5

42.1±8.3 20.2±5.4 21.9±3.8

1.9 (  1.3 to 5.1) 0.3 (  1.7 to 2.2)b  0.2 (  1.9 to 1.6)b

 2.1 (  5.3 to 1.1) 0.5 (  1.5 to 2.4)b  4.4 (  6.2 to  2.6)a,b

1.8 (  1.4 to 5.0)  1.8 (  3.7 to 0.1)  1.8 (  3.9 to 0.3)

6.1 3.6 3.8

5.2 2.5 5.8

4.4 4.4 4.4

Abbreviations: DEM, direct estimation method; ECF, extracellular fluid; ICF, intracellular fluid; IEM, indirect estimation method; MF-BIS, multi-frequency bioimpedance spectroscopy; RMSE, root mean squared error; SF-BIA, single-frequency bioimpedance analysis; TBW, total-body water. RMSE was computed for each of the comparisons. Data are reported as mean±s.d. or mean (95% confidence interval), as appropriate. TBW ¼ ICF TBK þ ECF Br; ICF ¼ TBW D2O  ECF Br; ECF ¼ TBW D2O  ICF TBK. a Po0.05 for systemic bias. b Po0.05 for proportional bias.

Kidney International (2014) 85, 898–908

899

clinical investigation

20

Male Female

Difference TBW (D2O–MF) (l)

Difference TBW (D2O–SF) (l)

20

JG Raimann et al.: Comparison of methods of fluid volume estimation

10

0

–10

Male Female

10

0

–10

–20

–20 30

40 50 60 TBW ((D2O+SF)/2) (l)

Difference TBW ((D2O+(Br+TBK))/2) (l)

20

20

30

20

70

40 50 60 TBW ((D2O+MF)/2) (l)

70

Male Female

10

0

–10

–20 20

30

40

50

60

70

TBW ((D2O+(Br+TBK))/2) (l)

Figure 2 | Bland–Altman plot (N ¼ 49). Bland–Altman plot showing the difference between total-body water (TBW) as estimated by a direct ‘gold-standard approach’ with deuterium dioxide dilution (TBW D2O) and TBW as assessed by whole-body (a) single-frequency and (b) multi-frequency bioimpedance analysis, and (c) an indirect ‘gold-standard approach’ using the sum of intracellular fluid volume (as assessed per total-body potassium (ICF TBK)) and extracellular fluid volume ((as per bromide dilution (ECF Br)). (a) TBW (D2O vs. SF-BIA); (b) TBW (D2O vs. MF-BIS); (c) TBW (D2O vs. (Br þ TBK)). Mean±2 s.d. is indicated; men are denoted by filled (m, K, and ’, respectively) and women by unfilled symbols (D, J and &, respectively).

900

40 ICF (SF and MF and (D2O–Br)) (l)

regression function and coefficient are provided in Supplementary Table S1 online. In BAA, comparison of ICF SF-BIA with ICF DEM showed a proportional error ((m ¼  0.24; Po0.001) with an R2 of 0.20 (Po0.001) (Figure 4a and b), which implies ICF SF-BIA underestimates ICF DEM at high values and overestimates it at low values. Comparison of ICF MF-BIS with ICF DEM showed a similar relationship (Figure 4b) (R2 ¼ 0.09; P ¼ 0.02). The difference in the comparison between ICF DEM and ICF IEM did not show a proportional error. For the comparison of ICF MF-BIS and ICF SF-BIA, both methods, when compared with DEM, showed better accuracy and precision as reflected by a RMSE of 3.6 and 2.5, respectively, than the comparison of IEM with DEM (RMSE of 4.4; Table 2). Extracellular fluid. There was no significant difference between ECF DEM and ECF MF-BIS. The difference between ECF SF-BIA (  4.4 (95% CI  6.2 to 2.6)) and ECF DEM was significant. The difference between ECF DEM and ECF IEM was not significant. The regression plot showing ECF estimations by all techniques vs. DEM is shown in Figure 5 and the regression function and coefficient are provided in Supplementary Table S1 online. It may be noted that the slopes of the measurements of all three methods strongly

ICF (SF vs. TBK) ICF (MF vs. TBK) ICF ((D2O–Br) vs. TBK)

35 30 25 20 15 10 10

15

20

25

30

35

40

ICF (TBK) (l)

Figure 3 | Linear regression plot: intracellular fluid (ICF) (singlefrequency (SF) and multi-frequency (MF) and (deuterium dioxide (D2O)  bromide (Br)) vs. total-body potassium (TBK)) (N ¼ 49). Linear regression of ICF volume as estimated by a direct ‘goldstandard approach’ with TBK (ICF TBK) vs. ICF as assessed by wholebody MF and SF bioimpedance analysis and an indirect ‘goldstandard approach’ using the difference between total-body fluid volume (assessed with D2O dilution (TBW D2O)) and extracellular fluid volume (as per bromide dilution (ECF Br)). Solid symbols (m, K, and ’, respectively) indicate those excluded for the sensitivity analysis. Kidney International (2014) 85, 898–908

clinical investigation

Male Female

10

Difference ICF (TBK–MF) (l)

Difference ICF (TBK–SF) (l)

JG Raimann et al.: Comparison of methods of fluid volume estimation

5 0

–10

–20 10

15

20

25

30

35

10

Male Female

5 0

–10

–20 10

40

15

25

30

35

40

ICF((TBK+MF)/2) (l)

ICF((TBK+SF)/2) (l) Difference ICF (TBK–(D2O–Br)) (l)

20

Male Female

10 5 0

–10

–20 10

15

20

25

30

35

40

ICF((TBK+(D2O–Br))/2) (l)

Figure 4 | Bland–Altman plot (N ¼ 49). Bland–Altman plot showing the difference between intracellular fluid as estimated by a direct ‘gold-standard approach’ with total-body potassium (ICF TBK) and ICF as assessed by whole-body (a) single-frequency and (b) multi-frequency bioimpedance analysis, and (c) an indirect ‘gold-standard approach’ using the difference between total-body fluid volume (assessed with deuterium dioxide dilution (TBW D2O)) and extracellular fluid volume (as per bromide dilution (ECF Br)). (a) ICF (TBK vs. SF-BIA); (b) ICF (TBK vs. MF-BIS); (c) ICF (TBK vs. (D2O  Br)). Mean±2s.d. is indicated; men are indicated by filled (m, K, and ’, respectively) and women by unfilled symbols (D, J, and &, respectively).

ECF (SF and MF and (D2O–TBK)) (l)

deviated from the line of identity (ECF SF-BIA vs. ECF DEM: m ¼ 0.49 (Po0.01); ECF MF-BIS vs. ECF DEM: m ¼ 0.45 (Po0.01); ECF DEM vs. ECF IEM: m ¼ 0.73 (Po0.01)). In BAA both bioimpedance techniques showed a significant proportional error, which implies that both bioimpedance techniques underestimate ECF DEM at low values and overestimate ECF DEM at high values (Figure 6a and b). Bland–Altman plot of the difference between ECF DEM and ECF IEM did not show a proportional error (Figure 6c). The RMSE for the comparison of DEM vs. MF-BIS was 3.8 and the lowest of all ECF estimations (Table 2).

ECF (SF vs. Br) ECF (MF vs. Br) ECF ((D2O–TBK) vs. Br)

40

30

20

10

0 0

10

20

30

40

ECF (Br) (l)

Figure 5 | Linear regression plot: extracellular fluid (ECF) (singlefrequency (SF) and multi-frequency (MF) (deuterium dioxide (D2O)  total-body potassium (TBK)) vs. bromide (Br)) (N ¼ 49). Linear regression of ECF as estimated by a direct ‘gold-standard approach’ with Br dilution (ECF Br) vs. ECF as assessed by whole-body MF and SF bioimpedance analysis, and an indirect ‘gold-standard approach’ using the difference between total-body fluid volume (assessed with deuterium dioxide dilution (TBW D2O)) and intracellular fluid volume (as measured using total-body potassium (ICF TBK)). Solid symbols (m, K, and ’, respectively) indicate those excluded from the sensitivity analysis. Kidney International (2014) 85, 898–908

Repeatability

The repeatability of all volume estimates was assessed by computing the volume using the extra- and intracellular resistance of each run for the MF-BIS estimates and calculating the coefficient of variation (TBW 1.2%, ICF 2.1%, and ECF 0.2%). For SF-BIA the same methodology was employed using reactance and impedance for the computation of volume estimates (TBW 0.1%, ICF 0.5%, and ECF 0.1%). 901

clinical investigation

JG Raimann et al.: Comparison of methods of fluid volume estimation

10

Male Female

Difference ECF (Br–MF) (l)

Difference ECF (Br–SF) (l)

10 5 0 –5 –10 –15 5

10

15

20

25

30

Male Female

5 0 –5 –10 –15 5

10

Difference ECF (Br–(D2O–TBK)) (l)

10

15

20

25

30

ECF ((Br+MF)/2) (l)

ECF ((Br+SF)/2) (l)

Male Female

5 0 –5 –10 –15 5

10 15 20 25 ECF ((Br+(D2O–TBK))/2) (l)

30

Figure 6 | Bland–Altman plot (N ¼ 49). Bland–Altman plot showing the difference between of intracellular fluid volume as estimated by a direct ‘gold-standard approach’ with bromide dilution (ECF Br) and ECF as assessed by whole-body (a) single-frequency and (b) multifrequency bioimpedance analysis, and (c) an indirect ‘gold-standard approach’ using the difference between total-body fluid (assessed with deuterium dioxide dilution (TBW D2O)) and intracellular fluid volume (as per total-body potassium (ICF TBK)). (a) ECF (Br vs. SF-BIA); (b) ECF (Br vs. MF-BIS); (c) ECF (Br vs. (D2O  BK)). Mean±2s.d. is indicated; men are indicated by filled (m, K, and ’, respectively) and women by unfilled symbols (D, J, and &, respectively).

Subset analyses

Results differed between sexes: The systematic bias for ECF SF-BIA (when compared to ECF DEM) was consistent for both sexes; however, a systematic bias for ICF SF-BIA (as compared with ICF DEM) was found in women only and was not present for men (Table 3). The proportional error in the comparison of ICF and ECF MF-BIS with ICF and ECF DEM was consistent for both sexes (Table 3). For the estimation of ICF with IEM there were sex-specific differences that showed a systematic bias between ICF DEM and IEM in women only. However, estimations of TBW did not show any sex-specific differences. Differences in accuracy and precision were also found in the RMSE values, which showed particularly distinguished differences for the estimation of ICF between men and women. In contrast to the obvious effect of sexes (in particular in the comparison between DEM and SF-BIA), the stratification into older (454 years) and younger (o54 years) only showed differences between the estimation of ICF and ECF MF-BIS compared with DEM (Supplementary Table 2 online). Notably, the accuracy and precision as per RMSEs after stratification into older and younger patients showed results consistent with the primary analysis. Sensitivity analysis. Exclusion of subjects showing implausible values resulted in a subset with three excluded subjects: One of these three subjects had an ECF/TBW DEM 902

ratio that exceeded the mean 0.45±0.10 by more than 2 standard deviations (s.d.); one subject had an ECF/ICF DEM ratio that exceeded the mean by more than 2 SDs and one had an ECF/TBW DEM ratio below 2 standard deviations of the mean. Analysis of this subset showed consistent results with the primary analysis. The proportional error in the ECF MF-BIS as compared with ECF DEM was not present in this secondary analysis. The regression plots depicted in Figures 1, 3, and 5 indicate the excluded subjects (solid symbols m, K, and ’, respectively). Error analysis

The relative differences between DEM and IEM estimations, and the lack of absolute accuracy and precision (Tables 2 and 3) led us to investigate the cause of these differences. Only the difference in ECF had a significant regression (m ¼ 0.27) in a modified Bland–Altman plot (R2 ¼ 0.09, P ¼ 0.02) when analyzed as a function of ECF DEM (as per Krouwer8). The same relation holds true when analyzed as a function of the IEM value (R2 ¼ 0.18, Po0.01; Figure 7). Regression analysis showed no significant association between the errors between DEM and IEM, and anthropometric parameters (height and weight) and demographic parameters (age, sex and race), respectively. The only significant predictor of this error in the model was the Kidney International (2014) 85, 898–908

clinical investigation

JG Raimann et al.: Comparison of methods of fluid volume estimation

Table 3 | Comparison of measures of distribution of fluid volume by direct or indirect estimation methods (DEM, IEM) or singleor multi-frequency bioimpedance in subjects stratified into (a) males and (b) females

(a) TBW (l) ICF (l) ECF (l)

DEM

IEM

MF-BIS

SF-BIA

Difference (DEM–MF-BIS)

Difference (DEM–SF-BIA)

Difference (DEM–IEM)

RMSE (DEM-MF-BIS)

RMSE (DEM-SF-BIA)

RMSE (DEM–IEM)

43.7±7.0 23.1±3.2 19.3±4.6

42.5±6.4 24.4±4.8 20.6±5.8

41.9±6.4 22.4±4.7 19.6±2.5

47.0±5.8 23.9±3.3 23.1±3.3

1.8 (  1.8 to 5.3) 0.8 (  1.4 to 2.9)a  0.2 (  2.2 to 1.8)a

 3.3 (  6.7 to  0.1)  0.8 (  2.5 to 0.9)  3.8 (  5.9 to  1.6)b

1.3 (  2.3 to 4.8)  1.3 (  3.4 to 0.9)a  1.3 (  4.0 to 1.5) a

5.9 2.9 4.2

6.1 2.0 5.7

4.1 4.1 4.1

DEM

IEM

MF-BIS

SF-BIA

(b) TBW (l) 34.7±5.6 32.1±6.5 32.6±7.3 35.0±5.7 ICF (l) 17.0±2.8 19.6±4.4 17.5±5.2 14.7±2.3 ECF (l) 15.0±4.5 17.6±3.9 15.1±3.1 20.2±3.9

(DEM–MF-BIS)

(DEM–SF-BIA)

(DEM–IEM)

2.1 (  2.1 to 6.2)  0.4 (  3.1 to 2.3)a  0.1 (  2.6 to 2.4)a

 0.3 (  3.9to 3.3) 2.3 (0.6 to 3.9)b  5.2 (  7.9 to  2.5)b

2.6 (  1.3 to 6.5)  2.6 (  5.0 to  0.2)a,b  2.6 (  5.3 to 0.1)

RMSE RMSE RMSE (DEM–MF-BIS) (DEM–SF-BIA) (DEM–IEM) 6.5 4.5 3.2

3.7 3.0 6.0

4.8 4.8 4.8

Abbreviations: DEM, direct estimation method; ECF, extracellular fluid; ICF, intracellular fluid; IEM, indirect estimation method; MF-BIS, multi-frequency bioimpedance spectroscopy; RMSE, root mean squared error; SF-BIA, single-frequency bioimpedance analysis; TBW, total-body water. RMSE was computed for each of the comparisons. Data are reported as mean±s.d. or mean (95% confidence interval), as appropriate. TBW ¼ ICF TBK þ ECF Br; ICF ¼ TBW D2O  ECF Br; ECF ¼ TBW D2O  ICF TBK. a Po0.05 for proportional bias. b Po0.05 for systemic bias.

Table 4 | Multivariable regression analysis of the difference in extracellular fluid as estimated by the direct estimation method using bromide dilution and the indirect estimation method using the difference between total-body water by D2O dilution and total-body potassium counting

Difference ECF (Br–(D2O–TBK)) (l)

10 5 0 –5 –10 –15 5

10

15 20 ECF (D2O–TBK) (l)

25

30

Figure 7 | Bland–Altman plot: extracellular fluid (ECF) (bromide (Br) vs. (deuterium dioxide (D2O)  total-body potassium (TBK)) (N ¼ 49). Modified Bland–Altman plot (only using the indirect estimation method on the x-axis) showing the comparison between ECF volume (ECF) as per Br dilution (ECF Br) and ECF as per the indirect estimation method (ECF ¼ TBW D2O  ICF TBK). Mean±2s.d. is indicated.

magnitude of ECF DEM, suggesting that the error is associated with the accuracy and/or precision of this method (Table 4). DISCUSSION Statement of principal findings

We aimed to compare the errors between DEM and MF-BIS, SF-BIA, and IEM in hemodialysis patients to assess the agreement between the methods used to estimate fluid volumes in this population. Except for the biased assessment of ECF SF-BIA (Table 3), volumes were not significantly biased when bioimpedance estimates were compared with DEM. Both bioimpedance methods appeared to be equally precise in measuring TBW (with an almost identical error in Kidney International (2014) 85, 898–908

Parameter

Estimate (s.e.)

Constant Age (years) Male (yes/no) Race (non-black/black) Height (cm) Pre-hemodialysis weight (kg) ECF bromide (l)

2.74  0.02 0.31 1.71 0.05 0.02 0.38

(15.11) (0.05) (1.66) (1.36) (0.10) (0.04) (0.16)a

Abbreviation: ECF, extracellular fluid. a Po0.05.

the difference from DEM). Despite a proportional error, ECF MF-BIS appeared to be slightly more accurate and precise than ECF SF-BIA (and ECF IEM) based on the RMSE when both were compared with ECF DEM. ECF appeared to not have a systemic bias when bioimpedance methods were compared with DEM; however, a proportional error was found for both (reflected by a regression in the Bland–Altman plot). Although we cannot establish the mechanism for this divergence at this time, the difference between both bioimpedance methods and the ECF DEM results was not significant compared with the difference between ECF DEM and ECF IEM. A better performance is also suggested by a RMSE lower for ECF MF-BIS than for ECF IEM and ECF SF-BIA. The RMSE of the ICF estimation by SF-BIA suggests slightly better accuracy and precision of SF-BIA over all other methods in comparison with ICF DEM (RMSE of 2.5 for SF-BIA; 3.6 for MF-BIS; and 4.1 for IEM; Table 2). D2O, Br, and TBK are commonly considered as ‘goldstandard’ methods for measurements of TBW, ECF, and ICF, 903

clinical investigation

JG Raimann et al.: Comparison of methods of fluid volume estimation

respectively, but, clearly, these fluid volumes are interrelated. If each were absolutely precise and accurate, then the values measured for the corresponding spaces should be independent of the marker used. For example, the TBW DEM should be identical to TBW IEM-, but clearly that was not what we found in the comparison of all of these methods (Table 2). We found the accuracy and precision of MF-BIS estimates to be comparable to that of the comparison of DEM with IEM results. It may be argued that the confidence limits in BAA are wide, which casts doubt on the use of all methods used. However, also this accounts equally for IEM and both BIA methods. We did not systematically investigate the reproducibility and longitudinal performance of the volume estimations in the current analysis. In the treatment of chronic hemodialysis patients this may be of particular interest for longitudinal observance of body volume content and also of nutritional status. The subgroup analysis of the effect of sex on accuracy and precision appeared to favor MF-BIS due to the smaller differences in comparison with DEM. However, a proportional error of ECF MF-BIS and ICF MF-BIS (compared with the corresponding DEM values) was present in men and women, whereas SF-BIA did not show this proportional error but showed accentuated systematic bias in ECF SF-BIA in women. Consistent with the primary analysis, there was no systematic bias in the estimation of all three fluid volume estimates by using MF-BIS as compared with DEM in either men or women. This was in contrast with the comparison to SF-BIA, which showed a substantially accentuated bias in women. The subset analysis of subjects with biologically extreme values did not show altered results, thus suggesting possible pre-analytic and/or analytic errors did not affect the validity of our results. The error analysis using potential predictors of the error between DEM and IEM suggests the magnitude of ECF DEM as being the driving determinant of the error (Figure 7).

and claimed high correlation between the resistances assessed with SF- and MF-BIS at different frequencies. Although, as suggested by Bland and Altman,13 linear regression is not an appropriate tool to test for agreement between methods, because the correlation coefficient is most strongly affected by the range of the analyzed data points (obviously quite substantial in the analysis of resistances, with the unit O commonly ranging between 400 and 700), an R2 larger than 0.9 needs to be acknowledged. This suggests that the methods are equally accurate in measuring electrical resistance. This emphasizes the necessity of most accurate and precise regression models to compute volumes without biases. General strengths and weaknesses

The main strength of this study is without doubt the possibility to compare so-called ‘gold-standard’ methods with bioimpedance methods and also against each other. This, however, leads to more fundamental questions such as: ‘What actually can be called a ‘gold-standard’? Substantial differences between the methods (IEM vs. DEM; see Table 2 and Table 3; even gaining significance for ICF in women (Table 3b)) suggest inability to estimate volumes with absolute accuracy. These findings are in agreement with a recent publication that reported biases between dilution methods for the estimation of ECF.5 An additional strength is the considerably large sample size enabling to draw solid conclusions. A predominance of black subjects in the study cohort needs to be acknowledged. However, in the error analysis using multivariable regression analysis, race appeared not be significantly associated with the error in the comparison of DEM and IEM. It also needs to be noted that no assessment of residual urine volume was performed to adjust the volume estimation for diluted tracers lost. This, however, has been determined to be negligible.14,15 Meaning of the study

Strengths and weaknesses in relationship to other studies

Bipedal SF-BIA has been suggested to be superior to anthropometric formulas for the estimation of TBW and more convenient as compared with urea kinetic modeling.9 Further it has been suggested that ICF MF-BIS offers slightly better precision as compared with ICF SF-BIA.3 Gudivaka et al.10 reported a randomized cross-over study investigating the effects of Ringer’s lactate infusion and administration of a diuretic on the agreement of seven clinically established BIA models to DEM in 27 patients. Consistent with our results, the authors report better agreement with models based on the results acquired by MF-BIS. The model by De Lorenzo et al.11 (which we used to compute our data in the current analysis—see Appendix) was also part of this analysis and performed most accurately compared to other models.10 The authors also report similar results after volume administration (potentially resembling a fluid-overloaded state) and euvolemic state.10 Piccoli et al.12 reported a correlation between resistances and reactances at different frequencies 904

Many factors may have caused the proportional errors and biases found. Firstly, the models by Kotler et al. and Wang et al. were developed in patients without known kidney disease. This clearly distinguishes the derivation cohort from the presently analyzed population. As weight and height are important determinants in both models, weight changes in fluid-overloaded patients are conceivable confounders in volume estimations in dialysis patients. Furthermore, it appears that the model developed by Kotler et al. does not sufficiently account for sex effects to compensate for the bias introduced by fluid overload, which was absent in their derivation cohort (Table 3). In general, it is difficult to determine the error sources causing biases in cross-sectional data. Five sources of errors in bioimpedance measurements have been identified (summarized by Piccoli et al.12): (1) the measurement error, (2) the regression error (standard error of the estimate against the reference method), (3) the intrinsic error of the reference method (assumptions and measurements error of the Kidney International (2014) 85, 898–908

JG Raimann et al.: Comparison of methods of fluid volume estimation

clinical investigation

dilution reference method and models), (4) the electric-volume model error (anisotropy of tissues and human body geometry other than one cylinder), and (5) the biological variability among subjects (different BC and body geometry).12,16,17 The non-uniform occurrence of the error after stratification of the study population into men and women suggests that despite two entirely separate models being developed for men and women, there is a sex-related bias for SF-BIA volume estimation, which may affect either the accuracy of the dilution method or the regression model using SF-BIA values. The error is present and more pronounced in men than in women for the comparison between DEM and IEM, respectively, with SF-BIA, but almost entirely disappears in women (see Table 3b). This sex-specific bias was not seen for MF-BIS (of note despite a preponderance of men in DeLorenzo et al.’s data (63 men vs. 10 women)). This suggests bias is related to either the sex or the BC of the studied subjects. Some simulations performed employing the model by Kotler et al. showed that the model is highly sensitive to height (e.g. a 1-cm increase in height changes the estimated volume between 150 and 200 ml). As 50 kHz current can go through extracellular and partly ICF spaces, the accuracy and precision of using the SF-BIA method cannot avoid the influence of interindividual differences on BC. However, both techniques need to be largely improved in order to apply bioimpedance methods in clinical routine. In addition to fluid overload as a contrasting difference between the studied hemodialysis population and the respective derivation cohorts of the respective regression models, other factors may be responsible—e.g. the binding properties of the used tracers may be different over a range of pH (largely different in dialysis patients as compared with healthy subjects). In addition, cell membrane permeability and protein concentrations may be altered (as necessary for Gibbs–Donnan and void volume adjustments done for the dilution methods with constant fudge factors derived from the literature). An additional evidence for the derivation cohort affecting the performance of the methods is the difference in the presence of proportional errors in BAA for MF-BIS. A proportional error is only present for the estimation of ICF and ECF BIS in comparison with DEM in subjects older than the median age (Supplementary Table S1 online). That is despite a (non-significantly) larger magnitude of ICF and ECF in the older population and thus suggests influences beyond the magnitude of the measured volume only. A possible explanation could be that the average age of DeLorenzo’s derivation cohort was 36±10 years and thus substantially younger than the current study population, which would imply that age-related changes in BC confound the accuracy and precision of the methods.

kinetic principles and binding properties of the administered compounds in connective tissue or plasma constituents. Furthermore, it has to be evaluated if the distribution of these molecules differs between hemodialysis patients and the general population. All these potential differences between dialysis patients and the general population need to be evaluated in more detail and can, at this point, not yet be determined with absolute certainty. One conclusion can be drawn at this point: Bioimpedance can be of great help in clinical medicine for the monitoring of body fluid volumes and nutritional markers such as muscle mass and intracellular volume. The errors in precision and accuracy are evident, but are of comparable magnitude to the errors found between the measurements of so-called ‘gold-standard’ techniques. The segmental bioimpedance technique may potentially increase the accuracy and precision of the estimations,7,18 but improvements may also be made by revisiting the approaches and models developed in the general population and generating new equations based on the data obtained in dialysis patients. This will likely affect the coefficients and change their magnitude and may likely result in models with improved accuracy and precision relevant to this specific population of interest. However, in the light of these data it needs to be acknowledged that more research is needed to evaluate the possibilities of reducing the errors in fluid volume estimation. METHODS The entire study was carried out at the Body Composition Unit at St Luke’s-Roosevelt Medical Center and in the dialysis clinics of the Renal Research Institute. Subjects’ demographics were evaluated in the former institution. Race and ethnicity were recorded based on the subject’s four grandparents’ self-description.

Unanswered questions and future research

Patient selection Maintenance hemodialysis patients dialyzing thrice weekly were enrolled at Renal Research Institute (RRI) clinics in an urban area. Subjects with data for isotope dilution and TBK, and available SF-BIA and MF-BIS volume estimations were included in this comparative analysis. A fit error in the Cole–Cole model for the computation of MF-BIS values was evaluated and those not meeting the quality criteria commonly applied at RRI were excluded. The study was approved by a local Institutional Review Board and conducted according to the Declaration of Helsinki. Subset analysis. For verification of our findings, we conducted three subset analyses: One stratified the entire population according to sex and repeated all analyses in a sex-specific fashion. In the second subset analysis subjects with biologically implausible values were excluded to test the correctness of our results. Ratios of ECF to ICF and ECF to TBW were calculated using the DEM methods and subjects showing implausible ratios (defined as a value exceeding the population mean plus two standard deviations) were excluded. The third stratified the subjects according to the median age of the entire population.

More research is needed to develop methods for fluid volume estimation without bias. These data suggest that there is no real ‘gold standard’ with absolute accuracy (at least in dialysis patients). Research is needed for the evaluation of the

Measurements Dilution methods. These studies were all carried out at the Body Composition Unit at St Luke’s-Roosevelt Medical Center

Kidney International (2014) 85, 898–908

905

clinical investigation

JG Raimann et al.: Comparison of methods of fluid volume estimation

approximately 3 h prior to the bioimpedance measurements preceding a midweek hemodialysis treatment. Dilution methods of D2O and NaBr were used to directly estimate TBW and ECF (TBW D2O, Br), respectively.19 Each patient was given an oral dose of 10 g of D2O (ICON, Summit, NJ) and 5 g of 4 M NaBr solution. The coefficient of variation for the D2O measurement using infrared spectroscopy is 1.2% and the coefficient of variation for the measurement of Br dilution using measurements by high-performance liquid chromatography system is 1.4%.19,20 Blood samples were collected immediately before and prior to the hemodialysis treatment at least 3 h after intake of these substances, assuming that total equilibration had been reached. This approach has been termed as DEM throughout the manuscript. Estimation of TBK was carried out by measurement of total endogenous 40K, emission from a natural radioisotope of K, by assessment of g-ray emission of 40K counts accumulated over 9 min adjusted for body size on the basis of a 42K calibration formula reported by Pierson et al.19,21,22 The ICV estimation based on TBK (mmol) is conducted under the assumption that intracellular K þ concentration is considerably stable at around 152 mEq/l and allows, in the knowledge of TBW, the estimation of the intracellular K40 distribution volume.23 Analogously to the dilution methods this approach is also termed DEM. To gauge the internal accuracy of the dilution methods, TBW was also estimated as the sum of ICF TBK and ECF Br; ICF as the difference between the TBW D2O and ECF Br; and ECF as TBW D2O minus ICF TBK. These estimations are collectively termed IEM. Bioimpedance. TBW, ICF, and ECF were estimated by MF-BIS and by SF-BIA using the data obtained only at 50 kHz to allow comparison between SF BIA and MF BIS using the same device. An MF-BIS device (Xitron 4200, Xitron Technologies, CA) was used for measurements of BC using wrist-to-ankle measurements on the non-dialysis access side, with frequencies ranging from 5 kHz to 1 MHz. The current was injected through two electrodes placed on one wrist and the ipsilateral ankle. To allow equilibration of body fluids, patients were asked to position themselves supine for at least 15 min before the measurement. Each measurement was repeated at least 10 times, and the average was used in subsequent computations.11,24 According to the manufacturer’s manual the device that was used to obtain data from the studied subjects measures the resistance and the reactance for each of the 10 conducted measurement runs at all frequencies (from 5 kHz to 1 MHz) and calculates the reciprocal impedance at each measured frequency using the relationship

The resistance and reactance at 50 kHz (extracted from the raw data of the device output) were used to estimate TBW using an algorithm for SF-BIA published by Kotler and Wang.23,27 Estimation of TBK allowed, under the assumption of a constant intracellular potassium concentration of 152 mEq/l, the assessment of ICF as the estimated TBK divided by the assumed concentration. The coefficient of variation for each volume estimate by both methods was calculated based on estimations for each of the runs. The bioimpedance measurements were conducted at the same time as the blood draw for the DEM. The Appendix shows a detailed description of the regression models used for SF and MF BIA volume estimation.

Z 2 ¼ R2 þ Xc2 :

ð1Þ

In a next step the impedance derived from the measurement was fit to the Cole–Cole Model using iterative non-linear curve fitting software,11 employing the equation  Zobs ¼

RE RE þ RI

 RI þ

RE 1 þ ðjoCm ðRE þ RI ÞÞa



  joT  D : e

ð2Þ

where Zobs is the observed complex impedance, RE, RI, and CM are the component values of this circuit (extra- and intracellular resistance; cell membrane capacitance), TD is the frequency invariant time delay, caused by the speed at which electrical information is transferred through the conductor, o the frequency in radians/s ( ¼ 2pfrequency), and j is O  1,25,26 to identify extra- and intracellular resistance. 906

Statistical methods Shapiro–Wilk test was employed to test for normality. Parametric tests, paired and unpaired T-test as appropriate for continuous variables, and w2 test for categorical variables were employed. Each fluid compartment measured by both MF and SF BIA were compared with DEM based on the approach suggested by Bland and Altman.13 T-test was used for the assessment of potential systematic biases. Error analysis between DEM and IEM was performed using a modified BAA8 (plotted against the respective DEM value) and multivariable linear regression analysis. In the BAA we defined a significant proportional error as a significant regression of the difference between the respective analyzed methods as a function of the average. This error indicates an increasing error with an increasing or decreasing magnitude of the estimated volume. Fluid compartment measurements by DEM were compared with those obtained using IEM in a similar fashion. A P-value below 0.05 was considered significant. Analyses were conducted with R 3.0.128 additionally using the package hydroGOF to compute the RMSE, a marker combining accuracy and precision. DISCLOSURE

NWL and PK hold stock in Fresenius Medical Care NA. The remaining authors declared no competing interests. ACKNOWLEDGMENTS

The authors wish to acknowledge the scientific input of Professor Dympna Gallagher, St Luke’s-Roosevelt Medical Center, New York, USA, in her field of expertise.

AUTHOR CONTRIBUTIONS FZ, JW, MKK, GAK, PK, and NWL were involved in study design and initiation. FZ, GAK, ST, and JW were involved in the data acquisition and logistics of the study. JGR, NWL, PK, and GAK initiated the current analysis. JGR and GAK conducted the analysis and wrote the manuscript. All authors revised the manuscript and gave scientific input in their area of expertise. SUPPLEMENTARY MATERIAL Table S1. Regression equation and correlation coefficients of a) Regression and b) Bland–Altman Analysis for Total Body Water (TBW), Intra- and Extracellular fluid (ICF, ECF) as estimated by Multi- and Single-Frequency Bioimpedance (MF BIS, SF BIA), and the direct and indirect estimation methods (DEM, IEM). Table S2. Comparison of measures of distribution of fluid volume by direct or indirect estimation methods (DEM, IEM) or single or multifrequency bioimpedance in subjects stratified into a) older and b) younger than the median age of 54 years. Supplementary material is linked to the online version of the paper at http://www.nature.com/ki Kidney International (2014) 85, 898–908

clinical investigation

JG Raimann et al.: Comparison of methods of fluid volume estimation

REFERENCES 1.

2.

3.

4.

5.

6.

7.

8. 9.

10.

11.

12.

13.

14.

15. 16. 17.

18.

19.

20. 21.

22. 23.

24.

25.

Cha K, Chertow GM, Gonzalez J et al. Multifrequency bioelectrical impedance estimates the distribution of body water. J Appl Physiol 1995; 79: 1316–1319. Donadio C, Halim AB, Caprio F et al. Single- and multi-frequency bioelectrical impedance analyses to analyse body composition in maintenance haemodialysis patients: comparison with dual-energy x-ray absorptiometry. Physiol Meas 2008; 29: S517–S524. Quirk P, Ward LC, Thomas BJ et al. Multiple frequency bioelectrical impedance for the prediction of total body potassium in cystic fibrosis. Clin Nutr 1995; 14: 348–353. Simpson JA, Lobo DN, Anderson JA et al. Body water compartment measurements: a comparison of bioelectrical impedance analysis with tritium and sodium bromide dilution techniques. Clin Nutr 2001; 20: 339–343. Zdolsek JH, Lisander B, Hahn RG. Measuring the size of the extracellular fluid space using bromide, iohexol, and sodium dilution. Anesth Analg 2005; 101: 1770–1777. Carter M, Morris AT, Zhu F et al. Effect of body mass index (BMI) on estimation of extracellular volume (ECV) in hemodialysis (HD) patients using segmental and whole body bioimpedance analysis. Physiol Meas 2005; 26: S93–S99. Zhu F, Kuhlmann MK, Kaysen GA et al. Segment-specific resistivity improves body fluid volume estimates from bioimpedance spectroscopy in hemodialysis patients. J Appl Physiol 2006; 100: 717–724. Krouwer JS. Why Bland-Altman plots should use X, not (Y þ X)/2 when X is a reference method. Stat Med 2008; 27: 778–780. Kushner RF, Roxe DM. Bipedal bioelectrical impedance analysis reproducibly estimates total body water in hemodialysis patients. Am J Kidney Dis 2002; 39: 154–158. Gudivaka R, Schoeller DA, Kushner RF et al. Single- and multifrequency models for bioelectrical impedance analysis of body water compartments. J Appl Physiol 1999; 87: 1087–1096. De Lorenzo A, Andreoli A, Matthie J et al. Predicting body cell mass with bioimpedance by using theoretical methods: a technological review. J Appl Physiol 1997; 82: 1542–1558. Piccoli A, Pastori G, Guizzo M et al. Equivalence of information from single versus multiple frequency bioimpedance vector analysis in hemodialysis. Kidney Int 2005; 67: 301–313. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: 307–310. Hannan WJ, Cowen SJ, Plester CE et al. Comparison of bio-impedance spectroscopy and multi-frequency bio-impedance analysis for the assessment of extracellular and total body water in surgical patients. Clin Sci (Lond) 1995; 89: 651–658. Yasumura S, Wang JH, Pierson RN. In vivo body composition studies. New York Academy of Sciences: New York, 2000. Ellis KJ. Human body composition: in vivo methods. Physiol Rev 2000; 80: 649–680. Heymsfield SB, Wang Z, Visser M et al. Techniques used in the measurement of body composition: an overview with emphasis on bioelectrical impedance analysis. Am J Clin Nutr 1996; 64: 478S–484S. Zhu F, Schneditz D, Wang E et al. Validation of changes in extracellular volume measured during hemodialysis using a segmental bioimpedance technique. Asaio J 1998; 44: M541–M545. St-Onge MP, Wang J, Shen W et al. Dual-energy x-ray absorptiometrymeasured lean soft tissue mass: differing relation to body cell mass across the adult life span. J Gerontol A Biol Sci Med Sci 2004; 59: 796–800. Heymsfield S. Human body composition, 2nd edn. Human Kinetics: Champaign, IL, 2005. Pierson RN Jr., Wang J, Thornton JC et al. Body potassium by four-pi 40 K counting: an anthropometric correction. Am J Physiol 1984; 246: F234–F239. Pierson RN Jr., Lin DH, Phillips RA. Total-body potassium in health: effects of age, sex, height, and fat. Am J Physiol 1974; 226: 206–212. Wang Z, St-Onge MP, Lecumberri B et al. Body cell mass: model development and validation at the cellular level of body composition. Am J Physiol Endocrinol Metab 2004; 286: E123–E128. Matthie JR. Second generation mixture theory equation for estimating intracellular water using bioimpedance spectroscopy. J Appl Physiol 2005; 99: 780–781. Cole KS. Membranes, Ions and Impulses; A Chapter of Classical Biophysics. University of California Press: Berkeley, 1968.

Kidney International (2014) 85, 898–908

26. 27.

28.

Cole KS, Cole RH. Dispersion and absorption in dielectrics II. Direct current characteristics. J Chem Phys 1942; 10: 98–105. Kotler DP, Burastero S, Wang J et al. Prediction of body cell mass, fat-free mass, and total body water with bioelectrical impedance analysis: effects of race, sex, and disease. Am J Clin Nutr 1996; 64: 489S–497S. R Development Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing: Vienna, Austria, 2010.

APPENDIX Multi-frequency bioelectrical impedance spectroscopy

According to the model developed by De Lorenzo et al., based on the Cole–Cole Model and the Hanai Theory, calculation of intracellular fluid volume (VICF) (l) and total-body fluid (TBW) (l) requires the calculation of extracellular volume (VECF) (l) as

VECF ¼

40:5KB height2  1025Re

pffiffiffiffiffiffiffiffiffiffiffiffiffi!2=3 weight

ð3aÞ

for men and

VECF ¼

39:0KB height2  1025Re

pffiffiffiffiffiffiffiffiffiffiffiffiffi!2=3 weight

ð3bÞ

for women, where RE is the extracellular resistance (O) derived from model fitting using the Cole–Cole model, height the height of the subject (cm), weight the body mass (kg), and KB the geometry constant (assumed to be 4.3).11 Intracellular volume is then calculated as

VICF ¼ VECF

!   rMIX ðRi þ Re Þ 2=3 1 40:5Ri

4a

for men and VICF ¼ VECF

!   rMIX ðRi þ Re Þ 2=3 1 39:0Ri

ð4bÞ

for women, where 

rMIX

Ri ¼ 273:9  ð273:9  40:5Þ Ri þ Re

2=3 ð5aÞ

for men and 

rMIX

Ri ¼ 264:9  ð264:9  39:0Þ Ri þ Re

2=3 ð5bÞ

for women, where Re and Ri are the intra- and extracellular resistance, respectively, derived from model fitting using the Cole–Cole model. 907

clinical investigation

JG Raimann et al.: Comparison of methods of fluid volume estimation

In a final step TBW is calculated as (Equation 6) TBW ¼VECF þ VICF

Total-body potassium (TBK) (mmol) can be calculated as ð6Þ

"

# L1:60 TBK ¼ 0:76 0:50 ð59:06Þ þ 18:52weight  386:66 Xcp

8ðaÞ

Single-frequency bioelectrical impedance analysis

According to the model by Kotler et al.27 total-body water (TBW) can be calculated as  1:62  L 1:0 TBW ¼ 0:58 0:70 þ 0:32weight  0:86 1:35 Z

ð7aÞ

for men, and  1:99  L 1:0 TBW ¼ 0:76 0:58 þ 0:14weight  0:86 18:91 Z

ð7bÞ

for women, where L is the height (cm), Z the electrical impedance, and weight the body mass (kg).27

908

for men, and "

# L2:07 TBK ¼ 0:96 0:36 ð1:30Þ þ 5:79weight  230:51 Xcp

ð8bÞ

for women, where L is the height (cm), Xcp the reactance (O), and weight the body mass (kg).27 The ICV estimation based on the TBK (mmol) is conducted under the assumption that intracellular K þ concentration is considerably stable at around 152 mEq/l and allows, with the knowledge of TBW, the estimation of the intracellular K þ distribution volume.23

Kidney International (2014) 85, 898–908