Abstract
Abstract. Whether multifrequency bioelectrical impedance analysis (MF-BIA), a relatively new method for measuring body composition, is also applicable for accurate body composition measurements in renal transplant (RTx) patients is not known. Therefore, the use of MF-BIA is validated in 77 RTx patients with a stable renal function at least 2 yr posttransplantation. MF-BIA is compared to isotope dilution techniques for measurement of body water compartments, and to dual energy x-ray absorptiometry (DEXA) and anthropometry for measurement of fat and fat free mass. Finally, DEXA and anthropometry are compared to each other. Method agreement is assessed by intraclass correlation coefficients (ICC) and plotted by Bland and Altman analysis. MF-BIA significantly underestimates total body water (TBW, 0.7 ± 2.1 L) and overestimates the extracellular water (ECW, 3.3 ± 1.8 L) compared to isotope dilution; the ICC between both techniques is 0.943 for TBW and 0.846 for ECW. The percentage body fat (BF) measured by MF-BIA is significantly higher than both BF measured by DEXA (3.4 ± 4.7%) or by anthropometry (5.5 ± 5.2%). The ICC between MF-BIA and DEXA is 0.887 and between MF-BIA and anthropometry 0.856. BF measured by DEXA is significantly higher than BF measured by anthropometry (2.1 ± 4.4%); their ICC is 0.913. In conclusion, MF-BIA seems to be suitable for measurement of TBW in RTx patients; however, method agreement between isotope dilution and MF-BIA for the measurement of ECW is not satisfactory. In the assessment of fat and fat free mass, the reliability of MF-BIA appears to be questionable. Method agreement between DEXA and anthropometry seems to be slightly better.
Renal transplant patients are at risk for increased weight, centripetal obesity, and muscle atrophy because of their long-term glucocorticoid requirements (1,2,3,4,5). Such changes in body composition are associated with an increased risk of cardiovascular complications (1,6,7,8,9,10), a major cause of morbidity and mortality in renal transplant patients (11,12,13,14,15,16). Body composition data might provide insight into the relation with outcome, survival, and posttransplant complications; it might also affect approaches to nutritional therapy and to therapy in the field of physical activity.
Various techniques are available to measure body composition, among them isotope dilution, anthropometry, dual energy x-ray absorptiometry (DEXA), and multifrequency bioelectrical impedance analysis (MF-BIA). Multifrequency bioelectrical impedance analysis is recently introduced for measuring body composition. It is a relatively inexpensive, noninvasive, easy to use, and portable technique, and therefore suitable for routine clinical use.
MF-BIA is considered a useful technique for body composition analysis in healthy individuals (17). However, it is not known whether MF-BIA is also applicable for accurate body composition measurements in renal transplant patients. In this study, we validated MF-BIA for its use in renal transplant patients by comparing MF-BIA to isotope dilution, which is considered the gold standard for measuring body water compartments in healthy individuals. Furthermore, we compared MF-BIA, DEXA, and anthropometry to one another for measurement of the body fat and fat free compartment.
Materials and Methods
Patients
Renal transplant patients with a stable renal function and maintenance immunosuppressive therapy for at least 2 yr were studied. Exclusion criteria were insulin-dependent diabetes mellitus, metal implants (prostheses or pacemakers), and recent complications (e.g., malignancies or surgery). All patients were receiving immunosuppressive therapy, which consisted of combinations of prednisolone, azathioprine, and cyclosporine.
Informed consent was obtained from each patient, and the study was approved by the Ethics Committee of the Maastricht University Hospital.
Methods
In this cross-sectional study, body composition was measured by isotope dilution (deuterium (D2O) and potassium bromide (KBr)), DEXA, anthropometry, and MF-BIA (Figure 1).
Study design.
Isotope Dilution
In the early morning after an overnight fast, patients received an orally administered dose of D2O of 25 ml (99%, Sigma Chemicals, St. Louis, MO) and KBr of 30 ml (150 mM). Dose bottles were rinsed out and the rinse water was also ingested by the patients to ensure that all D2O and KBr was consumed. Enrichments of D2O and KBr in the body fluid were measured in serum. Immediately before D2O and KBr intake, a (background) blood sample was taken. After an equilibration time of 4 h, a second blood sample was collected. Food intake was not allowed until the second blood sample was collected; nevertheless, drinking some small amounts of water was allowed as all patients had to take their immunosuppressive medication just before or during measurements.
The concentration of deuterium and bromide in serum was determined by isotope ratio mass spectrometry and ion chromatography, respectively (18,19). D2O and bromide dilution spaces were calculated from, respectively, the enrichment of D2O and bromide after 4 h (20). Total body water (TBW) was calculated as the deuterium dilution space corrected for the exchange of D2O with nonaqueous compartments and for the concentration of water in the serum by first dividing the dilution space by 1.04 and thereafter multiplying it by 0.94 (20,21). The extracellular water compartment (ECW) was calculated as the bromide dilution space corrected for intracellular penetration of bromide in erythrocytes, leukocytes, and secretory cells; for unequal bromide concentrations in the extracellular fluids (Gibbs-Donnan effect); and for the concentration of water in the serum. Therefore, bromide dilution space was multiplied by 0.90 × 0.95 × 0.94, respectively (20,22,23,24,25).
Dual Energy X-Ray Absorptiometry
DEXA was used for measurement of whole-body composition, including fat mass, lean soft tissue mass (comprising muscle, inner organs, and body water), and bone mineral density. The equipment used in this study was DPX-L (Lunar Radiation Corp., Madison, WI).
DEXA measurements were performed in a standard manner while the patient was lying in a supine position on a table. From an x-ray source and K-edge filter below the patient, x-ray beams of stable energy radiation of 38 and 70 KeV were emitted. Attenuation of the x-rays was measured with a detector situated above the patient. Transverse scans of the body were made from top to toe. For each transverse scan, about 120 pixel elements with a size of approximately 5 × 10 mm yield data on the attenuation ratio. Approximately 40 to 45% of the pixels over the body contain bone and soft tissue, and 55 to 60% contain soft tissue alone (26). Bone mass was estimated from the ratio of the attenuation at low energy peaks relative to attenuation at high energy peaks through bone containing pixels after correction for the overlying soft tissue. The composition of the soft tissue was estimated by the ratio of beam attenuation at lower energy relative to the higher energy in soft tissue pixels; this ratio is inversely and linearly related to the percentage of fat (26,27).
Anthropometry
Body weight was measured to the nearest 0.1 kg using a balance scale. Body height was measured to the nearest 0.5 cm with the patient standing, back to a stadiometer.
Skinfold measurements were used to estimate total body fat mass. Skinfold thickness was measured to the nearest 0.1 mm by a skinfold caliper. Measurements were made at the nondominant side of the body at four sites: biceps, triceps, subscapula, and iliac crest. Skinfolds were measured three times. Thereafter, the sum of the individual skinfolds was averaged. The logarithm of the sum of the four skinfolds was used in age- and gender-specific regression equations of Durnin and Womersly (28) to compute body density (D). Fat mass was computed as: Body weight (kg) × [(4.95/D) - 4.5]; fat free mass (kg) was calculated as: Body weight (kg) - Fat mass (kg) (28,29).
Waist circumference was measured midway between the lower rib margin (costal margin) and the superior anterior iliac spine (iliac crest). Hip circumference was measured at the level of the great trochanters. Both waist and hip circumference were measured to the nearest millimeter with a flexible tape, with the patient in the standking position.
Multifrequency Bioelectrical Impedance Analysis
Bioelectrical impedance measurements were performed in a standard manner while the patient was lying supine on a flat, nonconductive bed. Multifrequency (5 to 500 kHz), imperceptible currents were introduced at distal electrodes on the hands (just proximal to the phalangeal-metacarpal joint in the middle of the dorsal side of the hand) and the feet (just proximal to the transverse [metatorsal] arch on the superior side of the foot), and resistances were measured by proximal electrodes (to the wrist midway between the styloid process, to the ankle midway between the malleoli). TBW and ECW were predicted from a general mixture theory (theory of Hanai) (30,31,32): Water compartments are directly calculated from resistance values, assuming specific resistances of ECW and ICW. Specific resistances of ECW and ICW are provided by the manufacturer (for men: ρECW = 215.0, ρICW = 824.0; for women: ρECW = 206.0, ρICW = 797.0). The bioelectrical impedance analyser used in this study was the Xitron 4000B (Xitron Technologies, San Diego, CA). Fat free mass (FFM) is derived from the ECW and ICW volume; fat mass is calculated as: Body weight - FEM (33).
To determine the reproducibility of MF-BIA measurements, the MF-BIA measurement was performed twice in a small number of patients (n = 10); precision was assessed by calculating the mean coefficient of variation of duplicate measurements (Coefficient of variation = [SD of the two measurements/Mean of the two measurements] × 100%).
Statistical Analyses
Paired t tests were used to compare TBW and ECW measured by isotope dilution to TBW and ECW measured by MF-BIA and to compare body fat (BF) measured by MF-BIA to BF measured by DEXA and anthropometry, respectively. Correlations between body compartments measured by different methods were estimated by the use of Pearson product moment correlations. A P value < 0.05 was considered significant. Bland and Altman plots (34,35) were used to visually assess agreement between the different methods to measure TBW, ECW, and body fat, respectively. In addition to these, the intraclass correlation coefficient (ICC) was calculated using variance components generated from repeated measures ANOVA. The intraclass correlation coefficient can be interpreted as a measure of agreement between methods and will vary from 0 (no agreement at all) to 1 (perfect agreement). Lowest acceptable method agreement was defined as 0.85. According to Fleiss (36), an approximate one-sided 95% lower confidence level for the ICC can be calculated to ascertain whether it satisfies the predetermined level of agreement. Smallest detectable differences are used to calculate the limits of agreement at a 5% level. To ascertain whether method agreement varied over important covariates or factors of the patient population, the included (among other things creatinine clearance, the use of calcium antagonists, and the presence of cystic kidneys in patients with polycystic kidney disease) intraclass correlation coefficients were also calculated for subcategories of covariates or factors to see whether method agreement increases or declines in certain subcategories. For this purpose, covariates were categorized into three classes containing one-third of the patients (about 22 to 25 per class), using cutting points as defined in Table 1. Statistical analysis was performed by SPSS for Windows, version 7.5.
Definition of cutting points used to obtain three subcategories in covariatesa
Results
Patient Characteristics
Seventy-seven renal transplant patients (35 women, 42 men) with a stable renal function participated in the study. Patient characteristics are shown in Table 2. Mean age of the population was 51.0 ± 11.7 yr. Body mass index (BMI) of the transplant patients was 24.8 ± 4.8 kg/m2. Mean follow-up time after transplantation was 9.1 ± 4.3 yr, and the mean creatinine clearance was 60.2 ± 20.5 ml/min. On average, patients used a cumulative prednisolone dose of 19.3 ± 21.5 g.
Patient characteristics (n = 77)
Reproducibility of MF-BIA Measurements
Repeated MF-BIA measurements were performed in 10 renal transplant patients (five women, five men), with a mean age of 59.0 ± 13.1 yr and a mean BMI of 23.7 ± 5.1 kg/m2. The mean coefficient of variation (CV) of the TBW measurements is 1.0% (range individual CV, 0.0 to 2.9%) and of the ECW measurements 0.6% (range individual CV, 0.0 to 1.9%).
Total Body Water
Of 73 renal transplant patients, data of D2O dilution are available; data of four patients were lost during laboratory analysis. Descriptive statistics (mean, SD, and range) of TBW measured by D2O dilution (TBWD2O) and TBW measured by MF-BIA (TBWMF-BIA) are shown in Table 3. TBWD2O compared with TBWMF-BIA is 34.2 ± 6.1 L versus 33.5 ± 5.9 L (P < 0.05). TBWD2O is highly and significantly correlated to TBWMF-BIA (r = 0.943, P < 0.001). The regression plot of TBWD2O and TBWMF-BIA is given in Figure 2. In Figure 3, the Bland and Altman analysis is plotted. The mean difference between TBWD2O and TBWMF-BIA is 0.7 L (95% confidence interval [CI], 0.2 to 1.2 L). The limits of agreement (mean ± 2 SD) show that TBWMF-BIA may be 3.4 L higher or 4.8 L lower than TBWD2O.
TBW, ECW, and BF measured by different techniques in renal transplant patients (mean, SD, and range)a
Regression plot of TBWD2O and TBWMF-BIA. Outliers (standardized z-residue scores < -2 or > 2) are given as open circles.
Assessment of TBW by D2O dilution against MF-BIA (Bland and Altman analysis). The relative bias (dilution — MF-BIA) plotted against the size of the measurement (mean of dilution and MF-BIA).
Table 4 shows the intraclass correlation coefficient for measurement of agreement between TBWD2O and TBWMF-BIA (ICCTBW) and the calculation of this ICC. ICCTBW is 0.943 (95% limits of agreement, -3.3 to 4.8 L). The approximate one-sided 95% lower confidence level for the ICC according to Fleiss (36) turns out to be 0.916. Table 6 shows the ICC values for the subcategories. The agreement between the methods appears to vary little over age, BMI, waist-to-hip ratio, hydration status of the lean body mass (LBM), the ratio ECW to ICW, creatinine clearance, gender, or the use of calcium antagonists. All ICC values are larger than 0.892. The ICCTBW in patients with cystic kidneys is 0.886, whereas the ICCTBW in patients without cystic kidneys seems to be higher (0.950).
Analysis of variance for 73 renal transplant patients measured twice on TBW (by D2O dilution and MF-BIA) supplemented with estimation of variance componentsa
Subcategory intraclass correlation coefficients for measurement of TBWa
Extracellular Water
Data of bromide dilution are available for 72 renal transplant patients (data of five patients were lost during laboratory analysis). Descriptive statistics of extracellular water measured by bromide dilution (ECWKBr) and ECW measured by MF-BIA (ECWMF-BIA) are shown in Table 3. ECWKBr compared with ECWMF-BIA is 15.5 ± 2.9 L versus 18.7 ± 3.6 L (P < 0.05). ECWKBr is highly and significantly correlated to ECWMF-BIA (r = 0.865, P < 0.001). The regression plot of ECWKBr and ECWMF-BIA is given in Figure 4. The Bland and Altman analysis is plotted in Figure 5. The mean difference between ECWKBr and ECWMF-BIA is 3.3 L (95% CI, -3.7 to -2.8 L). Limits of agreement show that ECWMF-BIA may be 0.4 L lower or 6.9 L higher than ECWKBr.
Regression plot of ECWKBr and ECWMF-BIA. Outliers (standardized z-residue scores < -2 or > 2) are given as open circles.
Assessment of ECW by KBr dilution against MF-BIA (Bland and Altman analysis). The relative bias (dilution — MF-BIA) plotted against the size of the measurement (mean of dilution and MF-BIA).
The ICC for measurement of ECW by KBr dilution and MF-BIA (ICCECW) plus the way it is calculated is given in Table 5; ICCECW turns out to be 0.846 (95% limits of agreement, -6.8 to 0.3 L). This is below the predetermined acceptable method agreement. The approximate one-sided lower 95% confidence level for the ICC according to Fleiss (36) is 0.779. Subcategory ICC values are given in Table 7. Many of the subcategories do not meet the lowest acceptable method agreement. Remarkably, method agreement appears to be acceptable in female patients, but not in male patients.
Analysis of variance for 72 renal transplant patients measured twice on ECW (by KBr dilution and MF-BIA) supplemented with estimation of variance componentsa
Subcategory intraclass correlation coefficients for measurement of ECWa
Body Fat and Fat Free Mass: MF-BIA Compared to DEXA
For 75 renal transplant patients, DEXA total body scans are available. Descriptive statistics of body fat (BF, percentage of total body weight) measured by DEXA (BFDEXA) and BF measured by MF-BIA (BFMF-BIA) are shown in Table 3. BFDEXA compared with BFMF-BIA is 30.3 ± 10.5% versus 33.7 ± 9.2% (P < 0.05). BFDEXA and BFMF-BIA are highly and significantly correlated (r = 0.895, P < 0.001). The regression plot of BFDEXA and BFMF-BIA is given in Figure 6. In Figure 7, the Bland and Altman analysis is plotted. The mean difference between BFDEXA and BFMF-BIA is 3.4% (95% CI, -4.5 to -2.3%). Limits of agreement show that BFMF-BIA may be 6.0% lower or 12.8% higher than BFDEXA.
Regression plot of BFDEXA and BFMF-BIA. Outliers (standardized z-residue scores < -2 or > 2) are given as open circles.
Assessment of BF by DEXA against MF-BIA (Bland and Altman analysis). The relative bias (DEXA — MF-BIA) plotted against the size of the measurement (mean of DEXA and MF-BIA).
The intraclass correlation coefficient for method agreement between BFDEXA and BFMF-BIA (ICCBF-D) is 0.887 (95% limits of agreement, -12.6 to 5.8%). According to the Fleiss formula (36), the approximate one-sided 95% lower confidence level for the ICC turns out to be 0.836. The subcategory ICC values are shown in Table 8. Method agreement appears to be acceptable in most of the subcategories. Especially in patients with cystic kidneys, in patients with a BMI < 22.5 kg/m2, and in patients with a high waist-to-hip ratio, method agreement appears to be below the predetermined lowest acceptable method agreement. Because fat free mass is calculated as 100% minus fat mass both in DEXA and in MF-BIA, the results of ICC and thus of method agreement with regard to fat free mass are identical.
Subcategory intraclass correlation coefficients for measurement of percentage body fat by DEXA and MF-BIAa
Body Fat and Fat Free Mass: MF-BIA Compared to Anthropometry (Skinfolds)
Anthropometric data are available for 74 renal transplant patients (in three patients it was not possible to measure skinfold thickness of the iliac crest because of multiple scars). Descriptive statistics of the percentage body fat measured by anthropometry (BFanthr) and BFMF-BIA are shown in Table 3. BFanthr compared with BFMF-BIA is 27.9 ± 10.2% versus 33.4 ± 9.2% (P < 0.05). BFanthr is highly and significantly correlated to BFMF-BIA (r = 0.860, P < 0.001). The regression plot of BFanthr and BFMF-BIA is given in Figure 8. The Bland and Altman analysis is plotted in Figure 9. The mean difference between BFanthr and BFMF-BIA is 5.5 ± 5.2% (95% CI, -6.7 to -4.3%). Limits of agreement show that BFMF-BIA may be 15.9% higher or 4.9% lower than BFanthr.
Regression plot of BFanthropometry and BFMF-BIA. Outliers (standardized z-residue scores < -2 or > 2) are given as open circles.
Assessment of BF by anthropometry against MF-BIA (Bland and Altman analysis). The relative bias (anthropometry — MF-BIA) against the size of the measurement (mean of anthropometry and MF-BIA).
The intraclass correlation coefficient for method agreement of BF by anthropometry and by MF-BIA (ICCBF-A) is 0.856; because this ICC lies only a fraction above the predetermined lowest acceptable method agreement, method agreement between anthropometry and MF-BIA for measuring body fat appears to be questionable. The approximate one-sided lower 95% confidence level for the ICC turns out to be 0.793. Table 9 shows the subcategory ICC values. Also in most subcategories method agreement is poor. Because fat free mass is calculated as 100% minus fat mass both in anthropometry and in MF-BIA, the results of ICC and thus of method agreement with regard to fat free mass are identical.
Subcategory intraclass correlation coefficients for measurement of percentage body fat by anthropometry (skinfolds) and MF-BIAa
Body Fat and Fat Free Mass: DEXA Compared to Anthropometry (Skinfolds)
DEXA total body scans and anthropometric data are available for 72 renal transplant patients. Descriptive statistics of BFDEXA and BFanthr are shown in Table 3. BFDEXA compared with BFanthr is 30.2 ± 10.7% versus 28.1 ± 10.3% (P < 0.05). BFDEXA is highly and significantly correlated to BFanthr (r = 0.913, P < 0.001). The regression plot of BFDEXA and BFanthr is given in Figure 10. In Figure 11, the Bland and Altman analysis is plotted. The mean difference between BFDEXA and BFanthr is 2.1% (95% CI, 1.1 to 3.1%); limits of agreement show that BFanthr may be 6.7% higher or 10.9% lower than BFDEXA.
Regression plot of BFDEXA and BFanthropometry. Outliers (standardized z-residue scores < -2 or > 2) are given as open circles.
Assessment of BF by DEXA against anthropometry (Bland and Altman analysis). The relative bias (DEXA — anthropometry) plotted against the size of the measurement (mean of DEXA and anthropometry).
The intraclass correlation coefficient for method agreement between BFDEXA and BFanthr (ICCBF-DA) turns out to be 0.913 (95% limits of agreement, -6.5 to 10.7%). The approximate one-sided 95% lower confidence level for the ICC is 0.872. The subcategory ICC values are given in Table 10. In most of the subcategories, the ICC is larger than 0.85; however, especially in patients with a BMI < 22.5 kg/m2, patients with a high waist-to-hip ratio, and patients with a low hydration status of the LBM, method agreement does not meet the predetermined acceptable level. The results of ICC and thus of method agreement with regard to fat free mass are identical.
Subcategory intraclass correlation coefficients for measurement of percentage body fat by DEXA and anthropometry (skinfolds)a
Discussion
In this study, we assessed in the first place the reproducibility of MF-BIA measurements, in the second place method agreement between MF-BIA and isotope dilution, DEXA, and anthropometry, and finally method agreement between DEXA and anthropometry in renal transplant patients.
Reproducibility of MF-BIA Measurements
The reproducibility of the MF-BIA measurements was assessed by calculating coefficients of variation (CV) of duplicate measurements. The CV of the TBW measurements was 1.0%, and the CV of the ECW measurements was 0.6%. This suggests a high reproducibility of the TBW and ECW measurements by MF-BIA, also reported by others (38,39,40).
Body Water Compartments: MF-BIA Compared to Isotope Dilution
MF-BIA is based on the different conductive and dielectric properties of various biologic tissues at various frequencies of current. Blood, muscles, and other tissues that contain a lot of water and electrolytes are highly conductive; in contrast, bone, fat tissue, and air-filled spaces are highly resistive. Currents will flow predominantly through materials with higher conductivities (41). The volume of the various tissues can be deduced from measurement of their resistances. In MF-BIA, the human body is considered as five conductive cylinders (i.e., the trunk, two arms, and two legs), which are connected in series. In conductors connected in series, the conductor with the smallest cross-sectional area (i.e., the arm) will determine most of the resistances in series. Although the trunk comprises nearly 50% of the total body weight, the trunk (conductor with the largest cross-sectional area) may have only little influence on the whole body resistance (40,41).
MF-BIA measures body impedance at a wide range of frequencies. The advantage of MF-BIA above the classic single frequency bioimpedance measurement is the possibility of MF-BIA to discriminate between the extracellular and intracellular water compartment. At low frequencies, cell membranes act like a condenser and completely block the flow of the current through the intracellular pathway. At higher frequencies, the condenser function of the cell membranes is lost, and the current flows to both the extra- and intracellular compartments and thus to the total body water (30,31,42). Another advantage of MF-BIA is that MF-BIA predicts the volume of TBW and ECW from a general mixture theory (theory of Hanai) (30,31,32). Water compartments are directly calculated from resistance values, assuming specific resistances of ECW and ICW. So, empirically derived prediction formulas are no longer necessary for the estimation of TBW and ECW.
In this study, we validated MF-BIA against D2O and KBr isotope dilution, a technique that is considered to be the gold standard for measurement of body water compartments in healthy subjects. The accuracy of isotope dilution for measurement of TBW is excellent; the accuracy is dependent on the uncertainty of the estimate of nonaqueous exchange, which is about 1% (20). For TBW measurements by continuous flow isotope ratio mass spectrometry, used in this study, the accuracy is 1.6% (18). The accuracy for the measurement of ECW by isotope dilution is about 5% (19).
Compared with D2O dilution, MF-BIA underestimates TBW by 0.7 ± 2.1 L. ICCTBW and the ICC values of the various subgroups are all >0.886. These high ICC values suggest good method agreement between D2O dilution and MF-BIA for measuring TBW; therefore, for measuring TBW the D2O dilution technique might be substituted by MF-BIA.
Dihydropyridine calcium antagonists, frequently prescribed for treatment of hypertension in renal transplant patients, can lead to peripheral edema. We investigated whether the use of these calcium antagonists influenced the reliability of MF-BIA; the ICCTBW values in the group of patients using calcium antagonists and in the group of patients not using this medication are almost equal: 0.943 and 0.940, respectively. So, the use of calcium antagonists does not correlate with the method agreement for TBW.
We also investigated whether the presence of polycystic kidneys in the abdomen is related to method agreement. In patients with polycystic kidneys, the trunk contains a higher amount of body water than in patients without cystic kidneys. As mentioned earlier, in MF-BIA the trunk contributes hardly to the whole-body resistance, which is a major parameter in the estimation of body water compartments by MF-BIA (40,41). Therefore, in patients with polycystic kidney disease, MF-BIA might underestimate the total body water compared to isotope dilution, and method agreement might be lower. The ICCTBW between isotope dilution and MF-BIA in patients without polycystic kidneys is excellent (0.950); in patients with cystic kidneys the ICCTBW appears to be somewhat lower, but still acceptable (0.886). So, the presence of cystic kidneys seems to be related to the method agreement between isotope dilution and MF-BIA for TBW.
Compared to KBr dilution, MF-BIA overestimates ECW by 3.3 ± 1.8 L. ICCECW is 0.846 and thus below the lowest acceptable method agreement. The ICC values in most of the subgroups of patients do not meet the acceptable level of method agreement either. The low method agreement in male patients (ICCECW = 0.817) compared to the acceptable method agreement in female patients (ICCECW = 0.877) is remarkable. An explanation for this acceptable method agreement in female transplant patients cannot be given as yet. So, in renal transplant patients method agreement for measuring ECW by KBr dilution and by MF-BIA is not acceptable; method agreement is only reached in some very specific subgroups of patients. However, these results should be interpreted with caution. As discussed before, TBW determination by MF-BIA is satisfactory. TBWMF-BIA is calculated as the sum of ECW and ICW. In contrast to the ICW, the ECW is measured directly by MF-BIA (33). The excellent agreement between TBWD2O and TBWMF-BIA suggests that some of the disagreement between ECWKBr and ECWMF-BIA might be due to errors with the bromide dilution technique. Errors with the bromide isotope dilution technique, especially in patients with an abnormal ECW, might be due to the uncertainty in the correction constants for plasma water and Gibbs-Donnan equilibration as well as to penetration into the intracellular space due to changes in plasma protein concentrations and hematocrit (20).
Body Fat: MF-BIA Compared to DEXA
MF-BIA derives fat free mass from the measured volume of TBW. Fat mass is calculated as body weight minus fat free mass. In MF-BIA, it is assumed that fat is anhydrous and that the lean body mass contains a relatively constant proportion of water (73.2%), a proportion that is assumed to be true in healthy individuals (43). A disturbed water status can significantly reduce the accuracy of predictions of fat free mass and fat mass by MF-BIA. In renal transplant patients, the assumption of a constant hydration status is not invariably true.
As mentioned earlier, MF-BIA measures TBW with sufficient accuracy in renal transplant patients. The hydration status of the lean body mass (calculated as TBWD2O divided by lean body mass DEXA) in our study population is 0.709 ± 0.04 L/kg LBM (range, 0.596 to 0.836 L/kg LBM) and significantly different from the assumed hydration status. Therefore, MF-BIA probably overestimates the actual percentage of body fat.
We compared the values of BFMF-BIA to the values of BFDEXA. DEXA is a very useful technique for directly assessing soft tissue as well as bone; it has been shown to be of relatively high accuracy and very high precision (26). However, DEXA is not assumption-free: DEXA too assumes that the hydration status of the LBM is uniform and fixed at 0.732 L/kg LBM (26,27,44). The degree to which DEXA measurements of soft tissue are sensitive to changes in hydration in adult humans is unknown, and therefore further research is needed. Hence, DEXA cannot be considered as the gold standard for measuring body fat at this moment.
Compared to DEXA, MF-BIA overestimates the percentage of body fat by 3.7 ± 4.7%. The original ICC for method agreement between BFDEXA and BFMF-BIA is 0.887. So, method agreement seems acceptable. However, when our renal transplant population is divided into males and females, method agreement is unacceptable in both groups; the reason for this is that given an equal patient-by-method error, the ICC is generally lower within relatively homogeneous subgroups. The ICC values in some other specific subcategories also do not meet the acceptable level of method agreement. An explanation for the low method agreement in these groups cannot be given as yet. So, in renal transplant patients, method agreement between DEXA and MF-BIA for measuring body fat is not acceptable; therefore, results of DEXA and MF-BIA fat measurements may not be substituted for each other.
Body Fat: MF-BIA Compared to Anthropometry
We also compared MF-BIA to anthropometry (skinfolds) for measuring body fat. Skinfold thickness measurements are rapid, simple, noninvasive, and inexpensive. In skinfold thickness measurements, it is assumed that the majority of body fat resides in subcutaneous regions, that there is a consistent relationship between subcutaneous and visceral fat, and that body fat distribution is stable (45,46,47). It is also assumed that the density of the lean body mass is normal, i.e., that the hydration status of the LBM is normal (0.732 L/kg LBM) and that the bone mineral content represents a fixed fraction of the LBM (47,48). The reliability of skinfold measurements is approximately 5%; it largely depends on the sites of the skinfold measurements and the experience of the examiner (45). Anthropometric evaluation of body fat is most appropriate for population surveys. Skinfold measurements cannot be considered a gold standard for measurement of body fat content.
Compared to anthropometry, MF-BIA overestimates the percentage body fat in renal transplant patients by 5.5 ± 5.2%. The original ICC for measurement of agreement between BFanthr and BFMF-BIA is 0.856 and thus only a fraction above the lowest acceptable level of method agreement. Therefore, method agreement seems to be questionable. Both in men and women the ICC for method agreement is unacceptable. In most of the other subgroups, ICC values also indicate questionable, poor, or even very poor method agreement. The reason(s) for the low method agreement in these subgroups are not known.
Body Fat: DEXA Compared to Anthropometry
At present, DEXA is used more and more for the measurement of total body fat content. However, as mentioned earlier, DEXA is not considered the gold standard for measurement of body fat. We finally compared DEXA to the relative simple anthropometric (skinfold thickness) measurements. Compared to DEXA, anthropometry underestimates the percentage of body fat in renal transplant patients by 2.1 ± 4.4%. Hart et al. (49) also reported that skinfold measurements underestimate the percentage of body fat compared to DEXA in renal transplant patients. The original ICC for method agreement between DEXA and anthropometry found in our study is high, i.e., 0.913. So, method agreement seems to be very good. Nevertheless, there are a few subcategories of renal transplant patients in which the ICC is below the predetermined acceptable level and in which method agreement thus is questionable. In patients with a low (<22.5 kg/m2) and high (>26.0 kg/m2) BMI, the poor method agreement can be explained by the fact that measurement of skinfolds in very lean and very obese subjects (who have respectively a low or high BMI) is technically difficult; therefore, the validity of skinfold measurements in these groups could be affected (50). In patients with a high waist-to-hip ratio (apple configuration), the abnormal body fat distribution might play a role. Skinfold measurements, in which a stable body fat distribution is assumed, might be invalid in this patient group. In patients with a low hydration status (<0.6900 L/kg), method agreement is probably low because both anthropometry and DEXA assume a normal hydration status (0.732 L/kg LBM).
Method Comparison for the Measurement of Fat Free Mass
In MF-BIA and anthropometry (skinfold thickness measurements), the body is considered to be composed of two compartments: the fat compartment and the fat free compartment (bone mass included) (41,47). In DEXA, the body is divided into the fat compartment, the lean soft tissue compartment, and the bone mineral content (fat free mass equals lean soft tissue mass plus bone mineral content) (27). Thus, in MF-BIA as well as in DEXA and anthropometry the percentage of fat free mass equals 100% minus the percentage of body fat. Therefore, when the percentage of body fat is overestimated in the method comparisons, the percentage of fat free mass will be underestimated, and conversely, when the percentage of body fat is underestimated, the percentage of fat free mass will be overestimated. Furthermore, the conclusions about the degree of method agreement between MF-BIA, DEXA, and anthropometry for the measurement of body fat are the same for the measurement of fat free mass.
In conclusion, MF-BIA seems to be suitable to measure TBW with sufficient accuracy in renal transplant patients. The agreement between ECWKBr and ECWMF-BIA is not satisfactory. However, possible limitations of the bromide isotope dilution technique have to be taken into account.
Because neither DEXA nor anthropometry can be considered a gold standard, the interpretation of the results of the method comparisons regarding fat and fat free mass should be done with caution. Still, the poor agreement for measuring body fat and fat free mass between MF-BIA and DEXA and between MF-BIA and anthropometry, and the reasonable agreement between DEXA and anthropometry, makes the reliability of MF-BIA in the assessment of fat and fat free mass in renal transplant patients questionable.
Footnotes
American Society of Nephrology
- © 1999 American Society of Nephrology
References
