Over the last few decades, continuous advances in dialysis technologies have increased the safety and efficacy of hemodialysis (HD) . At present, many of the common acute complications during HD are related to fluid status and impaired compensatory mechanisms. In particular, the accuracy to measure changes in intradialytic fluid status and to assess the fluid overload remains a problem. Inaccuracy might result in excessive volume depletion during treatment, and possibly, the development of intradialytic hypotension (IDH) due to excess drop in systolic blood pressure. IDH increases patient morbidity (e.g. via nausea, vomiting, headaches and muscle cramps), reduces treatment efficacy, and increases the risk of mortality .
In recent years, multi-frequency bioimpedance spectroscopy (BIS) techniques have become more clinically attractive as they enable non-invasive, simple, and relatively inexpensive tools to assess continuously the fluid status during HD . A high level of accuracy for whole-body BIS (whBIS) compared with gold standards (e.g. dilution methods) have been reported in literature [4,5]. As the accuracy and reproducibility of whBIS can be influenced by external factors, such as changes in body position, the sum of segmental BIS bioimpedance (sBIS) technique was introduced [6,7]. Yet, the application of whBIS or sBIS for routine clinical use is sometimes impractical, as it requires a relatively complex implementation. Patients have to be motionless in a fixed position with arms and legs extended during the whole treatment and extra cables and electrodes are required. This is uncomfortable for many patients and costly, especially as it requires continuous surveillance to ensure intraindividual reproducibility. Moreover, correct placement of electrodes can sometimes be difficult due to the dialysis fistula or other medical conditions. As extremities account for up to 88% of the total body impedance , alternative configuration of BIS, namely knee-to-knee BIS (kkBIS)  was introduced. This new configuration should be reliable for continuous applications while avoiding the need for complex setups as required for whBIS or sBIS. To our knowledge, the adequacy of kkBIS so far has only been assessed in hemodynamically stable patients.
This study aimed at comparing the adequacy of kkBIS with whBIS in respect to continuous monitoring of extracellular fluid volume in hemodynamically unstable HD patients, and the usefulness of kkBIS for detecting the patient’s risk of IDH. The correlation between whBIS and kkBIS on the basis of changes in absolute extracellular resistance (Re) and Re normalized for ultrafiltration volume (ΔRe/UFV) has been investigated. Finally, volume calculations from both methods were compared and their cumulative changes were investigated against collected ultrafiltrate.
Materials and methods
Twenty HD patients (13 males and 7 females; mean ages 62 ± 16 years) were studied twice in the dialysis unit of the Department of Nephrology, RWTH Aachen University Hospital. All patients underwent HD for four hours with a fixed ultrafiltration rate of 560 ± 265 ml/h and a dialysate fluid composition of 138 mmol/L sodium (Na+), 2.1 ± 0.6 mmol/L potassium (K+), 1.25 mmol/L calcium (Ca2+), 5.5 mmol/L glucose (Glu), and 32 mmol/L bicarbonate (HCO3-). The ultrafiltration rate was only changed if required for clinical reasons. The inclusion criteria were: in- and outpatients ≥ 18 years of age on chronic HD three times a week, who had been on dialysis for a minimum of six months. The exclusion criteria were HIV or hepatitis virus C infection, pregnancy, pacemakers, amputation of a limb, artificial joints, or receiving a blood transfusion within a week before the measurement.
Obtained measurements included the patient’s pre- and post-dialysis weight, height, thigh length and circumferences, systolic and diastolic blood pressure, ultrafiltration rate and volume (UFV), plasma electrolytes, and BIS measurements.
Bioimpedance measurements were obtained using a multi-frequency bioimpedance device (Xitron Hydra 4200, Xitron Technologies Inc., San Diego, CA, USA) and standard hydrogel-aluminium BIS electrodes placed on the non- dialysis access side of the patient according to the manufacturer’s instruction as given in Fig. 1. For whBIS measurement, the position of the electrodes was chosen according to the manufacturer’s instruction, while for kkBIS, a distance of 5 cm was always kept between the voltage and current electrodes to ensure reproducibility of the measurement. Impedance data for kkBIS and whBIS were recorded sequentially with an interval of less than one minute between the two measurements.
Extracellular resistance (Re) for each BIS method were extracted by fitting the measured impedance data to the Cole model , while total body extracellular fluid (ECFwhBIS) was calculated as described elsewhere [11,12]. Knee-to-knee extracellular fluid was calculated according to either Matthie  and Medrano et al.  (ECFkkBIS), or as described elsewhere in [15,16] (ECFkkBIS,Z).
Blood samples were drawn from the arterial line and analyzed using an automated benchtop blood gas analyzer (ABL800 Basic, Radiometer Medical ApS, Brønshøj, Denmark) for plasma electrolytes and glucose concentrations. Systolic and diastolic blood pressures were measured using the blood pressure monitor integrated in the dialysis machine. All measurements were recorded every 30 min in the supine position with patients resting on a bed with their legs apart and arms placed away from the body.
Study Protocol and Data Analysis
Patients were post-dialytically divided into a hemodynamiccally stable group, which included patients with a stable systolic blood pressure throughout HD (drop < 25 mmHg), whereas the unstable group included patients with a continuous symptomatic drop in systolic blood pressure of ≥ 25 mmHg. For this study, the clinical symptoms considered relevant in the context of IDH were diaphoresis, nausea, vomiting, edema, cramps, blurred vision, presyncope, headache, and fatigue. Medical interventions such as fluid infusion, switching off ultrafiltration, and/or raising of the patient’s legs were applied as necessary.
The difference between the groups at baseline was tested by Student’s t-test or Pearson’s Chi-square test, as appropriate. Regression analysis was applied to study the agreement between the whBIS and kkBIS method, and to assess the correlation between the pre- and post-dialytic changes in Re and patient’s weight for both methods. For comparison between the patient groups, relative changes in extracellular resistance normalized for the UFV (ΔRe/UFV in [%/L]) were used. In all analyses, a p value ≤ 0.05 was considered statistically significant.
Informed consent has been obtained from all individuals included in this study.
The research related to human use has been complied with all relevant national regulations, institutional policies and in accordance with the tenets of the Helsinki Declaration, and has been approved by the local medical ethics committee.
The mean age of the patients was 62 ± 16 years with a body height of 170 ± 9 cm and 35% of patients being female. Twenty patients each were categorized as stable or unstable during HD. Table 1 summarizes the baseline clinical characteristics of the study population and their treatments. Apart from the whole-body extracellular volume ECFwhBIS (p = 0.009), there was no significant differences between the groups at baseline with regard to their clinical characteristics and dialysis settings (Table 1) or composition of dialysate fluid (data not shown). If ECFwhBIS were to be normalized to baseline patient’s weight, no significant difference could be observed (p = 0.928). Baseline thigh’s ECF value did not contribute as well to the hemodynamic stability of the patient with p = 0.769, or if normalized to weight, with p = 0.200.
Pre-dialysis physical and clinical characteristics. Data are means ± SD from 20 patients (40 measurements).
|Stable (n = 10)||Unstable (n = 10)||p|
|Gender, % female||20.0||50.0||0.350|
|Age, years||63.3 ± 15.8||60.7 ± 16.4||0.722|
|Height, cm||174.7 ± 7.9||165.8 ± 9.0||0.030|
|Weight, kg||83.6 ± 8.3||69.5 ± 17.3||0.038|
|BMI, Kg / m2||27.5 ± 3.4||25.4 ± 6.8||0.398|
|Dialysis duration, min||234.0 ± 19.0||222.0 ± 29.0||0.290|
|Ultrafiltration Rate, ml / h||557.5 ± 305.7||564.2 ± 235.1||0.957|
|UFV, mL||2252.0 ± 1226.9||2300.0 ± 1091.4||0.927|
|SysBP, mmHg||123.7 ± 17.2||125.4 ± 19.1||0.836|
|DiaBP, mmHf||71.5 ± 14.6||64.1 ± 11.6||0.226|
|ECFwhBIS, L||21.6 ± 2.4||17.3 ± 3.9||0.009|
|ECFkkBIS, L||2.7 ± 0.5||2.6 ± 0.7||0.769|
|Body temperature, °C||36.0 ± 0.6||35.9 ± 0.4||0.688|
|O2 saturation, %||96.9 ± 2.4||96.6 ± 2.7||0.796|
|Plasma Na+, mmol / L||136.8 ± 2.1||136.7 ± 2.7||0.928|
|Plasma K+, mmol / L||4.4 ± 0.7||4.7 ± 1.0||0.530|
|Plasma ionized Ca2+, mmol / L||1.1 ± 0.1||1.1 ± 0.1||0.461|
|Plasma glucose, mmol / L||121.4 ± 27.8||124.6 ± 31.3||0.809|
Clinical and physical assessment of the patients revealed that 35% of the patients showed signs of pre-dialysis overhydration mainly in the thigh and lower leg with a mean systolic blood pressure of 135 ± 9.6 mmHg. Intradialytic mean weight loss was 2.1 ± 1.0 kg in the stable group and 1.8 ± 0.9 kg in unstable patients (p = 0.499). Systolic pressure decreased by 14.8 ± 9.5 mmHg in stable and by 20.1 ± 14.8 mmHg in unstable patients (p = 0.356).
The correlation of relative changes in ΔRe and in ΔRe/UFV between the whBIS and kkBIS method is given in Fig. 2(a-b). Intradialytic changes in ΔRe (%) correlated very well between the methods (R2 = 0.86, p < 0.001) for both groups. If normalized for UFV, the correlation was slightly improved to R2= 0.92, p < 0.001. Loss in the thigh’s extracellular fluid (ECFkkBIS) also correlated with the total body loss in ECF (R2 = 0.88, p < 0.001) for stable patients and to a lesser extent (R2 = 0.51, p < 0.001) for unstable patients. This correlation persisted, when the thigh ECF was estimated using a specific resistivity value for the legs (ECFkkBIS,Z). Nevertheless, ECFkkBIS,z values for all patients were slightly higher (0.25 ± 0.17 L) in comparison to those (ECFkkBIS: 0.20 ± 0.07 L) calculated using a fixed resistivity value for the whole body.
The relative change in ΔRe/UFV in [%/L] during treatment is shown in Fig. 3(a) for kkBIS and in Fig. 3(b) for whBIS; their ratio is presented in Fig. 3(c). The relative changes in thigh and total body ECF with respect to pre-HD values are given in Fig. 3(d). Within the first hour of HD, the loss of thigh ECFkkBIS was 7.7 ± 3.3 % in stable patients compared to 3.5 ± 2.3 % loss of total body ECF. In unstable patients, thigh ECFkkBIS decreased by only 1.8 ± 1.5 % compared to 6.4 ± 2.4 % in total body ECF. Only at the end of HD, the decrease in thigh ECFkkBIS (9.1 ± 3.7 %) approached that of total body ECF (11.1 ± 3.4%) for unstable patients.
Linear regression analysis revealed a good relationship between relative intradialytic weight loss and changes in Re (for both whBIS and kkBIS), Fig. 4(a-b). Similar results were obtained for ΔECF compared to the collected UFV as shown in Fig. 4(c-d). Nevertheless, data revealed a significantly larger relative change in Re normalized to wt (pre-HD to post-HD) for unstable patients. Moreover, if we consider the difference in (ΔRe,whBIS - ΔRe,kkBIS)/Δwt between pre-HD to post-HD, stable patients would be characterized with a significant lower mean difference of 0.51 ± 0.89 compared to 1.1 ± 1.20 for unstable ones (p < 0.001).
Compared to stable patients, the accumulative loss in thigh ECFkkBIS was significantly lower in unstable patients especially within the first hour of the treatment (p = 0.004 after 60 min). Fig. 5(a-b) depicts the accumulative changes in ECF, estimated either from kkBIS or whBIS method, and compared to the accumulative UFV during HD treatments in both patient groups. Fig. 5(c) presents their ratio for both patient groups.
Discussion and conclusion
In the present study, we compared the application of the kkBIS with the standard whBIS method to assess the extracellular fluid status in a cohort of symptomatic unstable versus stable HD patients. kkBIS not only increases the comfort of the patient but was also found to correlate very well with whBIS. Furthermore, the intradialytic changes in parameters (as ΔRe/UFV and ΔRe/wt) obtained via kkBIS might be used to identify patients at risk of IDH.
Similar to previous work , our data confirms that baseline characteristics do not allow differentiating stable from unstable patients. In order to compare kkBIS with the traditional method whBIS, the first step was to compare the measured intradialytic changes in Re and Re normalized to UFV. Regression analysis revealed an excellent correlation between both methods, which is in agreement with the work of Cox-Reijven et al. . The next step was to compare the estimated ECF from measured Re values. Here, lower agreement (R2 > 0.50) between the methods was observed for the unstable group in particular. This could be due to the model applied to make segmental volume calculations. The estimation of ECF volumes is subject to some uncertainty concerning interpretation of the collected impedance data and a number of simplifications as the use of constant resistivity derived from a healthy population [11, 19]. Nevertheless, it is worth noting that in our study, segmental volume calculations using a specific resistivity value for legs fully correlated with those done using a constant resistivity value for the whole body, Fig. 2(d).
The relative change in extracellular resistance (Re) normalized to UFV could be another way to identify unstable patients, without being subject to possible inaccuracies arising from calculating ECF volumes, Fig. 3(a-b). This could be of particular interest especially when profiled ultrafiltration rate is applied. The ECF transport rate from legs to the trunk plays a pivotal role in this refilling process [20, 21]. Shulman et al.  found that legs contributed up to 0.69 L to the change in total body ECF of 1.33 L as measured by whBIS during 18 dialysis sessions with a mean UFV of 2.0 L, while the trunk contributed only 0.17 L. Others observed that during HD, major changes in total body extracellular resistance localized to the legs as compared to the arms and torso [7, 22, 23]. Fig. 3(c) partially reflects the contribution from the upper parts of the legs (thighs) to the total changes in Re. The lower ratios of ΔRe,kkBIS/ΔRe,whBIS for unstable patients, compared with stable patients, may indicate a reduced ability of the leg compartments to keep up with the volume changes in the trunk, possibly leading to a significant reduction in blood volume and, thus, to development of IDH, Fig. 3(d). Only later during dialysis, this ratio starts to catch up with that of stable patients possibly due to clinical intervention. If relative changes in ECF with respect to pre-HD values were considered, a similar observation held true. In stable patients, the losses of ECFkkBIS from leg compartments followed those of the whole-body with an equivalent slope throughout the treatment. On the other hand, fluid losses from the lower extremities fell significantly behind those observed from whole-body measurements for unstable patients. In the present study, the rapid change in knee-to-knee Re,kkBIS within the first hour of treatment was followed by a slower change until the end of HD, regardless of the applied ultrafiltration-rate. This suggests that the fluid contribution from the legs is not only affected by the applied ultrafiltration but also by other physiological mechanisms. The latter slow changes in leg resistivity may continue up to 48 h after dialysis, as demonstrated in .
Relative changes in parameters extracted from kkBIS correlated well with intradialytic weight loss for stable patients. This correlation was weaker for unstable patients, likely due to medical interventions for IDH. In particular, interventions involving switching patients to the Trendelenburg position, administering parenteral fluid, or decreasing the ultrafiltration rate would generate a new fluid distribution across the body and impair the ability of BIS to correctly estimate fluid volumes [15, 25]. Pre- to post-dialysis, the mean difference in change between ΔRe,whBIS and ΔRe,kkBIS, normalized to weight loss, was significantly lower in stable patients compared to unstable ones (p < 0.001), Fig. 4. This further indicates that a smaller contribution from lower extremities is expected for IDH-prone patients.
In Fig. 5, we investigated the relationship between the accumulative reduction in ECF and the collected volume of the ultrafiltrate during the period of HD. For stable patients, whole-body ECFwhBIS is clearly reduced by UFV in a very linear 1:1 relationship throughout the HD treatment. For unstable patients, ECFwhBIS was further depleted by an amount of ~469.4 mL above UFV. This could be explained as some of the ECF may be shifted toward the intracellular compartment driven by osmotic gradient built up during HD. Another possible explanation is due to a BIS measurements error occurred when a great amount of fluid is removed from the trunk during HD. The thigh’s ECFkkBIS is depleted as well by UFV; however, with a significantly higher slope for unstable patients (slope: -0.17 ± 0.05) than for stable ones (slope: -0.11 ± 0.01). If the ratio of thigh’s ECFkkBIS to whole-body ECFwhBIS was to be considered, unstable patients are obviously characterised by significantly lesser values compared to stable one especially within the first half of the treatment. This particular parameter could potentially be used in the future to identify patients at risk of IDH early during the HD treatment.
A limitation of this study is the small number of recruited patients. Nevertheless, having an equal number of patients in each group prevented a strong bias toward any group. In addition, findings were relatively consistent among patients within groups, suggesting that increased group sizes are unlikely to lead to different conclusions. Other errors are inherent to the BIS method itself, hence may affect the accuracy of kkBIS method as well.
The difference in cross-sectional area between different body segments, the regional fluid accumulation in certain body segments, and the alterations of body fluid distribution due to HD are all known to reduce the accuracy of BIS [19, 27, 28]. Nevertheless, the question of the accuracy of both methods, especially for ECF estimation, could not be adequately addressed in this study due to the lack of comparison against gold standards. Hence, this remains a field for further research.
In summary, due to some measurement limitations imposed by the standard electrode configurations of whole-body BIS method, alternative electrode configurations (e.g. dividing the body into segments in series) were suggested. These new configurations may further impairs patient comfort and increases the complexity of the method. Here we demonstrate that a BIS method measuring from knee to knee can monitor changes in body fluid volumes with a comparable accuracy as standard whBIS even in hemodynamically unstable patients. Knee to knee BIS thus be a simplified and more comfortable alternative to whole body BIS.
This work has been supported by the RWTH ExploratoryResearch Space (ERS) under the grant MTBo05.
Kuhlmann MK, Zhu F, Seibert E, Levin NW. Bioimpedance, dryweight and blood pressure control: new methods andconsequences. Curr. Opin. Nephrol. Hypertension 2005;14:543–549.https://doi.org/10.1097/01.mnh.0000185983.48319.00
Moissl UM, et al. Body fluid volume determination via bodycomposition spectroscopy in health and disease. Physiol.Meas. 2006:27: 921-33.
Heitman BL. Impedance: a valid method in assessment ofbody composition?. Eur. J. Clin. Nutr. 1994;48: 228-40.
Cole K, Cole R. Dispersion and absorption in dielectrics. I.alternating-current characteristics. J. Chem. Phys. 1941;9:341-51.
Ismail AH et al. Usefulness of bioimpedance spectroscopy fordetection of hypotensive episode during dialysis. ASAIO J.2014;60: 570-5.
Scharfetter H, Monif M, László Z, Lambauer T, Hutten H,Hinghofer-Szalkay H. Effect of postural changes on thereliability of volume estimations from bioimpedancespectroscopy data. Kidney Int. 1997;51: 1078-87.https://doi.org/10.1038/ki.1997.150
Baumgartner RN, Chumlea WC, Roche AT. Estimation of bodycomposition from segmental impedance. Am. J. Clin. Nutr.1989;50: 221-5.
Fuller NJ, Elia M. Potential use of the bioelectrical impedanceof the ‘‘whole body’’ and of body segments for theassessment of body composition: comparison withdensitometry and anthropometry. Eur. J. Clin. Nutr. 1989;43:779-91.