No full-text available
To read the full-text of this research,
you can request a copy directly from the author.
Adaptive Neuro Fuzzy system as a novel approach for predicting post-dialysis urea rebound
Abstract
Total dialysis dose (Kt/V) is considered to be a major determinant of morbidity and mortality in haemodialysed patients. The continuous growth of the blood urea concentration over the 30–60-min period following dialysis, a phenomenon known as urea rebound, is a critical factor in determining the true dose of haemodialysis (HD). The misestimation of the equilibrated (true) post-dialysis blood urea or equilibrated Kt/V results in an inadequate HD prescription, with predictably poor clinical outcomes for the patients. The estimation of the equilibrated post-dialysis blood urea (Ceq) is therefore crucial in order to estimate the equilibrated (true) Kt/V. Measuring post-dialysis urea rebound (PDUR) requires a 30- or 60-min post-dialysis sampling, which is inconvenient. This paper presents a novel technique for predicting equilibrated urea concentration and PDUR in the form of a Takagi–Sugeno–Kang fuzzy inference system. The advantage of this neuro-fuzzy hybrid approach is that it does not require 30–60-min post-dialysis urea sample. Adaptive neuro-fuzzy inference system (ANFIS) was constructed to predict equilibrated urea (Ceq)taken at 60 min after the end of the HD session in order to predict PDUR. The accuracy of the ANFIS was prospectively compared with other traditional methods for predicting equilibrated urea (Ceq), PDUR and equilibrated dialysis dose (eqKt/V). The results are highly promising, and a comparative analysis suggests that the proposed modelling approach outperforms other traditional urea kinetic models.
... FIS can interpret relationship between inputsoutputs by human knowledge in the form of "if-then" rules (Nasr et al., 2014). ANFIS is an effective method developed to combine the learning and connectionist structure of neural networks with the humanlike reasoning style of fuzzy systems (Azar, 2011). Additionally, ANFIS requires fewer adjustable parameters than those required in other ANN structures. ...
... The applicability of pseudo-second-order model indicated that the mechanism of adsorption was chemical interaction (i.e. chemisorption) (Atkins, 1995). The pseudo-second-order kinetics also suggests that the key interaction between the metal-ions and the biosorbent involves their sharing of valence electrons (Saleem et al., 2016). ...
... The checking data were used to cross-validate the fuzzy inference model for the detection and prevention of model overfitting. The hold-out cross-validation method was adopted for ANFIS because of its accuracy, relative stability and low computational time requirements (Azar, 2011). To validate the ANFIS model, the training and checking data errors were calculated in the form of root mean square error. ...
This study was carried out to examine the Typha domingensis as a biosorbent for the removal of Ni(II) and Cd(II) ions from aqueous solution, and to determine the experimental parameter that mostly influenced the biosorption process. The sorption mechanism was discussed in combination with the results obtained from Fourier-transform infrared spectroscopy (FTIR), scanning electron microscopy (SEM), and energy-dispersive X-ray spectroscopy (EDX). Results revealed that, at initial metal-ions concentration: 50 mg L−1, biosorbent dosage: 10.0 g L−1 and particle-size: 0.5 mm, the highest Ni(II) removal efficiency of 31.0% was obtained at pH: 6.0 within 30 min, whereas the maximum Cd(II) removal efficiency was 78.1% at pH: 7.0 within 15 min. The adsorption equilibrium data of Ni(II) fitted well to the Freundlich isotherm (r2: 0.9202), while Cd(II) adsorption was described well with both isotherms of Langmuir (r2: 0.9504) and Freundlich (r2: 0.9480). Kinetic studies showed that pseudo-second order described well the biosorption experimental data (r2 > 0.98). Furthermore, through modeling of an adaptive neuro-fuzzy inference system (ANFIS), the sequence of experimental parameters affecting metal-ions removal efficiency for Ni(II) ions was: Ni(II) concentration > contact time > pH > biosorbent dosage > particle-size, while for Cd(II) ions, the sequence was: pH > Cd(II) concentration > contact time > particle-size > biosorbent dosage.
... This algorithm showed several disadvantages such as slow convergence, sensitivity to local minima and the difficulty to adjust the learning parameters (the number of neurons in the hidden layers, learning step etc…). In some networks using Hebbian learning where the synaptic weights can be adjusted during a learning phase patterns through Hebb's formula leads to a formula expressing the weights based on grounds recognized [4,28,51]. ...
In this work, we use radial basis function neural network for modeling nonlinear systems.Generally, the main problem in artificial neural network is often to find a better structure.The choice of the architecture of artificial neural network for a given problem has long been a problem.Developments show that it is often possible to find architecture of artificial neural network that greatly improves the results obtained with conventional methods.We propose in this work a method based on No Sorting Genetic Algorithm II (NSGA II) to determine the best parameters of a radial basis function neural network.The NSGAII should provide the best connection weights between the hidden layer and output layer, find the parameters of the radial function of neurons in the hidden layer and the optimal number of neurons in the hidden layers and thus ensure learning necessary.Two functions are optimized by NSGAII: the number of neurons in the hidden layer of the radial basis function neural network, and the error which is the difference between desired input and the output of the radial basis function neural network.This method is applied to modeling Box and Jenkins system.The obtained results are very satisfactory.
This article describes how supplier selection is a multi-index problem which affects the efficiency of the whole supply chain in both manufacturing and service industries. Considering the importance of selecting effective suppliers, this article aims to integrate two well-known techniques, Quality Function Deployment (QFD) and The Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) in order to evaluate suppliers and rank them based on their merits. In order to handle the inherent uncertainty in the process of experts' judgments, fuzzy logic is involved in the methodology applied in this study. The validity of the utilized integrated approach is demonstrated through conducting a case study in the detergent manufacturing industry in Iran.
Nitrification at a full-scale activated sludge plant treating municipal wastewater was monitored over a period of 237 days. A combination of fluorescent in situ hybridization (FISH) and quantitative real-time polymerase chain reaction (qPCR) were used for identifying and quantifying the dominant nitrifiers in the plant. Adaptive neuro-fuzzy inference system (ANFIS), Pearson’s correlation coefficient, and quadratic models were employed in evaluating the plant operational conditions that influence the nitrification performance. The ammonia-oxidizing bacteria (AOB) abundance was within the range of 1.55 × 108–1.65 × 1010 copies L−1, while Nitrobacter spp. and Nitrospira spp. were 9.32 × 109–1.40 × 1011 copies L−1 and 2.39 × 109–3.76 × 1010 copies L−1, respectively. Specific nitrification rate (qN) was significantly affected by temperature (r 0.726, p 0.002), hydraulic retention time (HRT) (r −0.651, p 0.009), and ammonia loading rate (ALR) (r 0.571, p 0.026). Additionally, AOB was considerably influenced by HRT (r −0.741, p 0.002) and temperature (r 0.517, p 0.048), while HRT negatively impacted Nitrospira spp. (r −0.627, p 0.012). A quadratic combination of HRT and food-to-microorganism (F/M) ratio also impacted qN (r
2 0.50), AOB (r
2 0.61), and Nitrospira spp. (r
2 0.72), while Nitrobacter spp. was considerably influenced by a polynomial function of F/M ratio and temperature (r
2 0.49). The study demonstrated that ANFIS could be used as a tool to describe the factors influencing nitrification process at full-scale wastewater treatment plants.
Fault location and distance protection in transmission lines are essential smart grid technologies ensuring reliability of the power system and achieve the continuity of service. The objective of this chapter is to presents an accurate algorithm for estimating fault location in Extra High Voltage (EHV) transmission lines using Artificial Neural Networks (ANNs) for high speed protection. The development of this algorithm is based on disturbed transmission line models. The proposed fault protection (fault detection/classification and location) uses only the three phase currents signals at the one end of the line. The proposed technique uses five ANNs networks and consists of two steps, including fault detection/classification and fault location. For fault detection/classification, one ANN network is used in order to identify the fault type; the fault detection/classification procedure uses the fundamental components of pre-fault and post-fault sequence samples of three phase currents and zero sequence current. For fault location, four ANNs networks are used in order to estimate the exact fault location in transmission line. Magnitudes of pre-fault and post-fault of three phase currents are used. The ANNs are trained with data under a wide variety of fault conditions and used for the fault classification and fault location on the transmission line. The proposed fault detection/classification and location approaches are tested under different fault conditions such as different fault locations, different fault resistances and different fault inception angles via digital simulation using MATLAB software in order to verify the performances of the proposed methods. The ANN-based fault classifier and locator gives high accuracy for all tests under different fault conditions. The simulations results show that the proposed scheme based on ANNs can be used for on-line fault protection in transmission line.
Dialysis dose (Kt/V) is mostly dependent on dialysis kinetic variables such as pre-dialysis and post-dialysis blood urea nitrogen concentration (Cpost), ultrafiltration (UF) volume, duration of the dialysis procedure, and urea distribution volume. Therefore, post-dialysis blood urea concentration is used to assess the dialysis efficiency. It gradually decreases to about 30% of the pre-dialysis value depending on the urea clearance rate during the period of dialysis. If the urea removal is inadequate, then dialysis is inadequate. This paper proposes a novel method, Adaptive Neuro-Fuzzy Inference System (ANFIS) to predict the post-dialysis blood urea concentration. The advantage of this neuro-fuzzy hybrid approach is that it does not require the model structure to be known a priori, in contrast to most of the urea kinetic modelling techniques. The accuracy of the ANFIS was prospectively compared with other traditional methods for predicting single pool dialysis dose (spKt/V). The results are highly promising, and a comparative analysis suggests that the proposed modelling approach outperforms other traditional urea kinetic models (UKM).
Post-dialysis urea rebound (PDUR) has been attributed mostly to redistribution of urea from different compartments, which is determined by variations in regional blood flows and transcellular urea mass transfer coefficients. PDUR occurs after 30–90 min of short or standard hemodialysis (HD) sessions and after 60 min in long 8-h HD sessions, which is inconvenient. This paper presents adaptive network based on fuzzy inference system (ANFIS) for predicting intradialytic (Cint) and post-dialysis urea concentrations (Cpost) in order to predict the equilibrated (Ceq) urea concentrations without any blood sampling from dialysis patients. The accuracy of the developed system was prospectively compared with other traditional methods for predicting equilibrated urea (Ceq), post dialysis urea rebound (PDUR) and equilibrated dialysis dose (eKt/V). This comparison is done based on root mean squares error (RMSE), normalized mean square error (NRMSE), and mean absolute percentage error (MAPE). The ANFIS predictor for Ceq achieved mean RMSE values of 0.3654 and 0.4920 for training and testing, respectively. The statistical analysis demonstrated that there is no statistically significant difference found between the predicted and the measured values. The percentage of MAE and RMSE for testing phase is 0.63% and 0.96%, respectively.
The blood urea concentration has been used as a surrogate marker for toxin elimination in hemodialysed patients, and several indices based on it have been proposed in recent years for monitoring treatment adequacy. Measuring urea nitrogen concentrations at the inlet and outlet of dialyser is crucial to evaluate the in-vivo blood side dialyser urea clearance during hemodialysis. Although frequent measurement is needed to avoid inadequate dialysis efficiency, artificial intelligence can repeatedly perform the forecasting tasks and may be a satisfactory substitute for laboratory tests. Neuro-fuzzy technology represents a promising forecasting application in clinical medicine. In this study, two fuzzy models have been proposed to predict dialyser inlet and outlet urea concentrations in order to estimate dialyser clearance without blood sampling. The model is of multi-input single-output (MISO) type. Multi-adaptive neuro-fuzzy inference system (MANFIS) technique of fuzzy-based systems has been employed. The performance of the model is authenticated by evaluating the predicted results with the practical results obtained by conducting the confirmation experiments. The results suggest that the neuro-fuzzy technology, based on limited clinical parameters, is an excellent alternative method for accurately predicting arterial and venous urea concentrations in hemodialysis patients. The proposed model can be used for intelligent online adaptive system.
Soft computing techniques are known for their efficiency in dealing with complicated problems when conventional analytical methods are infeasible or too expensive, with only sets of operational data available. Its principal constituents are fuzzy logic, Artificial Neural Network (ANN) and evolutional computing, such as genetic algorithm. Neuro-fuzzy controllers constitute a class of hybrid soft computing techniques that use fuzzy logic and artificial neural networks. The advantages of a combination of ANN and Fuzzy Inference system (FIS) are obvious. There are several approaches to integrate ANN and FIS and very often it depends on the application. This chapter gives an overview of a neuro-fuzzy system design with novel applications in dialysis using an adaptive-network-based fuzzy inference system (ANFIS) for the modeling and predicting important variables in hemodialysis process.
Urea kinetic modelling (UKM) has been generally accepted as a method for quantifying hemodialysis (HD) treatment. During hemodialysis, reduction in the urea concentration in the intracellular fluid (ICF) compartment will lag behind that in the extra cellular fluid (ECF) compartment, and following the end of dialysis, a ”rebound” in the blood level of urea will occur where it continues to rise due to diffusion of urea from the ICF to ECF to establish an equilibrium state. Because of compartment effects, the dose of dialysis with regard to urea removal is significantly overestimated from immediate post-dialysis urea concentrations, because 30 to 60 min are required for concentration gradients to dissipate and for urea concentrations to equilibrate across body water spaces during the post-dialysis period. To avoid the delay of waiting for an equilibrated post-dialysis sample, it became necessary to describe and to quantitate effects causing the urea compartmentalization during dialysis; two-pool modeling approaches have been developed that more accurately reflect the amount of urea removed. This in turn gives more adequate measures not only of dialysis adequacy, but also of the protein catabolic rate, an important nutritional measure that is clinically monitored in dialysis patients. This chapter discusses the double pool urea kinetic models and regional blood flow models in order to understand the concept of urea rebound.
The post-hemodialysis rebound: Predicting and quantifying its effect on Kt/V. Immediately after hemodialysis, the urea concentration rebounds upwards as urea continues to be transferred into the arterial circulation from peripheral body compartments. This rebound takes at least 30 minutes to complete. Hemodialysis is quantified as the Kt/V, calculated prom pre- and post-dialysis urea samples. Unless the post-dialysis sample is taken at least 30 minutes after dialysis, the Kt/V will be overestimated. This overestimation will be relatively greater in short high-efficiency dialyses, which have greater post-dialysis rebounds. We propose a method of correction that uses only the conventional pre- and immediate post-dialysis samples and is based on the physiologically-appropriate patient clearance time (tp). This is the time needed to clear all body compartments when the dialyzer clearance is infinite. The tp can be calculated from the pre-, immediate post- and 30-minute post-dialysis urea concentrations and was 35 minutes (SD 16) in 29 patients undergoing short (149 min) hemodiafiltration and standard (243 min) hemodialysis the following week. There was no significant difference between tp values calculated during the two treatments. Standard Kt/V can be corrected by multiplying by t/(t + tp) and dialysis time should be increased by tp Kt/V minutes to compensate for the rebound. Despite individual variations in tp, a value of tp = 35 was sufficient to correct Kt/V in all patients. Kt/V corrected in this way agreed with Kt/V calculated using a 60-minute post-dialysis sample (r = 0.856, P < 0.001). The method predicted the 60-minute post-rebound concentration (SE 0.5 mM, r = 0.983, P < 0.001) and the addition of 35 minutes to the treatment time corrected for the rebound in both conventional and short treatments. Similar simple equations corrected the error in V caused by rebound effects.
Total dialysis dose (Kt/V) is considered to be a major determinant of morbidity and mortality in hemodialyzed patients. The continuous growth of the blood urea concentration over the 30- to 60-min period following dialysis, a phenomenon known as urea rebound, is a critical factor in determining the true dose of hemodialysis. The misestimation of the equilibrated (true) post-dialysis blood urea or equilibrated Kt/V results in an inadequate hemodialysis prescription, with predictably poor clinical outcomes for the patients. The estimation of the equilibrated postdialysis blood urea (Ceq) is therefore crucial in order to estimate the equilibrated (true) Kt/V. Measuring post dialysis urea rebound (PDUR) requires a 30- or 60-minute post-dialysis sampling, which is inconvenient. In this work a supervised Artificial Neural Network (ANN) is proposed for predicting equilibrated urea concentration taken at 30 min after the end of the hemodialysis (HD) session in order to predict PDUR. The advantage of ANN approach is that it doesn't require 30-60-minute post-dialysis urea sample. This approach is compared experimentally with other traditional methods for predicting equilibrated urea concentration (C eq), PDUR and equilibrated dialysis dose (eqKt/V). The results are highly promising, and a comparative analysis suggests that the proposed modeling approach outperforms other traditional urea kinetic models (UKM) and the previous work using ANN.
The recent introduction of urea sensors for dialysis monitoring has made possible new approaches to urea kinetic modelling. In this study we show how the equilibrated postdialysis urea concentration (Ceq) and Kt/V corrected for double-pool urea kinetics (Kt/Vdp) can be accurately determined using an on-line sensor providing a continuous measure of blood water urea. A modification of the Smye constant volume double-pool theory led to the following equations for Ceq and Kt/Vdp [formula: see text] where Cpre is the blood concentration measured at the start of dialysis, t is the length of the dialysis session (in min) and S(ex) is the constant slope of the blood urea logarithm concentration decline following development of the intercompartmental urea concentration gradient in the first 30-60 min of dialysis.
These equations were tested in 11 patients undergoing 165-240 min of paired filtration dialysis with continuous monitoring of blood urea concentration. Cpre was determined as the plateau concentration during a preliminary period of 15-20 min of slow isolated ultrafiltration. S(ex) was accurately determined from linear regression applied to the urea sensor data from the 80-min point to the end of dialysis.
Ceq and Kt/Vdp determined from the above equations compared closely to values determined from 25-40 min of urea rebound monitoring with the urea sensor: 10.6 +/- 3.0 versus 10.8 +/- 2.7 mmol/l (mean +/- SD) for Ceq and 1.21 +/- 0.24 versus 1.18 +/- 0.20 for Kt/Vdp, compared to single-pool values of Kt/V = 1.34 +/- 0.23.
This technique may be readily programmed into on-line urea monitors to provide current and extrapolated values of Ceq and Kt/Vdp from about the first hour of dialysis.
The effect of increasing dialysis efficiency magnifies rebound urea and the error in Kt/V determinations from single pool urea kinetics. Several formulae have been developed to calculate Kt/V taking into account the rebound urea (Kt/Vr). Smye et al. proposed a method whereby the equilibrated BUN is predicted by an additional intradialytic urea sample (Kt/VrSmye). Daugirdas et al. proposed a method where a single pool Kt/V is modified according to the speed of dialysis to obtain a double pool Kt/V (Kt/VrDaug). Maduell et al. developed a method based on analysis of post-dialysis urea rebound whereby the Kt/Vr is predicted according to the single pool Kt/V and K/V (Kt/VrMad).
We compared Kt/Vr estimated by these three formulae (Smye, Daugirdas, and Maduell) in 384 patients consisting of 211 males and 173 females, who received dialysis according to their regular protocols. Plasma urea was measured at the beginning, 90-100 min following the start of dialysis, the end, and 45 min post-dialysis.
Post-dialysis rebound urea was 22.4 +/- 9.7%. Kt/V and Kt/Vr obtained with rea kinetic model Kt/V 1.184 +/- 0.22 and 0.984 +/- 0.20, respectively. These was a good correlation between Kt/Vr and the Smye formula (Kt/VrSmye = 0.956 +/- 0.21, r = 0.729, P < 0.001), and a better one for Daugirdas (Kt/VrDaug = 0.984 +/- 0.18, r = 0.931, P < 0.001), and Maduell formulae (Kt/VrMad = 0.980 +/- 0.18, r = 0.946, P < 0.001). Limits of agreement and percentage of error estimated according to Bland and Altman show that Kt/Vr estimated by Daugirdas and Maduell formulae could be used in place of the Kt/Vr. The degree of agreement with the Smye method is not clinically acceptable.
Our results suggest that the use of a single pool Kt/V is not adequate to estimate the haemodialysis dose delivered and Kt/V taking rebound urea in consideration. Kt/Vr estimated by Daugirdas or Maduell formulae are a simple and accurate method for use in clinical practice.
In Europe, especially in Germany, little is currently known about the relationship between delivered and predicted haemodialysis doses for patients on maintenance haemodialysis. We compared delivered and predicted Kt/V in patients of an outpatient dialysis centre in Berlin by calculating the ratio of delivered and predicted Kt/V, resulting in the efficacy quotient, QE. Moreover, we studied the influence of technical and anthropometric parameters on both delivered Kt/V and QE under routine clinical conditions.
Blood samples were taken after the long interval in a thrice-weekly regimen before and 10 min after ultrafiltration and 100 ml/min slow-pump method. Delivered Kt/V was computed using the Daugirdas III formula. Predicted Kt/V was estimated from the dialysis filter urea clearance given by the manufacturer, treatment time and the total body water (V) computed by the Watson formula and was corrected for real blood flow. As and when appropriate, bivariate and multivariate regression analyses were used to make comparisons.
The mean quotient (QE) between delivered and predicted Kt/V was 1.02+/-0.20. Mean delivered Kt/V in 377 treatments of 128 patients was 1.28+/-0.27. Delivered Kt/V and QE were positively associated (P<0.001). QE was significantly associated with post-HD urea, body mass index (BMI) and sex, but not with session time. Significant positive predictors for delivered Kt/V were post-dialysis urea, sex, session time, blood flow and kind of vascular access. BMI was inversely related to delivered Kt/V.
In this study, the relationship between delivered and predicted Kt/V (QE) was reproducible and close to the ideal value of 1.0. In contrast to delivered Kt/V, QE was not influenced by session time, and positively by BMI. Since QE gives a valid measure of technical dialysis efficacy we suggest the use of this parameter in addition to delivered Kt/V to monitor HD adequacy in clinical routine more comprehensively.
Total dialysis dose (Kt/V) is considered to be a major determinant of morbidity and mortality in hemodialyzed patients. The continuous growth of the blood urea concentration over the 30- to 60-min period following dialysis, a phenomenon known as urea rebound, is a critical factor in determining the true dose of hemodialysis. The misestimation of the equilibrated (true) postdialysis blood urea or equilibrated Kt/V results in an inadequate hemodialysis prescription, with predictably poor clinical outcomes for the patients. The estimation of the equilibrated postdialysis blood urea (eqU) is therefore crucial in order to estimate the equilibrated (true) Kt/V. In this work we propose a supervised neural network to predict the eqU at 60 min after the end of hemodialysis. The use of this model is new in this field and is shown to be better than the currently accepted methods (Smye for eqU and Daugirdas for eqKt/V). With this approach we achieve a mean difference error of 0.22 +/- 7.71 mg/ml (mean % error: 1.88 +/- 13.46) on the eqU prediction and a mean difference error for eqKt/V of -0.01 +/- 0.15 (mean % error: -0.95 +/- 14.73). The equilibrated Kt/V estimated with the eqU calculated using the Smye formula is not appropriate because it showed a great dispersion. The Daugirdas double-pool Kt/V estimation formula appeared to be accurate and in agreement with the results of the HEMO study.
The double-pool urea kinetic model requires the measurement of the blood urea concentrations 30 min after haemodialysis (C(t+30)) to calculate equilibrated Kt/V. However, it has been suggested that urea concentrations 30 min before the end of dialysis (C(t-30)) may be representative of C(t+30). The aim of this study was to validate this suggestion.
Twenty-two patients underwent haemodialysis for 180, 210, and 240 min. For each patient in each dialysis session, urea exponential decay curve was calculated. Because we measured C(t+30), we calculated the time (T(c)) before the end of dialysis that blood urea concentrations would be the same as C(t+30). In an additional 33 patients, we measured blood urea concentrations at T(c) and in C(t+30).
We found that C(t-30) was significantly lower than C(t+30) independent of the duration of dialysis. However, there was a significant correlation between Kt/V(t-30) and Kt/V(t+30). The T(c) was 45 min before the end of dialysis. In the additional 33 patients, C(t-45) and C(t+30) were 54+/-17 and 52+/-17 mg/dl (NS), and Kt/V(t-45) and Kt/V(t+30) were 1.27+/-0.21 and 1.29+/-0.18 (NS), respectively. There were significant correlations between C(t-45) and C(t+30) (r=0.96; P<0.001), and between Kt/V(t-45) and Kt/V(t+30) (r=0.82; P<0.001). However, when measurements were analysed individually, 48% of the data points from C(t-45) vs C(t+30), and 42% of the data points from Kt/V(t-45) vs Kt/V(t+30) fell out of the 95% confidence interval of regression line.
Although C(t-45) is useful to estimate Kt/V when assessing mean values, it is not suitable when assessing patients individually. This study demonstrates that the best method to calculate equilibrated Kt/V was a blood sample for urea concentrations 30 min after haemodialysis.
Previous studies have suggested that postdialysis urea rebound is related to K/V, the rate of dialysis, but a systematic analysis of factors that affect rebound has not been reported. With the use of 30-min and, in a subset, 60-min postdialysis samples, postdialysis urea rebound was measured to (1) determine how well previously proposed equations based on the rate of dialysis (K/V) predict rebound in a large sample of patients with varying characteristics, (2) determine whether other factors besides K/V affect rebound, and (3) estimate more precise values for coefficients in prediction equations for rebound. Rebound was calculated relative to both immediate and 20-s postdialysis samples to study early components of rebound unrelated to access recirculation. The equilibrated Kt/V (eKt/V) computed by fitting the two-pool variable volume model to the 30-min postdialysis sample agreed well with eKt/V based on the 60-min postdialysis sample. Using the pre-, post-, and 30-min postdialysis samples for 1245 patients with arteriovenous (AV) accesses, the median intercompartmental mass transfer coefficient (Kc) was 797 ml/min for rebound computed relative to the 20-s postdialysis samples and 592 ml/min relative to the immediate postdialysis samples. K/V was the strongest predictor of rebound among 22 factors considered. Other factors associated with greater rebound for 1331 patients using AV accesses or venous catheters included access type, black race, male gender, absence of congestive heart failure, greater age, ultrafiltration rate, and low predialysis or intradialysis systolic BP. Equations of the form eKt/V = single-pool Kt/V - B x (K/V) were fit to the data. With AV access, the optimum values for the slope term (B) were 0.39 and 0.46 (in h(-1)) for single-pool Kt/V calculated based on 20-s postdialysis or immediate postdialysis samples, respectively. For patients using venous catheters, the respective values for B were 0.22 and 0.29. Postdialysis urea rebound can be predicted with acceptable accuracy from a postdialysis sample using a zero-intercept, K/V-based rate equation. Several patient or treatment-specific factors predict enhanced or reduced rebound. Rate equation slope coefficients for K/V of 0.39 (AV access) and 0.22 (venous access) are proposed when a 15- to 20-s slow-flow method is used to draw the postdialysis blood. Slightly higher K/V slope coefficients (0.46 and 0.29, respectively) should be used if a shorter (e.g., 10 s) slow-flow period is used.
The equilibrated Kt/V (eKt/V), a clinical index of hemodialysis adequacy, can be calculated by several means. The commonly used methods are second generation of natural logarithm by Daugirdas and rate adjustment methods. However these two methods used immediate post-dialysis blood urea nitrogen (BUN) (C0). The authors investigated the accuracy of 30-minute post-dialysis BUN (C30), equilibrated BUN at the end of urea rebound time, to determine the eKt/V MATERIAL AND METHOD: We measured 7 values of eKt/V in 30 hemodialysis sessions by using the 5 following methods: 1). Empirical method (Emp) 2). Smye method (Sm) 3). Daugirdas method using C0 and C30 (Dau0, Dau30) 4). Rate adjustment method using C0 and C30 (Rate0, Rate30) 5). Double Pool urea kinetic model (DP), and compared with the gold standard values calculated by the modified Direct Dialysate Quantitative method (mDDQ).
All patients had adequate hemodialysis with eKt/V (mDDQ) = 1.80 + 0.22. Compared with mDDQ, the median of absolute difference of eKt/V were Rate30 (0.10); Dau30 (0.11), Rate0 (0.11), Sm (0.11); Emp (0.13); DP (0.15); Dau0 (0.35) while the correlation coefficient (r2) were 0.926, 0.948, 0.785, 0.629, 0.551, 0.833,and 0.806 respectively.
By using 30-minute post-dialysis BUN to calculate, the values of eKt/V by Daugirdas and rate adjustment methods were associated with better accuracy and correlation than immediate post-dialysis BUN. In the demand of the accurate eKt/V measurement, the Dau30 and Rate30 may be the suitable method to determine the eKt/V in clinical hemodialysis.
Fundamental and advanced developments in neuro-fuzzy synergisms
for modeling and control are reviewed. The essential part of neuro-fuzzy
synergisms comes from a common framework called adaptive networks, which
unifies both neural networks and fuzzy models. The fuzzy models under
the framework of adaptive networks is called adaptive-network-based
fuzzy inference system (ANFIS), which possess certain advantages over
neural networks. We introduce the design methods for ANFIS in both
modeling and control applications. Current problems and future
directions for neuro-fuzzy approaches are also addressed
The history of many disciplines is fragmented and often subjective, but the history of dialysis quantitation is held together by the contributions of Frank Gotch. The story is unfolded in Frank's many publications and unpublished memoranda, including personal correspondence, some of which has graciously shared with me over the past 20 years. Frank Gotch is so interwoven into the history of dialysis quantitation that simply reading his papers provides more than half of the story. In fact he actually wrote a short history of dialysis solute kinetics, documenting historical milestones up to 1986 (1). The final chapter in this history is not yet written and,will probably not appear for many years to come. Meanwhile, Frank Gotch continues to be a part of it, serving as a consultant To the HEMO Study. In recent years, those who have had an opportunity to see a bit beyond the fringes of established knowledge in this field have done so only from the shoulders of dedicated investigators like Frank Gotch.
This paper describes an experiment on the “linguistic” synthesis of a controller for a model industrial plant (a steam engine), Fuzzy logic is used to convert heuristic control rues stated by a human operator into an automatic control strategy. The experiment was initiated to investigate the possibility of human interaction with a learning controller. However, the control strategy set up linguistically proved to be far better than expected in its own right, and the basic experiment of linguistic control synthesis in a non-learning controller is reported here.
Predictive modeling is the process of identifying a model of an unknown or complex process from numerical data. Due to the inherent complexity of many real processes, conventional modeling techniques have proved to be too restrictive. Recently, the hybrid approach to predictive modeling has become a popular research focus. A novel hybrid system combining different soft computing paradigms such as neural networks and fuzzy systems has been developed for predictive modeling of dialysis variables in order to estimate the equilibrated dialysis dose (eqKt/V), without waiting for 30-60 min post-dialysis to get the equilibrated urea sample which is inconvenient for patients and costly to the dialysis unit. The aim of using a neuro-fuzzy network is to find, through learning from data, a fuzzy model that represents the process underlying the data. In neuro-fuzzy models, connection weights, propagation and activation functions differ from common neural networks. Although there are a lot of different approaches, the term neuro-fuzzy is restricted to systems which display the following properties: · A neuro-fuzzy system is a fuzzy system that is trained by learning algorithm (usually) derived from neural network theory. The (heuristic learning procedure operates on local information, and causes only local modifications in the underlying fuzzy system. The learning process is not knowledge based, but data driven. · A neuro-fuzzy system can be viewed as a special 3-layer feedforward neural network. The units in this network use t-norms or t- cononrms instead of the activation functions usually used in neural networks. The first layer represents input variables, the middle (hidden) layer represents fuzzy rules and the third layer represents output variables. Fuzzy sets are encoded as (fuzzy)
The 2-pool urea kinetic model has been developed analytically and applied to the description of the observed increase in blood levels of urea following dialysis (urea rebound), assuming that the dialyser urea clearance K less than 0.4X where X is the urea mass transfer coefficient between the intracellular and extracellular pools (volumes V1, V2 respectively). Urea generation was also neglected. Measurements were made in a group of six children suffering from chronic renal failure. From the model X, the efficiency of dialysis, and the equilibrium urea concentration C infinity were estimated in the presence of urea rebound using a blood urea measurement taken 90 min following start of dialysis, in addition to the conventional samples taken immediately pre- and post-dialysis. In three of the patients agreement between the experimental value of X derived from a multi-blood-sample technique post-dialysis, and the model value, was within 10%, for the range V1 = 0.4 W - 0.38 W, V2 = 0.2 W - 0.238 W, (W = patient's weight). Experimental values of X were in the range 93 - 300 ml min-1. Model estimates of C infinity were accurate to within 10%. An approximate technique was also developed which permitted an estimate of C infinity which was independent of V1, V2, K. The results indicated that C infinity was estimated to within 10% of the true equilibrium urea concentration. The error in the estimate of dialysis efficiency based on a single pool model was reduced by at least 50% using the model. The model may be applied clinically to the estimation of dialysis efficiency in the presence of significant urea rebound.
The purpose of the NCDS was to determine the probability of clinical failure (PF) as a function of the level of dialysis and protein catabolic rate (pcr, g/kg/day). The level of dialysis prescribed in the NCDS was mechanistically defined as Kt/V (product of dialyzer urea clearance and treatment time divided by body urea volume), which exponentially determines decrease in BUN during dialysis and is also a mathematical analogue of pcr, BUN. Mechanistic analysis (MA) showed that PF was a discontinuous function of Kt/V as it was prescribed in the NCDS and that a dependence of PF on pcr could not be assessed because of the study design. The MA results were compared to those reported with statistical analysis (SA) that used BUN and pcr. The SA predicts PF is strongly dependent on pcr with nutrition-dependent high PF for pcr less than or equal to 0.8 and low PF with high pcr and intensive dialysis. The MA suggests SA results may not be valid because a continuous outcome function is assumed and, due to study design, Kt/V was a dependent variable of pcr and these two variables cannot be clearly separated by analysis of BUN and pcr alone.
The Smye method has been proposed to estimate the equilibrated post-dialysis BUN based on an additional intradialytic sample obtained approximately one hour into dialysis. However, the effects of access recirculation (AR) and cardiopulmonary recirculation (CPR) on the Smye computation and the corresponding details of how blood is sampled have not been studied. We examined the accuracy of two variations of the Smye technique. In one method, the intradialytic and postdialysis blood samples were obtained at full blood flow. In the other, the samples were obtained after two minutes of slow flow, to obviate the effects of both AR and CPR. Seventeen patients undergoing high efficiency dialysis and three- to four-hour treatment times were studied, in whom substantial AR was excluded based on two-minute slow flow urea rebound measurements during and just after dialysis. In this group equilibrated Kt/V (eKt/V) values computed using the Smye-derived equilibrated postBUN estimates (full flow samples, 1.22 +/- 0.058 SEM, slow flow samples, 1.23 +/- 0.064) were similar to eKt/V calculated from the 30-minute postdialysis BUN specimen (1.23 +/- 0.049, P = NS). In eight other patients with severe AR (mean 35% +/- 4.5), the accuracy of the full flow Smye estimates was poor when the degree of AR was not constant throughout the dialysis session. Accuracy of the slow flow Smye estimates of eKt/V was unaffected by the presence of severe AR. One advantage of using the full flow Smye method, however, was that a large delta Kt/V value ( < -0.40) was highly suggestive of severe AR.(ABSTRACT TRUNCATED AT 250 WORDS)
The original description of urea kinetic modeling in hemodialysis was based on a single pool of fluid equal to total body water. This assumption is valid only if the rate of transport between compartments is sufficiently rapid compared with the rate of urea removed by hemodialysis. We have reexamined the issue of urea compartmentation using 10 patients with initial BUN values ranging from 28 to 101 mg/dL who were studied while on high flux hemodiafiltration. Sampling was carried out at times as short as 1 minute after the beginning of the treatment. In 4 patients, samples were also drawn after the dialysis ended. The measured BUN values were used to fit a two-compartment, variable-volume kinetic model. The mean urea diffusive clearance of the dialyzer during the treatment was 503 mL/minute, and the bidirectional urea clearance between compartments was 1,282 mL/minute. The percentage of total body water in the rapidly exchangeable compartment was 35.4%. Comparison with a one-compartment, variable-volume model showed a statistically better fit for all patients with the more complex model. In the 4 patients studied in the postdialysis period, the rebound in BUN ranged from 22% to 24%. These studies show that, at high urea clearances, the urea distribution space and, therefore, dialysis modeling requires two compartments. These results explain the majority of the rebound in urea concentration observed 30 to 60 minutes following the discontinuation of the dialysis treatment and point out an improved strategy for monitoring the efficiency of the treatment delivered to the patient.
Urea rebound has been documented to occur after hemodialysis, but the magnitude and causes are not clearly defined. In this study we evaluated the effect of high-flux hemodialysis on urea rebound and Kt/V. Blood urea nitrogen samples were obtained before, immediately after, and 30 minutes after hemodialysis in 49 patients. Rebound was evaluated with respect to dialysis efficiency, dialysis treatment time, the occurrence of hypotension, and hematocrit. Urea rebound was significant and resulted in an overall decrease in Kt/V from 1.2 +/- 0.3 to 1.0 +/- 0.2 (P < 0.001). Of the 45 patients with a measured Kt/V of greater than 1.0, 40% had an actual delivered Kt/V of less than 1.0 once rebound was taken into account. Urea rebound correlated strongly with dialysis efficiency but not with hypotension, suggesting that rebound resulted primarily from delayed urea mass transfer across cell membranes. We conclude that increasing dialysis efficiency increases urea rebound and increases the error in Kt/V determinations from single pool urea kinetics.
Blood urea concentration is artificially low immediately following high-efficiency dialysis of duration T and will rise ('rebound') due to the continued diffusion of urea from the intracellular to the extracellular space. This leads to an overestimate of the efficiency of the dialysis given by KT/V (where V is the total distribution volume of urea and K is the urea clearance of the dialyser) if the true equilibrium blood concentration of urea is not used in the calculation of KT/V by the single-pool urea kinetic model (UKM). The measurement of the equilibrium urea concentration entails an additional blood sample 60 min after dialysis, but an estimate may be calculated using a blood sample taken 80 min following the start of dialysis together with an approximate formula derived from the 2-pool model. In 14 patients, an average error in the calculation of KT/V of 35% (range 19-75%) by the single-pool UKM was reduced to 13% (range 1-55%, but 8 measurements to < 7%) using the approximate technique. It is concluded that the approximate technique significantly improves the accuracy of dose estimation in high-efficiency dialysis without inconveniencing the patient.
Kidney International aims to inform the renal researcher and practicing nephrologists on all aspects of renal research. Clinical and basic renal research, commentaries, The Renal Consult, Nephrology sans Frontieres, minireviews, reviews, Nephrology Images, Journal Club. Published weekly online and twice a month in print.
Assuming that the clearance of urea from total body water (TBW) is flow limited, the authors developed a parallel flow model using physiologic data. Organ systems with a blood flow to water volume ratio of greater than 0.2 min-1 were allocated to the high flow system. Remaining organs were represented in the low flow system. In end-stage renal disease patients with minimal renal blood flow, the high flow system contained 20% TBW and received 70% of the systemic blood flow. The authors used this flow heterogeneity to predict the post-dialysis urea rebound (R) in 12 patients after 1 hr of hemodialysis. Dialyzer clearance was 248 +/- 14.5 ml/min (mean +/- SEM) Access recirculation was obviated by returning cleared blood into a central vein. In these patients, R at 1, 3, 5, 7, 10, and 15 minutes. after slowing dialyzer blood flow (Qb) from 383 +/- 18 to 50 ml/min was 3.8 +/- 2.9, 6.2 +/- 3.4, 7.6 +/- 3.1, 8.8 +/- 3.9, 9.0 +/- 4.1, and 9.9 +/- 4.4%, respectively. CO and QAc were modeled with values of 5.5 and 0.5 L/min, respectively. The modeled TBW was 35 L. Total body water derived by nomogram was 38.1 +/- 2.0 L. Our results suggest that the parallel-flow model for urea transport can be used to explain the amount and time course of post dialysis R on a physiologic basis.
The original formula proposed to estimate variable-volume single-pool (VVSP) Kt/V was Kt/V = -In(R - 0.008 * t - f * UF/W), where in the Kt/V range of 0.7 to 1.3, f = 1.0 (* denotes multiplication). This formula tends to overestimate Kt/V as the Kt/V increases above 1.3. Because higher Kt/V values are now commonly delivered, the validity of both the urea generation term (0.008 * f) and correction for UF/W were explored by solving VVSP equations for simulated hemodialysis situations, with Kt/V ranging from 0.6 to 2.6. The analysis led to the development of a second-generation formula, namely: Kt/V = -In(R - 0.008 * t) + (4-3.5 * R) * UF/W. The first and second generation formulas were then used to estimate the modeled VVSP Kt/V in 500 modeling sessions in which the Kt/V ranged widely from 0.7 to 2.1. An analysis of error showed that this second-generation formula eliminated the overestimation of Kt/V in the high ranges found with the first-generation formula. Also, total error (absolute value percent error + 2 SD) was reduced with the second-generation formula. These results led to the proposal of a new formula that can be used for a very wide range of delivered Kt/V.
The single-pool urea kinetic model assumes that urea is distributed within the body in a volume Vf at a uniform concentration. It may be used to describe the clearance of urea during and following haemodialysis, and to determine the value of the urea concentration at the end of dialysis Cf or the urea generation rate G, and the urea distribution volume Vf. The protein catabolic rate (PCR) is obtained using the ratio G/Vf. The sensitivity of the predictions of the model to small errors in the experimentally estimated model parameters is evaluated and suggests that the model prediction of Cf is relatively insensitive to errors in the estimate of G, but more sensitive to errors in Vf and dialyser clearance K. The determination of G from urea concentration measurements made during dialysis alone is very sensitive to errors. The accurate estimation of G requires the use of concentration measurements made between dialyses. The calculation of the PCR is very sensitive to errors in G and Vf since these are not independent and must be used with caution, particularly when used to compare values between patients.
In 26 patients, a linear relationship between delta Kt/V (equilibrated minus single pool) and dialysis efficiency K/V was noted (r = -0.72). To determine if such a relationship would be supported by formal urea kinetic analysis, t, Kd, and V were randomly varied in 1,400 simulations using both intracellular/extracellular and regional blood flow 2 pool variable volume models. In the intracellular/extracellular model, delta Kt/V was best correlated with Kd/Kc (r = -0.96), where Kc is the intercompartmental clearance. Kc was not correlated with V, which translated into a lack of correlation between delta Kt/V and V, and a better correlation between delta Kt/V and Kd than between delta Kt/V and K/V. In the regional blood flow model delta Kt/V was best correlated with Kd/QL (r = -0.99), where QL is the perfusion of the low flow compartment. QL was correlated with V because QL is a function of cardiac output, which varies with surface area and therefore with V. In the regional blood flow model, delta Kt/V did correlate with V (r = 0.49), and better with K/V (r = -0.76) than with K (r = -0.47), similar to the results in patients. The slope of delta Kt/V on K/V depended upon fQL (the fractional perfusion of the low flow compartment) and on cardiac index. At an fQL of 0.15 and a cardiac index of 2.85, the theoretical slope was similar to that seen in observational data: delta Kt/V = -0.6 x K/V + 0.03. The results show that the regional blood flow model predicts the observed relation between delta Kt/V and K/V, whereas the intracellular/extracellular model fails in this task unless one arbitrarily ties Kc to V.
We measured postdialysis urea rebound (PDUR) 30 minutes after dialysis in 92 chronic hemodialysis patients. The impact of PDUR on the estimation of dialysis delivery assessed by urea reduction ratio and Kt/V was evaluated. Total recirculation, access plus cardiopulmonary, was measured at the end of dialysis with the two-needle low blood flow method. The mean age of the 92 patients (49 men and 43 women) was 59.6 +/- 1.4 years. Thirty-eight patients had been receiving erythropoietin therapy for more than 3 months. Fifteen patients had central venovenous access and 77 had peripheral arteriovenous access. Sixty-five patients were dialyzed using hemophan membranes and 27 were dialyzed using polyacrylonitrile membranes. The mean blood flow rate was 240 +/- 28 mL/min and the mean length of the hemodialysis sessions was 3.6 +/- 0.1 hours. Kt/V was calculated with Daugirdas' second-generation formula. The mean PDUR was 16.6% +/- 0.8% (range, 2% to 44%) (n = 92), and significantly decreased the mean urea reduction ratio from 61.7% +/- 0.8% to 55.5% +/- 0.9%, the mean Kt/V from 1.14 +/- 0.03 to 0.97 +/- 0.02, and the mean protein catabolic rate from 1.06 +/- 0.04 to 0.98 +/- 0.02 (P = 0.0001). The effective Kt/V at 30 minutes postdialysis was well predicted by using a recently proposed equation: eKt/V30 = Kt/Vsp - (0.6 x Kt/Vsp/t) + 0.03, with a mean value corresponding also to 0.97 +/- 0.02. However, this estimation was less predictive in patients with very high PDUR. Moreover, PDUR showed only a weak negative correlation with dialysis session length (r = -0.28) and predialysis patient weight (r = -0.29), and showed no correlation with predialysis serum urea level or with blood flow rate. However, dialysis efficiency, as assessed by K/V, presented a correlation of 0.54 with both PDUR and the difference in Kt/V when using urea immediately postdialysis and at 30 minutes. The mean total recirculation was 7.4% +/- 0.6% (n = 86). Postdialysis urea rebound, calculated between 30 or 120 seconds and 30 minutes after dialysis to deduce the influence of recirculations, was reduced but remained important with a mean of 11.8% +/- 0.7%. Thus, total recirculation contributed to nearly 30% of PDUR. The 24 patients with PDUR > or = 20% were compared with the 68 patients with PDUR lower than 20%: women and patients with higher K/V and higher total recirculation presented greater PDUR. Because of relatively few predictive factors for PDUR, its potential considerable impact on dialysis delivery estimation, and the influence of recirculations on the total PDUR amount, total recirculation and PDUR should be determined on an individual basis in chronic hemodialysis patients. The equation proposed to estimate effective Kt/V at 30 minutes is accurate in most patients with PDUR lower than 30% and is a simple alternative.
A model of urea distribution comprising well-mixed intracellular and extracellular compartments, with diffusive transfer of urea between compartments, is used to study blood urea variation during haemodialysis. Assuming that a typical adult patient (weight 70 kg and urea generation rate 5 mg min-1) is dialysed using a high-efficiency dialyser (urea clearance 0.2- 0.251 min-1) at an ultrafiltration rate of 10 ml min-1, it is shown that a perturbation analysis can account for the effects of ultrafiltration and urea generation. This permits an analytical solution of the equations which describe the variation in solute concentration in each of the compartments, which may be compared with the solution obtained when urea generation and ultrafiltration are neglected. For a typical adult patient with urea distribution volume in the range 25-40 l undergoing high-efficiency haemodialysis, the analysis suggests that the contribution of ultrafiltration to the variation in urea concentration during dialysis is similar in magnitude to the experimental errors in measuring the blood concentration of urea and that a constant volume model will be sufficiently accurate to describe urea clearance in many patients.
The mathematics used for urea kinetic modeling are currently based on a single-pool distribution of urea throughout the body. In this study, we evaluated which one of a single- or a two-pool model would be more appropriate for the prediction of directly measured urea decay during hemodialysis. A numerical method was used which minimizes the relative root mean square (RMS) error between a calculated single- or two-pool urea decay curve and the measured intradialysis decay in 13 equilibrated dialysis patients. Using a two-pool model, the RMS error was markedly lower (1.27 +/- 0.72%) than the values obtained with a single-pool model, either based on multiple urea concentrations (RMS error 3.14 +/- 1.36%; p < 0.01 vs. two-pool model) or only on pre- and postdialysis urea (RMS error 5.00 +/- 2.38%; p < 0.001). This resulted for the single-pool model in an overall underestimation of urea generation, distribution volume (V) and protein catabolic rate and in an overestimation of Kt/V versus the two-pool model. In individual cases, the difference reached up to 18.7%. Comparison of V calculated from the two-pool model versus V values determined from anthropometric formulae (Watson) resulted in similar mean values (34.05 +/- 4.87 vs. 33.09 +/- 4.19 liters; p = NS), with a weak correlation (n = 13, r = 0.75, p = 0.003). Individual values, however, again differed by up to more than 20%. In conclusion, the use of single-pool kinetic models, as well as of anthropometric estimations of V, should be regarded with care, especially when individual patients are considered instead of groups. The two-pool model follows the directly measured urea decay more exactly which results in substantial differences in calculated kinetic parameters.
When accurate, non-urea-based methods of measuring recirculation are used, recirculation is usually absent in arteriovenous (AV) accesses. When urea-based methods are used to measure recirculation in AV accesses, falsely elevated recirculation rates are common. These errors are due to AV and venovenous disequilibrium (peripheral vein method), delayed systemic sampling (two-needle methods), and errors in urea measurement (all methods). The literature suggests that recirculation in central venovenous (CV) catheters is approximately 5%. The methods used for these determinations have all been urea based. However, there are few theoretical problems in using urea-based measurements for measuring recirculation in this setting, making it more likely that these values are accurate. When hemodialysis via CV and AV accesses are compared, equilibrated Kt/V values differ significantly for the same single-pool Kt/V when 15-second postdialysis blood urea nitrogen values are used for modeling, but differ minimally when 2-minute postdialysis samples are used. The impact of transient retrograde blood flow in the superior vena cava on recirculation and whether dialysis efficiency is influenced by the exact site of CV catheter placement (superior vena cava v right atrium) is uncertain.
A two-pool, variable volume urea kinetic model for estimation of solute removal in hemodialysis is solved analytically, and closed form expressions are presented for urea concentration in both compartments, both during dialysis and between dialyses. This approach also includes an estimation of the extent of the post dialysis rebound phenomenon of urea concentration. A method is presented to estimate values for the urea generation rate G, the distribution volume V and its partition in two compartments with volumes alpha 1V and alpha 2V (alpha 1-alpha 2-1), the total clearance K, and intercompartmental transfer coefficient X. To apply this analysis, several measurements are needed as input; the urea concentration at the end of a dialysis, the evolution of this concentration during the next dialysis, with at least four measurements including the initial and the final concentration, the volume of the dialysate, and its urea concentration. The main results are: the magnitude of the rebound is approximately proportional to alpha 2(2) K/X; the accuracy of the parameter estimation does not improve much further by taking more than six measurements during dialysis.
To look for patients with extreme urea rebound, we drew intradialytic samples one third of the way into dialysis during routine modeling for 3 months. The samples taken postdialysis were obtained after stopping the blood pump, without any slow flow period. Using the Smye equations, the intradialytic urea level was used to predict urea rebound, expressed as Kt/V-equilibrated minus Kt/V-single pool (deltaKt/V). Results were averaged for the 3-month period in 369 patients. Mean estimated deltaKt/V was -0.20 +/- 0.13, which was similar to but slightly higher than the predicted value (-0.6 x K/V + 0.03) of -0.19 +/- 0.04. In 27 patients, extreme rebound (mean deltaKt/V < -0.40) was found. Sixteen of these patients consented to further study, but only after access revision in four patients. In these patients, additional slow flow samples after 15 seconds and 2 minutes of slow flow, respectively, were drawn one third of the way into dialysis and postdialysis, and a sample was drawn 30 minutes after dialysis. On restudy, postdialysis rebound was still high with full flow samples deltaKt/V = -0.40 +/- 25, but was much lower (-0.18 +/- 0.07) and similar to predicted rebound (-0.19 +/- 0.05; P = NS) when based on 15-second slow flow samples. Eight of the 16 had marked (>15%) access recirculation by urea sampling, and deltaKt/V based on full flow post samples correlated with access recirculation (r = -0.91). The results suggest that the Smye method is valuable for identifying patients with aberrantly large postdialysis rebound values. When the postdialysis samples are drawn without an antecedent slow flow period, most patients with extreme rebound values turn out to have marked access recirculation.
Kidney International aims to inform the renal researcher and practicing nephrologists on all aspects of renal research. Clinical and basic renal research, commentaries, The Renal Consult, Nephrology sans Frontieres, minireviews, reviews, Nephrology Images, Journal Club. Published weekly online and twice a month in print.
The ongoing HEMO Study, a National Institutes of Health (NIH) sponsored multicenter trial to test the effects of dialysis dosage and membrane flux on morbidity and mortality, was preceded by a Pilot Study (called the MMHD Pilot Study) designed to test the reliability of methods for quantifying hemodialysis. Dialysis dose was defined by the fractional urea clearance per dialysis determined by the predialysis BUN and the equilibrated postdialysis BUN after urea rebound is completed (eKt/V). In the Pilot Study the blood side standard for eKt/V was calculated from the predialysis, postdialysis, and 30-minute postdialysis BUN. Four techniques of approximating eKt/V that eliminated the requirement for the 30-minute postdialysis sample were also evaluated. The first adjusted the single compartment Kt/V using a linear equation with slope based on the relative rate of solute removal (K/V) to predict eKt/V (rate method). The second and third techniques used equations or mathematical curve fitting algorithms to fit data that included one or more samples drawn during dialysis (intradialysis methods). The fourth technique (dialysate-side) predicted eKt/V from an analysis of the time-dependent profile of dialysate urea nitrogen concentrations (BioStat method; Baxter Healthcare, Inc., Round Lake, IL, USA). The Pilot Study demonstrated the feasibility of conventional and high dose targets of about 1.0 and 1.4 for eKt/V. Based on the blood side standard method, the mean +/- SD eKt/V for patients randomized to these targets was 1.14 +/- 0.11 and 1.52 +/- 0.15 (N = 19 and 16 patients, respectively). Single-pool Kt/Vs were about 0.2 Kt/V units higher. Results were similar when eKt/V was based on dialysate side measurements: 1.10 +/- 0.11 and 1.50 +/- 0.11. The approximations of eKt/V by the three blood side methods that eliminated the delayed 30-minute post-dialysis sample correlated well with eKt/V from the standard blood side method: r = 0.78 and 0.76 for the single-sample (Smye) and multiple-sample intradialysis methods (N = 295 and 229 sessions, respectively) and 0.85 for the rate method (N = 295). The median absolute difference between eKt/V computed using the standard blood side method and eKt/V from the four other methods ranged from 0.064 to 0.097, with the smallest difference (and hence best accuracy) for the rate method. The results suggest that, in a dialysis patient population selected for ability to achieve an equilibrated Kt/V of about 1.45 in less than a 4.5 hour period, use of the pre and postdialysis samples and a kinetically derived rate equation gives reasonably good prediction of equilibrated Kt/V. Addition of one or more intradialytic samples does not appear to increase accuracy of predicting the equilibrated Kt/V in the majority of patients. A method based on dialysate urea analysis and curve-fitting yields results for equilibrated Kt/V that are similar to those obtained using exclusively blood-based techniques of kinetic modeling.
The steady decline in blood urea during high efficiency hemodialysis is followed by a rebound phase after dialysis in which the level of urea rises to an equilibrium value (Ct + 30) that may be up to 20% higher than the immediate post dialysis (Ct) concentration. The artificially low urea concentration immediately after dialysis leads to an overestimate of the efficiency of the dialysis calculated by Kt/V if the true equilibrium blood concentration of urea is not used in the calculation by the single-pool urea kinetic model. The measurement of equilibrium urea concentration requires a blood sample approximately 30 min after hemodialysis, which is an encumbrance on dialysis patients. This study was undertaken to determine whether an intradialytic sample taken 30 min before the end of dialysis (Ct - 30) may be representative of the equilibrium sample, and to compare the Kt/V using the Ct - 30 and Ct + 30 samples. Thirty-six patients were studied and blood urea concentrations were measured half an hour before the end of dialysis (Ct - 30), at the end of dialysis (Ct), and half an hour after the end of dialysis (Ct + 30). Kt/V (Daugirdas method) was calculated using urea concentration 30 min before the end of dialysis (Kt/Vt - 30) and was compared with Kt/V calculated using equilibrium urea concentration (Kt/Vt + 30). There were no significant differences between the Kt/Vt - 30 and the KtVt + 30 (1.25 versus 1.22, p = 0.65). The correlation between Kt/Vt - 30 and Kt/Vt + 30 was excellent with r2 = 0.93, regression y = 1.05 x -0.033. Kt/Vt - 30 also compared favorably with the Kt/V double pool method (Kt/Vdp) described by Daugirdas (1.25 versus 1.19, p = 0.23). Using the Ct - 30 to calculate Kt/V by the percent urea reduction methods of jindal (Kt/Vpru) decreases the Kt/V value by 0.14 on average, but it remains significantly higher than the Daugirdas method. The authors conclude that calculations using urea concentration 30 min before the end of dialysis improves the accuracy of dose estimation in high efficiency dialysis, without inconveniencing the patient.
According to previous studies, postdialysis urea rebound (PDUR) is achieved within 30-90 min, leading to an overestimation of Kt/V of between 15 and 40% in 3- to 5-hour dialysis. The purpose of the study was to assess the impact of PDUR on the urea reduction ratio (URR), Kt/V and normal protein catabolic rate (nPCR) with long 8-hour slow hemodialysis.
This study was performed in 18 patients (13 males/5 females), 62.5 +/- 11.7 years of age, hemodialyzed for 3-265 months. Initial nephropathies were: 3 diabetes; 2 polycystic kidney disease; 3 interstitial nephritis; 2 nephrosclerosis; 3 chronic glomerulonephritis, and 5 undetermined. Residual renal function was negligible. The dialysis sessions were performed using 1- to 1.8-m2 cellulosic dialyzers during 8 h, 3 times a week. Blood flow was 220 ml/min, dialysate flow 500 ml/min, acetate or bicarbonate buffer was used. Serial measurements of the urea concentration were obtained before dialysis, immediately after dialysis (low flow at t = 0), and at 5, 10, 20, 30, 40, 60, 90 and 120 min, and before the next session. The low-flow method was used to evaluate the access recirculation, second-generation Daugirdas formulas for Kt/V, and Watson formulas for total body water volume estimation. The difference between the expected urea generation (UG) and urea measured after dialysis (global PDUR) defines net PDUR (n-PDUR).
The n-PDUR usually became stable after 58 +/- 25 (30-90) min. Its mean value was 17 +/- 10% of the 30-second low-flow postdialysis urea (3.9 +/- 2 mmol/l). This small postdialysis urea value and the importance of UG in comparison with shorter dialysis justify the use of n-PDUR. Ignoring n-PDUR would lead to a significant 4% overestimation (p < 0.001) of the URR (79 +/- 7 vs. 76 +/- 8%), 12% of Kt/V (1.9 +/- 0.4 to 1.7 +/- 0.38) and 4% of the nPCR (1.1 +/- 0.3 to 1.05 +/- 0.3). n-PDUR correlated negatively with postdialysis urea (r = 0.45 p = 0.05), positively with URR (r = 0.31 p = 0.01) and Kt/V (r = 0.3 p = 0.03) but not with K, and negatively with the urea distribution volume (r = 0.33 p = 0.05). Mean total recirculation, ultrafiltration rate, predialysis urea levels and urea clearance did not correlate with n-PDUR.
We found a significant PDUR in long-slow hemodialysis after a mean of 1 h after dialysis. This PDUR has a less important impact upon dialysis delivery estimation than short 3- to 5-hour hemodialysis, especially for the lower Kt/V or URR ranges. This is explained by the low-flux, high-efficiency, and long-term dialysis. Its inter-individual variability incites us to calculate PDUR on an individual basis.
Postdialysis urea rebound (PDUR) is a cause of Kt/V overestimation when it is calculated from predialysis and the immediate postdialysis blood urea collections. Measuring PDUR requires a 30- or 60-minute postdialysis sampling, which is inconvenient. Several methods had been devised for a reasonable approach to determine PDUR-equilibrated Kt/V in short dialysis without the need for a delayed sample. The aim of our study was to compare these different Kt/V methods during the longer eight-hour hemodialysis sessions, and to determine the optimum intradialytic urea sample time that fits best with PDUR.
The study included 21 patients (mean age 71.9 years) who were hemodialyzed for 60+/-60 months at three times eight hours weekly, using bicarbonate dialysate and cellulosic membranes. Blood urea samples were obtained at onset, and then at 17, 33, 50, 66, 75, 80, 85, and 100% of the dialysis session times, after 30 seconds of low flow, and then at 60-minutes postdialysis. All patients had a meal during dialysis. We compared four different formulas of Kt/V [(a) Kt/V-Smye with a 33% dialysis time urea sample, (b) two-pool equilibrated eKt/V, (c) Kt/V-std (Daugirdas-2) obtained with an immediate postdialytic sample, and (d) the different intradialytic urea samples for Kt/V (50, 66, 75, 80, and 85% of dialysis time)] with the equilibrated 60-minute PDUR Kt/V (Kt/V-r-60) formula as the reference method.
The mean PDUR was 17.2+/-9%, leading to an overestimation of Kt/V-std by 12.2%. Kt/V-r-60 was 1.68+/-0.34. Kt/V-std was 1.88+/-0.36 (Delta = 12.2+/-4.8%, r = 0.8). eKt/V was 1.77+/-0.3 (Delta = 5+/-5%, r = 0.96), and Kt/V-Smye was 1.79+/-0.47 (Delta = 5.2+/-14%, r = 0.9). The best time for the intradialytic sampling was 80% (that is, at 6 hr and 24 min). The Kt/V-80 was 1.64+/-0.3 and was best fitted with Kt/V-r-60 (Delta = -1.8+/-8%, r = 0.91). The mean intradialytic urea evolution showed a three-exponential rate, in discrepancy with the two-exponential rate theoretical model.
These results confirm that a significant postdialysis rebound exists in an eight-hour dialysis. An intradialytic urea sample taken at 80% of the total session time permits an estimation of the 60-minute Kt/V-rebound without the necessity of taking a delayed sample, with better accuracy than eKt/V or especially Kt/V-Smye. This may be related to a particular urea kinetics curve on the longer dialysis duration, which needs to be studied further.
The volume of urea distribution (V) is usually derived from single-pool variable volume urea kinetics. A theoretical analysis has shown that modeled single-pool V (Vsp) is overestimated when the urea reduction ratio (URR) is greater than 65 to 70% and is underestimated when the URR is less than 65%. The "true" volume derived from double-pool kinetics (Vdp) does not exhibit this effect. An equation has been derived to adjust Vsp to the expected Vdp.
To validate these theoretical predictions, we examined data from the Hemodialysis (HEMO) Study to assess the performance of Vdp as estimated from Vsp using the previously published prediction equation. For increased precision, both Vsp and Vdp were factored by anthropometric volume (Va). Patients were first dialyzed with a target equilibrated dialysis dose (eKt/V) of 1.45 during a baseline period and were then randomly assigned to eKt/V targets of either 1. 05 (a URR of approximately 67%) or 1.45 (a URR of approximately 75%). A blood sample was obtained one hour after starting dialysis during one dialysis in each patient.
Vsp/Va was (mean +/- SD) 1.014 +/- 0.127 in 795 patients during the baseline period when the URR was approximately 1.45. During the first modeled dialysis after randomization, the Vsp/Va fell to 0.961 +/- 0.138 in the group with an eKt/V target of 1.05, but did not change significantly under the high eKt/V goal. The correction of Vsp to Vdp using the prediction equation resulted in a Vdp/Va ratio of 0.96 to 0.98 in all three circumstances without significant differences. When a blood sample was drawn one hour after starting dialysis, the apparent Vsp/Va ratio at one hour was much lower at 0.708 +/- 0.139. However, the mean Vdp/Va ratio, computed using the correction equation, was 0.968 +/- 0.322, which was similar to the Vdp/Va ratio calculated from the postdialysis blood urea nitrogen.
These data suggest that the previously derived formula for adjusted Vsp is valid experimentally. The Vsp/Vdp correction should be useful for prescribing hemodialysis with either a very low Kt/V (for example, daily and early incremental dialysis) or a very high Kt/V.
Blood urea nitrogen (BUN) concentration rebounds logarithmically for 1 hour after a hemodialysis treatment. We have previously devised and evaluated an equilibrated Kt/V (eqKt/V) estimation method using logarithmic extrapolation of the BUN increase from 30 seconds to 15 minutes postdialysis in six pediatric hemodialysis patients. The current study evaluates logarithmic extrapolation in 15 additional pediatric patients. Mean measured equilibrated BUN (eqBUN) and estimated BUN at equilibrium (estBUN) using logarithmic extrapolation were 23.1 +/- 9.2 and 23.0 +/- 9.4 mg/dL, respectively. The mean absolute difference between estBUN and eqBUN was 0.7 +/- 0. 4 mg/dL (range, 0.1 to 1.55 mg/dL). All treatments had an absolute difference less than the SD of the laboratory measurement itself. The mean absolute percentage of difference between eqKt/V using eqBUN and estimated double-pool equilibrated Kt/V (estKt/V) using estBUN from logarithmic extrapolation was 3.4% +/- 2.3% and did not vary as a function of patient size, urea generation rate, dialyzer urea clearance, Kd/V, or ultrafiltration fraction. Mean absolute percentages of difference between eqKt/V and Kt/V estimated by the Tattersall, Daugirdas, or Maduell formulas were 4.5% +/- 3.9%, 4.4% +/- 3.7%, and 6.7% +/- 8.3%, respectively. Total percentages of error (absolute mean percentage of error + 2 SD) between eqKt/V and estKt/V by logarithmic extrapolation or the Tattersall, Daugirdas, or Maduell formulas were 8.0%, 12.3%, 11.8%, and 22.3%, respectively. The greater accuracy of logarithmic extrapolation compared with other methods of double-pool Kt/V estimation held true for patients weighing less than 35 kg. We have validated the use of an easy and accurate method requiring only an additional 15-minute posttreatment BUN level to estimate double-pool Kt/V in children.
Compartment effects in hemodialysis are important because they reduce the efficiency of removal of the compartmentalized solute during dialysis. The dialyzer can only remove those waste products that are presented to it, and then only in proportion to the concentration of the solute in the blood. Classically a two-compartment system has been modeled, with the compartments arranged in series. Because modeling suggests that the sequestered compartment is larger than the accessible compartment, an assumption has been made that the sequestered compartment is the intracellular space. For urea and other solutes that move easily across many cell membranes, compartmentalization may be flow related, that is, related to sequestration in organs (muscle, skin, bone). Although mathematically urea rebound and mass balance can be described with either model, the flow-related model best explains data showing that urea rebound after dialysis is increased during ultrafiltration, diminished during high cardiac output states, and also reduced during exercise. Whether compartmentalization is increased in vasoconstricted intensive care unit patients receiving acute dialysis remains an open question.
Our interest in urea kinetic modeling (UKM) was stimulated some 30 years ago at the time of the advent of hollow fiber kidneys with greatly improved urea transport. This led to examination of the interaction between time and clearance in computing the dialysis dose. In early studies a fixed-volume single-pool UKM was used but this frequently gave spurious high volumes and led to the advent of the variable-volume single-pool model. The role of volume calculation in assessment of the delivered dialysis dose and the value of normalized protein catabolic rate (nPCR) calculation are reviewed. More recently quantification of double-pool effects has become simplified and now is widely used for UKM. The National Cooperative Dialysis Study (NCDS) resulted in the concept of dose quantification by Kt/V. This is reviewed, including the controversy surrounding interpretation of the NCDS. Currently there is great interest in more frequent dialysis, 4-6 days/week. The development of a new dose parameter, the standard Kt/V (stdKt/V), to enable quantitative comparison of dose with widely varying dose schedules is discussed.
A regional flow model (RFM) can establish the missing link between hemodynamics and solute removal. We tried to simulate post-dialysis urea rebound using a RFM for the purpose of evaluating the validity of this model.
Eight patients on maintenance hemodialysis with negligible renal function were investigated. The parameters of the RFM were estimated so as to fit the calculated values of urea nitrogen to the measured values during a dialysis session. The estimated parameters were total urea distribution volume (TUV), systemic blood flow (Qsys), flow fraction (fQH) and volume fraction (fVH) of the high-flow system. Thirteen types of parameter sets were used for the estimation. The urea rebound at 60 min after a dialysis session (Creb) and the rebound ratio (RR) were calculated using these estimated parameters. The accuracy of the calculated Creb and RR was assessed.
The accuracy of Creb and RR determined using estimated TUV, by taking Qsys as systemic blood flow calculated from ultrasonic echo cardiogram (Qucg), fQH as 0.8, and fVH as 0.2, was insufficient (method 1a). The accuracy of these values was significantly increased by taking fQH as 0.85 (method 1b). The estimation of Qsys with TUV did not improve the accuracy of Creb and RR (methods 2a and 2b). The estimation of fQH, fVH, and TUV (method 8) increased the accuracy of Creb and RR significantly compared with method 1a, but not compared with method 1b. Even with method 1b or method 8, the percentage RR was less than 90% in two patients.
By taking fQH as 0.85, an acceptably accurate simulation of urea rebound can be accomplished with the necessity to estimate only TUV. The simulation was not significantly improved by the estimation of Qsys, fQH, and fVH. The RFM is useful in practice, although it has some limitations.
Hemodialysis urea rebound: the effect of increasing dialysis efficiency
- D M Spiegel
- P L Baker
- S Babcock
- R Cantiguglia
- M Klein