Prediction of cardiovascular risk in hemodialysis patients by data mining.

OBJECTIVES: The objective of this work was to contribute to the development, validation and application of data mining methods for prediction in decision support systems in medicine. The particular focus was on the prediction of cardiovascular risk factors in hemodialysis patients, specifically the interventricular septum (IVS) thickness of the heart of individual patients as an important quantitative indicator to diagnose left ventricular hypertrophy. The work was based on data from 63 long-term hemodialysis patients of the KfH Dialysis Centre in Jena, Germany. METHODS: The approach applied is based on data mining methods and involves four major steps: data based clustering, cluster based rule extraction, rulebase construction and cluster and rule based prediction. The methods employed include crisp and fuzzy algorithms. At each step, logical and medical validation of results was carried out. Different sets of randomly selected patient data were used to train, test and optimize the clusterbases and rulebases for prediction. RESULTS: Using the best clusterbase/rulebase combination designed, the IVS thickness cluster ('small' or 'large') was predicted correctly for 30 of the 35 patients with known IVS values in the training data set; no patient was predicted incorrectly and 5 were parity predicted. For the test data set, 4 of the 6 patients with known IVS values were predicted correctly, no patient incorrectly and 2 parity. These results did not substantially differ from those obtained using the second best clusterbase/rulebase combination which was finally recommended for use based on further performance criteria. The prediction of the IVS thickness clusters of the 22 patients with unknown IVS values also yielded good results that were (and could only be) validated by a medical individual risk assessment of these patients. CONCLUSIONS: The approach applied proved successful for the cluster and rule based prediction of a quantitative variable, such as IVS thickness, for individual patients from other variables relevant to the problem. The results obtained demonstrate the high potential of the approach and the methods developed and validated to support decision-making in hemodialysis and other fields of medicine by individual risk prediction.

Inflammatory response of a new synthetic dialyzer membrane. A randomised cross-over comparison between polysulfone and helixone.

Hemodialysis patients suffer from chronic inflammation due to intradialytic contact of blood with artificial materials. The FX 60 dialyzer which belongs to the new FX-class series of dialyzers is composed of the new membrane Helixone. This membrane is derived from the original Fresenius Polysulfone membrane. The FX-class design is based on modified geometry of fibres and housing and has resulted in a new dialyzer with improved efficiency, safety and ease of handling compared to the F series (F 60S) dialyzer. The aim of the study was to investigate whether the biocompatibility pattern in terms of inflammatory parameters of the new type of polysulfone dialyzer has changed compared to the standard. A clinical in vivo study was conducted to compare the intradialytic inflammatory response of the two dialyzers, FX 60 and F 60S. Eight chronic dialysis patients were selected for the study: mean age 65.5 +/- 15.5 years, mean time on dialysis 100 +/- 95 months. The randomized cross-over study involved a treatment period of 2 weeks (total 6 sessions), one week with each dialyzer, starting with one or the other according to the randomization scheme. Blood samples were taken at 0 (T0), 15, 60, and 240 minutes to evaluate white blood cell (WBC) count, complement factor C5a, leukocyte elastase, soluble intercellular adhesion molecule 1 (sICAM-1), platelet count, C-reactive protein (CRP). At 15 min, WBC count showed a comparably, low decrease for both dialyzers: -7.6% for FX 60 versus -6.6% for F 60S, p=not significant (ns). At the same time the C5a concentration decreased from 15.0 +/- 7.5 ng/ml to 13.5 +/- 6.7 ng/ml (p=ns) for FX 60, and from 15.1 +/- 12.5 ng/ml to 14.9 +/- 25.0 ng/ml for F 60S (p=ns). The elastase concentration progressively increased over time with no statistical difference between the two dialyzers. The levels of sICAM-1, CRP, and platelet count were similar at each time point for both dialyzers, varying around the baseline values (p=ns). No significant difference emerged in terms of inflammatory response between the two dialyzers, hemo demonstrating that the biocompatibility of the F-series was maintained in the FX-class series of dialyzers and is independent of design factors.

In vitro evaluation of the hydraulic permeability of polysulfone dialysers.

An in vitro set-up has been designed to study the hydraulic permeability of hollow fiber dialysers. Forward and reverse dialysate ultrafiltration were determined using both sterile dialysers and samples with a protein layer settled on the membrane (Fresenius F6, F8, F60 and F80). The ultrafiltration coefficient KUF (ml/h.mmHg) was calculated as the ratio of volumetrical flow (QUF) and transmembrane pressure (TMP) measurements. The protein layer on the membrane was induced either by recirculating human plasma through the dialysers (in vitro) or by a standard hemodialysis session (in vivo). KUF is largely independent of TMP up to 600mmHg (low flux) and 60mmHg (high flux) for forward and reverse flow In sterile dialysers, backfiltration yields a significantly different KUF except for the F80. An in vitro induced protein layer on the membrane decreases KUF15-30% (forward) and 4-12% (backward) in low flux and 45-70% (forward) and 65-73% (backward) in high flux dialysers.

Phosphate removal and hemodialysis conditions.

Hyperphosphatemia is frequently found in hemodialysis patients, and the association with an increased risk of mortality has been demonstrated. Other authors have linked hyperphosphatemia to increased cardiovascular mortality. The normalization of phosphate plasma levels is therefore an important goal in the treatment of end-stage renal disease patients. Absorption of phosphate from the food exceeds the elimination through a hemodialysis treatment, and this leads to a chronic phosphate load for the majority of hemodialysis patients. This imbalance should be improved by either a reduction of phosphate absorption or an increased removal of phosphate. A reduction of phosphate absorption can be achieved by reducing the amount of phosphate in the diet or by the administration of phosphate binders. Unfortunately, these measures imply practical difficulties, for example, a lack of patient compliance or other side effects. When considering modifications of the hemodialysis treatment, an essential understanding of the kinetics of dialytic phosphate removal is mandatory. Phosphate is unevenly distributed in different compartments of the body. Only a very small amount of phosphate is present in the easily accessible plasma compartment. The major part of phosphate removed during hemodialysis originates from the cytoplasm of cells. A transfer from intracellular space to the plasma and further from the plasma to the dialysate is necessary. However, if we consider improvement to phosphate removal by dialysis procedures, full dialyzer clearance is effective in only the initial phase of the dialysis treatment. After this initial phase, the transfer rate for phosphate from the intracellular space to the plasma becomes the rate-limiting step for phosphate transport. Attempts to improve this transfer rate have recently been investigated by acidosis correction, but turned out not to be consistently successful. Furthermore, modifications of the treatment schedule have been described in the literature as measures to influence the phosphate balance consistently. Successful improvements of the phosphate balance can be achieved specifically through increasing the frequency of the dialysis treatments.

The influence of skull-conductivity misspecification on inverse source localization in realistically shaped finite element head models.

The electric conductivities of different tissues are important parameters of the head model and their precise knowledge appears to be a prerequisite for the localization of electric sources within the brain. To estimate the error in source localization due to errors in assumed conductivity values, parameter variations on skull conductivities are examined. The skull conductivity was varied in a wide range and, in a second part of this paper, the effect of a nonhomogeneous skull conductivity was examined. An error in conductivity of lower than 20% appears to be acceptable for fine finite element head models with average discretization errors down to 3 mm. Nonhomogeneous skull conductivities, e.g., sutures, yield important mislocalizations especially in the vincinty of electrodes and should be modeled.