AQP1 Promoter Variant, Water Transport, and Outcomes in Peritoneal Dialysis.

BACKGROUND: Variability in ultrafiltration influences prescriptions and outcomes in patients with kidney failure who are treated with peritoneal dialysis. Variants in AQP1, the gene that encodes the archetypal water channel aquaporin-1, may contribute to that variability. METHODS: We gathered clinical and genetic data from 1851 patients treated with peritoneal dialysis in seven cohorts to determine whether AQP1 variants were associated with peritoneal ultrafiltration and with a risk of the composite of death or technique failure (i.e., transfer to hemodialysis). We performed studies in cells, mouse models, and samples obtained from humans to characterize an AQP1 variant and investigate mitigation strategies. RESULTS: The common AQP1 promoter variant rs2075574 was associated with peritoneal ultrafiltration. Carriers of the TT genotype at rs2075574 (10 to 16% of patients) had a lower mean (±SD) net ultrafiltration level than carriers of the CC genotype (35 to 47% of patients), both in the discovery phase (506±237 ml vs. 626±283 ml, P = 0.007) and in the validation phase (368±603 ml vs. 563±641 ml, P = 0.003). After a mean follow-up of 944 days, 139 of 898 patients (15%) had died and 280 (31%) had been transferred to hemodialysis. TT carriers had a higher risk of the composite of death or technique failure than CC carriers (adjusted hazard ratio, 1.70; 95% confidence interval [CI], 1.24 to 2.33; P = 0.001), as well as a higher risk of death from any cause (24% vs. 15%, P = 0.03). In mechanistic studies, the rs2075574 risk variant was associated with decreases in AQP1 promoter activity, aquaporin-1 expression, and glucose-driven osmotic water transport. The use of a colloid osmotic agent mitigated the effects of the risk variant. CONCLUSIONS: A common variant in AQP1 was associated with decreased ultrafiltration and an increased risk of death or technique failure among patients treated with peritoneal dialysis. (Funded by the Swiss National Science Foundation and others.).

Role of hemostatic factors on the risk of venous thrombosis in people with impaired kidney function.

BACKGROUND: Factors explaining the association between impaired kidney function and venous thrombosis have not been identified so far. The aim of our study was to determine whether the association between impaired kidney function and venous thrombosis can be explained by the concurrent presence of genetic or acquired venous thrombosis risk factors. METHODS AND RESULTS: The glomerular filtration rate was estimated (eGFR) in 2473 venous thrombosis patients and 2936 controls from a population-based case-control study. Kidney function was grouped into 6 categories based on percentiles of the eGFR in the controls (>50th [reference], 10th-50th, 5th-10th, 2.5th-5th, 1st-2.5th, and <1st percentile). Several hemostatic factors showed a procoagulant shift with decreasing kidney function in controls, most notably factor VIII and von Willebrand factor. Compared with eGFR >50th percentile, factor VIII levels (adjusted mean difference, 60 IU/dL for the <1st eGFR percentile category) and von Willebrand factor levels (adjusted mean difference, 60 IU/dL for the <1st eGFR percentile category) increased with each percentile category. The odds ratios for venous thrombosis similarly increased across the categories from 1.1 (95% confidence interval, 0.9-1.3) for the 10th to 50th percentile to 3.7 (95% confidence interval, 2.4-5.7) for the <1st percentile category. Adjustment for factor VIII or von Willebrand factor attenuated these odds ratios, indicating an effect of eGFR on thrombosis through these factors. Adjustments for other risk factors for venous thrombosis did not affect the odds ratios. CONCLUSION: Impaired kidney function affects venous thrombosis risk via concurrently raised factor VIII and von Willebrand factor levels.

Type of arteriovenous vascular access and association with patency and mortality.

BACKGROUND: There are only a few risk factors known for primary patency loss in patients with an arteriovenous graft or fistula. Furthermore, a limited number of studies have investigated the association between arteriovenous access modality and primary patency loss and mortality. The aim of this study was to investigate risk factors for patency loss and to investigate the association between graft versus fistula use and outcomes (patency loss and mortality). METHODS: We prospectively followed 919 incident hemodialysis patients and calculated hazard ratios (HRs) for putative risk factors of primary patency loss using Cox regression. Furthermore, HRs were calculated to study the association between graft versus fistula use and two-year primary patency loss and two-year mortality. RESULTS: Cardiovascular disease, prior catheter use, lowest tertile of albumin, highest tertile of hsCRP, and lowest tertile of fetuin-A were associated with primary patency loss in both patients with grafts and fistulas. Increased age, female sex, and diabetes mellitus were only associated with primary patency loss in patients with a fistula. We did not observe an association between primary patency loss and BMI, residual GFR, levels of calcium, phosphorus, and total cholesterol. Furthermore, graft use as compared with fistula use was associated with an 1.4-fold (95% CI 1.0-1.9) increased risk of primary patency loss and with an 1.5-fold(95% CI 1.0-2.2) increased mortality risk. CONCLUSION: Cardiovascular disease, prior catheter use, albumin, hsCRP, and fetuin-A are risk factors for patency loss. Graft use as compared with fistula use was associated with an increased risk of patency loss and mortality.

Short-term versus long-term effects of depressive symptoms on mortality in patients on dialysis.

OBJECTIVE: Depressive symptoms seem to pose a risk factor for mortality among patients on dialysis. It is currently unknown whether the association is only short-lived and whether associations over time depend on specific causes of mortality. METHODS: In a prospective nationwide cohort study, 1528 patients with end-stage renal disease starting on dialysis completed the Mental Health Inventory. Patients were observed up to 5 years or until the end of follow-up in April 2011. Cox regression analyses were used to calculate associations between depressive symptoms and short-term (0-6 months), medium-term (6-24 months), or long-term (24-60 months) cardiovascular and noncardiovascular mortality. RESULTS: The adjusted hazard ratio (HR) was 1.43 (95% confidence interval [CI] = 1.08-1.88) for cardiovascular mortality and 2.07 (95% CI = 1.62-2.64) for noncardiovascular mortality. Depressive symptoms posed a strong risk factor for noncardiovascular mortality at the short term (HR = 2.82, 95% CI = 1.58-5.05), medium term (HR = 2.08, 95% CI = 1.40-3.09), and long term (HR = 1.84, 95% CI = 1.26-2.69), whereas the association between depressive symptoms and cardiovascular mortality was not observed during the first 6 months of follow-up (HR = 1.03, 95% CI = 0.49-2.15). CONCLUSIONS: Depressive symptoms at the start of dialysis therapy are associated with short-, medium-, and long-term mortality. The cause-specific mortality risk over time may help clinicians to understand multifactorial causes of the association between depressive symptoms and survival.

The -174G/C variant of IL6 as risk factor for mortality and technique failure in a large cohort of peritoneal dialysis patients.

BACKGROUND: Functional variants in the IL6 gene, in particular the -174G/C polymorphism (rs1800795), affect the mortality risk in dialysis patients. Peritoneal dialysis (PD) patients harbouring the C allele of the -174G/C polymorphism of IL6 showed faster peritoneal transport. The aim of this study was to investigate this IL6 variant as risk factor for mortality and technique failure in a large cohort of Caucasian PD patients. METHODS: A Dutch multicentre cohort of 398 incident PD patients (NECOSAD) was analysed. Survival analysis was performed for death and technique failure with a maximum follow-up of 5 years. A combined PD cohort from Amsterdam (Academic Medical Center, N = 71) and Brussels (Université catholique de Louvain Medical School, N = 102) was used for independent replication. RESULTS: In NECOSAD, 105 patients died on dialysis [incidence rate 10.3/100 person-years (py)], and 138 patients experienced technique failure (16.2/100 py), with peritonitis as important cause. Patients with the C/C genotype had a 71% increased mortality risk compared to patients with the G/G genotype (95% confidence interval 0.98-2.98); this effect was mainly a long-term effect: a 2.7-fold increased mortality risk was found in patients having survived 2 years since the start on dialysis, and a 1.7-fold increased risk for the combined end point (mortality or technique failure). In the combined replication cohort, no increased risks were found in patients with the C/C genotype. CONCLUSIONS: The C/C genotype of the -174G/C polymorphism was associated with an increased mortality risk in 398 Dutch incident PD patients. The existence of substantial differences between the two academic replication cohorts and the discovery cohort from NECOSAD and the limited power of these cohorts prevented an independent replication of the NECOSAD findings.

The role of psychosocial factors in ethnic differences in survival on dialysis in the Netherlands.

INTRODUCTION: Ethnic minority patients on dialysis are reported to have better survival rates relative to Caucasians. The reasons for this finding are not fully understood and European studies are scarce. This study examined whether ethnic differences in survival could be explained by patient characteristics, including psychosocial factors. METHODS: We analysed data of the Netherlands Cooperative Study on the Adequacy of Dialysis study, an observational prospective cohort study of patients who started dialysis between 1997 and 2007 in the Netherlands. Ethnicity was classified as Caucasian, Black or Asian, assessed by local nurses. Data collected at the start of dialysis treatment included demographic, clinical and psychosocial characteristics. Psychosocial characteristics included data on health-related quality of life (HRQoL), mental health status and general health perception. Cox proportional hazards analysis was used to explore ethnic survival differences. RESULTS: One thousand seven hundred and ninety-one patients were Caucasian, 45 Black and 108 Asian. The ethnic groups differed significantly in age, residual glomerular filtration rate, diabetes mellitus, erythropoietin use, plasma calcium, parathormone and creatinine, marital status and general health perception. No ethnic differences were found in HRQoL and mental health status. Crude hazard ratios (HRs) for mortality for Caucasians compared to Blacks and Asians were 3.1 [95% confidence interval (CI) 1.6-5.9] and 1.1 (95% CI 0.9-1.5), respectively. After adjustment for a range of potential explanatory variables, including psychosocial factors, the HRs were 2.5 (95% CI 1.2-4.9) compared with Blacks and 1.2 (95% CI 0.9-1.6) compared with Asians. CONCLUSIONS: Although patient numbers were rather small, this study demonstrates, with 95% confidence, better survival for Black compared to Caucasian dialysis patients and equal survival for Asian compared to Caucasian dialysis patients in the Netherlands. This could not be explained by patient characteristics, including psychosocial factors.

Using a genetic, observational study as a strategy to estimate the potential cost-effectiveness of pharmacological CCR5 blockade in dialysis patients.

BACKGROUND AND OBJECTIVE: Randomized clinical trials are expensive and time consuming. Therefore, strategies are needed to prioritise tracks for drug development. Genetic association studies may provide such a strategy by considering the differences between genotypes as a proxy for a natural, lifelong, randomized at conception, clinical trial. Previously an association with better survival was found in dialysis patients with systemic inflammation carrying a deletion variant of the CC-chemokine receptor 5 (CCR5). We hypothesized that in an analogous manner, pharmacological CCR5 blockade could protect against inflammation-driven mortality and estimated if such a treatment would be cost-effective. METHODS: A genetic screen and treat strategy was modelled using a decision-analytic Markov model, in which patients were screened for the CCR5 deletion 32 polymorphism and those with the wild type and systemic inflammation were treated with pharmacological CCR5 blockers. Kidney transplantation and mortality rates were calculated using patient level data. Extensive sensitivity analyses were performed. RESULTS: The cost-effectiveness of the genetic screen and treat strategy was &OV0556;18 557 per life year gained and &OV0556;21 896 per quality-adjusted life years gained. Concordance between the genetic association and pharmacological effectiveness was a main driver of cost-effectiveness. Sensitivity analyses showed that even a modest effectiveness of pharmacological CCR5 blockade would result in a treatment strategy that is good value for money. CONCLUSION: Pharmacological blockade of the CCR5 receptor in inflamed dialysis patients can be incorporated in a potentially cost-effective screen and treat programme. These findings provide formal rationale for clinical studies. This study illustrates the potential of genetic association studies for drug development, as a source of Mendelian randomized evidence from an observational setting.

The analysis of competing events like cause-specific mortality--beware of the Kaplan-Meier method.

Kaplan-Meier analysis is a popular method used for analysing time-to-event data. In case of competing event analyses such as that of cardiovascular and non-cardiovascular mortality, however, the Kaplan-Meier method profoundly overestimates the cumulative mortality probabilities for each of the separate causes of death. This article provides an introduction to the problem of competing events in Kaplan-Meier analysis. It explains cumulative incidence competing risk analysis and demonstrates on a cohort of elderly dialysis patients that, in contrast to the Kaplan-Meier method, application of this method yields unbiased estimates of the cumulative probabilities for cause-specific mortality.

Vitamin D status and clinical outcomes in incident dialysis patients: results from the NECOSAD study.

BACKGROUND: The majority of dialysis patients suffer from vitamin D deficiency, which might contribute to an adverse health outcome. We aimed to elucidate whether European dialysis patients with low 25-hydroxyvitamin D (25(OH)D) levels are at increased risk of mortality and specific fatal events. METHODS: This was a prospective cohort study of incident dialysis patients in the Netherlands (the NECOSAD). We selected all patients with measured 25(OH)D at 12 months after the start of dialysis, the baseline for our study. By Cox regression analyses, we assessed the impact of 25(OH)D levels on short-term (6 months of follow-up) as well as longer-term mortality (3 years of follow-up). Associations of 25(OH)D levels with cardiovascular and non-cardiovascular mortality were also determined. RESULTS: The data from 762 patients (39% females, age 59 ± 15 years, 25(OH)D = 18 ± 11 ng/mL) were available. Fifty-one and 213 patients died during a follow-up of 6 months and 3 years, respectively. After adjustments for possible confounders, the hazard ratio (HR) (with 95% CI) for mortality was 2.0 (1.0-3.8) for short-term and 1.5 (1.0-2.1) for longer-term mortality when comparing patients with 25(OH)D levels ≤ 10 ng/mL with those presenting with 25(OH)D levels > 10 ng/mL. Adjusted HRs for cardiovascular mortality were 2.7 (1.1-6.5) and 1.7 (1.1-2.7) for short- and longer-term mortality, respectively. For non-cardiovascular mortality, we observed no relevant association overall. The impact of 25(OH)D levels on clinical events was modified by parathyroid hormone (PTH) status, with low 25(OH)D levels meaningfully affecting outcomes only in patients with PTH levels above the median of 123 pmol/L. CONCLUSIONS: Vitamin D deficiency in dialysis patients is associated with an adverse health outcome, in particular with short-term cardiovascular mortality. Intervention studies are urgently needed to evaluate whether vitamin D supplementation improves health outcomes of dialysis patients.

Low cholesterol in dialysis patients--causal factor for mortality or an effect of confounding?

BACKGROUND: The association between cholesterol and mortality is reversed in end-stage renal disease (ESRD). This phenomenon has many potential explanations, one of them being 'time discrepancy of competing risks'. It states that hypercholesterolaemia is beneficial only in the short term, while it worsens survival over a long-term interval. It is also proposed that the reversed relationship between cholesterol and outcome is due to confounding protein-energy wasting. The aim of the study was to verify the hypothesis of time discrepancy of competing risks in 1191 incident dialysis patients (Netherlands Cooperative Study on the Adequacy of Dialysis). METHODS: Conditional Cox proportional hazards analysis was applied, where associations between cholesterol level and short-term versus long-term mortality were being compared. Furthermore, to evaluate associations between cholesterol and outcome free from confounding, Mendelian randomization was introduced, using apolipoprotein E (apoE) genotype. RESULTS: Hypercholesterolaemia (>240 mg/dL) was associated with improved 5-year survival when compared to the low cholesterol group (<200 mg/dL), hazard ratio (HR) = 0.62 (0.47-0.82), P < 0.001. However, conditional Cox proportional hazards analysis revealed that the reverse association between high cholesterol and all-cause mortality was evident only during the first year of follow-up, HR = 0.43 (0.23-0.80), P < 0.01, and then, gradually, declined. The apoE genotype significantly affected cholesterol concentration. The ε2 carriers, associated with low cholesterol, had significantly increased risk of non-cardiovascular mortality. CONCLUSIONS: Reverse association between cholesterol concentration and mortality in dialysis patients is short-termed, consistent with the hypothesis of time discrepancy of competing risks. Low cholesterol appeared to affect non-cardiovascular mortality in ESRD patients free from confounders.

Is fetuin-A a mortality risk factor in dialysis patients or a mere risk marker? A Mendelian randomization approach.

BACKGROUND: Low levels of circulating fetuin-A are associated with increased mortality in dialysis patients. This study aimed to examine a potential causative role for fetuin-A on mortality by investigating whether a functional polymorphism in the alpha2-Heremans-Schmid glycoprotein (AHSG) gene associates with mortality, and by estimating the causative effect of fetuin-A levels on mortality using a Mendelian randomization design. METHODS: One thousand and forty-three incident dialysis patients were genotyped for the Thr256Ser polymorphism (rs4918) and followed up for 5 years; in 549 patients, serum fetuin-A levels were measured. RESULTS: Carriers of a serine allele displayed lower fetuin-A levels (-0.07 g/L per allele, P < 0.001). A small increased mortality risk was observed for the Thr/Ser and Ser/Ser genotype compared with the Thr/Thr genotype (HR 1.03, 95% CI 0.83-1.28 and HR 1.10, 95% CI 0.78-1.55, respectively). Using the AHSG genotype as an instrumental variable, the causative HR of fetuin-A levels on mortality was estimated as 1.01 per 0.1-g/L increase. Inflammation and diabetes partially modified the association of fetuin-A levels with outcome. CONCLUSIONS: The Thr256Ser polymorphism was weakly associated with mortality, and no causative effect of fetuin-A levels on this outcome was observed. Other risk factors, including inflammation and diabetes, might lead to lower fetuin-A levels, and/or modify the effect of low fetuin-A on mortality in end-stage renal disease patients.

Haemodialysis catheters increase mortality as compared to arteriovenous accesses especially in elderly patients.

BACKGROUND: Catheter use has been associated with an increased mortality risk in haemodialysis patients. However, differences in the all-cause and cause-specific mortality risk between catheter use and arteriovenous access use in young and elderly haemodialysis patients have not yet been investigated. METHODS: In this prospective cohort study of 1109 incident haemodialysis patients from 38 centres in the Netherlands, hazard ratios (HRs) with 95% confidence intervals (95% CIs) were calculated for 2-year all-cause, infection-related and cardiovascular mortality in patients with a catheter as compared to patients with an arteriovenous access stratified for age (< 65 years and ≥ 65 years). RESULTS: Of the 1109 patients, 919 had an arteriovenous access and 190 had a catheter. The mortality rate was 76 per 1000 person-years in young patients with an arteriovenous access, 129 per 1000 person-years in young patients with a catheter, 222 per 1000 person-years in elderly patients with an arteriovenous access and 427 per 1000 person-years in elderly patients with a catheter. The adjusted HR was 3.15 (95% CI: 2.09-4.75) for elderly patients with a catheter as compared to young patients with an arteriovenous access. The adjusted HRs in elderly patients with a catheter as compared to elderly patients with an arteriovenous access were 1.54 (95% CI: 1.13-2.12) for all-cause mortality, 1.60 (95%: CI 0.62-4.19) for infection-related mortality and 1.67 (95% CI: 1.04-2.68) for cardiovascular mortality. CONCLUSIONS: Especially, elderly haemodialysis patients with a catheter have an increased all-cause, infection-related and cardiovascular mortality risk as compared to patients with an arteriovenous access.

Genetic association studies: discovery of the genetic basis of renal disease.

Genetic association studies are a means to investigate the causal role of genes in diseases in order to unravel pathways involved in the etiology of disease. There are two types of genetic association studies: hypothesis-driven studies, i.e. candidate gene studies, targeting genes with a known or presumed role in pathways or diseases of interest, and non-hypothesis-driven studies, i.e. genome-wide association studies, aiming for the discovery of new genetic associations. This educational article is an introduction to genetic association studies for nephrologists and researchers in the domain of kidney disease.

Vitamin D deficiency is associated with sudden cardiac death, combined cardiovascular events, and mortality in haemodialysis patients.

AIMS: Dialysis patients experience an excess mortality, predominantly of sudden cardiac death (SCD). Accumulating evidence suggests a role of vitamin D for myocardial and overall health. This study investigated the impact of vitamin D status on cardiovascular outcomes and fatal infections in haemodialysis patients. METHODS AND RESULTS: 25-hydroxyvitamin D [25(OH)D] was measured in 1108 diabetic haemodialysis patients who participated in the German Diabetes and Dialysis Study and were followed up for a median of 4 years. By Cox regression analyses, we determined hazard ratios (HR) for pre-specified, adjudicated endpoints according to baseline 25(OH)D levels: SCD (n = 146), myocardial infarction (MI, n = 174), stroke (n = 89), cardiovascular events (CVE, n = 414), death due to heart failure (n = 37), fatal infection (n = 111), and all-cause mortality (n = 545). Patients had a mean age of 66 ± 8 years (54% male) and median 25(OH)D of 39 nmol/L (interquartile range: 28-55). Patients with severe vitamin D deficiency [25(OH)D of ≤ 25 nmol/L] had a 3-fold higher risk of SCD compared with those with sufficient 25(OH)D levels >75 nmol/L [HR: 2.99, 95% confidence interval (CI): 1.39-6.40]. Furthermore, CVE and all-cause mortality were strongly increased (HR: 1.78, 95% CI: 1.18-2.69, and HR: 1.74, 95% CI: 1.22-2.47, respectively), all persisting in multivariate models. There were borderline non-significant associations with stroke and fatal infection while MI and deaths due to heart failure were not meaningfully affected. CONCLUSION: Severe vitamin D deficiency was strongly associated with SCD, CVE, and mortality, and there were borderline associations with stroke and fatal infection. Whether vitamin D supplementation decreases adverse outcomes requires further evaluation.

Mendelian randomization: use of genetics to enable causal inference in observational studies.

The aim of aetiologic studies in epidemiology is to investigate whether factors are causally related to diseases and therefore become a potential target for therapeutic interventions. Mendelian randomization enables estimation of causal relationships in observational studies using genetic variants as instrumental variables. An instrumental variable is a variable that can be considered to mimic the coin toss in a randomized study. Given the random assignment of alleles in gamete formation, the use of genetic variants is an alternative method to control for confounding. This educational article describes the approach of Mendelian randomization, its underlying rationale and its necessary assumptions.

CCR5 deletion protects against inflammation-associated mortality in dialysis patients.

The CC-chemokine receptor 5 (CCR5) is a receptor for various proinflammatory chemokines, and a deletion variant of the CCR5 gene (CCR5 Delta 32) leads to deficiency of the receptor. We hypothesized that CCR5 Delta 32 modulates inflammation-driven mortality in patients with ESRD. We studied the interaction between CCR5 genotype and levels of high-sensitivity C-reactive protein (hsCRP) in 603 incident dialysis patients from the multicenter, prospective NEtherlands COoperative Study on the Adequacy of Dialysis (NECOSAD) cohort. CCR5 genotype and hsCRP levels were both available for 413 patients. During 5 yr of follow-up, 170 patients died; 87 from cardiovascular causes. Compared with the reference group of patients who had the wild-type CCR5 genotype and hsCRP 10 mg/L (n = 108) had an increased risk for mortality (HR: 1.82; 95% CI: 1.29 to 2.58). However, those carrying the deletion allele with hsCRP > 10 mg/L (n = 25) had a mortality rate similar to the reference group; this seemingly protective effect of the CCR5 deletion was even more pronounced for cardiovascular mortality. We replicated these findings in an independent Swedish cohort of 302 ESRD patients. In conclusion, the CCR5 Delta 32 polymorphism attenuates the adverse effects of inflammation on overall and cardiovascular mortality in ESRD.

Individual and joint expert judgments as reference standards in artifact detection.

OBJECTIVE: To investigate the agreement among clinical experts in their judgments of monitoring data with respect to artifacts, and to examine the effect of reference standards that consist of individual and joint expert judgments on the performance of artifact filters. DESIGN: Individual judgments of four physicians, a majority vote judgment, and a consensus judgment were obtained for 30 time series of three monitoring variables: mean arterial blood pressure (ABPm), central venous pressure (CVP), and heart rate (HR). The individual and joint judgments were used to tune three existing automated filtering methods and to evaluate the performance of the resulting filters. MEASUREMENTS: The interrater agreement was calculated in terms of positive specific agreement (PSA). The performance of the artifact filters was quantified in terms of sensitivity and positive predictive value (PPV). RESULTS: PSA values between 0.33 and 0.85 were observed among clinical experts in their selection of artifacts, with relatively high values for CVP data. Artifact filters developed using judgments of individual experts were found to moderately generalize to new time series and other experts; sensitivity values ranged from 0.40 to 0.60 for ABPm and HR filters (PPV: 0.57-0.84), and from 0.63 to 0.80 for CVP filters (PPV: 0.71-0.86). A higher performance value for the filters was found for the three variable types when joint judgments were used for tuning the filtering methods. CONCLUSION: Given the disagreement among experts in their individual judgment of monitoring data with respect to artifacts, the use of joint reference standards obtained from multiple experts is recommended for development of automatic artifact filters.

Prognostic Bayesian networks II: an application in the domain of cardiac surgery.

A prognostic Bayesian network (PBN) is new type of prognostic model that implements a dynamic, process-oriented view on prognosis. In a companion article, the rationale of the PBN is described, and a dedicated learning procedure is presented. This article presents an application here of in the domain of cardiac surgery. A PBN is induced from clinical data of cardiac surgical patients using the proposed learning procedure; hospital mortality is used as outcome variable. The predictive performance of the PBN is evaluated on an independent test set, and results were compared to the performance of a network that was induced using a standard algorithm where candidate networks are selected using the minimal description length principle. The PBN is embedded in the prognostic system ProCarSur; a prototype of this system is presented. This application shows PBNs as a useful prognostic tool in medical processes. In addition, the article shows the added value of the PBN learning procedure.

Prognostic Bayesian networks I: rationale, learning procedure, and clinical use.

Prognostic models are tools to predict the future outcome of disease and disease treatment, one of the fundamental tasks in clinical medicine. This article presents the prognostic Bayesian network (PBN) as a new type of prognostic model that builds on the Bayesian network methodology, and implements a dynamic, process-oriented view on prognosis. A PBN describes the mutual relationships between variables that come into play during subsequent stages of a care process and a clinical outcome. A dedicated procedure for inducing these networks from clinical data is presented. In this procedure, the network is composed of a collection of local supervised learning models that are recursively learned from the data. The procedure optimizes performance of the network's primary task, outcome prediction, and handles the fact that patients may drop out of the process in earlier stages. Furthermore, the article describes how PBNs can be applied to solve a number of information problems that are related to medical prognosis.

Temporal abstraction for feature extraction: a comparative case study in prediction from intensive care monitoring data.

OBJECTIVES: To compare two temporal abstraction procedures for the extraction of meta features from monitoring data. Feature extraction prior to predictive modeling is a common strategy in prediction from temporal data. A fundamental dilemma in this strategy, however, is the extent to which the extraction should be guided by domain knowledge, and to which extent it should be guided by the available data. The two temporal abstraction procedures compared in this case study differ in this respect. METHODS AND MATERIAL: The first temporal abstraction procedure derives symbolic descriptions from the data that are predefined using existing concepts from the medical language. In the second procedure, a large space of numerical meta features is searched through to discover relevant features from the data. These procedures were applied to a prediction problem from intensive care monitoring data. The predictive value of the resulting meta features were compared, and based on each type of features, a class probability tree model was developed. RESULTS: The numerical meta features extracted by the second procedure were found to be more informative than the symbolic meta features of the first procedure in the case study, and a superior predictive performance was observed for the associated tree model. CONCLUSION: The findings indicate that for prediction from monitoring data, induction of numerical meta features from data is preferable to extraction of symbolic meta features using existing clinical concepts.