Disability transitions in Dutch community-dwelling older people aged 75 years or older

BACKGROUND

Recent world population predictions show that the world population aged >=65 years will increase from 10% in 2022 to 16% in 2050. Population aging is accompanied by an increase in people with disability. It is important to pay special attention to people with disability, as these people are at high risk of adverse outcomes. Our study aimed to investigate the transitions of disability among Dutch community-dwelling older people aged 75 years or older, using a follow-up of nine years. We used socio-demographic factors gender, age, marital status, education, and income, but also lifestyle, diseases, and life events to predict the disability transitions over time.

METHODS

We used a sample of 484 people that was randomly drawn from the municipality of Roosendaal (the Netherlands), a municipality with 78,000 inhabitants. A subset of people who completed part A of the Tilburg Frailty Indicator (TFI) at baseline and the Groningen Activity Restriction Scale (GARS) questionnaires was used with a nine-year follow-up. Paired Wilcoxon tests were used to compare the consecutive measurements. Socio-demographic factors gender, age, marital status, education, and income, but also lifestyle, diseases, and life events were included to predict the disability transitions over time. For the univariable and multivariable analysis of the measurements over time with the predictor variables, we used generalized estimation equations (GEE). A p-value <0.05 was considered significant. R version 3.4.4 was used for all analyses.

RESULTS

Of the participants, 65% were younger than 80 years, 50% were married or cohabiting, 87% reported a healthy lifestyle, and 63% had no diseases or chronic disorders. Each year, more participants changed from status not disabled to disabled than vice versa. The GEE analyses showed that lifestyle ('not healthy') and diseases or chronic disorders ('two or more') were significant in the multivariable analysis for the disability score and only diseases or chronic disorders ('two or more') was significant in the multivariable analysis for the dichotomous disability score.

CONCLUSIONS

The transition of the disability score is strongly influenced by lifestyle and diseases or disorders. This applies to a lesser extent to the dichotomous disability score. There, only diseases or disorders are an important predictor. For health care professionals our study provides starting points for interventions focused on the prevention of worsening disability and for community-dwelling older people >= 75, the most important recommendation is: live healthy!

External Validation of Models for Predicting Disability in Community-Dwelling Older People in the Netherlands: A Comparative Study

Background

Advanced statistical modeling techniques may help predict health outcomes. However, it is not the case that these modeling techniques always outperform traditional techniques such as regression techniques. In this study, external validation was carried out for five modeling strategies for the prediction of the disability of community-dwelling older people in the Netherlands.

Methods

We analyzed data from five studies consisting of community-dwelling older people in the Netherlands. For the prediction of the total disability score as measured with the Groningen Activity Restriction Scale (GARS), we used fourteen predictors as measured with the Tilburg Frailty Indicator (TFI). Both the TFI and the GARS are self-report questionnaires. For the modeling, five statistical modeling techniques were evaluated: general linear model (GLM), support vector machine (SVM), neural net (NN), recursive partitioning (RP), and random forest (RF). Each model was developed on one of the five data sets and then applied to each of the four remaining data sets. We assessed the performance of the models with calibration characteristics, the correlation coefficient, and the root of the mean squared error.

Results

The models GLM, SVM, RP, and RF showed satisfactory performance characteristics when validated on the validation data sets. All models showed poor performance characteristics for the deviating data set both for development and validation due to the deviating baseline characteristics compared to those of the other data sets.

Conclusion

The performance of four models (GLM, SVM, RP, RF) on the development data sets was satisfactory. This was also the case for the validation data sets, except when these models were developed on the deviating data set. The NN models showed a much worse performance on the validation data sets than on the development data sets.

Predicting Disability Using a Nomogram of the Tilburg Frailty Indicator (TFI)

Disability is associated with lower quality of life and premature death in older people. Therefore, prevention and intervention targeting older people living with a disability is important. Frailty can be considered a major predictor of disability. In this study, we aimed to develop nomograms with items of the Tilburg Frailty Indicator (TFI) as predictors by using cross-sectional and longitudinal data (follow-up of five and nine years), focusing on the prediction of total disability, disability in activities of daily living (ADL), and disability in instrumental activities of daily living (IADL). At baseline, 479 Dutch community-dwelling people aged ≥75 years participated. They completed a questionnaire that included the TFI and the Groningen Activity Restriction Scale to assess the three disability variables. We showed that the TFI items scored different points, especially over time. Therefore, not every item was equally important in predicting disability. 'Difficulty in walking' and 'unexplained weight loss' appeared to be important predictors of disability. Healthcare professionals need to focus on these two items to prevent disability. We also conclude that the points given to frailty items differed between total, ADL, and IADL disability and also differed regarding years of follow-up. Creating one monogram that does justice to this seems impossible.

Bayesian Techniques in Predicting Frailty among Community-Dwelling Older Adults in the Netherlands

Background Frailty is a syndrome that is defined as an accumulation of deficits in physical, psychological, and social domains. On a global scale, there is an urgent need to create frailty-ready healthcare systems due to the healthcare burden that frailty confers on systems and the increased risk of falls, healthcare utilization, disability, and premature mortality. Several studies have been conducted to develop prediction models for predicting frailty. Most studies used logistic regression as a technique to develop a prediction model. One area that has experienced significant growth is the application of Bayesian techniques, partly due to an increasing number of practitioners valuing the Bayesian paradigm as matching that of scientific discovery. Objective We compared ten different Bayesian networks as proposed by ten experts in the field of frail elderly people to predict frailty with a choice from ten dichotomized determinants for frailty. Methods We used the opinion of ten experts who could indicate, using an empty Bayesian network graph, the important predictors for frailty and the interactions between the different predictors. The candidate predictors were age, sex, marital status, ethnicity, education, income, lifestyle, multimorbidity, life events, and home living environment. The ten Bayesian network models were evaluated in terms of their ability to predict frailty. For the evaluation, we used the data of 479 participants that filled in the Tilburg Frailty indicator (TFI) questionnaire for assessing frailty among community-dwelling older people. The data set contained the aforementioned variables and the outcome "frail". The model fit of each model was measured using the Akaike information criterion (AIC) and the predictive performance of the models was measured using the area under the curve (AUC) of the receiver operator characteristic (ROC). The AUCs of the models were validated using bootstrapping with 100 repetitions. The relative importance of the predictors in the models was calculated using the permutation feature importance algorithm (PFI). Results The ten Bayesian networks of the ten experts differed considerably regarding the predictors and the connections between the predictors and the outcome. However, all ten networks had corrected AUCs >0.700. Evaluating the importance of the predictors in each model, "diseases or chronic disorders" was the most important predictor in all models (10 times). The predictors "lifestyle" and "monthly income" were also often present in the models (both 6 times). One or more diseases or chronic disorders, an unhealthy lifestyle, and a monthly income below 1800 euro increased the likelihood of frailty. Conclusions Although the ten experts all made different graphs, the predictive performance was always satisfying (AUCs >0.700). While it is true that the predictor importance varied all the time, the top three of the predictor importance consisted of "diseases or chronic disorders", "lifestyle" and "monthly income". All in all, asking for the opinion of experts in the field of frail elderly to predict frailty with Bayesian networks may be more rewarding than a data-driven forecast with Bayesian networks because they have expert knowledge regarding interactions between the different predictors.

Prediction of COVID-19 Infections for Municipalities in the Netherlands: Algorithm Development and Interpretation

Background

COVID-19 was first identified in December 2019 in the city of Wuhan, China. The virus quickly spread and was declared a pandemic on March 11, 2020. After infection, symptoms such as fever, a (dry) cough, nasal congestion, and fatigue can develop. In some cases, the virus causes severe complications such as pneumonia and dyspnea and could result in death. The virus also spread rapidly in the Netherlands, a small and densely populated country with an aging population. Health care in the Netherlands is of a high standard, but there were nevertheless problems with hospital capacity, such as the number of available beds and staff. There were also regions and municipalities that were hit harder than others. In the Netherlands, there are important data sources available for daily COVID-19 numbers and information about municipalities.

Objective

We aimed to predict the cumulative number of confirmed COVID-19 infections per 10,000 inhabitants per municipality in the Netherlands, using a data set with the properties of 355 municipalities in the Netherlands and advanced modeling techniques.

Methods

We collected relevant static data per municipality from data sources that were available in the Dutch public domain and merged these data with the dynamic daily number of infections from January 1, 2020, to May 9, 2021, resulting in a data set with 355 municipalities in the Netherlands and variables grouped into 20 topics. The modeling techniques random forest and multiple fractional polynomials were used to construct a prediction model for predicting the cumulative number of confirmed COVID-19 infections per 10,000 inhabitants per municipality in the Netherlands.

Results

The final prediction model had an R² of 0.63. Important properties for predicting the cumulative number of confirmed COVID-19 infections per 10,000 inhabitants in a municipality in the Netherlands were exposure to particulate matter with diameters <10 μm (PM10) in the air, the percentage of Labour party voters, and the number of children in a household.

Conclusions

Data about municipality properties in relation to the cumulative number of confirmed infections in a municipality in the Netherlands can give insight into the most important properties of a municipality for predicting the cumulative number of confirmed COVID-19 infections per 10,000 inhabitants in a municipality. This insight can provide policy makers with tools to cope with COVID-19 and may also be of value in the event of a future pandemic, so that municipalities are better prepared.

A Comparison of Different Modelling Techniques in Predicting Mortality with the Tilburg Frailty Indicator (Preprint)

Background

Objective

Methods

Results

Conclusions

The Development of Multidimensional Frailty Over Seven Years A longitudinal study among Dutch community-dwelling older people using the Tilburg Frailty Indicator

PURPOSE

To examine the development of multidimensional frailty, including physical, psychological and social components, over a period of seven years. To determine the effects of sociodemographic factors (gender, age, marital status, education, income) on the development of frailty.

METHODS

This longitudinal study was conducted in sample of 479 community-dwelling people aged ≥ 75 years living in the municipality of Roosendaal, the Netherlands. The Tilburg Frailty Indicator (TFI), a self-report questionnaire, was used to collect data about frailty. Frailty was assessed annually.

RESULTS

Frailty increased significantly over seven years among the people who completed the entire TFI all years (n = 121), the average score was 3.75 (SD 2.80) at baseline and 5.05 (SD 3.18) after seven years. Regarding frailty transitions, most participants remained unchanged from their baseline status. The transition from non-frail to frail was present in 8.3% to 12.6% of the participants and 5.1% to 10.7% made a transition from frail to non-frail. Gender (woman), age (≥80 years), marital status (not married/cohabiting), high level of education, and incomes from €601-€1800 were significantly associated with a higher frailty score.

CONCLUSION

This study showed that multidimensional frailty, assessed with the TFI, increased among Dutch community-dwelling people aged ≥75 years using a follow-up of seven years. Gender, age, marital status, education, and income were associated with frailty transitions. These findings provide healthcare professionals clues to identify people at increased risk of frailty, and target interventions which aim to prevent or delay frailty and its adverse outcomes, such as disability and mortality.

The Prediction of Mortality by Disability Among Dutch Community-Dwelling Older People

Objective

To predict mortality by disability in a sample of 479 Dutch community-dwelling people aged 75 years or older.

Methods

A longitudinal study was carried out using a follow-up of seven years. The Groningen Activity Restriction Scale (GARS), a self-reported questionnaire with good psychometric properties, was used for data collection about total disability, disability in activities in daily living (ADL) and disability in instrumental activities in daily living (IADL). The mortality dates were provided by the municipality of Roosendaal (a city in the Netherlands). For analyses of survival, we used Kaplan–Meier analyses and Cox regression analyses to calculate hazard ratios (HR) with 95% confidence intervals (CI).

Results

All three disability variables (total, ADL and IADL) predicted mortality, unadjusted and adjusted for age and gender. The unadjusted HRs for total, ADL and IADL disability were 1.054 (95%-CI: [1.039;1.069]), 1.091 (95%-CI: [1.062;1.121]) and 1.106 (95%-CI: [1.077;1.135]) with p-values <0.001, respectively. The AUCs were <0.7, ranging from 0.630 (ADL) to 0.668 (IADL). Multivariate analyses including all 18 disability items revealed that only “Do the shopping” predicted mortality. In addition, multivariate analyses focusing on 11 ADL items and 7 IADL items separately showed that only the ADL item “Get around in the house” and the IADL item “Do the shopping” significantly predicted mortality.

Conclusion

Disability predicted mortality in a seven years follow-up among Dutch community-dwelling older people. It is important that healthcare professionals are aware of disability at early stages, so they can intervene swiftly, efficiently and effectively, to maintain or enhance the quality of life of older people.

Gender differences in retention rate of tumor necrosis factor alpha inhibitor treatment in ankylosing spondylitis: a retrospective cohort study in daily practice

AIM

To assess gender differences in ankylosing spondylitis (AS) patients in relation to tumor necrosis factor alpha inhibitor (TNFi) drug survival and occurrence of adverse events in daily practice in a large peripheral hospital.

METHOD

Retrospective data were collected from AS patients treated with etanercept, infliximab and adalimumab between January 2004 and January 2014. Kaplan-Meier survival curves were conducted to describe the drug survival and occurrence of adverse events in time.

RESULTS

Overall, 122 AS patients (60.7% male) were included over a 10-year time period, with a mean treatment period of 51 months (1-127 months). In total, 21 (17.2%) patients stopped the TNFi, mainly due to inefficacy (52.4%). Female patients showed a significant shorter treatment period compared to males (33.4 vs. 44.9 months). In addition, female patients switched more between TNFi compared to males (26.9% vs. 16.3%) and had a significantly higher risk at developing infections compared to male patients (26% vs.19%).

CONCLUSION

Females stayed on the same TNFi for a significantly shorter period compared to males (33.4 vs. 44.9 months) and the most important reason to stop or switch the drug was inefficacy. Moreover, females seemed to be more prone to infections during TNFi treatment than males.

Exercise Electrocardiogram Neither Predicts Nor Excludes Coronary Artery Disease in Women with Low to Intermediate Risk

AIM

The value of exercise electrocardiogram (ExECG) in symptomatic female patients with low to intermediate risk for significant coronary artery disease (CAD) has been under debate for many years, and nondiagnostic or even erroneous test results are frequently encountered. Cardiac-CT may be more appropriate to exclude CAD in women. This study compares the results of ExECGs with those of cardiac-CTs, performed within a time frame of 1 month in an all-comers female chest pain population.

PATIENTS AND METHODS

Five hundred fifty-one consecutive female patients from a patient registry were included. ExECGs were negative in 324 (59%), positive in 14 (3%), and nondiagnostic in 213 (39%) patients. CAD was revealed by cardiac-CT in 57% of the women with negative ExECG. No signs of CAD were present on cardiac-CT in 64% of the women with a positive ExECG. Cardiac-CT showed presence of CAD in 268/551 (49%) patients, of whom 56/268 (21%) was diagnosed with ≥50% stenosis. The ExECG of the latter group was negative in 26 (46%), inconclusive in 29 (52%), and positive in 1 (2%). Considering ≥50% stenosis at cardiac-CT as the reference, sensitivity, specificity, PPV, and NPV of ExECG for the present population were 3.7%, 95.7%, 7.1%, and 91.7%, respectively. Similar diagnostic performance was calculated when considering ≥70% stenosis at cardiac-CT as the reference.

CONCLUSION

ExECG failed to detect CAD in more than half of this cohort and in almost half of women with >50% stenosis at cardiac-CT. Importantly, no CAD was detected by cardiac-CT in 64% of women with a positive ExECG. ExECG is therefore questionable as a diagnostic strategy in women with low-to-intermediate risk of CAD, although prospective studies are warranted to determine whether replacing ExECG by cardiac-CT provides better prognoses.

In patients undergoing fast track total knee arthroplasty, addition of buprenorphine to a femoral nerve block has no clinical advantage: A prospective, double-blinded, randomized, placebo controlled trial

BACKGROUND

Several adjuvants have been proposed to prolong the effect of peripheral nerve blocks, one of which is buprenorphine. In this randomized double blinded placebo controlled trial we studied whether the addition of buprenorphine to a femoral nerve block prolongs analgesia in patients undergoing total knee arthroplasty in a fast track surgery protocol.

METHODS

The treatment group (B) was given an ultrasound-guided femoral nerve block with ropivacaine 0.2% and 0.3mg buprenorphine. We choose to use 2 control groups. Group R was given a femoral nerve block with ropivacaine 0.2% only. Group S also received 0.3 mg buprenorphine subcutaneously. Only patients with a successful block were enrolled in the study.

RESULTS

We found no difference in our primary outcome parameter of time to first rescue analgesic. We found lower opioid use and better sleep quality the first postoperative night in patients receiving buprenorphine perineurally or subcutaneously. Buprenorphine did not lead to any significant change in pain or mobilization. We found a high overall incidence of nausea and vomiting.

CONCLUSION

In patients undergoing total knee arthroplasty, in the setting of a fast track surgery protocol, the addition of buprenorphine to a femoral nerve block did not prolong analgesia.

Dementia and delirium, the outcomes in elderly hip fracture patients

BACKGROUND

Delirium in hip fractured patients is a frequent complication. Dementia is an important risk factor for delirium and is common in frail elderly. This study aimed to extend the previous knowledge on risk factors for delirium and the consequences. Special attention was given to patients with dementia and delirium.

METHODS

This is a retrospective cohort study performed in the Amphia Hospital, Breda, the Netherlands. A full electronic patient file system (Hyperspace Version IU4: Epic, Inc., Verona, WI, USA) was used to assess data between January 2014 and September 2015. All patients presented were aged ≥70 years with a hip fracture, who underwent surgery with osteosynthesis or arthroplasty. Patients were excluded in case of a pathological or a periprosthetic hip fracture, multiple traumatic injuries, and high-energy trauma. Patient and surgical characteristics were documented. Postoperative outcomes were noted. Delirium was screened using Delirium Observation Screening Scale and dementia was assessed from medical notes.

RESULTS

Of a total of 566 included patients, 75% were females. The median age was 84 years (interquartile range: 9). Delirium was observed in 35%. Significant risk factors for delirium were a high American Society of Anesthesiology score, delirium in medical history, functional dependency, preoperative institutionalization, low hemoglobin level, and high amount of blood transfusion. Delirium was correlated with a longer hospital stay (P=0.001), increased association with complications (P<0.001), institutionalization (P<0.001), and 6-month mortality (P<0.001). Patients with dementia (N=168) had a higher delirium rate (57.7%, P<0.001) but a shorter hospital stay (P<0.001). There was no significant difference in the 6-month mortality between delirious patients with (34.0%) and without dementia (26.3%).

CONCLUSION

Elderly patients with a hip fracture are vulnerable for delirium, especially when the patient has dementia. Patients who underwent an episode of delirium were at increased risk for adverse outcomes.

Pitfalls in interpretation of CT-values of RT-PCR in children with acute respiratory tract infections

Background

The relation between viral load and disease severity in childhood acute respiratory tract infections (ARI) is not fully understood.

Objectives

To assess the clinical relevance of the relation between viral load, determined by cycle threshold (CT) value of real-time reverse transcription-polymerase chain reaction assays and disease severity in children with single- and multiple viral ARI.

Study design

582 children with ARI were prospectively followed and tested for 15 viruses. Correlations were calculated between CT values and clinical parameters.

Results

In single viral ARI, statistically significant correlations were found between viral loads of Respiratory Syncytial Virus (RSV) and hospitalization and between viral loads of Human Coronavirus (HCoV) and a disease severity score. In multiple-viral ARI, statistically significant correlations between viral load and clinical parameters were found. In RSV-Rhinovirus (RV) multiple infections, a low viral load of RV was correlated with a high length of hospital stay and a high duration of extra oxygen use. The mean CT value for RV, HCoV and Parainfluenza virus was significantly lower in single- versus multiple infections.

Conclusion

Although correlations between CT values and clinical parameters in patients with single and multiple viral infection were found, the clinical importance of these findings is limited because individual differences in host-, viral and laboratory factors complicate the interpretation of statistically significant findings. In multiple infections, viral load cannot be used to differentiate between disease causing virus and innocent bystanders.

Single- and multiple viral respiratory infections in children: disease and management cannot be related to a specific pathogen

Background

The number of viral pathogens associated with pediatric acute respiratory tract infection (ARI) has grown since the introduction of reverse transcription real-time polymerase chain reaction (RT-PCR) assays. Multiple viruses are detected during a single ARI episode in approximately a quarter of all cases. The clinical relevance of these multiple detections is unclear, as is the role of the individual virus. We therefore investigated the correlation between clinical data and RT-PCR results in children with single- and multiple viral ARI.

Methods

Data from children with ARI were prospectively collected during two winter seasons. RT-PCR testing for 15 viruses was performed in 560 ARI episodes. In the patients with a single-viral etiology, clinical data, laboratory findings, patient management- and outcome data were compared between the different viruses. With this information, we compared data from children of whom RT-PCR data were negative, with children with single- and multiple viral positive results.

Results

The viral detection rate was 457/560 (81.6%) of which 331/560 (59.1%) were single infections and 126/560 (22.5%) were multiple infections. In single viral infections, some statistically significant differences in demographics, clinical findings, disease severity and outcome were found between children with different viral etiologies. However, no clinically recognizable pattern was established to be virus-specific. In a multivariate analysis, the only variables that were correlated with longer hospital stay were the use of oxygen and nebulizer therapy, irrespective of the viral pathogen. Children with RT-PCR positive test results had a significant higher disease severity, fever, length of hospital stay, days of extra oxygen supply, and days of antibiotic treatment than children with a negative RT-PCR test result. For children with single- versus children with multiple positive RT-PCR test results, these differences were not significant.

Conclusions

Disease (severity), management and outcome in pediatric ARI are not associated with a specific virus. Single- and multiple viral ARI do not significantly differ with regard to clinical outcome and patient management. For general pediatrics, RT-PCR assays should be restricted to pathogens for which therapy is available or otherwise may have clinical consequences. Further research with an extended panel of RT-PCR assays and a larger number of inclusions is necessary to further validate our findings.

Electronic supplementary material

The online version of this article (doi:10.1186/s12879-016-2118-6) contains supplementary material, which is available to authorized users.

A novel troponin I rule-out value below the upper reference limit for acute myocardial infarction

OBJECTIVE

To determine cut-off values for a recently introduced high sensitive cardiac troponin assay (hs-cTnI) which provide similar sensitivity, specificity, negative predictive value (NPV) and positive predictive value (PPV) for acute myocardial infarction (AMI) as known cut-off values for an hs-cTnT assay.

METHODS

A prospective observational study was performed. Hs-cTnT (Roche) and hs-cTnI (Abbott) were measured in consecutive patients with symptoms suggestive of AMI. Representative measurements (obtained at least 3 h after chest pain has started) and serial measurements with a time delay between 2.5 h and 4.5 h were used to determine cut-off levels. Two independent clinicians adjudicated the final diagnosis.

RESULTS

1490 patients were included in the study of whom 114 (8%) received a final diagnosis of AMI. Receiver operating characteristics analysis showed no statistically significant differences in the areas under the curve between the two assays. Cut-off values for representative hs-TnI were found to be as follows: rule-out: 10 ng/L (sensitivity: 98.2%; 95% CI 95.7% to 100.0% and NPV: 99.8%; 99.5% to 100.0%); rule-in: 70 ng/L (specificity: 90.8%; 89.3% to 92.4% and PPV: 39.7%; 36.1% to 43.3%). For serial measurements we found a Δ rule-out cut-off value of 20 ng/L (sensitivity: 94.9%; 88.0% to 100.0% and NPV: 98.7%; 96.9% to 100.0%) and Δ rule-in cut-off values of 100 ng/L (specificity: 92.7%; 87.9% to 95.8% and PPV: 57.6%; 39.4% to 74.0%) and 300% (specificity: 93.8%; 90.4% to 97.2% and PPV: 61.3%; 51.1% to 71.5%).

CONCLUSIONS

Cut-off values for hs-cTnI measurements are determined which allow a similar diagnostic classification as compared with hs-cTnT. Importantly, for a rule-out paradigm this cut-off value is unmistakably lower than the upper reference limit.

Feature selection and validated predictive performance in the domain of Legionella pneumophila: a comparative study

Background

Genetic comparisons of clinical and environmental Legionella strains form an essential part of outbreak investigations. DNA microarrays often comprise many DNA markers (features). Feature selection and the development of prediction models are particularly challenging in this domain with many variables and comparatively few subjects or data points. We aimed to compare modeling strategies to develop prediction models for classifying infections as clinical or environmental.

Methods

We applied a bootstrap strategy for preselecting important features to a database containing 222 Legionella pneumophila strains with 448 continuous markers and a dichotomous outcome (clinical or environmental). Feature selection was done with 50 bootstrap samples resulting in a top 10 of most important features for each of four modeling techniques: classification and regression trees (CART), random forests (RF), support vector machines (SVM) and least absolute shrinkage and selection operator (LASSO). Validation was done in a second bootstrap re-sampling loop (200×) for evaluation of discriminatory model performance according to the AUC.

Results

The top 5 of selected features differed considerably between the various modeling techniques, with only one common feature (“LePn.007B8”). The mean validated AUC-values of the SVM model and the CART model were 0.859 and 0.873 respectively. The LASSO and the RF model showed higher validated AUC-values (0.925 and 0.975 respectively).

Conclusions

In the domain of Legionella pneumophila, which comprises many potential features for classifying of infections as clinical or environmental, the RF and LASSO techniques provide good prediction models. The identification of potentially biologically relevant features is highly dependent on the technique used, and should hence be interpreted with caution.

Electronic supplementary material

The online version of this article (doi:10.1186/s13104-016-1945-2) contains supplementary material, which is available to authorized users.

Diagnosing perforated appendicitis in pediatric patients: a new model

OBJECTIVE

Studies have investigated sensitivity and specificity of symptoms and tests for diagnosing appendicitis in children. Less is known with regard to the predictive value of these symptoms and tests with respect to the severity of appendicitis. The aim of this study was to determine the predictive value of patient's characteristics and tests for discriminating between perforated and nonperforated appendicitis in children.

PATIENTS AND METHODS

Pediatric patients who underwent an appendectomy at Spaarne Hospital Hoofddorp, the Netherlands, between January 1, 2009 and December 31, 2013, were included. Baseline patient's characteristics, history, physical examination, laboratory data and results of ultrasounds were collected. Univariate and multivariate logistic regressions were used to determine predictors of perforation.

RESULTS

In total, 375 patients were included in this study of which 97 children (25.9%) had significant signs of perforation. Univariate analysis showed that age, duration of complaints, temperature, vomiting, CRP, WBC, different findings on ultrasound and the diameter of the appendix were good predictors of a perforated appendicitis. The final multivariate prediction model included temperature, CRP, clearly visible appendix and free fluids on ultrasound and diameter of the appendix and resulted in an area under the curve (AUC) of 0.91 showing sensitivity and specificity of respectively 85.2% and 81.2%.

CONCLUSION

This prediction model can be used for identification of 'high-risk' children for a perforated appendicitis and might be helpful to prevent complications and longer hospitalization by bringing these children to theater earlier.

Assessing Discriminative Performance at External Validation of Clinical Prediction Models

Introduction

External validation studies are essential to study the generalizability of prediction models. Recently a permutation test, focusing on discrimination as quantified by the c-statistic, was proposed to judge whether a prediction model is transportable to a new setting. We aimed to evaluate this test and compare it to previously proposed procedures to judge any changes in c-statistic from development to external validation setting.

Methods

We compared the use of the permutation test to the use of benchmark values of the c-statistic following from a previously proposed framework to judge transportability of a prediction model. In a simulation study we developed a prediction model with logistic regression on a development set and validated them in the validation set. We concentrated on two scenarios: 1) the case-mix was more heterogeneous and predictor effects were weaker in the validation set compared to the development set, and 2) the case-mix was less heterogeneous in the validation set and predictor effects were identical in the validation and development set. Furthermore we illustrated the methods in a case study using 15 datasets of patients suffering from traumatic brain injury.

Results

The permutation test indicated that the validation and development set were homogenous in scenario 1 (in almost all simulated samples) and heterogeneous in scenario 2 (in 17%-39% of simulated samples). Previously proposed benchmark values of the c-statistic and the standard deviation of the linear predictors correctly pointed at the more heterogeneous case-mix in scenario 1 and the less heterogeneous case-mix in scenario 2.

Conclusion

The recently proposed permutation test may provide misleading results when externally validating prediction models in the presence of case-mix differences between the development and validation population. To correctly interpret the c-statistic found at external validation it is crucial to disentangle case-mix differences from incorrect regression coefficients.

Modern modelling techniques are data hungry: a simulation study for predicting dichotomous endpoints

BACKGROUND

Modern modelling techniques may potentially provide more accurate predictions of binary outcomes than classical techniques. We aimed to study the predictive performance of different modelling techniques in relation to the effective sample size ("data hungriness").

METHODS

We performed simulation studies based on three clinical cohorts: 1282 patients with head and neck cancer (with 46.9% 5 year survival), 1731 patients with traumatic brain injury (22.3% 6 month mortality) and 3181 patients with minor head injury (7.6% with CT scan abnormalities). We compared three relatively modern modelling techniques: support vector machines (SVM), neural nets (NN), and random forests (RF) and two classical techniques: logistic regression (LR) and classification and regression trees (CART). We created three large artificial databases with 20 fold, 10 fold and 6 fold replication of subjects, where we generated dichotomous outcomes according to different underlying models. We applied each modelling technique to increasingly larger development parts (100 repetitions). The area under the ROC-curve (AUC) indicated the performance of each model in the development part and in an independent validation part. Data hungriness was defined by plateauing of AUC and small optimism (difference between the mean apparent AUC and the mean validated AUC <0.01).

RESULTS

We found that a stable AUC was reached by LR at approximately 20 to 50 events per variable, followed by CART, SVM, NN and RF models. Optimism decreased with increasing sample sizes and the same ranking of techniques. The RF, SVM and NN models showed instability and a high optimism even with >200 events per variable.

CONCLUSIONS

Modern modelling techniques such as SVM, NN and RF may need over 10 times as many events per variable to achieve a stable AUC and a small optimism than classical modelling techniques such as LR. This implies that such modern techniques should only be used in medical prediction problems if very large data sets are available.

Cognitive coping style (monitoring and blunting) and the need for information, information satisfaction and shared decision making among patients with haematological malignancies

OBJECTIVE

A haematological malignancy is a serious, life-altering disease and may be characterised as an uncontrollable and unpredictable stress situation. In dealing with potentially threatening information, individuals generally utilise two main cognitive coping styles: monitoring (the tendency to seek threat-relevant information) and blunting (avoiding threatening information and seeking distraction). The aim of this study was to obtain insight into the association between cognitive coping style and (a) need for information, (b) satisfaction with information, (c) involvement in decision making, and (d) quality of life (QoL).

METHODS

In this cross-sectional study, coping style was assessed among adult patients diagnosed with a haematological malignancy, using an adapted version of the Threatening Medical Situations Inventory. Information need, information satisfaction, decision-making preference and QoL were measured with validated questionnaires.

RESULTS

In total, 458 patients returned the questionnaire (66%). A monitoring coping style was positively related to need for both general and specific information. Blunting was positively and QoL was negatively related to need for information. Monitoring was positively related to involvement in decision-making and negatively to information satisfaction. Using multivariate analysis, this relation between monitoring and information satisfaction disappeared, and for blunting, we found a negatively significant relation. QoL was not related to coping style.

CONCLUSIONS

Among patients with haematological malignancies, coping style is related to a need for information, information satisfaction, and involvement in treatment decision-making. Therefore, it is important for health care professionals to be aware of individual differences in cognitive coping style.

Perceived need for information among patients with a haematological malignancy: associations with information satisfaction and treatment decision-making preferences

For patients with haematological malignancies, information on disease, prognosis, treatment and impact on quality of life is of the utmost importance. To gain insight into the perceived need for information in relation to sociodemographic and clinical parameters, comorbidity, quality of life (QoL) and information satisfaction, we compiled a questionnaire based on existing validated questionnaires. A total of 458 patients diagnosed with a haematological malignancy participated. The perceived need for information was moderate to high (40-70%). Multivariate regression analyses showed that a higher need for information was related to younger age, worse QoL, being member of a patient society and moderate comorbidity. The need for disease and treatment-related information was higher than the need for psychosocial information. A higher need for disease and treatment-related information was associated to being diagnosed with multiple myeloma. A higher need for psychosocial information was related to a lower educational level. The information provision could be improved according to 41% of the patients. Higher satisfaction with provided information was associated with better QoL. Most patients (62%) reported that they wanted to be fully informed about their illness and actively involved in treatment decision-making. The results contribute to improving patient-tailored information provision and shared decision-making in clinical practice.

Risk prediction with machine learning and regression methods

This is a discussion of issues in risk prediction based on the following papers: "Probability estimation with machine learning methods for dichotomous and multicategory outcome: Theory" by Jochen Kruppa, Yufeng Liu, Gérard Biau, Michael Kohler, Inke R. König, James D. Malley, and Andreas Ziegler; and "Probability estimation with machine learning methods for dichotomous and multicategory outcome: Applications" by Jochen Kruppa, Yufeng Liu, Hans-Christian Diener, Theresa Holste, Christian Weimar, Inke R. König, and Andreas Ziegler.

Evaluation of mitotic activity index in breast cancer using whole slide digital images

Introduction

Mitotic Activity Index (MAI) is an important independent prognostic factor and an integral part of the breast cancer grading system. Thus, correct estimation of this prognostically relevant feature is essential for guiding treatment decision and assessing patient prognosis.

The aim of this study was to validate the use of high resolution Whole Slide Images (WSI) in estimating MAI in breast cancer specimens.

Methods

MAI was evaluated in 100 consecutive breast cancer specimens by three observers on two occasions, microscopically and on WSI with a wash out period of 4 months. MAI was also translated to mitotic scores as in grading. Inter- and intra-observer agreement between microscopic and digital MAI counts and scores was measured.

Results

Almost perfect inter-observer agreements were obtained from counting MAI using a conventional microscope (intra-class correlation coefficient (ICCC) 0.879) as well as on WSI (ICCC 0.924). K coefficients reflected good inter-observer agreements among observers' microscopic mitotic scores (average kappa 0.642). Comparable results were also observed among digital mitotic scores (average kappa 0.635). There was strong to perfect intra-observer agreements between MAI counts and mitotic scores for the two diagnostic modalities (ICCC 0.716–0.863, kappa 0.506–0.617). There were no significant differences in mitotic scores using both diagnostic modalities.

Conclusion

Scoring mitoses using WSI in breast cancer seems to be just as reliable and reproducible as when using a microscope. Further development of software and image quality will definitely encourage the use of WSI in routine pathology practice.

Prediction of intracranial findings on CT-scans by alternative modelling techniques

Background

Prediction rules for intracranial traumatic findings in patients with minor head injury are designed to reduce the use of computed tomography (CT) without missing patients at risk for complications. This study investigates whether alternative modelling techniques might improve the applicability and simplicity of such prediction rules.

Methods

We included 3181 patients with minor head injury who had received CT scans between February 2002 and August 2004. Of these patients 243 (7.6%) had intracranial traumatic findings and 17 (0.5%) underwent neurosurgical intervention. We analyzed sensitivity, specificity and area under the ROC curve (AUC-value) to compare the performance of various modelling techniques by 10 × 10 cross-validation. The techniques included logistic regression, Bayes network, Chi-squared Automatic Interaction Detection (CHAID), neural net, support vector machines, Classification And Regression Trees (CART) and "decision list" models.

Results

The cross-validated performance was best for the logistic regression model (AUC 0.78), followed by the Bayes network model and the neural net model (both AUC 0.74). The other models performed poorly (AUC < 0.70). The advantage of the Bayes network model was that it provided a graphical representation of the relationships between the predictors and the outcome.

Conclusions

No alternative modelling technique outperformed the logistic regression model. However, the Bayes network model had a presentation format which provided more detailed insights into the structure of the prediction problem. The search for methods with good predictive performance and an attractive presentation format should continue.

Mortality associated with delirium after hip-surgery: a 2-year follow-up study

BACKGROUND

delirium after hip-surgery is associated with poor outcome. Few studies examined the mortality risk associated with delirium in elderly hip-surgery patients after 1 year or more. Aim of this study was to examine the hazard risk associated with delirium in elderly hip-surgery patients at 2-year follow-up, controlling for baseline risk factors and interaction effects.

METHODS

this is a secondary analysis based on data from a controlled clinical trial evaluating efficacy of haloperidol prophylaxis for delirium conducted in a large medical school-affiliated general hospital in Alkmaar, The Netherlands. Randomised and non-randomised patients (n = 603) were followed-up for 2 years. Predefined risk factors and other potential risk factors for delirium were assessed prior to surgery. Primary outcome was time of death during the follow-up period. Cox proportional hazards were estimated and compared across patients who had postoperative delirium during hospitalisation and those who did not.

RESULTS

a total of 90/603 patients (14.9%) died during the study period and 74/603 (12.3%) had postoperative delirium. Incidence of delirium was higher in patients who died (32.2%) compared with those who survived (8.8%). The interaction effect of delirium by illness severity on mortality was significant after adjusting for predefined delirium risk factors and other potential covariates including study intervention (adjusted Hazard risk = 1.05, 95% CI 1.02-1.08). A total of 14/27 delirium patients who were severely ill on admission died during follow-up versus 15/47 delirium patients who were not (RR 1.63 CI 0.93-2.83).

CONCLUSIONS

delirium does not independently predict mortality at 2-year follow-up in elderly hip-surgery patients. However, outcome from delirium is particularly poor when other risk factors are present.

Alcohol intoxication among Dutch adolescents: acute medical complications in the years 2000-2010

OBJECTIVE

To investigate acute medical complications of alcohol intoxication among adolescents presenting to Reinier de Graaf Hospital, a major nonacademic hospital in Delft, the Netherlands.

METHODS

The frequency of reduced consciousness, hypothermia, electrolyte disturbances, hypoglycemia, and acidosis along with alcohol intoxication was retrospectively studied in the medical records of patients aged 11 to 17 years admitted between January 1, 2000, and April 30, 2010.

RESULTS

Reduced consciousness (45%) and hypothermia (43.1%) were the most common clinical findings. More boys than girls experienced hypothermia (P = .028), probably because of the higher blood alcohol concentration level ( P = .026), which was associated with lower temperature (Pearson's r = -.37; P < .001). The most frequent electrolyte disturbance was hyperchloremia (31.1%), followed by low bicarbonate (22%), hypokalemia (11.9%), and hypernatremia (7.7%). Strikingly, none of the patients in this study suffered hypoglycemia. On the contrary, 13.6% of the patients were slightly hyperglycemic. Mild acidosis, both metabolic and respiratory, was observed among 28.8% of the patients, although the metabolic component predominated.

CONCLUSIONS

Acute medical complications because of alcohol intoxication appeared to be mild, and most remarkably none of the patients experienced hypoglycemia.

Long-term cognitive outcome of delirium in elderly hip surgery patients. A prospective matched controlled study over two and a half years

OBJECTIVE

To study the outcome of delirium in elderly hip surgery patients.

DESIGN

Prospective matched controlled cohort study. Hip surgery patients (n = 112) aged 70 years and older, who participated in a controlled clinical trial of haloperidol prophylaxis for delirium, were followed for an average of 30 months after discharge. Patients with a diagnosis of dementia or mild cognitive impairment (MCI) were identified using psychiatric interviews. Proportions of patients with dementia/MCI were compared across patients who had postoperative delirium and selected control patients matched for preoperatively assessed risk factors who had not developed delirium during index hospitalization. Other outcomes were mortality rate and rate of institutionalization.

RESULTS

During the follow-up period, 54.9% of delirium patients had died compared to 34.1% of the controls (relative risk = 1.6, 95% CI = 1.0-2.6). Dementia or MCI was diagnosed in 77.8% of the surviving patients with postoperative delirium and in 40.9% of control patients (relative risk = 1.9, 95% CI = 1.1-3.3). Half of the patients with delirium were institutionalized at follow-up compared to 28.6% of the controls (relative risk = 1.8, 95% CI = 0.9-3.4).

CONCLUSION

The risk of dementia or MCI at follow-up is almost doubled in elderly hip surgery patients with postoperative delirium compared with at-risk patients without delirium. Delirium may indicate underlying dementia.

Reasons for psychiatric consultation referrals in Dutch nursing home patients with dementia: a comparison with normative data on prevalence of neuropsychiatric symptoms

OBJECTIVE

To study psychiatric consultation referrals of nursing home patients with dementia and to compare referral reasons with normative data on prevalence of neuropsychiatric symptoms.

METHODS

This is part of a cross-sectional study of 787 patients residing in 14 nursing homes in the Netherlands. Nursing home physicians (NHP) noted the primary reasons for psychiatric consultation according to the Neuropsychiatric Inventory items and two extra domains. Patients were subsequently assessed by an old-age psychiatrist. Eligible patients were those that had dementia. Reasons for referral were compared with independent data on prevalence of neuropsychiatric symptoms in nursing home patients with dementia.

RESULTS

A total of 325/787 (41.3%) patients had dementia. Agitation, disinhibition and aberrant motor behaviour were frequent reasons for referral (>25%). Psychotic symptoms, apathy and eating behaviour changes were infrequent reasons (<10%) for seeking consultation. Agitation and disinhibition were more often primary reasons for consultation than would have been expected based on normative prevalence estimates of these symptoms. In contrast, delusions, euphoria, apathy, irritability and eating behaviour changes were less often reasons for referral compared with prevalence estimates.

CONCLUSIONS

This study is the first to examine psychiatric consultation for dementia patients in Dutch nursing home. Large differences exist between referral reasons and normative data on symptom prevalence. Specialized mental health service was provided for the agitated and disinhibited patient in particular. Chances are that this is at the expense of the apathetic, retarded and quietly 'not causing any trouble' patient.

Early symptoms in the prodromal phase of delirium: a prospective cohort study in elderly patients undergoing hip surgery

OBJECTIVES

The authors investigated prodromal delirium symptoms in elderly patients undergoing hip surgery.

METHODS

This was a prospective cohort study in the setting of a large medical school-affiliated general hospital in Alkmaar, The Netherlands. Participants were patients undergoing hip surgery aged 70 and older at risk for delirium. Before surgery, patients were randomized to low-dose prophylactic haloperidol treatment or placebo. Daily assessments were based on patient interviews with the Mini-Mental State Examination and Digit Span test. The Delirium Rating Scale-Revised (DRS-R-98) was used to measure early symptoms during the prodromal phase before the onset of delirium.

RESULTS

Data of 66 patients with delirium were compared with those of 35 at-risk patients who did not develop delirium: 14 of 66 patients (21%) had delirium on the day of surgery or early the day after, 32 of 66 (48%) on the second day, 14 of 66 on the third, and six of 66 (9%) on the fourth. The average DRS-R-98 total scores on day -4 to day -1 before delirium were 1.9 for the comparison group patients and 5.0, 4.3, 5.8, and 10.7 for patients with postoperative delirium. Multivariate analysis showed that the early symptoms memory impairments, incoherence, disorientation, and underlying somatic illness predict delirium.

CONCLUSIONS

Most elderly patients undergoing hip surgery with postoperative delirium already have early symptoms in the prodromal phase of delirium. These findings are potentially useful for screening purposes and for optimizing prevention strategies targeted at reducing the incidence of postoperative delirium.

Risk Factors and Prediction of Postoperative Delirium in Elderly Hip-Surgery Patients: Implementation and Validation of a Medical Risk Factor Model

OBJECTIVES

To evaluate risk factors for postoperative delirium in a cohort of elderly hip-surgery patients and to validate a medical risk stratification model.

DESIGN

Prospective cohort study.

SETTING

Medical school-affiliated general hospital in Alkmaar, the Netherlands.

PARTICIPANTS

Six hundred three hip-surgery patients aged 70 and older screened for risk factors for postoperative delirium.

MEASUREMENTS

Predefined risk factors for delirium were assessed on admission. One point was assigned for each of four risk factors present, resulting in three groups: low, intermediate, and high risk. Baseline screening and assessment included the Mini-Mental State Examination, the standardized Snellen test for visual impairment, chart review to determine Acute Physiological and Chronic Health Evaluation II score, and blood urea nitrogen to creatinine ratio. The primary outcome was postoperative delirium, as defined using Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, and Confusion Assessment Method criteria. All patients were screened daily for delirium.

RESULTS

Incidence of delirium was 3.8% in the low-risk group (P<.001), 11.1% in the intermediate-risk group (P=.27, relative risk (RR)=3.0), and 37.1% in the high-risk group (P<.001, RR=9.8). Cognitive impairment at admission had the highest predictive value for postoperative delirium (coefficient of determination=0.15). Contrary to previous findings, age was an independent predictive factor for delirium. Moreover, postoperative delirium was four times as frequent in acute patients as in elective hip-replacement patients.

CONCLUSION

The medical risk factor model is valid for elderly hip-surgery patients. Cognitive impairment, age, and type of admission are important risk factors for delirium in this surgical population.

Influence of Different Haemodialysis Modalities on AGE Peptide Levels: Intradialytic versus Long-Term Results

BACKGROUND

Peptide-linked degradation products of advanced glycation end products (AGE peptides) accumulate in chronic haemodialysis (HD) patients and may contribute to a number of HD-related long-term complications, such as accelerated atherosclerosis.

METHODS

The influence of a single HD session versus long-term HD on serum AGE peptides was determined. The patients were randomized to HD with a low-flux polysulfone (PS; F 6HPS), a high-flux PS (F 60S), a superflux PS (F 500S), or a superflux cellulose triacetate (CTA; Tricea 150G) dialyzer.

RESULTS

During a single HD session, both AGE peptides and reference peptides decreased significantly (AGE peptides: Tricea 150G -37.0 +/- 2.9%; F 6HPS -35.5 +/- 2.4%; F 60S -39.5 +/- 4.7%, and F 500S -43.3 +/- 2.1%, p = 0.005; reference peptides: Tricea 150G -73.2 +/- 8.8%; F 6HPS -73.2 +/- 7.9%; F 60S -72.5 +/- 8.2%, and F 500S -74.1 +/- 7.3%, p = 0.005). After 12 weeks of HD with the superflux CTA, the AGE peptide levels decreased significantly (week 1: 2.7 +/- 1.1 arbitrary units, week 12: 2.5 +/- 1.2 arbitrary units, decrease 7.4%; p = 0.01), whereas the AGE peptide levels remained unchanged after HD with each of the other three modalities. The reference peptide levels did not change after 12 weeks of HD.

CONCLUSION

Although AGE peptides can be effectively removed during a single HD session, superflux CTA seems to be the only modality capable of reducing AGE peptides in the long term.

Enhanced long-term reduction of plasma leptin concentrations by super-flux polysulfone dialysers

BACKGROUND

Hyperleptinaemia in chronic haemodialysis (CHD) patients has been associated with malnutrition, which is an independent predictor of morbidity and mortality in this patient group.

METHODS

To assess the influence of HD on plasma leptin, 10 CHD patients were crossover randomized to low-flux polysulfone (PS: F 6HPS), high-flux PS (F 60S), super-flux PS (F 500S) or super-flux cellulose-tri-acetate (CTA: Tricea 150G) for 12 weeks each. Blood samples were collected at the start of the study and each 12-week period. In addition, the relationship between patient characteristics, inflammation and leptin was analysed.

RESULTS

At baseline, all groups showed similar leptin concentrations (mean 33.6+/-21.7 ng/ml). After a single HD session, a significant (P<0.01) decrease was observed with all three high permeable devices (Tricea 150G -52.7+/-6.4%; F 60S -63.1+/-5.7%; F 500S -68.7+/-8.2%), whereas leptin remained stable with low-flux PS. After 12 weeks, a marked increase was observed with low-flux PS (week 1, 30.4+/-23.0; week 12, 40.5+/-5.4 ng/ml, P = 0.05), no change with super-flux CTA and high-flux PS (Tricea 150G week 1, 29.4+/-23.7; week 12, 32.0+/-27.9 ng/ml, P = ns; F 60S week 1, 36.0+/-31.8; week 12, 33.0+/-31.2 ng/ml, P = ns), and a significant decrease with super-flux PS (week 1, 38.3+/-33.0; week 12, 29.5+/-31.9 ng/ml, P = 0.02). The change in leptin after 12 weeks was significantly different between super-flux PS, and both low-flux PS (P = 0.009) and super-flux CTA (P = 0.01). Besides interleukin-6 (IL-6) at the start of the study (P = 0.006), no correlations were observed between patient characteristics, parameters of inflammation and plasma leptin levels.

CONCLUSIONS

Apart from low-flux PS, plasma leptin decreased considerably with all three high permeable dialysers after a single HD session. In the long run, leptin levels were lower with high-flux PS than with low-flux PS. Moreover, after switching from high-flux PS to super-flux PS (but not super-flux CTA), an additional decrease in leptin was observed. Apart from IL-6 at the start of the study, neither patient characteristics nor inflammatory parameters correlated with plasma leptin levels in this patient group.

Intercurrent clinical events are predictive of plasma C-reactive protein levels in hemodialysis patients

BACKGROUND

In chronic hemodialysis (HD) patients, the repetitive induction of the acute phase response (APR) may induce a chronic micro-inflammatory state, leading to various long-term complications.

METHODS

The present prospective study was designed to assess the alterations in the APR in 74 patients who were randomized to HD with a high-flux polysulfone (PS; F 60S), a super-flux PS (F 500S), or a super-flux cellulosic tri-acetate (CTA and CTA with filtered dialysate, CTA(f)) dialyzer. Blood samples collected at the start of the study and after twelve weeks were analyzed for interleukin-6 (IL-6) and C-reactive protein (CRP). In addition to the microbiological quality of the dialysate, the appearance of a "clinical event" was assessed.

RESULTS

At baseline, mean IL-6 levels were within the reference range whereas mean CRP levels were slightly elevated. Mean values did not change after 12 weeks of HD with either modality. After subdividing the patients in quartiles with increasing change in plasma CRP, 23.0% of the patients showed a change of more than 8.0 mg/L. In a multiple regression analysis, CRP levels appeared to be independent of the degree of dialysate contamination, the material and the flux characteristics of the devices. In fact, the variable "clinical events" was the only significant predictor of the plasma CRP levels (P < 0.001).

CONCLUSIONS

Based on these results, both PS and CTA super-flux dialyzers appear safe for clinical use. Whether changes in CRP values, which are associated with intercurrent clinical events, influence the long-term prognosis of chronic HD patients remains to be established.