Please activate JavaScript!
Please install Adobe Flash Player, click here for download


ISCB 2014 Vienna, Austria • Abstracts - Poster Presentations 125Wednesday, 27th August 2014 • 11:00-11:30 Monday25thAugustTuesday26thAugustThursday28thAugustAuthorIndexPostersWednesday27thAugustSunday24thAugust mance can then be summarised across the omitted studies. Application: Utilising IPD from seven studies to develop a prognostic model for VTE recurrence in unprovoked patients, we describe strengths and complications of the IECV approach, and subsequent adaptations. Challenges include dealing with studies with small numbers of events, missing data from one or many studies and heterogeneity across study populations. Conclusions: There are strong benefits to using the IECV approach for prognostic model building, and using IPD from several studies can pro- duce more robust models. However challenges remain to be addressed. P3.5.55 Creation and validation of a predictive model to assess poor outcomes in acute decompensated heart failure AG Unzurrunzaga1 , S Garcia-Gutierrez1 , I Arostegui2 , I Barrio-Beraza2 , U Aguirre3 , A Antón1 , JM Quintana1 1 Hospital Galdakao-Usansolo, Osakidetza-REDISSEC, Usansolo, Spain, 2 Universidad del País Vasco-REDISSEC, Leioa, Spain, 3 Hospital Galdakao-Usansolo, REDISSEC, Usansolo, Spain   Objective: To create a predictive model to assess severity in acute decom- pensated heart failure at acute setting. Methodology: Prospective cohort study. Patients over 18 years who were seen between April 2011 and April 2013 for acute decompensated heart failure were included. Sociodemographic, cardiovascular risk factors, co- morbidities, functional status, and general medical history of heart disease and analytical and echocardiographic data were collected. Dependent variable was“poor evolution”, a composed variable defined as: 1) mortality in acute period (one week); 2) admission to an intensive care unit (ICU); 3) need for invasive mechanical ventilation (IMV); 4) cardiac arrest; 5) use of non-invasive mechanical ventilation (NIMV). Statistical analysis: Categorical variables are expressed as frequencies and percentages and continuous variables as means and standard deviations. Model was created by means of multilevel logistic regression model ad- justed by hospital . The model as well as its AUC was validated by split validation techniques. Bootstrapping was used to internally validate the models’AUC. Results: The 51.35% of the 1856 patients of our sample were women, mean age was 79.6(9.7) with more than two comorbidities in the 61.69% of the cases. Background of acute myocardial infarction, previous visits to ED in the two previous years, glycemia and BUN entered into the model (AUC=0.78, 95%CI 0.73-0.83). Split validation techniques showed simi- lar AUC (0.79, 95%CI 0.74-0.84) as bootstrapping techniques (AUC=0.97, 95%CI 0.75-0.83). Conclusions: We achieve a simple to use predictive model to assess poor evolution in acute heart failure. P3.5.62 Model building using learning methods to identify SNPs related to the pharmacokinetics of high-dose methotrexate in pediatric acute lymphoblastic leukemia A Harnos1 , K Csordas2 , O Lautner-Csorba3 , AF Semsei3 , M Hegyi2 , DJ Erdélyi2 , OT Eipel2 , C Szalai3 , GT Kovács2 1 Dept. of Biomathematics, Szent István University, Budapest, Hungary, 2 2nd Dept. of Pediatrics, Semmelweis University, Budapest, Hungary, 3 Dept. of Genetics, Semmelweis University, Budapest, Hungary   High-dose methotrexat plays an important role in the consolidation ther- apy of acute lymphoblastic leukemia. We investigated the influence of single nucleotide polymorphisms (SNPs) in genes of the folate metabolic pathway, transporter molecules and transcription proteins on the phar- macokinetics of methotrexate (MTX) and 7-hydroxy-methotrexate (7-OH- MTX). 63 SNPs of 14 genes were genotyped. Totally 463 treatment courses were analyzed (4 measurements of each patient). As a first step, random forest method (RF) was applied to calculate vari- able importance measures because of the large number of explanatory variables and the relatively small sample size. With the RF, we selected the important ones from all explanatory variables for further analyses. In the second step, we built classification and regression trees (CART) with the preselected explanatory variables to generate clinical decision rules and to explore the relationship between the response and the explana- tory variables. In the last step general linear mixed model (testing the relationship be- tween SNPs and the logarithmic transformed serum levels) was applied to prove the significance of the selected variables and their interactions and to estimate effect sizes. SNPs (rs4948502, rs4948496, rs4948487) of the ARID5B gene were associated with the serum levels of MTX (P<0.02), se- rum levels and AUC of 7-OH-MTX (P<0.02). The rs4149056 of the SLCO1B1 gene showed also a significant association with the serum levels of MTX (P<0.001). Our findings confirm the association of novel genetic variations in folate- related and ARID5B genes with the serum MTX levels. P3.5.82 Prediction of outcome after severe and moderate head injury by classification and regression tree technique VK Kamal1 , RM Pandey1 , D Agrawal1 1 All India Institute of Medical Science, New Delhi, India   Traumatic brain injury is the leading cause of disability and death all over the Globe. Our aim is to develop and validate a prognostic model, which is simple and easy to use for In-hospital mortality and unfavourable out- come at 6-months in patients with moderate and severe head injury in- volving rapidly and easily available variables in daily routine practice. For this, a classification and regression tree (CART) technique was em- ployed in the analysis by using trauma database (n=1466 patients) of con- secutive patients. A total of 24 prognostic indicators were examined to predict In-hospital mortality and outcome at 6 months after head injury. For In-hospital mortality, there were 7 terminal nodes and the area under curve was 0.83 and 0.82 for learning and test data sample respectively. The overall classification predictive accuracy was 82% for learning data sample and 79% for test data sample, with a relative cost 0.37 for learning data sample. For 6-months outcome, there were 4 terminal nodes and the area under curve was 0.82 and 0.79 for learning and test data sample re- spectively.The overall classification predictive accuracy was 79% for learn- ing data sample and 76% for test data sample, with a relative cost 0.40 for learning data sample. Methodologically, CART is quite different from the more commonly used statistical methods with the primary benefit of illustrating the important

Pages Overview