Workpackage Leader: Charité
This work package will be led by the Charité - University Medicine Berlin, which has a long-standing expertise and publication record with regards to prospective and retrospective (database-driven) analyses of small-scale and large-scale clinical studies focusing on questions of the impact of body weight and body weight changes as well as metabolic alterations, exercise capacity assessments and survival analyses in patients with heart failure.
A joint database will be created from data received from the different members of the SICA-HF consortium including patients' personal information (recruiting center, date of recruitment, patient initials, date of birth, sex, age) and clinical information (aetiology, weight, height, body mass index, medication, echo parameters, body composition parameters, and work-package specific parameters, etc.) (Objective 1). A full-time statistician will be employed to perform necessary database programming, to elaborate and write the appropriate Statistical Analysis plans and to then help in solving the questions of descriptive and analytical statistics during the whole study period of SICA-HF (objectives 2 to 12).
All statistical analyses will be carried out using either the Statistical Package for the Social Sciences (SPSS) version 15 for Windows (SPSS Incorporated, Chicago, Illinois, USA), StatView 5.0 software for Windows (Abacus Concepts, Berkley, CA), or MedCalc for Windows version 8.2.0.3 (Broekstraat, Mariakerke, Belgium). All continuous data will be checked for normal distribution using the Kolmogorov-Smirnoff test. Non-normal distributed data will be treated as such or transformed to achieve normal distribution (e.g. log-transformation). Statistical analysis will make use of Student's paired and unpaired t test and analysis of variance with Fisher's post hoc test to compare differences between groups for normally distributed variables and using Mann Whitney U-test, Wilcoxon-test, and Kruskal-Wallis-test for non-normal distribution variables, as appropriate. Associations between variables will be assessed using univariate or multivariate (step-wise, where appropriate) regression analyses. A value of p<0.05 will be considered significant. To compare different predictive values, areas under the curve (AUC) for sensitivity and specificity will be constructed for relevant variables such as novel biomarkers (WP 10, WP 11) in relevant large patient subgroups. The best prognostic cutoff for survival status is defined as the highest product of sensitivity and specificity. To contrast prognostic accuracy, statistical comparison of receiver operating characteristic curves (ROC) will be performed using the method for paired receiver operating curves described by Hanley and McNeil (Hanley and McNeil, 1983). The relationship of baseline variables with survival will be assessed by Cox proportional-hazards analysis (single predictor and multivariable analysis). Hazard ratios and 95% confidence intervals for risk factors and significance level for c2 (likelihood ratio test) will be given. To estimate the influence of risk factors on survival, Kaplan-Meier cumulative survival curves will be constructed and compared by the Mantel-Haenszel log-rank test.
The SICA-HF consortiums further aims to elucidate the role of type 2 diabetes, obesity, and cachexia in patients with heart failure by analysing existing databases from large-scale clinical trials. These trials, whose databases are available for or from members of the consortium, include the Cardiac Insufficiency Bisoprolol Studies II and III (CIBIS II and CIBIS III), the Proactive Prospective Pioglitazone Clinical Trial In Macrovascular Events (PROactive), the Eplerenone Post-AMI Heart failure Efficacy and Survival Study (EPHESUS), the Valsartan Heart Failure Trial (Val-HeFT), the Optimal Therapy in Myocardial Infarction with the Angiotensin II Antagonist Losartan study (OPTIMAAL), and the Evaluation of Losartan in the Elderly (ELITE II) (objective 3). An statistical analysis plan has been developed that includes the following aims whose specific focus will be on type 2 diabetes, obesity and cachexia:
Body weight changes will be analysed in absolute terms and relative terms from baseline (of the respective trial or SICA-HF) and for the last 12-month period. At baseline and during follow-up we will consider body weight in 2 scenarios, if oedema is recorded: i) regardless of oedema status, as it was done for CHF patients in SOLVD [see Anker et al., Lancet 2003] in order to avoid any potential bias for over-detection of weight loss (i.e. to present conservative estimates of weight loss), and ii) with adjustment for oedema status by only accepting weights in oedema free status.
Importantly, in time-dependent analyses (to analyze the immediate impact of weight changes or changes on survival and/or hospitalization) one must consider body weight increases and body weight decreases independently of each other, as otherwise one never knows whether weight change predicting survival is driven by negative weight change (i.e. weight loss) or positive weight change (i.e. weight gain). It is clinically important to separate the two issues.
These statistical issues need to be also applied to changes in glucose metabolism, time to changes in HbA1c levels. At this stage, we are not able to perform formal power calculations for most of these analyses. , Hence, we regard these database analyses as valuable but exploratory, unless internal validation by creating derivation and validation datasets is possible. We aim to perform this strategy as much as possible as done for the SOLVD treatment trial (Anker et al, Lancet 2003).
Consortium
Rationale & Design Paper
Search
Members
Login
