Article type
Year
Abstract
Introduction/Objective: Meta-analyses can be based on aggregate data extracted from the reports of published trials only, aggregate data from all trials or centrally collected updated individual patient data (IPD) from all randomised trials. The last of these has been described as the 'gold standard' against which other meta-analyses should be measured. A major advantage of collecting IPD is that it enables more detailed and flexible analyses (e.g. time-to-event, intention-to-treat and subgroup analysis). No commercial software package exists to analyse and plot the results of meta-analyses of randomised trials that use individual patient data (IPD). We have therefore developed SCHARP, a SAS application with a point-and-click interface that produces publication-quality graphs and appropriate summary statistics for time-to-event data.
Methods: SCHARP was initiated by the MRC Cancer Trials Office, Cambridge and has been developed collaboratively with Istituto "Mario Negri", Milan. It has been designed under the Windows operating system using the SAS/BASE, SAS/STAT and SAS/GRAPH modules. The interface has been developed using standard PROC PMENU and %WINDOWS commands with the addition of macro statements to automate certain processes. All actions can be performed by following a series of menu-driven dialogs, without the input of any SAS statements so that SCHARP can be used by those unfamiliar with SAS.
Results: Major features of SCHARP are that it: processes "raw" data on individual patients and uses the individual times-to-event; provides estimates of the hazard ratio and associated 95% and 99% confidence intervals for each individual trial and combined over all trials (fixed-effects model); performs subgroup analyses based on factors such as age, disease stage or histology; produces hazard ratio plots and survival curves by trial or subgroup and performs test of significance for treatment effect, heterogeneity, interaction and trend. Although SCHARP has been developed specifically for the purposes of IPD meta-analysis, the design also permits the analysis of single randomised trials.
Discussion: Despite greater availability of statistical software that performs or can be adapted for meta-analysis, the major packages do not yet support specific procedures. SCHARP specifically supports both the manipulation and range of procedures for the appropriate analysis of IPD. The latest version will be demonstrated.
Methods: SCHARP was initiated by the MRC Cancer Trials Office, Cambridge and has been developed collaboratively with Istituto "Mario Negri", Milan. It has been designed under the Windows operating system using the SAS/BASE, SAS/STAT and SAS/GRAPH modules. The interface has been developed using standard PROC PMENU and %WINDOWS commands with the addition of macro statements to automate certain processes. All actions can be performed by following a series of menu-driven dialogs, without the input of any SAS statements so that SCHARP can be used by those unfamiliar with SAS.
Results: Major features of SCHARP are that it: processes "raw" data on individual patients and uses the individual times-to-event; provides estimates of the hazard ratio and associated 95% and 99% confidence intervals for each individual trial and combined over all trials (fixed-effects model); performs subgroup analyses based on factors such as age, disease stage or histology; produces hazard ratio plots and survival curves by trial or subgroup and performs test of significance for treatment effect, heterogeneity, interaction and trend. Although SCHARP has been developed specifically for the purposes of IPD meta-analysis, the design also permits the analysis of single randomised trials.
Discussion: Despite greater availability of statistical software that performs or can be adapted for meta-analysis, the major packages do not yet support specific procedures. SCHARP specifically supports both the manipulation and range of procedures for the appropriate analysis of IPD. The latest version will be demonstrated.