Contents © 2004-2011 Massachusetts
General Hospital


Capsule Summary

Cluster analysis of microarray data

The Clinical Biostatistics Core (CBC) provides clinical data management and statistical support to the Program. Within its clinical data management activities, the Core insures proper collection and flows of data into TrialDB (the clinical data collection application) and insures completeness and accuracy of the TrialDB data. The CBC assists the participating investigators in the design and development of experiments for primary analysis, and performs statistical analysis and report writing for manuscripts. The CBC serves as a link between the Information Dissemination and Data Coordination Core (IDDC) and the clinical and scientific investigators.

The Challenge

A major challenge in clinical research is obtaining a clinical data set large enough and of highest quality to elucidate the etiology of what is being studied. The research subjects must be heterogeneous enough to generalize to a wide patient population, and the data must be consistent enough to be pooled for analysis. The program leaders felt strongly that it was critical that those investigators involved ultimately in the analysis phase of the program be involved in the upfront design phase of the databases and the activities of the other cores. As a result, the CBC statisticians and the clinical data manager have been involved in the interactive design as well as the management of the clinical study data system.

The Approach

The Clinical Biostatistics Core maintains a steadily growing clinical database of completed case report forms from the trauma and burn studies. Subsets of “clean” clinical study data from TrialDB are produced by the CBC and made available to our participating investigators for analysis purposes in multiple formats, including a Microsoft Access database and an SQL script for producing a generic relational database. The Trauma-Related Data Base (TRDB) web-based interface provides de-identified, easily queried, relational table access to the data.

The CBC biostatisticians provide statistical support across the entire spectrum of research being conducted by our participating investigators. The group works in close collaboration with the Proteomics and Genomics Cores, and the PORC. The group provides leadership and expertise in all phases of the research enterprise from clinical data management, design of experiments, development and execution of comprehensive data analysis plans, and production of thoughtfully written statistical reports on original research and development of novel statistical methodology.

A Core innovation in the write-up of these plans is to present the analytic methods both in a non-technical discursive style, easily accessible to a non-statistician, which explains the intuitive idea underlying each method, as well as in the more formal technical style suitable for inclusion in the statistical methods section of the final manuscript. After the investigator reads and "signs off" on the data analysis plan, the Core staff, together with the IDDC Core, cleans and locks an analysis dataset. The statistician completes the analysis based on the locked data set, forwards the results to the investigator, and remains available to help with the subsequent interpretation of the results. An additional collaborative effort between the CBC and the investigators is the reduction of the raw clinical data to easily interpretable and clinically meaningful measures of baseline injury and clinical outcome.