GCRC Biostatistics
The goals for GCRC Biostatistics are to:
1 - provide input on statistical design to all proposed GCRC studies and ensure that appropriate power calculations, sample sizes, stratifications, and statistical analyses are employed;
2 - assist GCRC investigators with interim and final analyses of their studies;
3 - train GCRC investigators, fellows, and clinical faculty in experimental design, statistical concepts and statistical methods
4 - guide GCRC investigators in their preparation of statistical design sections of manuscripts and grant proposals; and
5 - provide GCRC investigators access to new statistical methods needed for specific projects.
GCRC Biostatistical Services
Dr. Hoffmann assists investigators with study design, power calculations, and the selection of appropriate statistical analysis methods. Investigators seek Dr. Hoffmann's advice prior to submission of their protocol to the GCRC Research Review Committee. To improve the efficiency of the Committee meetings, Dr. Hoffmann and Mr. Pajewski triage each proposal prior to its presentation at these meetings.
Dr. Hoffmann serves in an advisory capacity to the GCRC Research Review Committee (a subcommittee of the GAC). He prepares a detailed written report on the study design, statistical analysis, statistical power and evaluation of the statistical aspects of the DSMP/DSMB for each submitted protocol. Dr. Hoffmann's reviews constitute a portion of the overall review (disposition form) communicated to the investigators. When a protocol requires design modification, based upon Committee review, Dr. Hoffmann and Mr. Pajewski work with the investigator to incorporate the suggested changes. Dr. Hoffmann and Mr. Pajewski are particularly attentive to junior investigators who are conducting their first (pilot or feasibility) studies and meet with them individually. The availability of Mr. Pajewski has enabled the GCRC Biostatistics service to extend these one-on-one sessions from junior faculty to research fellows. In addition, Dr. Hoffmann and Mr. Pajewsi are available for consultation on statistical issues which arise during the course of the study. Dr. Hoffmann recommends interim analyses when appropriate, and he suggests changes in study design or recommends additional analytic approaches to account for missing data or changes. He also provides assistance for the design, statistical methods, and sample size determinations of NIH grant submissions by GCRC investigators.
Dr. Hoffmann oversees all of Mr. Pajewski's work, and they meet on an almost daily basis to discuss protocol design and statistical approaches. As part of his training, Mr. Pajewski may draft a protocol review and present it to Dr. Hoffmann for critique and clarification prior to submission to the Committee. For studies that require more involved statistical techniques, Dr. Hoffmann either supervises Mr. Pajewski in the performance of the analysis or else performs the analysis himself.
Some examples of methods that have been used over the last five years include:
• Generalized linear models and generalized estimating equations for non-normal data
• Logistic regression analysis
• Bootstrap and randomization tests to account for dependency among segments in the heart
• Proportional hazards survival models with the appropriate diagnostics
• Support Vector Machine (SVM) modeling to discriminate between types of timing task
• Time series for correlated fMRI data
• Spatial methods for correlated fMRI data
• Multivariate repeated measures and multivariate analysis of variance
• Review of analysis of variance results and advice on multiple comparisons procedures
• Variance-stabilizing transforms to improve the sensitivity of the repeated measures analysis
• Use of multivariate regression analysis or regression diagnostics in protocols
• Use of appropriate analyses in genetic association studies
• Determining the optimal dose testing regimen for a probiotic clinical trial through simulation
• PK/PD modeling to allow comparison of the effective dose of methylphenidate and cocaine
• Stopping rules for early termination for futility or early termination due to success of phase I/phase II effectivness trials
Additionally, situations arise where Dr. Hoffmann needs to adapt or translate new statistical approaches, e.g. estimating connectivity between regions of the brain, developing more efficient methods for combining results across subjects, carrying out multivariate spatial analysis to allow combination of information from different methods of obtaining information (fMRI, spectroscopy, DTI images of the brain, spatio-temporal models of changes in activation of the visual field diagram over task conditions or over changes in clinical conditions, such as stroke recovery).