The BF provides statistical expertise in study design and data analysis for members of UPMC Hillman Cancer Center. Statisticians within the unit collaborate on developing protocols for clinical trials, and research proposals for laboratory-based investigations, epidemiologic studies, and clinical investigations. We perform statistical analyses for such studies, substantively interpret their results and author or co-author papers for publication. The BF is committed to developing statistical methodology that aids cancer research, and to applying statistical and computational methods to accomplish this.
Study design
- Advise on statistical aspects of the design of laboratory-based studies, clinical trials, clinical investigations, epidemiologic studies, and population-based studies such as screening programs
- Formulate study objectives and endpoints in terms that are appropriate for statistical analysis
- Recommend and implement novel statistical methods, such as adaptive dose-finding techniques for phase I clinical trials
- Propose and implement alternative designs for phase II clinical trials
- Determine sample size needed to address study objectives at an appropriate level of significance, power or false-discovery rate
- Develop and write plans for statistical analyses
- Develop and implement randomization procedures
Data analysis
- Analysis of diagnostic & screening tests
- Assessment of prognostic markers
- Bayesian estimation
- Evaluation of cancer patients’ quality of life in clinical trials
- Interim monitoring of clinical trials using frequentist or Bayesian methodology
- Kaplan-Meier estimation
- Linear, nonlinear and multiple regression
- Logistic, polytomous, and Poisson regression
- Logic regression
- Meta-analysis
- Mixed models
- Nonparametric estimation
- Proportional hazards regression
- Repeated measures analysis
Grant Proposals, Manuscripts and Presentations
- Write analysis plans and design justifications for clinical and preclinical experiments
- Review study objectives and align them with analysis plans and experimental designs
- Scientific editing
- Write Statistical Methods and Results sections of manuscripts and posters
- Graphics for manuscripts, posters and presentations
Equipment
Each biostatistician has a workstation equipped with the latest software for statistical analysis (SAS, R, S-Plus, Python, StatXact, Stata, NCSS/PASS) and office productivity (LaTeX, Access, Excel, Outlook, PowerPoint, Word, MathType, Reference Manager). Their workstations are linked through a firewall-protected local area network to servers at the Pitt Network Operations Center which provides data backup, file sharing, a web-enabled patient randomization service, and a web-based tool for sample size calculations for in vivo experiments.
Shared Resource statisticians have access to the Pittsburgh Supercomputing Center for computing-intensive applications.
The BF also supports a legacy data system for raw data, analysis code, analysis results, graphics and reports associated with peer-reviewed publications.
The BF no longer supports the Java Bryant-Day clinical trial design tool. The function freq_binom_two_bryantday_twostage in the R package EurosarcBayes, available from CRAN, provides these designs.