Statistics Collaborative - Design and analysis for biomedical research


Statistical Programming

Preparing clinical study data and conducting statistical analyses are essential components of what we do at Statistics Collaborative, Inc. (SCI). Our staff of programmers and statisticians has many years of experience using commercially validated software, ranging from sample size calculations at the beginning of a study to sensitivity analyses following a regulatory submission, and all stages in between.

Much of our programming is done in SAS®, making use of our proprietary macros for version control and archiving of log and output files for each run. Our programs are independently reviewed by a more senior programmer or statistician using validation approaches customized to reflect the complexity of each program. Beyond SAS®, our programmers are also familiar with S-Plus®/R, nQuery Advisor®, PASS®, CART®, and other specialized statistical packages. In addition, we routinely use coding dictionaries (MedDRA, WHOdrug, and ICD9) to summarize adverse events and concomitant medications.

Examples of our programming capabilities and practices include:

  • Graphics: Using SAS®, we have developed in-house, validated macros for a variety of tabular and graphical presentations, examples of which can be viewed on our Special Presentations page.
  • Randomization: Our programmers are experienced in generating randomization schedules and in auditing the randomization processes of ongoing trials that implement either static or dynamic allocation methods.
  • Defensive programming: Interim data are by their nature “messy” and quite different from the final, locked databases used to generate clinical study reports. Our programming staff understands the unique issues posed by these data and is skilled at examining multiple sources in a clinical database to prepare thorough and complete interim monitoring reports.
  • External review: We have worked with code prepared by other programmers to serve as an “independent” reviewer or double-programmer to verify and validate results.

More specialized programming capabilities include simulations for study design, bootstrapping, randomization tests, and other non-standard statistical analyses.