Stats
Seminar notes "S-PLUS Clinical Safety Miner" (March 10, 2005).
Category: Statistical computing
I attended a web seminar by Michael O'Connell, "Applications in Drug Discovery and
Development. S-PLUS' Clinical Safety Miner." Michael O'Connell is the Director, Life Science
Solutions at Insightful Corporation, the company that
produces S-plus software.
Insightful also produces a range of other products that work with S-plus, such as
-
S+SeqTrial, for design
and analysis of group sequential trials with interim stopping rules.
-
S+ArrayAnalyzer, which analyzes data from microarray experiments.
This talk focused on S-plus Clinical Safety Miner which integrates features from
The seminar showed three applications of Clinical Safety Miner that emphasized the ability
to produce interactive and easily updateable web reports. These web reports allow you to
drill down from aggregated measures of adverse event risk to get subgroup information or data
on individual patient events. The web report has an Rich Text Format (RTF) template which
makes it easy for you to produce high quality printed reports.
Dr. O'Connell also demonstrated the capability for this software to
document validation as required in 21
CFR 11. This rule presents
criteria under which FDA will consider electronic records to be equivalent to
paper records, and electronic signatures equivalent to traditional handwritten
signatures.
This rule has six components:
-
validation;
-
the ability to generate accurate and complete copies of records;
-
archival protection of records;
-
use of computer-generated, time-stamped audit trails;
-
use of appropriate controls over systems documentation; and
-
a determination that persons who develop, maintain, or use electronic records and
signature systems have the education, training, and experience to perform their assigned
tasks.
Anyone submitting data to FDA needs to know about 21 CFR 11.
Perhaps the most interesting of the three applications involved data from the FDA AERS
database on four COX-2 inhibitors. The software computed and displayed observed versus
expected counts of adverse events for each drug. It used a Bayesian Poisson model implemented
with a Markov Chain Monte Carlo (MCMC). This model relied on a newly released S+Bayes
library. I asked at the end of the seminar whether this software could be used for reporting
to IRBs that are providing continuing review of research studies. Dr. O'Connell said that
this would be an excellent application of the software and would allow the IRBs to better
understand the flow of data.
The Insightful web site has materials from the talk at
07/08/2008.
