Quality Control in Statistics
As an integral part of Good Laboratory Practices (GLPs), the Quality Assurance Unit (QAU) confirms that no unapproved protocol deviations occurred in a study and assures that the methods and results are accurately reported in the final study report. Yet, unless formally trained in statistics, most Quality Assurance (QA) personnel are not qualified to confirm the accurate conduct and reporting of statistical analyses. Likewise, validation of statistical software requires a thorough comprehension of statistics to verify that requirements are being met.
Methods of Quality Control (QC) in statistics fall into one of the following 4 categories:
No Statistics Quality Control
The QAU has the responsibility of assuring that protocol statistical analysis methods are conducted correctly and that the results of those analyses are accurately reported in the final study report. Therefore, QC of statistics is essential to any study conducted under GLP regulations.
Verification of Statistical Output with Summary Tables
It is common practice for a QAU to verify reported summary table results against those of a raw dump of statistical output (eg. SAS® output). However, this type of QC does not address the following questions: 1) Was the statistical methodology conducted as described in the protocol; and 2) Does the programming code accurately perform the planned analyses? Many different methodologies and programming codes produce statistical output that may appear, to an untrained statistical reviewer, to be correct. Consequently, summary table results may be verified against statistical output that in itself is performed incorrectly or is invalid.
Independent Reproduction of Statistical Analyses
In this QC method, an independent statistician reproduces summary tables and results of statistical analyses. This requires that a QC statistician performs the interpretation and conduct of planned statistical methods, independent of personnel conducting the primary statistical analyses. Just as the QAU independently monitors a given study, the QC statistician independently verifies the statistical analyses.
Quality Control of Statistical Methodology
This type of QC requires an independent review of the planned statistical methodology for a given study. Although there is a certain amount of subjectivity in the choice of “correct” statistical methods, this type of QC may prevent inappropriate methods or produce an alternative methodology that is better suited to address the objectives of the study protocol.
The BioSTAT Solution
To assure the quality of its service, all statistical results and summary reports produced by βioSTAT, on behalf of its clients, are subject to the independent reproduction of statistical analyses QC methodology. Contact us for more detail.