Regulatory agencies are encouraging risk-based monitoring as a way to both improve data quality and streamline clinical trial costs. Risk-based monitoring should be driven by indicators of the quality and performance of investigational sites. A common way to evaluate site performance is to monitor predefined metrics, often called “Key Risk Indicators” (KRIs). These typically include variables known to be relevant indicators of quality (e.g. accrual rate, frequency of adverse and serious adverse events, frequency of data queries and time taken to resolve them). A robust complementary approach is Central Statistical Monitoring (CSM), which compares sites to each other and detects abnormal data patterns that may be indicative of errors, misunderstandings, sloppiness, data fabrication or fraud.

In this talk, we will describe an implementation of central statistical monitoring using a software that automatically performs a large number of statistical tests on all available data in multicenter clinical trials, resulting in a high-dimensional matrix of P-values which are summarized in an overall score for each center. The score is analogous to an average P-value, with low scores indicating centers that are statistically most different from all others. Centers are ranked for each statistical test performed as well as for their overall score. A permutation technique is used to assign statistical significance to the overall score of each center. Outlying centers, i.e. centers that have low ranks and/or a statistically significant overall score, are flagged for further scrutiny since their data are not statistically consistent with the data from all other centers. A risk-based monitoring strategy can use this information, along with other performance indicators, to prioritize on-site monitoring for sites identified at higher risk. In multinational clinical trials using regional monitoring teams, the approach can also unveil data patterns that reflect unanticipated differences in patient management or other important aspects of trial conduct. Finally, when a trial is completed, the same approach can be used to check that the data are statistically consistent across all sites, regions or countries.

Reference: Venet D, Doffagne E, Burzykowski T, et al. A statistical approach to central monitoring of data quality in clinical trials.Clin Trials 2012; 9: 705-13.