Using a Statistics-Based Tool to Enhance Risk-Based Monitoring Approaches

Centralized Statistical Intelligence and Data-Driven Risk-Based Monitoring

With complex clinical trial regulations and astonishing costs for monitoring and operations, how can a statistical approach to monitoring, such as Centralized Statistical Intelligence, enhance outcomes?

The clinical phase is the most complex part in a drug development process. It requires efficient planning and monitoring of clinical trials in order to achieve the desired quality and reliable study data for regulatory submissions. As Sponsors continue to confront clinical trial complexities, the costs of clinical monitoring and trial management have risen in an effort to achieve higher data quality and better patient safety.

Clinical monitoring is one of the largest expenses in a clinical trial accounting for 9-14% of the overall budget. Source Data Verification (SDV) can account for an additional 15% of the budget depending on the trial phase. Other budget concerns include clinical procedure costs (15-22%) and administrative staff costs (11-29%).

The most recent addendum to the ICH Guideline for Good Clinical Practice encourages “the implementation of improved and more efficient approaches to clinical design, conduct, oversight, recording and reporting while continuing to ensure human subject protection and reliability of trial results”. Risk-Based Monitoring is an approach supported by the FDA and EMA to protect the safety of trial subjects and reliability of the data collected.

More specifically, section 5.18.3 (Extent and Nature of Monitoring) has been updated to recommend a risk-based monitoring approach which includes centralized monitoring. According the addendum, centralized monitoring is “a remote evaluation of accumulating data, performed in a timely manner, supported by appropriately qualified and trained persons (e.g. data managers, biostatisticians)”.

The addendum goes on to say that centralized monitoring can reduce the extent and/or frequency of on-site monitoring and help distinguish between reliable data and potentially unreliable data by using statistical analyses of accumulated data to select sites and/or processes for targeted on-site monitoring.

Traditional monitoring can be costly, time consuming and difficult when it comes to detecting patterns across visits or subjects. In a Risk-Based Monitoring approach using metrics, the Sponsor and Data Managers set thresholds to flag sites or subjects. However, these thresholds can sometimes be subjective and fixed, and the Sponsor may have a long list of indicators to review.

Centralized Monitoring using statistical-based tools is another approach to RBM. This tactic includes:

  • Data-driven results rather than thresholds set at the beginning of the study
  • Data analyzed and collected with appropriate statistical models
  • Each subject or site is assigned a risk score based on the results of the statistical modelling
  • Sites or subjects with high risk are the focus of monitoring efforts (including corrective actions)

This approach allows for exceptional data quality checks. Risky sites and/or subjects can be flagged sooner rather than waiting for a subjective threshold. Using this approach, fraud can be easily detected.

How does CROS NT approach Centralized Monitoring?
We use Centralized Statistical Intelligence

CROS NT has developed a solution which includes a tool incorporating statistical methodology developed around Principal Component Analysis (PCA).

  • It assigns a risk score to each site by taking into consideration various indicators
  • The distribution of all recorded variables at each investigative site are compared with that of the other sites
  • Critical data can be selected by the Statistician, agreed by the Sponsor and the tool is configured by a Programmer

Principal Component Analysis is run on a matrix of indicators. Components are selected based on how much variance is explained and what the components represent. Thresholds are then used to classify sites into low, medium and high risk categories.

With ongoing studies, missing data tends to cause problems with traditional PCA models. Our solution is to implement either a Bayesian PCA or SDV imputation followed by traditional PCA. PCA can be carried out in SAS / IML®.

Centralized Statistical Intelligence – part of CROScheck®

CSI gives sponsors oversight of the quality of data coming from sites. It is part of our CROScheck clinical analytics solution which allows companies to:

  • Follow ICH guidelines and meet regulatory requirements related to a risk-based approach
  • Increase safety for patients, increase trial quality, spot potential misconduct and optimize spend
  • Conduct independent verification of the quality and integrity of the data collected from sites