CluePoints’ Statistical Strategies Enhance Risk-Based Monitoring

By Ann Neuer 
September 18, 2013 | Francois Torche, CEO of CluePoints, a statistical monitoring provider, has a front row vantage point on the ongoing industry-wide shift to risk-based monitoring (RBM).  He sees RBM as critical to improving the quality of clinical trial data through the use of statistical methods that help define more effective on-site monitoring strategies.
“There is tremendous interest in risk-based monitoring right now. All of the large sponsors have in-house initiatives, and just look at the number of conferences on the subject.  It’s amazing to see the number of people attending those sessions as they start to understand the huge benefit of moving away from the practice of 100% source data verification (SDV),” Torche says.  
TorcheHe is referring to the monitoring methods used extensively today, which date back to 1988, when the Food and Drug Administration (FDA) issued a guidance on the monitoring of clinical trials.  That guidance left many life sciences companies feeling compelled to do 100% source data verification, meaning that each and every data point reported in some or all aspects of a clinical trial is compared manually to the primary record of study subjects.  But with the withdrawal of that guidance in 2011 and the release of a new one in August 2013, entitled Oversight of Clinical Investigations —A Risk-Based Approach to Monitoring, Torche has seen a distinct trend toward acceptance of the risk-based approach.  
As described in the new guidance, FDA recognizes that there are statistically meaningful ways to comply with monitoring requirements that can yield better data than the 100% SDV technique.  Specifically, to help sponsors improve their oversight of clinical trials, the guidance encourages greater use of technology for statistical sampling, as well as centralized monitoring methods, where appropriate, replacing the exhausting and cost-prohibitive traditional methods of regular onsite visits by monitors.  
The opportunity for cost savings is substantial, as reported in a limited amount of published research.  One study in the Quality Assurance Journal indicated that for a typical Phase III trial, approximately one quarter of the budget was spent on SDV activities.  For a $100 million Phase III program, this represents $25 million.  In addition, a 2010 article in the Drug Information Journal on alternatives methods of monitoring suggested that cost reductions of 14.3% in Phase III cardiology studies and 23.5% in Phase III oncology trials were possible by moving away from 100% SDV. 
CluePoints is smack in the middle of this SDV to RBM transition.  Using SMART, a patent-pending software solution, clients can upload their clinical trial data periodically on the CluePoints web portal.  Based on various algorithms, SMART determines the best statistical techniques for evaluating the data and identifying the best way to supplement monitoring activity.  Next, CluePoints’ analysts review the resulting output and generate a report that identifies all the signals that highlight anomalous and inconsistent data from investigative sites and suggests why this could have occurred.  With the web portal, the report is interactive, so clients can directly react to the results and receive guidance to best exploit the report’s findings.
Torche comments, “We look at all the variables and endpoints that are patient related, and compare the values of one site against all the values in the others.  Using statistical methods, we can detect signals to better manage risk and determine variations from one site to another, and the differences between what is normal and the values that are being recorded.  This allows us to identify which sites are outliers or are at risk so we can send monitors to those sites sooner.  This improves data quality.”    
Statistical methods are critical when large volumes of data are involved because the review of all the data manually can easily introduce error.  Torche likens this effort to the childhood game of comparing two similar images and asking the participants to identify where they differ.  “This game is played using your eyes.  With two images, this is possible, but when the volume of information is massive, you can no longer use your eyes,” he says.  He points to one example in which CluePoints was involved in a study with 1,200 sites.  Approximately 800,000 p values were calculated, a measure of statistical signficance among study variables.
In modifying monitoring methods, there is a tendency to focus on key risk indicators, such as the number of queries, number of adverse events, or number of serious adverse events, all of which are pre-determined factors.  But as Torche explains, “that approach can give you a hint of possible issues, but we look at all the data, an approach that allows us to focus on the issue itself.  This provides the most objective and most meaningful interpretation of the available information.”    


Add Comment

Text Only 2000 character limit

Page 1 of 1