With the rise of Risk-Based Monitoring as an efficient and cost-effective clinical data strategy, regulatory authorities and clinical trial sponsors have raised concerns over data fraud. The FDA, for example, reviews and audits several clinical trial sites for “fraud, incompetence and misconduct”.
When it comes to clinical data, a clinical trial database can never be completely free from errors. The ICH-GCP requirements stipulate that data needs to be “accurate, complete and verifiable from source documents”. Therefore, where do we draw the line between error and fraud?
Data Fraud is thought to be rare in clinical studies (estimated to be less than 1% of cases). In addition, identifying and documenting fraud can be time consuming and expensive and can obviously potentially damage the reputation of the research company.
But is the importance of fraud detection underestimated? Why should we even bother looking for fraud?
- Patients’ lives and health could be at risk. Regulatory authorities are constantly looking to improve the authenticity of clinical trial data to protect the well-being of patients enrolled in a trial.
- To reduce failed outcomes. If data quality analysis is carried out often, problems can be addressed and resolved early on in the drug development process.
- Preserve the integrity of clinical research.
Types of Clinical Data Fraud
All of these “errors” can have devastating effects on trial outcomes especially if they results in exaggerated treatment effects.
- Fabricated Data: missing or outlying values replaced by plausible values or fabricating trial participants and all associated data values
- Data falsified to reach a desired objective: e.g. making patients eligible, showing a treatment effect
Fraud Detection in Clinical Trials
The conventional approach to fraud detection is to visit medical centers in person using Source Data Verification plus additional investigations if necessary. This method can be both costly and drag out development timelines.
The alternative method is to reply on statistical methods to detect abnormal data patterns which can be translated into graphical checks to identify potential fraudulent data. Data that is typically susceptible to fraud include: eligibility criteria, repeated measurements, adverse events, assessment of medical compliance, assessment dates and patient diaries.
If fraud is suspected or indeed confirmed, monitors need to be notified and the appropriate on-site audit and investigation should be scheduled.
CROS NT and Fraud Detection
CROS NT has developed a graphical tool to analyze data patterns from clinical sites. Different quantitative variables can be analyzed at the same time as well as different types of graphs. Data fraud generally appears at the site level, and CROS NT statisticians can use statistical methods and tools to analyze inliers, incorrect dates, under-reporting of adverse events, integer rounding, digit preference, extreme variances and unusual correlation structures. Interactive softwares and visualization tools are the most appropriate instruments for performing such analysis.
In conclusion, there is a fine line between data recording errors and fraud. However, fraud can through the entire integrity of a trial into jeopardy and risk the health and well-being of patients. Even if fraud is identified, it may not be enough to simply correct or discard that data. The performance of the trial will inevitably be called into question. More importantly, the European Union has taken measures to make clinical trial data accessible in a public database by June 2016, and therefore the quality and accuracy of data is more imperative than ever.