25. How important is the accuracy of the clinical trial data?
It is not the accuracy of the individual trial data that is important, but the reliability and robustness of the trial results. The aim of the management, monitoring and data management activities is recommended to focus on the data and activities that are critical to the reliability of the trial results, for example, the endpoint for the primary objective of the trial or key design aspects (e.g. randomisation). These would be identified during a risk assessment of the trial. It is recommended to aim for a high level of accuracy in these areas identified and potentially accept some degree of error in other areas. Consideration for defining such acceptability in terms of tolerance limits is recommended
Concern may be raised by sponsors about the reliability of the data where no SDV is conducted, for example, if the primary endpoint has not been verified or if there has been no review of patient notes to detect unreported SAEs or if the data submitted in the CRF is found to be incorrect. These concerns arise from the traditional monitoring approach that all the data has to be accurate and that any error is unacceptable. The design of the trial can assist in reducing or mitigating the impact of missing or incorrect data, for example, the results of large blinded, randomised trials with high power are unlikely to be affected by increased variability/omissions of the data, particularly as the errors/omissions would not be differential on a treatment basis (biased). Small blinded and randomised trials may suffer from reduced power with increased data variability/omissions and there is potential to increase the risk of a false negative result. Open trials are more at risk from bias, as errors and omissions could be potentially differential for the treatment groups. This issue is recommended to be evaluated as part of the risk assessment to determine what level of SDV (and other monitoring checks) is needed to mitigate any concerns about the reliability of the trial results. The monitoring plan may take a conservative approach initially, then reduce the monitoring intensity if the concerns are not realised. The data accuracy and proper conduct of the trial can be influenced not only by the monitor detecting and the investigator correcting errors retrospectively (where possible), but by prevention of such errors in the first place, for example, by appropriate trial design, training, communication and systems that facilitate the conduct of the trial.
Version 1: 22 February 2013