Impact of monitoring approaches on data quality in clinical trials.

Abstract

Source data verification (SDV) has been reported to account for up to 25% of the budget in clinical trials (CT) and cost-benefit of SDV has been questioned. Guidelines for risk-based monitoring (RBM) were published in 2013 by agencies and in 2016, ICH-GCP-E6-(R2) added a requirement for risk-based approaches. This report will perform a comparison of the impact of RBM vs classic monitoring (CM) on data quality (defined as accuracy of data reporting from source data to final trial data) and expected impact on costs of CTs. Data on residual errors from four, large comparable randomised CTs were examined by post-trial SDV. Observed discrepancies were analysed in the categories of "overall" data, "major efficacy" and "major safety". In each category, the residual error rate was calculated as the number of discrepancies divided by the number of data-fields verified. A total of 1 716 087 data points were verified using CM and 323 174 using RBM. The overall error rate was 0.40% for RBM and 0.37% for CM (P < 0.01). For major efficacy, defined by risk assessment, the error rate was 0.15% and 0.28% (P < 0.0001); in major safety, defined by risk assessment, the error rate was 0.49% and 0.67% (P = 0.15), both in favour of the RBM approach. These empirical data, directly comparing RBM with CM, suggest that RBM improves data quality regarding data-points of major importance to trial outcomes, efficacy and major safety. Overall, the RBM approach showed a correlation to reduced amount of data collection errors with major relevance for interpretation of study results and subject safety as well as reducing on-site monitoring and data cleaning resources.