From: Principles and procedures for data and safety monitoring in pragmatic clinical trials
Design features intended to increase generalizability and/or efficiency | Effect on data and safety monitoring procedures |
---|---|
Enrollment of all eligible patients in a defined clinical population | Recruitment or enrollment rate is more predictable More vigorous advertising or outreach is less likely to increase enrollment |
Cluster-level randomization | Statistical power is influenced by within-cluster correlation of outcomes Monitoring enrollment may include monitoring of cluster size and within-cluster correlation |
Safety profiles of study treatments may already be established | Between-group comparisons to establish safety or risks may be less useful |
Treatments often delivered by community providers, with little direct involvement by investigators | Fewer data available regarding adverse events or relationship between events and study treatment Investigators may have less ability to monitor or assure safe delivery of study treatments |
Frequency of contact with study staff may vary between treatment groups | Comparison of adverse event rates between treatment groups may be biased |
Fidelity of or adherence to treatments may be variable | Relationship between adverse events and study treatments may be more difficult to evaluate Valid inference may require some monitoring of (and corrective actions to improve) fidelity or adherence |
Potential adverse events often ascertained from health system records | Identification of adverse events may be significantly delayed |
Study outcomes often ascertained from health system records | Changes in health system record-keeping may affect integrity of study data Delays in access to records data may interfere with interim analyses of study outcomes |
Study questions may involve wider range of outcomes, such as effects on health service use or cost | Longer follow-up periods may be necessary to assess some “downstream” outcomes of study treatments |
Study questions may focus on implementation or policy, rather than individual clinical decisions | Decision thresholds for early termination of enrollment or intervention delivery may be asymmetric with respect to evidence for benefit or harm |