Your Easy Guide to Clinical Studies

Statistical analyses can be subjected to Quality Controls (QC). This entails that an independent statistician reviews the work of the statistician who initially performed the analysis (main statistician).

QCs support the accuracy and reliability of study results, which again provide confidence in study results.

 

Depending on the significance and complexity of the study, different types of QCs exist, such as:

• Reprogramming the analysis (double programming): an independent statistician performs the statistical analysis according to specifications (i.e. Statistical Analysis Plan (SAP) or study protocol), and compares the results with the results obtained by the main statistician.
• A review of the implemented statistical tests: QCs assess whether the implemented tests (e.g. descriptive, T-test, regression) used for the statistical analysis are consistent with the specifications outlined in the SAP or study protocol?
• A review of the statistical report: an assessment of the report based on credibility, plausibility, and layout)

As a SP-INV, implement in collaboration with a statistician QCs as described in the SAP of the study.

 

In the event the study has no SAP, apply QCs with regard to specifications given in the study protocol, or assess the extent of required QCs. Discuss with a statistician:

• The complexity of the study: How complex are the implemented statistical methods? More complex methods may require more rigorous validation techniques
• The importance of the study: How critical are the obtained study results? High-impact studies may require more stringent QCs (i.e. high-impact studies are expected to have a higher impact on society, public health recommendations, public policies)
• Available resources: the performance of QCs require additional resources (e.g. qualified staff, budget).

 

Ensure planned QCs are covered by the study budget. If it not covered ensure to include it in lessons learned.

References

ICH Topic E9 – see in particular

• 5.2.1 Full analysis set
• 5.2.2 Per protocol set