Quality expectations and tolerance limits of trial master files (TMF) - Developing a risk-based approach for quality assessments of TMFs.

This article addresses the question of when a trial master file (TMF) can be considered sufficiently accurate and complete: What attributes does the TMF need to have so that a clinical trial can be adequately reconstructed from documented data and procedures? Clinical trial sponsors face significant challenges in assembling the TMF, especially when dealing with large, international, multicenter studies; despite all newly introduced archiving techniques it is becoming more and more difficult to ensure that the TMF is complete. This is directly reflected in the number of inspection findings reported and published by the EMA in 2014. Based on quality risk management principles in clinical trials the authors defined the quality expectations for the different document types in a TMF and furthermore defined tolerance limits for missing documents. This publication provides guidance on what type of documents and processes are most important, and in consequence, indicates on which documents and processes trial team staff should focus in order to achieve a high-quality TMF. The members of this working group belong to the CQAG Group (Clinical Quality Assurance Germany) and are QA (quality assurance) experts (auditors or compliance functions) with long-term experience in the practical handling of TMFs.

Comparative judgments with missing information: a regression and process tracing analysis.

We had participants decide which one of two applicants was better qualified for a scholarship. They also judged the difference between them (comparative judgment). The applicants were described by features (grades) in different subjects (dimensions). The grades on some dimensions were missing (unique dimensions) for an alternative while all the grades were available on other dimensions (common dimensions). In a conventional regression analysis, we found that decision makers gave more weight to dimensions when they were common than when they were unique. However, this commensurability effect was limited to medium important dimensions and did not apply to dimensions of high or low importance. We also observed how participants retrieved information for the choice alternatives and analysed how importance and commensurability are reflected in the processing prior to the decision. Features on more important or common dimensions were inspected earlier than features on less important or unique dimensions. Participants preferred dimensional transitions and inspected features on unique dimensions longer than their common counterparts. This finding suggested that participants used inferences when features were missing. We propose an outline of a decision heuristic to describe decision making with missing information.