Data management in clinical investigations - Part 4: Statistical considerations for pivotal investigations
Medical device manufacturers planning their first pivotal premarket investigations for CE-marking benefit greatly from early collaboration with statistics professionals and an understanding of statistical principles as presented in ICH's ”Statistical Principles for Clinical Trials”. Labquality offers data management, biostatistics and data protection services as part of the wider CRO clinical investigations offering.
Clinical benefit claims are quantifiable and comparative claims regarding the impact of the device on the health of the device user. Both quantification (how much, specifically) and comparator (compared to what) are required in the statistical design of a clinical investigation. The manufacturer establishes from previous knowledge the difference in the performance of their device and the comparator (typically, a difference in means in a given endpoint measurement) and also the respective standard deviations, since pivotal investigation essentially aims to replicate and rigorously demonstrate what is already known by the manufacturer. This information is also needed for power calculations, which set the objective for a number of subjects needed, which in turn drive the complexity and cost of the investigation.
To build accumulative evidence and convince potential customers, a clinical data strategy is designed. It consists of a series of clinical investigations and other high-quality publications. For the evidence to be accumulative, the clinical benefit claims must be stable and well-grounded with existing evidence and a consistent method for end-point measurements is needed. To support generalization, the investigations are often conducted in different countries and/or widening the inclusion and exclusion criteria to approximate real users in real clinical settings. A common pitfall is too narrow a focus on getting permission to sell and neglecting how the investigation fits into the wider clinical data strategy.
Pivotal clinical investigations are confirmatory investigations with a pre-set hypothesis. Compared to more exploratory types of research done before pivotal investigations, clinical investigations have a specific outcome. That is, typically, whether the null hypothesis can be rejected. Pre-setting means that analysis objectives are documented in the protocol and preregistered to clinical investigation databases before data collection. This limits bias arising from the attempt to design objectives after seeing the data.
The statistical analysis objectives and methods are planned and documented in the statistical analysis plan. Controls are designed to reduce bias. Commonly, the treatment arms are made as homogenous as possible regarding the presence of relevant background factors by using randomization or propensity matching. Blinding is done to avoid subjects being selected for the study or end-point evaluation is done in a biased way. The identification and mitigation of scientific risks related to end-point measures, samples, confounding factors, execution of the study assessments etc. is a significant effort in the development of the statistical analysis plan SAP.
This post was written by Labquality‘s Data Manager Markus Vattulainen. You can contact him for more information.