By Maureen Lyden, M.S., president/CEO, BioStat International
Throughout my years of work in biostatistics, there are certain problems I see recurring repeatedly in clinical trials. The most common problems are incorrect study designs and a lack of sufficient exploration into study conduct processes that could affect analyses and interpretation.
I have seen examples of companies that have reached the end of a study but have an unacceptable proportion of patients with missing data. They find that there are problems in analysis. Conclusions drawn from the data may be less precise and likely biased. Even if not biased, they may not have enough evaluable patients to meet statistical objectives and project goals.
The handling of missing data must be discussed in the statistical analysis plan of a study. To provide for accurate trial results, the missing data discussion should occur during the study planning stages, not after a study has veered out of control with gaps in key information. The statistician can plan the appropriate analyses for the most plausible missing data process that may occur.
When assumptions are being developed for sample size and analysis planning, information can also be inadequate or missing. While we cannot name any specific companies, we will use “Company XYZ” as an example.
Representatives of Company XYZ have data they believe is all they need to show that their product will be successful in a comparison pivotal study, but they do not have any comparator data. Against statistical advice to do a pilot study to obtain real data for study planning on sample size and endpoints, they use what they “believe” from this incomplete data instead of what they could have observed (evidence) in a pilot. Sample sizes are estimated based on this data. When the study starts, many issues arise, such as not being able to show statistical significance or clinical benefit.
No statistician likes to find out that data they have been asked to analyze cannot support research objectives. By nature, we like to investigate data to find answers to questions, not come back and say, “I just can’t know from this data.” Sometimes we find it is impossible to analyze objectives due to the information that was collected. Sometimes this happens due to how the data was collected — and in the worst-case scenario, not knowing why it was collected in the first place.
Innovation Key to Better Clinical Trials
The word innovation derives from the Latin word innovatus, which is the noun form of innovare, meaning “to renew or change.” Our industry cannot fight innovation, so we must embrace technologies and tools that will improve our processes and renew our thinking when planning our clinical trials.
Innovation is not always about reinventing the wheel, but instead, finding better solutions for the wheel. An example of innovation improving the process of interpreting clinical trial data is the growing field of data standards for collection and regulatory submission. It is a human and technological effort. Technology creates tools that assist in the implementation of standards, yet it is the collaboration of experts and stakeholders such as clinicians, academics, regulators, and industry researchers that is essential for success. Furthermore, the specific skills and insights that statisticians have concerning the use of data for obtaining valid evidence of clinical benefit or safety should be considered when forming these collaborations. They can identify early pitfalls in substandard design and data collection, which could hinder the research project, and they can bring critical thinking and objectivity to any collaboration.
To quote an old adage, experience does not cost but it pays. Using an experienced team and avoiding missteps early on in your trial pays off in the end. The result: a clinical trial with the potential for success.