What Type of Biostatistical Expertise do Biotechs Need?

Biotechs are often led by scientists and physicians who are in the leading edge of their field. You wouldn’t try to teach Mozart how to create a symphony, or Michelangelo how to paint. Likewise, CROs should not try to 'out-expert' experts in their own field, but rather they should work to provide sound advice in areas where these experts may not be as strong.

Biostatistics Expertise

An area that many biotechs often have not yet built up a level of expertise in is biostatistics. Most physicians have some level of statistical knowledge, and some biotechs may have even hired a junior-level 'eager-to-prove-themselves' biostatistician. But often, although likely quite smart and knowledgeable when it comes to statistics, that junior-level person may not have the background and experience in complex study designs and the types of questions that regulatory authorities may ask at a later date. This is why it is often advisable for biotechs to reach out to their partners (e.g. their CRO) for their assistance in this area. Some CROs, such as Linical, have a team of well-established biostatisticians with robust clinical biostatisical experience.

What Questions Should You Expect?

What experience, you may ask? Often, it is not knowing the answers to every question that demonstrates your knowledge and experience, but rather it is knowing or anticipating the questions that will be asked of you so that you can plan and prepare for a response. These experiences often come from FDA/EMA submissions, or regulatory advice meetings. So, when you (the biotech company) are discussing your study design with a biostatistics expert, you should expect to see questions such as:

• What is the over-reaching question (more specific than the study objectives) you are trying to answer with this study?
• If there is more than one question you are trying to answer, how would you categorize, weight, and/or rank them?
• What is the basis for your sample size determination, or what is needed to be known or assumed to properly determine the required sample size? (There are more sample size related questions, which we plan to discuss in a future article).
• What is the population of interest?
• What contributing factors and intercurrent events should be considered in your analysis methods?
• What multiplicity issues do you need to take into account (e.g. multiple primary endpoints, key secondary endpoints, interim analyses & whether they are blinded or unblinded) via alpha adjustment methods?
• If you are planning an interim analysis, what is the purpose of it (not just because you want an early look at the results), how many interims, and when to conduct them?
• Are there any potentially unblinding issues coming from this study design?
• Choice of endpoints, timepoints for those endpoints, frequency collected, and how they are to be collected.
• How do you plan to deal with missing data, and are there ways to minimize missing data, or to retain patients in the study?
• Is stratification necessary, and do you want a central-based, center-based, or other type of randomization?

Final Thoughts

The purpose and goal of all of these questions is to give your study a better chance at being able to meet the goals of the study, with minimal regulatory dissension on your design or methods. So, you should expect many, if not all, of these questions and more, to be asked by your friendly neighborhood biostatistician.

Author:
Jonathan White,
Senior Director, Statistical Strategy & Key Accounts,
Linical