Biostatistics and Clinical Trials: An Interview with BioRealm's Carolyn Ervin, PhD

In this interview with Carolyn Ervin, biostatistician, and co-founder of BioRealm, we discuss how to get the most value from experienced biostatisticians, developments in the industry and her associated concerns, a few of her favorite tools when working on clinical trials, and project management secrets.

Conor Ryan: When do clients usually come to you about clinical trials?

Carolyn Ervin: Anywhere from the very early conceptual stage all the way through the time when they need to submit their final reports to the FDA. We love working on "greenfield" projects, where there are no constraints imposed by prior work, but we also love helping clients that come to us later, even if they come to us after things aren't going well, they get overwhelmed, and the project needs to be rescued.

CR: When is the best time for biostatisticians to get involved?

CE: The earlier the better! We can provide more value and guidance when we're involved from the very beginning. Ideally, we would be there right from the start, at the conceptual stage, but if not, we should be brought in as soon as possible.

CR: How has biostatistics changed in the last decade?

CE: Consulting companies like BioRealm are much more popular. It can be very difficult and expensive to recruit, employ, and retain talented and experienced biostatisticians as full-time staff. BioRealm provides much more value, especially since we're a one-stop shop, providing not only biostatisticians, but a full suite of experts and services.

Another big change is that projects are more structured and automated, especially since we use biomarkers much more now.

[For more information on how biomarkers have changed clinical trials, see our interview with James Baurley.]

CR: What are your favorite software tools as a biostatistician for use on clinical trials?

CE: For many problems, the old favorites, like SAS, R, and SQL, are still good. I'm increasingly seeing more use of nQuery for sample size calculations.

Carolyn M. Ervin, PhD

Carolyn M. Ervin, PhD

When I'm listening to a potential client though, I'm listening to what they're saying, and what they're not saying. I'm not thinking about tools at all—it's way too early for that. People without proper biostatistical training, and inexperienced biostatisticians, often fall into the trap of learning a few tools and then trying to use them to solve everything. Their problem solving abilities are actually limited because they're always thinking about the tools first.

We have many more tools available to us today, not just because new tools have been developed, but even new approaches. We also have so much more computational power and storage available as well. This really empowers us to look at each problem individually and use the best tools for the job. That's one reason why today we'll often have complex workflows using many different software tools.

CR: Do clients ever request that you only use specific tools?

CE: Sometimes. When you really break that down, it turns out that they actually don't care what tools we use, as long as we return data in a specific format. We'll then use the tools we believe are best for the job, and when we give the data back to the client, in the format they need, we'll explain what we did, and what tools we used, and why we believed those tools were the best choice. Based on that recommendation, sometimes clients switch to those tools. That's just one of the reasons we believe transparency is so important: we don't just want to get the project out the door, but also leave the client in much better shape for the future. That's our real goal.

CR: Do you see any need for caution with some of the new tools available today?

CE: Fifteen or twenty years ago many companies were still using products like Microsoft Excel to manage clinical trials—and some still are. It's very easy for someone that isn't trained and experienced as a professional biostatistician to use a tool like Microsoft Excel. If you're not trained and experienced, it's too easy to use a piece of software and believe that if there aren't any error messages your report to the FDA must be correct.

We've also seen this with people using JMP, a statistical software tool developed by SAS. JMP can be a very powerful tool, but in the hands of someone that isn't a trained and experienced biostatistician, it can be very easy to quickly generate bad results.

The biggest reason we see this happening is trying to save money by using relatively inexpensive tools like Microsoft Excel, and JMP, instead of a much more expensive tool like SAS. Companies also try to save money by using staff that aren't trained and experienced biostatisticians. Both strategies are penny wise and pound foolish, since those projects often end up needing to be rescued, or, even worse, people and organizations end up damaging their reputations and careers when mistakes are only noticed when it's too late. Unfortunately, we've seen that happen as well.

CR: How do you make sure clinical trials stay on track?

CE: Communication! Participating in all of the project meetings is a good start, but we also have to stay engaged with every group and member in the project, including all of the different managers, scientists, physicians, and programmers.

We also do a lot of testing, to make sure the results we're getting look as expected. We're always looking out for potential problems, especially subtle problems. Our experience really helps prevent those, and it's so much easier to prevent problems than correct them. Of course, that also saves time, money, and reputations.

CR: How do you avoid "scope creep" on a project?

CE: One of my favorite ways to avoid scope creep is to use dry runs, where we set deadlines and milestones for the entire project team at various points. That really limits scope creep, since it keeps everyone focused on our stated goals, and making sure our results are accurate at each point along the way.