As a biostatistics consultant for Harvard Catalyst, Anna Modest, PhD, advises scientists how to build biostatistics into translational research–from the start–in order to definitively answer the scientific question presented. What she didn’t anticipate when she was recruited to the role was how much she herself would learn by interfacing with investigators in fields of science vastly different from her own specialty.

Modest is a reproductive epidemiologist and director of clinical research education in the Obstetrics and Gynecology (OB/GYN) Department at Beth Israel Deaconess Medical Center (BIDMC). We caught up with her just before she headed to a farm in Vermont for a family vacation.

Investigators request biostatistics consults on a wide range of studies at various stages. What advice would you offer young investigators looking to up-level their research protocol with a consult?

I would say the earlier you get us involved the better. We’re not just here to analyze data once you’ve collected it; by that time there’s only so much we can do. The sooner you contact us to lay out a biostatistics plan for your study, the more definitive your results are going to be. That’s important not just to individual researchers, but in the larger context of evidence-based medicine.

Many of the things that can go wrong in a study could be prevented if they were thought about ahead of time. When you’re thinking about study design, you have to think about how much data to collect, the kind of data, and how to collect it. If you don’t collect the right information or enough detail, there’s no magical math that’s going make it all better after the fact.

“Biostatistics is not magic. It’s mathematical formulas. You need to put the right things into the formula to get the right answer. If we don’t know what the right thing is, or don’t put the right thing in, we’re not going to get a great product.”

Biostatistics is not magic. It’s mathematical formulas. You need to put the right things into the formula to get the right answer. If we don’t know what the right thing is, or don’t put the right thing in, we’re not going to get a great product.

How did you get involved as a biostatistics consultant for Harvard Catalyst and how does that fit with your day job at BIDMC?

When I joined the Harvard Medical School faculty in 2020 after completing my PhD, I was invited to be a Harvard Catalyst consultant as part of a hospital-wide group that could provide this service. There aren’t many people in the hospital who are trained in biostatistics and epidemiology. Most of us have historically stayed in public health or other areas, so bringing that skill set into a clinical setting is relatively new. I think people are recognizing the importance of having a good study design and statistical analysis plan when they’re putting together their study.

As an epidemiologist, my role is to help with both the design and analysis of the study. Much of my training has to do with thinking critically about how to design a study to answer the right question. What are the things that can go wrong when you design and carry out a study, and how do we prevent them? Or if they weren’t prevented, how do we then deal with them? How do we account for bias, confounding factors, and other important issues that we worry about in designing human studies? This is the focus of most of my work. Studies involving human participants can be challenging.

Much of what I do day to day is very similar to my work with Harvard Catalyst, in terms of helping investigators design their studies and conduct effective statistical analyses, which is a big part of my role in the OB/GYN department. Outside of acting as a biostatistical consultant, I spend most of my time thinking about rare and serious pregnancy complications.

What stands out to you in the last few years of biostatistics consulting?

I have actually learned a great deal during that period that benefits me professionally, though I think the idea was that we’re supposed to be helping other people.

I’m a big believer in the idea that there’s always something to be learned, no matter what you do, but I’m not sure I expected to learn as much as I have. My focus has been reproductive epidemiology in an OB/GYN department, and I’m not clinically trained, so I don’t know a ton about other clinical areas.

“What is the optimal this or the optimal that? In the end, it is based on human thought and decision-making–all of those things that encompass the art and science of medicine–and we don’t have numbers for that.”

I’ve learned a great deal about cardiology and sleep apnea. I’ve learned about dermatology, particularly in the transgender population. I’ve worked with several researchers at Fenway Health, which has been really interesting. I’ve had the opportunity to use large datasets that I wouldn’t otherwise have had, such as cancer datasets. I definitely learned about other areas of the clinical world and how we can best study them.

I’ve also been in somewhat humbling situations where I learned that maybe there’s no great way to answer the question with the data available, despite how interesting the question may be. That’s always a tougher conversation and lesson, but sometimes we just don’t have the right methods to give a definitive answer.

How critical is this kind of data crunching to moving translational science forward?

I think it is very critical. There are so many nuances to healthcare and medicine that you need these large numbers to be able to answer some of that.

On the flip side, I think we need to be very careful when interpreting our work to understand the limitations of these large data sets. Medicine is very much both a science and an art, and the things we can’t answer no matter how hard we try are related to the art of medicine. There’s no way to capture and quantify the art of decision-making that goes on in healthcare.

I’ve run up against several investigators who are interested in how decisions are made within clinical settings, such as how people are triaged or assigned to different rooms within the hospital. People want to know: What is the optimal this or the optimal that? Many people have tried, and people have come very close with decision analyses and clinical-care roadmaps, but I have yet to find a dataset that can really answer those questions. It comes down to an individual decision by somebody who is looking at the patient in front of them with the information they have and making a clinical judgment.

In the end, it is based on human thought and decision-making–all of those things that encompass the art and science of medicine–and we don’t have numbers for that.

Outside of biostats, what do you do for fun?

I like spending time with my family, and going to farms to hang out with animals. I also love baking desserts. I’m not a huge fan of straight-up cakes and cupcakes, but prefer baking cookies or this chocolate mousse layer cake I’ve been experimenting with. In some ways, I guess baking is not dissimilar from research, right? You have a very precise formula, a very precise recipe, then you kind of put it all in, cross your fingers and hope it works out. So that probably tracks.