Avoid These Common Biostatistical Flaws When Conducting Clinical Trials

04 Feb Avoid These Common Biostatistical Flaws When Conducting Clinical Trials

Biostatistics is involved in every step of a clinical trial, including study design, protocol development, data management, analyses, and reporting of trial results.

Despite this wide use, biostatistics can sometimes be misunderstood or misinterpreted. This can lead to flaws and shortcomings that have a negative effect on the validity and reliability of study data.

In this article, Dr. Michele Shaffer, Co-Director of the Institute of Translational Health Sciences’ Children’s Core for Biomedical Statistics, shares her thoughts on the most common statistical pitfalls she observes and how to avoid them.

1. Failure to devote time and resources to the planning and study design phase.

I think the biggest mistake people make is failing to devote enough time to the study design up front. This includes finding people with the appropriate expertise, which might include a statistician to help with the study design.

There are so many components of the design. How do you select the patients that you are going to target? How do you select the appropriate endpoint? What is your control group? I tell people that a statistician is capable of providing critical care, but we would rather provide preventative medicine.

If there are preventable errors, we want to design them out at the beginning to get started on the right foot. It is about really making sure that the design is correct at the start.

How to avoid this: Give yourself enough time to design the study. Also, include the whole study team when you are making big decisions. If there is a statistician and a data manger involved, it is helpful to have them in those early design meetings. If you have already made decisions, it is much harder to convince people to change. If you don’t want to spend a lot of time redesigning, then having the whole research team involved in those early meetings and decision making points is important.

2. Failure to monitor the data in real time.

We find that people often wait until nearly the end of the study to begin looking at the data. This typically occurs when there are not specific safety rules in place or other reasons why the data would be monitored.

If, at the end of the study, you discover you have missing data, there may be less of a chance to resolve it. Trying to go back and find out why there is missing data can be challenging. It may have been years since the patient completed his or her part of the study.

If patients decided to drop out, there may have been an opportunity to ask them why. It may be something completely unrelated, for example they were moving. Or they may have felt that the treatment they were receiving had side effects. If you can get that information in real time, we have a better chance of understanding the trial results.

How to avoid this: Using the appropriate tools is the easiest way to monitor the data. For example, ITHS offers REDCap, a free electronic data capture tool. We also have the data visualization software Tableau.

If you collect your data in a spreadsheet, there is a lot of time involved to manipulate the data and look at it easily. But REDCap has some automatic reporting functions. Tableau allows you to make graphics of recruitment or how much missing data you have for a particular outcome. It is so much easier because it can be programmed from the start.

3. Failure to report the trial results completely.

When we are reporting clinical trials, there can be limitations or weaknesses of the study. For example, if it was difficult to recruit, you may have had to expand some of the inclusion criteria in order to be able to complete recruitment. But that may complicate interpreting the results of the study. That can be a limitation of the study that should be mentioned.

Other things, such as what assumptions you are making when doing the data analysis, are important to understand. If there is missing data, what assumptions are you making about the missing data when you conduct the analyses? Sometimes the information is not provided. When you don’t provide that information, the reader will fill in what is missing. They actually may jump to conclusions. It is helpful to make sure that information is recorded completely.

How to avoid this: Recommend that your research team keep a lab book while conducting the study, including the analysis phase.

There will always be some questions. The research team may have a question about how the statistician is conducting the analyses. I try to keep track of those questions. If the personnel involved in the study have questions about what is being done, then the manuscript reader will likely have the same questions.

One of the ways you can make sure you are reporting completely is to provide the tough questions your team faces, as well as the answers. Another way is to have your manuscript reviewed by someone who is not close to the study before you publish it.

Looking for biostatistics support?

ITHS facilitates biostatistics collaboration at every stage of the research study, from the initial design stage to the final reporting phase. Free initial consultations are offered to provide expert advice on topics such as study design, data collection and monitoring, and analysis methods.

Learn more about ITHS biostatistics services.