Three Challenges for the Non-Clinical Biostatistician

Posted on:

Statisticians are often categorized into two groups: clinical and nonclinical. The nonclinical statistician encompasses most areas where studies were not involved in a clinic. This wide variety of work means that nonclinical biostatisticians have a wide variety of challenges to face. Of all these challenges, three main ones were discussed at the 2015 Nonclinical Biostatistics Conference (NBC) at Villanova, PA this past fall. While no finite solution was reached, several statisticians presented possibilities to overcome these obstacles.

One of these challenges is assay development. Similar to a clinical trial study design, a nonclinical biostatistician uses different assay structures to reach a specified goal. The next hurdle is margin and threshold determination, which includes parallelism through equivalence testing. The third challenge faced is translating the results from statistical language into scientific or laymen’s terms.

Mathematics is a science and statistics is an art. Developing an assay requires both an analytical and creative mind. The end goal must be addressed throughout the entire development process in order to ensure that the data collected at the end will be practical in providing the answers desired. Many questions arise relating to the number of assays needed, the placement, number of operators, the days on which to run the assay, and the materials used in each assay to name a few. All these factors can create a large amount of variability if not carefully designed. A process simulation study was mentioned at the NBC where a Bayesian Generalized Multivariate Linear Mixed Model was implemented so that the measurement error was known at every step which reduced the amount of unknown error. While this method has not been perfected, the art of predicting where and how much variability will be present in an assay is a challenge non-clinical statisticians are striving to overcome.

Similar to referees for sporting events, a statistician’s margin for error is approximately 0. Determining a margin for parallelism, equivalence, or anything else is vital to the project’s future, yet there is not one concrete process. Throughout my experience in determining acceptable margins, each is evaluated on a case by case basis. This causes conflict with scientists who are used to other statistical methods always being one specific way. As a majority of statisticians spend their days deciding which method is “more right,” several non- clinical statisticians have their own thoughts on the method to use and were presented at the NBC. Some of these methods include the Limenatani Approach, a sample size and variance adjusted margin, or a fixed margin using the ratio of means. At BioStat Solutions, Inc., a method we recently implemented was a slight variation on the two-one sided test (TOST) approach. The log ratios of the slopes of vaccine reference material were used to set the margin of equivalence using the 2.5% and 97.5% quantiles and then back transformed. Using this equivalence test, the power was summarized using a statistician’s view of meaningful results. The final determination of an interval was a discussion between statisticians and scientists using both the statistical input and scientific input to decide what interval provides enough information. This brings us to the final challenge.

Meeting with a statistician can often feel like traveling to another country and not understanding the language. This can be intimidating and may lead scientists to avoid talking to statisticians or letting statisticians get away with whatever they say. One of the keynote speakers at the NBC discussed the importance of this communication and provided a two-fold solution. First, statisticians need to increase their ability to translate statistics to English and in terms of the specific study. Individualizing the meaning and methods behind the statistics will allow scientists to not only trust a statistician, but learn more in the process, creating a better final result. Second, more emphasis on statistics in the education system will allow future scientists and leaders to understand the importance of statistics as well as their implications in their field. Accomplishing these will strengthen the mutualistic relationship between statisticians (non-clinical and clinical) and scientists.

The Nonclinical Biostatistics Conference allowed non-clinical statisticians to discuss their challenges. As history has shown, when great minds meet and discuss openly, new solutions emerge.