“Historically, statistical tests were applied to differentiate between genuine effects and findings that could be caused by chance. This is still true in confirmatory clinical trials where pre-specified hypotheses are tested – for example, if a new treatment is associated with longer survival than the current standard of care,” explains Dr. Jan Rekowski, Senior Statistician in the ICR’s Clinical Trials and Statistics Unit (ICR-CTSU).

“But in hypothesis-generating work, where we search data for relationships or patterns that can subsequently be tested, the focus nowadays is more on estimation and precision – for example getting an initial idea of how many patients a new treatment appears to work for, and how confident one is about this initial estimate.”

Jan feels statistics often plays an under-appreciated role in science.

“The need for statistics doesn’t just arise at the analysis stage. Defining the research question and setting up an adequate experimental design involves statistical input as well, as does the definition of study endpoints, pre-planning appropriate analyses, and the timing of these analyses. Poorly designed studies lead to the waste of resources and don’t do justice to the patients taking part in the trial.

“Getting the statistics input right, early on, ensures we can help set up robust studies that will produce meaningful results.”

Among other trials, Dr. Rekowski is the statistician for the PEARLS trial, a combined phase II/III trial that opened in June this year and investigates whether moderately fractionated extended field intensity-modulated radiotherapy is safe and can improve outcome in prostate cancer patients.

A crucial role

He explains that in research today, there are more and more complex and sophisticated approaches being implemented, that ask for more intensive statistical assistance. However, statistics has always had an important role to play.

“In my humble opinion, statistics has always been crucial for cancer research. Without statistics we would not be able to perform any cancer research at all – one would be left with a bunch of data that no one would be able to summarise or analyse in a meaningful way.

“The main question is whether statistics is being acknowledged properly as an integral part of medical research, and the statistician as a research partner on eye-to-eye level with other researchers instead of only being seen as the person performing the analysis. Although I feel that improvements have been made in this regard in recent years, many young medical statisticians still prefer to work for well‑paying pharmaceutical companies instead of pursuing a career in academia.

“Better incentives are required to attract these highly-talented people to our cause – both academically in naming statisticians as co-investigators on grant applications and sharing lead and senior authorships, as well as better financial recognition.”

Getting into the field

Dr. Rekowski explained how he embarked on a career in statistics.

“It was misconception to be honest. I thought it would be less theoretical than mathematics and would offer options to work in sports. The latter is true though as the toolkit of statistical methods taught in university can be applied to data from essentially any field – I have friends from university working for banks, airlines, car manufacturers, sports data agencies, consultancies, and one of them even works for an archaeological institute.

“You would only need to set a focus relatively late in your studies and this flexibility may have been my initial motivation for studying statistics as a young indecisive man. Nevertheless, as time passed on, I quickly became more interested in medical statistics and eventually found my way to cancer research.” 

But, if someone were actively considering a career in this field, what advice does he have on the type of training they would need?

“One should have a degree in statistics, medical statistics, or a related field. There are some opportunities for people with undergraduate degrees, but most statisticians in medical research will at least have a master’s degree.

“The exact schedule varies by university, but training typically includes fundamentals in mathematics (in my case mathematical analysis and linear algebra), computer sciences (with a particularly focus on databases and statistical programming), and a number of statistics courses that introduce descriptive statistics, statistical tests, and regression models. These topics would be covered by the bachelor’s programme.

“The subsequent master’s programme would then focus in more detail on the theoretical background of the statistical methods introduced in the bachelor’s programme and would allow for further specialisation. My training saw me take courses in clinical trials, epidemiology, Bayesian statistics, and multivariate approaches to name a few.

“The broad spectrum of base knowledge opens the door not only for trained statisticians but also for mathematicians, physicists, computer scientists, and psychologists, who would then undergo additional training in, for example, clinical trials or epidemiology depending on their future field.”

Learning to speak across languages

For Jan, the interdisciplinary nature of science is one of his favourite things about his job.

“I am fond of languages and we all speak a different language depending on our background and training – statisticians, biologists, trial managers, physicists, chemists, bioinformaticians, clinicians, data managers, and so on. It’s really rewarding when these different backgrounds come together to find a common language to communicate with each other to successfully answer a specific research question.”

Ultimately, the role of the statistician is more than just statistics – it involves helping to plan, conduct, analyse, and publish an experiment, study, or trial: “With statistics being an important part of almost every phase and type of research, statisticians should be involved in a project from start to finish.”