Ronald Geskus: Sophisticated biostatistics for complex clinical research
The role of biostatisticians in clinical research is to contribute to trial design, by calculating sample size for example, and to help draw correct conclusions from the data, discriminating important information from noise. They are instrumental in the translation of a practical problem into a statistical model, and the translation of the result into practice.
My name is Ronald Geskus, I currently work at OUCRU in Ho Chi Minh City, Vietnam. We are a group of 6 biostatisticians and I am the head of that group. A biostatistician tries to prevent incorrect conclusions to be drawn from the data analysis. We are like the controllers of all of the quantitative analysis data done at OUCRU.
In a clinical trial we want to see whether some new treatment works better than the existing treatment, or maybe works better than no treatment at all, and it may be that the new treatment also has side effects or maybe it doesn’t work at all. Another thing is that a clinical trial in general is very expensive; there are lots of regulations around clinical trials, not without reason, but this requires a lot of logistics as well. So you don’t want to include too many patients, you want to include just enough patients to show that the effect works better. That is done in sample size calculation, finding out how many patients to include, and that is a typical job of a biostatistician.
Clinical trials are a basic example of an explanatory model; you want to explain the effect of the treatment. An observational study is different from a clinical trial: in general the sicker patients are more likely to receive a treatment. In both cases the purpose is explanation: what is the effect of the treatment? So explanation and effect or efficacy is basically the same. Whereas in a prediction model you want to find factors that help predict the outcome.
Research in biostatistics is important because nowadays so many data are collected, and it’s very important to do a proper analysis of these data, and this also holds for medical fields. Many more data are collected nowadays and this will generate a lot of noise. A lot of data is not relevant so it has become more important to find methods that are better to discriminate between signal - important information and noise.
As statisticians, we work at the basic part of the scientific research. We don’t work at bedside, but if you take translational medicine literally, it’s about translation of results into practice. Statisticians also play a role as they know very well the structure behind the model and they are in general very good at explaining results. In general what I like about doing biostatistical research is the translation of a practical problem into a statistical model, and then the translation of the result to practice.