Statistics has been part of the work of the MRC for almost our entire history ― the MRC Biostatistics Unit is 100 years old this year. Here Dr Howard Thom, who did his PhD at the unit, describes how important it is to remember who statistics is for: patients.
Psoriatic arthritis is a combination of two very unpleasant conditions: the rashes associated with the skin condition psoriasis and the painful inflammation of joints found in arthritis.
How and why some patients rapidly deteriorate ― making it more and more difficult to complete everyday tasks such as preparing food, making a bed, or even sitting in a chair ― while others remain stable, is of great interest to doctors and, of course, to patients.
Medical researchers have tried to answer these questions by building statistical models, called multistate models, to predict the progression of the disease.
Multistate models are used to understand the progression of a wide range of conditions, from depression to heart disease, so the question of whether or not your model is the best impacts on almost the whole field of medical research.
But how do the researchers know these models are true? Are they even sensible? Are there better models that could be built? It’s in answering these questions that the work of statisticians, and my research in particular, enters the picture.
I became interested in comparing various models during my PhD at the MRC Biostatistics Unit in Cambridge. During my time there, I read many papers by a group of researchers in Bordeaux, France, led by Daniel Commenges.
The group had made some breakthroughs in the comparison of multistate models for disease progression by gaining a deep enough understanding of the statistical theory to directly measure which model’s predictions best matched reality. However, they had only looked at a few simple cases.
I very much wanted to extend their methods to other more complicated disease progression models such as those used in psoriatic arthritis. An MRC Centenary Award came at just the right moment, as I was finishing my PhD, to allow me to extend my research to figure out the details. The award also provided funds for me to spend two months in Bordeaux and collaborate directly with the group.
I found Daniel and his department to be incredibly welcoming ― and fiercely intelligent. Despite being such a strong statistical theorist, Daniel was very keen for me to understand that the design of statistical models should always be driven by the real-world questions that are most important to doctors.
I learned that it doesn’t matter how realistic and complicated your model is if it doesn’t answer the important questions well. This helped me to better understand his work but also to become a better medical statistician. It taught me that as a statistician building models, I should always remember the point of the work ― in this case to help doctors make decisions about treatment.
With this in mind, Daniel and I developed our methods to find the best model for progression of psoriatic arthritis. We made sure it could answer questions posed by doctors who knew which levels of disability mattered most to patients. As well as eventually guiding treatment choice, our research could be used by doctors to help patients plan for the future by knowing when they’re likely to need to make lifestyle adjustments.
We gained valuable insights into how the disease progressed and now our work can be built on to further understand the type of patients who will suffer further deterioration and those who will stabilise.
Together with my supervisors Chris Jackson and Linda Sharples at the Biostatistics Unit, we have built a software package so that statisticians the world over, working in dozens of diseases, can apply our methods and come to a clearer understanding of how and why diseases progress.
The MRC Centenary Awards were provided to the very best MRC-funded early-career researchers to give them extra time and resources to build on their achievements and learn new skills. £12m was made available for the awards, which marked 100 years of the MRC in 2013.
Learn more about the work of the MRC Biostatistics Unit in the Autumn 2014 edition of our Network magazine.
* This article has been originally published on insight.