The Numbers Game

If you have a love of mathematics then a career in pharma, at first glance, may not seem like a dream job, but dig deeper and you’ll find that effective pharmaceutical development is built on maths – particularly statistics. Stephen Pyke has a degree in mathematics and first decided to go into the actuary business, before quickly deciding that it was the wrong choice. After returning to university to complete a Masters in statistics, he gradually became interested in biology and pharmaceuticals. Today, Pyke is Senior Vice President of Clinical Projects and Quantitative Sciences at GlaxoSmithKline, UK, but he is also Vice President for Professional Affairs at the Royal Statistical Society (RSS). At the latter, he helps thrust statisticians into the limelight with the Award for Statistical Excellence in the Pharmaceutical Industry, a prize jointly sponsored by the RSS and the Statisticians in the Pharmaceutical Industry (PSI) group. So, are pharmaceutical statisticians finally getting the recognition they deserve?

You don’t have a science degree. How did you become interested in pharma?

I’ll be honest – I wasn’t that interested in biology at school! But I ended up getting a job at the Medical Research Council in North London where I worked in the laboratory of mathematical biology – and the application of mathematical models to biological systems intrigued me. I worked in a really nice group led by Tom Kirkwood, who later went on to work in the area of gerontology; today, he’s the Associate Dean for Ageing at Newcastle University’s Faculty of Medical Sciences. After that, I got the chance to work at the London School of Hygiene and Tropical Medicine, where I focused on clinical trials, epidemiology, and public health. I was  able to work with Simon Thompson (my boss at the time) and Stewart Pocock, who are both very well known in the field of medical statistics. I loved my time there, but I was also gradually getting interested in the pharma industry – and eventually I took the plunge. The rest is history... It’s certainly a great industry to work in. People come into the pharma industry for all sorts of reasons, but I think there are very few of us who don’t get a sort of warm glow about the end result – treating and preventing disease, or at least mediating the symptoms. It makes you feel really good about your work and its impact.

And the complexity makes it really interesting. The industry is working on everything from human biological targets, through to selecting molecules (small or large), to clinical experimentation, animal studies... and then, when you get into the clinic, there are even more questions to answer. We’re famous as an industry for having agonizingly high attrition rates and yet there are times when you wonder how on earth we managed to pull it off for a single molecule, let alone many. But of course the key is collaboration – collaboration across different disciplines, and amongst industry and academia. 

How important are statisticians in the pharma industry?

Designing and analyzing clinical trials is a fundamental activity of pharma, and we statisticians are essential to that activity. In fact, the basis of clinical trial design is randomization, which is a statistical concept used to randomly allocate patients in a trial so as to enable fair comparison. But statistical expertise is needed far beyond clinical trial design; the pharma industry is based on the generation of data (and it generates buckets of it), and statistics is about understanding data. To truly understand data, you need a statistician.

Statistics can also forecast the future and process the past. Indeed, statistical techniques have been used to develop predictive modeling systems that have transformed the efficiency of the pharma industry. Predictive models allow us to deduce that, if a drug has the expected mechanism of action, then we should see certain outcomes. Comparing statistically predicted results with actual results can give you a degree of comfort (or a clear warning) about a drug before huge sums of money are spent.

Generating a medicine is a complex process that involves thousands of people, and often it’s the people involved at the end of the process who get the glory – the tremendously valuable role played by statisticians is sometimes forgotten. ‘Unsung heroes’ is a phrase we often use – so I think that the Award for Statistical Excellence in the Pharmaceutical Industry is a nice way to give credit to at least some of these people and the fantastic statistical techniques they develop. 

How did you get involved with the award?

The Statistical Excellence in the Pharmaceutical Industry Award is  jointly sponsored by the RSS and the PSI organizations. I joined the RSS when I was at university, but now I’m at a stage in my career where I want to give something back, so I got involved with the RSS Professional Affairs committee, which is responsible for certifying statisticians as being professionally competent. It also organizes various events and activities for professional fellows and, in particular, it supports members who work for organizations that are not primarily statistical in terms of their activities, such as the pharma industry. In fact, pharma-employed statisticians have a significant presence in the UK and many of them are RSS or PSI members. With that background, there was a sense in the RSS and PSI that we should be sponsoring an award to celebrate what’s best about statistics in the pharma industry. We were giving out other awards for journalism and for statistics work emanating from government bodies and non-government organizations, so an award for statistics in pharma was an obvious gap. The Award was born about eight years ago and has been an annual event ever since.

The reaction from industry has been very positive. And personally I think that it’s great for us statisticians to be able to celebrate our field. Statistics are well respected in the pharma industry, but it’s one thing to respect them and another to love them. When I was at Pfizer, I nominated one of the early winners of the award, and it was great to be involved in that way. Last year, the prize was jointly awarded to teams from Pfizer and from GlaxoSmithKline, and I know that everyone involved was delighted about that. I’m a huge supporter of the award and I’m encouraging teams from my organization to put their nominations in. I hope others will too! The deadline for the 2016 award is fast approaching but there’s always 2017, if you think you’re running out of time.

What does it take to win the award?

Nominators are asked to complete a form that describes the nature of the work they’re putting forward. The work is then evaluated by a small committee, made up of representatives from PSI and RSS. What we’re looking for is evidence of impact. As lovely as it is to have done a really thoughtful, clever piece of work, it counts for rather less if it’s just published and forgotten about. We want to see evidence that it has an application in the real world. Ideally, we want independent commentary from others – not statisticians per se – indicating that the work has made a difference; that’s the hallmark of a winner.

One of the winners from last year used a Bayesian approach to determine the likelihood of clinical trial success. Their system permits the incorporation of various soft and hard information – for example, literature data, pre-existing evidence relating to similar APIs, clinical judgements as to whether a drug would work in the way that the developer anticipates, effects of other treatments previously evaluated in the target population – that can be combined objectively and transparently to provide a level of assurance; ultimately providing information that is of practical use to the people who make the critical investment decisions. The judges’ view was that this method was also beginning to make a difference in terms of the way government bodies evaluated those investments.

Another recent winner was a checklist tool to help people design statistically meaningful and reproducible preclinical experiments. It’s intended to support scientists who don’t necessarily have a statistician at their elbow all the time. This is not to say that scientists don’t think about the issues of reproducibility and statistical power, but they’re not usually experts in these issues. This entry got the prize because it was rolled out as a kind of kit, bundled with education and training, and was being broadly adopted for an organization-wide impact. That’s exactly what we were looking for.

Industry evolves rapidly; will it continue to need statisticians?

Undoubtedly! Actually, I believe there is increasing recognition of the need for more trained statisticians to meet future recruitment requirements. A recent report from the UK’s Association of the British Pharmaceutical Industry on skill gaps in the pharma industry identified lack of statistical expertise as one of the biggest needs. The shortfall in statisticians for pharma is exacerbated by the current emphasis on big data and digital devices. The explosion of data associated with these devices, and our massively increased ability to access and integrate the data, are of very limited value if we can’t analyze and make sense of it all. And for big data you probably need not just statistical skills, but also some facility with informatics, computer science and mathematics. It’s hard to find people with that mixture of skills, and that’s another gap in the pharma skill base.

Nevertheless, universities are starting to recognize the need – there’s now an abundance of courses and research aimed at dealing with big data. Pharma may have influenced that to some degree; many companies, including GSK, interact with leading universities by sponsoring PhD students and post-doctoral fellows. Pharma’s connections with academia are getting stronger all the time, including in the field of statistics. But pharma won’t be the only industry wanting those skills, and I’m sure that we’ll be in a fierce battle to recruit the very best.