<img alt="" src="https://secure.lote1otto.com/219869.png" style="display:none;">
Skip to content

Group Sequential Designs and Sample Size Re-estimation

In this blog, we speak with Christopher Jennison, Professor of Statistics at the University of Bath, UK. Professor Jennison provides us insights on group sequential methods, the origin of their implementation and the value they have been adding over the years.

Cytel is hosting a new webinar series that introduces clinical fellows, early career biostatisticians, and others interested in clinical research, to some of the more commonly used complex innovative trial designs. In our previous blog, we spoke with Zoran Antonijevic about adaptive design methods.

Join us for a complimentary webinar on June 3, 2020 where Professor Jennison is going to introduce us to the basics of group sequential designs and sample size re-estimation. Learn how to use these methods and understand how they can improve trial design.

Register

Cytel: You have been involved in research on clinical trial design for over 35 years. How did you get involved in Statistics and this research area in particular?

Christopher Jennison (CJ): I was an undergraduate studying mathematics at Cambridge, UK, and I took statistics courses, which I enjoyed. For my PhD, I went to Cornell University in upstate New York. I was studying statistics and found topics such as, survival data, experimental design and sequential analysis very interesting. I had the good fortune of working as an intern in Marvin Zelen’s* Biostatistics group at the Dana-Farber Cancer Institute. This is when I received my first-hand experience in clinical trial analysis. Putting those ideas together helped me define my PhD topic of repeated confidence intervals, a new approach to interim monitoring.

I have tried to retain my mathematical foundations and do interesting methodological work, while I always keep applications in mind. I had a great supervisor, Professor Bruce Turnbull, at Cornell. We first met in 1978, and we have been working together ever since. Bruce is great at spotting important problems to work on and has been well connected with the world of applications. We continue to work together on theory and methods, with a strong bent on applications.

Cytel: What are the methods that you are advancing? Tell us a bit about their history: what is the ethical and commercial value of these methods?

CJ: In my upcoming webinar with Cytel, I shall be primarily talking about group sequential monitoring of clinical trials. Sequential analysis is the business of looking at your data as they come in and seeing when you have enough information to stop the trial. When you are testing a null hypothesis then as soon as you can reject it, you may say, I don't need any more data”.

The first applications of this method go back to the early 1900s in industrial quality control, for example, inspection sampling. There were some very important developments in the mid-1900s. Abraham Wald's sequential probability ratio test was a very nice piece of theory and methodology. However, this still applied to engineering and industrial applications only.

Clinical trials did not really come along until the mid-20th century. Once they were in place, statisticians started looking at the possibility of bringing this methodology to medical studies and clinical trials. Peter Armitage talked about repeated significance testing. The problem is when you look at your data over and over again, there is a danger that you will find a false positive just by looking often enough, and you need to control for that. Armitage showed how one could do that. But these methods were not taken up very quickly, and it was really in the late 1970s that we saw two key pieces of work, a paper by Stuart Pocock and a paper by Peter O'Brien and Thomas Fleming, introducing the idea of group sequential methods. 

In a clinical trial, we do not look at the data every time we get a new observation. Instead, we look every six months, say, and so have a small number of analysis where we may make another decision. These ideas that came around in the late 1970s saw embellishments, such as ways of dealing with unpredictability in the numbers of observations at each analysis. After moving on from my PhD thesis, I worked in these areas.

The benefit of these methods is that you can run your clinical trial and reach a conclusion sooner. If you compare a sequential trial with a fixed sample study, on an average, you may save 20 to 30 per cent of the observations. Sequential trials help reduce costs and the number of patients involved, but finding a positive conclusion earlier is really important too. In the drug development process, where patent lifetime is limited, reaching a decision six months or a year earlier is a big advantage.

Accelerating the drug development process is also very important for patients. If you have a new treatment that has been proved to be effective, you would want to bring it to the market and make it widely available as soon as possible. We are also seeing this now in the COVID-19 area where people are racing against time in the search for vaccines and treatments. As soon as we find anything effective, the rollout is going to be phenomenal as there are patients contracting the virus every day.

Cytel: How did you get excited about these methods?

CJ: At heart, I am a mathematician and a statistician, and I realize that I have been able to use all kinds of mathematics in my work. I would not be so excited if it weren’t for their application. In a research department of mathematicians and statisticians, it is something of a rare event when you can say, "Look, here is something I did and here is someone actually using it." In my career, I went through a progression where I worked on methods and published papers, and then became more involved in the applications of these methods. It is when Bruce and I wrote our book, Group Sequential Methods with Applications to Clinical Trials which came out in the year of 2000, we found that we really had made an impact. The book introduces clinical trials in general, explains how group sequential methods can be applied to a variety of testing problems and different data types, and demonstrates the benefits that can be gained from these methods.  

After our book became well known, I became more involved in discussions with investigators on designing trials and showing how to take advantage of group sequential methodology. Seeing the consequences of that is exciting, when things that you worked on are implemented, and they influence the way medicine develops.

Cytel: Can you think of a memorable success story involving Group Sequential Designs?

CJ: The trial for Beta-Blockers, a popular treatment for hypertension, is one of the first applications of group sequential methods and a great success story. It was set up as a group sequential design including seven possible analyses to be conducted six months apart. It was stopped at analysis 6 out of 7. For this treatment to reach the large number of patients who would take Beta-Blockers six months sooner was a significant achievement.

Cytel: Have there been others who needed persuading and how did you persuade them?

CJ: I see this more as a community-wide issue and not a discussion to persuade an individual. It is important to put relevant information out there and help people see the usefulness of group sequential methods. In our book on group sequential methods for clinical trials, Bruce and I present a coherent picture of this methodology. Before that, a lot of the things that we were talking about were available in a wide collection of papers and journals, using different notation and different views of how to do things. We put those together in a common framework and worked hard to make it accessible.

When the book came out, Cytel was releasing East® 3. We got together with Cyrus Mehta and presented four short courses at conferences and to a company. Since then, I have taught a further 25 short courses for companies and at conferences. I also attend many applied conferences where I get the chance to interact with people and understand the problems that need solving. In my opinion, the question is not how to persuade people, but how to help them understand what is possible and explain the methods to do this.

Cytel: How should these methods be used given the industry’s current challenges?

CJ: A basic benefit of using sequential methods is that it allows you to collect the right amount of data to answer your questions. That alone of course will not help invent a new drug. The challenge is to bring these methods forward in a way that makes people understand and see the benefits. We need to provide the applied statisticians with tools that are necessary for implementation of these methods.

Cytel: Where can a beginner learn more?

CJ: There is a wide range of materials available now. I will be somewhat immodest and say that my book with Bruce Turnbull is a good place to start. It is 20 years old now, but it is quite relevant and up to date on group sequential methods. I believe we did a good job of explaining the basics of these methods and how to use them. You can also find courses and webinars for instruction on running trials with group sequential designs.

One should also look more widely at ongoing developments. The scope of adaptive clinical trial design has really grown in the last 20 years. There are new approaches to old problems, and there are a lot of new methods addressing novel issues such as, enrichment (focusing on a sub-population of patients), seamless trial designs that combine phase 2 and phase 3 trials, multi-arm multi-stage trials. I am sure there is more to come, so my advice is to keep in touch with developments and watch out for those that may be useful to you.

 

*Marvin Zelen was a close friend of Cytel’s and his Oncology Group at Dana Farber inspired the set of questions which eventually led to the founding of Cytel. Read more about Cytel’s early history here.


 

Register for Cytel's Introduction to Complex Innovative Trial Designs webinar series.

register

 

About Professor Christopher Jennison

christopher jennisonChristopher Jennison is Professor of Statistics at the University of Bath, UK. His PhD research at Cornell University concerned the sequential analysis of clinical trials and he has continued to work in this area for over 35 years. His book with Professor Bruce Turnbull, "Group Sequential Methods with Applications to Clinical Trials", is a standard text on this topic and is widely used by practicing statisticians. More recently, he has written with a variety of co-authors on adaptive trial design and over-arching optimization of the drug development process.

Professor Jennison's research is informed by experience of clinical trial analysis at the Dana Farber Cancer Institute, Boston and a broad range of consultancy with Pharmaceutical companies. 

Click here to learn more about Professor Jennison and access his publications.

 

contact iconSubscribe back to top