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

Monte Carlo Simulations for Patient Enrollment: A Presentation

Enforesys: Learn More Recently, we published an interview with Chris Conklin, the Director of the Center for Feasibility Excellence at Pfizer. During the interview, Chris spoke about how his team navigates the complex terrain of trial planning and patient recruitment, and achieves those high flying enrollment milestones for each and every trial. His key message was to utilize modern methods in data-driven feasibility studies, augmenting historic and site-level data with new techniques in forecasting. 

Since our interview, Chris gave a talk at the annual SCOPE conference, in which he divulged a few more tips on obtaining consistent patient enrollment figures. An important feature was the use of Monte Carlo simulations, a popular tool from industrial and business operations, which is now gaining popularity amongst clinical operations specialists.

Monte Carlo simulations are easy to implement with the right tools, and yet can achieve target enrollment with 99% confidence. You can find below, a simple explanation of how this method works. Chris's slides (attached) contain a case study.    

What is Monte Carlo Simulation: 

Imagine that at the start of feasibility planning, you have quite a lot of historic data about your potential trial sites. You might know, for example, that certain sites are very good at achieving target milestones in particular therapeutic areas, or that there are region specific benefits and challenges to using a given site. We often use this information to try and make informed guesses about how enrollment might look.

What Monte Carlo Simulation aims to do is to give a quanitifiable probability of how likely your guesses are to what will actually happen. Taking all of the data available, enrollment experts can create a model of inputs and outputs. Inputs include things like site-level data, information about region and country, and other pertinent factors affecting patient enrollment. An output might be the likelihood that a particular site enrolls 200 patients within 5 months. 

Essentially, if you can take these inputs and simulate what would happen in 1000 cases, then you can obtain a probability of what will likely happen in one case. For example, suppose you want to determine how many months it will take to randomize the last subject (i.e. calculate the time from trial start to LSR, as in the figure above). When provided with all of the input information, simulation software like Cytel's Enforesys can give you a plot like the one above. 

A close examination of the plot shows that we have reason to believe that there is a 95% chance that enrollment will complete in 25 months. This is because 95% of the one thousand simulations completed before this time. Given this information, it is possible to reformulate all of the information in the chart above into a cumulative probability distribution like in the one below.

The figure above represents similar information to that in the subplot, but in a way that stresses the relationship between the cumulative probability that a trial will complete by a certain date, and the date of LSR for a trial with 690 patients. The figure below shows clearly that there is a 70% probability that the last subject will be randomized by July 12, 2015 for at trial that began in May 2014. In fact, it also shows that there is around 40% chance that the last subject will be randomized by the beginning of July 2015, and a 90% chance that enrollment will complete by the end of July. 

According to Chris Conklin, this information can be crucial to correlating clinical operations with business needs. Says Chris: 

"We might want to be aggressive and select a target that we have a 10% chance of achieving. The difference here is that we can communicate that to the organization so that it understands it’s an aggressive target. Everyone is going in with the same expectations. We also have great insight into the conditions that need to be present in order to achieve that target and the levers that we can pull to make hitting that target more likely. If there isn’t a compelling business need to be that aggressive, we can choose a more conservative target."

Related Items of Interest

 Data-Driven Trial Planning: An Interview with Pfizer's Chris Conklin

View Conklin's Slides



contact iconSubscribe back to top