Fifteen percent of clinical trial sites never enroll a single participant. And the sites that do find participants often take years to complete enrollment. Almost half the time spent on bringing a drug through clinical trials is during the enrollment phase. This causes serious delays in getting potential new drugs to patients who need them now.
Amgen’s solution? Follow the data.
A highly collaborative, cross-functional group of data scientists, engineers and analysts at Amgen have teamed up to create a tool that can sort through and analyze vast amounts of data to find potential clinical trial sites that are most likely to complete a study more quickly through faster enrollment of patients. The project is called the Analytical Trial Optimization Module, better known as ATOMIC. It leverages machine learning (ML) to pull hundreds or even thousands of pieces of data related to clinical trial sites and determines which data is most relevant to high-enrolling sites. It then puts out a ranked list of sites, predicted enrollment rates at those sites and other relevant data about the country, site and related investigators.
“We wanted to be able to design clinical trials to be faster and have a higher likelihood of success,” said Matt Austin, executive director of Data Sciences in the Center for Design & Analysis (CfDA). “ATOMIC allows us to make predictions in an informed way and build a ranked list of sites that may perform best.”
Amgen’s clinical trial teams have been able to sort through parts of this information successfully, but site selection has traditionally been a manual, time-consuming, inconsistent process that isn’t always able to follow all the data available. Often it is based on the trial team member’s experience with a site or investigator, combined with user-directed computer research. There is no centralized data source and no consistent method for analyzing that data.
“There are only so many parameters that the human mind can calculate,” said Scott Skellenger, vice president of Information Systems. “And often what happens is you’re selecting sites based on human relationships, past experience and anecdotal input,” he added. “ATOMIC represents a new class of technology that's involved in using both leading-edge data and predictive modeling to be able to make choices that are more difficult for people to make on their own.”
Improving, accelerating and diversifying trials
ATOMIC can help identify key drivers of enrollment, leading to the prediction of the most successful sites. By integrating so much disparate data, ATOMIC also provides a single, comprehensive, automated information source for clinical trial team members.
“We are truly at a historic moment, where we have a confluence of unprecedented access to data and advanced analytics,” said Rob Lenz, senior vice president of Global Development. All this data calls for new analytic approaches to predict its value and improve decision-making, said Lenz.
The ATOMIC project team has been working with Amgen’s Global Development Operations (GDO), Center for Design & Analysis (CfDA), Center for Observational Research (CfOR) and Representation in Clinical Research (RISE) teams on ways to increase the diversity and improve the representation of clinical trial investigator and participant populations. They have pulled anonymized and/or aggregated data such as racial and ethnic demographics and geography about clinical sites from Amgen and other sponsors’ trials and aggregated it into a scorecard. The data are then incorporated into ATOMIC’s pipeline for use by any team looking to diversify or increase representation in a trial. This may support the enrollment of a patient population that is more closely representative of the real-world population typically afflicted by the disease being studied, including race, ethnicity, sex and age. It is also important more broadly across Amgen to help improve patient access to our medicines.
Many kinds of data can be fed into ATOMIC to help inform site selection, such as anonymized real-world data (RWD), which includes vast amounts of data derived from electronic health records (EHRs) and medical claims. These data can be leveraged to better understand where populations with certain clinical and demographic characteristics seek care. This information can be included with other data types in the ATOMIC platform. For example, the ATOMIC team incorporated RWD that included diversity data on providers and their patients in the RISE diversity scorecard.
By examining RWD, the ATOMIC project team and CfOR were also able to identify trial sites likely to be found near clusters of patients with high Lipoprotein(a), or Lp(a) levels. Lp(a) levels are associated with risk of cardiovascular disease and levels differ between some populations. Using these data can help site investigators, who are testing an investigational Lp(a)-lowering drug, screen 50% fewer patients to find one participant with elevated Lp(a) for the trial.
A collaborative, flexible model tailored to users’ needs
ATOMIC was first tested as a pilot for an ulcerative colitis trial and has since been used for site selection on trials to test drugs for cardiovascular disease, atopic dermatitis, gastric and lung cancer. By the end of 2022, ATOMIC will be able to support most late phase studies at Amgen.
When the ATOMIC team begins the site selection process for a clinical trial, they work closely with the clinical trial teams, who make sure the data is curated for their trial’s specific protocol. That means ATOMIC must be a flexible model that can incorporate new and different data as it comes up. The team then collects user feedback and further refines the model, which is part of the agile principles the team applies to how they work.
“A lot of cross functional planning and partnering goes into making a clinical trial successful, drawing on expertise within and outside the company,” said Sheryl Jacobs, vice president of Global Development Operations. “ATOMIC allows us to streamline that process by finding sites, including those “hidden gems” that are most likely to enroll participants quickly and efficiently.”
Ultimately ATOMIC’s streamlined results are meant as a guide to the clinical trial teams – more to support decision-making than make the actual decision, said Skellenger. “At the end of the day, ATOMIC is there to help people take this data and couple it with their experience and judgement,” said Mike Zahigian, senior vice president and Chief Information Officer.
And ATOMIC will continue to evolve, added Zahigian. “It’s not like you build ATOMIC, you use it, and it applies everywhere all the time. It’s not a one-and-done project,” he said. “It’s an ongoing capability and that’s true of most ML work, because the world keeps changing on us.”
Learn more about clinical trial innovation at Amgen.
For more about careers in R&D, visit careers.amgen.com.
