First installment in a four-part special report

By Deborah Borfitz

June 24, 2019 | Study sponsors are experimenting with strategies to improve the site selection process, leaning more heavily than ever before on evidence and machine learning techniques, and each other, to solve the problem child of clinical research—the low or no enrollers. So far, there are no clear winners.

The statistics have long been grim by most every site-related yardstick, including patient recruitment and retention rates and investigator turnover. Under the best of circumstances, finding volunteers for trials is challenging because of rising protocol complexity, says Jim Kremidas, executive director of the Association of Clinical Research Professionals. "You can design a protocol where there are literally no patients to enroll because they’re not out there."

Ironically, several positive developments for the population at large—personalized medicine and immunotherapies—are contributing factors to dismal trial-related metrics in many disease areas, says Laura Galuchie, director of global clinical trial operations at Merck. Modern-day protocols are "operationally more similar to rare diseases than protocols of the past." More studies also look at root causes of disease rather than symptomatology.

The "brilliant clinical minds" writing study protocols may not have a real-world understanding of operational feasibility for both sites and patients, says Lorena Gomez, director of global study startup and essential documents at Allergan. Private practice sites may not be appropriately staffed for round-the-clock blood draws, and Alzheimer's patients may not have the mental stamina to spend hours completing outcomes surveys without an impact to the data, she cited as examples. When procedures are critical to the success of a trial but impractical for sites or subjects to complete, sponsors need to be prepared from the get-go to offer support, be it a supplemental budget or additional staffing.

It would also help to remember that the practice of medicine isn't the same everywhere in the world, or even within the same state, says Kremidas. Regional differences exist even when it comes to what type of insulin is considered the "standard of care" for common conditions such as diabetes.

The trend among sponsors to select sites based more on evidence than the opinions of people on the ground is overall a good one, Kremidas adds. A certain number of key opinion leaders will always be involved because their support is needed for the study's success. More problematic is when choices are routinely made based on site monitoring capacity in a given geography simply to keep hired staff occupied, he says.

Data Aplenty

When it comes to selecting sites, "past performance is a pretty darn good indicator of future success," Kremidas says, citing an analysis done more than a decade ago at Lilly. All sites that enrolled at least one patient were ranked as low, medium, or high performers based on a cluster analysis. The analysis found that investigators classified as a high enroller in one trial would 70% of time be a high enroller all the time relative to their colleagues—and the low enrollers were similarly consistent.

"I think all of us believe historical site performance matters," says Allen Kindman, M.D., FACC, IQVIA's vice president of clinical planning and analytics. "But I think how we use other investigator and site variables differs from pharma to pharma and CRO [contract research organization] to CRO." IQVIA has developed a machine learning process to exploit these other variables and is "accumulating the data to demonstrate value to this process."

Just because a site performs well in a therapeutic area like diabetes doesn't mean it will fare well in another, such as oncology, Kremidas notes. "In fact, it probably wouldn't, based on the type of patients drawn to the practice."

Identifying potential sites and investigators should not be a problem, says Kremidas. The WCG Knowledge Base contains trial and site performance data for 95% of all industry-sponsored protocols.  And TransCelerate sponsors an Investigator Databank, the result of a collaboration between sponsor companies that allows them to share investigator and site information.

Finding Investigators

Investigators are best matched to the needs of a study and, in some instances, that means experience and published work in a specific indication may matter more than other variables, says Kindman. Another important consideration is how many patients with a specific diagnosis are seen at a site—information that has been easier to come by since the 2017 Quintiles-IMS merger that created IQVIA.

"We can use administrative claims data and prescription data to come up with accurate estimates of protocol-specific patient counts," says Kindman. "We have an investigator database of over 350,000 physicians worldwide. In addition, we now leverage access to OneKey, a worldwide database of healthcare providers, which includes almost five million physicians, and their clinical interests."

Investigator profiles also uncover the referral network of investigators and other, potentially competing, studies they are conducting, Kindman continues. "We know who is practicing in a certain area and can determine their institutional affiliations. Typically, physicians will refer within an institution." Alternatively, notably in the U.S., insurance claims are an easy referral tracking mechanism and the same form contains patients' diagnosis and procedures performed. ClinicalTrials.gov, Informa and Citeline all have databases that can be mined to learn of studies investigators have already committed to.

IQVIA's evidence-based approach has resulted in a 24% median increase in enrolling days and a median of 44 days saved in completing site identification, as measured from the time it starts looking at prospective sites to final approval of those chosen for a site selection visit, relative to historic data, says Kindman.

Gomez says the TransCelerate Investigator Registry is a favorite tool. The registry pools investigator metrics and experience data from TransCelerate membership sponsor companies to surface the most accurate information and remove the duplicates, making it a big time-saver, she says.

The TransCelerate database was originally built and maintained by DrugDev. While DrugDev is now an IQVIA company, and TransCelerate data sits within IQVIA's servers, Kindman notes that the CRO side of the house has no access to all that rich investigator data. It's heavily firewalled due to contractual considerations. "So sponsors who are participants in the TransCelerate collaboration can access that data and we can help them ask the right questions of the data, but we can't get it ourselves."