Hospitals are still overwhelmed with COVID-19 cases. Patients have postponed diagnostic tests and screenings. Researchers are still focusing their attention on combatting the virus. It’s no surprise that clinical trials have plunged.
It would be a mistake to consider the challenges that clinical trials face today without remembering those of the past, however. Sponsors, cancer center research staff and clinicians know that matching eligible patients to trials is costly and burdensome even in the best of times. Before the pandemic, only 5 percent of oncology patients participated in clinical trials.
If we want to really expand patient recruitment and enrollment in trials, offer trials as therapeutic options and make up for time lost during the pandemic, we must overcome the unprecedented challenges of COVID-19 without replicating the inefficiencies that formerly hindered research.
Today, many cancer centers use a time-consuming but necessary processes to identify eligible patients, including manually cross-referencing records to determine who might participate in trials. Natural language processing and other AI-related tools, in contrast, can parse through electronic medical records in vast health networks and match eligible patients to trials quickly.
It’s already happening.
In the Southwestern US last year, feasibility study questionnaires (FSQs) for clinical trials inundated a community oncology network. FSQs spell out the criteria that make patients eligible or ineligible for trials. They are an essential step in connecting trial sponsors with patients in cancer centers. Gathering the requisite information to answer FSQs, however, is a daunting, time-intensive task that requires cancer centers to vet their patient population – often manually – to determine if they will fit the highly specific needs of proposed cancer studies.
This network included nine clinics and two large cancer centers that covered 80 percent of the state. On average, it took the staff as long as three hours to complete a single FSQ. Some took much longer, however. Filling out FSQ forms, which generally must be returned within two days, took between 24 and 48 hours in total per week to complete. Many of the network’s patients may have been primed to benefit from new treatment options, but finding them was taking too long.
Additionally, after all the time and effort of set up, clinical trial can close at cancer center due to insufficient enrollment – even though, through the manual FSQ process, the center expected to enroll enough patients in the trial. The institution’s reputation, financial resources and, most importantly, its capacity to benefit patients suffered.
The network’s research director wanted to improve this unsustainable process that required significant time but frequently yielded less-than-optimal returns.
They turned to use a suite of precision matching technology that included natural language processing and collects unstructured data, or differently formatted data not generally included in EHRs like doctors’ notes, lab and pathology reports. As much as 80 percent of the data that determines whether patients are eligible for trials is unstructured data and not easily accessible in standard EHRs.
Reviewing this data is the most time-consuming step in finding feasible patients for trials. Aggregating and normalizing all patient data into a single universe, the technology quickly determined who among their patient populations might be suited for clinical trials on offer. After automation, staff spent only around 30 minutes on FSQs a week.
Instead of fumbling with file folders, digging through EHRs and double checking with doctors, staff could run a query of patients potentially eligible for a study and let the sponsor know almost immediately. Doctors and staff still needed to vet and ultimately approve patient matching. But they bypassed the laborious work. The technology did it for them. As a result, they had more time to spend on patient care and could present clinical trials to patients as early as possible.
At the same Southwestern practice, only three breast cancer patients were enrolled in clinical trials from November 2019 to March 2020. But after deploying the same system, the number of potential patients for the trial grew to 100. Between March and June 2020, the height of the pandemic when most cancer clinical trial enrollment was down, 10 patients were enrolled – a 233 percent increase – with more pending.
These case studies shows how using new and, importantly, scalable technology to sift through EHRs can make a tedious, time-consuming task fast and easy, even in a crunch.
The days of frantically searching for patients to match to clinical trials are over. Quick and easy searches through vast databases of cancer patients can save doctors’ time and help pharmaceutical companies find discrete populations of patients for potentially life-saving clinical trials. It’s a solution that can jumpstart research if we can apply the lessons that we’ve already learned.