Contributed Commentary By Karim Damji

July 26, 2018 | If there was a 50% chance you would fail to meet a key metric for a project that would cost millions, would you procced?

If your projected timeline was almost certain to double, would you still be all in?

You don’t have to be a master strategist to recognize a potentially losing proposition. The game is seemingly over before it’s even begun.

Yet these are the odds that the majority of clinical trial sites face during start-up. And almost all forge ahead. According to the Tufts Center for the Study of Drug Development, nearly half of clinical trial sites either fail to enroll a single patient or under-enroll, and trial timelines typically run twice as long as initially projected, just to meet established enrollment levels. The life sciences industry knows that these all too common shortcomings lead to delays in clinical trial initiation, not to mention cost overruns, premature trial termination, loss of critical statistics/data, and delayed time to market launch.

Since clinical trials are the backbone of therapeutic development, as well as the source of new hope for patients facing devastating diseases and the HCPs who treat them, shouldn’t it be easy—in theory—to find willing and appropriate patient participants?

As evidenced by the statistics – and experience – the answer is no.

For years, many factors have contributed to the problem of identifying and recruiting the right kind and numbers of patients for any given clinical trial: lack of physician and patient awareness about available trials, unclear and/or unrealistic patient inclusion/exclusion criteria, and trial location, to name a few. Despite historical efforts to address these issues, only incremental progress has been made. Barriers remain, resulting in delayed trial initiation.

Fortunately, the advent of artificial intelligence (AI) has opened the door to more effective patient recruitment solutions than ever before. AI has ushered in a new era of clinical trial feasibility, informing and optimizing the patient identification, recruitment, and retention continuum. AI-informed data analytics solutions are producing game-changing results at lightning speed, leveraging machine learning (ML) to accelerate clinical trial timelines. Such feasibility solutions can optimize decision parameters and predict patient recruitment outputs, answering critical questions such as:

• Are the protocol inclusion/exclusion criteria practical?
• Are enough patients available for the planned protocol?
• Are patient locations feasible to conduct a successful study?
• Is the study relevant to the proposed region/demographic?

When clinical trial protocols are informed by such data, the time and costs associated with patient recruitment are minimized, the numbers of trial-appropriate patients are maximized, start-up is streamlined, and trials can proceed as planned. Common patient recruitment problems of enrollment exceeding timeline goals, screen failures, and patient attrition are avoided.

Most AI-based trial feasibility solutions leverage a Natural Language Understanding (NLU) engine that enables them to sift through structured and unstructured Real World Data (RWD) in a manner and speed that was previously both unattainable and unimaginable. NLU engines make it possible for trial feasibility solutions to process plain text in protocol documents and analyze inclusion/exclusion criteria. In the US, almost 80% of the data in 1.2 billion clinical records is unstructured. NLU facilitates the mining of the enormous amount of useful, but text-heavy, information contained in Electronic Health Records (EHR) such as physician notes, images, and scanned documents, matching patients with trial sponsors based on geography and demographics. If context and history is what you need, AI-based feasibility solutions can locate and list past trials with similar parameters, including selection criteria, number of patients enrolled, and trial status. Going one step further, these platforms can develop a list of recommended inclusion/exclusion criteria, based on therapeutic area and indication, to include in a clinical trial protocol document.

Any advance in the life sciences creates a ripple effect, either directly or indirectly effecting patients, caregivers, HCPs, and industry. Employing AI-based solutions to optimize clinical trial recruitment is unique however, because instead of a ripple it creates a tsunami of advantages that result in more effective site start-up, reduced trial costs, better data generation, timely trial completion and, ultimately, greater opportunities for new therapeutic options. Those are odds I can get behind.

Karim Damji is SVP Product Management & Marketing at Saama Technologies. Damji is responsible for the product management, strategy, and marketing of Saama Technologies’ analytics solutions. He can be reached at karim.damji@saama.com.