AI Shrinks Cancer Trial Screening to Seconds With Proven Accuracy

Validated AI now screens cancer trial patients in seconds with 98.6% accuracy, speeding studies and widening access

19 Dec 2025

American healthcare has heard many promises about artificial intelligence. In cancer research, some are starting to stick. A new system from Memorial Sloan Kettering Cancer Center shows how AI can solve a problem that has quietly slowed clinical trials for decades: finding eligible patients quickly enough.

The tool, called MSK-MATCH, scans electronic patient records and checks them against trial criteria. In a retrospective test across six breast-cancer studies, it matched patients to trials with 98.6% accuracy. That level of performance suggests something more than a laboratory exercise. It looks fit for routine use.

The bigger gain is speed. Even when human review was still required, screening took about 43 seconds per patient-trial pair. Manual chart reviews often take 20 minutes or more. For research teams with thin staffing and strict deadlines, that difference is not marginal. It changes how work gets done.

Recruitment has long been the hidden weakness of clinical trials. Many studies fail to meet enrolment targets, pushing up costs and delaying results. Early eligibility checks consume time and attention but add little scientific value. Automating that step allows teams to review far more patients without hiring more staff. Trials can open sooner and proceed with fewer delays.

This reflects a wider shift. AI in healthcare is moving upstream. Instead of being used mainly to analyse data after a study ends, it is shaping how trials are designed and run. Sponsors face pressure to move faster without sacrificing rigour. Tools that reduce friction at the start of a trial are therefore attractive.

There is also an equity angle. Automated screening can search broadly across patient populations rather than relying on referrals from a handful of doctors. That may help address persistent gaps in representation, particularly in cancer research, where minorities are often under-recruited.

Risks remain. Patient privacy must be protected. Algorithms can reflect biases in the data they are trained on. Regulators will need to decide how much oversight is enough. These concerns are real, but they are no longer enough to stop adoption. MSK’s partnership with Triomics to expand AI-supported screening suggests institutional confidence.

For patients waiting on new treatments, faster trials matter. AI is not curing cancer. But by clearing old administrative hurdles, it may help cures arrive sooner.