A sponsor budgets an imaging screen failure rate based on the last trial in the indication, opens enrollment, and watches the real rate come in well above it. The sites are not the problem. The imaging is not the problem. The problem is usually one sentence in the eligibility criteria that made sense to the clinician who wrote it but cannot be applied consistently to an image by a reader who has never met the patient.
Eligibility reads are where imaging first touches trial economics, and they are the part of the imaging plan most often settled last.
When central eligibility reads reject more candidates than expected, the instinct is to question the readers. That is usually the wrong place to look. Reader disagreement produces scatter around a threshold. A high rejection rate produces something else: a systematic gap between what the protocol was intended to capture and what its words actually permit. The site and the central reader are both applying the same criterion. They diverge because the criterion admits more than one reading.
The criteria that cause this tend to fail in a handful of recognizable ways. For example, they specify a threshold measurement without specifying how the measurement is made or at what location. They name an index joint, lesion, or segment without giving a selection rule for patients who have several that qualify. They call for a grade without specifying the view, so a site acquires a projection the grade was never validated on. They combine an imaging finding with a clinical qualifier that a blinded reader has no way to evaluate. They rely on a published grading system that is inherently subjective and has documented low agreement between readers.
None of these are foreign. All of them are discoverable before the protocol is finalized, simply by asking an imaging core lab to apply the criterion to a handful of real images and report where it breaks.
A site investigator assessing eligibility knows the patient, the history, and the referral question. A central reader has the images and the written criterion. That asymmetry is deliberate, and it is the reason central eligibility reads are worth doing, but it means any criterion that quietly depends on clinical correlation will fail at the core lab.
This is worth stating plainly during protocol development, because clinicians writing criteria do not naturally imagine the read. They imagine the clinic.
The last point matters beyond enrollment. Baseline eligibility discordance does not stay at baseline. A subject enrolled on a site assessment that the central reader would not have supported carries that discrepancy into every subsequent timepoint and into the endpoint analysis.
Medical Metrics operates as an independent imaging core lab across multiple therapeutic areas, supporting drug, biologic, and medical device sponsors. Eligibility reads are treated as a design question during protocol development rather than a throughput question during enrollment, because the cost of an ambiguous criterion is paid in screen failures, not in read time.
Want to know where your eligibility criteria will break? Let’s find out before enrollment opens!