The clinical-trial record can already show you who medicine leaves out. It just has to be read right

Long before a clinical trial enrols its first participant, it has already settled a question almost nobody examines: who is physically close enough to take part. The answer sits in the registry, in the list of study sites, in public, in a structured field.

"Site geography is the most under-read field in the registry," says Yaswanth Sai Pogadapula. "Where a trial places its sites determines who can physically reach it." You can see the catchment of a study before you see a single enrolment figure, he argues, because "travel distance is an eligibility criterion that appears nowhere in the eligibility section."

He makes that claim as someone who builds the records rather than analyses them. Pogadapula is a clinical trial disclosure specialist who, over his career, has been responsible for registration and results disclosure across roughly 40 global, regional and national registries, among them ClinicalTrials.gov, the EU register, and country-level registries with their own formats and deadlines. He has been lead disclosure coordinator for immunology, medical devices and consumer studies, with backup coverage across neuroscience, oncology, cardiovascular and metabolism, and pulmonary hypertension.

The work is specific. He has set up, maintained and closed study records, carrying studies from first registration through results posting, and has handled the quality comments registry operators send back, the NIH among them, the step that decides whether a public record is accurate rather than merely present. He has produced disclosure records for audits and regulatory inspections.

His argument: the record can answer questions about exclusion, provided you accept which parts of it bear weight. Registration data shows what was studied, where, and who was eligible. Results data carries baseline characteristics and participant flow. Together they describe the shape of who was enrolled and, by implication, who was not.

The trap is expecting it to behave like a research dataset. "It is a compliance record maintained to a standard," he says, "and the standard governs completeness, not comparability."

Three things follow. The first is that the most consequential field in the record is prose. Eligibility criteria, the terms that decide who may enter a study, are written as free text. "Prose does not aggregate," he says. "You can count trials; you cannot easily count exclusions." Any analysis of systematic exclusion therefore rests on interpretation that someone has to defend. His view of the remedy is structural rather than exhortative: "If the field were structured, questions about systematic exclusion would be answerable in an afternoon rather than in a research programme."

The second is the error he considers most common, and the one he is most insistent about. Where a demographic field is optional or inconsistently completed, an empty field says nothing about the people in the trial. "Absence in the record is not evidence of absence in the trial," he says. "It is evidence that nobody was required to report it." Treating a blank as a zero is, in his experience, the most common error made with this data, and it is made in good faith. He applies the caution symmetrically: he would distrust any claim built on missing registry fields, "including claims that support a conclusion I agree with."

The third is a seam invisible unless you have worked both sides of it. A registration record is created early, while the protocol is still moving; results are added years later, by different people. "Registration and results are maintained by different people at different times, and it shows," he says. Anyone reading across the two halves should know the join is there.

Those cautions carry weight because he has worked the production side of the problem. A single study may appear in a dozen registries, each with its own data model, required fields, language and deadline, so the same trial has to be expressed accurately several different ways, and small differences accumulate into records that will not line up later. His response was to treat the record "as a product with a standard rather than a form to be filled", the precondition, he says, for anyone reading across trials afterwards.

Statutory deadlines apply under FDAAA 801 in the United States and equivalent obligations in the EU, and missed deadlines are publicly visible. He treats on-time disclosure as the baseline of the job rather than its achievement. The deadline, though, is the easy half. "Compliance is often discussed as a deadline problem; in practice it is a data quality problem." The harder work was routing recurring registry findings back to study teams so the same correction did not have to be made twice.

Which leads to the tension he thinks explains most of the record's failings. "A compliance record and a public service are not the same artefact, and we ask one document to be both."

He is more optimistic than that sounds. Obligations are tightening and converging across jurisdictions, and every increment of required structured reporting makes representativeness more answerable. His prescription for policymakers is correspondingly blunt: "If policy wants exclusion to be visible, the lever is the required-fields list, not exhortation."

His advice to sponsors costs little and is easy to overlook. Treat the record as a product with an owner, not a filing obligation discharged at a deadline. Complete the optional fields anyway. Assume somebody will one day read across the portfolio and ask who was missing, and write the record now so the honest answer is visible then.