The Registry

For most of the twentieth century, a company running clinical trials of its own product had no obligation to report the ones that did not go well. A drug could be tested repeatedly, and the trials that produced a favourable result could be published while the others stayed in a filing cabinet.

This is the file drawer problem with a financial interest attached, and the consequence is not an abstract distortion of the literature. It is that prescribing decisions were made on a body of evidence that had been selected by the party with the most to gain.

Specific cases established the pattern beyond dispute. Litigation and regulatory review in the 2000s surfaced unpublished trial data on several widely prescribed drugs where the complete picture differed materially from the published one, on both efficacy and harm. In several instances the regulator had the fuller data and the prescribing clinician did not.

The structural fix was registration. From 2005, the leading medical journals, coordinated through the International Committee of Medical Journal Editors, refused to publish trials that had not been registered in a public database before enrolment began. Registration records what will be measured, in advance. Legislation in the United States and later in Europe added reporting requirements.

The principle is exactly the one this archive has been circling. If you must declare in advance what you are testing and what counts as success, you cannot decide afterwards that the study was always about the secondary outcome that happened to work.

Compliance has been the difficulty. Audits of results reporting against the legal deadlines have repeatedly found substantial non-reporting, including by large commercial sponsors and by major universities. Projects set up specifically to track this have published league tables, and the enforcement of penalties has been limited.

So the mechanism exists, it demonstrably works where it is followed, and it is followed unevenly. That is a more useful thing to know than either the optimistic or the cynical version.

What makes this the most hopeful entry in the series is that it is the one case in this archive where a suppression mechanism was identified, named, and structurally addressed within a generation. The Index ran for four centuries. The scrolls embargo ran for forty-four years. This was diagnosed and partially fixed in about a decade.

The fix was not persuasion and it was not ethics training. It was a requirement, enforced at the point of publication, by editors who could simply refuse. A structural problem yielded to a structural remedy, which is the lesson worth carrying.

It also means the outstanding problem is now measurable, which is a different situation from not knowing. Somebody can count the unreported trials, and does.