Every later safeguard inherits whatever the trial evidence got wrong

The argument

Proponents of this position begin from where the licensing decision sits in the sequence. Everything a regulator does afterwards treats that decision as its starting point. An inspector checks that a medicine is made to the specification the dossier set out. A surveillance system watches for harms against the expectations the trials established. A prescriber weighs a benefit the trials reported. If the evidence underneath was wrong, each of these functions continues to operate correctly and arrives at the wrong place, because none of them is asking the question the trial has already answered. The argument is that trial design is where this goes wrong in ways that are hard to see later. A trial can measure a surrogate endpoint, a laboratory value that moves reliably, in place of the outcome patients care about, and report a clear effect on something that may not translate. It can compare a medicine against placebo where a treatment already exists, which establishes only that it is better than nothing. It can be powered to detect benefit and not harm, stopped early at a favourable moment, or conducted alongside others that are never published, so that the licensed picture is assembled from the subset which happened to be flattering. Proponents stress that each of these produces evidence that is accurate as far as it goes and misleading about what was actually asked. Holders of the view conclude that a regulator's leverage is greatest here and nowhere else. The design of a trial is fixed before it begins, in a protocol a regulator can require, amend, register publicly and hold a sponsor to, and the comparator, the endpoint, the duration, the population and the rules for stopping are all decided in advance. Once the trial has run that leverage is gone, and no amount of inspection or surveillance can retrieve the question that was never asked.

Premises

Counter-arguments

Critics argue that the asymmetry is overstated. Surveillance does revise the trial picture, sometimes decisively, and some of the clearest reversals of an approved medicine's benefit and harm profile have come from observing it in use rather than from a better protocol. They also doubt that regulators can specify good trials in advance as reliably as the position assumes, since the endpoint that matters may not be known until later and a regulator that fixes a protocol too firmly slows the arrival of medicines whose value is real. Others note that the position says little about the stages after approval, which is where most of the medicines in circulation actually sit.

Rejecting the premises

[Rejecting P1] Surveillance in use does not simply inherit the trial picture but has repeatedly revised it, which weakens the claim that later functions cannot correct the evidence. [Rejecting P3] The endpoint or population that turns out to matter is often unknown when a protocol is written, so fixing design in advance does not reliably secure the right question. [Rejecting C] Concentrating effort before approval leaves the stage at which most medicines in circulation are actually used comparatively unattended.