If Julia Hinde has reported Dr Chris Palmer correctly ("Trials with too much error", THES, April 17), he believes that it is the failure of trialists to abandon traditional statistical approaches in favour of more innovative ones that prevents "initial results of trials to be acted on while the trials are still ongoing". In practice, there are more formidable obstacles.
Right or wrong (right in my view), drug regulatory agencies require good evidence that treatments are safe, effective and of good quality before registering them. The Food and Drug Administration in the United States generally requires two significant phase III trials for acceptability.
Even after the last trial is complete, it can take months to compile a regulatory dossier. After that, it usually takes the regulators between one and three years to deliver their verdict. Manufacturing capacity for the drug must then be created and, in any case, physicians can take years to abandon tried favourites for better therapies. Thus, even if researchers convince themselves, using data from their own trial, that a treatment is efficacious, most patients will continue to receive conventional treatment for many years.
Where, as in the case of early trials of Aids therapies, the rules were bent to encourage rapid introduction, later doubts about efficacy led this haste to be criticised subsequently by groups that had initially been impatient about the pace of change. At the moment there are influential calls for a fourth hurdle, that of value for money, to be added to the drug regulatory process. There are also increased requirements to demonstrate efficacy in subgroups. Thus, elsewhere, critics of the pharmaceuticals industry appear to think that new therapies are introduced too quickly rather than too late. This is not to say that data-dependent trials may not have a role in some cases. In general, however, trials that promise to deliver a fixed, rather than a random, amount of information will continue to be useful.
Stephen Senn Professor of pharmaceutical and health statistics University College London