Statistics from Altmetric.com
The results of clinical trials may not be reflected in ordinary clinical practice. Strict regulation of the conditions under which trials occur may mean that both intervention and control groups are cared for in ways that are different from those that apply outside such trials. Standards of practice may be different in units that undertake clinical trials and may differ between trial and non-trial groups in the same units.
At McMaster University in Hamilton, Canada a trial of antithrombin treatment in premature infants with respiratory distress syndrome (Barbara Schmidt and colleagues. American Journal of Respiratory and Critical Care Medicine1998;158:470–6) showed an adverse effect of antithrombin. A total of 198 babies was eligible for entry into the trial but 76 were not randomised because parental consent was withheld (38), because consent was not asked for, usually for administrative reasons (24), or because the trial was closed temporarily because of production problems with the antithrombin (14). These eligible, non-randomised babies did worse than those randomised to placebo (Barbara Schmidt and colleagues. Journal of Pediatrics 1999;134:151–5; see also editorial, Ibid: 130–1). The median duration of mechanical ventilation was 7.1 days in the antithrombin group, 4.8 days in the placebo group, and 6.2 days in the eligible, non-randomised group. Median duration of supplemental oxygen administration was 7.9 days, 5.5 days, and 4.9 days, respectively. Intraventricular haemorrhage was somewhat more frequent in the eligible, non-randomised group compared with the placebo group (32%v 22%, p = 0.22). The authors argue that the differences seen might not simply be a placebo effect as ordinarily understood.
If patients allocated to placebo do better than those not randomised there are several implications. Placebo groups are likely to be treated according to strict protocol and therefore the finding might be seen as providing justification for protocol or guidelines based practice. Second, the true benefit of the treatment being investigated might be greater than that shown by the difference between treatment and placebo groups (an editorial writer suggests that a less rigorous standard of statistical significance might be acceptable). Third, participation in clinical trials might be regarded as beneficial in itself and the need for ethical standards to protect trial patients might be seen as less compelling. (The same editorialist rejects this proposition on the grounds that patients in trials need protecting not because trials are inherently dangerous but because they alter the doctor–patient relationship.)
The changes brought about by setting up a clinical trial may be complex. They need to be taken into consideration in assessing trial results and the implications for clinical practice in general.