Five-year change in statistical designs of phase II trials published in leading cancer journals.

This study compares the evolution in statistical design reporting for phase II cancer clinical trials published in the six following leading journals: American Journal of Clinical Oncology, Annals of Oncology, British Journal of Cancer, Cancer, European Journal of Cancer and Journal of Clinical Oncology. Only articles where tumour response was considered as the primary endpoint were selected. A total of 393 phase II trials published in 1995 (n=185) and 2000 (n=208) were reviewed. Neither sample size nor design parameters were specified in 157 (85%) and 113 (46%) papers in 1995 and 2000, respectively. 28 (15%) and 95 (46%) papers included at least some information on the statistical designs used: Gehan (4.3% and 3.3%), Fleming (2.2% and 4.3%), and Simon (2.7% and 11.0%). Ad hoc, non-referenced methods were used in 5.9% and 27.3% articles in 1995 and 2000, respectively. Although there is an increase in the mention of at least some statistical design parameters in phase II cancer clinical trials over a 5-year period in these selected cancer journals, the use of referenced methods is still short or often inadequate.