More on program-survival bias:
[Thanks to ‘tuck’ on SI for this citation, which is apropos to a topic much discussed here on iHub.]
>>
Journal of Clinical Oncology, Vol 23, No 28 (October 1), 2005: pp. 6982-6991
Comparison of Outcomes of Phase II Studies and Subsequent Randomized Control Studies Using Identical Chemotherapeutic Regimens
Mohammad I. Zia, Lillian L. Siu, Greg R. Pond, Eric X. Chen
Department of Medical Oncology and Hematology, Princess Margaret Hospital and University of Toronto, Toronto, Ontario, Canada
PURPOSE: To determine whether promising results from phase II studies could be reproduced in phase III studies, and to examine which characteristics of phase II studies might be of predictive value for subsequent phase III studies.
METHODS: We searched for all phase III studies of chemotherapy in advanced solid malignancies, published in the English language literature from July 1998 to June 2003. Each phase III study was reviewed to identify preceding phase II studies. Phase II and phase III studies included in this analysis must have used identical regimens. Data were extracted from both phase II and phase III studies.
RESULTS: Of 181 phase III studies identified, 43 used therapeutic regimens identical to those in 49 preceding phase II studies. Twelve phase III studies (28%) were "positive." The vast majority (81%) of phase III studies have lower response rates than preceding phase II studies, with a mean difference of 12.9% among all studies analyzed. None of the phase II study characteristics evaluated significantly predicted for "positive" phase III studies, but the sample size of phase II studies demonstrated a trend toward being predictive (P = .083).
CONCLUSION: Promising results from phase II studies frequently do not translate into "positive" phase III studies. Response rates in most phase III studies are lower than those in preceding phase II studies.
<<
“The efficient-market hypothesis may be
the foremost piece of B.S. ever promulgated
in any area of human knowledge!”