1 Technische Universitaet Berlin, Institute of Health Sciences, Department of Health Care Management, Berlin, Germany
*E-mail: marcial.velasco@tu-berlin.de
In the March issue of Annals of Oncology, Fossati et al. [1] presented what they called indirect evidence for the existence of a wish bias in oncology trials that may lead to overestimation of treatment results when a drug is being considered an innovation. The authors observed a decrease in the overall response rate and partial response rate to treatments of metastatic breast cancer, including doxorubicin, over time. They explained this finding with the conscious or unconscious overestimation of treatment effects by non-blinded investigators when efficacy is assessed using end points exposed to some subjectivity. The authors did not, however, consider the, at least theoretical, possibility of confounding through a shift over time towards patients with worse prognosis being included in the trials. The first trials could have concentrated on highly selected patient groups with a better prognosis. With widespread use of chemotherapy, trials conducted later would have progressively opened to less selected patients, thus including also women with a worse prognosis. If this had happened, one would expect to observe a progressive decrease in the overall response rates not only in the doxorubicin arm, but also in the group treated with the alternative drugs, whichever these were.
To explore this possibility we reproduced the investigation of Fossati et al. to some extent, concentrating on the doxorubicin-free arm. From the trials cited in their paper we extracted the overall response rates observed within the doxorubicin-free groups and the randomisation year, and plotted them in a similar way to their paper. The median response rate was 49%, ranging from 22% to 68%. Figure 1 shows a similar correlation between randomisation year and overall response rates for the doxorubicin-free arm, as did the picture reported by Fossati et al. for the doxorubicin-containing arm, thus indicating also a decrease in the efficacy of the drugs in the doxorubicin-free arm. Further, we found a moderate but significant positive correlation (r = 0.5558, P <0.01) between the overall response rates in both arms of the trials. In the trials in which the doxorubicin-containing combination did better, the alternative drug combination was also doing better. These findings suggest that overall response rates in chemotherapy trials for metastatic breast cancer have been decreasing over time independently of the drugs under investigation. As stated above, this general decrease of the efficacy could be better explained by other factors (such as the inclusion of patients with worse prognosis or the more stringent application of remission criteria in newer trials) affecting the results on both arms of a trial (i.e. the one including the therapy considered to be new and promising and the one including the drugs considered to be old and standard) rather than by distortion of the results in favour of the new drug.
|
1Technische Universitaet Berlin, Institute of Health Sciences, Department of Health Care Management, Berlin, Germany (*E-mail: marcial.velasco@tu-berlin.de)
References
1. Fossati R, Confalonieri C, Apolone G et al. Does a drug do better when it is new? Ann Oncol 2002; 13: 470473.