Reply to comparative risk-adjusted mortality outcomes after primary surgery, radiotherapy, or androgen-deprivation therapy for localized prostate cancer
Article first published online: 14 DEC 2010
Copyright © 2010 American Cancer Society
Volume 117, Issue 11, pages 2577–2578, 1 June 2011
How to Cite
Cooperberg, M. R., Carroll, P. R. and Vickers, A. J. (2011), Reply to comparative risk-adjusted mortality outcomes after primary surgery, radiotherapy, or androgen-deprivation therapy for localized prostate cancer. Cancer, 117: 2577–2578. doi: 10.1002/cncr.25817
- Issue published online: 19 MAY 2011
- Article first published online: 14 DEC 2010
- Manuscript Accepted: 22 OCT 2010
- Manuscript Received: 12 OCT 2010
Dr. Froehner suggests our article1 claimed unmeasured confounding is not a partial explanation for the differences observed among treatment modalities. In fact, we explicitly considered that possibility, devoting multiple sensitivity analyses and a significant portion of the discussion to this very question. The question is not whether there is unmeasured confounding underlying the analysis—we will never know for sure but must assume this possibility. The question is whether unmeasured confounding can explain a 2-fold difference in mortality between radiation and surgery and a 3-fold difference between androgen deprivation and surgery.
The presumptions in Dr. Froehner's argument are that radiation and androgen-deprivation patients have inherently higher-risk disease than surgery patients and that this difference in risk is not captured by multivariate risk-stratification systems. The sensitivity analysis presented in Table 5 of the article1 was designed to quantify the unmeasured confounding one would have to assume to explain away the findings. We artificially raised the Kattan scores for each of the surgery patients by 5-point increments, in effect giving them progressively less “credit” for their baseline disease characteristics in the risk-adjusted survival analysis.
The mortality difference between surgery and radiation remained statistically significant until the surgery patient scores were raised by 20 points and did not attenuate completely until the scores had been raised by 30 points. Thus, to attribute the observed mortality difference to unmeasured confounding, we would have to assume that a radiation-therapy patient with a predicted survival, for example, of 80% would have the same biologic risk as a surgical patient with a predicted survival of 60%. Moreover, these imbalances in prediction would have to be both pervasive and consistent. Given the established accuracy of risk prediction for both surgical and radiation-therapy patients, such a consistent imbalance seems highly unlikely.
We reiterate that randomized trials for men with high-risk disease are essential and that these must include surgical arms. In the interim, our study, together with others recently published and/or presented, points to a larger role for surgery—often as a part of a multimodal approach—for high-risk prostate cancer.
- 1Comparative risk-adjusted mortality outcomes after primary surgery, radiotherapy, or androgen-deprivation therapy for localized prostate cancer [published online ahead of print November 15, 2010.] Cancer. 2010; 116: 5226-5234. doi: 10.1002/cncr.25456., , , .