Association Between Primary Local Treatment and Non-Prostate Cancer Mortality in Men with Non-Metastatic Prostate Cancer

The question of the optimal treatment for patients with non-metastatic prostate cancer remains one of great interest in the urologic and oncologic communities. Indeed, it has been included as one of the top priorities for comparative effectiveness research. Randomized controlled trials (RCTs) are widely accepted among researchers, clinicians and policymakers as the highest quality study design to guide practice¹. However, there are substantial barriers to the conduct and implementation of RCTs including ethical constraints to randomization, lack of generalizability, and high costs^2,3. Further, RCTs may be out of date prior to completion due to their significant periods of accrual and follow-up¹. This is particularly challenging in studies that require randomization to surgery, which have been difficult to execute⁴. Thus, there remains a significant role for well-designed observational studies to both answer research questions efficiently and address those for which other study designs may be inappropriate or difficult to conduct (e.g. studies of harms or the case of rare diseases)⁵. Well-designed observational studies produce results that are very similar to comparable RCTs^6,7 and often produce more precise estimates⁶, owing to their larger size.

Residual confounding, due in part to selection bias, is a concern in observational comparisons of surgery and radiotherapy in the treatment of prostate cancer⁸. Residual confounding occurs where there has been an inadequate adjustment for the effects of a variable which confounds the relationship between exposure and outcome. This may arise due to unmeasured confounders, inadequate adjustment due to substantial heterogeneity among covariate categories (as in the use of a binary age classification⁹), or misclassification.

Therefore, in this study, we sought to examine rates of non-prostate cancer mortality among older men treated for non-metastatic prostate cancer with surgery or radiotherapy. Further, we compared these men treated for prostate cancer to the general population. Finally, in order to better examine the role of selection bias in observational studies of this sort, we used a previously published technique to quantify the effect of residual confounding¹⁰.

After accounting for exposure to androgen deprivation therapy, men treated with radiotherapy were significantly more likely to experience non-prostate cancer-related mortality (1.57, 95% CI 1.35-1.83). When we stratified by radiotherapy modality, this effect was restricted to patients who received external beam radiotherapy, not those who received brachytherapy. When compared to matched members of the general population, both men who received surgery and those who received radiotherapy had significantly decreased rates of non-prostate cancer death (see Supplementary Figure 1).

Finally, we examine the differential incidence and magnitude of potential confounders necessary to eliminate the observed effect. In other words, if there is really no difference between patients treated with surgery and radiotherapy, what are the characteristics of the confounders necessary to produce the observed effect. As can be seen in manuscript Figure 2, there needs to be a very large differential prevalence (that is substantially more frequent among patients receiving radiotherapy) and a very large effect (hazard ratios in excess of 2.5, and often in excess of 10, depending on the absolute and differential prevalence).

Therefore, we can conclude that, while unmeasured confounding or selection bias is almost certainly present, it is highly unlikely that they entirely account for differences in survival between patients treated with surgery and radiotherapy in observational studies. Thus, the present results may be explained by a true difference in mortality between patients treated with surgery and radiotherapy, magnified by residual confounding.

References:

Bothwell LE, Greene JA, Podolsky SH, Jones DS. Assessing the Gold Standard--Lessons from the History of RCTs. The New England journal of medicine. 2016;374(22):2175-2181.
Stuart EA, Bradshaw CP, Leaf PJ. Assessing the generalizability of randomized trial results to target populations. Prev Sci. 2015;16(3):475-485.
Shore BJ, Nasreddine AY, Kocher MS. Overcoming the funding challenge: the cost of randomized controlled trials in the next decade. J Bone Joint Surg Am. 2012;94 Suppl 1:101-106.
McCulloch P, Taylor I, Sasako M, Lovett B, Griffin D. Randomised trials in surgery: problems and possible solutions. Bmj. 2002;324(7351):1448-1451.
Black N. Why we need observational studies to evaluate the effectiveness of health care. Bmj. 1996;312(7040):1215-1218.
Concato J, Shah N, Horwitz RI. Randomized, controlled trials, observational studies, and the hierarchy of research designs. The New England journal of medicine. 2000;342(25):1887-1892.
Benson K, Hartz AJ. A comparison of observational studies and randomized, controlled trials. The New England journal of medicine. 2000;342(25):1878-1886.
Williams SB, Huo J, Chamie K, et al. Discerning the survival advantage among patients with prostate cancer who undergo radical prostatectomy or radiotherapy: The limitations of cancer registry data. Cancer. 2017;123(9):1617-1624.
Chen CL, Gilbert TJ, Daling JR. Maternal smoking and Down syndrome: the confounding effect of maternal age. American journal of epidemiology. 1999;149(5):442-446.
Lin DY, Psaty BM, Kronmal RA. Assessing the sensitivity of regression results to unmeasured confounders in observational studies. Biometrics. 1998;54(3):948-963.

Written by: Christopher Wallis, MD, PhD, Division of Urology, Sunnybrook Health Sciences Centre, Sunnybrook Research Institute, University of Toronto, Toronto, Ontario, Canada; Institute of Health Policy, Management, & Evaluation, University of Toronto, Toronto, Ontario, Canada.

Read the Abstract

Login