In this presentation, the authors aim to emphasize not the outcomes of these comparisons, but rather the residual confounding present in these studies. They wanted to assess the residual confounding in comparisons between these two modalities and against men without prostate cancer.
To that effect, they utilized a province-wide (Ontario) cohort of men with localized prostate cancer from 2002-2009 and compared them to men without prostate cancer. Patients treated with surgery and radiotherapy were matched on demographics, comorbidity, and cardiovascular risk factors to men without prostate cancer. Of 20,651 eligible men with prostate cancer, 10,786 (5393 pairs) were matched (1 radiotherapy to 1 surgery patient). Patients were matched based on age, comorbidity, hypertension, statin use, diabetes diagnosis, previous MI, previous CVA, geographic region.
The primary outcomes were non-prostate cancer mortality and cardiovascular mortality. They used a previously published technique to quantify the prevalence and strength of residual confounding necessary to account for observed results.
The demographics of the cohort before and after propensity matching:
Prior to matching, patients undergoing surgery were younger and generally healthier, as can be expected.
Of the surgery and radiotherapy matched pairs, the 10-year cumulative incidence of non-prostate cancer mortality was higher among patients who underwent radiotherapy (12%) than surgery (8%; adjusted subdistribution hazard ratio 1.57; 95% confidence interval [CI] 1.35-1.83). Patients treated with radiotherapy also had an increased risk of cardiovascular mortality (adjusted HR 1.74, 95% confidence interval 1.27-2.37). Both groups had significantly lower rates of non-prostate cancer mortality than matched men without prostate cancer (18%; p<0.001).
Men treated for non-metastatic prostate cancer have significantly lower non-prostate cancer mortality than men in the general population – suggesting they are healthier men, to begin with.
Based on their subsequent sensitivity analysis assessing the effect of a hypothetical residual confounder, hypothetical residual confounders would have to be both strongly associated with non-prostate cancer mortality (HRs in excess of 2.5) and have highly differential prevalence in order to nullify the observed effect. Hence, selection bias does exist in choosing surgery or radiation and may account for some of the differences in RCT data – but these biases are unlikely to fully account for the observed survival differences between these treatment modalities.
References:
Wallis CJD, Satkunasivam R, Herschorn S, Law C, Seth A, Kodama RT, Kulkarni GS, Nam RK. Association Between Primary Local Treatment and Non-prostate Cancer Mortality in Men With Nonmetastatic Prostate Cancer. Urology. 2018 Apr;114:147-154. doi: 10.1016/j.urology.2017.12.013. Epub 2018 Jan 2. PMID: 29305198
Presented by: Wallis, Christopher J. Resident, University of Toronto, Toronto, Canada
Co-Authors: Christopher Wallis1, Raj Satkunasivam2, Sender Herschorn1, Calvin Law3, Arun Seth4, Ronald Kodama1, Girish Kulkarni1, Robert Nam1
Author Information:1Urology, University of Toronto, Toronto, ON, Canada; 2Urology, Houston Methodist Hospital, Houston, TX, United States; 3General Surgery, University of Toronto, Toronto, ON, Canada; 4Anatomic Pathology, Sunnybrook Health Sciences Centre, Toronto, ON, Canada
Written By: Thenappan Chandrasekar, MD, Clinical Fellow, University of Toronto Twitter: @tchandra_uromd at the 73rd Canadian Urological Association Annual Meeting - June 23 - 26, 2018 - Halifax, Nova Scotia