First dibs.

Where there is more than one standard treatment approach, the selection of treatment modality becomes a huge source of selection bias in any look at real-world outcomes. To measure just how big this bias can be, this novel SEER analysis looked at patients with low-risk prostate where we know any treatment—including none—results in a near absent prostate cancer mortality rate. The primary endpoint was something they termed the “treatment selection bias” effect, which was essentially unadjusted overall survival not due to prostate cancer. As (everyone) expected, the prostate cancer mortality at 10 years among the over 50K patients included was only 1%. At the same time, as (rad oncs) expected, there was a clear discrimination in overall survival at 10 years per treatment modality: 93% after radical prostatectomy, 84% after brachytherapy, and 77% after external beam radiation. What’s more, this discrimination persisted after the notorious propensity score matching technique. TBL: The selection of treatment itself is a huge source of selection bias that is not accounted for in propensity score matching, meaning survival outcomes stratified by treatment modality are not appropriate outcomes for registry studies. | Miccio, Prostate Cancer Prostatic Dis 2020

Comments

Popular Posts