Association between author specialty and reported outcomes in non-randomized comparisons of prostatectomy versus radiation.

Abstract

Background: While the ProtecT randomized trial demonstrated comparable oncologic outcomes between radical prostatectomy (RP) and radiation therapy (RT) for localized prostate cancer, several retrospective studies have reported mixed findings. In this study, we investigated whether non-randomized studies tend to favor RP or RT and assessed whether the specialty of leading authors-urologists or radiation oncologists-correlates with study conclusions.

Methods: We identified non-randomized clinical studies published between 1993 and 2024 that compared RP and RT for localized prostate cancer and reported oncologic outcomes including biochemical, metastasis, and/or survival outcomes. Author specialty was determined based on whether the first and senior authors were affiliated with urology or radiation oncology departments. Studies in which a urologist and radiation oncologist were first and senior authors were excluded. Each study's conclusion was categorized as favoring surgery, favoring radiation, or reporting no conclusive difference in oncologic outcomes. Multivariate logistic regression was used to evaluate predictors of study conclusions.

Results: A total of 105 studies met inclusion criteria. Nearly half (44%) reported no conclusive difference in outcomes, while 42% favored RP and 14% favored RT. Multivariate analysis showed that use of national databases (P = .01) and urology author specialty (P = .01) predicted conclusions aligning with the author's specialty. In a subgroup excluding studies relying solely on biochemical recurrence-free survival outcomes, author specialty remained associated with conclusions, with urologists more likely to favor RP (P = .006).

Conclusion: Author specialty was associated with the direction of study conclusions. Further research is needed to understand how these patterns may influence interpretation and decision-making in prostate cancer treatment.