This study estimated individualized absolute benefit of adding and extending androgen deprivation therapy (ADT) alongside radiation therapy (RT) for localized prostate cancer. The primary metric was the modelled 10-year absolute risk reduction (ARR) in distant metastasis (DM) when short-term ADT (STADT) is added to RT, or when STADT is prolonged to long-term ADT (LTADT).
The intent was to integrate three components that determine absolute benefit: cancer-specific prognosis, competing (other-cause) mortality, and relative efficacy of ADT. The authors emphasize that this integrated approach reveals heterogeneity in benefit that is not captured by conventional risk groupings alone.
Individualized cancer-specific and non-cancer hazards were obtained from validated models. Cause-specific hazard estimates for prostate cancer outcomes were taken from the STAR-CAP staging system. Non-cancer mortality probabilities were estimated using the validated OCCAM model. Relative treatment efficacy of ADT was incorporated using published hazard ratios from the MARCAP meta-analysis.
These components were combined to produce individualized estimates of absolute DM risk under guideline-concordant care and the counterfactual risk after adding or prolonging ADT. The methodology links validated cause-specific hazard models with external estimates of relative treatment effect to convert relative efficacy into absolute risk change for individual patients.
Model utility and output distributions were illustrated using the Prostate, Lung, Colon, and Ovarian (PLCO) cancer screening trial cohort. The analytic sample included 5,468 prostate cancer patients with complete covariates required for the integrated models.
The primary outcome calculated for each patient was the estimated 10-year ARR in DM from either adding STADT to RT, or from extending STADT to LTADT, depending on clinical subgroup and guideline indications.
Model-estimated 10-year risk of distant metastasis (DM) under guideline-concordant care varied widely across the PLCO cohort, ranging from less than 1% to 56%. Independently, modeled other-cause mortality (OCM) risk at 10 years ranged from 3% to 80%.
In the subgroup with NCCN unfavorable intermediate-risk disease (n=1,814), adding STADT to RT produced a median estimated 10-year ARR in DM of 4%. Individual ARRs within this group varied substantially, from under 1% to nearly 15%.
Among NCCN high-risk patients (n=1,289), prolonging ADT treatment yielded a median ARR of 9% with observed individual ARRs spanning 0.4% to 17.2%.
Quartile-level results highlight that 25% of unfavorable intermediate-risk patients had ARRs less than 2.4%, and 25% of high-risk patients had ARRs less than 5%, demonstrating substantial within-group heterogeneity.
The integrated model shows that conventional risk categories (for example NCCN unfavorable intermediate and NCCN high risk) include patients with a broad range of expected absolute benefits from ADT. Some patients categorized within the same NCCN group have very small estimated ARRs, while others in the same group have much larger expected absolute benefit.
This heterogeneity arises because ARR depends not only on tumor aggressiveness but also on competing mortality risk and the baseline absolute risk of metastasis; two patients with the same NCCN stage can have different comorbidities and baseline hazards that substantially change the net absolute benefit of adding or extending ADT.
The approach provides individualized absolute benefit estimates that may support shared decision-making about adding or prolonging ADT in patients receiving RT for localized prostate cancer. Presenting absolute risk reductions can help contextualize the trade-offs of ADT given its side effects and duration.
The authors caution that the analysis is based on an integrated modelling framework combining validated prognostic models and published relative treatment effects; it is a preprint and has not been peer-reviewed. They explicitly state that external validation is required prior to clinical implementation.
Other limitations inherent to the report include reliance on available covariates in the PLCO cohort and the use of published hazard ratios to represent ADT efficacy; any unreported sources of heterogeneity in treatment effect or changes in practice patterns were not detailed in the source.
PLCO data are available on request through the National Cancer Institute repository (https://cdas.cancer.gov/learn/plco/instructions/). The cause-specific hazards used in the analysis and the code to generate estimates are available on reasonable request to the authors.
The authors conclude that an integrated model combining cancer aggressiveness and comorbidity demonstrates substantial heterogeneity in expected absolute benefit of ADT within conventional risk groups and that individualized estimates could inform treatment discussions, but that external validation is needed before clinical use.