Related Experiment Videos
Analyzing predictive models following definitive radiotherapy for prostate carcinoma
B Movsas1, A L Hanlon, T Teshima
1Fox Chase Cancer Center, Department of Radiation Oncology, Philadelphia, Pennsylvania 19111, USA.
Cancer
|September 26, 1997
Summary
Predictive models for prostate cancer outcomes are crucial for treatment stratification. This study found that after logarithmic transformation, several prostate-specific antigen (PSA)-based models demonstrated equal predictive accuracy for biochemical recurrence-free survival.
Area of Science:
- Urology
- Oncology
- Radiotherapy
Background:
- Clinically useful predictive models for prostate carcinoma are needed for patient stratification in treatment strategies.
- Existing models often use prostate-specific antigen (PSA)-based constructs to predict outcomes after radiotherapy.
- This study aimed to evaluate the predictive accuracy of different PSA-based models against clinical outcome data.
Purpose of the Study:
- To determine which of the six analyzed predictive models best fits independent clinical outcome data for biochemical freedom from failure (bNED control) in prostate cancer patients.
- To assess the predictive performance of various PSA-based constructs and prognostic groupings in a definitive radiotherapy series.
Main Methods:
- Analysis of six predictive models in a cohort of 421 patients with localized prostate carcinoma treated with definitive radiotherapy (median dose 74 Gy).
- Stepwise Cox proportional hazards multivariate analysis (MVA) was used to predict bNED control using covariates including PSA, Gleason's score, stage, and dose.
- Model adequacy was assessed using residual plots and Akaike's Information Criteria (AIC); logarithmic transformation of PSA was applied due to its log-normal distribution.
Main Results:
- Initially, the Pisansky et al. model showed the highest predictive value with the lowest AIC.
- However, following a logarithmic transformation analysis of PSA, all evaluated models demonstrated equivalent predictive accuracy for bNED outcome.
- Biochemical failure was defined as two consecutive PSA elevations ≥ 1.5 ng/mL, with a median follow-up of 34 months.
Conclusions:
- Multiple accurate models for predicting prostate carcinoma outcomes after radiotherapy have been developed.
- In this dataset, after logarithmic conversion, these models are essentially equally predictive of treatment success.
- Further validation in larger radiotherapy series is recommended to corroborate these findings.