Metastatic carcinoma of the prostate: Identifying prognostic groups using recursive partitioning

Tracy R. Glass, Catherine M. Tangen, E. David Crawford, Ian Thompson

Research output: Contribution to journalArticlepeer-review

124 Scopus citations


Purpose: While in patients with metastatic prostate cancer median survival is approximately 30 months when treated with hormonal therapy, there is substantial interpatient variation. To explain better the outcome in patients with advanced disease we developed a set of prognostic groups within a large-scale clinical trial. Materials and Methods: Southwest Oncology Group Study 8894 was a randomized prospective clinical trial that compared orchiectomy and flutamide to orchiectomy and placebo. Using the technique of recursive partitioning we analyzed 5-year survival outcomes using a substantial number of patient, treatment and disease related variables to develop a set of prognostic groups with significant differences in survival. Results: Of 1,286 eligible patients 1,076 had sufficient data for analysis. The patient data set was split to allow prognostic group development in the first half of patients, followed by validation in the second half of patients accrued to the study. After pruning the regression tree 4 factors had a major impact on outcome, namely appendicular versus axial disease, performance status 0 versus 1 to 3, prostate specific antigen less than 65 versus 65 ng./ml. or greater and Gleason score less than 8 versus 8 or greater. Using these criteria 3 prognostic groups were developed, including a good (hazards ratio 1), intermediate (hazards ratio 1.8) and poor (hazards ratio 2.8) group. Five-year survival estimates in the 3 groups were 42%, 21% and 9%, respectively. Using the validation test set similar 5-year survival estimates for the 3 groups of 46%, 25% and 14%, respectively. Conclusions: These data from a large-scale randomized clinical trial provide a validated set of easily applied prognostic groups. We hope that our results may be validated by other investigators and we would encourage the future reporting of outcomes in patients with advanced prostate cancer using these prognostic groupings. Risk stratification is helpful for designing clinical trials and individual treatment, and it should be incorporated into future reports of outcomes of patients with metastatic disease.

Original languageEnglish (US)
Pages (from-to)164-169
Number of pages6
JournalJournal of Urology
Issue number1
StatePublished - Jan 1 2003
Externally publishedYes


  • Flutamide
  • Neoplasm metastasis
  • Orchiectomy
  • Prostate
  • Prostatic neoplasms

ASJC Scopus subject areas

  • Urology


Dive into the research topics of 'Metastatic carcinoma of the prostate: Identifying prognostic groups using recursive partitioning'. Together they form a unique fingerprint.

Cite this