OBJECTIVES: Determine the impact of correcting for different definitions of prostate-specific antigen (PSA) recurrence when analyzing relative effectiveness of seven treatments for localized PC; Emphasize the importance of clinical input and real-world data in ensuring modeling accuracy. METHODS: We conducted two lifetime cost-utility analyses comparing men undergoing open, laparoscopic, or robot-assisted radical prostatectomy, 3-dimensional conformal or intensity-modulated radiation therapy, brachytherapy, or combined external-beam radiation and brachytherapy.  We constructed a Markov model quantifying lifetime costs and quality-adjusted life years (QALYs) for men with localized PC in low-, intermediate-, and high-risk strata.  Post-treatment PSA recurrence is defined differently for surgical versus radiation treatments.  In Study 1, we made clinically-based corrections in time between recurrence and metastasis, the basis for time-to-disease-progression (biochemical failure [BCF]) calculations, to account for the different definitions.  In Study 2, these corrections were not made.  RESULTS: In “corrected” analyses, surgery tended to yield more QALYs and lower costs than radiation, with minimal differences among surgical modalities. Compared to a base case assuming a 4-year increment between the surgery and radiation modalities in the median time from BCF to metastasis, “uncorrected” analyses yield surgical costs relatively overstated by 3 to 11%; QALYs understated by 1 to11%; life expectancy understated by <1% to 8.5%; metastasis overstated by 17 to 36%; and PC death overstated by 17% to 43%.  Results are generally greater with larger corrections and outcomes discounted to net present value.  CONCLUSIONS: “Without correction” results do not affect conclusions about relative costs for all patients or QALYs for low-risk patients.   They do affect QALY conclusions for intermediate- and higher-risk patients, yielding erroneous conclusions on the relative superiority of radiation versus surgical treatments.  Without correction, the same incorrect conclusions would be reached for survival, metastasis, and PC death.  Literature and real-world data were used to validate the corrections.