Predicting the Risk of Pelvic Node Involvement Among Men With Prostate Cancer in the Contemporary Era

Purpose

The “Roach formula” for the risk of pelvic lymph node metastases [(2/3) ^∗ PSA + (Gleason score − 6) ^∗ 10] was developed in the early prostate-specific antigen (PSA) era. We examined the accuracy of this formula in contemporary patients.

Methods

We included men in the Surveillance, Epidemiology, and End Results Registry with a diagnosis of clinical T1c–T4 prostate cancer in 2004 who had a surgical lymph node evaluation, Gleason score (typically from prostatectomy), and baseline PSA level (n = 9,387). Expected and observed rates of node positivity were compared.

Results

Ninety-eight percent were clinical T1c/T2, and 97% underwent prostatectomy. Overall, 309 patients (3.29%) had positive lymph nodes. Roach scores overestimated the actual rate of positive lymph nodes in the derivation set by 16-fold for patients with Roach score less than or equal to 10%, by 7-fold for scores greater than 10–20%, and by approximately 2.5-fold for scores greater than 20%. Applying these adjustment factors to Roach scores in the validation data set yielded accurate predictions of risk. For those with Roach score less than or equal to 10%, adjusted expected risk was 0.2% and observed risk was 0.2%. For Roach score greater than 10–20%, adjusted expected risk was 2.0% and observed risk was 2.1%. For Roach score greater than 20–30%, adjusted expected risk was 9.7% and observed risk was 6.5%. For Roach score greater than 30–40%, adjusted expected risk was 13.9% and observed risk was 13.9%.

Conclusion

Applied to contemporary patients with mainly T1c/T2 disease, the Roach formula appears to overestimate pelvic lymph node risk. The adjustment factors presented here should be validated by using biopsy Gleason scores and extended lymphadenectomies.

Introduction

The “Roach formula” [(2/3) ^∗ PSA + (Gleason score − 6) ^∗ 10], where PSA is prostate-specific antigen level, is a widely used means of estimating the risk of pelvic lymph node involvement in men with prostate cancer (1). Initially based empirically on the 1993 Partin tables for patients treated with radical prostatectomy at Johns Hopkins University, Baltimore, MD, in the pre-PSA era (2), the formula was then validated for 282 patients who underwent radical prostatectomy at five San Francisco area hospitals from 1987–1991. In men with a calculated Roach score less than 15%, the observed incidence of nodal involvement was 6%, and in those with a Roach score of 15% or greater, the observed incidence of nodal involvement was 40% (1).

Based on these results and its ease of clinical use compared with the Partin tables, the Roach formula was widely adopted by radiation oncologists for decisions regarding elective nodal irradiation in patients with prostate cancer. For example, the Phase III randomized trial Radiation Therapy Oncology Group (RTOG) 94-13, which was designed in part to determine whether elective whole pelvic radiation would improve progression-free survival compared with prostate-only radiation, used Roach score greater than 15% as one of its entry criteria.

However, because of the advent of PSA screening, there has been a dramatic shift of patients with prostate cancer into earlier stages and lower PSA levels during the nearly 15 years since the Roach formula initially was derived (3). It is possible that the stage migration resulting from PSA screening and early detection has decreased the risk of occult lymph node metastasis for every PSA level and Gleason score. Therefore, the aim of this study is to determine whether the Roach formula accurately predicts the risk of lymph node involvement in a contemporary cohort of men with a diagnosis of prostate cancer in 2004 and included in the Surveillance, Epidemiology, and End Results (SEER) Registry.