The objective is to compare the performance of immunohistochemistry (IHC) with that of reverse transcription (RT)-PCR for detecting clinically significant micrometastases in histopathologically normal archival pelvic lymph nodes (PLN) removed at radical prostatectomy from men with locally advanced nonmetastatic prostate cancer. We stained 1864 fixed, paraffin-embedded PLNs from 199 pT(3)N(0)M(0) prostate cancer patients for prostate-specific antigen (PSA) and cytokeratin. We also assessed human glandular kallikrein (hK2) expression in a subset of 164 patients. In addition, all PLN specimens were assayed for hK2 mRNA using a previously described RT-PCR assay. PSA and cytokeratin were expressed in the same 13 of 199 (7%) cases; hK2 was expressed in 3 of 164 (2%) cases. PSA/cytokeratin and hK2 expression were associated with cancer involvement of seminal vesicles, higher Gleason sum, and a positive RT-PCR-hK2 assay result. In standard postoperative multivariable models, IHC-PSA/IHC-Cytokeratin or IHC-hK2 was associated with prostate cancer progression, development of distant metastases, and prostate cancer-specific survival. However, when RT-PCR-hK2 assay result was added to the models, it was the sole predictor of clinical outcomes. Although IHC-PSA/IHC-Cytokeratin and IHC-hK2 were more specific for identifying patients who would suffer biochemical progression and develop metastases and who would ultimately die of prostate cancer, RT-PCR-hK2 was more sensitive and accurate. Although IHC for PSA, cytokeratin, and hK2 appear to be more specific methods for detecting biologically and clinically significant prostate cancer micrometastases in histopathologically normal PLN, RT-PCR-hK2 appears to be a more sensitive method that maintained a reasonable specificity. In pT(3)N(0) patients, a positive RT-PCR-hK2 assay result when performed on PLN was the strongest predictor of clinical outcomes after radical prostatectomy.