Background: The immunostaining features of the androgen receptor (AR) have been studied in prostate cancer (CaP) to predict the outcome of androgen deprivation therapies. We have developed an automatic video color image analysis system for quantitation of AR expression in large samples of prostatic nuclei. Methods: Essential criteria of immunostaining have been examined to establish a linear relationship between AR protein content and mean optical density (MOD) of the immunoperoxidase-substrate reaction product. Titration of monoclonal AR antibody, F39.4.1, and concentration and reaction time of substrate were optimized using color video image analysis. The methodology was tested twice. First, CWR22 human CaP xenograft specimens, harvested from testosterone (T)-stimulated, castrated and T-resupplemented mice, were immunostained to demonstrate the dependence of AR expression on serum androgen levels. Second, AR expression was measured in archived clinical specimens. Results: In CWR22 tumor-bearing mice castrated for 6 days, AR MOD decreased to 57% of T-stimulated, intact mice. After 72 hrs of T treatment, AR MOD returned to the level measured in T-stimulated, intact mice. Sixteen radical prostatectomy specimens and 16 transurethral resection of prostate (TURP) specimens were double-labeled with F39.4.1 and anti-cytokeratin MAb (13 beta E12) specific for basal epithelial cells. Benign epithelial cells exhibited lower AR MOD in prostatectomy compared to TURF specimens (P < 0.01). Differences in AR immunostaining intensity may have resulted from differences in tissue fixation of whole organ versus small tissue specimens. Conclusions: AR immunostaining can be quantitated accurately using optimized immunohistochemical criteria and video image analysis. Cytometry 35:2-10, 1999. (C) 1999 Wiley-Liss, Inc.