Thirty-one patients with bidimensionally measurable hormone-refractory prostatic cancer received trimetrexate (TMTX). Serial values of prostate-specific antigen (PSA) and acid phosphatase (SAP) were correlated with response. Five patients (17%; 95% confidence interval, 3% to 30%) achieved a partial remission for a median of 3 months (range, 3 to 7.5 months). Marker levels showed large variations with no discernible patterns. Serial PSA and SAP in 19 patients with abnormal baseline values showed a correlation with measurable disease response in only 68% (13 of 19) and 47% (nine of 19) of patients, respectively. Values were then smoothed using an exploratory data analysis technique of running medians and averages. Trends in marker changes were much more apparent. Several "decision rules" were evaluated for use of markers as indices of disease progression. A 50% increase from the patient's minimum value in either PSA or SAP on two successive determinations correlated with progression in 90% of cases in this trial. TMTX has modest activity in prostatic cancer, and further trials are not warranted. Biochemical markers do not uniformly reflect disease activity in hormone-refractory disease, and changes in biochemical markers must be interpreted cautiously when used as the sole end point to assess efficacy in clinical trials.