Prostate-specific antigen: A surrogate endpoint for screening new agents against prostate cancer?

BACKGROUND. An endpoint for clinical trials of prostate cancer which simpli fies traditional endpoints (response of measurable lesions, progression rat es, and death) is urgently needed. This is especially true for hormone-unre sponsive disease, for which many new drugs are presently in a development p hase. This paper presents a rationale for the use of prostate-specific anti gen (PSA) in clinical trials of progressive prostate cancer under endocrine treatment. METHODS. The study is based on 84 patients who progressed after radical pro statectomy or node dissection, of whom 24, showed increasing PSA levels und er subsequent endocrine treatment. An average linear relationship between ( log-transformed) PSA and time and a subject-specific deviation from this av erage relationship were assessed. The predictive value of the subject-speci fic parameters of the linear fit with respect to time to prostate cancer-sp ecific death was determined. The outcomes of the fitting procedure were use d to calculate sample sizes for future studies (duration, 6 months) using P SA increase over time in hormone-unresponsive prostate cancer as a marker f or treatment efficacy. RESULTS. The average PSA doubling time in this population was 4 months (cor responding time constant = 0.25). The assessed variance of the time constan ts equalled 0.04; the overall residual variance equalled 0.265. The subject -specific rate of change of the log-transformed PSA value in hormone-unresp onsive prostate cancer was a highly significant predictor of prostate cance r-specific death. This suggests the potential usefulness of PSA as an endpo int in trials of hormone-unresponsive prostate cancer. Depending on conditi ons chosen (e.g, desired power and changes in log PSA slope), 18-70 partici pants per arm will be necessary in future phase III studies. A suggestion ( algorithm) for the use of PSA in drug development is presented. CONCLUSIONS. Relatively small PSA-based trials in patients with hormone-unr esponsive prostate cancer are possible if a similar patient population is u tilized. As long as surrogacy is not established, such studies cannot be co nsidered conclusive with respect to effectiveness of treatment, but are lik ely to be useful as a screening tool for new drugs. Experimental confirmati on in human prostate cancer model systems of synergism between PSA decrease and tumor control by a given test treatment is likely to enhance the level of certainty of PSA-based drug evaluation. (C) 2000 Wiley-Liss, Inc.