Medically significant concentrations of prostate-specific antigen in serum assessed

We used the method of Rudolph et al. (Clin Chem 1988; 34:2031-8) to find information in the data from correlated determinations of acid phosphatase (PAP, EC 3.1.3.2; DuPont aca) and prostate-specific antigen (PSA, Hybritech). We described there how we assign medical decision limits for two or more correlated variables and convert the database to a binary coded message, allowing separation of a selected disease class with minimum error. The decision point, analogous to a percentile upper limit on the ordered values of each variable in the reference group, satisfies the maximum entropy constraints of reference, producing a minimum entropy for the binary coded patient database. We found maximum entropy decision points at PAP = 0.75 U/L and PSA = 22.8 micrograms/L. Patients with PSA values exceeding 22.8 micrograms/L had no benign prostatic disease except for five patients with benign prostate hyperplasia (BPH) with adjacent colon carcinoma (95.3), BPH with infarction (27.6), BPH (23.4) 28.1), or acute prostatitis (34.6). We consider PSA exceeding 22.8 micrograms/L as indicative of carcinoma of the prostate, stage C or D, in the absence of disconfirming evidence. Another decision value for PSA is 11.3 micrograms/L. This bounds the region between 11.3 and 22.8 micrograms/L, where the frequency of BPH is 1.5 times that for adenocarcinoma. At PSA less than 11.3 micrograms/L there is a high frequency of BPH. PSA concentration is not correlated with prostatic size (mass) or with prostatitis. A metastatic carcinoma is as likely to be nonprostatic as prostatic when the PSA concentration is less than 11.3 micrograms/L.