A direct comparison of the analgesic activities of heroin and hydromorphone was carried out in cancer patients with postsurgical pain. Intramuscular doses of 5 and 10 mg of heroin were compared with 1 and 2 mg of hydromorphone in a randomized, double-blind, 4-point parallel group assay. Design innovations in the study provided that about half the patients would receive prior repeated doses of the same drug as the test medication and half would receive the alternate medication. Both test drugs were found to be potent, relatively short acting analgesics with similar profiles of action. Hydromorphone was about 5 times as potent as heroin on a milligram basis. The comparison of those patients who had repeated doses of the same treatment prior to the test dose and those who had repeated doses of the alternate drug demonstrated no significant effect on the relative potency estimates. Side effect occurrence was similar for both drugs, with sleepiness the most prominent effect. The study supports the view that hydromorphone and heroin produce similar clinical effects and that either drug may adequately substitute for the other. Covariate analysis indicated that time since last analgesic was positively related to analgesia and amount of prior opioid had a negative relationship. To a lesser extent, increase in patient age was associated with an increase in analgesic scores. Taking these covariates into account served to increase the sensitivity of the analysis.