A statistical model was developed: 1) to compare the efficacy of patient-controlled analgesia (PCA) and traditional intramuscular (IM) opioids for pain relief in 40 patients after total knee replacement and, 2) to evaluate pain cycles associated with each technique. Hourly visual analog pain scores were subjected to two-way analysis of variance (ANOVA) and time-series analysis. Hourly verbal analog pain scores were used to determine predominant pain levels. According to ANOVA, PCA was no more effective than were IM opioids. Time-series analyses documented a complete cycle of pain every 5.3 hours in patients receiving IM opioids but no pain cycle with use of PCA. Analysis of PCA verbal analog scores demonstrated self-administration of opioids to "moderate" levels of pain relief with use of PCA, not to complete analgesia. These results suggest that certain patients may not envision complete postoperative analgesia as being possible. Hence, they self-administer opioids for pain relief with PCA according to their expectations. Population characteristics may modify PCA efficacy. These characteristics should be delineated and the use of PCA targeted to appropriate patients.