This randomized double blind study investigates the relative efficacies of controlled analgesia (PCA) regimens in three different patient groups: epidural diamorphine 2.5 mg followed by PCA bolus 1 mg with a 20-min lockout (Gp1), subcutaneous diamorphine 2.5 mg followed by PCA bolus with a 10-min lockout period (Gp2) and epidural diamorphine 2.5 mg in 4 mL of 0.125% (w/v) bupivacaine followed by a PCA bolus of 1 mg diamorphine in 4 mL 0.125% (w/v) bupivacaine with a 20-min lockout (Gp3). Patients were evaluated at 0, 1, 2, 3, 4, 8, 12, 16, 20, 24 and 48 h. Patients in Gp2 consumed significantly more diamorphine than those in Gp1 or Gp3 (P < 0.05), but their pain scores were higher only at 1, 2 and 3 h (P < 0.05) with respect to Gp3 and at 1 h with respect to Gp1. Fewer side effects (sedation, pruritus and nausea as assessed by anti-emetic requirements) occurred in Gp2 compared to Gp1 (P < 0.05). Fewer patients in Gp2 required catheterization than in Gp3 (P < 0.05). This study indicates that the use of PCA epidural diamorphine, either alone or in combination with bupivacaine, reduces the dose requirement for analgesia but offers little clinical advantage over subcutaneous PCA diamorphine.