The lung single photon emission tomographic (SPET) images of 985 consecutive patients referred for suspected pulmonary embolism were correlated with clinical outcome and angiography to evaluate the clinical usefulness of lung SPET compared to conventional planar ventilation/perfusion lung imaging. SPET interpretations followed the revised PIOPED criteria and clinical outcome was determined from referring physicians, hospital records, direct patient contact and county hall records. Patients were deemed to have had no clinically significant pulmonary embolism at the time of the SPET examination if, within the following 3 months: (1) the patient was alive and had no clinical evidence of pulmonary embolism or, (2) if deceased, pulmonary embolism was unlikely to have been the cause of death. Operating characteristics were based on the methods of Choi and of Simel. SPET interpretation was categorized as follows: high probability, 143 (14%); low probability, 840 (82%); intermediate, 41 (4%) (in contrast to PIOPED, with 39% intermediate interpretations). Pulmonary angiography was performed in only 4% of patients. Adequate follow-up data were available for 97% of patients. To facilitate comparison with PIOPED, either a high-probability or an intermediate-probability or an intermediate-probability study was considered to be a positive test, and either a low-probability or a normal study was considered to be a negative test. The sensitivity was 83% (PIOPED 82%), specificity 92% (PIOPED 52%), positive predictive value 62% (PIOPED 47%) and negative predictive value 97% (PIOPED 85%). The positive and negative predictive values have not been corrected for prevalence, which was approximately twice as high in the PIOPED study. Lung SPET provided accurate diagnostic information in 96% of patients and specificity was greatly improved compared to planar lung imaging reported in PIOPED. The diminished need for angiography greatly reduced the cost of evaluating patients suspected of having pulmonary embolism.