To assess the use of modified PIOPED scintigraphic criteria for lung scan (V/Q) interpretation to detect pulmonary embolism (PE), we prospectively applied these criteria in suspected PE patients referred for V/Q from 9/1/92 to 2/7/94. PIOPED criteria were modified by placing a moderate segmental perfusion mismatch in the intermediate instead of low probability of PE category and using the "stripe sign." METHODS Patients were studied by six-view V/Q imaging using 74 MBq (2 mCi) 99mTc-MAA followed by 148-370 MBq (4-10 mCi) 99mTc-DTPA aerosol, contrast pulmonary selective angiography and Doppler sonography with leg compression as needed. Patients underwent follow-up (mean 13.9 mo) to detect subsequent thromboembolic events. In this study group, 1000 patients were studied by V/Q followed by angiography in 133 patients. RESULTS The distribution of V/Q-assigned PE probabilities was: high probability 5.7%, intermediate 17.4%, low 41.4% and normal 35.5%. Group A patients (133) underwent angiography, which resulted in the determination of a 27.1% PE prevalence. Group B patients (867) did not have angiograms; the clinical prevalence of PE was 7.5%. In the total study population, the positive predictive value of a high probability V/Q study for PE (10.1% prevalence) was 98.2%, intermediate probability V/Q study for PE was 24.1% and a low probability study for PE was only 0.5%. CONCLUSIONS Modified PIOPED V/Q interpretation criteria afford better angioproven PE discrimination between intermediate (31.8% PE prevalence) and low (5.5% PE prevalence) probability V/Q results than reported for PIOPED intermediate (32.6% PE prevalence) and low (16.3% PE prevalence) probability V/Q interpretation criteria.