Background: It remains unknown whether estimation of the relative stress perfusion deficit offers added value in the prediction of significant coronary artery stenosis in myocardial perfusion imaging with [¹⁵O]H₂O positron emission tomography (PET) in a population with high prevalence of established cardiac disease. Methods: During eight months, we consecutively included all patients undergoing [¹⁵O]H₂O PET and subsequent invasive coronary angiography (ICA). Significant stenosis was defined from ICA as fractional flow reserve ≤.8 or coronary artery narrowing of ≥70%. We calculated absolute and relative total perfusion deficits (aTPD and rTPD, respectively) as semiquantitative measures of the extent and severity of reduced stress perfusion. A multivariate logistic regression analysis was performed to test the adjusted associations (odds ratio (OR) with 95% CI) with significant coronary artery stenosis. Results: Of 800 patients undergoing [¹⁵O]H₂O PET, 144 underwent ICA, where 142 patients had aTPD of ≥3% and 79 (55%) of these had at least one significant stenosis. In an adjusted analysis, rTPD (OR_{10% increase} = 2.12 (1.44-3.12), P < .001), previous coronary artery bypass grafting (CABG) (OR = .11 (.03-.36), P < .001) and reduced left ventricular ejection fraction (LVEF) (OR = .25 (.08-.84), P = .02) were independently associated with significant stenosis, whereas the association with aTPD (OR_{10% increase} = 1.14 (.98-1.32), P = .08) was modest. Conclusions: In the presence of an absolute perfusion deficit (aTPD of ≥3%), rTPD may improve the prediction of significant stenosis in a heterogeneous population of patients examined with [¹⁵O]H₂O PET. Furthermore, previous CABG and reduced LVEF are associated with nonstenotic perfusion deficiencies, suggesting caution when interpreting myocardial perfusion imaging in such patients.