Background Myocardial ischaemia occurs principally in the subendocardial layer, whereas conventional myocardial perfusion imaging provides no information on the transmural myocardial blood flow (MBF) distribution. Subendocardial perfusion measurements and quantification of the transmural perfusion gradient (TPG) could be more sensitive and specific for the detection of coronary artery disease (CAD). The current study aimed to determine the impact of lesion severity as assessed by the fractional flow reserve (FFR) on subendocardial perfusion and the TPG using [O-15]H2O positron emission tomography (PET) imaging in patients evaluated for CAD. Methods and results Sixty-six patients with anginal chest pain were prospectively enrolled and underwent [O-15] H2O myocardial perfusion PET imaging. Subsequently, invasive coronary angiography was performed and FFR obtained in all coronary arteries irrespective of the PET imaging results. Thirty (45%) patients were diagnosed with significant CAD(i.e. FFR <= 0.80), whereas on a per vessel analysis (n = 198), 53 (27%) displayed a positive FFR. Transmural hyperaemic MBF decreased significantly from 3.09 +/- 1.16 to 1.67 +/- 0.57 mL min(-1) g(-1) (P < 0.001) in non-ischaemic and ischaemic myocardium, respectively. The TPG decreased during hyperaemia when compared with baseline (1.20 +/- 0.14 vs. 0.94 +/- 0.17, P < 0.001), and was lower in arteries with a positive FFR (0.97 +/- 0.16 vs. 0.88 +/- 0.18, P < 0.01). ATPG threshold of 0.94 yielded an accuracy to detect CAD of 59%, which was inferior to transmural MBF with an optimal cutoff of 2.20 mL min(-1) g(-1) and an accuracy of 85% (P < 0.001). Diagnostic accuracy of subendocardial perfusion measurements was comparable with transmural MBF (83 vs. 85%, respectively, P = NS). Conclusion Cardiac [O-15]H2O PET imaging is able to distinguish subendocardial from subepicardial perfusion in the myocardium of normal dimensions. Hyperaemic TPG is significantly lower in ischaemic myocardium. This technique can potentially be employed to study subendocardial perfusion impairment in more detail. However, the diagnostic accuracy of subendocardial hyperaemic perfusion and TPG appears to be limited compared with quantitative transmural MBF, warranting further study.