During a three month period, we prospectively enrolled consecutive patients scheduled for coronary x-ray angiography (CA) who had previously undergone adenosine stress CMR examination and shown subendocardial perfusion deficit. Patients without perfusion deficit and patients with more than subendocardial perfusion deficit in CMR were enrolled in equal proportions to serve as control groups. The exclusion criteria were an internal pacemaker or defibrillator, contraindications for adenosine infusion, or inability to give written informed consent. Written informed consent was obtained from all patients. Patients with a history of myocardial infarction or in whom Late Gadolinium Enhancement (LGE) could be visualized were excluded from the study. All anti anginal medication and caffein containing beverages were stopped at least 24 hours before CMR examination.

A 12-lead surface ECG was obtained for each patient. All patients were examined clinically and cardiovascular risk factors such as hypertension, diabetes mellitus, hypercholesterolemia, smoking and family disposition for CAD were assessed. In case of claustrophobia mild sedation with midazolame was offered.

All CMR studies were performed with a 1.5T magnetic resonance system (Signa Excite^®, GE Medical Systems, Milwaukee, USA) using an 8-element phased array surface coil (Cardiac coil, GE Medical Systems). Left ventricular (LV) parameters were measured using long-axis (two-chamber and four-chamber-views) functional steady-state free precession (SSFP) sequence images as part of our clinical routine protocol. After infusion of adenosine at a constant rate of 140 μg/kg per minute over three minutes (Spectris MR injector, Medrad, Indianola, USA) first-pass kinetic of a gadolinium-based contrast agent (Omniscan^®, GE Healthcare Buchler, Germany; 0.1 mmol/kg) was measured in 4 contiguous short axis orientations at every heart beat using a hybrid gradient echo/echo-planar pulse sequence (echo time 1.2 ms, flip angle 25°, slice thickness 8 mm, field of view 32–34 × 24–25.5 cm, matrix 128 × 96) as previously described 720. Echo time was reduced to 1.2 ms for reducing susceptibility artifact as sometimes seen in gradient echo sequences 21. Ten minutes after stress perfusion a second perfusion study with the same orientation and with the same setting was performed at rest without adenosine infusion. Ten minutes after this second bolus, LGE images were acquired by using an inversion-recovery prepared gated fast-gradient echo-pulse sequence (repetition time 6.7 ms; echo time 3.3 ms; flip 20°; inversion time individually adjusted; slice thickness 8 mm; rectangular field of view 30 to 34 cm; matrix 256 × 160). Again, three long axes, 4–5 short axes as planned in the perfusion study as well as contiguous short axes views using a 3D 20 slice sequence were acquired.

Two experienced investigators evaluated all CMR studies in consensus. If consensus could not be achieved, a third opinion was included. Image analysis was performed with the standard software provided by the CMR system manufacturer (Advantage Workstation, GE Medical System). Image analysis was performed visually for reducing the rate of false positive results due to rim artifacts as previously reported 22. Secondly, we compared stress to rest perfusion to reduce the potential rate of artifacts. If a deficit was equally present at stress and rest, if it did not follow the subendocardial border, if ghosting artifacts could be seen or if it "blinked" bright and dark it was not regarded as an evident hypoperfusion, but a potential artifact. Such cases were not included into the study. Segments were classified according to AHA recommendations 23 and evaluated for inducible hypoperfusion during the stress first-pass sequence and classified as "no hypoperfusion", "subendocardial hypoperfusion" 12 or "relevant hypoperfusion" indicative of relevant coronary artery stenosis 67 in comparison to rest perfusion images according to following criteria:

- Patients were classified as having small vessel disease, if diffuse subendocardial hypoperfusion 12 (affecting ≤ 1/3 of myocardial wall thickness and myocardial areas supplied at least by two different coronary arteries or circumferential perfusion deficit) and lasting for maximum five heart beats after maximal signal peak intensity in the LV cavity 24. Patients with a lesser degree of hypoperfusion (such as ≤ 1/3 of myocardial wall thickness in only one territory) were not included.

- Patients with regional perfusion deficit of > 1/3 wall thickness and lasting for more than five heart beats were classified as having CAD with ≥ 70% luminal narrowing 7. Patients with a lesser degree of regional hypoperfusion (such < 5 beats of stress perfusion defect regardless of affected myocardial wall thickness or ≥ 5 beats but ≤ 1/3 of myocardial wall thickness in one territory) were not included. See figure 1 for examples of our patients groups.

Patients with small vessel disease were analyzed as study patients, patients with no perfusion deficit as "control 1" and patients with perfusion deficit consistent with CAD as "control 2". Classification was performed before CA.

All patients underwent CA within 48 hours after CMR examination. Angiographic case study data were collected and analyzed for affected coronary arteries and degree of luminal reduction.

Afterwards stored angiograms were assessed for corrected TFC for each coronary vessel 14. TFC analysis was performed by the physician in charge with CA evaluation before CMR data disclosure. The corrected TFC is the number of cine frames required for contrast to first reach standardized distal coronary landmarks 1415. In our patients, CA was performed by hand injection of contrast medium via 5 French catheters and filmed at 12.5 frames/s. Since the original TFC is described for a cinefilm speed of 30 frames/sec, an adaptation was performed as previously reported 15. Therefore, frame counts were multiplied by 2.4 (correction factor). TFC for coronary arteries with normal flow is different in the left anterior descending (LAD) compared to the left circumflex (LCX) and right coronary artery (RCA) because of the larger vessel length. To adapt those differences and to provide comparable results of TIMI frame count for the different vessels, TIMI frame count of the LAD was divided through a correction factor of 1.7 15.

Data are reported as mean ± standard deviation. Continuous variables between groups were compared by t-test for unpaired observations. Nominal data were compared by Fisher's exact test. Categorial data were compared by Wilcoxon signed rank test for matched pairs. Correlation was assessed by means of correlation coefficient κ and regression coefficient R². In all cases, a p value < 0.05 was considered statistically significant. 95% confidence intervals (CI 95%) are given. Analyses were performed with commercially available statistic software (StatView 5).

CMR examination was performed in all 73 patients without relevant complications or adverse events. Image quality was sufficient for further analysis in all patients with primary investigators consensus in 71/73 cases and inclusion of a third opinion in two cases. Subendocardial perfusion deficit during stress perfusion in comparison to rest perfusion was found in 22 (30%) patients, who formed the study group. In 27 (37%) patients no perfusion deficit could be observed (control 1), a relevant perfusion deficit was visualized in 24 (33%) patients (control 2). Patient groups did not differ significantly in age or gender (see table 1).

CMR-derived LV parameters in study patients compared to controls were as follows: LV mass (g) 135 ± 36 (control 1: 119 ± 31, p: 0.17; control 2: 129 ± 36, p: 0.65), LV ejection fraction (%) 60.3 ± 8.8 (control 1: 61.7 ± 8.1, p: 0.60; control 2: 60.5 ± 7.2, p: 0.92), LV wall stress (N/m²x1000) 43.3 ± 8.8 (control 1: 40.0 ± 7.8, p: 0.25; control 2: 41.3 ± 8.6 g, p: 0.46).

Perfusion deficits in control 2 patients were detected in the LAD perfusion territory in 13 [54%], in the LCX territory in 15 [63%] and in the RCA perfusion territory in 12 [50%] cases, respectively.

CA was performed in all patients without relevant complications. Coronary one-vessel disease (≥ 70% luminal narrowing) was observed in 15 [21%], two-vessel disease in 9 [12%] and three-vessel disease in 6 [8%]. Mean corrected TFC for the LAD was 21.5 ± 4.6 frames, for the LCX 33.0 ± 7.3 and for the RCA 25.9 ± 4.9.

In our study patients (only subendocardial perfusion deficit on CMR exam) no coronary stenosis ≥ 70% could be shown. In contrast, all control 2 patients (relevant perfusion deficit in CMR) had coronary stenoses as visualized by CA. A highly significant correlation between classification of CMR perfusion deficit and degree of coronary luminal narrowing was found (see table 2).

Corrected TFC in all coronary arteries was significantly increased in study patients compared to both controls groups: Study patients vs. control 1 (no perfusion deficit): 25.1 ± 4.9 frames vs. 20.9 ± 4.2 frames in the LAD, p = 0.002; 39.1 ± 7.7 vs. 30.1 ± 6.1 in the LCX, p < 0.0001; 29.1 ± 5.5 vs. 24.4 ± 3.8 in the RCA, p = 0.001) and vs. control 2 (relevant myocardial ischemia): 18.7 ± 2.0 in the LAD, p < 0.0001; 30.7 ± 4.5 in the LCX, p < 0.0001; 24.6 ± 4.5 in the RCA, p = 0.004 (figure 2). Prolonged corrected TFC (values above mean of control 1) were present in all 22 study group patients (per vessel analysis: in 60/66 (91%)), in contrast to control 2 patients (per vessel analysis: in 12/72 (17%)). A good correlation between corrected TFC in LAD and LCX was found in our study patients (κ = 0.87; 95% CI [0.72–0.95]; p < 0.0001). Corrected TFC in control 1 and control 2 patients showed a lower correlation for LAD and LCX (κ = 0.71; 95% CI [0.45–0.86]; p < 0.0001 and κ = 0.50; 95% CI [0.11–0.75]; p = 0.01), respectively (figure 3). No correlation between LAD and RCA or LCX and RCA TIMI frame count was observed, respectively.

Study patients had more often hypertension (15 [68%] versus 10 [37%], p = 0.03) and diabetes (9 [41%] versus 4 [15%], p = 0.04) than control 1 patients. Control 2 patients also had more commonly hypertension (16 [67%], p = 0.03) and diabetes (11 [46%], p = 0.02) than control 1 patients. Patient groups did not differ for hypercholesterolemia or smoking nor did study and control 2 patients differ for the above mentioned cardiovascular risk factors.