We prospectively and consecutively included patients referred from the Department of Cardiology, Copenhagen University Hospital Bispebjerg from January to April 2022. The study was initiated when [15O]H2O production was approved at our site and during a four month overlap with continued clinical use of 82Rb. Thus all [15O]H2O PET scans were performed during the startup period. The center has a high-throughput with around 1,500 heart PET examinations per year16 and during the startup period, a number of exams were co-reviewed by experts from Turku PET Centre to ensure quality. To increase the likelihood of ischemia in the cohort, we included patients with typical anginal chest pain and with at least one of the following risk factors: a family history of cardiac disease (< 55 years for male and < 65 years for female family members), smoking, diabetes, hypertension, BMI > 30 or hyperlipidemia. Exclusion criteria were unstable angina, significant chronic obstructive lung disease or asthma, claustrophobia, acute severe illness or a significant language barrier. The study was approved by the Research Ethics Committee of the Capital Region of Denmark (ID: H-21016899) and written consent to participate were obtained from all individuals after receiving oral and written information according to the Helsinki declaration. All data were handled according to regulations by The Danish Data Protection Agency.

[15O]H2O was produced in two steps. First, a target gas mixture (97.5% 15N2, 2.5% O2) was continuously bombarded for a few minutes with a 7.8 MeV proton beam in a GenTrace cyclotron (GE, Uppsala, Sweden) dedicated to the production of 15O. Second, for administration of [15O]H2O, the target was mixed with the transport gas (N2 with 4% H2) and pushed into an oven with a 400 °C hot palladium wire. The resulting radioactive water vapor was fed into a bedside automated production system (Hidex RWG, Hidex Oy, Turku, Finland) consisting of a dual membrane system to mix physiological saline with [15O]H2O. The resulting radioactive saline solution was injected into the patient without further user interaction.

All subjects refrained from using caffeine-containing beverages and food or theophylline-containing medication for 24 hours before examination. Furthermore, phosphodiesterase type 5 inhibitors were withheld five days before examination, antithrombotic medicine containing Dipyridamole or Nicorandil two days before, extended-release nitrates 12 hours before and short-acting nitroglycerin two hours before examination. All patients were scanned using a Discovery 710 PET/CT scanner (GE Healthcare, Milwaukee, WI, USA). After a CT scan for attenuation correction and for anatomical localization, a 5 minutes dynamic emission scan in list mode was performed during resting condition after intravenous injection of 1100 MBq of 82Rb eluted from an 82Sr/82Rb generator (CardioGen-82; Bracco, Princeton, NJ). [15O]H2O PET in resting condition was mean performed earliest 10 minutes after 82Rb PET, and a dose of 394 MBq (range: 345-563 MBq) [15O]H2O was injected intravenously using a synthesis and injection system, Hidex RadioWaterGenerator (Hidex, Turku, Finland) and a 5-minutes scan was initiated simultaneously with the bolus arrival. After a 10 minutes interval to allow for decay of radioactivity, an identical PET sequence was performed during stress conditions induced by intravenous adenosine infusion (140 μg/kg/min) for 6 minutes. Adenosine was started 2 minutes prior to the stress PET scans to achieve maximum hyperaemia. We did not randomize the order of 82Rb and [15O]H2O PET as several patients could not cooperate to two adenosine infusions and we aimed to ensure a clinically useful 82Rb PET examination. Reconstruction of dynamic PET images was performed using ordered subset expectation maximization (OSEM) with Time of Flight (ToF) (2 iterations, 24 subsets and 6.4 mm in-plane filtering). For 82Rb a static reconstruction for the last 150 s was used. The Corridor 4DM software version 2018 (Invia Medical Imaging Solutions, Ann Arbor, MI, USA) was used for the analysis of 82Rb PET data, while CarimasCE software version 1.3.1. (Turku, Finland) was used for [15O]H2O PET data. Corridor 4DM estimates quantitative MBF from 82Rb K1 measurements using a modified version of Equation 2 according to Lortie17:

$$ K_ = MBF \cdot \left( }}} } \right) $$

(3)

82Rb PET was evaluated as part of daily clinical routine by a nuclear medicine specialist with > 10 years of experience in myocardial imaging. Previous studies have shown good inter-observer agreement.18,1982Rb PET were assessed visually using the ‘splash’ images using polar plots with relative differences compared to a normal database. The degree of relative defects within the myocardium was rated using 17 segments and 4 degrees of reduction, i.e. maximum total score 68. A stress defect score of 7 or more (i.e. ≈10%) of the myocardial wall was considered significant. Based on the clinical readings, the 82Rb PET was classified into four groups (normal, regional ischemia, globally reduced myocardial perfusion and myocardial scarring, see Table 1 for definitions). To further simplify data, the classifications were reduced to two groups of normal and regional ischemia, the latter included global reduction with suspicion of triple-vessel disease. The vascular territory involved was noted as left anterior descending artery (LAD), left circumflex artery (LCX) or right coronary artery (RCA).

[15O]H2O PET was assessed according to Danad15 with a cut-off of two neighboring segments with MBFstress≤ 2.3 mL/(min·g). Additionally, in line with 82Rb, total perfusion deficit (TPD) was calculated as a score of segmental reduction in percentage according to the following limits: normal (MBFstress > 2.3; score 0), mildly reduced (2.0< MBFstress ≤ 2.3; score 1), moderately reduced (1.7< MBFstress ≤ 2.0; score 2), severely reduced (1.4< MBFstress ≤ 1.7; score 3), and very severely reduced (MBFstress ≤ 1.4; score 4). TPD ≥ 10% was tentatively considered significant for ischemia. TPD is not a validated measure in [15O]H2O PET but was introduced in our clinic during the training period as we quickly realized that the threshold of two segments ≤ 2.3 mL/(min·g) identified too many patients with ischemia in the current patient population, which was also demonstrated in a recent study in patients with previous myocardial infarction or PCI.20 All [15O]H2O PET images were evaluated after the startup period had ended and [15O]H2O PET had been clinical routine for 2 months. Readers had > 10 years of experience in nuclear cardiology (UT and MK). [15O]H2O PET images were randomized and readers were blinded to the result of the 82Rb PET but with full access to all clinical data to simulate daily clinical routine. As for 82Rb, each [15O]H2O PET examination was classified into four groups based on the two segment threshold with MBFstress≤ 2.3 mL/(min·g) and two groups based on the aforementioned TPD of 10% (see Table 1 for definitions). The involved vascular territories (LAD, LCX and/or RCA) were noted. In case of discrepancy between the readers, consensus reading was performed. If a scan was classified with both a regional ischemia and global perfusion reduction or scarring, the regional defect overruled the other findings. To assess how previous heart disease affected the agreement, we performed a subgroup analysis of patients without known heart disease.

Data are reported with mean and standard deviation or median and interquartile range. Differences in heart rate response during adenosine infusion were compared using a paired Student’s t-test. Using SPSS (IBM SPSS Statistic version 25). Agreement was defined as the number of identical classification divided by the total number of patients. Cohen’s kappa was used to compare agreement between methods and between readers for the two and four group comparisons when sample sizes were sufficient. Kappa-values were evaluated according to Altman: κ ≤ .2: poor, .2 < κ < .4: fair, .4 < κ < .6: moderate, .6 < κ < .8: good, κ > .8: very good.21 McNemar’s test were used to test for agreement between tracers with the 2 × 2 contingency tables. Pearson’s correlation coefficient was used to compare the measured perfusion with the two methods. The K1 values were calculated from 82Rb MBF estimates using Equation 3 to compare the apparent 82Rb uptake in static images to the parametric [15O]H2O PET MBF images.