Study population and study design

The study was approved by the Regional Ethics Committee at Lund University Sweden (Dnr 2015/507) and conformed to the Declaration of Helsinki. Written informed consent was given by all participants. A subset of male participants in the previous MEDIM cohort without any signs or established risk factors for CVD were invited to participate in the study. The cohort has been described in detail previously¹⁴. Briefly, the MEDIM cohort is a cross-sectional study conducted 2010–2012 including 2155 Iraqi or Swedish born residents of Malmö 30–75 years of age. In total 259 healthy, never smokers, non-obese males without CVD risk factors were identified from the baseline study and invited by mail to participate in the present study. A total of 18 Iraqi born and twelve native Swedish born males, of Caucasian origin and none had parents originating outside of Europe, fulfilled inclusion criteria and accepted to participate in this CMR sub-study, see Fig. 1A. Exclusion criteria included known CVD, diabetes, obesity, kidney disease, history of asthma, smoking or any active medication. Also, subjects with clear regional perfusion deficits on CMR were excluded. Subjects were recruited between 2017 and 2019 and examined at Lund University Hospital, Lund, Sweden. All patients underwent rest and adenosine stress CMR and adhered to a 24-h caffeine restriction prior to scanning.

Baseline characteristics and blood samples

Prior to the CMR exam, height, weight, waist circumference and blood pressure (in the supine position) were assessed. Fasting glucose, hemoglobin A1c, creatinine, cholesterol, triglyceride, high-density lipoprotein, low-density lipoprotein and apolipoprotein B/A1 in serum and urine albumin/creatinine were analyzed fasting as previously described^5,14. Homeostasis model assessment (HOMA) was used to estimate both insulin resistance (HOMA-IR) and beta cell function (HOMA-β)⁵. An electrocardiogram was acquired prior to the CMR exam.

For objectively calculating the ten year risk of fatal CVD of study participants, the risk scoring systems “Framingham risk score for hard coronary heart disease”¹⁵ and the “systematic coronary risk evaluation (SCORE)”¹⁶ were used including the variables sex, age, systolic blood pressure, total cholesterol and smoking status.

CMR image acquisition

All images were acquired on a Magnetom Aera 1.5 T CMR system (Siemens Healthcare, Erlangen, Germany), see Fig. 1B for an overview of the CMR protocol.

Left ventricular volumes and function

Left ventricular volumes and function were assessed by cine imaging using a steady-state free precession sequence (SSFP) in breath hold both in short-axis and long-axis projections (2-, 3- and 4-chamber views). Typical imaging parameters included repetition time (TR) = 2.7 ms, echo time (TE) = 1.2 ms, flip angle 60°, spatial resolution 1.5 × 1.5 × 8 mm with no slice gap and field of view (FOV) 270 × 320 mm².

Quantitative first pass perfusion

A basal, a mid-ventricular and an apical short-axis image were acquired at rest and during adenosine stress (Adenosin, Life Medical AB, Stockholm, Sweden, 140 µg/kg/min infusion) using qFPP imaging during administration of an intravenous bolus of contrast agent (0.05 mmol/kg, infusion rate 4 ml/s, Gadobutrol, Gadovist, Bayer AB, Solna, Sweden). Stress images were acquired first during 60 heart beats, starting three minutes after initiation of intravenous adenosine infusion and rest images were acquired approximately 15 min later. Measurements in low-resolution proton density images optimized for high gadolinium concentration in the LV blood pool were used to calculate the arterial input function¹⁰. After motion correction and conversion of signal intensities to gadolinium concentration, myocardial perfusion in ml/min/g was derived on a per-pixel basis^10,13. Typical imaging parameters were: SSFP single shot readout, TE 1.0 ms, TR 2.5 ms, flip angle 50°, FOV 360 × 270 mm², slice thickness 8.0 mm, parallel acquisition technique factor 3, acquisition time per single shot slice 142 ms and saturation delay 105 ms.

Coronary sinus flow

Images of the coronary sinus were acquired at rest and during adenosine stress for quantification of global perfusion. Rest images were acquired first and stress images were acquired immediately after acquisition of the qFPP images, approximately 5 min later. A breath-hold phase-contrast CMR with retrospective ECG triggering was used. Typical imaging parameters were: TR 5 ms, TE 2.8 ms, flip angle 20°, parallel imaging factor 2, a reconstructed spatial resolution of 1.6 × 1.6 × 8 mm and velocity encoded (VENC) factor 80 cm/s for rest and 120 cm/s for stress.

Fibrosis and extracellular volume

A modified Look-Locker (MOLLI) T1 sequence was used to generate T1 maps before and after gadolinium contrast agent administration. An inline extracellular volume map (ECV-map) was created after manual input of the hematocrit. Three short axis-slices were acquired at a basal, mid-ventricular and apical level. Macroscopic fibrosis was studied using a free breathing, motion corrected, late gadolinium enhancement (LGE) sequence acquired both in short and long axis projections. The inversion time was chosen to null remote myocardium. Specific parameters for the LGE sequence were: TR 2.8 ms, TE 1.2 ms, flip angle 50°, FOV 360 × 270 mm, resolution 1.4 × 1.4 × 8 mm, no slice gap.

CMR image analysis

All images were analyzed using the software Segment (v2.0 R5378), Medviso AB, Lund, Sweden)¹⁷, with the observer blinded to subject identity. For all images, endo- and epicardial borders were manually delineated. Left ventricular (LV) volumes, ejection fraction and left ventricular mass (LVM) were quantified from the short-axis cine stack. Interobserver variability for LVM was analyzed in six Iraqi and six Swedish controls. Rate pressure product (RPP) was calculated as heart rate × systolic blood pressure for both rest and stress.

Quantitative first pass perfusion and ECV

For qFPP and ECV images region of interest (ROI) were drawn 10% away from the endo- and epicardial borders to avoid inclusion of blood pool or extra-cardiac structures. Myocardial perfusion (ml/min/g)^10,13 and ECV^11,12 (%) were assessed manually by drawing a ROI in each short-axis slice. The absolute perfusion could then be extracted directly from the ROI as each pixel contained information on absolute perfusion. Myocardial perfusion and ECV were assessed globally by averaging all acquired short-axis slices. Each short-axis slice in qFPP images at rest and stress was also divided as endocardial (inner 50%) and epicardial (outer 50%) regions. Resting perfusion was corrected for rate pressure product (RPP) as resting perfusion × 10,000/((heart rate at rest) × (systolic blood pressure at rest). Myocardial perfusion reserve (MPR) was calculated as the ratio of stress to rest perfusion and MPR using RPP corrected resting perfusion was also calculated. Transmural endocardial-to-epicardial MPR gradients were also calculated. Recently, a cutoff of global stress qFPP < 2.25 ml/g/min was proposed to be suggestive of microvascular disease¹⁹. We also calculated the myocardial perfusion per mass of myocytes as qFPP/(1-ECV).

Sinus coronary flow

Global myocardial perfusion (ml/min/g) was also quantified as CSF (ml/min)/LVM (g)²⁰. LVM for CSF quantification was calculated by including papillary muscles and trabeculae by visual thresholding. Interobserver variability for CSF at rest and stress was analyzed in six Iraqi and six Swedish controls and for the addition of papillary muscles and trabeculae to LVM in all subjects.

Late gadolinium enhancement

Focal fibrosis was assessed on short-axis LGE images by the EWA (expectation maximization weighted intensity a priori information) algorithm [21].

Statistical analysis

Continuous data are expressed as mean ± standard deviation (SD). Mean values between groups were assessed using Fischer’s exact test, paired or unpaired t-test as appropriate in normally distributed data. The relationship between continuous variables were assessed with Pearson’s correlation coefficient. Bias according to Bland–Altman was used to compare qFPP and sinus coronary flow and for interobserver analysis. Univariable association with myocardial perfusion for covariables that differed between groups was analyzed using linear regression. All statistical analyses were performed using IBM SPSS Statistics (IBM SPSS Statistics 23, IBM, New York, USA) and Graph Pad Prism 7.0 software (Graph Pad Software, Inc., La Jolla, CA, USA). Differences with a P value < 0.05 were considered statistically significant.