This project was performed in cooperation with a research group working with the TwinsUK Registry. As part of this cooperation, the TwinsUK Registry was systematically screened for subjects with significantly different BMIs. Thirteen identical twin pairs (n = 26 subjects) with differences in BMI were recruited from more than 10,000 twin pairs from the TwinsUK Registry . As a threshold, a difference in BMI between twin pairs of 1.6 kg/m2 was chosen. The average difference in BMI between the twin pairs in this study was 6.3 ± 4.4 kg/m2. This small study population is due to the rarity of identical twins with significant differences in BMI (thirteen out of more than 10,000) due to the strong genetic contribution and matching lifestyles. The average difference between identical twins is usually less than one kg. The strength of this study, however, is that monozygote twin pairs were investigated. Twins from the registry were recruited from the general population through national media campaigns in the UK. St Thomas' Hospital research ethics committee approved the study (ethics number EC95/041), and informed consent was obtained from participating twins. None of the subjects in this study had a known history of cardiovascular disease or reported symptoms relating to cardiovascular disease. Zygosity was established in all subjects by standardized questionnaire and was confirmed by DNA fingerprinting. Exclusion criteria included: twins younger than 18 years, subjects with mental disorders or unable to give consent, claustrophobia, known allergies or contraindications to gadolinium, pregnancy or breastfeeding subjects, unstable angina pectoris (CCS IV), heart failure (NYHA IV), significant cardiac arrhythmia or impaired renal function (GFR < 30 ml/min).
Risk factor assessment
Demographics, medical history and medication were available as part of the Twins UK cohort study. Subjects were considered current smokers if they had smoked at least 1 cigarette a day within the last 30 days. Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg or current use of antihypertensive therapy. Diabetes was defined by self-report, a haemoglobin A1c level ≥6.5 %, or current use of hypoglycemic medication. Hypercholesterolemia was characterized by a fasting serum low-density lipoprotein cholesterol level ≥140 mg/dl on direct measurement or current use of lipid-lowering agents. The Framingham Risk Score (FRS)  was used to predict the 10-year risk of coronary heart disease in subjects without overt congenital heart disease. The BMI was calculated as body weight divided by body height squared (kg/m2). The Short Form 36 (SF-36) Health Survey questionnaire was used (see supplementary material). Intrapair differences in BMI between twin pairs were on average higher than 2.5 kg/m2. In all twins with an increase in BMI, an increase in FRS was measured. Mean BMI was 29.6 (range 20.1–40.5) versus 35.8 (range 26.3–52.5) and mean FRS was 13.4 (range 5–18) versus 16.4 (range 10–20). For details, please see Table 1.
MRI imaging and coronary vessel wall imaging
MRI examinations were performed using a 1.5-Tesla (T) MR system (Achieva, Philips Healthcare, Best, Netherlands). The system was equipped with high-performance gradients with a maximum amplitude of 30 mT/m and a slew rate of 150 mT/m/msec. All subjects were examined in the supine position with electrocardiographic leads placed on the anterior left hemi thorax for vector ECG triggering. A 32-element cardiac phased-array receiver coil was used. All MRI examinations were performed free-breathing without the use of sedation or general anaesthesia. First, a coronary MR angiography (MRA) was performed with the previously described navigator-gated free-breathing and cardiac-triggered three-dimensional (3D) steady-state free precession sequence . Imaging parameters of the coronary MRA included: a matrix of 256 x 256, a field of view of 320 x 320 mm, and the resulting acquired in-plane resolution was 1.25 x 1.25 mm. The acquired slice thickness was 3 mm and the reconstructed slice thickness was 1.5 mm. The acquisition window was limited 80 to 100 ms. Further parameters included a repetition time/echo time (TR/TE) of 4.2/2.1 ms and a flip angle of 110 degrees. The acquired number of slices was 24. Based on the MRA, a coronary navigator-gated, vector ECG-triggered, fat-suppressed, T1-weighted, 3D gradient-echo inversion recovery late gadolinium enhancement sequence (Fig. 1) was planned. Image acquisition was performed in mid-diastole. To compensate for breathing, navigator gating and tracking with a 5-mm gating window was performed [8, 9, 11, 15]. In all subjects, imaging using a 3D IR TFE was performed prior to and after administration of 0.2 mmol/kg of gadolinium-diethylenetriamine pentaacetic acid. Image parameters of the 3D IR TFE included: spatial resolution 1.25 x 1.25 x 3 mm; 20 slices were acquired; the repetition time was 6.1 ms; the echo time was 1.9 ms; and the flip angle was 30°. The acquisition window was limited to 80–100 ms. The inversion time (TI) was derived from a Look-Locker sequence aiming for maximal suppression of the blood pool signal. Parallel imaging was not applied. Overall scan time was 55 ± 3 minutes as imaging was performed prior to and 30 to 40 minutes following the administration of gadolinium-diethylenetriamine pentaacetic acid [8, 9, 11, 15].
For the analysis of the MRI data sets, the coronary artery tree was broken down into eight segments after a modified American Heart Association (AHA) classification. All segments were analysed individually (Fig. 2) [8, 16, 17]. MR images were reformatted with a semiautomatic vessel analysis tool (Soap-Bubble, release 5.0) . A technique, described by Oakes at al. , was adapted for the quantification of coronary vessel wall enhancement with OsiriX version 3.9.1 . Late gadolinium enhancement (LGE) was defined as areas in the vessel wall with a signal intensity higher than two standard deviations compared to the normal non-enhancing vessel wall. With the coronary MRA used as an overlay roadmap (automatic image fusion, Fig. 3), coronary vessel wall enhancement was measured. As the anatomy of the coronary artery tree slightly varies between twin pairs, regions of interest (ROIs) had to be adjusted manually for each twin pair. The variation of size for each segment was less than 10 % between twin pairs. ROIs were exactly copied within twin pairs to enable a high reproducibility, as the anatomical tree did not show variations within twin pairs. To determine the signal intensity (SI) of blood, a contour was drawn in the ascending aorta. Noise was measured in the area anterior to the chest wall by drawing a contour manually. Contrast enhancement within the coronary wall and the aortic wall was defined as the contrast to noise ratio CNR = ((SIwall - SIblood)/SDnoise) . Prior to the analysis, a trial assessment of five separate MR data sets for quality assurance was performed by the two readers together. For the analysis of the data sets, readers were blinded to the clinical information regarding the different twin groups. All data sets were analyzed independently in a blinded and random order. Late gadolinium enhancement measurements were repeated to calculate the intraobserver and interobserver variability. Two observers with >8 years of cardiovascular MRI experience analyzed data in an independent manner.
For statistical analysis, SPSS software (SPSS 17.0, SPS Inc., Chicago, IL, USA) was used. Results were expressed as mean ± SD. The Shapiro–Wilk test and visual confirmation was used to confirm the normal distribution of the data. A Student’s t test was applied for the comparison of continuous and discrete variables. Bonferroni correction was used to correct for repeated comparisons. A p value < 0.05 was considered to indicate statistical significance.