Study Design and Population
The Netherlands Epidemiology of Obesity study (NEO) is a population-based cohort study including 6671 individuals. Men and women aged between 45 and 65 years with a BMI ≥ 27 kg/m2 living in the area of greater Leiden (The Netherlands) could participate in the NEO study. In addition, all inhabitants aged between 45 and 65 years from one municipality (Leiderdorp) were invited irrespective of their BMI, allowing for a reference distribution of BMI. Of the participants without contra-indications for MRI (most notably metallic devices, claustrophobia, or a body circumference of more than 1.70 m), a random subsample of approximately 20% of participants underwent cardiac MRI. Individuals completed a questionnaire with demographic, lifestyle, and clinical information. At the study center in the Leiden University Medical Centre (LUMC), all individuals underwent an extensive physical examination, including anthropometry, blood sampling (after an overnight fast), and electrocardiography. The present analysis is a cross-sectional analysis using the baseline measurements of the NEO study. We excluded participants in whom no cardiac MRI was performed or measurement of left ventricular mass (LVM) was missing, participants with abnormalities that could interfere with the electrocardiographic detection of LVH or the assessment of the spatial QRS-T angle, namely individuals with complete bundle branch block, ventricular pre-excitation (Wolff–Parkinson–White syndrome), previous myocardial infarction or a paced rhythm, and also individuals with missing values of the spatial QRS-T angle. Further details of the study design and population have been described in detail elsewhere . The Medical Ethical Committee of the LUMC approved the design of the study and all individuals gave their written informed consent.
Ethnicity was self-identified in eight categories and grouped into white and other. Body height and weight were measured without shoes and 1 kg was subtracted from the weight to correct for clothing. Waist circumference was measured with a horizontally placed flexible tape in the middle of the distance between the lowest rib and the iliac crest. Hip circumference was measured at the maximum circumference of the buttocks. Brachial blood pressure was measured in a seated position on the right arm using a validated automatic oscillometric device (OMRON, Model M10-IT, Omron Health Care Inc, Chicago, IL, USA). Blood pressure was measured three times with 5 min of rest between consecutive measurements. The mean systolic and diastolic blood pressure was calculated. Blood samples were drawn after an overnight fast of 10 h. Fasting glucose was measured with the enzymatic colorimetric method (Roche Modular Analytics P800, Roche Diagnostics Mannheim, Germany).
After a resting period of at least 10 min, 12-lead electrocardiograms were obtained using a Mortara Eli-350 (Mortara Instrument Inc., Milwaukee, WI, USA). The raw data were extracted and transferred to the University of Glasgow electrocardiogram (ECG) core lab where ECGs were automatically processed and Minnesota codes were assigned using the University of Glasgow ECG analysis program . We investigated four conventional electrocardiographic criteria for LVH (continuous variables): two widely used voltage index electrocardiographic criteria, namely Sokolow–Lyon voltage and Cornell voltage, and two voltage-duration product criteria, namely Sokolow–Lyon product, and Cornell product [15,16,17,18]. Sokolow–Lyon voltage was defined as |SV1| + RV5/6 and Sokolow–Lyon product as Sokolow–Lyon voltage × QRS duration. Cornell voltage was defined as RaVL + |SV3| with 600 μV added for women and Cornell product was defined as Cornell voltage × QRS duration. T-wave abnormalities were defined as Minnesota Codes 5-1 or 5-2.
Standard 10-s ECGs were each stored in an 8-lead (I, II, II, V1–V6), 5000 sample comma-separated-value file. The Kors matrix was used to calculate vector cardiograms from the eight independent ECG leads . ECGs and vector cardiograms were analyzed using the automatic MATLAB-based (The MathWorks, Natick, MA, USA) program BEATS and the semiautomatic program LEADS [20, 21]. BEATS was used to detect the timings of all QRS complexes and calculated R–R intervals (ms). The QRS and T integral vectors were approximated by calculating the numerical sum of x–y–z deflections (amplitudes of positive deflections are added and those of negative deflections subtracted). The spatial QRS-T angle was defined as the angle (°) between the integral QRS vector and the integral T vector.
Magnetic Resonance Imaging
In 1150 participants, LVM was assessed using cardiac magnetic resonance imaging. The heart was imaged in the short-axis orientation by using ECG gated breath-hold balanced steady-state free precession imaging. Using in-house-developed software packages (MASS and FLOW; LUMC, Leiden, The Netherlands), image postprocessing was performed and decisions were based on consensus between two experienced observers.
LVM was indexed by height1.7 to obtain left ventricular mass index (LVMI). LVM was not indexed by body surface area to prevent underestimation of the prevalence of LVH in the NEO study population, which has a high prevalence of overweight and obese individuals . Cut-offs for LVH were based on the sex-specific upper limits of normality (95th percentile) from a subgroup of 252 healthy individuals from the NEO study, with a BMI < 30 kg/m2, normal blood pressure (< 135/< 85 mmHg and no use of antihypertensive medication), no history of cardiovascular disease and normal glucose metabolism (no self-reported diabetes mellitus I or II or medication and fasting plasma glucose < 7 mmol/l). LVH was defined as LVMI > 51.9 g/m1.7 in men and LVMI > 41.8 g/m1.7 in women.
Adjustments for the oversampling of individuals with BMI ≥ 27 kg/m2 in the NEO study were made to correctly represent baseline associations in the general population. This was done by weighting individuals towards the BMI distribution of participants from the Leiderdorp municipality, whose BMI distribution was similar to the BMI distribution of the general Dutch population. Baseline characteristics are presented as mean (SD), median (IQR), or as percentage.
First, univariate discriminative performance for LVH of the conventional electrocardiographic criteria, namely Sokolow–Lyon voltage, Sokolow–Lyon product, Cornell voltage, and Cornell product was assessed using the area under the curve (AUC) of the receiver operating characteristic (ROC) curve. The AUC reflects how well individuals are classified as having LVH or no LVH. Also, sensitivities of the conventional electrocardiographic criteria were determined at a specificity of 90%, since especially specificities of 90–100% are clinically relevant for detection of LVH. Second, univariate discriminative performance for LVH of age, sex, BMI, waist circumference, and waist:hip ratio was assessed with the AUC. Then, stepwise logistic regression analysis with an entry criterion of p < 0.05 and removal criterion of p > 0.10 was performed with LVH as dependent variable and each conventional electrocardiographic criterion separately with addition of the variables age, sex, BMI, waist circumference, and waist:hip ratio as independent variables. AUC, R2, and sensitivity at 90% specificity of the selected models were assessed. Univariate discriminative performance for LVH of T-wave abnormalities (dichotomous) and the spatial QRS-T angle was also assessed using the AUC. Finally, for each conventional electrocardiographic criterion separately, the best performing models were determined, consisting of a combination of the best performing measure of body fat and the best of T-wave abnormalities and spatial QRS-T angle. AUC, R2, sensitivities at a specificity of 90% and calibration plots were reported for the new models for LVH detection. Furthermore, the internal validity of the estimated AUC values was assessed using bootstrapping. Data were analyzed using STATA (StataCorp, College Station, TX, USA), version 14.
Compliance with Ethics Guidelines
All procedures performed in studies involving human participants were in accordance with the ethical standards of the Medical Ethical Committee of the LUMC and the study conformed with the Helsinki Declaration of 1964, as revised in 2013. Informed consent was obtained from all participants.