Latent profiles of global electrical heterogeneity: the Hispanic Community Health Study/Study of Latinos

Abstract

Aims: Despite the highest prevalence of stroke, obesity, and diabetes across races/ethnicities, paradoxically, Hispanic/Latino populations have the lowest prevalence of atrial fibrillation and major Minnesota code-defined ECG abnormalities. We aimed to use Latent Profile Analysis in the Hispanic Community Health Study/Study of Latinos (HCHS/SOL) population to obtain insight into epidemiological discrepancies.

Methods and results: We conducted a cross-sectional analysis of baseline HCHS/SOL visit. Global electrical heterogeneity (GEH) was measured as spatial QRS-T angle (QRSTa), spatial ventricular gradient azimuth (SVGaz), elevation (SVGel), magnitude (SVGmag), and sum absolute QRST integral (SAIQRST). Statistical analysis accounted for the stratified two-stage area probability sample design. We fitted a multivariate latent profile generalized structural equation model adjusted for age, sex, ethnic background, education, hypertension, diabetes, smoking, dyslipidaemia, obesity, chronic kidney disease, physical activity, diet quality, average RR' interval, median beat type, and cardiovascular disease (CVD) to gain insight into the GEH profiles. Among 15 684 participants (age 41 years; 53% females; 6% known CVD), 17% had an increased probability of likely abnormal GEH profile (QRSTa 80 ± 27°, SVGaz -4 ± 21°, SVGel 72 ± 12°, SVGmag 45 ± 12 mVms, and SAIQRST 120 ± 23 mVms). There was a 23% probability for a participant of being in Class 1 with a narrow QRSTa (40.0 ± 10.2°) and large SVG (SVGmag 108.3 ± 22.6 mVms; SAIQRST 203.4 ± 39.1 mVms) and a 60% probability of being in intermediate Class 2.

Conclusion: A substantial proportion (17%) in the Hispanic/Latino population had an increased probability of altered, likely abnormal GEH profile, whereas 83% of the population was resilient to harmful risk factors exposures.

Keywords: Cardiovascular risk; ECG; Electrophysiology; Hispanic or Latino ethnicity; Latent class analysis; Vectorcardiogram.

Graphical Abstract

Figure 1

Representative example of vectorcardiogram (A) and measured global electrical heterogeneity phenotype, presented by five metrics: spatial QRS-T angle, 3D spatial ventricular gradient magnitude and direction (azimuth and elevation), and spatial ventricular gradient scalar sum absolute QRST integral (B). A 3D spatial ventricular gradient vector is a QRST integral on orthogonal XYZ leads. Spatial ventricular gradient scalar sum absolute QRST integral is an absolute QRST integral on orthogonal XYZ leads or a vector magnitude signal. VMQTi = 0.62 ∗ SAIQRST. Spatial ventricular gradient elevation indicates spatial ventricular gradient direction relative to a horizontal plane on a Y-axis, upward from 0. Spatial ventricular gradient azimuth indicates spatial ventricular gradient direction relative to a frontal plane on a Z-axis, with positive backward and negative forward values. AUC, area under the curve; SVG, spatial ventricular gradient.

Figure 2

Predicted three latent profiles of global electrical heterogeneity variables. Histograms of (A) Spatial QRS-T angle, (B) spatial ventricular gradient azimuth, (C) spatial ventricular gradient elevation, (D) spatial ventricular gradient magnitude, (E) sum absolute QRST integral, and (F) average RR’ interval in non-overlapping three classes of participants with a posterior probability > 0.5 of belonging to Class 1 (green), Class 2 (yellow), and Class 3 (red). SAIQRST, sum absolute QRST integral; SVG, spatial ventricular gradient.

Figure 3

Estimated adjusted (for age, sex, Hispanic/Latino background, education attainment, hypertension, diabetes, smoking, dyslipidaemia, obesity, chronic kidney disease, physical activity, diet quality, average RR’ interval, median beat type, and cardiovascular disease prevalence) differences and 95% confidence interval of global electrical heterogeneity variables, associated with demographic and clinical cardiovascular risk factors and prevalent cardiovascular disease as compared with a reference category in non-overlapping three classes of participants with a posterior probability > 0.5 of belonging to Class 1 (green square), Class 2 (brown diamond), and Class 3 (red circle). CKD, chronic kidney disease; CVD, cardiovascular disease; HTN, hypertension; SVG, spatial ventricular gradient.

Figure 4

Fractional polynomial fit with 95% confidence interval of age (A–E), body mass index (F–J), and height (K–O) for prediction of spatial QRS-T angle (A, F, K), spatial ventricular gradient azimuth (B, G, L), spatial ventricular gradient elevation (C, H, M), spatial ventricular gradient magnitude (D, I, N), and sum absolute QRST integral (E, J, O). Unadjusted relationships between continuous variables age, body mass index, height (X-axes), and vectorcardiogram global electrical heterogeneity variables (Y-axes) are illustrated for populations of the latent Class 1 (green), Class 2 (brown and yellow), and Class 3 (red). BMI, body mass index; SAIQRST, sum absolute QRST integral; SVGaz, spatial ventricular gradient azimuth; SVGel, spatial ventricular gradient elevation; SVGmag, spatial ventricular gradient magnitude.