Regional Circumferential Strain is a Biomarker for Disease Severity in Duchenne Muscular Dystrophy Heart Disease: A Cross-Sectional Study

Abstract

The aim of this study is to determine the contribution of strain ε _cc in mid left ventricular (LV) segments to the reduction of composite LV circumferential ε _cc in assess severity of duchenne muscular dystrophy (DMD) heart disease as assessed by cardiac magnetic resonance imaging (CMR). DMD patients and control subjects were stratified by age, LV ejection fraction, and late gadolinium enhancement (LGE) status. Tagged CMR images were analyzed for global ventricular function, LGE imaging, and composite and segmental ε _cc. The relationship between changes in segmental ε _cc changes and LGE across patient groups was assessed by a statistical step-down model. LV ε _cc exhibited segmental heterogeneity; in control subjects and young DMD patients, ε _cc was greatest in LV lateral free wall segments. However, with increasing age and cardiac disease severity as demonstrated by decreased EF and development of myocardial strain the segmental differences diminished. In subjects with advanced heart disease as evidenced by reduced LV ejection fraction and presence of LGE, very little segmental heterogeneity was present. In control subjects and young DMD patients, ε _cc was greatest in LV lateral free wall segments. Increased DMD heart disease severity was associated with reduced composite; ε _cc diminished regional ε _cc heterogeneity and positive LGE imaging. Taken together, these findings suggest that perturbation of segmental, heterogeneous ε _cc is an early biomarker of disease severity in this cross-section of DMD patients.

Appendix

Descriptive Phase

We computed a difference in segmental ε _cc across groups and segments via the Eq. (1) as

$$\varDelta \varepsilon_{{{\text{cc}}_{i} }} = \sum\limits_{j = 1}^{5} {\left( {\varepsilon_{{{\text{cc}}_{ij} }} - \varepsilon_{{{\text{cc}}_{{ij^{\prime}}} }} } \right)} \quad j \ne j^{\prime};\quad i = 1,2 \ldots 6$$

(1)

To determine if segmental ε _cc could predict the presence of MDE, we specifically focused on ∆ε _cc by segment from group B subjects (no MDE—early disease) to group E subjects (positive MDE state, advanced disease).

Inferential Phase

The full model consisted of main effects and interaction effects between disease state (subject groups A-E) and LV ε _cc in segments 7–12. As such, a 5 × 6 (disease state x segment) split-plot design with unbalanced cells was used for this analysis with both group and segment modeled as fixed effects and subjects nested within groups as our lone random effect (see Eq. 2).

$$Y_{ijk} = \mu + \rho_{i(j)} + \alpha_{j} + \beta_{k} + \left( {\alpha \beta } \right)_{jk} + \omega_{ijk},$$

(2)

where α _j  = main effect of group; ρ _i(j) = whole-plot error; β _k  = main effect of segment; (αβ) _jk  = interaction term; and ω _ijk  = split-plot error.

The two error terms, ρ _i(j) and ω _ijk , were assumed to be normally, independently, and identically distributed with mean 0 and variance components σ ² _ρ and σ ² _ω , respectively. Moreover, it was assumed that the whole-plot error variance (σ ² _ρ ) and split-plot error variance (σ ² _ω ) are independent of each other.