Study Design and Population

We performed a cross-sectional observational study of 230 healthy volunteers, with a wide age range ( Figure 1 , Table 1 ), prospectively recruited from October 2011 to July 2013. The present work is a substudy of a large project (Padua 3D Echo Normal) aimed at providing normative data on quantitative parameters obtained with three-dimensional (3D) echocardiography. Volunteers were recruited among hospital employees, fellows in training, their parents, and people who underwent medical visits for driving or working licenses and met the following inclusion criteria: age > 17 years, no history of cardiovascular or lung disease, no symptoms, absence of cardiovascular risk factors (i.e. systemic hypertension, smoking, diabetes, dyslipidemia) assessed during the screening visit according to established diagnostic criteria, no cardioactive or vasoactive treatment, and normal results on electrocardiography and physical examination. Exclusion criteria were status as a professional athlete, pregnancy, body mass index > 30 kg/m ² , no knowledge of serum lipid or glucose levels at any time, more than mild tricuspid or pulmonary valve regurgitation, pulmonary artery systolic pressure > 36 mm Hg (estimated by peak tricuspid regurgitation velocity taking into account mean right atrial pressure), and poor 2D echocardiographic image quality from the apical approach.

Enrollment flowchart. DBP , Diastolic blood pressure; LDL , low-density lipoprotein; SBP , systolic blood pressure; 2DE , 2D echocardiographic.

The study was approved by the University of Padua ethics committee (protocol no. 2380 P, approved on October 6, 2011), and all volunteers provided written informed consent before undergoing physical examination, blood pressure and anthropometric measurements, and echocardiography. Body surface area was calculated according to the formula of Du Bois and Du Bois.

Image Acquisition

All subjects underwent comprehensive 2D echocardiography using a commercially available Vivid E9 ultrasound machine (GE Vingmed Ultrasound AS, Horten, Norway) equipped with an M5S probe. The electrocardiographic tracing on the monitor was selected to show a well-defined P wave (typically lead I, II, or aVF) by selecting the most suitable lead either using the dedicated knob on the echocardiographic machine or changing the position of electrodes on the chest of the patient. All patients were examined in the left lateral decubitus position using grayscale second-harmonic 2D imaging, with adjustment of image contrast, frequency, depth, and sector size for adequate frame rate and optimal LA and LV endocardial border visualization. Echocardiographic studies included three apical views (four-chamber, two-chamber, and long-axis) optimized for LV global LS (GLS) analysis followed by dedicated (i.e., nonforeshortened) apical four- and two-chamber views for LA LS analysis. For each view, three consecutive heart cycles were recorded with a frame rate ranging between 50 and 80 frames/sec.

At the end of the 2D echocardiographic examination, four or six consecutive electrocardiographically gated subvolumes were acquired from the apical approach during a breath-hold to generate separated 3DE full-volume data sets of the LV and the LA, using a 4V matrix-array transducer (GE Vingmed Ultrasound AS) and taking care to encompass the entire cavities in the data sets. Data sets were stored digitally and exported to a separate workstation for offline analysis.

Image Analysis

LA and LV myocardial LS was measured using the Q-Analysis software package in EchoPAC version 201 (GE Vingmed Ultrasound AS). The LA LS analysis was obtained from the dedicated apical four- and two-chamber views (biplane analysis). The peak of the P wave on the electrocardiographic trace was used as the zero-reference point for the generation of the strain curves, and the LA endocardium was manually traced at LA minimal volume. For each view, a 15-mm-wide region of interest was automatically generated and divided into six segments of equal length. Then a cine loop preview feature allowed visual confirmation that the internal line followed the LA endocardium movements throughout the cardiac cycle. If tracking of the LA endocardium was unsatisfactory, manual adjustments of the region-of-interest size and position were performed to ensure optimal tracking. When more than two LA segments were inadequately tracked, the 2D speckle-tracking echocardiographic software package no longer computed LS, and these data sets were excluded from further analyses. The analysis was then approved, and an average GLS curve was automatically generated ( Figure 2 ). The LS curve included an initial negative deflection (LSneg), representing LA active contraction, followed by a positive deflection (LSpos) during LA filling ( Figure 2 ). LSneg, LSpos, and their summation (LStot) values were recorded using the biplane approach by averaging individual values from 12 LA segments.

LA LS measurement using 2D STE. Color coding of the six myocardial segments of the LA region of interest for the measurement of LA strain in dedicated four- and two-chamber views ( A and D , respectively). Individual LA segment LS values in dedicated four- and two-chamber views ( B and E , respectively). Average LA LS/time curve obtained (white dotted lines) from the four- and the two-chamber view using the peak of the P wave on the electrocardiographic tracing (yellow dots and yellow vertical lines) as the zero-reference point for the generation of the strain curves ( C and F , respectively). Measurements (double-headed arrows) of LSneg, LSpos, and LStot are shown for the dedicated four-chamber view (C) .

To obtain robust data for LA size and function, we measured LA volume using 3D echocardiographic data sets and dedicated software (LA Analysis, TomTec Imaging Systems, Unterschleissheim, Germany), validated against cardiac magnetic resonance. The end-systolic (just before the mitral valve opening), pre-A (peak of the P-wave frame), and end-diastolic (just after mitral valve closure) time cycles were marked in the corresponding frame on the electrocardiographic tracing. Then, rapid manual data set alignment was performed by translating and rotating the four-chamber plane to obtain orthogonal nonforeshortened planes of the left atrium in all three apical views. In each apical view, the LA blood-tissue interface was manually initialized on 2 frames identifying LA maximal and minimal volumes. Then the initialized LA endocardial boundaries were used to reconstruct the LA endocardial surface. For each consecutive frame, the voxel count inside the 3D echocardiographic LA surface was used to measure LA volumes, resulting in a smooth interpolated time-volume curve from which maximal LA volume, minimal LA volume, and pre-A volume were measured and phasic function parameters calculated.

Transmitral flow velocities were measured using pulsed-wave Doppler echocardiography, and mitral flow parameters, including peak velocities during early (E) and late (A) diastole and the deceleration time of the E wave, were measured. Doppler tissue imaging was performed with a sample volume placed at the medial and lateral annulus in the apical four-chamber view. The peak myocardial velocities during systole (s′) and early (e′) and late (a′) diastole were measured as the average of the septal and lateral values, and the E/e′ ratio was calculated.

LV volumes were measured using dedicated 3D echocardiographic data sets and 4D AutoLVQ software package (EchoPAC version BT12). Initialization of LV endocardial border tracing was manually performed by the operator, by identifying the middle of the mitral annulus and the LV apex at both end-diastole and end-systole on the four-chamber view frames. Manual editing of the semiautomatically generated endocardial contours was routinely applied to include the LV outflow, as well as papillary muscles and trabeculae within the LV cavity.

Finally, speckle-tracking analysis for GLS measurements of the left ventricle was performed using the three apical 2D views (e.g., four-chamber, two-chamber, and long-axis). The endocardial border of the left ventricle was manually traced slightly inside the myocardium. A second, larger, concentric circle was then automatically generated near the epicardium to define a region of interest that included the LV myocardium. Then the software automatically divided each LV view into six equal segments and performed the speckle-tracking on a frame-to-frame basis. The peak values of LS obtained from 18 LV segments were averaged to compute LV GLS.

Statistical Analysis

Normal distribution of variables was verified using the Kolmogorov-Smirnov test. Continuous variables are summarized as mean ± SD, while categorical variables are reported as percentages. The unpaired t test was used to compare variables between men and women, as well as between subjects younger and older than 50 years. Pearson correlation was used to analyze the relationships between LA LS and demographic and echocardiographic variables.

Except for reporting reference values (when LSneg was considered negative), all LS parameters and GLS elsewhere were interpreted as absolute values, and comparisons were based on strain magnitude (with lower strain values indicating worse deformation).

The sample was divided into two subgroups on the basis of gender and the younger or older than 50 years to develop the reference limits for LSneg, LSpos, and LStot. Linear quantile regression was used to estimate the 2.5th and 97.5th quantiles for each LA LS measure. Both quantiles and ±2 SDs are metrics used in descriptive statistical analysis to provide the 95% interval of the sample data. The main difference is that the use of ±2 SDs is based on the assumption of a symmetric (normal) distribution of the data. Multivariate stepwise logistic regression was performed to identify the determinants of LA LS parameters among demographic and echocardiographic variables. The multivariate models included the following covariates: age, LV GLS, LV stroke volume, LV end-systolic volume, e′ (average value between septal e′ and lateral e′), systolic and diastolic blood pressure, mitral E-wave velocity, E/A ratio (having P < .10 on bivariate analysis), plus gender, with model selection using the backward likelihood ratio. In addition, to investigate whether the determinants of LA LS parameters were the same for both genders and in the young and old subjects, we repeated the multivariate analyses after stratifying our study population in men and women and in subjects younger or older than 50 years.

Intra- and interobserver variabilities for LA LS parameters were analyzed in 15 random subjects using the Bland-Altman method and are reported as bias ± 2 SDs. To obtain intraobserver variability, one observer (M.H.M.) evaluated the same studies on a separate occasion 2 weeks apart. For the interobserver variability assessment, two independent observers (M.H.M. and S.M.) measured LS parameters on the same 15 subjects. Both observers were allowed to choose the cardiac cycle on which to perform the measurements among the three recorded in the loop.

Analyses were performed using R System (Statistical Package 2009) rms and quantreg libraries and MedCalc version 10.0.1 (MedCalc Software, Mariakerke, Belgium). Significance was assigned to test statistics with P values < .05.

Normal Values of LA LS Parameters

We obtained the reference values and the limits of normality for LA LSpos, LSneg, and LStot separately for men and women ( Table 3 ). Overall, women had higher magnitudes of LSpos and LStot than men, this effect was statistically significant in subjects <50 years of age. ( Table 3 ).

LSpos and LStot values were significantly lower in the elderly than in the young ( Table 3 , Figures 3 and 4 ). However, these changes were more pronounced and started earlier (by the third decade) in women than in men ( Figure 4 ). In men, changes can be noted by the fourth decade. By the fifth decade, the gender differences in LA LS tend to disappear ( Figure 4 ).

Correlations between LA 2D speckle-tracking echocardiographic LS parameters and age using the biplane method.

Gender-related LA LS parameter changes across the different age groups.

Except for LA LSneg, absolute values of LA LS parameters were significantly higher in the younger age groups ( Table 3 ), and there was an inverse correlation between LA LS parameters and age in both genders ( Figure 3 ).

Effects of the Echocardiographic Method on LA LS Parameter Normal Values

Demographic, physiologic, and echocardiographic characteristics of the 171 subjects in whom LA LS was measured in both four- and two-chamber views were similar to those of the 230 subjects in whom LA LS was measured in the four-chamber view only. Although the median and 25th and 75th quartiles of LA LSpos obtained from the four-chamber view (19% [15% to 24%]) were similar to those obtained with the biplane method (19% [15% to 24%]) ( P = NS), the absolute value of LA LSneg (and, consequently, LA LStot) was significantly lower using the four-chamber (−13% [−11% to −14%]) than using the biplane (−15% [−13 to −16%]) method ( Tables 3 and Supplemental Table 1 ). The difference was explained by an average −3.47% (−2.86% to −4.08%) more negative value of LA LSneg obtained from the two-chamber view compared with the value obtained from the four-chamber view.

The correlations of LA LS values with 3D echocardiographic LA volumes and volumetric phasic function parameters and LV diastolic and systolic function parameters were similar using both the single-plane and biplane methods of measuring LA LS.

Demographic and Echocardiographic Correlates to LA Longitudinal Strain Parameters

On bivariate analysis, LA LSpos and LStot showed weak correlations with body surface area ( r = −0.26, P = .001, and r = −0.25, P < .001, respectively), body mass index, ( r = −0.35, P < .001 for all), and systolic ( r = −0.32, P < .001, and r = −0.19, P = .004, respectively) and diastolic ( r = −0.24, P < .001, and r = −0.18, P = .008, respectively) blood pressure. Conversely, LA LSneg correlated only with systolic blood pressure ( r = −0.26, P < .001). LSpos and LStot showed weak positive correlations with all phasic LA volumes and LA active emptying volume ( Supplemental Table 2 ). Moreover, all LA LS parameters correlated positively with LA passive emptying fraction and negatively with LA active emptying fraction ( Supplemental Table 2 ).

LA LS parameters showed significant correlations with indexes of LV diastolic and systolic function. LA LSpos and LStot presented positive correlations with E-wave velocity, E/A ratio, e′ velocity, and E/e′ ratio, showing a direct relationship with indexes of LV compliance. LA LSneg showed a negative correlation with A-wave velocity ( Supplemental Table 3 ).

LSpos and LStot, but not LSneg, increased significantly with the increase of myocardial s velocity and LV GLS ( Supplemental Table 3 ).

Independent Correlates of LA Longitudinal Strain Parameters

The multivariate linear regression analysis performed on the whole study population showed that age, E/A ratio, and e′ average and LV stroke volume and GLS were independently correlated with LA LS values and accounted for 57%, 20%, and 40% of the variance for LSpos, LSneg, and LStot, respectively ( Table 4 ). However, when we analyzed separately subjects younger or older than 50 years of age, we found that in addition to age, both LV systolic and diastolic function parameters were independently correlated with LA LS in the subjects younger than 50 years, but not in the oldest, in whom age and systolic and diastolic blood pressures were the main determinants of LA LS parameters ( Supplemental Table 4 ).

Finally, age was the only independent covariate that correlated with LA LSpos and LStot in men ( Supplemental Table 5 ). Conversely, age, diastolic function indexes, and LV stroke volume were independently correlated with LA LS parameters in women ( Supplemental Table 5 ).

Reproducibility

LA LS parameters showed low intra- and interobserver variability: LSpos, 0.0 ± 2.5% and 0.1 ± 3.5%, respectively; LSneg, −0.1 ± 1.9% and 1.0 ± 3.0%, respectively; and LStot, 0.1 ± 3.0% and 1.1 ± 2.4%, respectively.