We prospectively and consecutively enrolled the inpatients in sinus rhythm from the Cardiology Department of our hospital from January 2019 to October 2019.

In accordance with the 2016 recommendations for the evaluation of LVDF from the ASE and EACVI [4], patients with specific cardiovascular disease in which the evaluating algorithm had significant limitations were excluded, such as the patients with atrial fibrillation, sinus tachycardia, atrioventricular block, left bundle branch block, hypertrophic and restrictive cardiomyopathy, significant mitral annular calcification, mitral stenosis, moderate and severe mitral regurgitation, severe aortic regurgitation, noncardiac pulmonary hypertension, right ventricular pacing, cardiac resynchronization therapy, valvular heart surgery or intervention and heart transplantation. Moreover, to obtain accurate measurements of LAV and LAAP, patients with poor quality of echocardiography, hyper-mobile interatrial septum or interatrial septal aneurysm, dilated aortic root or ascending aortic encroaching on the LA anterior wall, dilated or tortuous descending aorta indenting the LA posterior wall, and severe calcification at aortic sinuses and aortic valve were also excluded from this study.

A total of 552 subjects (mean age 58.4 ± 11.1 years, 221 women) met the eligibility criteria during the study period. These subjects were randomly divided into two groups: (1) derivation set (n = 276, mean age 59.1 ± 10.8 years, 114 women), in which best-fitting regression models and equations of LAVI–LAAPI were developed; and (2) validation set (n = 276, mean age 57.7 ± 11.3 years, 107 women), in which concordance for evaluating LVDF between using estimated and observed LAVI was verified.

Written informed consent was obtained from all patients before enrollment. The study protocol was approved by the China Medical University Ethics Committee and was conducted in line with the ethical guidelines of the 1975 Declaration of Helsinki.

All subjects were examined at rest using a Vivid E9 ultrasound system (GE Healthcare, Waukesha, WI, USA) equipped with M5S phased-array probe. Standard two-dimensional cine loops, including at least 3 consecutive cardiac cycles, and Doppler spectrum during normal respiration (< 20 breaths/min) were recorded for offline analysis using an EchoPAC workstation (GE Healthcare). Both image acquisition and quantitative analysis were performed according to the recommendations of transthoracic echocardiographic examination and cardiac chamber quantification from ASE by two experienced cardiologists who were blinded to any clinical data [15, 16].

LVEF was measured using the biplane Simpson’s disk summation method from both the apical two- and four-chamber views during LV end-diastole. LVEF < 52% in males or < 54% in females was considered depressed [15]. LAAP was measured during LV end-systole from the parasternal long-axis view [15]. LAV was measured using the biplane Simpson’s disk summation method from both the apical two- and four-chamber views during LV end-systole (Fig. 1). LAVI and LAAPI were LAV and LAAP divided by body surface area, respectively. Mitral inflow early diastolic (E) and late diastolic (A) peak velocity by pulsed flow Doppler, the septal and lateral early diastolic mitral annulus peak velocity (e′) by pulsed tissue Doppler, and tricuspid regurgitation (TR) systolic jet velocity by continuous Doppler were measured in the apical four-chamber views. Then E/A, average e′ and average E/e′ were calculated.

The diagnosis and grading of LVDD was performed according to 2016 ASE/EACVI recommendations [4]. For patients with normal LVEF, four variables were recommended for diagnosing LVDD (septal e′ < 7 cm/s or lateral e′ < 10 cm/s, average E/e′ > 14, LAVI > 34 mL/m², TR velocity > 2.8 m/s). Patients were diagnosed with normal DF if none or only one variable met the cutoff values, DD if more than two of the variables met the cutoff values, and indeterminate if just two variables met the cutoff values. In patients with normal LVEF and myocardial disease or DD, and in those with depressed LVEF, grade I DD was diagnosed if E/A ratio ≤ 0.8 and E ≤ 50 cm/s and grade III DD if E/A ratio ≥ 2. If E/A ≤ 0.8 and E > 50 cm/s, or 0.8 < E/A < 2, additional variables are needed (average E/e′ > 14, LAVI > 34 mL/m², TR velocity > 2.8 m/s). Grade I DD was diagnosed if only one of the variables met the cutoff values, and grade II DD if more than two of the variables met the cutoff values. However, if only one variable is available, or only two variables are available but discrepant, LV diastolic dysfunction grade could not be determined.

Inter- and intra-observer agreement of LAAP and LAV measurements were examined in 10 randomly selected patients. To assess intra-observer variability, the same doctor repeated the measurements after > 4 weeks. In addition, to assess inter-observer variability, the two doctors independently repeated the measurements twice.

Statistical analyses were performed using SPSS version 21.0 statistical software (IBM Corp, Armonk, NY) and R version 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria). Normality plots with tests were performed using the Shapiro–Wilk test. Continuous data were expressed as mean ± standard deviation (SD) or median (interquartile range), and categorical variables as numbers and percentages. In the comparison between derivation set and validation set, differences in continuous variables were assessed by independent samples t-test or Mann–Whitney U test where appropriate, and categorical variables by chi-square test or Fisher exact test. Correlation between LAVI and LAAPI was sought using Pearson correlation analyses. In the derivation set, the best-fitting regression models between LAVI and LAAPI were developed, including linear, logarithmic, quadratic, cubic, compound, power, S, growth, and exponential. The F-values, which express the statistical significance of the different models, of these models were compared. The regression equations with the highest F-values were chosen to calculate the estimated LAVI in the validation set. The comparison between the observed LAVI with estimated LAVI by each equation was performed by paired samples t-test or Wilcoxon test, where appropriate. The variability of estimated and observed LAVI was assessed by calculating the intra-class correlation coefficient and 95% confidence interval (CI) using a two-way random-effects model. Receiver operating characteristic (ROC) curve analysis and area under curve (AUC) were used to evaluate the performance for identifying observed LAVI > 34 mL/m². Bland–Altman analyses were used for assessing the bias and limits of agreement (LOA) between the observed and estimated LAVI, and estimating the inter-observer and intra-observer reproducibility of LAAP and LAV. The concordance for evaluating LV diastolic function between using estimated and observed LAVI were tested by calculating the κ coefficient and the overall proportion of agreement (proportion of subjects identically classified). The calculated κ coefficient was judged as follows: 0 ≤ κ < 0.2 slight; 0.2 ≤ κ < 0.4 fair; 0.4 ≤ κ < 0.6 moderate; 0.6 ≤ κ < 0.8 substantial, and 0.8 ≤ κ < 1 perfect [17].

There was a significant correlation between LAVI and LAAPI (r = 0.67, P < 0.001), and the correlation coefficient between LAVI and LAAPI in the subjects with depressed LVEF (r = 0.68, P < 0.001) was larger than that in subjects with normal LVEF (r = 0.53, P < 0.001) (Fig. 2).

Table 2 displays the best-fitting regression models for computation of LAVI from LAAPI in derivation set. The linear model had the highest F-value in all subjects (r² = 0.44, F = 218.60, P < 0.001) and in the subjects with normal or depressed LVEF (r² = 0.28, F = 83.06, P < 0.001; r² = 0.46, F = 48.65, P < 0.001). Therefore, we estimated LAVI from LAAPI stratified by LVEF using two regression equations (first equation: LAVI = 2.05 × LAAPI − 13.86 in all subjects; second equation: LAVI = 1.54 × LAAPI − 3.97 in the subjects with normal LVEF and LAVI = 2.15 × LAAPI − 12.83 in the subjects with depressed LVEF).

Before estimating LAVI, we first analyzed the identifying performance of LAAPI for LAVI > 34 mL/m² [AUC, 0.85 (0.80–0.90); P < 0.001], and the concordance between LAAPI > 23 mm/m² with LAVI > 34 mL/m² (overall proportion of agreement, 77.0%; κ = 0.47).

After estimating LAVI using the abovementioned two regression equations in the validation set, we compared the observed LAVI with estimated LAVI by each regression equation respectively, and found that there were no significant differences between the observed LAVI [29.03 (23.59–37.13) mL/m²] and estimated LAVI by the first equation [30.79 (26.88–35.42) mL/m²] and the second equation [29.98 (27.23–33.98) mL/m²] (Fig. 3).

Table 3 displays the intra-class correlations between estimated and observed LAVI, identifying performance of estimated LAVI for observed LAVI > 34 mL/m², and concordance between estimated and observed LAVI > 34 mL/m², respectively. The estimated LAVI using the first equation showed the similar intra-class correlations coefficient [ρ, 0.63 (0.55–0.70); P < 0.001] and overall proportion of agreement (78.98%) with observed LAVI, but the slightly higher AUC [0.85 (0.80–0.90); P < 0.001] for identifying observed LAVI > 34 mL/m². In Bland–Altman analysis, the estimated LAVI using the first equation showed the slightly narrower range of difference compared with observed LAVI (Fig. 4). Therefore, we finally used the first equation to estimate LAVI from LAAPI and further evaluate LVDF.

Concordance for diagnosis (overall proportion of agreement, 88.4%; κ = 0.79) and grading (overall proportion of agreement, 84.8%; κ = 0.74) of LVDD was substantial between using estimated and observed LAVI (Table 4 and Fig. 5). The overall reclassification rate was 15.2% (42 patients). Seven patients (2.5%) had normal DF according to observed LAVI but indeterminate on the basis of estimated LAVI. One patient (0.4%) and 9 patients (3.3%) with grade I DD according to observed LAVI were reclassified to normal DF and grade II DD using estimated LAVI, respectively. One patient (0.4%) and 3 patients (1.1%) with grade II DD according to observed LAVI were reclassified to grade I DD and indeterminate according to estimated LAVI. And Seventeen patients (6.2%) and 4 patients (1.4%) with indeterminate according to observed LAVI were reclassified to normal DF and grade II DD using estimated LAVI, respectively.

Inter-observer and intra-observer reproducibility for LAAP and LAV was high. The inter-observer reproducibility for LAAP and LAV had a bias of 0.5 mm (LOA, − 0.5 to 1.4 mm), and 0.6 mL (LOA, − 5.3 to 6.4 mL), respectively. The intra-observer reproducibility for LAAP and LAV had a bias of − 0.1 mm (LOA, − 1.1 to 0.9 mm), and 0.6 mL (LOA, − 4.7 to 5.8 mL), respectively.

The major limitation of our study was related to the exclusion of the patients with several specific cardiovascular diseases, such as atrial fibrillation, mitral valve dysfunction, left bundle branch block, hypertrophic cardiomyopathy, and so on, in which the evaluation of LVDF required the determination of additional specific parameters according to the 2016 recommendations [4]. This may limit the generalizability of our findings in these patients.

Another study limitation was the relatively low percentage of patients with depressed LVEF and limited range of LAVI, which was mainly attributable to the unselected and consecutively recruited population for reducing selection bias in the present study, and a multicenter study with a larger number of patients with depressed LVEF and a wide range of LAVI should be designed to verify the findings.