RESULTS
TTE and TEE measurements of LVOTD . The 53 patients in whom LVOTD was measured during TTE and TEE had a mean age of 64 ± 16 years (40 males and 13 females). Mean LVOTD was 21.2 ± 2.2 mm at TTE and 21.2 ± 2.1 mm at TEE (P=0.774). Correlation between the TTE and TEE techniques was excellent (r=0.921, P<0.001) (Supplementary Figure 1). The Bland-Altman method showed no trend for under- or over-estimation using TTE (Supplementary Figure 1).
Derivation and validation group patients . Characteristics of the derivation and validation group patients are reported in Table 1. No significant difference was found for age, gender, anthropometric data, LV volumes and EF, SVD, LVOTD and LVEDD between the derivation and validation groups. Prevalence of coronary artery disease and chronic obstructed pulmonary disease were higher in the derivation group (Table 1).
Determinants of LVOTD in the derivation study . Spearman correlations between LVOTD and prespecified covariates for the derivation group are reported in Table 2. Patient height, BSA, SVD, LVEDD and LVEDV significantly correlated with LVOTD (Table 2, Figure 1). SVD showed the highest correlation (R=0.568, P<0.001), followed by anthropometric variables (BSA R=0.532, height R=0.525; all P values <0.001) and LV measures (LVEDD R=0.357, LVEDV R=0.363; all P values <0.001). All correlations were confirmed by univariate linear regression analysis (Table 2). On multivariate linear regression analysis for the primary endpoint in the derivation study, SVD (beta=0.392, P<0.001), BSA (beta=0.229, P<0.001), LVEDD (beta=0.145, P=0.001) and height (beta=0.125, P=0.037) were independently associated with LVOTD (Model 1, Table 3). No significant multicollinearity was found (all variance inflation factors <0.3). An analogous linear regression model including SVD and LVEDD as indexed values was performed, showing independent correlation of SVDI (beta=0.448, P<0.001) and LVEDD index (beta=0.138, P=0.002) with LVOTD. A regression equation was derived from linear regression analysis including SVD, BSA, LVEDD and height for predicting LVOTD dimension (LVOTDRE1, Table 4).
Determinants of discrepancy between LVOTDM and LVOTDBSA . The difference between LVOTDMand LVOTDBSA was correlated at multivariate linear regression analysis with SVD and LVEDD as indexed values and height in the derivation group, with the aim to corroborate primary findings (Model 1, Table 5; Figure 2). SVDI (beta=0.499, P<0.001), LVEDD index (beta=0.151, P=0.002) and height (beta=0.2, P<0.001) were all confirmed as independent determinants of the discrepancy between LVOTDM and LVOTDBSA(Table 5).
Validation of the regression equation for predicting LVOTD . The regression equation for predicting LVOTD (LVOTDRE1) and LVOTDBSA were tested in the validation group (Table 6). The LVOTDRE1 showed a high correlation (R=0.739, P<0.001) and the lowest mean difference with LVOTDM (0.02± 1.57 mm), whereas LVOTDBSA showed a moderate correlation with LVOTDM (R=0.531, P<0.001) and the highest mean difference (-0.51± 1.95 mm) (Table 6). Paired Student’s t-test revealed that LVOTDRE1 did not differ from LVOTDM, whereas LVOTDBSA was statistically different from LVOTDM (P<0.001, Table 6). Figure 3 shows the distribution of differences between LVOTDM and LVOTDBSA (left panel) and LVOTDRE1 (right panel). The use of our regression equation allowed to predict LVOTDM with less than± 2 mm of error in 84% of cases, whereas it was 65.5% with the BSA-based formula (Figure 3).
Results with LVEDV instead of LVEDD. Similar results were obtained using LVEDV instead of LVEDD for all the analyses in the derivation group (univariate analysis, Table 2; multivariate model 2, Table 3; LVOTDRE2, Table 4; multivariate model 2 for discrepancy between LVOTDM and LVOTDBSA, Table 5) and in the validation group (LVOTDRE2, Table 6), respectively. LVOTDRE2 had just a slightly lower absolute correlation than LVOTDRE1 with LVOTDM (Table 6).
Effects of gender . When gender was added to multivariate analysis in the derivation group, SVD was confirmed as the main determinant of LVOTD (beta=0.351, P<0.001), followed by BSA (beta=0.258, P<0.001), female sex (beta=-0.175, P<0.001) and LVEDD (beta=0.134, P=0.003) or LVEDV (beta=0.122, P=0.001), whereas height was not associated with LVOTD (model 3 and 4, Table 3). The regression equations (Table 4) derived from the models with gender, BSA, SVD and LVEDD (LVOTDRE3) or LVEDV (LVOTDRE4) maintained high correlation and no significant difference with LVOTDM in the validation study, but showed a slightly lower absolute correlation than LVOTDRE1 with LVOTDM (Table 6).
Intra-observer and inter-observer variability. Intra-observer and inter-observer variabilities are reported in Table 7. While similar good percentage variability was observed between SVD, LVEDD, LVEDV and LVOTDM, variability of LVOT area derived from regression equations was lower than that of the measured LVOT area.