RESULTS
Population
One hundred and one children with structurally normal hearts were
included in the LV analysis and 98 in the LA analysis. Of the subjects,
51 (51%) were female and 49 (49%) were male. The median and range for
height was 127cm (57 – 195cm) and for weight was 25kg (5.2 – 139kg).
Body surface areas between 0.5 and 1kg/m2 composed the greatest
proportion of the group (42%) whereas BSA <0.5kg/m2 were less
well represented (12%) (Table 1).
Assessment of regression
The end-diastolic and end-systolic measurements were first regressed
against BSA and tested for heteroscedasticity. All measurements
demonstrated a strong correlation with BSA. Both TT derived and VMS
derived LV and LA EDV showed significant heteroscedasticity using the
Breusch-Pagan F-statistic. Power transformation provided the best model
for the relationship between BSA and all volumes, however,
heteroscedasticity was accentuated. A log-log transformation of volumes
and BSA eliminated heteroscedasticity with acceptable overall residuals
(Table 2). Average, standard deviations and upper and lower normal
ranges for values indexed using the optimal power transformation are
shown in Table 3 and 4. Calculation of z-scores for study values using
the best fit power regression revealed that 5 - 6% of diastolic and
systolic study values using either technique fell beyond the 2 standard
deviation range (Table 3).
Agreement and reliability
LA Volumes: There was very good to excellent correlations between VMS
and Tomtec measuring LA systolic and diastolic volumes: LAESV ICC 0.94
(95% CI 0.90,0.96), LAEDV ICC 0.87 (95% CI 0.81, 0.91). On average
Tomtec LA measurements were greater than VMS measurements with a small
mean bias: LAESV (mean difference 6.6±1.9% (limits of agreement 44%,
-31%), LAEDV (mean difference 6.5% (95% CI 47%, -60%). The Pearson
correlation coefficient for LA ESV was 0.94 and for LA EDV was 0.87
LV Volumes: Tomtec derived ventricular measurements were higher than VMS
measurements on average, with no difference in variability. The mean
bias between Tomtec and VMS measurements of LVEDVi was 7.5ml±9.2ml or
15.5±1.9% (limits of agreement: 52.9%, -22.3%), and LVESV was
1.7ml±4.2ml or 7.3±2.0%, (limits of agreement 47.1, -32.5%) (Table 3,
Figure 1). There was very good to excellent correlations between VMS and
Tomtec measurements of LV systolic and diastolic volumes: LV systolic
ICC 0.93 (95% CI 0.89, 0.95), LV diastolic ICC 0.92 (95% CI 0.79,
0.96). The Pearson correlation coefficient for LVEDV was 0.94, and for
LVESV was 0.94. LV and LA linear regression scatter plots are displayed
in Figure 2.
Efficiency :
VMS analysis time was shorter than Tomtec for both the LA and LV (Table
5). LV analysis time (total EDV and ESV) using TomTec vs VMS analysis
time had an estimated difference of 0.75 minutes (95% CI 0.56, 0.94,
p<0.0001) For LA analysis time, there was an estimated
difference between TomTec and VMS of 1.23 minutes (95% CI 1.01,1.45,
p<0.0001).
Reproducibility :
Interobserver Agreement (IOA): We
used a two-way agreement model with 95% confidence interval to report
IOA completed on 23 Tomtec analyses and 18 VMS analyses of the LV and
LA. There was excellent IOA for Tomtec LAESV (ICC 0.96 (95% CI 0.90 –
0.98)), VMS LAESV (ICC 0.90 (95% CI 0.75 – 0.96)) and VMS LAEDV (ICC
0.92 (95% CI 0.64 – 0.98). There was good IOA for Tomtec LA EDV (ICC
0.84 (95% CI 0.13, 0.95), with several outliers accounting for wide
confidence intervals. There was excellent IOA for TomTec LVEDV (ICC 0.95
(95% CI 0.56 – 0.98)), VMS LVEDV (ICC 0.97 (95% CI 0.90 – 0.98)) and
VMS LVESV (ICC 0.96 (95% CI 0.90 – 0.99)) and good IOA for LVESV (ICC
0.77 (95% CI 0.53 – 0.89). LA & LV interobserver regression analysis
and Bland-Altman plots of agreement are displayed on Figures 3 and 4
respectively.
Intra-observer agreement (IAOA): We used two-way agreement model with
95% confidence interval to report IAOA completed on 21 studies.
There was excellent IAOA for VMS
LAESV (ICC 0.99 (95% CI 0.95 – 1.0) and VMS LAEDV (ICC 0.96 (95% CI
0.76 – 0.99). There was also excellent IAOA agreement for TomTec LAESV
ICC 0.96 (95% CI 0.91 – 0.98) and good agreement for LAEDV (ICC 0.86
(95% CI 0.52 – 0.95)). VMS and TT LA intra-observer regression
analysis and LA Bland-Altman plots of agreement are displayed on Figure
5 and 6 respectively. TomTec LV intra-observer agreement was performed
as part of a multicenter study13.