DISCUSSION
This study reports the largest multicenter data collection for development of 3D echocardiographic z-scores for LV volumes and function in healthy North American children using the semi-automated quantification method. The feasibility was 79% analyzable datasets at the core laboratory. We demonstrated that the semi-automated software quantification with LV border adjustment was more reliable between centers than automated software quantification. Although automated software quantification took less time, the greater variability in measurement rendered it less reliable, and therefore the normative Z-score values were derived from the semi-automated method.
Our study had feasibility of 79% of obtaining adequate 3D data for analysis at the core laboratory, similar to another multicenter study in the pediatric age group 8. Krell et al reported having a feasibility of 74% in their multicenter study but our study population was larger with more patients. Kuebler et al reported normative LV volume and functional values in 238 pediatric subjects of different age group and body surface area 9. Only 14% of their subjects were under the age of 5 (34/238) 9. Our study is notable in that 27% (141/523) of the subjects were under the age of 5 and 18% were less than 3 years old. This is the population that is technically challenging to acquire 3D volumes as young patients are unable to hold their breath or stay relatively motionless during the acquisition, obviating “stitch artifact”. On the other hand, they may have the best imaging window due to smaller chest walls and ability to incorporate the entire ventricular volume into the dataset. Cantinotti et al studied 800 Italian healthy children and reported excellent feasibility of 91%; however, the feasibility for smaller children with smaller BSA less than 0.5 was 68% to 80% respectively10.
We did not find that the automated quantification software improved the reproducibility in volume and function analysis. There was wide LOA between automated and semiautomated LV quantification in all centers. The automated software frequently did not track the LV endocardial borders. This became apparent in the intraobserver reproducibility in the core lab when the same LV datasets were re-measured using the semi-automated contouring method. The intraobserver reproducibility was excellent when semi-automated method was used and showed small bias and LOA. However, the intraobserver reproducibility did not have good agreement when automated versus semi-automated quantification comparison was used on the same LV datasets. This study highlights the difference between pediatric and adult studies of using automated versus semi-automated contouring of LV volumes and function. We postulate that the current automated contouring does not adequately track LV contours in pediatric 3D datasets because the fully automated algorithms are built based on adult 3D datasets. The time for semi-automated contouring of the LV was not more than a mean of 2-3 minutes in all centers, which should be acceptable for routine clinical use.
Our study is unique in that we performed extensive analysis of reproducibility by comparing all 3D datasets from 4 different centers to the core lab using the semi-automated and automated methods. The semi-automated methods between centers were reproducible. The percent difference in the LV EDV was ± 30% indicating that if an individual measured the LV EDV at 10ml, the difference could be ± 3ml. This is an acceptable range of measures between centers. The percent difference for LV ESV measurement was also about ± 30% (Figure 3). The LOA of LV volumes and EF were wider using the automated methods compared to the semi-automated methods between the core lab to other centers, indicating that LV volumes and EF were not reproducible using this method. Prior studies from Kuebler et al and Cantinotti et al have described pediatric normative LV volumes and function derived from single centers9,10. Similar to previous studies, LV volumes indexed to the BSA demonstrated gradual increase from childhood to adolescent years 8-12.
The mean absolute percent difference in 3D LV EF between all centers in our study was 7% with the observed difference of 4.2% if the LV EF was measured at 60%. This is better than the reported 2D LV EF absolute percentage difference of 11.2% in the Pediatric Heart Network study13. 3D LV EF has been reported to be more accurate and reproducible than 2D LV EF in adults and children because 3DE does not rely on geometric assumptions and is less affected by acquisition technique such as foreshortening 3,14-18. Thus, it is recommended that 3D LV EF should be reported in clinical echocardiograms in centers with experience in 3DE 18. Our study demonstrated that the reproducibility of 3D LV EF is good to excellent and can be used in multiple centers.