Intervendor Agreement for Right Ventricular Global Longitudinal Strain in Children.

Doaa Aly, Children's Mercy Hospital
Sandhya Ramlogan
Rita France, Children's Mercy Hospital
Stephanie Schmidt
Julie Hinzman
Ashley K. Sherman, Children's Mercy Hospital
Suma P Goudar
Daniel Forsha, Children's Mercy Hospital


BACKGROUND: Right ventricular global longitudinal strain (RVGLS) has emerged as an important technique for clinical evaluation of (RV) function. The routine application of RVGLS in pediatrics remains limited by a lack of data on agreement between vendors. The aim of this study was to investigate intervendor agreement for RVGLS between the two commonly used analysis vendors in pediatrics, hypothesizing that RVGLS has good intervendor agreement, although it is likely lower than intravendor agreement (inter- and intraobserver reproducibility).

METHODS: Seventy infants and children with normal cardiac anatomy and varying ventricular function were included after prospectively obtaining RV-focused four-chamber apical images on the GE Vivid E95. Images were analyzed for RVGLS at acquired frame rates in EchoPAC (GE) and TomTec (TT) and in the compressed Digital Imaging and Communications in Medicine format in TT. Intraclass correlation coefficients and Bland-Altman plots were used to test intervendor agreement and intravendor reproducibility.

RESULTS: RVGLS measurements were equally feasible using TT and EchoPAC analysis (92%). There was good to excellent agreement for RVGLS between TT and EchoPAC analysis, with a relatively higher intraclass correlation coefficient between GE and TT at the acquired frame rate (0.85) than between GE and TT at the compressed frame rate (0.75) and significantly higher agreement in patients with abnormal RV function (0.7-0.9) than those with normal function (0.4-0.6). Intra- andinterobserver reproducibility for RVGLS was excellent (intraclass correlation coefficient = 0.74-0.96). Heart rate ≥ 100 beats/min and acquisition frame rate/heart rate ≤ 0.7 were associated with diminished agreement, especially when compressed data were involved.

CONCLUSIONS: RVGLS analyzed using EchoPAC and TT show good agreement, especially when analyzed at acquisition frame rates and in the setting of abnormal RV function. Otherwise, RVGLS should ideally be analyzed using the same vendor, and intervendor comparisons should be undertaken with caution, particularly if data are in a compressed format.