Background
Complex anatomy and limited windows complicate echocardiographic assessments of ventricular function in Fontan patients. For the Pediatric Heart Network Fontan Cross-Sectional Study, data were acquired from which mean ventricular pressure change during isovolumetric contraction (dP/dt ic ), Tei index, and maximal systolic annular velocity (S′) could be measured. The purpose of this study was to compare these nongeometric indices of ventricular function to cardiac magnetic resonance (CMR) measurements of ventricular ejection fraction (EF).
Methods
Echocardiographic and CMR studies were performed prospectively using standardized protocols; measurements were completed by core laboratories. Data from both modalities were available from 137 patients.
Results
A weak but statistically significant correlation was observed between mean dP/dt ic and CMR-derived EF ( r = 0.20, P = .022). This correlation was strengthened when preload was taken into account ( r = 0.30, P = .001). Statistically significant correlations did not exist between CMR-derived EF and the Tei index or S′.
Conclusions
Among Fontan patients, the correlation between CMR-derived EF and nongeometric echocardiographic indices of ventricular function is not strong. Of the indices evaluated, however, mean dP/dt ic appears to be the best.
Patients with single-ventricle physiology who have undergone the Fontan procedure often have complex abnormalities of ventricular anatomy and segmental wall motion. Consequently, obtaining a quantitative, objective, and reproducible echocardiographic assessment of ventricular function in these patients is challenging. Standard echocardiographic indices, such as the shortening fraction and the ejection fraction (EF), rely on assumptions regarding ventricular geometry and symmetry that may not be valid for many Fontan patients. Furthermore, many of these patients have limited echocardiographic windows, and adequate visualization of the endocardial border at end-systole and end-diastole is often difficult or impossible. Indeed, perhaps as a result of these limitations, many past clinical studies of Fontan patients have not included echocardiographic assessments of ventricular function or have reported only subjective and semiquantitative assessments.
In contrast, cardiac magnetic resonance (CMR) imaging can provide accurate measurements of ventricular volume and EF, even in the presence of complex anatomy and segmental wall motion abnormalities. However, CMR expertise and facilities are not widely available. The procedure is relatively expensive and may be contraindicated in patients with pacemakers and implantable cardioverters. Image artifacts from metallic implants also limit quantitative volumetric analyses in some Fontan patients. Furthermore, in young patients, the acquisition of CMR images often requires sedation. Hence, there still exists a need for a reliable, quantitative echocardiographic index of ventricular function that would be valid for patients with Fontan physiology.
For the National Heart, Lung and Blood Institute–sponsored Pediatric Heart Network Fontan Cross-Sectional Study, systemic blood pressures, pulsed Doppler echocardiographic tracings of systemic ventricular inflow and outflow, and tissue Doppler tracings of annular velocities were recorded according to a standardized protocol (these tracings can usually be obtained even in patients with poor echocardiographic windows). From these data, the following geometry-independent indices of ventricular function may be calculated: mean ventricular pressure change during isovolumetric contraction (dP/dt ic ), the Tei index, and the maximum annular velocity during systole (S′). Furthermore, CMR studies were obtained in approximately one third of study subjects. The primary purpose of this analysis was to assess the value of these easily obtained, geometry-independent echocardiographic indices of ventricular function by comparing them with CMR-based estimates of ventricular EF and to examine the relationship of these indices to one another. In addition, in light of Margossian et al. ‘ recent observation from the Fontan Cross-Sectional Study database that the mean difference between echocardiographic and CMR-based estimates of EF is small (although with wide limits of agreement), we also compared the geometry-independent indices of ventricular function with echocardiography-based estimates of EF.
Methods
Patient Sample
The Fontan Cross-Sectional Study was undertaken by the Pediatric Heart Network, which consists of 7 pediatric cardiac centers in the United States and Canada and performs multicenter clinical studies with funding from the National Heart, Lung, and Blood Institute of the National Institutes of Health. The purpose of the Fontan Cross-Sectional Study was to prospectively assess the correlation between measures of functional health status, ventricular function, and exercise performance in young Fontan survivors. A total of 546 subjects aged 6 to 18 years were enrolled in the Fontan Cross-Sectional Study, at the time of their routine cardiologic follow-up, in 2003 and 2004. The Fontan study design has been described. Two analytic cohorts from the study sample were used in this report: (1) subjects who had both CMR and echocardiographic data permitting the calculation of the echocardiographic indices of interest (patients who could not lie still without sedation and patients with metallic artifacts that precluded acquisition of a full CMR volumetric data set were not included in this cohort) and (2) subjects with echocardiographic data irrespective of CMR completion. All institutions obtained institutional review board approval to conduct the study, and the parent or guardian of each participant provided written informed consent.
Echocardiographic Protocol
Two-dimensional and Doppler echocardiograms were obtained according to a standardized protocol. Data were recorded at a paper speed of ≥100 mm/sec, and all measurements were completed by a single observer at a core laboratory. The mean dP/dt ic was calculated as (aortic diastolic pressure − ventricular end-diastolic pressure)/isovolumetric contraction time.
Aortic diastolic pressure was obtained from an automated blood pressure monitoring system using an appropriately sized cuff, during the echocardiographic study. An average of at least three measurements was used for the calculation. Ventricular end-diastolic pressure was assumed to equal 5 mm Hg. Isovolumetric contraction time was calculated by subtracting the time interval between the onset of the QRS complex and the closure of the systemic atrioventricular (AV) valve on the systemic ventricular inflow pulse Doppler tracing from the time interval between the onset of the QRS complex and the opening of the aortic valve on the systemic ventricular outflow pulsed Doppler tracing ( Figure 1 ).
S′ was measured from tissue Doppler tracings of the systemic ventricular free wall at the level of the systemic AV valve. One measurement of an S′ velocity of 24 cm/sec was considered unphysiologic and was therefore excluded from analysis. All other S′ observations were <16 cm/sec.
The Tei index was calculated from Doppler tissue imaging time intervals as (isovolumetric contraction time + isovolumetric relaxation time)/ejection time. Whenever present, the average of measurements from the septum and lateral walls was used; otherwise, the single available site was used.
Isovolumetric contraction time was measured as the time from the end of the A′ wave to the beginning of S′. Isovolumetric relaxation time was measured as the time from the end of S′ to the beginning of the E′ wave. Ejection time was measured as the duration of S′ ( Figure 2 ).
To evaluate the reproducibility of the geometry-independent indices of ventricular function, 20 studies were randomly selected, and the relevant measurements were repeated twice by a second observer unaware of the results of the previous measurements.
Tracings of the ventricle’ endocardial border in the apical (ventricular long-axis) and parasternal short-axis views at end-diastole and end-systole were obtained and estimates of ventricular volumes generated using a biplane modified Simpson’s rule. Ventricular morphology was characterized as left dominant (e.g., tricuspid atresia), right dominant (e.g., hypoplastic left-heart syndrome), or mixed (e.g., unbalanced AV canal defect).
For each patient, a single observer graded each AV valve’ regurgitation qualitatively as absent or trivial, mild, moderate, or severe. Subjects were classified as having moderate or severe AV valve regurgitation if the right, left, or common AV valve regurgitation was moderate or severe or if both right and left AV valve regurgitation were graded mild.
CMR Protocol
CMR was performed using a standardized imaging protocol developed by the core laboratory. Imaging was performed using locally available 1.5-T whole-body scanners (Signa LX or TwinSpeed, GE Medical Systems, Milwaukee, WI; Intera, Philips Medical Systems, Andover, MA; and Sonata or Maestro, Siemens Medical Systems, Erlangen, Germany). The standardized imaging protocol included electrocardiographically gated segmented k-space fast (turbo) gradient (14% of studies) or steady-state free precession (86% of studies) cine magnetic resonance acquisitions in the vertical and horizontal long-axis planes and contiguous short-axis cine imaging from the AV junction through the cardiac apex. Studies were performed within 3 months of echocardiography. Deidentified CMR data were analyzed using commercially available software (MASS; Medis Medical Imaging, Leiden, The Netherlands) at the core CMR laboratory by a single observer.
Statistical Analysis
Summary statistics are presented as mean ± SD or as medians and interquartile ranges. Patient characteristics for the group with mean dP/dt ic values and those without this measurement were compared using Fisher’ exact test for categorical variables, the t test for continuous nonskewed variables, and Wilcoxon’ rank-sum test for other continuous variables. Mean dP/dt ic was divided by end-diastolic volume (EDV)/(body surface area [BSA]) 1.3 to account for preload factors. We used generalized additive modeling to assess whether EF was linearly associated with the nongeometric parameters. Log-transformed mean dP/dt ic had a linear association with EF and was therefore also used in correlation analyses. Linear regression of EF with an interaction term between valve regurgitation grade and predictor was used to identify differential correlations within patient subgroups defined by severity of valve regurgitation. An interaction term between ventricular morphology and dP/dt ic was also fit to identify differential correlations within patient subgroups defined by ventricular morphology. P values < .05 were considered significant. All analyses were conducted using SAS version 9.2 (SAS Institute Inc., Cary, NC) and S-Plus (Insightful Corporation, Seattle, WA).
Results
Study Subjects
Of the 546 patients enrolled in the Fontan Cross-Sectional Study, 449 had data from which the mean dP/dt ic could be calculated ( Figure 3 ). Of these, 137 had CMR data from which the CMR-derived EF could be measured. Sample size was similar for the other nongeometric indices of ventricular function. The 137 patients who had both CMR and mean dP/dt ic data were more likely to have had tricuspid atresia (30% vs 19%), were older (12.1 ± 3.4 vs 11.8 ± 3.4 years), and had undergone Fontan surgery less recently than the other patients in the cohort.
Mean dP/dt ic and Subgroup Factors
In the cohort of 137 patients with CMR data, the mean dP/dt ic was 1,224 ± 774 mm Hg/sec (median, 977 mm Hg/sec; range, 381–5,857 mm Hg/sec). The mean CMR-derived EF was 57 ± 9%. Twenty percent had moderate AV valve regurgitation (none had severe regurgitation). The distribution of ventricular morphology was 53% left (usually tricuspid atresia), 27% right (usually hypoplastic left-heart syndrome), and 20% mixed type (usually unbalanced AV canal variant). In the larger cohort of 449 patients with mean dP/dt ic data regardless of the availability of CMR data, the average value was slightly higher than that of the cohort with CMR data (1,410 ± 935 mm Hg/sec, P < .05; Table 1 ). For these patients, the distributions of valve regurgitation and ventricular morphology were similar to those of the smaller CMR cohort.
| Variable | n | Mean ± SD | Median | IQR | Range |
|---|---|---|---|---|---|
| Mean dP/dt ic (mm Hg/sec) | 449 | 1,410 ± 935 | 1,125 | 802–1,700 | 239–5,857 |
| Log mean dP/dt ic | 449 | 7.1 ± 0.6 | 7.0 | 6.7–7.4 | 5.5–8.7 |
| Mean dP/dt ic /EDV/(BSA) 1.3 | 359 | 26.1 ± 19.1 | 20.8 | 13.8–30.8 | 2.8–113.1 |
| Log mean dP/dt ic /EDV/(BSA) 1.3 | 359 | 3.0 ± 0.7 | 3.0 | 2.6–3.4 | 1.0–4.7 |
| echocardiographic ejection fraction (%) | 414 | 58.5 ± 10.4 | 59.3 | 52.4–66.0 | 25.9–85.3 |
| S′ (cm/sec) | 428 | 6.4 ± 2.0 | 6.2 | 5.0–7.5 | 2.4–14.8 |
| Tei index | 462 | 0.64 ± 0.19 | 0.61 | 0.52–0.71 | 0.27–1.62 |
| CMR ejection fraction (%) | 161 | 56.9 ± 9.5 | 57.7 | 51.4–64.2 | 14.2–78.0 |
Correlation between CMR-Derived Measurements of EF and Nongeometric Indices of Ventricular Function
Log mean dP/dt ic was weakly correlated with CMR-derived EF ( r = 0.20, P = .022; Table 2 ). It was not significantly correlated with CMR-based measurements of end-diastolic volume, end-systolic volume, mass, or stroke volume. Because the mean dP/dt ic may be influenced by preload, we attempted to correct for ventricular preload by indexing log (mean dP/dt ic ) using [EDV/(BSA)] 1.3 . This correction strengthened the correlation with the CMR-derived EF ( r = 0.30, P = .001; Table 2 ; Figure 4 ).
| Index | n | r | P |
|---|---|---|---|
| Log mean dP/dt ic | 137 | 0.20 | .022 |
| Log mean dP/dt ic /EDV/(BSA) 1.3 | 110 | 0.30 | .001 |
| Tei index | 130 | −0.12 | .16 |
| S′ | 124 | 0.07 | .41 |
The correlation between CMR-derived EF and log mean dP/dt ic was not affected by the presence or absence of moderate AV valve regurgitation ( P = .33). Similarly, the correlation between CMR-derived EF and log mean dP/dt ic was not influenced by ventricular morphology.
Statistically significant correlations did not exist between the CMR-derived EF and the other nongeometric echocardiographic indices of ventricular function (Tei index and S′; Table 2 , Figure 4 ). Once again, these correlations were not affected by ventricular morphology.
Comparison of Patients with Echocardiographic and CMR Studies Performed on the Same Day Versus Patients with Studies Performed on Different Days
Of the patients who had CMR studies, 100 had echocardiographic studies performed on the same day as CMR (group 1), and 90 had the studies on different days (within 3 months of each other, without any intervening major therapeutic intervention; group 2). Patients in group 1 did not differ from those in group 2 with regard to age at study (11.9 ± 3.5 vs 12.7 ± 3.1 years, P = .09) or age at Fontan completion (3.5 ± 2.2 vs 3.4 ± 1.9 years, P = .72). When the analyses were limited to patients who had studies on the same day, the correlation between log mean dP/dt divided by [EDV/(BSA)] 1.3 and CMR EF remained significant ( P = .036 rather than P = .001) despite the smaller sample size, because the correlation coefficient changed very little in the analysis restricted to this smaller subpopulation ( r = 0.27 rather than r = 0.30). In this subpopulation, the correlation coefficients between the CMR-based EF and Tei index and S′ were also similar to those encountered in the entire study population and remained statistically insignificant.
Correlation between Echocardiographically Derived Measurements of EF and Nongeometric Indices of Ventricular Function
Of the 546 patients enrolled in the Fontan Cross-Sectional Study, 414 (76%) had echocardiographic images from which the EF could be calculated (59 ± 10%). Of these, 359 had data from which the mean dP/dt ic could be calculated. Log mean dP/dt ic was correlated with the echocardiography-based estimates of EF ( r = 0.18, P < .001). Once again, this correlation was strengthened modestly when log mean dP/dt ic was corrected for preload ( r = 0.20, P < .001). Statistically significant correlations ( P < .001) also existed between S′, Tei index, and echocardiography-based estimates of EF, but these correlations were weaker than the correlation between EF and log mean dP/dt ic ( Table 3 ).
