Background
Some neonates with tetralogy of Fallot (TOF) have rapid progression of right ventricular outflow tract obstruction, requiring early repair irrespective of Doppler gradient as measured in the neonatal period. The aim of this study was to test the hypothesis that infundibular morphology in neonates with TOF is associated with the occurrence of hypercyanotic spells and need for neonatal surgery.
Methods
Fifty patients with TOF undergoing surgical repair from 2003 to 2009 were studied. Neonatal echocardiograms were retrospectively analyzed to measure conal septal angle (the angle between the conal septum and the horizontal plane passing through the center of the aortic valve in the parasternal short-axis view, with a larger angle denoting more anterocephalad deviation of conal septum), conal septal thickness and length, the degree of aortic dextroposition, and sizes and Z scores of the pulmonary annulus and the main and branch pulmonary arteries. Outcomes included the occurrence of hypercyanotic spells and the need for neonatal surgery.
Results
The median age at first echocardiogram was 2 days (range, 0–12 days). The median age at surgery was 94 days (range, 5–282 days); hypercyanotic spells occurred in 17 patients (34%), and nine (18%) underwent neonatal repair. The presence of a wider conal septal angle was significantly associated with the occurrence of hypercyanotic spells (59 ± 21° vs 48 ± 13°, P = .023) and the need for neonatal surgery (67 ± 13° vs 48 ± 16°, P = .004). The positive and negative predictive values of hypercyanotic spells for conal septal angles ≥60° were 64% and 78%, respectively. Importantly, Doppler right ventricular outflow tract gradient at initial echocardiography, degree of aortic dextroposition, and pulmonary or aortic valve size were not associated with these outcomes.
Conclusions
A wider conal septal angle is associated with the occurrence of hypercyanotic spells and the need for neonatal surgery.
Tetralogy of Fallot (TOF) is one of the most common cyanotic congenital heart diseases, with an incidence of about 0.2 per 1,000 live births. Abnormal neural crest migration is thought to result in an anterocephalad displacement of the infundibular septum in relation to the rest of the ventricular septum. This infundibular septal deviation results in a large, malaligned ventricular septal defect, aortic override, and right ventricular outflow tract (RVOT) obstruction.
Symptoms of untreated patients depend on the degree of the RVOT obstruction. Patients with minimal RVOT obstruction may have symptoms of a large ventricular septal defect, may be fully saturated, and may in fact have “congestive heart failure” from pulmonary overcirculation; at the other extreme, patients with critically severe RVOT obstruction will be profoundly cyanotic and may be dependent on patency of the ductus arteriosus for pulmonary perfusion. The natural history in TOF is for RVOT obstruction to progress with time, but the rate at which this happens is greatly variable in individual patients, and the ability to clinically predict how an individual patient will progress is currently limited. It is well recognized that in the early neonatal period, the peak Doppler gradient may underestimate the severity of RVOT obstruction because of elevated pulmonary vascular resistance. Some patients thought to have “mild” RVOT obstruction on the basis of low Doppler gradient in the early neonatal period become significantly cyanotic soon thereafter and may present with hypercyanotic spells that may be life threatening or at the very least result in unplanned admission and/or urgent need for surgical intervention. Although some centers have recommended early neonatal surgery to circumvent this unpredictability, data clearly suggest an increased incidence of junctional ectopic tachycardia, longer intensive care unit and hospital stays, more complicated recovery, and increased need for valve-sacrificing transannular patch repairs in patients who undergo surgery in the neonatal period. If clinicians had a simple and reliable echocardiographic tool to identify patients at higher risk for rapid progression of RVOT obstruction, it might allow safer surveillance of these patients before surgery.
We hypothesized in neonates with TOF, echocardiographic morphometrics of the RVOT may predict those who are at risk for hypercyanotic spells and those who require neonatal surgery.
Methods
Study Design
We retrospectively reviewed the medical records of patients with TOF who met our inclusion criteria and who underwent either TOF repair or shunt placement at our institution between 2003 and 2009. The study was approved by the institutional review board of the University of Arkansas Medical Center, Arkansas Children’s Hospital.
Study Population
Sixty-one subjects with TOF, pulmonary stenosis underwent surgical repair between 2003 and 2009 at our institution. We excluded subjects who were diagnosed with TOF after the neonatal period and those with pulmonary valve atresia. We also excluded patients with other associated cardiac lesions, such as double-outlet right ventricle, atrioventricular septal defect, or absent pulmonary valve syndrome. No patient had a larger than small patent ductus arteriosus at the time of initial echocardiography. Fifty patients were included for the final data analysis. We excluded 11 patients from the final analysis because of diagnosis in the postneonatal period ( n = 4), inadequate images ( n = 5), and missing data ( n = 2).
The following echocardiographic parameters were measured in these patients: (1) Conal septal angle: the conal septum in normal subjects divides the proximal aorta from the pulmonary artery, whereas in patients with TOF, the conal septum is displaced anteriorly and cranially in relation to the rest of the interventricular septum. The conal septal angle is the angle between the conal septum and the horizontal plane passing through the center of the aortic valve in the parasternal short-axis plane ( Figure 1 ). (2) Conal septal length: the greatest measurement of the conal septum was measured in the parasternal short-axis plane. (3) Degree of aortic dextroposition: rightward deviation of the aorta was measured in the parasternal short-axis plane, as described by Isaaz et al. The line passing through the atrial septum was extrapolated to intersect the aorta, as shown in Figure 2 . The proportion of the aorta lying to the right of this line was measured, and the ratio of this dimension to the aortic diameter was recorded as a measure of aortic dextroposition. (4) Anterior malalignment of the conal septum: anterior deviation of the conal septum was measured in the parasternal short-axis view. A line, a, was drawn from the center of the aorta to the tip of the conal septum; another line, b, was drawn from the center of aorta to the endocardial lining of the RVOT. Anterior malalignment was expressed as the ratio of these two lines, as shown in Figure 3 . (5) Sizes and Z scores of the aortic valve, pulmonary valve annulus, and main and branch pulmonary arteries at their origin. (6) The peak pressure gradients across the RVOT on initial and follow-up echocardiography up to initial surgery were recorded.
Medical records were reviewed for demographic information, serial pulse oximetry saturation until surgery, the presence of hypercyanotic spells, and the type and date of initial surgery. The initial echocardiogram and all follow-up echocardiograms until initial surgery were reviewed.
Echocardiographic measurements in all 50 subjects were made by a single observer (S.C.U.); the conal septal angle was analyzed for interobserver reliability by two observers (S.C.U. and H.V.V.) for all patients.
The outcomes analyzed included the occurrence of hypercyanotic spells and the need for surgery in the first month of life. For study purposes, a hypercyanotic spell was defined as an episodic decline in oxygen saturation of >10% from baseline lasting >20 sec without obvious respiratory etiology, the episodic occurrence of increased central cyanosis as perceived by parents or other nonmedical caregivers, or documentation in the medical record by the attending physician or pediatric cardiologist of the occurrence of a hypercyanotic spell if the pulse oximetry reading was not recorded in the chart.
Statistical Analysis
Summary statistics were estimated for demographic, echocardiographic, and clinical outcomes, with continuous variables reported as mean ± SD or as median (interquartile or absolute range) and categorical data reported as frequency (percentage). Wilcoxon’s rank-sum tests were used to test for associations between echocardiographic parameters or patient characteristics and the outcomes (i.e., hypercyanotic spells and neonatal surgery). Logistic regression models were used to further evaluate any association determined in the univariate analysis and outcomes adjusting for age at echocardiography. Explanatory variables used in the logistic regression were parameterized using restricted cubic splines with three knots to relax the strict linearity assumption implied within linear regression models. This parameterization allows for changes in the magnitude of an association between an explanatory variable and an outcome (e.g., similar to the change associated with a cut point), while maximizing power using a continuous explanatory variable. After estimating the model, goodness of fit was assessed using the Hosmer-Lemeshow test, and overall predictive capacity was estimated using the C-statistic. The Hosmer-Lemeshow test evaluates whether the observed event rate matches the predicted rate in quantiles of estimated risk. The area under the receiver operating characteristic (ROC) curve, also known as the C-statistic, was used to quantify the model’s ability to discriminate between outcomes. The area under the ROC curve was estimated in the same sample that was used to construct the logistic regression model; to correct for potential overestimation, we also used a bootstrap resampling method to estimate a bias-corrected area under the ROC curve. Although bootstrapping provides a bias-corrected C-statistic, estimates of sensitivity and specificity are not bias corrected and may be overly optimistic. Detailed predictive capacity summarized by sensitivity, specificity, and positive and negative predictive capacity was also estimated for varying conal septal angles for each outcome.
Interobserver reliability of conal septal angle measurements was estimated using Lin’s concordance correlation, a statistic that combines Pearson’s correlation with an estimate of accuracy. The representation of agreement can be visually assessed by comparing the estimated agreement with a line of perfect agreement represented by a 45° line.
Results
Data analysis was performed on 50 patients who met our inclusion criteria. Nine subjects (18%) underwent neonatal surgery ( Table 1 ), and 17 (34%) had hypercyanotic spells ( Table 2 ).
| Variable | Post–neonatal surgery ( n = 41) | Neonatal surgery ( n = 9) | P ∗ |
|---|---|---|---|
| Hypercyanotic spells [N (%)] | 9 (22%) | 8 (89%) | <.001 † |
| Age at initial echocardiographic study (d) | 2.0 (0.0 to 7.0) | 2.0 (0.0 to 2.0) | .242 |
| Weight at initial echocardiographic study (kg) | 3.2 (2.7 to 3.4) | 3.0 (2.4 to 3.2) | .261 |
| Age at surgery (d) | 98.0 (67.0 to 142.0) | 10.0 (7.0 to 18.0) | <.001 |
| Conal septal length (mm) | 7.6 (7.0 to 8.9) | 6.9 (6.0 to 7.4) | .054 |
| Aortic dextroposition | 5.1 (4.7 to 5.8) | 4.8 (3.2 to 5.4) | .184 |
| Conal septal angle (parasternal short-axis view) (°) | 49.3 (41.0 to 56.0) | 68.5 (59.5 to 77.0) | .004 |
| Anterior malalignment of conal septum | 0.6 (0.5 to 0.7) | 0.7 (0.6 to 0.7) | .181 |
| Peak gradient at initial echocardiographic study (mm Hg) | 36.0 (21.0 to 49.0) | 43.0 (41.5 to 49.0) | .344 |
| Aortic annular Z score | 2.0 (1.6 to 2.7) | 2.1 (0.9 to 3.1) | .605 |
| Main pulmonary artery Z score | −2.1 (−3.0 to −0.8) | −2.2 (−4.1 to −1.9) | .331 |
| Right pulmonary artery Z score | −1.0 (−2.2 to 0.0) | −0.9 (−1.9 to −0.3) | .752 |
| Left pulmonary artery Z score | −0.9 (−1.6 to 0.0) | −0.6 (−1.3 to −0.3) | .870 |
| Pulmonary annular Z score | −2.6 (−3.4 to −1.9) | −3.5 (−3.8 to −2.3) | .230 |
∗ Wilcoxon’s rank-sum test except as indicated.
| Variable | No hypercyanotic spells ( n = 33) | Hypercyanotic spells ( n = 17) | P ∗ |
|---|---|---|---|
| Age at initial echocardiographic study (d) | 2.0 (1.0 to 7.0) | 1.0 (0.0 to 2.0) | .027 |
| Weight at initial echocardiographic study (kg) | 3.2 (2.9 to 3.5) | 2.7 (2.4 to 3.1) | .017 |
| Age at surgery (d) | 103.0 (71.0 to 151.0) | 50.0 (16.0 to 85.0) | <.001 |
| Conal septal length (mm) | 8.0 (7.0 to 9.1) | 7.2 (6.3 to 7.6) | .037 |
| Aortic dextroposition | 5.4 (4.8 to 5.9) | 5.0 (3.9 to 5.2) | .101 |
| Conal septal angle (parasternal short-axis view) | 49.0 (45.0 to 54.0) | 60.2 (50.5 to 75.5) | .022 |
| Anterior malalignment of conal septum | 0.6 (0.5 to 0.6) | 0.7 (0.6 to 0.7) | .073 |
| Peak gradient at initial echocardiographic study (mm Hg) | 37.3 (30.0 to 54.0) | 29.0 (21.0 to 48.9) | .243 |
| Aortic annular Z score | 2.0 (1.6 to 3.3) | 2.1 (1.3 to 2.5) | .461 |
| Main pulmonary artery Z score | −2.1 (−3.0 to −0.6) | −2.2 (−3.6 to −1.9) | .357 |
| Right pulmonary artery Z score | −1.0 (−2.2 to 0.0) | −1.0 (−2.3 to −0.2) | .580 |
| Left pulmonary artery Z score | −0.8 (−1.6 to 0.0) | −0.9 (−1.5 to 0.2) | .830 |
| Pulmonary annular Z score | −2.5 (−3.4 to −1.8) | −3.2 (−3.8 to −2.7) | .076 |
There were no in-hospital deaths. The median age at first echocardiographic study was 2 days (range, 0–12 days). The median age at surgery in the postneonatal surgery group was 98 days (interquartile range, 67–142 days), whereas in the group requiring neonatal surgery, the median age was 10 days (interquartile range, 7–18 days). The median weight in the neonatal surgical group was 2.95 kg (interquartile range, 2.5–3.7 kg), whereas the median weight in the postneonatal patients was 5.6 kg (interquartile range, 4.5–6.2 kg) ( Table 1 ). Of the 17 patients who had hypercyanotic spells, eight (47%) required neonatal surgery, whereas only one patient with no hypercyanotic spells required neonatal surgery, and this patient underwent modified Blalock-Taussig shunt placement. This patient had critical RVOT obstruction due to valvular stenosis. Six of nine patients (67%) who underwent neonatal surgery underwent modified Blalock-Taussig shunt placement. Neonatal transannular patch repair was performed in three patients after the first week of life. There were no significant differences among the patients with hypercyanotic spells and those without spells with regard to height, weight, and oxygen saturation at initial presentation and serial measurements ( Table 2 ).
On univariate analysis, we found that the patients who underwent surgery in the first month of life had significantly wider conal septal angles (66.5 ± 13.2° vs 48.3 ± 15.5°, P = .003; Table 1 ). Patients with hypercyanotic spells also had significantly wider conal septal angles than those without spells (58.6 ± 20° vs 48 ± 12.8°, P = .021; Table 2 ). Patients with hypercyanotic spells had shorter conal septal lengths (7.2 ± 1.1 vs 8.0 ± 1.5 mm; Table 2 ). Hypercyanotic spells were significantly associated with conal septal angle parameterized as a restricted cubic spline ( P = .001) adjusting for age (also parameterized as a cubic spline). This regression model had a C-statistic of 0.82 (bootstrap bias-corrected estimate, 0.78), indicating good ability to discriminate between patients with and without hypercyanotic spells. The predicted probabilities from the model failed to show disagreement with the observed data (Hosmer-Lemeshow test P = .296). The predicted probabilities (with 95% confidence intervals) of hypercyanotic spells and neonatal surgery by conal septal angle are shown in Figure 4 . Sensitivity, specificity, negative predictive value, and positive predictive value for a variety of cutoffs can be found in Table 3 ; for one specific cutoff (conal septal angle ≥60°), negative and positive predictive values were 78% and 64%, respectively. Interobserver reliability of conal septal angle was assessed using Lin’s concordance correlation. Paired measurements were available for 44 observations, and the concordance correlation was 0.951 (95% confidence interval, 0.912–0.973). Using the Bland-Altman method, the bias, or average difference between raters, was −1.01 ± 4.773, and the 95% limits of agreement were −10.4 to 8.3. A plot showing the agreement of conal septal angle values between two observers is shown in Figure 5 .
