Objective. Echocardiography has been shown to be highly accurate in infants and children, but its accuracy has not been studied in detail in low birth weight (LBW) infants with structural congenital heart disease (CHD). This study was undertaken to examine the diagnostic accuracy of echocardiography in the evaluation of CHD in these infants.
Methods. All echocardiograms performed between 1995 and 2000 on infants who had structural CHD and weighed ≤2.5 kg were reviewed (n = 251). Infants who weighed >2.5 kg, matched for type of CHD, were control subjects (n = 319). The results of initial echocardiograms were compared for diagnostic accuracy with consensus diagnoses on the basis of all confirmatory data available. An observer who was blinded to patient group graded a sample of randomly selected echocardiograms (n = 100) for technical quality.
Results. There were 13 major diagnostic errors in the LBW infants (5.2%) and 6 in the control subjects (1.9%). There were 20 minor errors in the LBW group (8.0%) and 21 in the control subjects (6.6%). The technical quality scores of the 2 groups were not significantly different, but a higher proportion of the LBW studies were graded as “poor” or “borderline” quality, compared with the control subjects (40% vs 28%). The majority (54%) of major errors in the LBW infants could potentially be explained by alterations in blood flow secondary to pulmonary hypertension.
Conclusions. Although echocardiography is an accurate imaging modality in premature and LBW infants with structural CHD, the incidence of clinically important diagnostic errors is higher than in infants who weigh >2.5 kg. As surgical and transcatheter interventions are extended to this population, heightened awareness of clinicians to pitfalls of echocardiography in this group of patients is warranted.
Echocardiography has evolved to become the primary diagnostic imaging modality in the diagnosis of structural congenital heart disease (CHD) and is now used extensively as the sole pre- and postoperative diagnostic test for many patients.1–10 Several studies have demonstrated the accuracy of echocardiography in the assessment of CHD in infants and children.11–13 Tworetzky et al9 found a 2% rate of major errors in a study of 412 patients who underwent cardiovascular surgery based on echocardiography alone. Gutgesell et al14 showed a sensitivity of 87% for echocardiography in prospectively identifying 259 abnormalities that were diagnosed by catheterization.
Premature and low birth weight (LBW) infants constitute an important group of pediatric patients with CHD, in whom corrective cardiac surgery is increasingly performed early in life.15–17 Echocardiography is particularly important for preoperative diagnosis in these patients because diagnostic cardiac catheterization in premature and LBW infants is associated with relatively high rates of morbidity and mortality.18,19 However, echocardiography in premature and LBW infants presents additional technical challenges. Factors such as small body size, associated lung disease, pulmonary hypertension, occasional use of high-frequency mechanical ventilation, increased risk of sedation, and intolerance to the procedure can potentially limit the ability of echocardiography to accurately assess cardiovascular anatomy and physiology. As the diagnostic accuracy of echocardiography in this group of patients has not been systematically evaluated, this study was undertaken to address this question in a large referral center linked to 2 high-volume NICUs.
The study was designed as a retrospective case-control analysis to compare the diagnostic accuracy of echocardiography for premature and LBW infants (study group) with a contemporary group of full-term patients with CHD (control group). The database of the Cardiovascular Program at Children's Hospital Boston was searched for the records of all patients who weighed ≤2.5 kg at the time of their initial echocardiogram. Patients who fulfilled the following criteria were included: (1) diagnosis of structural CHD by any diagnostic modality, (2) initial echocardiogram in database performed at Children's Hospital or affiliated NICUs or nursery units, and (3) available follow-up data. Patients with an isolated patent ductus arteriosus (PDA) or an isolated patent foramen ovale were excluded. The study group, composed of infants who were imaged from October 1995 through December 2000, was compared with a control group of infants who were matched on the basis of anatomic type of CHD. For the control group, the database was searched on the basis of the following criteria: (1) date of study between 1995 and 2000, (2) weight at first study >2.5 kg, (3) age at study <6 months, and (4) diagnostic category. Patients were chosen at random from the resultant list of records.
For evaluating trends in use of diagnostic tests and mortality related to structural CHD in premature and LBW infants since the advent of 2-dimensional echocardiography in our laboratory, comparison groups of all such infants with structural CHD were drawn from the following periods: 1984–1986, 1987–1989, and 1990–September 1995. The Committee on Clinical Investigations at Children's Hospital Boston gave permission to conduct a medical record and database review.
Echocardiograms were performed using several commercially available cardiac ultrasound scanners. The same ultrasound equipment was used for examinations at the bedside (portable studies) and in the Echocardiography Laboratory at Children's Hospital. Transducer frequency was chosen to obtain an optimal balance between spatial resolution (higher frequency) and penetration (lower frequency) and was adjusted on an individual basis. The examination protocol included 2-dimensional and Doppler imaging from the subxiphoid, apical, parasternal, suprasternal, and, when necessary, modified views as previously described.20 Studies were recorded on 1.27-cm (0.5-inch) super-VHS videocassette tapes and reviewed off-line by an attending pediatric echocardiographer. All interpretations and diagnostic classifications were stored in an electronic database.
The original echocardiographic diagnoses were compared with consensus diagnoses on the basis of the following sources: (1) results of subsequent imaging tests, consisting of either repeat echocardiograms or cardiac MRI; (2) cardiac catheterization; (3) surgery; (4) clinical follow-up; and (5) autopsy.
Echocardiography was determined to have diagnosed a lesion when the finding was positively identified or described as suspected in the original echocardiography report. A lesion was considered undetected when not mentioned or reported as “ruled out.”
To compare the technical quality of echocardiography between the study and control groups, we randomly selected 50 study subjects and 50 control subjects. Their initial videocassette tapes were reviewed for technical quality by an echocardiographer (J.C.L.) who was blinded to the patient group assignments. A score was assigned to each study on a 1 to 4 scale as follows: (1) poor quality (key structures not seen), (2) borderline quality (key structures incompletely seen), (3) good quality (all key structures seen but not “perfect” image quality), and (4) excellent quality (all key structures clearly seen with good image quality).
The diagnoses recorded in the database for each initial echocardiogram were compared with consensus diagnoses derived from summation of all available sources, including subsequent diagnostic tests, surgical observations, clinical data, and findings at autopsy. All discrepancies were reviewed by 2 investigators and classified as (1) none, (2) minor (when no significant clinical impact was expected), or (3) major (when the discrepancy was judged to have a potentially significant impact on patient treatment or outcome). The analysis focused on anatomic diagnoses of structural CHD and did not include functional diagnoses such as grades of ventricular dysfunction or valve regurgitation. Diagnoses that are generally considered common in neonates such as patent foramen ovale and PDA were not included in the analysis. A secundum atrial septal defect was differentiated from a patent foramen ovale as described by Gnanapragasam et al.21
A commercially available statistical package was used for data analysis (STATA version 7.0; STATA Corp, College Station, TX). Study subjects and control subjects were compared using the Wilcoxon rank sum test for continuous variables and Fisher exact test for categorical variables. Differences between premature and LBW infants across time periods were evaluated using the Kruskal-Wallis test for continuous variables and Fisher exact test for categorical variables. Technical quality of echocardiograms was compared between study subjects and control subjects using Fisher exact test.
The database search identified 320 study patients. Of these, 69 patients were excluded from additional analysis for lack of follow-up. No patient was excluded on the basis of incomplete data on the initial echocardiogram. The final study group consisted of 251 infants. The control group comprised 320 patients, chosen on the basis of the initial diagnoses of the 320 study patients. One of those had no additional follow-up, leaving a final group of 319 control subjects. There was no significant difference in the proportion of anatomic diagnoses between the final study and control groups. Of the 251 echocardiograms performed in the study patients, 227 (90.4%) examinations were done at the bedside, most commonly in the ICU. Patient demographics and rates of catheterization, surgery, and mortality are summarized in Table 1. The consensus diagnoses are summarized in Table 2.
Table 3 summarizes the sources of data used to formulate the consensus diagnoses. A higher percentage of patients in the study group had autopsies, reflecting their overall higher mortality rate. A significantly higher number of control patients had clinical follow-up alone.
Major discrepancies were detected in 13 (5.2%) of the 251 study patients and in 6 (1.9%) of the 319 control patients (P = .035; Table 4). Of the 13 major discrepancies noted in the study patients, 9 were false negative (2 atrial septal defects, 2 ventricular septal defects, 2 pulmonary valve stenosis, 1 cor triatriatum, 1 aortic coarctation, and 1 mitral valve stenosis). The remaining 4 patients included a false-positive diagnosis of anomalous left coronary artery from the main pulmonary artery, a false-positive diagnosis of a large secundum atrial septal defect, and 2 errors in anatomic diagnoses (1 based on autopsy and 1 based on surgical findings). In the control group, 5 patients had false-negative diagnoses (1 atrial septal defect, 1 pulmonary valve stenosis, 1 scimitar syndrome, 1 major aortopulmonary collateral, and 1 aortopulmonary window). One patient had an incorrect anatomic diagnosis based on surgical findings. Minor discrepancies were detected in 20 (8.0%) of the study patients and in 21 (6.6%) of the control patients (P = .63; Table 5). No single type of cardiac anomaly was dominant in either group.
Figure 1 shows the distribution of scores of the 100 echocardiograms that were reviewed for technical quality. Overall, there was no statistically significant difference between the 2 groups. However, compared with control subjects, a higher proportion of echocardiograms in the study group were graded in the “poor” or “borderline” categories (40% vs 28%; P = .29).
Trends in Diagnosis and Outcome
Table 6 compares the demographic and clinical data of the contemporary study group with premature and LBW infants whose disease was diagnosed between 1984 and 1995. The proportions of patients who underwent cardiac catheterization and mortality rate decreased significantly during the study period.
The results of this study show that the rate of clinically important diagnostic errors of echocardiography is higher in LBW (≤2.5 kg) infants than in larger infants. However, the overall error rate of echocardiography in this population is low, and it is an excellent diagnostic tool when structural CHD is suspected. Indeed, during the study period, echocardiography largely replaced cardiac catheterization in this patient population, with the majority of premature and LBW infants who underwent cardiovascular surgery on the basis of echocardiography alone.
Potential Sources of Diagnostic Errors
Although it is difficult to determine with certainty which of the many factors that can potentially contribute to diagnostic errors in echocardiography dominates, the results of this study provide some clues. Variations in technical proficiency or interpretive skills were unlikely to have played a significant role in the differential error rates, as the echocardiograms were performed and interpreted on both study and control groups in the same period. Therefore, it is worth considering 2 factors that distinguish performance and interpretation of echocardiograms in premature and LBW infants: technical and physiologic considerations.
From a technical standpoint, premature and LBW infants pose a challenge to echocardiography as a result of restricted acoustic windows and/or cardiorespiratory instability. Acoustic windows in these patients are often limited as a result of the high prevalence of respiratory problems, including chronic lung disease, air leaks, and hyperinflation. Moreover, the use of high-frequency oscillatory ventilation further interferes with cardiac ultrasound imaging and Doppler evaluation. Cardiorespiratory instability poses a particular problem when patients do not tolerate the pressure exerted by the ultrasound transducer and leads to a greater reluctance to sedate LBW infants, potentially further limiting their studies. In addition, the small body size and presence of monitoring equipment, indwelling lines, and chest tubes further limits the space available to apply and manipulate the ultrasound transducer. Despite these considerations, this study did not detect a statistically significant difference between the study patients and control subjects with regard to technical quality. It is worth noting, however, that there was a trend toward a higher proportion of echocardiograms in the study patients with “poor” or “borderline” quality. These observations suggest that differences in technical quality may have contributed but probably did not play a major role in the higher rate of clinically important diagnostic errors in this study.
Premature infants often have important cardiorespiratory physiologic differences compared with full-term infants. This population has a high prevalence of pulmonary hypertension as a result of immaturity of the pulmonary vascular bed, parenchymal lung disease, and PDA. Pulmonary hypertension could potentially explain 7 (54%) of the 13 missed diagnoses in the study group, including missed ventricular septal defect in 2, atrial septal defect in 3, and pulmonary valve stenosis in 2 patients. This hypothesis is consistent with the findings of Moss et al,22 who assessed the reliability of echocardiography in the NICU. In that report, of the 7 discrepancies found between studies performed by neonatologists and cardiologists, 3 were misdiagnoses of pulmonary valve stenosis and 2 were missed PDA, all of which could potentially be explained by high pulmonary pressures.
The findings of this study highlight the need for particular vigilance when performing and interpreting echocardiograms on premature and LBW infants. In general, color Doppler flow mapping is a highly sensitive tool to detect intra- and extracardiac shunts and abnormal flow jets characteristic of valve stenosis. However, in the presence of pulmonary hypertension, an anatomic defect or shunt can be missed because the typical appearance of a high-velocity flow jet depends on significant pressure drop between adjacent chambers or vessels. These diagnoses are most reliably made when there is data agreement between 2-dimensional imaging and color Doppler flow mapping. In the absence of this confirmation, the diagnosis is less secure. In light of these results, it seems prudent to maintain a low threshold for repeat echocardiograms or to use alternative imaging tests in these patients whenever there is inadequate 2-dimensional or color Doppler imaging or inconsistency between results of the echocardiogram and clinical course.
The retrospective nature of this study did not allow complete case ascertainment, and some patients were lost to follow-up. However, it is reasonable to assume that most clinically significant missed diagnoses would likely have presented during follow-up, an assumption based on patterns of referral to cardiac care in New England. Anomalies that typically present at an older age, such as bicuspid aortic valve, were more likely not to be captured in this study. Finally, the study lacks a “gold standard” for diagnostic comparison. However, every method of diagnosis has limitations, and no technique has been documented to be completely accurate. Therefore, the use of “consensus diagnoses” based on summation of clinical course and all available diagnostic information provides the best opportunity to recognize clinically important missed diagnoses. A prospective study with long-term follow-up is warranted.
Echocardiography is an accurate diagnostic test for the evaluation of CHD but has a higher error rate in LBW infants. Particular vigilance is recommended in interpreting the echocardiograms of these infants with a low threshold for repeat examinations when image quality is suboptimal or when the clinical course is inconsistent with the imaging data.
- Accepted June 30, 2004.
- Reprint requests to (T.G.) Department of Cardiology, Children's Hospital Boston, 300 Longwood Ave, Boston, MA 02115. E-mail:
No conflict of interest declared.
- Leung MP, Mok CK, Lau KC, Lo R. The role of cross sectional echocardiography and pulsed Doppler ultrasound in the management of neonates in whom congenital heart disease is suspected. Br Heart J.1986;56 :73– 82
- Pasquini L, Sanders SP, Parness IA, Colan SD. Diagnosis of coronary artery anatomy by two-dimensional echocardiography in patients with transposition of the great arteries. Circulation.1987;75 :557– 564
- ↵van der Velde M, Parness I, Colan SD, et al. Two-dimensional echocardiography in the pre- and postoperative management of totally anomalous pulmonary venous connection. J Am Coll Cardiol.2003;18 :1746– 1751
- ↵Geva T. Echocardiography and Doppler ultrasound. In: Garson A, Bricker JT, Fisher DJ, Neish SR, eds. The Science and Practice of Pediatric Cardiology. Baltimore, MD: Williams & Wilkins; 1997:789–840
- ↵Gnanapragasam JP, Houston AB, Doig WB, et al. Influence of colour Doppler echocardiography on the ultrasonic assessment of congenital heart disease: a prospective study. Br Heart J.1991;66 :238– 243
- ↵Moss S, Kitchiner DJ, Yoxall CW, Subhedar NV. Evaluation of echocardiography on the neonatal unit. Arch Dis Child Fetal Neonatal Ed.2003;88 :F287– F289
- Copyright © 2005 by the American Academy of Pediatrics