- Top of page
Much effort has been put into predicting the postnatal outcome of cases with congenital diaphragmatic hernia (CDH). Despite major improvements in perinatal care, the mortality associated with isolated CDH remains high, due mainly to pulmonary hypoplasia and pulmonary hypertension1–4, with the prediction of the neonatal outcome of these cases remaining a challenge5–8.
As severe pulmonary hypoplasia is the main cause of neonatal mortality, measuring fetal lung volume could be useful in prognosis evaluation. On conventional two-dimensional (2D) ultrasound it is possible to evaluate fetal lung volume only indirectly9. In fetuses with CDH, the lung-to-head ratio (LHR) has been shown to be one of the best prenatal parameters correlating with prognosis10–12, but is by no means without its limitations. With advances in sonography, it is possible to identify the lung ipsilateral to the diaphragmatic defect in a few cases, which is not considered in the LHR. Therefore, many studies in cases with CDH have been concentrating on measuring fetal lung volume directly. As it is technically difficult on conventional ultrasound (2D), magnetic resonance imaging (MRI) has been used, with encouraging results for the prediction of neonatal pulmonary hypoplasia and outcome of cases with CDH13–19. So far, different nomograms of fetal lung volume measured by three-dimensional (3D) ultrasound using various methods have been published20–26. Recently, our group demonstrated that 3D ultrasound can be used to measure fetal lung volumes in CDH using the rotational multiplanar technique (virtual organ computer-aided analysis (VOCAL™))27, 28. One of the main advantages of 3D ultrasound over MRI is probably the lower cost. However, to our knowledge, there are no data about the accuracy and precision of fetal lung volume assessed by 3D ultrasound in vivo, despite the intra- and interoperator variability previously evaluated29, 30.
We studied prospectively cases of isolated CDH and controls (with no pulmonary malformation) that underwent termination of pregnancy. In all cases fetal lung volumes were measured by 3D ultrasound immediately before termination of pregnancy. These prenatal sonographic measurements were compared with postmortem fetal lung volumes in order to evaluate the precision of fetal lung volume measured by 3D ultrasound.
- Top of page
Between January and December 2002, a prospective study was conducted in a single tertiary care center. 3D ultrasound was performed to evaluate fetal lung volume immediately before termination of pregnancy in eight cases of CDH (six left and two right-sided) and in 25 controls.
Terminations of pregnancies were performed at the mother's request in accordance with French Law, after the parents had been fully informed on the expected neonatal outcome. The results of 3D ultrasound were not taken into account to counsel patients. All women were fully familiarized with the study protocol and the technique. They accepted to undergo a 3D ultrasound examination before induction of labor, and consented that a postmortem examination be performed following termination of pregnancy. This study protocol was approved by the review board of the Centre Pluridisciplinaire de Diagnostic Prénatal de l'Hôpital Necker.
Among the 25 controls, the indications for termination of pregnancy were major congenital brain abnormality (n = 7), severe congenital heart defect (n = 6), aneuploidy (n = 3), fetal akinesia sequence (n = 2), multiple malformation syndrome (n = 2) osteochondrodysplasia (n = 1), large thoracocervical lymphangioma (n = 1), Pierre–Robin syndrome (n = 1), Bourneville tuberous sclerosis (n = 1) and amniotic band syndrome (n = 1). Among cases with CDH, medical termination of pregnancy was decided in those cases with a poor prognosis, namely: LHR ≤ 1.0 (n = 1) or herniated liver into fetal thorax (n = 3) and both LHR ≤ 1.0 and herniated liver (n = 4)6, 10, 12, 31, 32.
Before induction of labor, patients underwent 3D ultrasound examinations. For assessment of fetal lung volumes, a Voluson 730 ultrasound machine (General-Electric, Zipf, Austria) with a 4–8-MHz transducer for 3D volume scanning was used. Fetal lung volumes were assessed by 3D ultrasound using a rotational multiplanar technique for volume measurements (VOCAL)29, 30, 33. A transverse section of the fetal thorax at the level of the four-chamber view, with the fetal heart proximal to the transducer, was identified by 2D ultrasound and the volume box was adjusted in order to scan the entire fetal thorax. The volume sweep angle was set between 45° and 75°, depending on gestational age. The slowest scan duration (high quality) was adjusted in order to obtain the best possible resolution. In the event of fetal movements, the procedure was repeated until adequate visualization of the lungs was observed in all three planes. After scanning the volume, the three orthogonal ultrasound sections (multiplanar imaging) were analyzed and stored on a removable hard disk (Figure 1). Each lung was identified carefully in the three orthogonal multiplanar images. A transverse section of the fetal thorax using multiplanar imaging was chosen and each lung volume was measured by the rotational technique (consisting of repeatedly outlining the contour of the lung manually after rotating its image by 30° six times).
Figure 1. Three-dimensional multiplanar imaging of the fetal thorax at 27 weeks of gestation. This image demonstrates that both lungs were included entirely in the volume sample. (a) Transverse plane; (b) sagittal plane; (c) frontal plane; (d) three-dimensional rendered image. FH, fetal heart; FL, fetal liver; LL, left lung; RL, right lung; s, stomach.
Download figure to PowerPoint
Reviewing parallel slices executed through the whole lung in the three perpendicular planes allowed us to check the adequacy of our rotational area tracing approach. This was made possible because the volume defined by rotational analysis could be reconstructed and displayed using multiplanar imaging. Left and right lung volumes were measured manually (Figure 2) four times by the same operator (R.R.), and the mean volume was considered, which took approximately 5 min. When it was impossible to identify the lung on the same side as the diaphragmatic defect, the lung volume was considered to be immeasurable and a value of zero was given. Each volume was measured four times in order to evaluate the intraoperator variability.
After the 3D ultrasound examination, medical termination of pregnancy was performed by injecting thiopental (Bellon Laboratories, Rhône-Poulenc Rorer, Neuilly-Sur-Seine, France, 10 mg/kg estimated fetal weight) and sufentanyl for fetal anesthesia (Jansen-Cilag Laboratories, Paris, 1 µg/kg estimated fetal weight) into the fetal circulation under ultrasound guidance, followed by 5–15 mL potassium chloride (1.34 mMol/mL) to induce fetal cardiac arrest. Cervical ripening was achieved using mifepristone and Dilapan in all cases. Labor was induced by misoprostol in 23 cases and, where necessary, the additional administration of oxytocin (in four cases). In 10 cases, labor was induced directly by oxytocin. In all cases, vaginal delivery occurred within 1–11 (mean, 6.23; SD, 2.67) h. All patients had epidural analgesia (Table 1).
Table 1. Characteristics of labor induction in congenital diaphragmatic hernia (CDH) cases and controls
|Characteristic||CDH cases (n = 8)||Controls (n = 25)|
|Gestational age at TOP (weeks, mean (range))||28.25 (23–34)||25.54 (15–38)|
|Cervical ripening||Mifepristone + Dilapan (n = 8)||Mifepristone + Dilapan (n = 25)|
|Labor induction||Vaginal misoprostol (n = 6);||Vaginal misoprostol (n = 17);|
|Oxytocin (n = 5)||Oxytocin (n = 9)|
|Epidural analgesia (n (%))||8 (100)||25 (100)|
|Labor duration (h, mean (range))||6.16 (3–11)||5.26 (1–10)|
During the autopsy, the right and left lungs were dissected, and their volumes were measured separately by displacement of water, before fixation in formalin. The pathologist (J.M.) was not aware of the 3D ultrasound lung volume measurements. Postmortem values were compared with the 3D ultrasound estimated fetal lung volumes in both controls and cases with CDH.
3D volumetric measurements and postmortem results were compared in both control and CDH groups using the intraclass correlation coefficient and Bland–Altman analysis, which calculated the bias (the mean of the volumetric difference between the two methods), the precision (the SD of the difference between the two methods) and the absolute limits of agreement (1.96 SD of the mean difference)34–36. The relative error (RE) was also estimated by the following equation:
where X is the postmortem measurement and X0 is the estimated ultrasound measurement. Accuracy (A) was calculated by the equation:
where PD is the percent difference (|X − X0|/X × 100). This is the absolute value of the difference between both volumetric measurements divided by the postmortem measurements, expressed as a percentage.
All 3D ultrasound examinations were performed by the same operator (R.R.). Intraobserver variability was assessed by analyzing the measurements obtained from 10 randomly selected controls and from all cases of CDH (n = 8). For each pregnancy, fetal lung volume was estimated four times by this same operator and the intraobserver variability was calculated by analysis of variance (Excel, Microsoft, Redmond, WA, USA).
- Top of page
The mean gestational age at termination of pregnancy was 25.54 (range, 15–38) weeks in controls and 28.25 (range, 23–34) weeks in cases with CDH (Table 1). Fetal lung volume was on average 9.66 (range, 5.52–14.72) cm3 in cases with CDH and 24.54 (range, 0.63–57.80) cm3 in controls (Table 2).
Table 2. Mean precision, error and intraoperator variability for total fetal lung volume in cases with congenital diaphragmatic hernia (CDH) and controls
|Parameter||CDH (n = 8)||Controls (n = 25)|
|Mean FLV on 3D ultrasound (cm3 (range))||9.66 (5.52–14.72)||24.54 (0.63–57.8)|
|Intraclass correlation coefficient for total FLV||0.95||0.99|
|Bias of total FLV (cm3)||0.35||0.08|
|Precision of FLV (cm3)||1.46||2.80|
|Mean relative error for total FLV (% (range))||−7.19 (−42.70 to + 18.11)||−0.72 (−30.25 to + 19.22)|
|Accuracy for total FLV (% (range))||84.86 (57.30–99.48)||91.38 (69.75–99.45)|
|Mean absolute error for total FLV (cm3 (range))||1.40 (0.71–2.52)||2.12 (0.05–4.98)|
|Intraoperator variability (cm3)||0.17||0.28|
The intraclass correlation coefficient of fetal lung volume estimated by 3D ultrasound and measured at postmortem examination was 0.95 in cases with CDH and 0.99 in controls. The mean RE of 3D ultrasound fetal lung volume measurement was −7.19 (range, −42.70 to + 18.11)% in cases with CDH and −0.72 (range, −30.25 to + 19.22)% in controls. The mean absolute error of 3D ultrasound fetal lung volume measurement was 1.40 (range, 0.71–2.52) cm3 in cases with CDH and 2.12 (range, 0.05–4.98) cm3 in controls. Accuracy of 3D ultrasound for measuring fetal lung volume was 84.86 (range, 57.30–99.48)% and 91.38 (range, 69.75–99.45)%, respectively, in cases with CDH and in controls (Table 2).
The scattergram of the difference against the average of these two measurements according to Bland and Altman is displayed in Figure 3. The bias and precision were 0.35 cm3 and 1.46 cm3, with the absolute limits of agreement between −2.51 and + 3.21 cm3, in cases with CDH. Among controls, the bias and precision were 0.08 cm3 and 2.80 cm3, respectively, with the absolute limits of agreement between −5.41 and + 5.57 cm3.
Figure 3. Absolute limits of agreement for comparison of fetal lung volumes (FLV) estimated by three-dimensional ultrasound and measured during autopsy. Continuous lines represent the mean difference in FLV estimated by these two methods in controls (heavy line) and in cases with congenital diaphragmatic hernia (CDH) (fine line). The heavy dashed lines represent the limits of agreement for normal fetal lungs (controls) and the fine dashed lines are the limits of agreement in cases with CDH. ●, controls; ○, CDH cases.
Download figure to PowerPoint
The mean intraobserver variability for lung volume measurements estimated by 3D ultrasound in controls and in cases with CDH was 0.28 cm3 and 0.17 cm3, respectively.
- Top of page
Our results suggest that total fetal lung volume can be estimated accurately by 3D ultrasound using a rotational multiplanar imaging technique. Good accuracy and low intraoperator variability, which were observed in this study, may be explained by the rigorously methodological standardization of the acquisition and measurement of fetal lung volumes.
In this study, postmortem measurement of fetal lung volume by water displacement was considered the ‘gold standard’ with which to evaluate the accuracy of 3D ultrasound in estimating fetal lung volume. At autopsy, fetal lungs were dissected and measured directly, with no fixation or air inflation, in order to avoid any influence on the volumetric measurements. As the mean duration of labor was approximately 6 h in both groups, the delay between postmortem and sonographic measurements was less than a day, avoiding any interference due to tissue necrosis. To our knowledge, this is the first study comparing actual with estimated fetal lung volumes by 3D ultrasound using a rotational multiplanar technique (VOCAL). Previously, the accuracy of estimating volumes by 3D ultrasound using a similar technique had only been evaluated in vitro, using phantoms33.
Considering the postmortem measurement by water displacement as the gold standard, we could estimate both the error and the accuracy of the 3D ultrasound measurement. However, in situations without such a gold standard (for example, comparing fetal lung volume estimated by 3D ultrasound and MRI), other statistical tests should be used, such as the correlation coefficient and Bland–Altman graphics30, 34–36.
The basic requirement for the clinical relevance of any biological measurement is to allow comparison of the normal situation to the pathological state. Nomograms could be useful to achieve this goal. Our previous study compared fetal lung volume estimated by 3D ultrasound in cases with CDH to a nomogram using the rotational technique28. However, in order to be clinically useful, it remains important to assess the accuracy of the measurement method in both the normal and the pathological situation, as demonstrated in this present study. It is also important that the method be reproducible, with minor intra- and interoperator variability, not only in normal fetuses under normal conditions, but also in those at risk for pulmonary hypoplasia and reduced optimal imaging29.
In our study, the accuracy of assessing fetal lung volume by 3D ultrasound was evaluated in fetuses with CDH and a gestational age of over 23 weeks, due to their referral to our unit after second-trimester sonographic diagnosis and a normal karyotype result. In our opinion, as prenatal diagnosis is generally made between 20 and 22 weeks in many of these cases, they will be referred to a specialized center at a gestational age similar to those observed here. Previous studies reported that fetal lung measurements become less precise with increasing gestational age26, 28.
In conclusion, fetal lung volume can be estimated accurately by 3D ultrasound using the rotation technique both in normal fetuses and in cases with CDH. Further studies are necessary to evaluate the clinical use of this method in the prediction of neonatal outcome or postnatal pulmonary hypoplasia.