In fetal lung development there is a critical interval, the canalicular phase, between 16 and 28 weeks' gestation. Preterm prelabor rupture of membranes (PPROM) during this interval can delay lung development, thus causing pulmonary hypoplasia1. Pulmonary hypoplasia poses a serious threat due to its high mortality and morbidity rates. It can occur as severe respiratory failure leading to early neonatal death, as respiratory insufficiency with pulmonary hemorrhage, bronchopulmonary dysplasia or subacute lung disease, or as mild and even transient respiratory disease2. Perinatal mortality is approximately 70% (55–100%) in most series3.
Once mid-trimester PPROM has occurred, assessment of the probability of pulmonary hypoplasia is important both for clinical decision-making and for counseling of parents. In a recent meta-analysis4 we assessed the predictive capacity for presence of hypoplasia of clinical parameters: gestational age at which PPROM occurred, latency period until delivery and degree of oligohydramnios. Gestational age at PPROM was a significantly better predictor than was latency period or degree of oligohydramnios. Accuracy in the prediction of the lethal variant of pulmonary hypoplasia was similar.
Biometric parameters assessed by ultrasound (two-dimensional (2D), three-dimensional (3D) or Doppler) or magnetic resonance imaging (MRI) have also been proposed in the prediction of pulmonary hypoplasia following PPROM. To our knowledge, the predictive capacity of these parameters for the presence of hypoplasia after mid-trimester PPROM specifically has not been assessed systematically. We therefore performed this meta-analysis to assess the capacity of biometric parameters assessed with ultrasound or MRI to predict pulmonary hypoplasia following mid-trimester PPROM.
We searched the literature for studies that reported on neonatal outcome after mid-trimester PPROM, using combinations of the following search terms: pregnancy, oligohydramnios, fetal membranes—premature rupture, diagnostic imaging, fetal diseases, respiratory system, fetal mortality, fetal death, infant mortality, pulmonary hypoplasia, lung hypoplasia, lung diseases and respiratory system. We performed an electronic search of MEDLINE (inception to 05-3-2011) and EMBASE (inception to 05-03-2011) and checked reference lists of known reviews and primary articles to identify cited articles not captured by electronic searches. Language restrictions were not applied.
The selection process was performed by one of the authors (A.S.P.v.T.). To be selected for inclusion, a study had to report on the outcome of pregnancies complicated by PPROM between 14 and 27 completed weeks of gestational age, in which any ultrasound or MRI parameter was used with the goal of predicting pulmonary hypoplasia. The diagnosis of pulmonary hypoplasia could be based either on clinical and radiological findings or on findings at autopsy. For the purposes of analysis we distinguished two types of hypoplasia: lethal hypoplasia and any form of hypoplasia. Lethal hypoplasia was defined as hypoplasia resulting in the death of the fetus or neonate due to hypoplasia. Fetuses with lung hypoplasia proven on autopsy after early pregnancy termination were also included in the lethal group. Any form of hypoplasia was defined as the sum of all cases of hypoplasia, both lethal and non-lethal. We chose not to include cases of oligohydramnios caused by conditions other than mid-trimester PPROM, since these are other entities, with their own specific pathophysiology which might have influenced the outcome of our review. Moreover, the biometric indices studied in this review might have been influenced by these conditions, for example measurements in fetuses with polycystic kidneys or obstructive uropathy might be influenced by subsequent abdominal enlargement5.
Studies had to report on any ultrasound or MRI parameter that was used with the goal of predicting pulmonary hypoplasia. The following characteristics of each study were noted: (1) sampling (consecutive or other); (2) data collection (prospective or retrospective); (3) study design (cohort study or case–control study); (4) blinding (present or absent); (5) verification bias (present or absent) and (6) selection bias (present or absent)6.
For each published study, its characteristics were scored by two of the authors (A.S.P.v.T. and J.v.d.H.), who each constructed independently a 2 × 2 table cross-classifying one or more of the imaging parameters with the presence of pulmonary hypoplasia. In case of disagreement, the judgement of a third author (B.W.J.M.) was decisive. It appeared that all but two studies found were aimed at diagnosing lethal rather than any form of pulmonary hypoplasia. Therefore, we limited the outcome in this review to lethal hypoplasia.
To visualize the data, for each model we combined sensitivity and specificity in the form of a receiver–operating characteristics (ROC) curve. A bivariate meta-regression model was used to calculate summary estimates of sensitivity and specificity for predictive values and to fit summary ROC (sROC) curves. The bivariate method has been described extensively elsewhere7–10. Briefly, rather than using a single outcome measure per study, such as the diagnostic odds ratio, the bivariate model preserves the two-dimensional nature of diagnostic data in a single model. This model incorporates the correlation that may exist between sensitivity and specificity within studies due to possible differences in threshold between studies. The bivariate model uses a random effects approach for both sensitivity and specificity, allowing for heterogeneity beyond chance due to clinical or methodological differences between studies. In addition, the model acknowledges the difference in precision by which sensitivity and specificity have been measured in each study. This means that studies with a larger number of pregnancies resulting in lethal pulmonary hypoplasia receive more weight in the calculation of the pooled estimate of sensitivity, while studies with more patients without hypoplasia are more influential in the pooling of specificity.
When individual study sensitivity–specificity points were grouped close to an imaginary underlying ROC curve (i.e. studies with high sensitivity had relatively low specificity and vice versa), an sROC-curve was drawn using parameter estimates from the bivariate model11.
Differences in the capacity of all parameters to predict lethal pulmonary hypoplasia were tested for statistical significance by entering the tests as covariates in the bivariate regression model. P < 0.05 was taken to indicate a significant difference of one parameter as compared with the other.
Figure 1 summarizes the identification and selection process of the thirteen published studies included in this meta-analysis12–24. All studies reported on ultrasound parameters, and one also evaluated (two-dimensional) MRI parameters.
Study characteristics of the 13 included studies are listed in Table S1 online. In two of the studies, sampling of data was consecutive. Data collection was prospective in all studies, and all studies were designed as cohort studies. Five studies were adequately blinded. Selection bias was present in eight studies. Selection biases most frequently seen were limitation of studies to pregnancies with established oligohydramnios and inclusion of pregnancies in which PPROM occurred over 28 weeks. In one study of 45 pregnancies with oligohydramnios, in eight cases the oligohydramnios was not caused by PPROM20. Since in the majority of cases in this study the cause was PPROM, we included it in our review. One study excluded pregnancies in which measurements were unsuccessful or in which the results of reference tests were not obtained15.
Verification bias was not present in any study. In six of the 13 studies, the diagnosis of lethal pulmonary hypoplasia was not always based on autopsy data; sometimes clinical and radiological data had to be used14. Table S2 gives characteristics of the biometric parameters that were used. The most commonly used ultrasound parameters were chest circumference (seven studies), chest circumference/abdominal circumference ratio (six studies) and chest circumference/femur length ratio (three studies). The MRI parameters used in the only study incorporating MRI were chest circumference and ratio of chest area minus cardiac area divided by cardiac area; volumes were not measured. The sensitivities and specificities for chest circumference, chest circumference/abdominal circumference ratio and chest circumference/femur length ratio, as well as some other parameters that were used less frequently in the diagnosis of lethal hypoplasia, are summarized in Tables S3 and S4, the former giving parameters derived from 2D ultrasound measurements, and the latter parameters derived from 3D ultrasound, 2D MRI or Doppler measurements.
A plot of sensitivity–specificity points for chest circumference, chest circumference/abdominal circumference ratio and chest circumference/femur length ratio for lethal pulmonary hypoplasia is shown in Figure 2. The study of Laudy et al.19 was the only one to report an optimal sensitivity for chest circumference, but this was at the expense of low specificity; the other six studies combined a high specificity with a sensitivity varying between 50% and 80%. The study of d'Alton et al.13 was the only one to demonstrate perfect sensitivity and specificity for chest circumference/abdominal circumference ratio; all other studies had either suboptimal sensitivity or suboptimal specificity.
Figure 3 shows the performance of chest circumference/abdominal circumference ratio, as in Figure 2, as compared with the sROC of our previous meta-analysis4 which assessed the predictive capacity of gestational age at PPROM, latency time between PPROM and delivery and amount of amniotic fluid.
The study of van Eyck et al.24 used Doppler measurements during (induced) periods of fetal breathing. All these studies were relatively small, and none indicated that any of the evaluated tests had a good accuracy. The study of Ohlsson et al.21 reported almost perfect accuracy for the chest circumference/femur length ratio, with a sensitivity of 100% and a specificity of 97%, but again the sample size in this study was rather low, as there were only 35 pregnancies in the cohort. Neither the amount nor the timing of measurements performed throughout the latency period was uniform, as can be seen from Table S2.
In this meta-analysis, we included 13 studies that reported on the prediction of lethal pulmonary hypoplasia. The estimates for sensitivity and specificity showed the capacity of imaging techniques to predict hypoplasia to be very limited.
Our review identified several weaknesses in the literature. First, the methodological quality of the studies was limited. Many suffered from verification bias, as a result of the test being used in the management of the patients or because the observers were not blinded. All studies were single-center, which is worrisome, since second-trimester PPROM is a rare condition (0.7% of pregnancies), making it unlikely that a single-center study would reach sufficient power. Indeed, the sample size of each of the studies was rather small, especially with respect to the number of cases of hypoplasia.
Most of the studies that we identified focussed on lethal pulmonary hypoplasia. This is important, as accurate prediction of lethal pulmonary hypoplasia before 24 weeks gives parents the chance to opt for termination of pregnancy. Moreover, this information could be of clinical relevance in women for whom discussions on intervention have to be made at gestational ages around that of viability of the child. In women relatively early on in pregnancy who are at high risk of fetal lethal pulmonary hypoplasia, the decision for an unnecessary Cesarean section could be delayed.
As we included studies performed between 1988 and 2006, there will have been differences between them in terms of treatments. Differences in antenatal management (tocolysis, corticosteroids, antibiotics) and advances in neonatal intensive care could have resulted in lower mortality figures in the more recent studies, with a subsequent lower incidence of lethal pulmonary hypoplasia.
The reference test used in the various articles showed strong variation, which is also a source of heterogeneity. In seven of the 13 studies, all cases with lethal pulmonary hypoplasia were proven by autopsy, albeit with different pathological standards. The technique to diagnose pulmonary hypoplasia in autopsy varies widely. The three criteria used to define pulmonary hypoplasia are lung weight/body weight ratio, radial alveolar count and amount of DNA detected in lung tissue (lung DNA (in mg)/body weight (in g) ratio). Each of these three criteria has its own disadvantage. (Wet) lung weight/body weight ratio is decreased in pulmonary hypoplasia; however, tissue edema could increase the ratio, confusing the diagnosis. Radial alveolar counts are difficult to interpret in the preterm lung before the development of alveoli. The numbers in a fixed expanded lung differ from those in a fixed collapsed lung. The lung DNA/body weight ratio is confounded by the presence of increased pulmonary interstitial inflammatory cells25. In six studies, some of the affected cases were identified by (predefined) clinical and radiological characteristics. However, a clinical diagnosis of pulmonary hypoplasia is difficult to establish since congenital pneumonia or infant respiratory distress syndrome sometimes occur simultaneously and have overlapping symptoms. The diagnostic meta-analysis we performed allows for control for heterogeneity in sensitivity and specificity, since both report on the same underlying test. When the cut-off for abnormality increases, sensitivity increases and specificity decreases, while a decrease in the cut-off for abnormality has the opposite effect. The summary ROC curves we estimated are a means of addressing this heterogeneity.
We previously assessed the predictive capacity of gestational age at PPROM, latency time between PPROM and delivery and the amount of amniotic fluid4. Gestational age at PPROM performed significantly better than did the two other parameters in the prediction of pulmonary hypoplasia. Future studies should focus on whether ultrasound parameters or other imaging techniques can further improve on the prediction of pulmonary hypoplasia that is possible using age at PPROM. In view of the current evidence, we feel that there is no indication to perform such tests in a clinical setting.
SUPPORTING INFORMATION ON THE INTERNET
The following supporting information may be found in the online version of this article:
Tables S1 and S2 Characteristics of studies (Table S1) and of biometric parameters used (Table S2) in studies comparing at least one biometric parameter with occurrence of lethal pulmonary hypoplasia.
Tables S3 and S4 Sensitivity and specificity of two-dimensional ultrasound (Table S3) and three-dimensional ultrasound, two-dimensional magnetic resonance imaging and Doppler (Table S4) parameters in detection of lethal pulmonary hypoplasia.