Indications for MRI in fetal isolated mild ventriculomegaly… ‘And then, there were none’

Fetal isolated mild ventriculomegaly (IMV) is a relatively common condition, complicating 1% of all pregnancies. It is a benign finding in most cases, but it may be associated with a variety of different etiologies, ranging from infections to chromosomal anomalies, and from cerebral clastic conditions to migration abnormalities1. The prognosis for cases with IMV is good, with an abnormal neurological outcome reported in roughly 10% (95% CI, 6.1–18.1%)1. Whether IMV should be considered an indication for magnetic resonance imaging (MRI) is a controversial issue, involving significant and delicate economic and medicolegal aspects. Most studies published over the last two decades seem to demonstrate that, when MRI is performed in cases of apparent IMV, it uncovers additional, clinically relevant, central nervous system (CNS) abnormalities in about 10% (range, 0–40%) of them2-5. However, the diagnostic yield of fetal brain MRI has been challenged by several fetal neurology experts6, 7. Some light was cast on this controversy by the recent meta-analysis of Di Mascio et al.8, which confirmed that, overall, clinically relevant information additional to that found at ultrasound was found on MRI (CRIMR) in roughly 10% of IMV cases. The finding of particular interest, however, is the fact that stratifying the analysis according to the type of ultrasound assessment revealed a considerable difference: the rate of CRIMR was 5.0% (95% CI, 3.0–7.0%) when dedicated neurosonography had been undertaken, compared with 16.8% (95% CI, 8.3–27.6%) in cases which had undergone a standard ultrasound assessment of the fetal brain in the axial plane. We believe that these results represent a first attempt at separating apples from oranges9. Until now, there has been a significant overestimation of the clinical usefulness of MRI in fetal brain anomalies, due to a worrying underperformance of what has been labelled ‘Level-II’ ultrasound but which was apparently inferior even to screening level in some studies10. We feel, however, that the meta-analysis of Di Mascio et al.8 did not complete the task, but instead has opened up a can of worms, the appetites of some of which look set for them to devour the conclusions of several papers, both old and new, on the topic. In this Opinion, we review our recent experience of IMV and cases published in the literature (including, but not limited to, virtually all of the articles included in the meta-analysis of Di Mascio et al.8). Our objectives were: (1) to ascertain whether there is any association between certain variables (specialty of first author, atrial width, sonographic route of examination) and the usefulness of prenatal MRI as an adjunctive diagnostic imaging modality in IMV (i.e. any association with CRIMR); and (2) in the subgroup of fetuses in which MRI was considered to have added clinically relevant information that had been missed on ultrasound (i.e. the subgroup in which CRIMR had been reported), to analyze in detail all clinical and diagnostic data, in order to evaluate possible inaccuracies or inconsistencies in the findings or reporting. For this, we focused specifically on the clinical relevance of CRIMR findings, adequacy of terminology used in the description of the CRIMR and visibility of findings at neurosonography. Details of our review methodology are summarized in Appendix S1. We retrieved a total of 1498 patients from 20 studies of fetuses with IMV (atrial width, 10.0–15.0 mm) which had undergone both ultrasound and MRI examination, to which we added a cohort of 29 patients seen at our institution over the last 5 years, giving a total of 1527 cases. There was CRIMR in 118 (7.7%) of these cases (Table S1). Of the three variables tested for association with the usefulness of MRI in IMV, all were found to be significantly related to CRIMR, though with rather different likelihood ratios (Table 1). Of note, CRIMR was found in 1.6% vs 9.2% of cases, depending on whether a transvaginal or a transabdominal ultrasound approach was used, respectively (P < 0.001; likelihood ratio, 26.4). The second part of the analysis, a detailed and critical evaluation of CRIMR in the pooled group of cases in which CRIMR was reported, included 118 cases from 16 studies (in five studies, including our current cohort, there was no CRIMR). The clinical relevance of CRIMR was graded according to whether the additional anomalies detected by MRI would be expected to be associated with: no or minimal risk of neurodevelopmental delay (e.g. connatal cyst, germinal matrix hemorrhage); possible but limited risk of major neurodevelopmental delay (e.g. isolated corpus callosal agenesis, intraventricular hemorrhage); or high risk or certainty of major neurodevelopmental delay (e.g. brain atrophy, schizencephaly). The additional findings on MRI had no impact on the neurodevelopmental prognosis in 13.6% (n = 16) of the cases, possible or limited impact in almost half (47%; n = 56) and a certain negative impact in 39% (n = 46) of them. The diagnostic terminology for MRI diagnoses was rated as being adequate or inadequate according to whether the term employed is found in at least two well-known reference textbooks11, 12. A list of the descriptors considered inadequate, with reasons, is given in Table 2. The terminology was considered inadequate or incomplete in nearly a quarter (28/118; 23.7%) of the cases, a regrettable finding considering that these studies were performed by fetal medicine and neuroradiology professionals. The most striking finding in this analysis, however, was with regard to the sonographic detectability of the CRIMR. We categorized if and how easily the MRI diagnoses should have been detected on neurosonography according to whether, on neurosonography, the lesion is: not diagnosable or diagnosable in a minority of cases (e.g. polymicrogyria, focal gyral abnormalities); usually diagnosable, with some exceptions or only for some subtypes (e.g. germinal matrix hemorrhage, heterotopia – subependymal heterotopia can be diagnosed on neurosonography, while band heterotopia cannot); or always diagnosable (e.g. corpus callosal agenesis, holoprosencephaly, Dandy–Walker malformation). We found that in two-thirds (67%; n = 79) of the cases, the lesions should have been easily detectable on ultrasound, even screening ultrasound (Figure 1), in around a quarter (24%; n = 28), the lesion would be expected to have been detected only in some subcategories, while in just 9% (n = 11) of the cases the lesion would have been undiagnosable by any means on ultrasound. The rationale for this analysis was to investigate possible factors responsible for the striking discrepancy (0–100%, Table S1) in the reported usefulness of MRI in the assessment of IMV2-5, 10, 13-27. Our analysis showed that having a radiologist/neuroradiologist rather than a fetal medicine specialist or obstetrician as first author, the atrial width being 12.1–15.0 mm rather than 10.0–12.0 mm and the ultrasound examination being by the transabdominal rather than the transvaginal route were all associated with a significant increase in the rate of CRIMR (Table 1). Notably, transabdominal ultrasound was associated with a 26-fold increased risk of CRIMR. In our opinion, this result alone should be sufficient to make mandatory a transvaginal approach for neurosonography. This has, in fact, been recommended since 200728, but apparently not all fetal medicine units, at least not most of the ones involved in the studies analyzed here, have complied so far, over a decade after publication of the ISUOG guidelines. The fact that breech presentation makes the transvaginal approach impossible is not a valid excuse. In fact, in most cases of breech presentation, an external cephalic version is feasible and safe. This should be attempted for any breech fetus in the second trimester and, in selected cases, up to 32 gestational weeks, in order to allow a transvaginal approach. With respect to our detailed examination of the 118 cases in which CRIMR was reported, while the findings of all three parts of this subanalysis were of interest, of particular concern is the fact that only 9% of the lesions that were detected only by MRI were in fact undiagnosable on ultrasound. This woeful statistic can only be explained by a significant lack of training and experience among the professionals involved, and is epitomized by the recently published MERIDIAN trial and previous publications from the same multicenter group4, 5, 10, 14. The abnormally high incidence of CRIMR (Table S1) in that work was apparently due to the suboptimal diagnostic performance of fetal medicine units involved in the study, which missed six of nine cases of complete agenesis of the corpus callosum10. This lesion is diagnosable indirectly, and rather easily, in the screening transventricular view, by the absence of the cavum septi pellucidi and the colpocephalic aspect of the lateral ventricles (Figure 1a2). It is important to underscore that this critical reappraisal of the current evidence on the role of MRI in IMV is based, in part, on a somewhat subjective interpretation of certain variables, i.e. visibility of anomalies missed at ultrasound on transvaginal neurosonography or the evaluation of the nomenclature employed in the description of anomalies. Even though both of these assessments were made following current published evidence and reference textbooks11, 12, it cannot be excluded that, in some cases, we might have inadvertently under- or overestimated the prognostic impact of some lesions. In addition to reinforcing the findings of Di Mascio et al.8, we wish to underscore inaccuracies and inconsistencies regarding the apparent need for MRI in cases of fetal IMV that were not highlighted by their meta-analysis8. Our analysis of the current evidence strongly supports the concept that the gold standard for assessment of pregnancies in which fetal IMV is suspected should be: (1) transvaginal neurosonography, (2) with (three-dimensional) multiplanar imaging, (3) carried out in referral centers for CNS malformations (which does not necessarily apply to all fetal medicine units). If our proposed approach is employed by adequately trained fetal medicine professionals, then the incidence of CRIMR would be only 1.6%. This points to an urgent need for improved training in neurosonography and fetal neurology in fetal medicine units, particularly those reporting an incidence of CRIMR > 1.6%. Table 3 proposes an outline for such training. Whether the possibility of missing 1.6% of serious cerebral lesions is sufficient to warrant MRI in all cases of IMV depends, in our opinion, on local financial resources and, above all, on the diagnostic level of the regional fetal medicine vs neuroradiology units. It is probable, in these times of spending reviews and global economic crisis, that, in most countries, this percentage will not warrant a fetal MRI examination, even on medicolegal grounds, particularly when there is the option of using stored three-dimensional volume datasets to obtain a second opinion and verify appropriate conduct of the examination. Thus, having carried out a careful, detailed and critical analysis and finding that evidence to support an indication for MRI in fetuses with IMV is lacking, we find ourselves turning to Agatha Christie's famous words: ‘And then there were none’. Yet, as for most areas of medicine, the situation is not black and white. It is important to emphasize that there will always be exceptions, namely cases in which MRI examination is of fundamental importance because of the technical limitations of ultrasound. Failed or impossible (twins) external cephalic version or advanced gestational age (> 34–35 weeks), especially if the mother is obese, represent two situations in which fetal MRI is truly indicated. Regardless of the recommended local health policy, it is important to stress that all fetuses with a suspicion of IMV should be referred for transvaginal neurosonography, as per international guidelines28. An indication to perform MRI should come only from the experts performing this examination, in agreement with local health policies, and not directly from the screening setting. Looking to the future, both diagnostic ultrasound, including neurosonography, and MRI are now exploring resolution capabilities never seen before. MRI is being applied to functional and metabolic aspects of the fetal brain, for example, in fetuses with congenital heart disease, in which significant alterations in brain metabolism and oxygen delivery and consumption have been demonstrated29, 30. It is not impossible that, in the not-too-distant future, fetal ultrasound may turn from diagnostic to therapeutic. In fact, experiments are underway using high- and low-intensity focused ultrasound, for treatment of brain lesions and non-invasive neuromodulation, respectively31, 32. These approaches are promising in the adult; their application in the fetus with a brain malformation or other acquired pathological condition may soon leave the realms of science fiction to become a reality. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.