Traditional cephalometric examinations comprised of linear and angular measurements are not sufficient in determining subtle changes in the craniofacial complex in children prenatally exposed to anticonvulsant medication. Geometric morphometric analysis of lateral cephalograms can therefore be a useful adjunct to determine how the development of the craniofacial structures such as the maxilla, posterior and overall cranial base affect the shape of the face.
In this study we used generalized Procrustes superimposition, Principal Component Analysis (PCA) and Discriminant Function Analysis (DFA). These allowed us to superimpose shapes of the bones studied, to explore the major features of shape variation, and to distinguish the separation between the study and the control groups. Hence, these geometric morphometrics tools are used for description and classification.
Our results do not show a sagittal shortening or lack of prominence of the maxilla. However, both orbitale and rhinion are posteriorly located, which could give an impression of a retrusive midface. This partially agrees with Holmes et al16, who found no shortening in the length of the maxilla (Ptm-A), palate (ANS-PNS), nasal bone (N – Na) or the anterior (Se – N; S – Se) or posterior (S – Ba) cranial base using traditional cephalometrics. Their analysis showed a significant decrease in only one measurement, the S-N-Na angle, which represents the angle between the anterior cranial base and the nasal bone and is in line with the graphical results of this study showing a retrusive position of Rh (Fig. 8a). This shortening would account for the upturned appearance of the nose called “anteverted nostrils.”
In contrast, Orup14 demonstrated a statistically significantly smaller maxilla in both height and length as measured by traditional cephalometric analyses that included the height from Nasion to ANS, the palatal length (ANS-PNS), and the maxillary projection compared to the anterior cranial base. We evaluated shape, not size, but our results are consistent with his in many aspects. Figure 6a shows ANS being placed more superiorly in the maxilla relative to the control group, which would account for Orup’s finding of shorter height from Nasion to ANS. ANS is also positioned slightly more posteriorly while PNS is somewhat more anterior, agreeing with the decreased ANS-PNS length. The decreased maxillary projection relative to the cranial base can also be seen in Fig. 8a, where orbitale and rhinion are placed more posteriorly relative to the anterior cranial base, namely glabella. Although A point is not more posterior, the overall impression would be a decreased midface.
Identifying the source of anomalies is important for understanding development. Nie25 points out the pivotal position the cranial base occupies between the neurocranium and the face, with the anterior and posterior portions having distinct embryological origins. The anterior cranial base comes from neural crest, and the posterior from the paraxial mesoderm. Kjaer26 details the role the anterior cranial base plays in development. Deficiency in the anterior cranial base growth is often accompanied by midface deficiency, and the length and inclination of the cranial base are controlling factors of jaw position. She concluded that cranial base development is probably more controlled by genetic than environmental factors.
Tepedino et al27 carried out morphometric analysis of sella turcica in patients with different sagittal malocclusions (Class I, II, III) and pointed out that anomalies of the anterior wall of the sella appear to be associated with alterations of the frontonasal area and defects of body axis, while anomalies of posterior wall seem to be related to brain alterations.
The pathogenesis of anticonvulsant anomalies can similarly be traced to genetic changes. In one study28, a mother with polymorphisms of microsomal epoxide hydrolase (EPHX1 113H and 139R) and who took phenytoin during pregnancy had a higher likelihood of having an infant with craniofacial abnormalities. In a second study29, children exposed during pregnancy to five different AEDs, including PHT, had a four-fold increase in their risk for having a child with a malformation when the mother was homozygous for the MTHFR (methylene-tetrahydrofolate reductase) polymorphism 677TT in comparison to mothers homozygous for another MTHFR polymorphism 677CC. Other suggested underlying defects include a deficiency of epoxide hydrolase30, the formation of free radicals31, and hypoxia-reperfusion damage resulting from the inhibition of potassium channels32.
Studies such as the current one will not by themselves lead to the discovery of the pathogenesis of these effects; however, they may help to differentiate one drug from another to establish patterns of effect that can help in determining cause. Indeed, GM has become increasingly used in the fields of medicine and in particular in studies of facial anomalies. Katsube33 explains how understanding the phenotype completely and accurately is necessary for understanding the pathogenesis of a disease. For example, Martinez-Abadias34 found that mice with a missense mutation of S252W of FGFR2 (which results in Apert syndrome) were born with a more severe palate dysmorphology compared to the P253R mutation.
GM can also be used to evaluate surgical outcomes in patients with clefts. Hoffmannova et al35 compared the palate morphology of children with clefts of the lip and palate who had lip and palate repairs in the first week of life with those whose repairs were carried out at 6 months. GM allowed the detection of subtle differences in shape and curvature that were not possible using conventional methods33. Likewise, GM has been used to study the pattern of sutural closure and the resulting bony growth in patients with craniosynostosis36–38 – as well as the shape variability of the cranial base, the maxilla, and the mandible of patients with beta thalassaemia based on principal components analysis.39 Moreover, applying partial least square analysis on subjects with achondroplasia allowed to identify a significant axis of covariation between shape and age. These results enabled to predict during the growth the increase of morphological differences on five anatomical regions located on the skull and the mandible.40
A recent study completed by Holmes et al. concluded that anticonvulsant-exposed children with the characteristic dysmorphic facial features, particularly a shorter nose, were more likely to have deficits in cognitive function.16 The IQ scores were subdivided into full scale IQ, verbal IQ and performance IQ. The 115 anticonvulsant-exposed children, considered together, had IQ scores that were about 6 points lower (range 5. 6–6.8) than those in the matched comparison children. There was a correlation between the presence of “the anticonvulsant face” and deficits in IQ. The full-scale IQ of all AED-exposed children who had a short nose had an IQ that was 6 points lower than the unexposed children with a short nose. Our study used shape, not linear measurements, to identify craniofacial anomalies. It is possible that more nuanced relationships between dysmorphology and function can be elucidated using geometric morphometrics.
Although the three drugs included in this study are only rarely prescribed in the US currently, there is evidence that they are frequently used in other countries41 due to lower cost and ready availability. Neurologists around the world need to become aware of this risk for women of child-bearing age and cease prescribing these particular medications. Furthermore, concerns over exposing children are compounded by evidence that research into growth effects of these drugs has shown significant evidence that valproate results in shorter stature after as little as one year.42,43 It seems that the teratogenic effects of this class of drugs are not limited to differences in facial morphology; more work needs to be done to fully characterize their effects.
One limitation of this study is that the effect of age and sex were not analyzed. It is unclear due to the small sample sizes whether the dysmorphic features increase with age or not. Orup et al14 evaluated 28 children who had been exposed to phenytoin monotherapy or polytherapy using traditional cephalometric analysis. Eleven individuals who exhibited one or more measurements that were significant for previously published effects of anticonvulsant medications were reevaluated after a mean of 7 years. He found that several measurements describing retrusion of the maxilla and mandible worsened over time. However, these children were already noted to be significantly affected, and the limitations of traditional cephalometrics blunt the significance of these results.
There have been no previous studies using GM to investigate the craniofacial features of children with prenatal exposure but our results show its value. Analyzing the shape of dysmorphic structures can give some insight into where the developmental damage first occurs by visualizing the graphical differences in the population. Future studies can explore craniofacial effects in three dimensions if such imaging becomes available, as well as determine whether specific drugs have different effects.