Craniofacial morphology in down syndrome: a systematic review and meta-analysis

The aim of this study was to evaluate the craniofacial cephalometric characteristics of individuals with Down syndrome (DS), comparing them with healthy subjects. An electronic search was made in Pubmed, Embase, Lilacs, Scopus, Medline and Web of Science without imposing limitations on publication date or language. Studies were selecting following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement. The PECO acronym was applied as follows: P (population), individuals with DS; E, (exposition) diagnosis of DS; C (comparison), individuals without DS; O (outcomes) craniofacial characteristics based on cephalometric measurements. Independent reviewers performed data extraction and assessed the methodological quality of the articles using the Newcastle–Ottawa Quality-Assessment-scale. Seven case–control studies were included in meta-analysis. Given the variability of the cephalometric measurements used, only those that had been reported in at least three or more works could be included. Anterior cranial base length (SN), posterior cranial base length (SBa), total cranial base length (BaN), effective length of the maxilla (CoA), sagittal relationship between subspinale and supramentale (ANB), anterior facial height (NMe), and posterior facial height (SGo) values were significantly lower in the DS population than among control subjects. No significant differences were found in sagittal position of subspinale relative to cranial base (SNA) and sagittal position of supramentale relative to cranial base (SNB). Summarizing, individuals with DS present a shorter and flatter cranial base than the general population, an upper jaw of reduced sagittal dimension, as well as a tendency toward prognatic profile, with the medium third of the face flattened and a reduced anterior and posterior facial heights.

Methodological quality assessment. The methodological quality of the articles selected was assessed using the Newcastle-Ottawa Quality Assessment scale (NOS) 18 by two examiners working together (A.V and A.L) and a third examiner (C.S) working independently. The degree of agreement between the examiners was analyzed with Cohen's kappa coefficient (κ).
The NOS gives a maximum score of nine; articles scoring seven or above are considered of good quality. Assessment is divided into three areas: selection, comparability and exposure. The selection criterion consisted of four items: adequate definition of cases, representativity of the cases, selection of controls, and definition of controls. Comparability consists of the comparability of cases and controls derived from study design, considering factors such as age or ethnicity. The last part is exposure, which is divided into three items: evaluation of exposure, equality of methods for cases and controls, and dropout rate.
There is a maximum score of one for each item, with the exception of comparability, which can award a score of two. Data extraction. The following data were extracted from each of the articles by two researchers (A.V. and C.S): authors, year of publication and for both cases and controls sample size, ethnicity, gender, and cephalometric measurements were registered. Given the wide variety of cephalometric measurements taken in the articles, for meta-analysis it was decided to include only those measurements that were repeated in at least three articles, these being: Anterior cranial base legth (SN, mm), posterior cranial base length (SBa, mm), total cranial base length (BaN, mm), posterior cranial base inclination (SNBa, °), sagittal position of subspinale relative to cranial base (SNA, °), effective length of the maxilla (CoA, mm), sagittal position of supramentale relative to cranial base (SNB, °), relative position of maxilla and mandible to each other (ANB, °), anterior facial height (NMe, mm), posterior facial height (SGo, mm) ( Fig. 2; Table 1). The mean value and standard deviation were recorded for each measurement.
The reliability of the codification process between examiners was analyzed by means of Cohen's kappa coefficient for qualitative variables and the intra-class correlation coefficient for quantitative variables.
Effect size index. For each outcome the effect size index was the mean difference (MD) between the DS group and the control group. A negative value for the MD indicated that a mean in the DS group was lower than that of the control group, and vice versa. www.nature.com/scientificreports/ Statistical analysis. Separate meta-analyses were carried out for each outcome. As different articles varied in the cephalometric measurements taken, the results could only be subjected to meta-analysis when a mean datum was reported in at least three studies. As some evidence of heterogeneity among the individual studies was expected, a random-effects model was assumed. This model implies to weight each effect size by its inverse variance, this being the inverse of the sum of two variances: within-study variance and between-studies variance. Between-studies variance was estimated using the DerSimonian and Laird estimator. For each outcome, a pooled effect size (in terms of mean difference) was calculated, as well as the 95% confidence interval 19 . Heterogeneity of the effect sizes was checked with the Q statistic and I 2 index. A statistically significant result for the Q statistic (p < 0.05) indicated the presence of heterogeneity. In addition, heterogeneity I 2 indices around 25%, 50%, and 75% were considered to reflect low, moderate, and large heterogeneity, respectively 20 . Finally, to assess the potential influence of the two age ranges found in the studies ( www.nature.com/scientificreports/ evidence of significant differences between the average effect sizes of the age ranges. In addition, a double forest plot was constructed for each outcome in order to illustrate the potential differences among the effect sizes as a function of the age range. All statistical analyses were carried out using Comprehensive Meta-analysis 3.3 software 21 .    www.nature.com/scientificreports/ the full texts were read and a further 69 excluded for failure to meet the review criteria, leaving a total of 9 works (Fig. 1). The Cohen kappa coefficient of inter-examiner reliability was κ = 0.890. The manual search in journals did not identify any additional works.

Results
Methodological quality assessment. The nine articles selected were assessed with the NOS 18 , obtaining a Cohen kappa coefficient of κ = 0.832 between evaluators.
Of the nine articles, seven underwent quality assessment; four obtained maximum scores of nine [22][23][24][25] , two a score of eight 26,27 , and one of seven 28 . The two articles considered of low quality 9,29 obtained scores of six (Table 2) and they were excluded from the meta-analysis.
Reliability of data extraction process. The degree of inter-examiner agreement was high. For all quantitative variables, except for the item "number of controls, " an intra-class correlation coefficient of one was obtained. Initially the intra-class correlation coefficient for "number of controls" was 0.77 and after finding the mistake in the number of controls between the examiners we corrected it and obtained a κ = 1. Qualitative variables all obtained values of κ = 1. Table 3 summarizes details of the seven case-control studies included in meta-analysis. Five articles were published in English [22][23][24][25]27 , one in French 26 and one in Spanish 28 . The earliest publication date was 1985 26 , while the most recent was published in 2014 25 . Studies were conducted in Canada 23 , France 26 , Colombia 28 , Spain 22,24 , Brazil 27 and Saudi Arabia 25 . The smallest sample size was 14 individuals 28 , while the maximum was 60 25 . The ages of the subjects investigated ranged from 5 27 to 22 years 25 . In five articles the individuals were Caucasian [22][23][24]26,27 , in one Columbian 28 , and in the other Arab 25 . All the works coincided in age range and ethnicity between case and control groups. Regarding gender, five works included individuals of both sexes [22][23][24][25][26] ; one included only males 27 , while another one did not report the sex of the subjects 28 .

Study characteristics.
Cephalometric characteristic of the cranial base were assessed by means of the measurements SN 22 23,25,27 . Anteroposterior relation between maxilla and mandible was determined by means of the ANB angle 23,25,27,28 ; anterior face height by NMe 23,27,28 ; and posterior face height by means of SGo 23,27,28 . Table 4, a total of five studies were found to report outcomes. With the exception of SNBa and SNB, lower mean outcomes were found among DS subjects than controls. Moreover, the DS group exhibited significantly lower mean values than the control group for outcomes SN, SBa, NBa, CoA, ANB, NMe, and SGo. Two outcomes (SNBa and SNB) obtained higher mean values for DS subjects than controls but only the difference in SNBa reached statistical significance.

Meta-analyses. As shown in
Due to the low number of studies identified, the interpretation of Q statistics must be treated with caution, due to problems of statistical power. The I 2 index offers a more adequate means of assessing heterogeneity among effect sizes. As shown in Table 3, moderate-to-high heterogeneity was found for all outcomes, with the exception of NBa and ANB.
The studies were grouped in two categories depending on participants' age ranges (6-11 and 5-22 years old), the former category being more homogeneous than the latter. To assess the potential influence of age ranges on effect sizes, subgroup analyses were carried out for each outcome. Forest plots (Figs. 3,4,5,6,7,8,9,10,11,12) show the results. In addition, the Q B statistic for testing statistical significance in differences between the two average effect sizes is shown in Table 5. Statistically significant differences were found between the average effect sizes of the two age range categories for outcomes SNBa (p = 0.039); SNB (p = 0.004) and NMe (p = 0.035). Regarding Table 2. Methodological quality for all 9 case control studies identified by the search strategy, assessed using the Newcastle-Ottawa-Scale. a (1) Adequate case definition, (2) Representativeness of the cases, (3) Selection of controls, (4) Definition of controls, (5) Comparability of Cases and Controls on the basis of the design or analysis (age y ethnicity), (6) Ascertainment of exposure, (7) Same method of ascertainment for cases and controls, (8) Non-response rate.  www.nature.com/scientificreports/ the outcome SNBa, both age ranges exhibited a larger mean effect size for DS cases than for controls, but the 5-22 year age range exhibited a larger difference (MD + = 8.854) than the 6-11 year age range (MD + = 6.500). For the outcome SNB, the significant difference found between average effect sizes of the two age ranges was due to a larger mean effect in the DS group than the controls for the category 5-22 years (MD + = 2.191), whereas the inverse was found for the category 6-11 years (MD + = −2.600). Lastly, for the outcome NMe, lower mean values were found in the DS group than the controls for both age ranges, but the average mean difference was more pronounced in the category 5-22 years (MD + = −15.510) than 6-11 years (MD + = −10.342).   www.nature.com/scientificreports/

Discussion
The aim of this meta-analysis was to determine whether the craniofacial cephalometric characteristics of individuals with DS differ from those of the general population. It was decided not to limit the literature search by language or date of publication in order not to miss any article that might provide information relevant to the work's hypothesis; as a result, articles in three languages were included.
When evaluating the comparability of the cases and controls by means of the NOS, the age and ethnicity of subjects were considered, as cephalometric measurements may be influenced by these variables. Although in the seven works included for meta-analysis, age ranges and ethnicity were the same in case groups and control   www.nature.com/scientificreports/ groups, there was some variability between the articles, especially regarding age, so in order to maintain homogeneity, the works were grouped according to age ranges into two groups: 6-11 years 27,28 and 5-22 years [22][23][24][25][26] .
The seven studies included in meta-analysis used different cephalometric measurements, and so it was decided to analyze only those measurements that were repeated in at least three articles. Nevertheless, the maximum number of works to coincide in any single measurement was five. For this reason, the results of the present metaanalysis should be interpreted with caution.
Cranial base length was evaluated by means of the measurements SN, BaS and BaN. It was found that anterior cranial base length (SN) was significantly shorter in DS subjects than among control subjects. Some authors have suggested that the lesser development of the anterior cranial base is related with smaller brain size 27 . But according to Enlow the growth of the cranial base is considered rather autonomous, as compared to the cranial vault, because of that, we think that the smaller brain size of DS individuals should not be considered the main cause or the only cause of the short anterior cranial base of these patients. Rather both facts could be the effect   www.nature.com/scientificreports/ of a different primary cause 30 . One study evaluated cranial base growth in subjects aged 8-18 years, observing that it grew in the same proportion among DS individuals as control subjects, showing that the deficit in size was produced before the age of 8 years 22 . Some authors have suggested that the deficit is of prenatal origin 9 , while others state that is develops during the first 4 years of life 31 . As for posterior cranial base (BaS), it was also found to be significantly reduced in size in DS groups. A histological study based on autopsies of individuals at different ages determined that growth in this region comes to an end at the age of 18 years in healthy individuals 32 . Alió et al. observed that, in individuals with DS, the rate of growth decreases gradually up to the age of 15 years, while in control subjects it continues to grow, so that spheno-occipital synchondrosis growth stops earlier in individuals with DS 22 .
Obviously, the reduced anterior and posterior cranial bases in cases of DS mean that the total cranial base length (BaN) is also significantly smaller than in the general population.   www.nature.com/scientificreports/ At the same time, the cranial base angle (SNBa) was significantly larger in DS than control subjects. It has been suggested that delay in intra-sphenoidal synchondrosis fusion in postnatal life is essential to the cranial base's flattening mechanism. Radiographic data obtained from individuals with DS show a delay in fusion, so that synchondrosis remains without fusing between the ages of 1-7 years, while in the general population it is fully obliterated by the end of the first year of life 33 . The present findings show that the difference in measurement between DS cases and control subjects was more acute in the 5-22 year age range. But the difference in effect size between the 6-11 year and the 5-22 year age group was 2.35°, so even though this difference was statistically significant, it was not clinically significant, as the standard deviation for this angle is around 4°3 4 . In addition, according to Bjork, the cranial base angle is 130.8° ± 4.2° at age 12 and 131.6° ± 4.5° at age 20 years, showing an insignificant change from the former age to the latter 34  The anteroposterior position of the maxilla was evaluated by means of the SNA angle, which, although smaller in cases of DS, did not show significant differences in comparison with control subjects. The SNA angle relates the maxilla to the cranial base, which, as stated above, is significantly shorter in individuals with DS. Anteroposterior and vertical cephalometric measurements for the maxilla and mandible based on an anomalous cranial base can lead to erroneous cephalometric interpretation. For this reason, Jesuino and Valladares suggest that in these cases cephalometric measurements should take the Frankfort plane as reference 27 . The reduced anterior cranial base in DS can make the SNA similar to that of control subjects due to a geometric effect, even though the maxilla is smaller 23 , because the N point will be located at a more posterior location 24 .
Maxillary length (CoA) was significantly smaller in DS than control groups, corroborating the existence of maxillary hypoplasia in the sagittal plane. Alió et al. showed that the maxilla grows in the same proportion between the ages of 8 and 18 years in DS cases as in healthy subjects 24 . Klingel et al. observed that at the age of 6-9 months the maxilla is already smaller in all three dimensions in DS infants 36 .
As for the mandible, no statistically significant differences were observed between DS and control subjects in SNB angle, although values were higher in DS subjects. This measurement is also related to the cranial base; in this case shortening produces larger SNB angles 23 . At the same time, the present work found that the cranial base angle was significantly larger in DS groups, which would lead to less symphysis projection, while in control subjects the angle was more acute, contributing to a more anterior mandibular position 35 . Again, it is clear that to assess cases of DS, it is necessary to use a reference plane other than the cranial base to determine the anteroposterior position of the mandible correctly, and to reach clear conclusions regarding the size of the mandible. Meanwhile, in the 6-11 year age range, SNB angle values were lower in DS subjects than controls, while in the 5-22 year range it was larger in DS groups. This could be a result of the fact that the mandible has not undergone full growth in the younger age group. Mandibular growth ends around the age of 17 in females and 19 in males 37 .
Relating maxillary and mandibular anteroposterior position by means of the ANB angle, it was found that this was significantly smaller among DS subjects, indicating a greater tendency among these individuals to present skeletal Class III malocclusions.
Both anterior (NMe) and posterior face height (SGo) were significantly smaller among individuals with DS. In these cases, maxillary hypoplasia in the vertical plane conditions developmental deficiency in the facial middle third 24 . For NMe, the difference between DS groups and control subjects was more pronounced in the 5-22 than the 6-11 age range, probably due to the fact that the latter group had still not undergone complete growth.
The tendency toward brachyfacial pattern in individuals with DS was notable, in spite of the muscular hypotonia that such cases usually present, the tendency to keep the mouth half open in repose, and frequent infection of the upper airway. This is perhaps due to mandibular autorotation in maximum intercuspation resulting from maxillary vertical hypoplasia. The meta-analysis by Oliveira et al. found a higher prevalence of anterior open bite among individuals with DS, although they reported a low level of evidence for an association between DS and anterior open bite 38 .
Although the present work suffered several limitations, it was possible to reject the hypothesis that there would be no differences in craniofacial characteristics between individuals with DS and a healthy population, as clear differences were found. Nevertheless, in order to reach clear conclusions it would be necessary to unify the cephalometric measurements taken in different studies, and to use different reference planes for SN in order to determine vertical and anteroposterior positions of the maxilla and mandible. The results must be interpreted with great caution due to the scarce number of studies that reported each outcome measure. In addition, the small number of studies included in meta-analysis did not make it possible to examine the potential effects of publication bias, as at least 10 studies are needed to apply the typical methods for assessing publication bias.

Conclusions
In spite of the limitations of the present meta-analysis, it may be concluded that individuals with DS present a cranial base of reduced length, which is more flattened than in healthy subjects. Moreover, the maxilla is of reduced size in the sagittal plane and there is a tendency toward skeletal Class III malocclusion. DS is also characterized by reduced facial height in both posterior and anterior regions.