Anthropometric craniofacial pattern profiles in Down syndrome

American Journal of M e d i c a l Genetics 47:748-752 (1993)

Anthropometric Craniofacial Pattern Profiles in Down Syndrome J.E. Allanson, P. O’Hara, L.G. Farkas, and R.C. Nair Children’s Hospital of Eastern Ontario, Ottawa (J.EA.); Departments of Pediatrics (J.E.A.) and Epidemiology and Community Medicine (R.C.N.), University of Ottawa; Department of Psychiatry, Carleton University (P.O.); and Hospital for Sick Children, Toronto (L.G.F.)

A series of 21 anthropometric craniofacial measurements was performed on 199 individuals with Down syndrome (DS), age 6 months to 61 years. These were compared to age and sex-matched normal standards, and Z score pattern profiles were constructed. These profiles confirmed brachycephaly and reduced ear length. With increasing age, maxillary growth was reduced in comparison to mandibular growth. Clinically, this was manifested by a change in facial shape from the characteristic round face of infancy to an oval shape in later life. Stepwise forward discriminant function analysis identified a subset of three variables (earlength, maxillary arc, and upper facial depth) which could accurately classify greater than 99%of the individuals in the combined sample of affected and unaffected individuals. Of the subjects with DS, 96.8% were classified correctly. These findings demonstrate the usefulness of anthropometric craniofacial pattern profiles in defining abnormal facial dimensions in particular syndromes and documenting the changes that occur with age. The technique should facilitate syndrome recognition, identification of carriers, and comparisons between syndromes. 8 ims wiley-~iss,IUC.

KEY WORDS: trisomy 21, Down syndrome, pattern profile analysis, craniofacial measurements LNTRODUCTION Syndrome diagnosis usually involves clinical observation of unusual morphological findings and abnormal proportions. Diagnosis of a specific syndrome enables Received for publication May 29,1993; revision received June 4, 1993. Address reprint requests to Dr. Judith Allanson, Division of Genetics,Children’sHospital of Eastern Ontario,401 Smyth Road, Ottawa, ON, K1H 8L1, Canada.

0 1993 Wiley-Liss, Inc.

the physician to provide the patient and family with discussion of treatment options, possible preventive measures, prognosis, and counseling regarding pathogenesis and recurrence risk. Inter-observer disagreement in syndrome identification is common and diagnostic accuracy would be enhanced by objective quantitative criteria and analytic methodology where possible. Since 1984,more than 30 anthropometric studies of various syndromes have been published [Meaney and Farrer, 19861,but only a few provide detailed evaluation of the craniofacies. Anthropometry has much to offer the clinical geneticist because it is simple and non-invasive, with minimal equipment costs, particularly in comparison to alternative objective techniques such as cephalometry and photogrammetry. The necessary measurement skills are readily learned with adequate training and practice. Measurement reliability (intraobserver error) and repeatability (inter-observer error) are quantifiable and controllable. Databases for a wide variety of normal craniofacial dimensions are well established in the Caucasian population [Farkas, 19811, although few racial and ethnic norms have been established. Several excellent reviews of the methodology and application of anthropometry have been published [Ward, 1989;Ward and Jamison, 19911. Anthropometry can be greatly enhanced by the production of pattern profiles after converting individual measurements into Z scores. These profiles are similar to those developed for hand radiographs by Poznanski et al. [19721 and by Garn et al. [19841for cephalometric roentgenograms. Any dimension characterised by a low or high Z score is immediately obvious, of potential diagnostic, or developmental significance, and serves to identify craniofacial areas that are most deviant from average. Pattern profiles allow comparison between parent and child, or between an individual and a group with known diagnosis. Finally, pattern profiles can be applied across age groups, and superimposition of profiles obtained at various ages may demonstrate persistence of or change in the pattern. The real success of this technique is attributed to a powerful set of multivariate statistical techniques such as cluster analysis, discriminant function analysis, and factor analysis. The correlation coefficient, rz,computed from paired Z scores of two pattern profiles, provides a numerical indication of the

Down Syndrome Pattern Profiles

similarity of any two patterns. The pattern variability index, uz,is a standard deviation of sets of 2 scored measurements expressed relative to norms for age and gender [Garn et al., 19851. It expresses the degree of dysmorphogenesis as a single number. The more highly patterned an individual, the greater the deviation from the reference population, and the greater the uz. With these considerations in mind we have used anthropometry to identify distinctive facial characteristics in a sample of individuals with trisomy 21 and to document changes that occur with increasing age. Obviously, the phenotype in Down syndrome (DS) is not difficult to recognise; rather, this condition was chosen because of the availability of large numbers of patients for study, with a diagnosis defined by karyotyping. The immediate goal of this research was to demonstrate that pattern profiles can be easily produced in a clinical setting, and can be employed as an aid to diagnosis. The long-term objective is to establish a series of syndromespecific pattern profiles or “signatures” to allow earlier diagnosis with more complete ascertainment, especially in rare syndromes, thereby improving the ability to provide accurate genetic counseling.

MATERIALS AND METHODS One hundred ninety-nine Caucasian individuals with DS, age 6 months to 61 years, were chosen as a defined syndrome population, with a proven diagnosis, to explore the development and application of anthropometric craniofacial pattern profiles. These individuals were drawn from clinic populations at Children’s Hospital of Eastern Ontario, Ottawa; The Hospital for Sick Children, Toronto; Hotel Dieu Hospital, Kingston; and Rideau Regional Centre, Smiths Falls. The study was approved by the Research Ethics Committee of Children’s Hospital of Eastern Ontario. A series of anthropometric measurements was obtained on each subject (Table I) following the methodology published by Farkas [1981], using sliding and spreading calipers and a tape measure. In a cooperative individual, the complete evaluation took 15 to 30 minutes, These dimensions were chosen to represent craniofacial widths, lengths, depths, and circumferences plus details of ear, eye, nose, and mouth morphology (Fig. 1). For each dimension age- and sex-matched normal standards were available. The population norms were derived from measurements of the head and face in 1538 6 to 18-year-oldNorth American and Western European Caucasian children and young adults [Farkas, 19811.In children under 6 years of age, unpublished norms of Hreczko and Farkas were used. Measurements were taken by two of the authors (LGF and JEA) after technique and consistency had been validated during a training period. The raw data were compared t o normal standards and converted to Z scores to control for age and sex differences. Pattern profiles were compiled for each age and sex. Because of small numbers a t certain ages, additional profiles were produced for chosen groups of subjects, for example 2 to 5,6 to 10, and 11 to 15. Subsequent statistical analysis consisted of correlation coefficient and pattern variability index formulation and stepwise

749

TABLE I. Anthropometric Measurements Employed in This Study Head width Skull base width Minimum frontal width Upper facial width Lower facial width Head length Upper facial depth Midfacial depth Lower facial depth Nasal protrusion Total facial height Upper facial height Nasal width Mouth width Inner canthal distance Outer canthal distance Ear width Ear length Maxillary arc

pra-pa sa-sba t-sn-t

Mandibular arc

t-gn-t

Head circumference

OFC

eu-eu t-t mf-mf ZY-ZY

go-go g-op t-n t-sn t-gn

sn-prn n-gn n-sn al-a1 ch-ch en-en ex-ex

Eurion to eurion Tragion to tragion Frontotemporale to frontotemporale Zygion to zygion Gonion to gonion Glabella to opisthocranion “ragion to nasion “ragion to subnasale Tragion to gnathion Subnasale to pronasale Nasion to gnathion Nasion to subnasale Alare to alare Cheilion to cheilion Endocanthion to endocanthion Exocanthion to exocanthion Preaurale to postaurale Superaurale to subaurale Tragion to subnasale to tragion Tragion to gnathion to tragion Maximum circumference in horizontal plane at level of glabella and opisthocranion

multivariate discriminant function analysis [SPSS,Nie et al., 19751, comparing the standardised data from the affected and control samples from 6 to 23 years of age.

RESULTS Figure 2 compares the pattern profiles in males (fine line) and females (thick line) at age 3 years. All profiles display Z scores on the vertical axis and craniofacial dimensions on the horizontal axis. Visually, these two profiles are fairly similar. The correlation coefficient is 0.88, providing a simple numerical confirmation of resemblance. The profiles have striking negative nasal protrusion (sn-prn). Ear length (sa-sba) is reduced, as expected from previous studies [Aase et al., 19731. The male pattern variability index is 5.2, whereas the female index is 2.8. This difference probably reflects one variable: nasal protrusion. There is a “peak and valley” appearance to three dimensions which is seen consistently at all ages. It represents modest increases in nasal width (al-all and inner canthal distance (en-en), with reduced mouth width (ch-ch). Between age 2 and 5 years (Fig. 3) nasal protrusion remains significantly negative. There is a more striking deficit in ear length than in the previous profile. The actual ear length Z score has been stated rather than plotted, because it falls well outside the profile scale. Head length (g-op) is shorter than width (eu-eu) confirming brachycephaly. The pattern profile between 6 and 10 years of age (Fig. 4) shows some new aspects. Lower facial width, depth, and height, plus mandibular arc (solid arrows) are mov-

Allanson et al.

750

Fig. 1. Dimensions chosen to represent craniofacial widths, lengths, depths, heights, and circumferences, and to evaluate the eyes, ears, nose, and mouth.

ing closer to zero, suggesting that mandibular growth velocity is increasing disproportionately, relative to maxillary dimensions (dotted arrows), and approaching normal. This continues to be evident from 11 to 15 years of age (Fig. 5). In fact, upper facial width (bizygomatic

diameter] in this study group changes little with time. Clinically, reduced maxillary growth in comparison to mandibular growth is evident as an alteration in facial shape, from round to oval, with increasing age. Figure 6 compares males in age groups 2 to 5 (fine

2

2

0

-4 -

Z

-6 -

S C

.I:i,

.

, , , , , ,

,y,

-2

0

R E

, , , , , , , , ,

-4

-6

-12

e t f z g g t t t s n n a u . 1 y o . . . . n . . l . I . o n s g . g s . i ; g p n n p n n a c e u t y o I I n

c e h n . . e e h n

e x . . x

p r a s p a

s a . n b a

t . s ; . t

t 0 . F ' . t

-8

e u

t ,

.

t

e u

~ y . . f z t y

z

~ . g o

. o p

g

. n

t

.

t

t

n

:

p I

n

Fig. 2. A craniofacialpattern profile which compares three year old males (fineline) and females(thick line) with Down syndrome. The two patterns are strikingly similar (correlation coefficient 0.88).

s

n

n

g

s

.

a

c

e

e

p

s

t

. n . . l h n x r a . . F

s

n

n

a

c I

.

.

e h

.

e n

a

. x

i p a

i i C b . . a c t

Fig. 3. A craniofacial pattern profile which compares males (fine line) and females (thick line) age 2 to 5 years.

t

0

Down Syndrome Pattern Profiles

751

I ,

1

0 Z .

-1

S

C 0 R E

-2

-3

-4

e

t

f

z

u

.

t

y

o

.

. e

t

. f

. z

. g

o n p

t

y

o

u

g

g

t

.

r

.

t

.

s g n n

s n n a c e e p n . . l h n x r . g s . . . . a p n n a c e e . i I I h n x p n a

s t t 0 a . . F .

i : C b . . a t t

-1

i

z

g

t

y

o

0 p

n

s n

g n

g ~. . . . a . n a c c e . s i : C I h n h p b . . a a t t

. n

p r

n

Fig. 4. A craniofacial pattern profile which compares males (fine line) and females (thick line) age 6 to 10 years.

7

. .

t

e u

Fig. 5. A craniofacial pattern profile which compares males (fine line) and females (thick line) age 11 to 16 years.

7

I

1

-

-8 e

t

f

z

g

u .

. t

t

y

g

t

t

o

.

t

o p

n

s n

g n

. .

e

i

z

g

u

t

y

o

.

.

.

s n . p n

n . g n

n . s n

a l

.

c h

.

a c l h

e n

.

e n

e x

.

e x

p r a . p a

s a . s b a

t t O . . F s n i C . . t t

4

e

~

"

'

~

t

f

z

g

u

.

t

y

o

.

.

f

i t

. z y

. g o

o p

e u

g

"

"

"

t

t

t

"

"

"

"

'

~

s n n a c e e p . . . n . . l h n x r n s g g s . . . . a n n p n n a c e e 1 I h n x p n a

s t t 0 a . . F . s s

n

b a

. . t t

i

C

Fig. 6. A compound craniofacial pattern profile which compares males of 2-5 (fine line), 6-10 (thick line), and 11-15 (dottedline) years of age.

Fig. 7. A compound pattern profile which compares adult males from 20-29 (fine line),30-39 (thick line), and 40-49 (dottedline) years of age.

line), 6to 10 (thick line), and 11 to 15 (dotted line). There is marked pattern similarity in facial widths and regional dimensions including ear length. The main area of discordance is facial depth. A similar compound pattern profile comparing adult males in 3 decade ranges (Fig. 7) demonstrates reasonable concordance, with stabilisation of facial depths. Mandibular dimensions (solid arrows) remain closer to normal than their maxillary counterparts (dotted arrows). In these adult profiles, ear lengths are no longer dramatically subnormal. This may be because adult norms are only available up to age 18. Subtle increases in ear length after age 18 in DS may reduce the discrepancy between the study group and the control group. We are attempting to validate this conclusion by measuring adult ears in Ottawa. Forward stepwise discriminant function analysis was

performed in the two groups, subjects and norms, in an attempt to define the combination of variables which best distinguished between them. Raw normative data were available from 6 to 23 years of age. Three variables were defined in the resulting discriminant function: ear length, maxillary arc, and upper facial depth. This function correctly classified 96.8% of the individuals with DS, and 99.9%of the normal subjects, with an overall correct classification rate of 99.78%.The standardised canonical coefficients and the F ratios associated with them, which provide some indication of the relative contribution of the three variables to the function, are indicated in Table 11. Thus, ear length, maxillary arc, and upper facial depth contribute most to separation of the two groups. All three predictors show univariate F ratios that are significant at the