Emaptm }<*nx& of OrAoiaaia 9 (19S7) 10*-108 1987 Euxopoo Orthodontic Society

A comparison of vertical and horizontal

cephalometric variables with regard to heritability
Anders Lundstrom and John S. McWilliam
Stockholm. Sweden

SUMMARY The aim of the present study was to compare horizontal and vertical cephalometric
distances with regard to heritability. The material used consisted of 56 pairs of twins of the same
sex ranging from 13-20 years of age. Heritability (h2) and cultural inheritance (c2) were
determined according to the path analysis method and quotients between genetic and
environmental standard deviations were compared. No systematic difference could be found
between horizontal and vertical measurements, but the highest heritability values were obtained
among four vertical variables.

Introduction Material and method

Twin studies may be used to compare different The study was based on lateral skull radiographs
variables with regard to their dependence on from a series of twins earlier described by Lund-

Downloaded from by guest on December 8, 2015

heredity versus environment As regards strom (1954,1967). This series included 50 mono-
cephalometric distances this was done by Hunter zygotic (MZ) and 50 dizygotic pairs (DZ),
(1965) when he compared vertical and horizontal however, pairs younger than 13 years of age or
dimensions among 37 monozygotic and 35 dizy- pairs subjected to orthodontic treatment were
gotic twin pairs. The significance of this compari- excluded The materialfinallyemployed is shown
son was based on a conclusion made by Osborne in Table 1.
and de George (1959) in their book, The Genetic The variables studied are presented in Fig. 1
Basis of Morphological Variation. They stated and consist of six horizontal and six vertical
that, 'In general,... measurements taken parallel cephalometric measurements. They were con-
with the longitudinal axis of the body show the structed from ten landmarks registered directly
strongest genetic component of variability'. on the lateral skull radiographs using digitizer
Hunter found that 11 of the 12 vertical and 8 of (CM-1, Central Institute for Industrial Research,
the 14 horizontal distances showed significantly Oslo). The digitizer was connected to a micro-
greater variation between dizygotic, than between computer programmed to calibrate incoming
monozygotic co-twins. The average quotient data for minor systematic instrument errors
between the variances of dizygotic and mono- caused by scale and x/y axis discrepancies
zygotic twins was also higher for vertical than (McWilliam, 1980). The x and y-coordinates of
horizontal distances (3.5 and 26, respec-
tively). Lundstrom (1984) tried to check Hunter's
findings for facial height (N-Gn/V) and depth Table 1 The age and sex distributions of the
(S-Pg/H) on limited material of adult twins (23 mono- and dizygotic twin pairs included in this
mono- and 19 dizygotic pairs of the same sex) study.
without being able to prove the hypothesis. A N Mean age Range
larger material was considered necessary in order
to check Hunter'sfindings.The aim of the present Monozygotic pairs 28 15.2 12.7-20.0
study was to include more variables in an Male 7 14.7 13.0-17.6
investigation of a larger sample of twins and in so Female 21 15.4 117-20.0
Dizygotic pairs 27 14.7 13J-19.2
doing further test the hypothesis that, vertical Male 13 14.3 13.2-17.7
cephalometric variables show stronger genetic Fondle 14 15.1 13.3-19.2
dependence than horizontal variables.
HERITABILITY OF CEPHALOMETRIC VARIABLES 105

HORIZONTAL VARIABLES tion for single observations of each variable

estimated (Table 2).
The distribution of each variable was studied.
All variables were found to be approximately
(S-N)-4° normally distributed, although one twin, a 17-
year-old boy, was found to have an extreme Gass
m basal bone relationship with an A-B distance
lying 3.5 x s.d. from the group mean. Since this
could be regarded as an outlier which could
possibly mask the results of any comparison
between the MZ and DZ groups it was decided to
omit this pair from further analysis.

Table 2 Method errors in mm calcu-

lated according to the formula: So) =
^/(Ld2/2n) where d is the difference
VERTICAL VARIABLES between the 1st and 2nd observations.

Variable s, Variable s,
(S-N)-4° S-N/H (U N-Sp/V 0.7
S-A/H 0.6 Sp-Gn/V 0.6
S-1JH 0.6 N-Gn/V 0.6
S-I^H OS N-VV 0.5
S-B/H 0.9 N-yv 0.6
S-Pg/H 0.7 S-Go/V 1.0

Each variable was tested for correlation with

age and sex. In the presence of such correlations
residuals corrected for these factors were em-
Figure 1 Definition of the horizontal and vertical variables
employed in this study.
ployed during further analysis. A subsequent
analysis of variance was used to determine the
"within' pair and 'between' pair variances for both
the MZ and DZ groups. These were employed to
all landmarks were first registered within a calculate the intra-class correlation coefficients
coordinate system defined by fourfiducialpoints (r,Mz and rIDZ) which could then be used to
placed near the corners of each radiograph estimate the genetic heritability (h2). The latter
(McWilliam, 1982). The x- and .y-axes were thus was done according to the Path Analysis model
mathematically adjusted so as to be parallel to (Rao et al^ 1976; Iselius, 1983):
the sides of the four point rectangle with origo
placed at its centre of gravity. Using a special
computer program all subsequent landmark co- The heritability coefficient indicates the propor-
ordinates were transformed to a new coordinate tion of the total variance which can be attributed
system based on sella as origo and with the x-axis to inherited factors. The greater degree of simi-
orientated parallel to the line SN-4° (Bjerin, larity found between first degree relatives com-
1957). From these coordinates the twelve vari- pared with non-related individuals may, apart
ables were determined, corrected for enlarge- from similarities in genetic configuration, be due
ment, and transferred to a main frame for to so-called cultural inheritance. Thisrefersto
statistical analysis. the similar environment shared by members of a
Analysis of the method error was based on family compared with the varying environments
duplicate radiographs of 23 twins. These were acting on norf-related individuals. The Path
registered independently and the standard devia- Analysis allows a separation of these factors and
106 ANDERS LUNDSTR6M AND JOHN S. McWILLIAM

an estimation of cultural heritability. In twin Table 4 Infra-class correlations (rj for MZ and DZ
studies c2 is obtained from the formula: twins
c = 2 r,DZ—TMZ Variable r Variable r^ r,DZ
IDZ
In addition to the Path Analysis, Dahlberg's S-N/H 0.72 0.46 N-Sp/V 0.62 0.54
analysis (see Lundstrom, 1948) was employed S-A/H 0.70 0.38 Sp-Gn/V 0.89 0.46
to calculate a quotient between heredity and S-yH 0.71 0.45 N-Gn/V 0.92 0.49
environment This approach utilizes the intra- S-VH 0.74 0.36 N-yv 0.93 0.48
pair variances (s.d.2^ and S.&DZ) f° r MZ and S-B/H 0.77 0.41 N-I,/V 0.88 0.50
S-Pg/H 0.79 0.44 S-Go/V 0.77 0.64
DZ twins:
Mean r, 0.74 0.42 0.84 0.52
c A
''•"• 'environment

The calculation shown above does not take into

account the effect of the measurement error tion within the MZ and DZ groups differed. The
which may increase the variances employed. influence of age and sex on each variable was thus
Consequently it would be interesting to eliminate determined using a multiple stepwise regression
this factor, (Si)2 for S-d.^ and for s.d.J3Z respect- analysis. Sex was found to be significant for all
ively. In this way a more reliable estimate of the but three variables (S-A/H, S-Is/H, S-Ii/H).
heredity/environment quotient may be obtained. Five vertical variables were also significantly
affected by age. For all variables showing cor-
relations with sex and/or age, subsequent analyses
Results were performed onresidualsindependent of both

Downloaded from by guest on December 8, 2015

The group means for all 12 variables studied factors. Motivation for including an age adjust-
showed no significant difference between the MZ ment even where this factor was not found to be
and DZ twins. Table 3 presents the associated of significance, was that such a relationship may
standard deviations calculated as the square root be expected within the interval 13-20 years in
of the total source variances obtained from the spite of its slightly positive but apparently non-
analysis of variance. significant influence in this material.
For eight variables the DZ standard deviation Since the Path Analysis is based on the intra-
was found to be greater than the corresponding class correlation coefficients, these were calcu-
MZ value. In two, similar values were obtained lated first and are presented in Table 4. As was
(S-Pg/H, N-GN/V) and in two the MZ distribu- expected the MZ pairs were in general, more
tion was greater than its DZ counterpart (S-IS/H, highly correlated than the DZ pairs. The hori-
Sp-GN/V). None of these differences were stat- zontal and vertical variables for the MZ pairs
istically significant gave mean coefficients of 0.74 and 0.84 respec-
Since age and sex differences may be expected tively. Corresponding values for the DZ pairs
to influence the material, the original measure- were 0.42 and 0.52, respectively. Four of the
ments are not particularly suitable for estimation vertical variables (Sp-Gn/V, N-Gn/V, N-I/V,
of heritability. This aspect was especially im- N-Ii/V) showed particularly high correlation
portant considering the fact that the sex distribu- coefficients (0.88-0.93 and 0.46-0.50 respectively)
suggesting strong dependence on inheritance
(genetic and/or cultural).
Table 3 Standard deviations in mm among MZ and On average h2 was 0.6 for both horizontal and
DZ twins for horizontal and vertical variables. vertical variables while c2 was lower, 0.1 for
horizontal and 0.2 for vertical distances (Table 5).
Variable »*« «>.„ Variable sd-uz "i-DZ
It should be observed, however, that the highest
S-N/H 2.3 3.1 N-Sp/V 12 2.6 h2 value (0.9), was obtained for four vertical
S-A/H 22 3.9 Sp-Gn/V 5.1 43 variables. The correspondence between means
S-I^H 4.1 35 N-Gn/V 5.6 5.6 was due to the fact that two of the vertical
S-I,/H 4.0 4.8 N-yv 3.7 3.8 variables, N-Sp/V and S-Go/V had the lowest h2
S-B/H 5.1 53 N-yv 3.8 4.0
S-Pg/H 55 55 S-Go/V 4.1 4.5 values. As regards the latter variable, this is
perhaps not so astonishing as the jaw angle is
HERITABILITY OF CEPHALOMETRIC VARIABLES 107

Table 5 Coefficients of heritability (h2) and cultural the lowest h2 values returned s.d.tcJs.±car
inheritance (c2). quotients of 1.0 and 1.1. This suggested a close
agreement between the Path Analysis and Dahl-
Variable h2 Variable berg's analysis for these variables. Comparison of
values corrected for the method error indicated
S-N/H 0.52 0.20 N-Sp/V 0.16 0.46
S-A/H 0.64 0.06 Sp-Gn/V 0.86 0.03 that
t\\at the
tka latter
tnai me lai
lot had little influence on the quotients
S-yH 0.52 0.19 N-Gn/V 0.86 0.06 studied.
S-VH 0.76 -0.02 N-yv 050 0.03
S-B/H 0.72 0.05 N-yv 0.76 0.12
S-Pg/H 0.70 0.09 S-Go/V 0.26 0.51 Discussion
Mean 0.64 0.10 0.63 0.20 A basic problem in twin research is the reliability
of the twin diagnoses. Zygosity determinations
were based on anthropological similarity deter-
minations including careful tooth anatomy
subject to functional influence from the masti- comparisons. In addition, 40 pairs underwent
catory muscles. The high c 2 value (0.5) for the serological and blood group determinations. The
same variable can, therefore, be explained to value of tooth morphology as a basis for zygosity
some extent as being due to intra-family simi- diagnosis, has been illustrated in an earlier
larity in the choice of a hard or soft diet The publication (Lundstrom, 1963) with a 94 per cent
corresponding values concerning the upper facial concordance between diagnoses based on com-
height is more difficult to explain. It may prob- prehensive comparisons including somatic, sero-
ably be due to random variation in intra-class logical and blood group data on one hand, and
correlation coefficients. Thus, it was observed tooth anatomy alone on the other. Discordance
that the two variables with the lowest h 2 and was found in 0.8 per cent where tooth anatomy
highest c 2 values were based on unexpectedly indicated MZ and 4.8 per cent where the same
high correlation coefficients of 0.54 and 0.64 indicated DZ. For 38 other pairs, diagnoses
found among the dizygotic pairs (Table 4). based on general similarity and tooth anatomy
Table 6 illustrates the relative importance of were found to coincide with serological and
heredity as calculated according to Dahlberg (see blood group determinations in 37 pairs (Lund-
Lundstrom, 1948). The same five variables which strom, 1967). The number of erroneous diagnoses
according to the Path Analysis, showed greatest in the present material should thus be small,
heritability also gave the highest s.d.fen/s.d.enT considering that a combination of general
quotients, 1.7-3.0 when using this test This somatic and tooth anatomy comparisons were
suggested two to three times more influence from employed in all pairs and that a considerable
heredity compared with environment The lowest number of pairs also underwent serological and
quotient was found for variable S-A/H (0.6). The blood group determinations.
two variables (N-Sp/V and S-Go/V) which gave For eight of the variables the DZ twins gave
higher standard deviations than the MZ twins.
Table 6 Quotients between heredity and environ-
One variable suggested a converse relationship.
ment (s.d.OBJs.d.Esy) with (+) and without (—) These differences were never significant, how-
correction for method errors (SJ. ever, and may well be of a random nature.
The material studied in this investigation was
Variable Correction Variable Correction relatively small. Consequently, it is possible that
the results include a certain degree of random
- + - + variation in the extent of influence exerted by
S-N/H 0.94 0.98 N-Sp/V 0.87 1.01
heredity on the different variables. This also
S-A/H 0.60 0.64 Sp-Gn/V 1.69 1.81 applies to the quotients which varied from 0.2-
s-yH 126 L31 N-Gn/V 134 2.53 0.9 for h2 and 0.6-3.0 for s-d^s-d.^,. A similar
s-yH 1.64 1.72 N-yv 2.62 3.05 argument may be used to explain the variability
S-B/H 1.33 1.43 N-yv 1.90 114 in c2. Most of the variables studied gave relatively
S-Pg/H 1.29 1J9 S-Go/V 0.94 1.09
low values for cultural inheritance. The high
Mean 1.18 1.25 1.72 1.94 value, (h2 = 0.5) found in connection with the
vertical development of the mandibular angle,
108 ANDERS LUNDSTR6M AND JOHN S. McWILLIAM

may be due to differences between families in the References

consistency of the diet which can have influenced Bjerin R 1957 A comparison between the Frankfort hori-
masticatory muscle function. The correspond- zontal and the sella turcica-naaon as reference planes in
ingly high c 3 value for upper facial height is caphalometric analysis. Acta Odontologica Scandinavka
difficult to explain since the total and lower facial 15: 1-12
heights gave low c2 values (0.0 and 0.1 respec- Hunter W S 1965 A study of the inheritance of craniofacial
tively). characteristics as seen in lateral cephalograms of 72 like-
Considering the s.d.ien/s.d.enT quotients, this sexed twins. Transactions of the European Orthodontic
Society 59-70
investigation would seem to give some support to
Hunter's findings which suggested that vertical Iselius L 1983 Genetic epidemiology of some major risk
variables were more influenced by heredity than factors for coronary heart disease. Thesis Karolinska Insti-
rutet 7-13
were horizontal ones. This is not confirmed,
however, by the mean h 2 values arising from the Lundstrom A 1948 Tooth size and occlusion in twins. Thesis
Path Analysis. The first model did not take into (Uppsala)
consideration, cultural inheritance, and the pos- Lundstrom A 1954 The importance of genetic and non-
sibile effects of family environment which may genetic factors in the facial skeleton studied in 100 pairs
have contributed to increase the s.d.ien/s.d.cn, of twins. Transactions of the European Orthodontic
Society 92-107
quotient Since c2 was somewhat higher for the
vertical than horizontal variables (0.2 and 0.1 Lundstrom A 1963 Tooth morphology as a basis for distin-
respectively) greater confidence should be given guishing mono- and dizygotic twins. American Journal of
to the h2 values particularly in the case of the two Human Genetics 15: 34-43
vertical variables giving high c2 values (N-Sp/V Lundstr6m A 1967 Genetic aspects of variation in tooth
and S-GofV). In conclusion, no general difference width based on asymmetry and twin studies. Hereditas 57:

Downloaded from by guest on December 8, 2015

403-410
between the horizontal and vertical variables
could be established. It should be noted, how- Lundstrom A 1984 Nature versus nurture in dento-facial
ever, that the highest h2 and s.d..eJs.d.enr variation. European Journal of Orthodontics 6: 77-91
quotients were found among four vertical vari- McWilliam J S 1980 Evaluation and calibration of X- Y-
ables (Sp-Gn/V, N-Gn/V, N-I./V and N-I/V). coordinatographs used in cephalometric analysis. Scan-
dinavian Journal of Dental Research 88: 496-504
Acknowledgement McWilliam J S 1982 Orientation of orthogonal coordinate
systems used for registration of cephalometric landmarks.
This study has been supported by the Swedish Scandinavian Journal of Dental Research 90: 145-150
Dental Society.
Osborne R H, de George F V 1959 Genetic basis of
morphological variation. Harvard University Press
Address for correspondence
Rao D C, Morton W E, Yee S 1976 Resolution of cultural and
Professor A Lundstrom biological inheritance by path analysis. American Journal
Odontologiska Klinikerna of Human Genetics 28: 228-242
Box 4064, S-14104 Huddinge
Sweden