ORIGINAL ARTICLE

Prediction of mandibular growth rotation: Assessment of the

Skieller, Bjrk, and Linde-Hansen method
Laurel R. Leslie, DDS, MS,a Thomas E. Southard, DDS, MS,b Karin A. Southard, DDS, MS,c
John S. Casko, DDS, MS, PhD,d Jane R. Jakobsen, MS,e Elizabeth A. Tolley, PhD,f
Stephen L. Hillis, PhD,g Chris Carolan, MS,g and Mark Logue, MS,g
Iowa City, Iowa, and Memphis, Tenn.
The purpose of this investigation was to assess the method proposed by Skieller, Bjrk, and Linde-Hansen
in 1984 to predict mandibular growth rotation. Our sample consisted of 40 randomly selected, untreated,
adolescent subjects representative of the patient population generally encountered in orthodontic practice.
The four independent variables identified in the Skieller study as having the highest predictive value
(mandibular inclination, intermolar angle, shape of the lower border of the mandible, and inclination of the
symphysis) were identified on initial lateral cephalograms. The proposed regression equations were applied
and predicted mandibular rotations obtained. Final lateral cephalograms made 6 years after the initial profile
radiographs were superimposed and actual mandibular rotation recorded. The observed and predicted
rotations were compared and regression analyses performed to determine the amount of variability in
observed values accounted for by the four variables individually and in combination. Only 5.6% of the
variability in mandibular growth rotation could be accounted for using the four variables individually. Only 9%
of the variability could be accounted for with a combination of the variables. In addition, we performed a
Monte Carlo analysis, which mirrored the Skieller analysis but used random numbers instead of actual
cephalometric data, to determine if the Skieller results may simply have capitalized on chance. Using the
same forward stepwise selection procedure with a rejection level of P > .1, we found after 5000 simulations
that a mean of 84% and a median of 94% of mandibular growth rotation variability could be accounted for
using meaningless data in the Skieller analysis. This result was comparable to the Skieller value of 86%. In
conclusion, information derived from pretreatment lateral cephalograms using the Skieller, Bjrk, and LindeHansen method does not permit clinically useful predictions to be made in a general population relative to
the direction of future mandibular growth rotation. (Am J Orthod Dentofacial Orthop 1998;114:659-67)

Magnitude and direction of mandibular

growth markedly impact orthodontic therapy in growing patients. Traditionally, it has been demonstrated
that the mandible grows in a downward and forward
direction via posterior growth and anterior displacement.1-8 However, it has also been shown to rotate over
the course of growth.9-12 This rotational aspect of
mandibular growth can greatly influence the therapeutic result of orthodontic treatment. In fact, a backward
rotation of the mandible is quite often regarded as
aIn private practice, Longmont, Colo; former orthodontic resident at The University of Iowa.
bAssociate Professor, The University of Iowa.
cAssociate Professor, The University of Iowa.
dProfessor and Head, Department of Orthodontics, The University of Iowa.
eDepartment of Preventive and Community Dentistry, The University of Iowa.
fDepartment of Preventative Medicine, The University of Tennessee.
gDepartment of Statistics and Actuarial Science, The University of Iowa.
Reprint requests to: John S. Casko, DDS, MS, PhD, Department of Orthodontics, College of Dentistry, The University of Iowa, Iowa City, IA 52242
Copyright 1998 by the American Association of Orthodontists.
0889-5406/98/$5.00 + 0 8/1/89302

being particularly difficult to treat.10-14 Close observation is often required throughout the course of therapy
to deal with effects of the rotational pattern.
The ability of an orthodontist to predict future
mandibular growth rotation would greatly aid in diagnosis and treatment planning. Better therapeutic decisions could be made regarding timing and length of
treatment, appliance selection, extraction pattern, and
possible need for surgery. Therapy could truly be tailored to the individual with the possibility of optimal
results in a shorter period of time.
A number of techniques have been proposed to predict growth based on cephalometric landmarks,6,15-20
and a few of these specifically address rotational
growth of the mandible. Skieller, Bjrk and LindeHansen16 proposed a method for predicting future
mandibular growth rotation from an initial prepubertal
lateral cephalogram. Of a group of 100 children who
had been enrolled in an implant growth study, 21 subjects were selected retrospectively on the basis of a pattern of extreme backward or forward growth rotation.
659

660

Leslie et al

Fig 1. Location of landmarks, lines, and independent

variables no. 2. (intermolar angle), no. 3 (shape of the
lower border of the mandible), and no. 4 (inclination of
the symphysis).

None of the subjects had received orthodontic treatment before or during the period of study. Metallic
implants had been placed in their mandibles, and jaw
growth was followed from childhood into early adulthood. Growth rotation was calculated on the basis of
the change in inclination of the implant line on radiographs. A 6-year observation period was selected,
with the initial lateral cephalogram made 3 years
before puberty and the final cephalogram made 3 years
after pubertal onset. Forty-four morphologic variables
were measured from each initial lateral cephalometric
radiograph and multivariate statistical methods were
used to identify which variables, alone or in combination, showed the highest predictive value for
mandibular growth rotation. The authors reported that a
combination of four variables accounted for 86% of the
variability observed. These included: mandibular inclination, intermolar angle, shape of the lower border of
the mandible, and inclination of the symphysis.
Regression equations were given based on these variables that could be applied to an initial prepubertal lateral cephalogram and were believed to predict future
mandibular growth rotation.
Limitations of this study are apparent. For instance,
the sample size was small relative to the large number
of measurements investigated. In order to provide some

American Journal of Orthodontics and Dentofacial Orthopedics

December 1998

assurance that multiple regression analysis is not simply

capitalizing on chance, a minimum of 5 to 10 subjects
are generally needed for each independent variable tested to predict an outcome variable.21 In other words, in
the Skieller et al study, 220 to 440 subjects in contrast
to 21 subjects would be needed to test the 44 independent variables. In addition, the authors themselves
pointed out the most serious limitation of the study; the
sample included extreme rotational patterns that may be
influential in determining the prediction equation. As a
consequence of having such a biased sample, the
applicability of the predictive variables to less extreme
rotational patterns could be questioned.
To address the latter concern, Lee et al22 tested this
predictive method with 25 subjects with metallic
implants with less extreme facial patterns than those in
the Skieller et al sample. Efforts were made to reproduce the procedures and statistical methods of the
Skieller study as closely as possible. Contrary to the
results reported by the Skieller et al study, the Lee et al
study found the combined predictive power of the four
independent variables to account for only 8% of the
variability observed in mandibular growth rotation.
A limitation in the Lee et al study was the inclusion
of treated subjects. It has been demonstrated that
orthodontic treatment affects mandibular rotation.10,23,24 Consequently, including treated subjects
could weaken the application of the Skieller et al predictive equation that was derived from a sample of
untreated individuals.
Because the Skieller, Bjrk, and Linde-Hansen
study is frequently quoted by orthodontists and even
considered a classic work,25 a need exists to examine
the four identified predictive variables with a large
sample of untreated subjects who present with facial
morphology more typically seen in an orthodontic
practice. The purpose of the present study is to assess
the ability of the Skieller, Bjrk, and Linde-Hansen
method in predicting the direction of mandibular
growth rotation using a sample of 40 untreated subjects
(10 subjects for each of the 4 identified variables) who
are more representative of a general patient population.
MATERIAL AND METHODS

The material for this research project was obtained

from the Iowa Facial Growth Study, a longitudinal
study composed of 183 whites (92 males and 91
females), 97% of whom are of northwest European
ancestry. Records from a set of 40 subjects (20 males
and 20 females) who had not received orthodontic
treatment were randomly selected. A 6-year time interval had been established in the Skieller et al study.
Because accompanying developmental data pinpoint-

American Journal of Orthodontics and Dentofacial Orthopedics

Volume 114, Number 6

Table I.

Variable description

Mandibular inclination determined by

Posterior facial height
Anterior facial height
Mandibular inclination determined by NSL
to mandibular line 1 (NSL -ML1)
Intermolar angle (MOLs-MOLi)

3
4

Shape of the lower border of the mandible (ML1-ML2)

Inclination of the symphysis (CTL-NSL)

1a

661

Definition of variables

Variable number
1

Leslie et al

ing the onset of pubescence were not available, an initial age of 9 years was selected with the follow-up lateral cephalogram at age 15. All subjects in the Iowa
Facial Growth Study had lateral cephalograms made at
these ages. It was assumed that puberty would fall
within this time period. The initial profile headfilms
were traced and the following anatomic landmarks
identified (Fig 1):
Nasion (N): The most anterior point of the nasofrontal
suture.
Sella (S): The center of the cavity outlined by sella turcica.
Menton (Me): The most inferior point on the mandible at
the symphysis.
Articulare (Ar): The point of the intersection of the inferior surface of the cranial base and the averaged posterior
surfaces of the mandibular condyles.
Tangential gonion (tGo): Found by bisecting the angle
formed by a line tangent to the averaged inferior borders
of the mandible and a line through articulare tangent to
the averaged posterior borders of mandibular rami.

Each landmark was then verified by another investigator. The following lines were drawn using definitions established by the Skieller et al study (Fig 1 and
Table I):
Nasion-sella line (NSL).
Mandibular line 1 (ML1): The tangential line formed by
the lower border of the mandible.
Mandibular line 2 (ML2): The line formed tangent to the
lower gonial border and passing through menton.
Molar long axis (MOL) inferior (i) and superior (s).
Chin Tangent Line (CTL).

The four variables of interest were calculated using

these landmarks and lines: mandibular inclination (Figs
2 and 3), intermolar angle (MOLs to MOLi, Fig 1),
shape of the lower border of the mandible (ML1 to
ML2, Fig 1), and inclination of the symphysis (NSL to
CTL, Fig 1). Two measurements were made for the
independent variable mandibular inclination (Figs 2
and 3), both of which had been designated by Skieller
et al as having high levels of predictability. Angular

Method of measurement

s-tgo distance
100 (Fig 2)
n-me distance
Angle between SN line and
mandibular line 1 (Fig 3)
Angle between the maxillary first molar long
axis and the mandibular first molar long
axis (Fig 1)
Angle between the two mandibular lines (Fig 1)
Angle between chin tangent line and SN line
(Fig 1)

measurements were made to within 0.5, and linear

measurements were made to within 0.5 mm. All measurements were repeated by another investigator and
any discrepancies between the independent sets of
measurements were re-measured by the principal
author to the prescribed limits. The Skieller et al
regression equations (Tables II and III) were used to
obtain the predicted mandibular growth rotation using
the four variables of interest. Thus, the initial calculations were done in a prospective manner with no prior
knowledge of the actual rotational outcome.
Next, the final (postpubescent) lateral cephalograms were traced and superimposed on the initial lateral cephalogram. There are many superimposition
techniques that can be used by the clinical orthodontist,
but there is no one best technique. In this study we
elected to superimpose by registering on sella and orienting along NSLa technique used by many clinicians. Actual mandibular rotation was then determined
as the change in mandibular line 2 (ML2) from the initial to final cephalogram. Again, rotational measurements were made by two independent investigators and
discrepancies resolved by the principal author. Forward
rotation (counterclockwise) was designated as negative
(-) and backward rotation (clockwise) was designated
as positive (+).
Statistical analysis was performed using the SAS
statistical software package (SAS Institute Inc,). The
regression equations given by the Skieller et al study
(Tables II and III) were used to calculate the predicted
mandibular rotation with the measurements obtained
for the independent variables. Two measurements were
made for the independent variable mandibular inclination (1 and 1a) and the appropriate regression equation
was used in each instance. Univariate and stepwise
regression analyses were used to evaluate the predictive value of the proposed method as it related to the
sample chosen for this study.
In cases where the number of subjects is small rel-

662

Leslie et al

American Journal of Orthodontics and Dentofacial Orthopedics

December 1998

Fig 2. Independent variable no. 1. Mandibular inclination determined by s-tGO 100

n-Me

Skieller et al regression equation with independent variables 1, 2, 3, and 4

Table II.

75.6 -

R2 =

0.59
(variable 1)
0.32
(variable 2)
0.98
(variable 3)
0.15
(variable 4)
predicted rotation
0.8612

+
=

ative to the pool of possible predictor variables, the forward stepwise selection procedure is unreliable in the
sense that the R-squared statistic will be quite high,
even when the independent variables have no predictive power whatsoever.
To illustrate this point, we performed a Monte
Carlo study that mirrored the number of subjects and
number of possible predictor variables in the Skieller
et al study; however, all the variables were unrelated.
For each simulation, 945 random numbers were generated from a normal distribution and then randomly
grouped into 45 columns with 21 random numbers per
column. The first column of 21 numbers was taken to
be the response variable and the remaining 44 columns
to be the candidate predictor variables. For each simulation a forward stepwise selection procedure was performed and the following quantities were recorded: (1)

Fig 3. Independent variable no. 1a. Mandibular inclination determined by the angle NSL-ML1.

Skieller et al regression equation with independent variables 1a, 2, 3, and 4

Table III.

18.3 +

R2 =

0.44
(variable 1a)
0.29
(variable 2)
0.94
(variable 3)
0.14
(variable 4)
predicted rotation
0.8201

+
=

the largest R-squared value between any predictor

variable and the response and the corresponding Pvalue; and (2) the overall R-squared value and the
number of predictors allowed into the model that
resulted from a forward stepwise selection procedure
with a .1 P-value threshold entry criterion (as in the
Skieller et al article). This procedure was repeated
5000 times.
RESULTS

Sample demographics from the Skieller et al, Lee et

al, and present studies are shown in Table IV. The differences in respective sample size (21, 25, and 40 subjects) should be noted. The inclusion of 40 untreated
subjects in the present study provided 10 subjects for
each of the 4 independent variables investigated. Also,
both Skieller et al and the present study used untreated
samples, whereas Lee et al used a mixture of treated
and untreated subjects.

Leslie et al

American Journal of Orthodontics and Dentofacial Orthopedics

Volume 114, Number 6

Table IV.

Comparison of sample demographics

Sample size (n)

Sex distribution
Males
Females
Baseline age
Treatment status
Population characteristics

Table V.

663

Skieller et al

Lee et al

Present study

21

25

40

12
9
11.3 1.00
Untreated
Extreme growth patterns

11
14
8.5 0.0
Mixed
Less extreme

20
20
9.0 0.0
Untreated
Less extreme

Skieller et al

Lee et al

Present study

63.6 6.4

63.4 2.8

63.5 4.1

35.9 8.2

35.3 4.3

36.2 5.3

173.6 5.6

177.7 3.5

173.3 4.2

2.4 1.8

1.1 0.9

1.4 0.9

87.4 12.2

98.7 7.4

83.6 8.4

Independent variables: mean standard deviation

Independent variables
1
Post face height
Ant face height
1a
(NSL - ML1)
2
(Intermolar angle)
3
(ML1 - ML2)
4
(CTL - NSL)

The means and standard deviations for the independent variables for each investigation are listed in Table
V. The means appear to be remarkably similar with the
exception of variable no. 4 (CTL-NSL). The standard
deviations are much greater in the Skieller et al sample
than either the Lee et al or the current sample. This is
consistent with the Skieller et al selection of a group of
subjects with extreme growth rotation. A comparison
of the ranges for the independent variables in the
Skieller et al study and the present study are shown in
Table VI. These data were not published for the Lee et
al study. Again, the disparity is indicative of the selection of a specific sample of extreme rotators by
Skieller.
The means, standard deviations, and ranges for the
dependent variable (mandibular rotation) in each of the
investigations, when available, are presented in Table
VII. It was found that both Skieller et al and Lee et al
had observed mandibular rotations greater than what
was recorded for this study; the Skieller et al measurements were the largest.
The means, standard deviations, and ranges for the
expected mandibular rotation are listed in Table VIII
(variables 1, 2, 3, and 4) and Table IX (variables 1a,
2, 3, and 4). The means, standard deviations, and
ranges of the differences between the predicted and
the actual observed mandibular rotations are also presented. As can be seen, the difference between the

predicted rotation of the mandible and the observed

rotation of the mandible is quite large with either
regression equation.
The results of simple linear regression analyses are
shown in Table X. The amount of variability accounted
for by each of the independent variables is much lower
in the current study than that found by the Skieller et al
study. For example, variable no. 1 (mandibular inclination) had the highest predictive value found in the
Skieller et al research. It accounted for 62% of the variability observed. The best value obtained in the current
investigation was for variable no. 2, which accounted
for only 5.58% of the variability observed.
The results of stepwise multiple regression analyses
to determine the combined predictability of all four
variables are shown in Table XI (variables 1, 2, 3, and
4) and Table XII (variables 1a, 2, 3, and 4). The order
in which the variables entered each analysis is also
given. In this procedure the goal was to find how much
of the variability could be explained when all four
independent variables were used in combination. The
variables are selected in the order of their ability to
explain additional variance. Using independent variables no. 1, 2, 3, and 4 (Table XI), Skieller et al
obtained a combined predictive ability of 86%. In contrast, the current investigation was only able to account
for a total of 7% of the observed rotation using those
four variables; Lee et al could account for only 8%. In

664

Leslie et al

Table VI.

Independent variables: range

American Journal of Orthodontics and Dentofacial Orthopedics

December 1998

Skieller et al
Independent variables
1
Post face height
Ant face height
1a
NSL-ML1
2
Intermolar angle
3
ML1 - ML2
4
CTL - SNL

Table VII.

Present study

Minimum

Maximum

Minimum

Maximum

54.4

76.8

57.1

74.0

19.0

52.5

23.3

45.0

159.8

180.1

163.0

181.5

0.0

5.5

0.0

4.5

65.0

116.0

65.5

104.5

Dependent variable

Observed mandibular rotation

Skieller et al

Lee et al

Present study

Mean standard deviation

Range: Minimum
Maximum

6.0 4.5
+5.3
16.4

4.9 3.0
NA
NA

1.9 2.1
+3.3
5.8

addition, the order of entry for the variables is different

for the present study compared with the other studies.
On the second stepwise multiple regression analysis
(Table XII), with the use of independent variables 1a,
2, 3, and 4, the combined predictive ability found by
the Skieller et al study was 82%. The current research
obtained only a 9% predictability for those variables.
Lee et al did not present comparable results for this
analysis.
Results of the Monte Carlo study with random
numbers instead of cephalometric measurements
showed that the empirical distribution of the largest
squared correlation between the response and any of
the 44 independent variables had a minimum value of
0.10130 and a maximum value of 0.64688. The mean
and standard deviation were 0.28048 and 0.08151,
respectively. For approximately 90% of the simulations, there was at least one variable that was significantly correlated with the response at the 0.05 level significance level.
The Skieller et al forward stepwise selection procedure used a rejection level of P > .1, and at this level
four variables entered their prediction model. In our
simulations that allowed exactly four predictor variables into the model (there were 393 cases), the overall
R-squared statistic ranged from 0.52473 to 0.86722,
with a median value of 0.67646.
In our simulations that did not restrict the number of
variables entering the model, but instead used the

Skieller et al rejection level of P > .1, the empirical distribution of the overall R-squared statistic for the forward stepwise selection procedure had a mean of
0.84328, a median value of 0.94741, and a standard
deviation of 0.21533. In approximately 20% of the simulations, the R-squared value was exactly 1. This happened whenever the procedure chose 20 variables to be
in the model. (For mathematical reasons, the overall Rsquared statistic will be 1 for any subset of 20 variables
selected from the 44 independent variables.)
DISCUSSION

Successful orthodontic treatment in growing

patients may depend heavily on the amount and direction of jaw growth. Lateral cephalometric radiographs
are indispensable tools in orthodontic diagnosis,
treatment planning, and quantifying dentofacial
changes over time. Attempts, such as that made by
Skieller and co-workers, to apply measurements
made from these films in predicting jaw growth are
necessary and commendable.
Difficulties can arise in the interpretation of these
studies. In our eagerness to find help in predicting jaw
growth, practicing orthodontists sometimes accept
findings from preliminary studies as infallible. The
four variables isolated from the Skieller study are often
referred to in orthodontic seminars to explain why a
patient grew in a given direction.
It should be kept in mind, as noted previously, that

Leslie et al

American Journal of Orthodontics and Dentofacial Orthopedics

Volume 114, Number 6

Predicted mandibular rotation with regression equation for independent variables 1, 2, 3, and 4
Table VIII.

Predicted mandibular rotation with regression equation for independent variables 1a, 2, 3, and 4
Table IX.

Range
Mean standard deviation

Predicted rotation
Difference from
observed rotation

6.71 2.57
4.76 2.94

Minimum

0.79
0.00

Maximum

12.99
11.41

the Skieller et al study sample was purposefully composed of extreme patterns of growth rotation. The
authors themselves pointed out this most serious limitation of their study. The Lee et al study selected a
sample to mirror the means recorded for the Skieller
et al independent variables but with less extreme patterns of growth rotation. As would be true of a typical
orthodontic practice, no predetermined rotational pattern was used in the random selection of the present
sample.
The baseline age of the sample from the present
study and from the Lee et al study differ from the baseline age of the Skieller et al study as exact data were
not available on the onset of puberty. Therefore, a
starting age of 9 years was chosen in an attempt to
incorporate puberty in the 6-year time frame set up by
Dr Skeiller. Certainly, clinical orthodontists would
only be able to make a best guess, based on the available information, as to the onset of puberty in their
patients. With further inquiry, actual pubertal onset
could be determined, but by then the growth spurt may
be well underway.
It is remarkable to note how similar the means for
the independent variables are for the three samples
(Table V), considering that there was no preconceived
selection on the part of the current study. The standard
deviations are quite different, which is reflective of the
differences in extremes between the samples selected.
The observed mandibular rotations (Tables VIII and
IX) are significantly different. This not only represents
the selection for extreme rotational patterns on the part
of Skieller et al, but also the absence of implants as reference points in the current study. As noted by Bjrk,6
half of the actual mandibular rotation that occurs is
masked by remodeling of the lower border of the
mandible. The mean rotation recorded using the
mandibular plane angle is consistent with Bjrks11
statement that, ...the mandible on the average rotates
forward a little during adolescence, and its inclination
thus decreases. Even though the absence of implants
may be viewed as a technical limitation of the present
study, our purpose was to assess the Skieller et al
method as applied to clinical practice where implants

665

Range
Mean standard deviation

Predicted rotation
Difference from
observed rotation

5.64 2.21
3.70 2.57

Minimum

0.57
0.20

Maximum

9.89
10.22

would not be used.

In contrast to the Skieller et al study, our results
demonstrate only minimal ability for any of the four
individual independent variables to predict mandibular
growth rotation. We found that the variable with the
highest predictive value was intermolar angle. This is
interesting considering the difficulty experienced by
most orthodontists in tracing molars on lateral cephalometric radiographs and on the questionable value of
this variable found in current cephalometric landmark
reliablity research.26 The Lee et al study was unable to
find significance at the P < .15 level for any of the four
independent variables.
Finally, in tabulating the combined predictability of
the four variables with regard to mandibular growth
rotation, not only was the level of variability predicted
significantly lower in the present study, but the variables entered the equation in different orders. Regardless of the measurement used to determine mandibular
inclination, a maximum of only 9% predictability was
achieved. These results with untreated subjects support
the results of the Lee et al study, which included treated subjects.
Because of the small sample size of the Skieller et al
study relative to the number of variables tested, there is
a lack of assurance that the multiple regression analysis
results were not simply based on a few extreme and
highly influential observations. Ideally, the Skieller et al
study should be redone prospectively with 220 to 440
subjects if all 44 measurements are to be tested. The
present study was specifically designed to address this
problem. Four independent variables were tested on a
sample of 40 randomly selected subjects.
In our Monte Carlo study, we mirrored the Skieller
et al study but used random numbers instead of actual
cephalometric data. In other words, the Skieller et al
analysis was run repeatedly with meaningless data to
see if the Skieller et al results could have capitalized on
chance. Using the Skieller et al rejection level of P > .1,
we found that a mean of 84% and a median of 94% of
mandibular growth rotation variability could be
explained using meaningless data. This compares favorably with the Skieller et al value of 86%. From these

666

Leslie et al

Table X.

Simple linear regression analysis: R2

American Journal of Orthodontics and Dentofacial Orthopedics

December 1998

Skieller et al
Independent variables
1
Post face height
ant face height
1a
(NSL - ML1)
2
(Intermolar angle)
3
(ML1 - ML2)
4
(CTL - ML1)

Present study

R2

Level of significance

R2

Level of significance

0.62

P < .001

0.0075

P < .59

0.61

P < .001

0.0192

P < .39

0.39

P < .005

0.0558

P < .14

0.27

P < .025

0.0012

P < .83

0.38

P < .005

0.0007

P < .87

Stepwise regression analysis for independent variables 1, 2, 3, and 4

Table XI.

Skieller et al

Lee et al

Present study

Independent variable

R2

Independent variable

R2

Independent variable

R2

1
2
3
4

0.62
0.76
0.81
0.86

1
2
3
4

0.05
0.08
0.08
0.08

2
4
1
3

0.06
0.06
0.07
0.07

Variables are listed in order of entry and R2 are cumulative for all variables included in the model.

Table XII.

Stepwise regression analysis for independent variables 1a, 2, 3, and 4

Skeiller et al

Present study

Independent variable

R2

Independent Variable

R2

1a
2
3
4

0.61
NS
NS
0.82

2
1a
4
3

0.06
0.07
0.09
0.09

Variables are listed in order of entry and R2 are cumulative for all variables included in the model.

results we conclude that stepwise regression should

only be viewed as an exploratory technique when the
number of independent variables is larger than the number of observations, because it is easy to obtain a model
with a high R-squared value even when none of the
independent variables is related to the response variable. In this situation stepwise regression will have a
strong tendency to over fit models, adding variables
with no predictive power. Therefore, the Skieller et al
study should simply be considered exploratory.

tion using the method proposed by Skieller et al.16

The results of this study indicate that the information extracted from a pretreatment lateral cephalogram with the use of this method is inadequate to
permit clinically useful predictions to be made relative to the magnitude or direction of future mandibular growth rotation.
We thank Dr Brent Bankhead, former student at the
University of Iowa College of Dentistry, for his support
during this project.

CONCLUSION

This study assessed the clinicians ability to

accurately predict the future mandibular growth
rotation of untreated persons from a general popula-

REFERENCES
1. Hunter CJ. The correlation of facial growth with body height and skeletal maturation
at adolescence. Angle Orthod 1966;36:44-54.
2. Humphrey GM. Results of experiments on the growth of the jaws. Br J Dent Sci

Leslie et al

American Journal of Orthodontics and Dentofacial Orthopedics

Volume 114, Number 6

1863;6:548-50.
3. Brash JC. The growth of the jaws: Normal and abnormal in health and disease. London: Dental Board of the United Kingdom; 1924.
4. Brodie AG. On the growth pattern of the human head from the third month to the
eighth year of life. Am J Anat 1941;68:209-62.
5. Robinson IB, Sarnat BG. Growth pattern of the pig mandible. A serial roentgenographic study using metallic implants. Am J Anat 1955;96:37-64.
6. Bjrk A. Variations in the growth pattern of the human mandible: longitudinal radiographic study by the implant method. J Dent Res 1963;42:400-11.
7. Enlow EH, Harris DB. A study of the postnatal growth of the human mandible. Am J
Orthod 1964;50:25-50.
8. McNamara JA Jr, Graber LW. Mandibular growth in the Rhesus monkey. Am J Phys
Anthropol 1975;42:15-24.
9. Brodie AG. Facial patterns: a theme on variation. Angle Orthod 1946;16:75-87.
10. Schudy FF. The rotation of the mandible resulting from growth: its implications in
orthodontic treatment. Angle Orthod 1965;35:36-50.
11. Bjrk A. Prediction of mandibular growth rotation. Am J Orthod 1969;55:585-99.
12. Bjrk A, Skieller V. Facial development and tooth eruption: an implant study at the age
of puberty. Am J Orthod 1972;62:339-83.
13. Schudy FF. Vertical growth versus anteroposterior growth as related to function and
treatment. Angle Orthod 1964;34:75-93.
14. Baumrind S, Korn EL, West EE. Prediction of mandibular rotation: an empirical test
of clincian performance. Am J Orthod 1984;86:371-85.
15. Ricketts RM. Planning treatment on the basis of the facial pattern and an estimate of

667

its growth. 1957;27:14-37.

16. Skieller V, Bjrk A, Linde-Hansen T. Prediction of mandibular growth rotation evaluated from a longitudinal implant sample. Am J Orthod 1984;86:359-70.
17. Johnston LE. A simplified approach to prediction. Am J Orthod 1975;67:253-7.
18. Bahti SN, Wright GW, Leighton BC. A proposed multivariate model for prediction of
facial growth. Am J Orthod 1979;75:264-81.
19. Singer CP, Mamandras AH, Hunter WS. The depth of the mandibular antegonial notch
as an indicator of mandibular growth potential. Am J Orthod 1987;91:117-24.
20. Aki T, Nanda RS, Currier GF, Nanda SK. Assessment of symphysis morphology
as a predictor of the direction of mandibular growth. Am J Orthod 1994;106:
60-9.
21. Norman GR, Streiner DL. Biostatistics: the bare essentials. St. Louis: Mosby; 1994.
22. Lee RS, Daniel FJ, Swartz M, Baumrind S, Korn EL. Assessment of a method for the
prediction of mandibular rotation. Am J Orthod 1987;91:395-402.
23. Schudy FF. The control of vertical overbite in clinical orthodontics. Angle Orthod
1968;38:19-39.
24. Odegaard J. Growth of the mandible studied with the aid of metal implant. Am J
Orthod 1970;57:145-57.
25. Casko J. Practical reviews in orthodontics. Vol. 3, number 1. Educational Reviews Ltd,
May 1991.
26. Baumrind S. The reliability of head film measurements: landmark identification. AJO
1971;60:111-27.

BOUND VOLUMES AVAILABLE TO SUBSCRIBERS

Bound volumes of the American Journal of Orthodontics and Dentofacial Orthopedics
are available to subscribers (only) for the 1998 issues from the Publisher, at a cost of $87.00
($104.86 Canada and $98.00 international) for Vol. 113 (January-June) and Vol. 114 (JulyDecember). Shipping charges are included. Each bound volume contains a subject and author
index and all advertising is removed. Copies are shipped within 60 days after publication of
the last issue of the volume. The binding is durable buckram with the journal name, volume
number, and year stamped in gold on the spine. Payment must accompany all orders. Contact
Mosby, Inc., Subscription Services, 11830 Westline Industrial Drive, St. Louis, MO 631463318, USA; telephone (314)453-4351 or (800)325-4177.
Subscriptions must be in force to qualify. Bound volumes are not available in place
of a regular Journal subscription.