979

Thoracic Kyphosis:
Downloaded from www.ajronline.org by 37.120.131.236 on 05/04/20 from IP address 37.120.131.236. Copyright ARRS. For personal use only; all rights reserved

Range in
Normal Subjects

Gerald T. Fon1 ‘ 2 Thoracic kyphosis was measured on chest radiognaphs of 31 6 “normal” subjects by
Michael J. Pitt1 means of a modification of the Cobb technique for measuring scoliosis. Patients were
A. Cole Thies Jr.3 accepted as “normal” if they had
no thoracic or spinal complaints or radiographic
abnormalities in the chest including
the thoracic spine. A total of 1 59 male and 157
female subjects 2-77 years old was studied. The relation among age, gender, and
kyphosis were determined using least squares fits of first-order linear mathematical
models. These results were also used to determine the expected ranges of kyphosis
for a “normal” patient of a given age and gender. The degree of kyphosis increased
with age and the rate of increase was higher in females than in males. Clinical
explanations for this differential increase are discussed.

Exaggeration of the normal thoracic curvature is associated with a variety of

conditions, such as Scheuerrnann disease, congenital spinal anomalies, and
paralytic and metabolic processes, as well as inflammatory or traumatic condi-
tions [1 An increased
]. incidence of spinal curvature in patients with cystic
fibrosis was recently reported [2]. Severe thoracic kyphotic curves may result in
pain, cardiopulmonary failure, and even paraplegia, in addition to cosmetic
deformities.
Some degree of thoracic convexity in the saggital plane is normally present
and a range of 20#{176}-40#{176}
has been suggested from limited studies [3-5]. While
there has been extensive research on the treatment and prevention of spinal
curvatures, we could find no studies specifically on the normal range of thoracic
kyphosis, although ranges are assumed in published studies for other purposes
[4-6]. The purpose of our study was to determine the expected range of thoracic
kyphosis in a group of patients of varying age and gender as seen on routine
chest radiographs.

Received June 1 2, 1 979; accepted after revi-

sion December 1 8, 1979.
. . Materials and Methods
, Department of Radiology, University of An-
zona, Health Sciences Center, Tucson, AZ 85724. Chest radiographs of 450 patients examined at the University of Arizona Health Sciences
Address reprint requests to M. J. Pitt. ‘ .

Center in Tucson from 1 976 to 1 978 were selected. These radiographs had bean inter-
2 Present address: Department of Radiological pretad as normal by a radiologist at the time of the examination and were classified
Sciences, UCLA School of Medicine, Los Angeles, . ‘ ‘ ‘

CA 90024. according to the American College of Radiology index [7]. The patient identification
. . . numbers and the code for a normal chest examination had been stored on microfiche. From
3 Division of Computer Systems and Biostatis- . .

tics, University of Arizona, Health Sciences Cen- this file 1 50 cases/year for the years 1 976-1 978 were sampled with about equal numbers
ten, Tucson, AZ 85724. for each age group and gender.
AJR 134:979-983, May 1980 Even though the chest radiographs had been classified as normal, they were reviewed
0361-803X/80/1345-0979 $00.00 again by the senior author (G.T.F.). Hence the films ware evaluated by two radiologists
© American Roentgen Ray Society over a 1 -3 year period. Of the 450 patients sampled, only 31 6 were selected for the study.
 980 FON ET AL. AJR:134, May 1980

These mat the two important criteria of: (1 ) no abnormality of heart,

lungs, or thoracic skeleton and (2) proper positioning and suitable
radiographic technique.
For study purposes, patients were accepted as ‘ ‘normal’ ‘ if no
thoracic or spinal complaints were noted in the request for radio-
logic examination and no radiologic abnormalities in the chest
Downloaded from www.ajronline.org by 37.120.131.236 on 05/04/20 from IP address 37.120.131.236. Copyright ARRS. For personal use only; all rights reserved

including the thoracic spine were identified. Medical charts were

not reviewed. All patients were ambulatory. Most ofthe radiographs
selected were of patients who had the examination as a preoperative
routine for elective surgery or preemployment. While the general
health status of some of these patients was not optimal, it was
assumed that the patients were sufficiently fit to be amublatory and
to meet the other criteria required for positioning as described
below.
During the second review of the radiographs by the senior author,
changes in the lungs or thoracic spine that could be considered
normal in the initial interpretation but might affect the thoracic
kyphotic curve were specifically sought. Changes that increase the
curvature of the spine include atelectasis, hyperaeration, or air-
trapping. Bony changes such as pectus excavatum, spinal deform-
ities, Schauermann disease, and scoliosis of the spine were ax-
cluded. It is well known that the musculature may be affected by
systemic diseases with resultant effect on spinal curve. By exclud-
ing the radiographs that showed any scoliosis patients with signifi-
cantly weak musculature were presumably excluded because it Fig. 1 -Method of measuring kyphosis.
would be very unusual for poor muscular support to affect the
kyphotic curve only without associated scoliosis.
The recognition of osteoporosis or loss of mineral content of ment error. Another source of error is the patient’s positioning at
bones especially in the milder forms from simple films is notably the time the radiographs were taken. A prospective study would be
difficult. However, by excluding any patients with secondary necessary to control this more completely than in this study.
changes in the spine resulting from demineralization, the effect of Statistical analysis. Whether a relation existed between age of
the loss of mineral content on spinal curvature would be reduced to patient and degree of kyphosis was determined by testing the
a minimum. Such secondary changes as biconcave depressions of statistical significance of the slope of a straight line (i.e., first-order
the vertebral end plates resulting in ‘ ‘fish vertebrae,’ ‘ vertebral linear) model fitted separately for males and females by linear
collapse, and significant vertebral wedging were reasons for exclu- regression. The adequacy of each straight line model was deter-
sion. For those patients over 50, mild degenerative changes in the mined by visual inspection of fits and by statistical tests. The
spine consisting of minor marginal osteophytes ware accepted as difference of slopes between males and females was also tested,
normal. For all selected patients these changes were not mentioned using an analysis of covarianca procedure. Confidence regions for
in the initial report and presumably were considered to be within kyphosis based on age and gender were then determined, using
the norm. the linear regressions [9]. Finally, a descriptive analysis of kyphosis
It is obvious that the degree of kyphosis is related to patient by age and gender was performed, computing means and standard
positioning. To minimize positioning variations, the radiographs deviations for each age-gender group. The results of the descriptive
included in this study met the following criteria: (1 ) they were made analysis were used to further investigate the adequacy of the first-
with the patient standing and with a focal-film distance of 1 .8 m; (2) order linear model.
the patient’s r.rms were above the shoulders in the lateral view, with
neither humerus below the horizontal position (this would help
reduce any increase in the thoracic curve due to positioning); and Results
(3) patients under 3 years of age were excluded since most of these
The composition of the patient population according to
patients were supine or strapped to the Brat Board (the only
age and gender is shown in table 1 . Linear regression of
exception was a 2-year-old whose radiograph was made erect).
The degree of kyphosis was measured by a modification of the kyphosis on age for both males and females resulted in an
Cobb technique [8] as used for scoliosis. The upper and lower adequate fit in the age range 6-75 years and a poor fit in
vertebral bodies defining the curve were selected and lines were the range 0-3 years. Only at age 6 and over did each model
drawn, extending along the superior border of the upper end indicate an equal likelihood of over- or underprediction of
vertebra as well as along the inferior border of the lower end kyphosis. For this reason, further regression analysis was
vertebra. Perpendiculars were drawn from these two lines and the limited to the age range of 6-75 years for males and
angle was measured at the intersection (fig. 1 ). An angle rule was females.
used to draw the perpendiculars and measure the angles. The Results of that analysis are shown in figures 2 and 3. In
measurements were done on one set of films per patient and in
each figure, the middle line represents the predicted value
most cases that was the only available set to fall within our definition
of kyphosis, and the outer lines delineate a 95% confidence
of normal. For each patient, one measurement of the thoracic
region for an observation of kyphosis for a patient of a given
kyphotic curve was made. To test measurement error, a sample of
30 cases out of the 31 6 was drawn, and the cases remeasured. age (parameters of both models are shown in table 2). For
The maximum difference between first and second readings was both males and females the slopes of these lines were
50, It should be noted that this is only one aspect of total measure- significantly different from zero (p < 0.01 ) indicating that
 AJR:134, May 1980 THORACIC KYPHOSIS 981

TABLE 1 : Distribution of Patients by Age and Gender

Male Female Totals

Age (years)
No. Patients Mean Age No. Patients Mean Age No. Patients Mean Age

2-9 . . . 26 5.62 23 5.43 49 5.53

10-19 . . . 28 14.21 22 14.73 50 14.44
Downloaded from www.ajronline.org by 37.120.131.236 on 05/04/20 from IP address 37.120.131.236. Copyright ARRS. For personal use only; all rights reserved

20-29 37 24.62 24 24.25 61 24.48

30-39 . . . . . 26 33.28 26 35.23 52 34.31
40-49 . 20 43.55 32 43.75 52 43.67
50-59 . . 10 53.20 17 54.41 27 53.96
60-69 . . . 9 63.33 7 65.29 16 64.19
70-79 . . 3 74.00 6 72.17 9 72.78
Totals . 1 59 28.42 1 57 32.88 31 6 30.63

kyphosis was related to the age of the patient. The ability of 60r
age to predict kyphosis is indicated by two measures in
table 2, standard error of the estimate and r2. Standard error 50
of the estimate is a measure of the model’s average mac-
curacy of prediction. The percentage of overall variation of a)
kyphosis explained by the model is measured by r2 (the
square
Analysis
and predicted
or overprediction
Parameters
of the product-moment
of residual values
kyphosis)
across
of the straight
the entire
line
correlation
(difference
indicated equal
age
models
coefficient
between observed
likelihood
ranges.
are
r).

shown
of under-

in table
I
2. Analysis of covariance indicated that females showed a
statistically significantly greater slope (parameter b in table
2) than males (p < 0.05). 0 10 20 30 40 5060 70 80

Descriptive profiles of kyphosis are shown in tables 3 and

4 (these tables include patients of ages 3-5 years, who Age (Years)
were not included in the linear regression models). The
Fig. 2.-Linear regression of kyphosis on age, including prediction and
adequacy of the linear model can be further investigated by
comparing: (1 ) the male-female differences at each age 95% confidence regions, for male patients. (See text.)
range and (2) the patterns of standard deviation across age
groups. The latter comparison indicates no immediate pat-
tern for females, but there exists a suggestive decrease in
standard deviation of kyphosis as age increases for male
patients. This decrease suggests that confidence bounds in
figure 2 may be slightly too wide for older patients and
slightly too narrow for younger patients. The former corn-
parison is depicted in figure 4, where mean kyphosis for
each age group is plotted against mean age within each age
group for males and females. An unpaired t test of differ-
ences between group means (males versus females) is
performed at each age group and the results of each test
displayed as significance (p)levels in figure 4. An interesting
result of this analysis is the similarity of males and females
until about age 40, after which females exhibit significantly
higher kyphosis for each group (the highest age group, 70-
Age (Years)
79 years, included a very small number of cases and did
not indicate statistical significance). These results may in- Fig. 3.-Linear regression of kyphosis on age, including prediction and
dicate a subtle departure from the straight line model for 95% confidence regions, for female patients. (See text.)

female patients that is not immediately apparent in figure 3.

kyphosis increases with age and the rate of increase is

Discussion
higher in females than in males; this appears to be more
This study shows that the normal range of kyphosis is obvious after age 40. In a limited study of a group of patients
related to both age and gender of a patient. The degree of 60-79 years old, Cowan [1 0] also found an increase in the
 982 FON ET AL. AJR:134, May 1980

TABLE 2: Parameters of Least Squares Fit For Mathematical

Models of Form Kyphosis = b (age) + a TABLE 4: Degree of Kyphosis in Females By Age

Gender No Cases b a SE r r Kypho sis ()

No
Age (years) Cases
Male 147 0.204 21.8 7.6 0.41 0.17 Mean SD Minimum Maximum
Female 148 0.345 19.2 7.9 0.62 0.38
2-9 23 23.87 6.67 8 36
Downloaded from www.ajronline.org by 37.120.131.236 on 05/04/20 from IP address 37.120.131.236. Copyright ARRS. For personal use only; all rights reserved

10-19 . 22 26.00 7.43 11 41

20-29 24 26.83 7.98 7 40
30-39 26 28.42 8.63 10 42
TABLE 3: Degree of Kyphosis in Males By Age 40-49 32 32.66 6.72 21 50
50-59 17 40.71 9.88 22 53
No Kyphosis (1
60-69 7 44.86 7.80 34 54
Age (years)
Cases . .
Maximum
. 70-79 6 41.67 9.00 30 56
Mean SD Minimum

2-9 26 20.88 7.85 5 40

10-19 28 25.11 8.16 8 39
20-29 37 26.27 8.12 13 48
49 60
30-39 26 29.04 7.93 13
40-49 20 29.75 6.93 17 44
50-59 10 33.00 6.46 25 45 50
60-69 9 34.67 5.12 25 62 .. . females
70-79 3 40.67 7.57 32 66

kyphotic
method
angle
for measurement.
with age, even though he used a different :
a
: pP.16 p.69 P-.79 p.79 pa.05 p.04 p.OO7
males

The increase in the kyphotic curve with age is not unex-

10
pected, because of the associated changes in the soft
tissues and mineral content of the bones with the progres-
sion of years. An association of progressive increase in 0 10 20 30 40 50 60 70 80
spinal curvatures with gradual compression wedging of the
vertebrae and its narrowing of the intervertebral discs have Mean Age (Years)

been described [1 1 , 1 2]. The difference in the rate of

Fig. 4.-Mean kyphosis for males and females by 1 0-year age intervals,
increase in the degree of kyphosis between genders in the (See text.)
older age groups is interesting. It is significant in that the
higher rate of increase in kyphosis females as shown in this
study where the measurements were done on radiographs
is in agreement with the findings of Milne and Lander [13] abnormals’ ‘ in the older age groups. As is the case for any
who used a surveyor’s flexicurve and did the measurements sample of patients from one location the results of this study
on the subjects themselves. They used an entirely different cannot be strictly applied to the ‘ ‘general population.’ ‘ How-
method and the kyphosis was measured indirectly by cal- ever, in the absence of any control study where the sample
culation of an ‘ ‘ index of kyphosis. ‘ ‘ These authors later is a true random sample from the ‘ ‘general population,”
showed that the difference in kyphosis between the two these results should serve as useful guidelines for prediction
genders in the older population was not due to a difference of normal thoracic kyphosis.
in the degree of wedging in the vertebral bodies as one
might expect [1 4]. There were other variables related to age
ACKNOWLEDGMENT
that were contributory to the thoracic kyphosis in females.
They postulated the possibility of poor posture and aging of We thank Janet Quinones for manuscript preparation.
soft tissues with resultant loss of muscle tone leading to
increased kyphosis in older females. It is also reasonable to
REFERENCES
postulate that the relative physical inactivity in females,
probably related to occupation, may decrease the tone on 1 . Keim H. Kyphosis and lordosis in the adolescent spine. New
the spinal ligaments and muscles. The dependent breasts York: Grune & Stratton, 1976:193-209
may further accentuate the kyphotic curve in these older 2. Erikkila JC, Warwick WJ, Bradford DS. Spinal deformities and
females. cystic fibrosis. Clin Orthop 1978;1 31 :146-150
3. Bosekar EH. The determination of the normal thoracic ky-
The association of kyphosis with age and gender darn-
phosis; a roentgenographic study of the spines of 1 21 “nor-
onstrated in our study clearly shows that the range of 20#{176}-
mal’ ‘ children. Presented at Gillete Children’s Hospital, St.
40#{176}
that has been reported [1 , 2, 6] as a reasonable normal
Paul, Minn., 1958
range for thoracic kyphosis is inadequate because this 4. Bradford DC, Moe JH, Winter AR. Kyphosis and postural
range of values in our population would have resulted in roundback deformity in children and adolescents. Minn Med
‘ ‘false normals’ ‘ in the younger age groups and ‘ ‘false 1973;56:1 14-1 20
 AJR:124, May 1980 THORACIC KYPHOSIS 983

5. Roaf A. Vertebral growth and its mechanical control. J Bone 1 0. Cowan NA. The frontal cardiac silhouette in older people. Br
Joint Surg (Br] 1 960;42 : 40-59 HeartJ 1965;27:231 -235
6. Bradford D. Editorial comment-kyphosis. Clin Orthop 1 1 . Nicholas JA, Wilson PD. Osteoporosis of the aged spine. Clin
1977;1 28:2-4 Orthop 1 963;26 :19-33
7. American College of Radiology. Index for roentgen diagnoses, 1 2. Rowe CA, Sorbie C. Fractures of the spine in the aged. Clin
3d ad. Baltimore: Waverly, 1975 Orthop I 963;26 : 34-39
Downloaded from www.ajronline.org by 37.120.131.236 on 05/04/20 from IP address 37.120.131.236. Copyright ARRS. For personal use only; all rights reserved

8. Cobb AJ. Outline for the study of scoliosis. Am Acad Orthop 1 3. Milne JS, Lander lJ. Age affects in kyphosis and lordosis in
Surg 1948;5:261 -275 adults. Ann Hum BioI 1974;1 :327-337
9. Draper N, Smith H. Applied regression analysis. New York: 14. Milna JS, Lander IJ. The relationship of kyphosis to the shape
Wiley, 1966:23-24 of vertebral bodies. Ann Hum Biol 1976;3: 173-179