ONLINE ONLY

Geometric morphometric evaluation of cervical vertebrae shape and its relationship to skeletal maturation
Athina Chatzigiannia and Demetrios J. Halazonetisb Thessaloniki and Athens, Greece Introduction: Cervical vertebrae shape has been proposed as a diagnostic factor for assessing skeletal maturation in orthodontic patients. However, evaluation of vertebral shape is mainly based on qualitative criteria. Comprehensive quantitative measurements of shape and assessments of its predictive power have not been reported. Our aims were to measure vertebral shape by using the tools of geometric morphometrics and to evaluate the correlation and predictive power of vertebral shape on skeletal maturation. Methods: Pretreatment lateral cephalograms and corresponding hand-wrist radiographs of 98 patients (40 boys, 58 girls; ages, 8.1-17.7 years) were used. Skeletal age was estimated from the hand-wrist radiographs. The rst 4 vertebrae were traced, and 187 landmarks (34 xed and 153 sliding semilandmarks) were used. Sliding semilandmarks were adjusted to minimize bending energy against the average of the sample. Principal components analysis in shape and form spaces was used for evaluating shape patterns. Shape measures, alone and combined with centroid size and age, were assessed as predictors of skeletal maturation. Results: Shape alone could not predict skeletal maturation better than chronologic age. The best prediction was achieved with the combination of form space principal components and age, giving 90% prediction intervals of approximately 200 maturation units in the girls and 300 units in the boys. Similar predictive power could be obtained by using centroid size and age. Vertebrae C2, C3, and C4 gave similar results when examined individually or combined. C1 showed lower correlations, signifying lower integration with hand-wrist maturation. Conclusions: Vertebral shape is strongly correlated to skeletal age but does not offer better predictive value than chronologic age. (Am J Orthod Dentofacial Orthop 2009;136:481.e1-481.e9)

hape of the cervical vertebrae is related to skeletal maturation, permitting skeletal age assessment from cephalograms, instead of subjecting the patient to the additional radiation of a hand-wrist radiograph.1-6 However, in spite of extensive research during the past years, statistical and biological details about the relationship between vertebral shape and skeletal maturation are still missing. Regarding statistics, the accuracy of prediction has not been quantied with a condence interval. The question, therefore, arises as to whether the cervical vertebral maturation (CVM) method (any of those proposed in the literature)

Private practice, Thessaloniki, Greece. Associate professor, Orthodontic Department, Dental School, University of Athens, Athens, Greece. Supported by the European Virtual Anthropology Network, a Marie Curie Research Training Network (MRTN-CT-2005-019564). The author DJH reports nancial interest in the software Viewbox 4 used for analysis of the data. The other author reports no commercial, proprietary, or nancial interest in the products or companies described in this article. Reprint requests to: Demetrios J. Halazonetis, 6 Menandrou St, Kissia 145 61, Greece; e-mail, dhal@dhal.com. Submitted, March 2009; revised and accepted, April 2009. 0889-5406/$36.00 Copyright 2009 by the American Association of Orthodontists. doi:10.1016/j.ajodo.2009.04.017

can provide more accurate information than an assessment based on comparatively simpler procedures such as chronologic age or the timing of secondary sexual characteristics. Regarding the biological issue, shape measurement has been mainly conned to the heightwidth ratio of the vertebral bodies and the depth of the inferior concavity.5-9 Such measurements correspond to a limited subset of the overall shape information. Conceivably, a more comprehensive shape evaluation could provide additional information to further show the underlying biologic relationship and presumably increase predictive power. Maturational status has been traditionally evaluated by hand-wrist radiographs, which are considered the gold standard, even though their clinical usefulness has been questioned.9-12 The most well-known methods include the atlases of Greulich and Pyle13 and Tanner and Whitehouse,14 and the methods of Bjo rk,15 Grave 16 17,18 Fishman,19,20 and Brown, Ha gg and Taranger, 21 . Recently, software packages for comand Sempe puter-aided evaluation of hand-wrist maturation have been developed.22,23 Cervical vertebrae methods are advantageous in orthodontics because they do not need additional patient
481.e1

481.e2

Chatzigianni and Halazonetis

American Journal of Orthodontics and Dentofacial Orthopedics October 2009

Table I.

Descriptive statistics and comparisons between the groups

Girls (n 5 58) Chronologic age (y) Mean SD Range Median SML Mean SD Range Median FMI Mean SD Range Median NS, Not signicant. Boys (n 5 40) t statistic (P value)

11.96 1.86 8.07-16.55 11.74 787 164 444-995 834 6.96 2.86 1-11 7

12.63 1.92 8.91-17.71 12.55 659 210 337-991 701 5.05 3.26 0-11 5

1.73 (0.09) NS

3.38 (0.001)

3.00 (0.004)

radiation exposure. They evaluate the rst 4 or 5 vertebrae, excluding the atlas, and distinguish 5 or 6 maturational stages.3-5 Their assessment is based on the change in shape of the vertebral bodies, mainly the height-width ratio and the depth of the inferior concavity. These features have been shown to progress during ontogeny in a caudal direction, from C2 to C6. Although the methods are simple to apply clinically, they have certain limitations, as described above. The aims of this study were to measure vertebral shape quantitatively and comprehensively, and to evaluate the correlation between vertebral shape and skeletal maturation, as assessed by hand-wrist radiographs. To measure shape, we used the tools of geometric morphometrics on the rst 4 vertebrae. We sought to establish prediction condence intervals for estimating maturation level from vertebral shape and compare the predictions with simpler maturation measures, such as chronologic age.
MATERIAL AND METHODS

Fig 1. Fixed landmarks (lled circles) and sliding semilandmarks (open circles) used in this investigation.

The sample consisted of pretreatment lateral cephalometric radiographs of 98 white patients of Greek ethnic origin (40 boys, 58 girls). Inclusion criteria were age between 8 and 18 years, good-quality radiographs with a reference ruler on the cephalostat for exact measurement of the magnication factor, the rst 4 cervical vertebrae in full view, and a hand-wrist radiograph taken on the same date as the lateral cephalogram. Patients with congenital malformations or syndromes were excluded. Descriptive statistics are shown in Table I. The radiographs were scanned at 150 dpi and digitized by using the Viewbox 4 software (dHAL software,

Kissia, Greece). Each of the 4 rst cervical vertebrae was traced by using cubic spline curves around their outlines. A total of 187 landmarks were automatically placed by the software on the vertebral outlines in predened positions (see below).24 Two types of landmarks were used: xed landmarks, generally placed on the corners of vertebral bodies or processes, and sliding semilandmarks, placed between the xed landmarks on the outline curves.25-27 For each vertebra, the landmarks were dened as follows (Fig 1). C1: 6 xed landmarks (at the most anterior and posterior points of the vertebra and at the most superior and inferior points of the upper and lower articular processes) and 31 sliding semilandmarks (total, 37 points). C2: 3 xed landmarks on the odontoid process (the most superior point and the 2 corners at the inferior surface), 3 xed landmarks on the vertebral process (the most anterior point, the most posterior point, and the most posterior-inferior point of the inferior articular surface), 17 sliding semilandmarks on the odontoid process, and 37 sliding semilandmarks on the remaining vertebra (total, 60 points). C3 and C4: 4 xed landmarks on the body (at the 4 corners), 12 sliding semilandmarks on the body, 4 xed landmarks on the articular process, 13 sliding semilandmarks on the articular process, 3 xed landmarks on the spinous process, and 9 sliding semilandmarks on the spinous process (total, 45 points on each vertebra). The x and y coordinates of all points were scaled according to the magnication of each radiograph to correspond to true life size.

American Journal of Orthodontics and Dentofacial Orthopedics Volume 136, Number 4

Chatzigianni and Halazonetis

481.e3

Table II.

Percent variance explained by the 10 rst PCs for each object studied in form space and shape space
C1v C2v Form Shape 22.8%* 14.4%* 13.5%* 7.9%* 5.9%* 5.4%* 3.4%* 3.3%* 3.1%* 2.6% 82.4% Form 29.1%* 16.2%* 14.7%* 10.9%* 4.9%* 4.4%* 3.9%* 2.9%* 2.0% 1.8% 90.8% C3v Shape 21.9%* 20.2%* 18.0%* 6.7%* 6.2%* 5.2%* 3.9%* 2.9% 2.5% 2.0% 89.6% Form 34.2%* 15.3%* 11.9%* 8.4%* 6.8%* 4.5%* 4.1%* 2.5%* 2.3% 1.5% 91.5% C4v Shape C2od Form Shape C3bd Form Shape C4bd Form Shape

Form

Shape

PC1 41.5%* 18.9%* 46.5%* PC2 11.3%* 15.7%* 12.5%* PC3 9.0%* 11.1%* 7.9%* PC4 6.6%* 8.8%* 6.4%* PC5 5.0%* 8.1%* 4.3%* PC6 4.7%* 5.2%* 3.2%* PC7 3.1%* 4.6%* 3.0%* PC8 2.7%* 3.6%* 1.9%* PC9 2.1% 3.3%* 1.8%* PC10 1.9% 2.9% 1.7%* Sum 87.9% 82.2% 89.2%

22.6%* 70.6%* 26.3%* 69.3%* 70.9%* 67.3%* 64.2%* 17.9%* 7.6%* 18.4%* 19.8%* 8.4% 20.4%* 10.2%* 16.6%* 5.5%* 14.7%* 3.2% 5.8% 3.6%* 6.8% 9.8%* 4.4%* 10.0%* 2.2% 4.4% 2.4% 4.9% 6.3%* 3.0%* 6.3%* 1.6% 2.6% 1.7% 3.7% 5.9%* 1.8%* 5.8%* 1.0% 2.2% 1.2% 2.4% 3.6%* 1.5%* 3.9% 0.8% 1.2% 0.8% 2.0% 3.3%* 1.2% 3.0% 0.4% 0.9% 0.7% 1.3% 2.2% 0.9% 2.4% 0.4% 0.7% 0.5% 1.0% 1.8% 0.7% 1.7% 0.3% 0.6% 0.4% 0.7% 89.9% 97.2% 92.5% 99.0% 97.8% 98.8% 97.2%

*PCs that were considered meaningful and retained for further analysis, as determined by at least 2 of the 3 criteria mentioned in the text.

We considered 7 objects for shape analysis: the 4 vertebrae including all points as described above (objects C1v, C2v, C3v, and C4v), the odontoid process of C2 (object C2od), and the bodies of C3 and C4 (objects C3bd and C4bd). Sliding of the semilandmarks and calculation of the average for each of the 7 objects were accomplished in an iterative fashion, as described by Bookstein25 and extended by Gunz26 and Gunz et al.27 For each object, at each iteration step, the points were superimposed on the already calculated average (at the rst iteration, the rst patient was used), and the semilandmarks were adjusted to minimize the bending energy by sliding on the outline curves and projecting. Ten cycles of sliding and projecting were done at each iteration step. Three iterations were found sufcient to converge. Principal component analysis was used for the 7 objects separately. The analysis was performed in shape spaceie, by using the residuals of the point coordinates after Procrustes superimposition on the consensus shapeand in form space (also called size-shape space28)ie, by including the logarithm of centroid size (CS) (we refer to the natural logarithm of base e).29 Sexual dimorphism was evaluated with permutation tests, by using Procrustes distance as the test statistic. Skeletal maturation (see below) was regressed on the principal components (PCs) as the predictors, by using multiple linear regression. The hand-wrist radiographs were evaluated by using 21 and the related software package the method of Sempe Maturos 40 (Groupe Franc ais dAuxologie, Ecully, France).22 This method evaluates the skeletal maturation level of each patient based on 22 features of the car maturation levels pal bones and phalanges. Sempe (SML) range from 0 at birth to 999 at the adult stage

when growth is completed. The reliability of the handwrist assessments were measured by repeating the evaluations on 25 randomly selected patients later. In addition to SML, skeletal maturation was assessed according to the simpler method of Fishman,19,20 with only 6 sites evaluated. The Fishman maturity indicators (FMI) have values from 1 to 11. We assigned the value of zero to patients who had not reached the rst level (width of the epiphysis as wide as the diaphysis at the proximal phalanx of the third nger). All evaluations and tracings were done by the same investigator (D.J.H.). Geometric morphometric routines and permutation tests were done with Viewbox 4. Statistical tests were performed by StatsDirect (StatsDirect, Cheshire, United Kingdom).
RESULTS

Replicate assessments of the 25 hand-wrist radiographs showed no systematic error in SML (paired t test: t value 5 1.49; P 5 0.15). The standard deviation of measurement error was 21.6, and the coefcient of reliability was 97.5%.30 Descriptive statistics relating to age and skeletal maturation of the groups by sex are shown in Table I. The male group was slightly ahead in chronologic age (by 8 months), but the difference was statistically significant only at the 10% level. Because of comparable chronologic ages, the 2 groups differed signicantly in skeletal maturation; the female group was more mature by approximately 2 FMI units and 130 SML units. The 2 methods of skeletal maturation assessment showed statistically signicant and high correlations between them. The coefcient of determination (r2) of FMI vs SML was calculated after rescaling FMI to maximize linearity against SML and was found equal to

481.e4

Chatzigianni and Halazonetis

American Journal of Orthodontics and Dentofacial Orthopedics October 2009

94.1% and 93.6% for the female and male groups, respectively. We used SML for the rest of the statistical tests, because it is based on more osseous sites than FMI, and it is graded on a ner scale. Table II shows the percentages of variation explained by the 10 rst PCs for each of the 7 objects studied. The number of PCs that summarize meaningful variations of the data, and should therefore be retained, is difcult to establish. We used 3 criteria31,32 to determine how many PCs to retain: the broken-stick criterion, the rnd-lambda, and the avg-rnd; the last 2 are randomization methods based on eigenvalues. We retained as many PCs as specied by at least 2 of the 3 criteria (Table II), and these will be called meaningful PCs. The randomization methods were run by using 1000 permutations. Figure 2 shows the consensus shape of each object and the shapes representing 3 and 13 SD along the PC1 axis in shape space and form space. CS was larger in the boys than the girls for all 7 objects (Table III). The difference in size ranged from 4.8% to 7.5%. Figure 3 shows regression plots of SML vs logarithm of CS for C3v and C3bd (plots for other objects were similar). The regression lines are parallel (P .0.05) but well separated, showing that boys are larger than girls at every maturation level. All 7 objects showed sexual dimorphism in form space (Table IV). This was mostly due to size differences (in form space, PC1 mostly incorporates size rather than shape), although shape was also a factor, as attested by the signicant differences in shape space for all objects except for C4v, and marginally for C2v. Shape differences between the sexes were an expected nding because of the difference in maturation levels between the 2 groups. SML was regressed on the meaningful PC scores of each object by using multiple linear regression. Table V shows the square of the correlation coefcient (r2, coefcient of determination) for each object and the 90% CI for prediction. Prediction intervals (PI) widen by increasing the distance from the mean; here, they are reported by using the consensus (average) shape as the reference value. The 90% PI was selected instead of the customary 95%, because this would be an acceptable level for clinical purposes. In retrospect, because of high variability in the data, a 95% level would prove

Fig 2. Variation in shape described by PC1. Middle shape in each row is the consensus shape. Left shape was constructed by setting PC1 at 3 SD below the average; right shape resulted by setting PC1 at 3 SD above the average. A, Shape space; B, form space.

American Journal of Orthodontics and Dentofacial Orthopedics Volume 136, Number 4

Chatzigianni and Halazonetis

481.e5

Table III. Object C1v C2v C3v C4v C2od C3bd C4bd

Comparison of natural logarithm of centroid size (ln[CS]) between the sexes for the 7 objects
Girls ln(CS): mean (SD) 4.461 (0.0548) 4.860 (0.0646) 4.391 (0.0605) 4.380 (0.0681) 4.021 (0.0753) 3.326 (0.1041) 3.319 (0.1028) Boys ln(CS): mean (SD) 4.533 (0.0652) 4.912 (0.0776) 4.443 (0.0677) 4.433 (0.0793) 4.067 (0.0822) 3.381 (0.1306) 3.385 (0.1257) Difference of CS 7.5% 5.3% 5.3% 5.5% 4.8% 5.6% 6.8% t test 5.91 3.62 3.97 3.57 2.89 2.29 2.83 P \0.0001 0.0005 0.0001 0.0006 \0.005 \0.05 \0.05

Table IV.

P values testing difference between the sexes for the 7 objects (permutation test using Procrustes distance [100,000 permutations])
Form space 0.0000 0.0002 0.0002 0.0017 0.0036 0.0074 0.0024 Shape space 0.0109 0.0759 (NS) 0.0294 0.4339 (NS) 0.0082 0.0107 0.0136

Object C1v C2v C3v C4v C2od C3bd C4bd

NS, Not signicant at the 5% level.

Fig 3. Regression plots of SML vs logarithm of CS for C3v and C3bd. Diamonds, Boys; circles, girls.

too strict, giving wide prediction intervals. To compare and evaluate the predictive value of chronologic age and CS, we also calculated the multiple regression of SML on the PCs and chronologic age, the regression of SML on chronologic age and logarithm of CS, and the regression of SML on chronologic age alone. In the female group, the width of the 90% PI for the regression of SML on age was 297 units. The shape of no object could produce a smaller PI, although C3v and

C2od came relatively close (Table V, column Shape space, PCs; and Fig 4). This implies that shape of the individual vertebrae could not give a better prediction of maturation level than chronologic age. By using the PCs of the bodies of C2, C3, and C4 together, corresponding to the qualitative CVM analysis, the calculated PI approached that of age, but still did not improve it.5,6 When chronologic age was added as a predictor to the shape variables, the prediction interval decreased, but the maximum improvement over age alone was small, leaving the PI above 250 units. In contrast to shape space, the form space PCs, which incorporate size as well, gave much better results; their PI was consistently approximately 100 units smaller than the shape space PI, and equal or better than the age PI. By adding age to the form space PCs, the PI was reduced to around 200 SML units, almost 100 units below the age PI. Slightly less predictive power was obtained by using CS and age, without incorporating shape information. The male group showed overall lower correlations, but the pattern was similar (Table V, Fig 4). The PCs in shape space gave the worst prediction intervals, most of them larger than the age PI. The best PIs were obtained for the combination of age and form space PCs, and these were approximately 100 units smaller than the age PI, as observed in the female group. The

481.e6

Chatzigianni and Halazonetis

American Journal of Orthodontics and Dentofacial Orthopedics October 2009

Table V.

Coefcient of determination (squared adjusted multiple regression correlation coefcient: r2) and width of 90% prediction interval (w90%PI) for multiple regression of SML on the meaningful PCs (Table II) of the 7 objects, in form and shape spaces
Shape space PCs r
2

Form space PCs and age PCs r

2

PCs and age r

2

ln(CS) and age r2 73.6% 77.8% 84.0% 82.2% 83.0% 78.3% 75.4% 82.6% 68.1% 79.2% 75.0% 73.6% 74.6% 73.9% 74.3% 75.0% w90%PI

w90%PI

w90%PI

w90%PI

w90%PI

Girls C1v C2v C3v C4v C2od C3bd C4bd C2od 1 C3bd 1 C4bd Boys{ C1v C2v C3v C4v C2od C3bd C4bd C2od 1 C3bd 1 C4bd

0.7% (NS) 27.6%* 59.4% 50.2% 57.6% 53.8% 47.0% 68.0% 28.1% 50.4% 72.8% 59.6% 45.0%* 55.0% 41.9% 75.0%

473 354 392 361 380 403 314 614 509 377 458 536 487 547 362

71.4% 77.6% 76.8% 77.7% 77.0% 75.8% 79.4% 64.6% 72.7% 76.4% 75.7% 72.2% 72.5% 68.5% 79.3%

297 263 267 262 266 272 253 432 378 352 356 381 376 403 330

28.8%* 63.3% 80.3% 74.4% 70.4% 73.2% 68.1% 80.2% 30.9%* 74.1% 80.5% 77.8% 63.7% 71.2% 68.5% 79.7%

470 337 247 282 303 287 313 248 605 370 322 343 437 387 404 398

74.2% 80.7% 88.8% 85.4% 84.8% 83.4% 80.6% 86.3% 67.1% 82.7% 82.2% 79.8% 75.7% 78.3% 77.3% 80.9%

283 244 187 213 217 226 245 207 418 303 307 327 358 336 343 321

284 261 221 234 229 258 275 231 405 327 358 368 361 366 363 359

Same data are shown for regression on PCs and chronologic age and for regression on natural logarithm of centroid size (ln[CS]) and chronologic age. PI was calculated for PCs equal to zero (consensus shape), age equal to average age of each group, and CS equal to average CS of each object. Results of linear regressions of SML and FMI on chronologic age are for comparisons. NS, Not signicant; *P \0.01; P \0.05; P \0.001. Female group (n 5 58): chronologic age vs SML: r2 5 71.7%, width of 90% prediction interval at age 11.96: 297 SML units; chronologic age vs FMI: r2 5 72.0%, width of 90% prediction interval at age 11.96: 5.2 FMI units; {Male group (n 5 40): chronologic age vs SML: r2 5 65.3%, width of 90% prediction interval at age 12.63: 427 SML units; chronologic age vs FMI: r2 5 59.2%, width of 90% prediction interval at age 12.63: 7.2 FMI units.

combination of size and age was slightly worse, but probably not clinically so.
DISCUSSION

The methodology of this investigation differed from previous research. Comprehensive and quantitative shape measurement was used, based on geometric morphometric methods.25-29,33 Densely placed points along the whole outline of each vertebra ensured that the biologic information inherent in vertebral outline shape was fully recorded. Therefore, the obtained correlations between vertebral shape and hand-wrist maturation should reect the total extent of the biological coordination between these 2 skeletal units. Each vertebra or vertebral body was independently evaluated, instead of using all vertebrae as a group. This gave information as to which vertebrae are the best predictors of maturation level and allowed potential clinical application by using a single vertebra (but not with equal accuracy among them).

Sexual dimorphism was found for both size and shape. Shape differences in form space were more pronounced, probably because form space incorporates size as well. CS was found to be larger in boys (Table III), even though the groups had comparable ages and, therefore, the girls were more advanced in development. When compared by maturation level, boys were even larger, by 10% or more (Fig 3). Hellsing34 found larger sizes for girls up to age 15, but most of these were not statistically signicant. In contrast, men had signicantly larger vertebrae. Larger sizes in males, as well as shape differences, were reported by Stemper et al,35 after adjusting for other variables that might inuence vertebral size, such as sitting height and head circumference. Similar results have been reported for lumbar vertebrae.36 These sex differences do not seem to be related to biomechanical factors of bone loading.37 Our statistical tests included correlations, reported here as coefcients of determination (r2), but emphasis was placed on the PI of the regression, because the main purpose was to evaluate the predictive power of

American Journal of Orthodontics and Dentofacial Orthopedics Volume 136, Number 4

Chatzigianni and Halazonetis

481.e7

Fig 4. Data from Table V. Width of 90% PI for the 7 objects plus the combination of the bodies of C2, C3, and C4 (smaller value signies better prediction). Age line corresponds to PI with chronologic age as the predictor. SS, Shape space; FS, form space; ln(CS), natural logarithm of CS.

vertebral shape on skeletal maturation.38 Additionally, to allow comparison, prediction intervals were calculated for size and age, as well as for various combinations of the independent variables. The female group showed the highest correlations and the most consistent results between the 7 objects examined. The PIs were smaller than the corresponding PIs of the male group by approximately 100 units. This is probably not due to the larger sample size of the female group, because PIs are not particularly sensitive to sample size but could be attributed to differences in the onset of puberty and the duration of the pubertal period between boys and girls.38 Other investigators have also reported higher correlations between vertebral shape and skeletal maturity in female subjects, but no denite explanation has been reached.1,39,40 Overall, shape alone (shape space PCs) showed a moderate to high correlation to SML, with r2 ranging from 27% to 75%. A wide range of correlations has been reported in the literature, but direct comparison is difcult because of the different methods of evaluating shape and skeletal maturity.1,40-42 Interestingly, as a predictor of skeletal maturation, vertebral shape (shape space PCs) was worse than chronologic age for all exam-

ined objects (even the combination of the bodies of C2, C3, and C4) in the girls and for most of the objects in the boys. This seems contrary both to the signicant body of literature that advocates cervical vertebrae maturation methods5,6 and to the assumed unreliability of chronologic age,43 which is considered not a valid predictor of skeletal growth velocity or skeletal maturity.44 However, not all data point to the same direction.11 Hunter45 concluded that skeletal and chronologic age were equally unreliable to predict the peak of pubertal growth spurt in girls, although skeletal age was better in boys. Uysal et al40 reported comparable correlations to hand-wrist maturation for both age and cervical vertebral development when the sexes were examined separately. It is unfortunate that most studies take the unreliability of age as an established fact; furthermore, they do not compute prediction intervals, for either for age. Haiter-Neto et al46 computed correlation coefcients between chronologic age and 3 methods of skeletal age estimation and reported r2 values from 71% to 86%. Soegiharto et al47 reported the r2 values and the 95% PIs for the regression of Fishmans maturation index on age, giving similar results to those we found for the girls, but not for the boys (Table V).

481.e8

Chatzigianni and Halazonetis

American Journal of Orthodontics and Dentofacial Orthopedics October 2009

The incorporation of CS in the shape measures (ie, the PCs of form space) was the factor that appreciably decreased the width of the PI by approximately 100 units. A comparable additional decrease was achieved by adding chronologic age to the predictor variables (Fig 4). Sorting the various factors based on predictive power, we get the following order, from best to worse: (1) form space PCs and age, (2) CS and age, (3) shape space PCs and age, (4) form space PCs, (5) age, and (6) shape space PCs. The combination of vertebral size and age provides the second-best predictive power, without directly incorporating shape information. An important nding is that even the best prediction interval was not smaller than approximately 200 SML units, which corresponds to approximately 1.3 years, if the patient is near the pubertal growth peak, and to a longer period if the patient is younger or older. This is not the PI for predicting the timing of the pubertal growth peak or the peak of facial growth, but for predicting the hand-wrist maturation level. Of all the objects examined, C1 gave the least predictive value. The biologic reason could not be determined from this study. Perhaps C1 is integrated more closely with the occipital bone than with the rest of the cervical vertebrae, and its ontogenetic trajectory is different. Further studies are needed in this area. The other objects were almost equally effective to each other (when form space PCs and age are the predictors), implying that any of them could be used. Assembling the 3 bodies of C2, C3, and C4 into 1 group of independent variables did not result in a signicantly smaller prediction interval, unless shape alone is considered. In that case, the shape of C3 gave very similar results. Thus, a further simplied version of the CVM method could be developed, by using C3 instead of all 3 vertebrae.5 However, our results show that even the current CVM method might not be more informative than simple chronologic age.5,6 On the other hand, regarding the morphologic factors that were evaluated, the CVM method seems to be supported by the ndings of this study. The PCA analysis showed 1 or 2 meaningful components of shape for the bodies of C3 and C4, signifying that most shape variability is concentrated along 1 axis, PC1. As Figure 2 shows, this axis describes the same morphologic variation that is assessed by the CVM method: the concavity of the lower border and the ratio of height to width. These same factors have, of course, been described by almost all investigators in the eld.3 Quantitative measurements of vertebral shape have been used before, but they were limited to height and width distances and ratios, and to the depth of the inferior concavity.5-7 For example, Chen et al5 used several

linear and angular measurements and arrived at a rstorder (linear) equation to predict 1 of 4 stages of growth velocity; these were concatenations of the 11 Fishman stages. Unfortunately, no prediction interval was reported, nor is the accuracy of the method compared with other evaluation systems, such as age.

CONCLUSIONS

Vertebral shape changes and hand-wrist ossication stages are both results of the ontogenetic developmental process, so their integration has been considered tighter than their relationship to chronologic age. However, the prediction of one from the other based solely on shape features was not better than using age; it could be improved if size and chronologic age are added to the independent variables. In orthodontics, hand-wrist skeletal maturation is an indirect way of estimating the peak of facial (maxillary and mandibular) growth and probably not very reliable.12,44 The direct relationship between vertebral size and shape and facial growth is of main interest, but research in this eld is limited because of the problems of longitudinal radiographic recordings. Perhaps reanalysis of available data along the lines presented here might provide new insights.4,48 We thank Philipp Gunz for helping with the sliding semilandmark procedure and Michael Coquerelle for validating the Viewbox code.
REFERENCES 1. Lamparski DG. Skeletal age assessment utilizing cervical vertebrae [thesis]. Pittsburgh: University of Pittsburgh; 1972. 2. OReilly MT, Yanniello GJ. Mandibular growth changes and maturation of cervical vertebrae. Angle Orthod 1988;58:179-84. 3. Hassel B, Farman AG. Skeletal maturation evaluation using cervical vertebrae. Am J Dentofacial Orthop 1995;107:58-66. 4. Franchi L, Baccetti T, McNamara JA Jr. Mandibular growth as related to cervical vertebral maturation and body height. Am J Orthod Dentofacial Orthop 2000;118:335-40. 5. Baccetti T, Franchi L, McNamara JA Jr. An improved version of the cervical vertebral maturation (CVM) method for the assessment of mandibular growth. Angle Orthod 2002;72: 316-23. 6. Baccetti T, Franchi L, McNamara JA Jr. The cervical vertebral maturation (CVM) method for the assessment of optimal treatment timing in dentofacial orthopedics. Semin Orthod 2005;11:119-29. 7. Caldas MP, Ambrosano GM, Haiter-Neto F. New formula to objectively evaluate skeletal maturation using lateral cephalometric radiographs. Braz Oral Res 2007;21:330-5. 8. Chen LL, Xu TM, Jiang JH, Zhang XZ, Lin JX. Quantitative cervical vertebral maturation assessment in adolescents with normal occlusion: a mixed longitudinal study. Am J Orthod Dentofacial Orthop 2008;134:720.e1-7. 9. Houston WJ, Miller JC, Tanner JM. Prediction of the timing of the adolescent growth spurt from ossication events in hand-wrist lms. Br J Orthod 1979;6:145-52.

American Journal of Orthodontics and Dentofacial Orthopedics Volume 136, Number 4

Chatzigianni and Halazonetis

481.e9

10. Houston WJ. Relationships between skeletal maturity estimated from hand-wrist radiographs and the timing of the adolescent growth spurt. Eur J Orthod 1980;2:81-93. 11. Smith RJ. Misuse of hand-wrist radiographs. Am J Orthod 1980; 77:75-8. 12. Turpin DL. British Orthodontic Society revises guidelines for clinical radiography. Am J Orthod Dentofacial Orthop 2008; 134:597-8. 13. Greulich WW, Pyle SI. Radiographic atlas of skeletal development of hand and wrist. 2nd ed. Stanford, Calif: Stanford University Press; 1959. 14. Tanner JM, Whitehouse RM. Atlas of childrens growth: Normal variation and growth disorders. 1st ed. London, United Kingdom: Academic Press; 1982. 15. Bjo rk A. Timing of interceptive orthodontic measures based on stages of maturation. Trans Eur Orthod Soc 1972;48:61-74. 16. Grave KC, Brown T. Skeletal ossication and the adolescent growth spurt. Am J Orthod 1976;69:611-9. 17. Ha gg U, Taranger J. Skeletal stages of the hand and wrist as indicators of the pubertal growth spurt. Acta Odontol Scand 1980;38: 187-200. 18. Ha gg U, Taranger J. Maturational indicators and the pubertal growth spurt. Am J Orthod 1982;82:299-309. 19. Fishman LS. Radiographic evaluation of skeletal maturation. A clinically oriented method based on hand-wrist lms. Angle Orthod 1982;52:88-112. 20. Fishman LS. Maturational patterns and prediction during adolescence. Angle Orthod 1987;57:178-93. M. La pe diatrie au quotidien. Analyse de la maturation 21. Sempe squelettique. Paris, France: Les Editions INSERM; 1987. Available at: http://www.auxologie.com/mediatheque/pdf/la_pediatrie_au_ quotidien.pdf. Accessed November 14, 2008. M. MATUROS. 4.0 CD: Un nouvel outil 22. Bouchard M, Sempe valuation de la maturation squelettique. Biom Hum et Anthrode pol 2001;19:9-12. 23. Gilsanz V, Ratib O. Hand bone age. A digital atlas of skeletal maturity. Berlin, Germany: Springer-Verlag; 2005. 24. Vastardis H, Evans AC. Evaluation of cervical spine abnormalities on cephalometric radiographs. Am J Orthod Dentofacial Orthop 1996;109:581-8. 25. Bookstein FL. Landmark methods for forms without landmarks: morphometrics of group differences in outline shape. Med Image Anal 1997;3:225-43. 26. Gunz P. Statistical and geometric reconstruction of hominid crania: reconstructing the ontogeny of Australopithecines [thesis]. Vienna, Austria: University of Vienna; 2005. 27. Gunz P, Mitteroecker P, Bookstein FL. Semilandmarks in 3D. In: Slice DE, editor. Modern morphometrics in physical anthropology. New York: Kluwer Academic, Plenum Publishers; 2005. 28. Mitteroecker P, Gunz P, Bernhard M, Schaefer K, Bookstein FL. Comparison of cranial ontogenetic trajectories among great apes and humans. J Hum Evol 2004;46:679-97. 29. Dryden IL, Mardia KV. Statistical shape analysis. Chichester, United Kingdom: John Wiley & Sons; 1998. 30. Houston WJB. The analysis of errors in orthodontic measurements. Am J Orthod 1983;83:382-9.

31. Jackson DA. Stopping rules in principal component analysis: a comparison of heuristical and statistical approaches. Ecology 1993;74:2204-14. 32. Peres-Neto PR, Jackson DA, Somers KM. How many principal components? Stopping rules for determining the number of nontrivial axes revisited. Comput Stat Data Anal 2005;49:974-97. 33. Halazonetis DJ. Morphometrics for cephalometric diagnosis. Am J Orthod Dentofacial Orthop 2004;125:571-81. 34. Hellsing E. Cervical vertebral dimensions in 8-, 11-, and 15-yearold children. Acta Odontol Scand 1991;49:207-13. 35. Stemper BD, Yoganandan N, Pintar FA, Maiman DJ, Meyer MA, DeRosia J, et al. Anatomical gender differences in cervical vertebrae of size-matched volunteers. Spine 2008;33:E44-9. 36. Gilsanz V, Boechat MI, Roe TF, Loro ML, Sayre JW, Goodman WG. Gender differences in vertebral body sizes in children and adolescents. Radiology 1994;190:673-7. 37. Gilsanz V, Kovanlikaya A, Costin G, Roe TF, Sayre J, Kaufman F. Differential effect of gender on the sizes of the bones in the axial and appendicular skeletons. J Clin Endocrinol Metab 1997;82: 1603-7. 38. Altman DG. Practical statistics for medical research. London: Chapman and Hall; 1991. n P, Palma JC, Oteo MD, Nevado E. Skeletal maturation 39. San Roma determined by cervical vertebrae development. Eur J Orthod 2002;24:303-11. 40. Uysal T, Ramoglu SI, Basciftci FA, Sari Z. Chronologic age and skeletal maturation of the cervical vertebrae and hand-wrist: is there a relationship? Am J Orthod Dentofacial Orthop 2006; 130:622-8. 41. Lai EH, Liu JP, Chang JZ, Tsai SJ, Yao CC, Chen MH, et al. Radiographic assessment of skeletal maturation stages for orthodontic patients: hand-wrist bones or cervical vertebrae? J Formos Med Assoc 2008;107:316-25. 42. Flores-Mir C, Burgess CA, Champney M, Jensen RJ, Pitcher MR, Major PW. Correlation of skeletal maturation stages determined by cervical vertebrae and hand-wrist evaluations. Angle Orthod 2006;76:1-5. 43. Alkhal HA, Wong RW, Rabie AB. Correlation between chronological age, cervical vertebral maturation and Fishmans skeletal maturity indicators in southern Chinese. Angle Orthod 2008;78: 591-6. 44. Flores-Mir C, Nebbe B, Major PW. Use of skeletal maturation based on hand-wrist radiographic analysis as a predictor of facial growth: a systematic review. Angle Orthod 2004;74:118-24. 45. Hunter CJ. The correlation of facial growth with body height and skeletal maturation at adolescence. Angle Orthod 1966;36:44-54. 46. Haiter-Neto F, Kurita LM, Menezes AV, Casanova MS. Skeletal age assessment: a comparison of 3 methods. Am J Orthod Dentofacial Orthop 2006;130:435.e15-20. 47. Soegiharto BM, Cunningham SJ, Moles DR. Skeletal maturation in Indonesian and white children assessed with hand-wrist and cervical vertebrae methods. Am J Orthod Dentofacial Orthop 2008;134:217-26. 48. Gu Y, McNamara JA. Mandibular growth changes and cervical vertebral maturation. A cephalometric implant study. Angle Orthod 2007;77:947-53.