Computer-Aided Skeletal Age Scores in Healthy Children, Girls with Turner Syndrome, and in Children with Constitutionally Tall Stature

Abstract

The manual Tanner-Whitehouse 2 method has recently been transformed into a computer-aided skeletal age scoring system (CASAS), which rates either the complete TW-RUS score (13b model) or a subset consisting of radius, ulna, and the four bones of the third finger (6b model). In this study the reliability of CASAS was evaluated in healthy children, and the 13b model was compared with the manual ratings and with the 6b model in (subgroups of) 151 healthy children, 87 girls with Turner syndrome, and 362 children with constitutionally tall stature. In addition, reference curves for bone maturation in Turner syndrome and constitutionally tall stature are presented. Some of mean differences in methods were statistically significant; however, because these mean differences were less than 0.4 bone age “year,” they are clinically not significant. In all comparisons the range of the difference between the methods (either with the 6b or the 13b model) was considerable, but the combined within- and between-components of variance(0.7%) were in the same order of magnitude as reported for the manual readings. In general, the percentage of equal stage ratings on duplicate assessments was high (±90%). Our data indicate that this computerized method is applicable in these groups of children. The use of the 6b model seems preferable because it is less time-consuming than the rating of 13 bones. In view of the percentages of manual insertions of a stage (up to 8% in all groups) the clinical use of this CASAS version (3.5) seems to be more efficient, particularly with longitudinal studies. Manual substitution of a stage should be avoided, and when performed its percentage and the limits for the acceptance of disagreement should be reported.

Main

It is common practice to determine the biologic maturity of children by means of an estimation of the BA from a radiograph of the hand and wrist. Tanner et al.(1) described maturity indicators for the epiphyses of each bone of the hand and wrist (TW2 method). Each bone progresses through a series of specific stages with attributed weighted scores. These scores are summed to form the MS, which in turn can be converted to a corresponding BA by means of a table.

Although the TW2 method is widely used, the reliability is limited(25). Furthermore, this method uses an interval scale and consequently a difference of one stage in the rating of a particular bone may result in an increase of 0.3 BA “years.” To diminish the errors in the interpretation of maturity stages and to improve the BA ratings, the TW2-(RUS) method has recently been transformed by the original author into a computerized image-analysis system using a continuous scale(6, 7). This system rates each of the 13 bones in the TW-RUS classification system (13b model). A shortened model, using a six-bone subset consisting of radius, ulna, and the four bones of the third finger, is also available (6b model).

This study evaluates 1) the reliability of CASAS in healthy children. 2) A comparison is made with manual ratings in healthy children, as well as in two groups of children with a particular growth condition, i.e. girls with TS and children with CTS. 3) To determine whether the easier 6b model could substitute the more laborious 13b model, the 13b model of CASAS is compared with the 6b model in these groups of children. 4) In addition, reference curves for bone maturation in TS and CTS as determined by CASAS are presented.

METHODS

Populations

Healthy children. Eighty-one boys and 70 girls with a CA of 2-22 y were recruited from the urban area of Rotterdam, The Netherlands. Children on medication or with conditions or disorders compromising growth and/or bone metabolism were excluded from the study.

TS. Eighty-seven girls with TS, aged 2-16 y, participated in this study before their enrollment in a drug intervention trial. Furthermore, 818 radiographs of the hand of 314 Dutch girls with TS, born between 1934 and 1992, were gathered from 11 pediatric centers in the Netherlands (see“Appendix”). The diagnosis was confirmed by chromosome analysis; approximately half of the girls had a 45,X karyotype. Women with conditions known to be related to TS and receiving adequate medical care were included. Three groups were detected: girls without spontaneous menarche or estrogen treatment (A, n = 224), girls with spontaneous menarche (B, n = 14), and those who received estrogen therapy (C,n = 76). About half of group C received 2.5-5.0 μg of ethinyl estradiol/d; some girls, however, received an ethinyl estradiol dosage of up to 50 μg/d.

CTS. One hundred and fifteen tall boys (aged 8.8-17.2 y) and 247 tall girls (aged 9.0-16.9 y) had visited our out-patient clinic for final height prediction. All children had heights at the 90th percentile or above. In total, 154 and 267 radiographs of the hand were gathered from boys and girls, respectively.

Study Designs

Reliability study in healthy children. To determine between- and within-observer variance of CASAS (13b model), two observers made duplicate assessments of 20 blinded radiographs, which were selected by taking every first healthy child in a CA class of 1 y.

CASAS in healthy children. In a subgroup of 40 children (20 boys and 20 girls) radiographs rated with the 13b model of CASAS were compared for assessment of agreement with the 6b model. The radiographs of all 151 children were rated manually and with CASAS (6b model). For each sex the ratings were compared for assessment of agreement.

CASAS in TS and CTS. Radiographs of 87 girls with TS were rated manually and with the 6b model of CASAS; a subgroup of 31 TS girls were rated with the 13b model and compared with the 6b model. Four hundred twenty-one x-rays of all children with CTS were rated manually and with CASAS (6b model). In a subgroup of 40 children (20 boys and 20 girls), radiographs rated by the 13b model of CASAS were compared with those rated by the 6b model. Ratings were compared for assessment of agreement.

Bone maturation curves in TS and CTS. In every yearly CA interval (rounded to the nearest integer year), only one x-ray of the hand from each patient was used, leaving 749 x-rays of girls with TS and 411 x-rays of all children with CTS (146 and 265 x-rays for boys and girls, respectively) for analysis with CASAS (6b model). Separate regression equations were estimated for group A and the combined groups B and C in TS, and for boys and girls in CTS.

All protocols were approved by the Medical Ethical Committee of the Erasmus University Medical School and the Academic Hospitals involved in these studies, and informed consent was obtained from the participants or the guardians.

Radiographs and Bone Ratings

All radiographs had been taken by standard radiologic techniques (TW2), and identification details were removed. Radiographs were rated manually (TW2 RUS method)(1) and by a computerized image analysis system(CASAS, version 3.5)(6, 7). In combination with the radius and ulna, the TW2-RUS score uses only three rays-each comprising a metacarpal and the accompanying phalanges-to avoid excessive weighting on the finger bones (13 bones-score). Apart from a 13 bones rating (13b), CASAS has a short model which analyses only six bones (6b; radius, ulna, and the bones of the third ray) to form a complete TW2-RUS score. The grades (A-I) from the third ray are then extrapolated to the first and fifth rays, and the matching maturity scores are calculated based on the original TW2 tables for that grade. If a grade allocated by CASAS does not conform to the expected grade of the observer, it can be inserted manually. In our studies a manual insertion of a stage was allowed only after three attempts with a difference between expected and determined stage of two integer stages or more.

For each design comparing methods two different observers performed the readings, except for both the manual versus 13b and manualversus manual designs in TS, which were rated by one observer. Hence, intermethod and interobserver variability may be inextricably compounded, as is often the real situation in practice.

Statistical Analysis

Data are expressed as MS or as RUS-BA years. To assess the degree of agreement between manual and CASAS ratings of maturity score (and subsequently BA), we used the approach described by Bland and Altman(8). In short, this method uses simple calculations and graphical techniques instead of correlation coefficients to describe the degree of agreement. Plots of the difference between methods against their average are given to illustrate the range of agreement by level.

Within and between observer variance of CASAS (13b model) was determined with an analysis of variance using the BMDP statistical package, module 8V. All other comparisons between methods were performed by paired t tests. Differences between sexes were tested with the Mann-Whitney U test. p values < 0.05 were considered significant.

The regression equations for the TS and CTS reference groups were calculated using a repeated measures analysis of variance (BMDP, module 5V). The within-subject covariance matrix was assumed to have a first order autoregressive structure after reordering the data on an equidistant time axis.

RESULTS

Reliability study on 13-bone model. The estimated total component of between-observer, within-observer, and observer/radiograph interaction variance was, for both CASAS 13b and the manual ratings, only 0.7% of the total variance, leaving 99.3% for the between-radiograph component of variance. The SD of the within-observer component of variance using CASAS was 0.36 BA years, for the manual ratings 0.25 BA years.

On average, a single observer using CASAS 13b gave the same stage rating(rounded to the integer stage) on two occasions in 88% of instances. For the manual readings the equal staging within observers was 90%. CASAS 13b showed most inconsistencies with regard to equal staging for the fifth metacarpal and proximal phalanx, and the fifth distal phalanx. Using CASAS 13b the average percentage of equal stage ratings between two observers was 88%; with the manual ratings, 89%.

Most inconsistencies between observers were noted for the first proximal and distal phalanx. The CASAS 13b ratings resulted occasionally in a difference of two stages (1%); the manual ratings showed differences only between adjacent stages.

The percentage of manual insertions per radiograph was 8.1 and 5.6% for the first and second rating, respectively, which is equivalent to about one bone in every radiograph. In 89% of instances, the same bone was manually inserted on both occasions. In the first rating period most manual insertions were performed for the radius, ulna, first metacarpal, and the distal phalanx of the fifth finger. In the second period only the first three bones accounted for most of the manual insertions. The bones which showed inconsistency with regard to manual insertions on either of the two rating sessions were the radius, ulna, the first metacarpal, and the distal phalanx of the fifth finger.

We then excluded x-rays of five fully matured children (manual MS = 1000), because of the excess of extreme stages (H or I) which are more difficult to rate by CASAS (see “Discussion”). The percentages of manual insertions decreased only in the first rating period with 1.7%, mainly due to the radius and the distal phalanx of the fifth finger.

CASAS in healthy children. In a subgroup of 40 children (20 boys and 20 girls) the correlation coefficient between the 13b and 6b CASAS model was 0.99 in MS as well as in BA years. The mean difference between the models was significantly different only in boys and only when expressed as MS(p = 0.002). This also resulted in a significantly smaller mean difference in MS in girls compared with boys (p = 0.03), seeTable 1.

Table 1 Mean difference and 95% limits of agreement between manual, 6b, and 13b CASAS ratings expressed in MS and in BA years from x-rays of healthy children, girls with TS, and children with CTS

In the x-rays of all 151 healthy children, the correlation coefficient between the manual and the CASAS 6b ratings was also 0.99 (p < 0.0001). The mean difference between the two rating methods in MS was only significantly different in girls: on average, manual ratings were 17.3 points higher than CASAS ratings (or 0.15 BA years); 95% of the differences in MS in girls are expected to lie between -88.2 and +122.8 (or between -1.08 and +1.38 BA years). Because the mean difference between the methods was also significantly higher in girls compared with boys (p = 0.003), values for boys and girls are given separately (see Table 1 and Fig. 1,A andB.

Figure 1
figure1

Plot of the difference between manual and CASAS 6b ratings against the average of these methods expressed as MS in healthy boys(A) and in healthy girls (B); mean difference (-), limits of agreement (-----).

The percentage of manual insertions using CASAS 6b was 5.1%; the radius and ulna were manually inserted most frequently (74% of all manual insertions).

CASAS in TS and CTS. In girls with TS, four separate comparisons between manual ratings and CASAS were performed (seeTable 1). In each comparison the correlation coefficient between the methods was very strong (r ≥ 0.97; p < 0.01). On average, the CASAS 6b ratings were significantly higher than the manual ratings (p < 0.0001), see Figure 2A.

Figure 2
figure2

Plot of the difference between manual and CASAS 6b ratings against the average of these methods expressed as MS in girls with TS(A), and in boys (B) and girls (C) with CTS; mean difference (-), limits of agreement (-----).

The percentage of equal stages on two occasions by a single observer was similar for the manual rating and CASAS 6b, 73 and 72%, respectively. Manually, the proximal phalanges were rated markedly better than the radius and the ulna, 90 and 56%, respectively. Using CASAS 6b there were hardly any differences between the stages of the six bones determined on two occasions. The percentage of equal stage ratings between two observers was 73%. The percentage of manual insertions was similar for two observers, about 8%. Most of the six bones had manual insertions on both occasions; however, the ulna showed a marked inconsistency: in about 20% of instances only on one manual insertion.

In a subgroup of 40 CTS children (20 boys and 20 girls) the correlation coefficient between the 13b and 6b model of CASAS was 0.96 in MS as well as in BA years. The mean difference between the models expressed as MS and in BA years was not significant for both sexes (Table 1). When the manual ratings were compared with CASAS 6b in all children with CTS, the mean difference between the methods was significantly different in boys, but not girls. In addition, the mean difference between the methods was smaller in boys compared with girls (p < 0.0001), seeFigure 2,B and2C.

The percentage of manual insertions using the 6b model (all CTS children) was low, 0.2% in boys and 1.9% in girls. Radius, ulna, and the third metacarpal were most frequently manually inserted (81% of total manual insertions). In the 40 radiographs rated using the 13b model of CASAS the percentage of manual insertions was 7.7% for boys and 13.8% for girls. The difference in percentage of manual insertions between the 13b and 6b model was explained mainly by insertions of the first and fifth ray. Using the 13b model, manual insertions of the first and fifth ray accounted for 60 and 86% of all manual insertions in boys and girls, respectively.

Bone maturation curves in TS and in CTS.Figure 3 depicts the CASAS 6b ratings versus CA in TS. A regression analysis of BA (dependent variable) on CA and the presence of estrogens (independent variables) in 749 radiographs of 314 girls with TS yielded the following results:

Figure 3
figure3

Plot of BA determined using CASAS 6b vs CA in girls with TS. Separate regression lines for group A (no estrogens, ,-----) and the combined groups B and C (with estrogens, , , ---) are depicted in comparison with the assumed y = x line for BA and CA (-).

Without estrogens (group A): Equation The difference in regression lines between groups A and the combined groups B and C was close to the 0.05 level of significance (likelihood ratio test p = 0.0694).

The first derivative of BA with respect to CA on these regression equations gives the biologic maturation which has passed during a full CA year (dBA/dCA in BA year/year):

Without estrogens (group A): Equation

Figure 4,A and b, depict the CASAS 6b ratings versus CA in CTS for boys and girls, respectively. The estimated regression models are given by the following equations:

Figure 4
figure4

Plot of BA determined using CASAS 6b vs CA in boys (A) and girls (B) with CTS. The regression line(-----) is depicted in comparison with the assumed y = x line for BA and CA (-).

For boys: Equation

For girls: Equation

DISCUSSION

CASAS in healthy children. The reliability study of CASAS 13b showed very small estimated between- and within-observer components of variance. However, when expressed in BA years the range seems considerable and appears even slightly higher compared with the manual ratings. In an earlier reliability study with the traditional manual x-ray ratings of healthy children from 7 to 17 y of age(3) the SD of the within-observer component of variance of two observers was similar (0.30-0.35 BA years) compared with those in our manual ratings (0.25 BA years) and with CASAS (0.36 BA years). It has been stated, though, that the reliability varies somewhat in different parts of the age span; as a result of a change of one stage of a single bone the BA may “jump” by 0.3 BA years(1). The interval scale in the manual method has been replaced by a continuous scale in CASAS and, therefore, reliability should be better and equal at all ages. A further improvement would be the use of MS instead of BA. The correlation between biologic maturation and CA varies between populations(9). Therefore, ideally every population should have its own standards for MS and BA. In fact, BA is a crude determination derived from the MS table.

A comparison of the staging in duplicate assessments by the same observers(within-observer comparison) showed similar percentages for CASAS 13b (88%) and the manual ratings (90%). Ten percent of the ratings differed by one stage, and using CASAS, occasionally a difference of two stages was observed. Noticeably, there were hardly any differences between the individual bones included in the 13b model of CASAS. Our percentage of different stages within observers using CASAS 13b is higher than reported by Tanner and Gibbons(5) (2-5%); however, in their study the CASAS 6b model was used. Using the manual method, Beunen and Cameron(3) also found that a single observer gave the same stage rating on two occasions in about 90% of the cases (within-observer comparison), whereas different observers gave the same stage in 75-85% of the cases. In our study, the latter(between-observer comparison) percentage was similar for both CASAS and the manual ratings, 88 and 89%, respectively. Again, using CASAS, occasional differences of two stages were observed.

As a part of the CASAS grading process each bone is compared with reference standards which represent each stage of bone maturity. The best fitting stage together with the four adjacent stages were compared with the standard by means of their root mean square errors. The extreme stages of each bone (A and H/I) do not have two adjacent stages on both sides. Although the analysis of covariance is adjusted for these extreme stages (R. D. Gibbons, personal communication), the extreme computer-rated stage scores, i.e. 1.0 and 8.0/9.0, are not always attained. To illustrate this, when all epiphyseal plates are visually closed and CASAS rates the stages with a deficit of minus 0.15, the difference with the expected MS of 1000 can amount to about 44 MS in boys and 33 MS in girls, i.e. 0.85 and 0.5 BA years, respectively. A comparable deficit of 0.15 of a grade to stage A results in a difference of about 15 MS in boys and girls; at this very young CA also a substantial difference. It seems less important, however, because skeletal maturity is less often determined in the first years of life.

The percentage of manual insertions was slightly lower in the second rating period compared with the first period. The data are in good agreement with the CASAS 6b model ratings performed with x-rays from all healthy children (second study), in which 5% manual insertions were performed. There was a high degree of consistency for each bone between the two rating sessions. When a smaller range was chosen for manual insertions, i.e. a difference between expected and determined stage of one integer stage or more, the percentage of manual insertions on the same x-ray series rated by the same observers increased to 11-12%, but the between- and within-observer variances did not show a marked change (data not shown).

The possibility of inserting a manual score with CASAS creates the problem of when to make use of this option. To what extent a difference between the expected and determined rating will be accepted is likely to depend on the experience of the user and on the reliability of the computer rating of the individual bones. In addition, the magnification of the individual bones using CASAS may induce distinct differences in expected ratings compared with manual ratings usually performed without magnification. Therefore, cut-off limits for such a manual expected rating remain very subjective and particularly when one keeps the evident within-observer variability of manual ratings in mind. Ideally, one should not give a manual insertion at all, and the computer should analyze the radiograph fully automatically. However, because CASAS is based on the same classification of developmental stages of each bone, one cannot omit to give a manual insertion whenever the computer rating is far from logical (e.g. stage B instead of G) due to, e.g. positioning, imaging problems, or software imperfectness. Consequently, CASAS is not yet completely independent of the observer. In addition, in our experience, individual CASAS ratings may sometimes vary considerably just by repositioning of the x-ray and without giving a warning to the user. The designers of the program underlined this by stating that “correct assessment depends crucially on correct positioning”(5). The choice of acceptance of a single bone rating by the user is also dependent on the subjectiveness (experience) of the user. Therefore, to compare study results we consider that the percentage of manual insertions and the limits for the acceptance of disagreement must be reported, at least until the computer operating system is considered to be the gold standard.

The designers of CASAS suggest that prior knowledge is not needed; however, one should familiarize oneself with the system, and some knowledge on the developmental stages of the various bones is needed to avoid evident mistakes in the rating. Although this is seldom due to the quality of the x-ray, it is obvious that the TW2 radiologic technique(1) should be used, in particular with regard to the positioning of the hand.

The use of the 6b model seems preferable because it is less time consuming than the rating of 13 bones, 5-6 min versus 10 min or more, respectively. Although experienced TW2 “raters” might be able to perform the manual procedure quicker, it must be noted that, after rating the stages, data entry in the computer and subsequent calculations still has to follow, in contrast to CASAS.

The mean difference between the 13b and the 6b model of CASAS in healthy children also showed a significant gender difference only when expressed as MS; however, 95% limits of agreement were again wide.

The mean difference between the manual ratings and CASAS in healthy children was significantly different only in girls, although the 95% limits of agreement were considerable. When x-rays of fully matured children were excluded from the analysis, the 95% limits of agreement were only slightly“narrower.” The mean difference between methods was significantly different between boys and girls. A comparable difference between the sexes was found in CTS children. We have no satisfying explanation for this phenomenon.

CASAS in TS and CTS. In TS, the mean difference between the manual and the 13b CASAS ratings was close to zero. However, there was a small difference between the manual ratings and CASAS 6b as well as between the duplicate manual ratings. Again, all comparisons-including the duplicate manual assessments-showed fairly large 95% limits of agreement.

The percentage of equal stages on duplicate assessments using the 6b model was lower in these TS girls than in healthy children, but identical to that using the manual rating method in girls with TS. The number of manual insertions was slightly higher (8%) compared with healthy children (5%). In view of the anatomical and structural abnormalities of the hand-wrist bones in these girls, this difference seems acceptable.

On average, ratings of the 13b model were comparable with those of the 6b model in both boys and girls with CTS, although individual differences may be considerable. In girls with CTS the mean difference between the manual ratings and CASAS was also almost zero. In boys, ratings with CASAS were significantly higher compared with the manual ratings. These mean differences between methods were also significantly different between the sexes. For both sexes, however, the 95% limits of agreement were of the same magnitude and comparable with that in healthy children.

The percentage of manual insertions in the 6b model of CASAS was remarkably low in CTS. Using the 13b model the percentage of manual insertions increased rapidly; it appeared that most of the manual insertions were due to the first and fifth rays, which are not included in the short 6b model. The use of the 6b model will thus reduce the number of difficulties in rating. This might explain the slight, but not significant, difference in BA compared with that of the 13b model.

Bone maturation curves in TS and CTS. We evaluated the development of bone maturation in TS using CASAS. Our results are comparable with the results described by Ranke et al.(10) using the manual TW2 method (400 observations), and by Brook et al.(11) and Rochiccioli et al.(12) using the Greulich and Pyle(13) method. The last named method is known to have a systematic diminution of BA progression compared with TW2-RUS BA(10).

The BA progression seems to decline until the CA of 3 y. Alternatively the BA is possibly already retarded at birth. It is of interest that in our study population there were seven girls with BA “0” until the CA of 1.5 y. From the CA of 3 y the rate of BA progression is about 1 BA year/year. Bone maturation reaches its peak (1.15 BA year/year) around the CA of 6-7 y. Thereafter, the dBA/dCA declines, resulting in a progressive difference between BA and CA from the CA of 10-11 y, as opposed to 12 y found in the earlier studies(1012). At the CA of 20 y, epiphyseal closure was reached in many of the girls with spontaneous menarche and/or estrogen treatment in contrast to those girls without this estrogen influence. Although the girls of the former group appeared to have a faster progression of BA than the girls without the estrogen influence, the difference between these groups was not significant. The x-ray sampling period spans half a century; however, a time-dependent effect for the bone maturation in girls with TS does not seem to be relevant, because separate regression analyses for women born before and after 1974 (i.e. the median year of birth) revealed similar curves (data not shown).

Our mixed cross-sectional and longitudinal data of bone age in children with CTS show that BA is advanced at younger ages in boys as well as in girls. This discrepancy weakens as CA progresses. To our knowledge, there are no cross-sectional or longitudinal data on bone maturation in children with CTS. However, our findings are in agreement with our own clinical experiences. In addition, in various clinical studies the mean manual TW2 BA before the start of treatment was always higher compared with the mean CA of the patients, whereas the SD of the BA was always smaller than the SD of the CA(1418). This seems to be in concert with our present findings.

Conclusions. In healthy girls, girls with TS, and boys with CTS, the semiautomated CASAS 6b ratings were significantly different from manual ratings. The difference between the 13b and 6b ratings was significant only in healthy boys and in girls with TS. In all comparisons the mean differences between methods (either with the 6b or the 13b model) were less than 0.4 BA year, the range of the difference between methods was considerable, but the within- and between-observer variations are in the same order of magnitude as reported for the manual readings. It should be noted that both methods only estimate BA; a measure of the “true” BA is unknown. Therefore, we think that this computerized method is applicable in these groups of children. The use of the 6b model seems preferable because it is less time consuming than rating 13 bones.

In view of the percentages of manual insertions the clinical use of this CASAS version seems to be with longitudinal studies of patients in particular. Manual substitution of a stage should be avoided and when performed its percentage and the limits for the acceptance of disagreement should be reported.

Our data indicate that CASAS is applicable in girls with TS and children with CTS. The bone maturation curve in TS indicates that from the CA of 9 y the CASAS BA ratings tend to deviate progressively from the theoreticaly = x line for BA and CA in healthy children. Estrogen treatment and endogenous estrogens result in a (statistically not significant) faster closure of the epiphyseal growth plates. The bone maturation curves in boys and girls with CTS tend to be advanced at younger ages; however, as CA progresses this discrepancy diminishes.

Abbreviations

CASAS:

computer-aided skeletal age scoring system

13b and 6b:

13- and 6-bone rating models, respectively, of CASAS

CA:

chronologic age

BA:

bone age

TW2:

Tanner and Whitehouse 2 method of bone age determinations

MS:

maturity score according to Tanner and Whitehouse 2 method

TS:

Turner syndrome

CTS:

constitutionally tall stature

REFERENCES

  1. 1

    Tanner JM, Whitehouse RH, Cameron N, Marshall WA, Healy MJR, Goldstein H 1983 Assessment of Skeletal Maturity and Prediction of Adult Height (TW2-Method). Academic Press, London

  2. 2

    Taranger J, Bruning B, Claesson I, Karlberg P, Landstrom B 1976 A new method for the assessment of skeletal maturity-the MAT method (mean appearance time of bone stages). Acta Paediatr Scand Suppl 258: 109–120

  3. 3

    Beunen G, Cameron N 1980 The reproducibility of TW2 skeletal age assessments by a self-taught assessor. Hum Biol 7: 155–162

  4. 4

    Wenzel A, Melsen B 1982 Replicability of assessing radiographs by the Tanner Whitehouse-2 method. Hum Biol 54: 575–581

  5. 5

    Tanner JM, Gibbons RD 1994 Automatic bone age measurement using computerized image analysis. J Pediatr Endocrinol 7: 141–145

  6. 6

    Tanner JM, Gibbons RG, Bock DR 1992 An image analysis system for TW2 skeletal maturity. Horm Res 37( suppl 4): 11.

  7. 7

    Tanner JM, Gibbons RD, Bock RD 1993 advantages of the computer-aided image analysis system for estimating tw skeletal maturity: increased reliability and a continuous scale. Pediatr Res 33( suppl): s35.

  8. 8

    Bland JM, Altman DG 1986 Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1: 307–310

  9. 9

    Eveleth PB, Tanner JM 1990 Worldwide Variation in Human Growth. University Press, Cambridge

  10. 10

    Ranke MB, Chavez-Meyer H, Blank B, Frisch H, Häusler G 1991 Spontaneous growth and bone age development in Turner syndrome: results of a multicentric study 1990. In: Ranke MB, Rosenfeld R(eds) Turner Syndrome: Growth Promoting Therapies. Elsevier Science Publishers B.V., Amsterdam, pp 101–106

  11. 11

    Brook CGD, Murset G, Zachmann M, Prader A 1974 Growth in children with 45,XO Turner's syndrome. Arch Dis Child 49: 789–795

  12. 12

    Rochiccioli P, Pienkowski C, Tauber MT 1991 Spontaneous growth in Turner syndrome: a study of 61 cases. In: Ranke MB, Rosenfeld R(eds) Turner Syndrome: Growth Promoting Therapies. Elsevier Science Publishers B.V., Amsterdam, 107–112

  13. 13

    Greulich WW, Pyle SJ 1959 Radiographic Atlas of Skeletal Development of the Hand and Wrist. Stanford University Press, California

  14. 14

    Zachmann M, Ferrandez A, Murset G, Prader A 1975 Estrogen treatment of excessively tall girls. Helv Paediatr Acta 30: 11–30

  15. 15

    Zachmann M, Ferrandez A, Murset G, Gnehm HE, Prader A 1976 Testosterone treatment of excessively tall boys. J Pediatr 88: 116–123

  16. 16

    John G, Schellong G 1980 Oestrogen therapie hochwuchsiger Madchen. Monatsschr Kinderheilkd 128: 545–550

  17. 17

    Stover B, Kollmann F 1983 Ergebnisse der Behandlung hochwuchsiger Madchen mit conjugierten Oestrogenen. Monatsschr Kinderheilkd 130: 36–40

  18. 18

    Brämswig JH, Hermeling F, Petrykowski W, Schellong WV 1984 Comparison of height predictions with final adult height in boys with constitutional tall stature. In: Borms J, Hauspie R, Sand A, Susanne C, Hebbelinck M (eds) Human Growth and Development. Plenum, New York, 423–429

Download references

Acknowledgements

The authors acknowledge the kind co-operation of Dr. Roland Aarsen for referring the CTS patients. We are indebted to Ares Services S.A. and Discerning Systems for supplying a CASAS system and Dr. Berthon Rikken for the development of additional software applications.

Author information

Correspondence to A van Teunenbroek.

Additional information

Manuscript dedicated to Professor H.K.A. Visser in honor of his retiremen

Supported by Novo Nordisk Pharma B.V.

Appendix

Appendix

Appendix. The members of the Dutch Working Group on Growth Hormone who participated in this study are:

  • M. Gons and T. Vulsma, Amsterdam

  • H. Delemarre-Van de Waal and G. Hoorweg, Amsterdam

  • J. Waelkens, Eindhoven

  • N. Drayer, Groningen

  • H. Reeser, Den Haag

  • J. M. Wit and W. Oostdijk, Leiden

  • J. Gosen, Leiderdorp

  • W. J. Gerver, Maastricht

  • B. Otten and C. Rongen-Westerlaken, Nijmegen

  • S. de Muinck Keizer-Schrama and S. Drop, Rotterdam

  • M. Jansen, Utrecht

Rights and permissions

Reprints and Permissions

About this article

Cite this article

van Teunenbroek, A., de Waal, W., Roks, A. et al. Computer-Aided Skeletal Age Scores in Healthy Children, Girls with Turner Syndrome, and in Children with Constitutionally Tall Stature. Pediatr Res 39, 360–367 (1996). https://doi.org/10.1203/00006450-199602000-00028

Download citation

Further reading