Introduction

Trisomy 21 is the most frequent chromosomal abnor-mality observed at birth. Because it is associated withsevere mental handicap and long survival, different pol-icies for prenatal diagnosis have been developed to de-tect pregnant women at risk for this syndrome. Thebest known risk factor is increased maternal age and, inFrance, fetal karyotype analysis is routinely offered topregnant women over 38 years of age. However, thisprocedure detects only a minority of cases, most affect-ed infants being born to younger women.

Other methods have therefore been proposed thatcan be applied on a larger scale. Maternal biologicaltests have recently been introduced. The triple test,based on beta-hCG, unconjugated oestriol, and serumalpha-fetoprotein levels in a very limited period, is areasonably reliable test if undertaken between 15 and17 weeks' gestation (WG), although with false-positiveand false-negative results [1].

Fetal US screening can be employed at any period ofgestation, but is particularly useful in the first trimesteror for late or missed diagnoses. Several studies have de-scribed various abnormal features, mainly skeletal, as-

Noelle StempfleYolene HutenCatherine FredouilleHerve BrisseCatherine Nessmann

Skeletal abnormalities in fetuses withDown's syndrome:a radiographic post-mortem study

Received: 19 May 1998Accepted: 19 February 1999

N. Stempfle ´ H. BrisseDepartment of Radiology,R. DebrØ Hospital, Paris, France

Y. Huten ´ C. Fredouille ´C. Nessmann ())Department of Developmental Biology,R. DebrØ Hospital, 48 boulevard SØrurier,F-75 019 Paris, France

Abstract Objective. To evaluateskeletal abnormalities on post-mor-tem radiographs of fetuses withDown's syndrome.Materials and methods. Biometricaland morphological criteria, whichare used for US prenatal detectionof trisomy 21, were assessed. Limblong bones, biparietal diameter(BPD)/occipito-frontal diameter(OFD) ratio, ossification of nasalbones and appearance of the middlephalanx of the fifth digit (P2) in 60fetuses with Down's syndrome wereanalysed and compared with 82 nor-mal fetuses matched for gestationalage (GA) from 15 to 40 weeks' ges-tation (WG).Results. We observed reducedgrowth velocity of limb long bonesduring the third trimester in bothgroups, but the reduction was morepronounced in the trisomic group.Brachycephaly was found as early as15 WG in Down's syndrome and

continued throughout gestation(sensitivity 0.28, specificity 1). Ossi-fication of the nasal bones, whichcan be detected in normal fetusesfrom 14 WG, was absent in onequarter of trisomic fetuses, regard-less of GA. The middle phalanx ofthe fifth digit was evaluated bycomparison with the distal phalanx(P3) of the same digit. We foundthat P2 was not ossified in 11/31 tri-somic fetuses before 23 WG, andwas either not ossified or hypoplas-tic in 17/29 cases after 24 WG (sen-sitivity 0.56, specificity 1).Conclusions. Three key skeletalsigns were present in trisomic fetus-es: brachycephaly, absence of nasalbone ossification, and hypoplasia ofthe middle phalanx of the fifth digit.All these signs are appropriate toprenatal US screening. When pre-sent, they fully justify determinationof the fetal karyotype by amniocen-tesis.

Pediatr Radiol (1999) 29: 682±688Ó Springer-Verlag 1999

sociated with trisomy 21 [2±10]. However, these obser-vations are somewhat controversial.

Consequently, we undertook a detailed analysis ofpost-mortem skeletal radiographs of 60 fetuses withDown's syndrome and compared them with a controlseries of 82 normal fetuses. By evaluating the skeletalbiometry and morphology between 15 and 40 WG, wehave reassessed the criteria which are used as thescreening tool for prenatal detection of trisomy 21,whatever the gestational age (GA).

Materials and methods

We analysed the skeleton on post-mortem radiographs of 60 fetus-es with Down's syndrome, aged from 15 to 40 WG. Medical termi-nation of pregnancy was based on fetal chromosomal analysisindicated by maternal age, biological screening, or a previous childwith trisomy 21. In 40 cases, US was not performed. In 20 cases, fe-tal karyotype was undertaken because of various US anomalies, ofwhich only 9 were related to the skeleton (7 short femora, 1brachymesophalangy, and 3 short nasal bones), either singly or incombination. These 60 trisomic fetuses were compared with 82normal fetuses matched for GA. Fetuses in the control groupwere selected using the following criteria for normality: (1) agree-ment between GA, morphological data (weight, height, footlength), external brain morphology, and histological status whichwas used as the 'gold standard' for fetal age; (2) absence of externalor internal malformation; (3) well-documented acute placental orumbilical cord cause of fetal death. In this population, we studiedvarious skeletal areas using quantitative and qualitative criteria inorder to establish radiographic scores for fetal bone age [11]. Inthe present study, we analysed in the same population additionalbiometrical and morphological skeletal data in order to comparethem to the trisomic group.

Radiographs were taken with a Faxitron (Hewlett-Packard)which produced radiographs without significant magnification.Kilovoltage and exposure time were adjusted using a previously es-tablished table according to fetal dimensions. Screenless X-OmatReady Pack films (Kodak) were used. The fetus was fixed with ad-hesive bandage and laid over the film. Frontal and lateral views ofthe whole fetus were taken. For the frontal view limbs were ex-tended and for the lateral view they were side by side and half ex-tended.

Biometrical and morphological evaluation

In both groups the following biometrical and morphological skele-tal evaluations were performed on each radiograph independentlyby two observers: measurements of the diaphysial length of alllimb long bones (LB) [1±6], the biparietal and occipito-frontal di-ameters (BPD and OFD, respectively), and the nasal bone length(NBL). The ratio BPD/OFD was calculated. In addition, the mor-phological appearance of the middle phalanx (P2) of the fifth digitand its length were compared to those of the terminal phalanx (P3)of the same digit between 15 and 40 WG.

Statistical analysis

In both groups, the following statistical analyses were undertaken:1. A second-order polynomial regression analysis to evaluate the

relationship between all measurements (LB, BPD, OFD, NBL)and GA

2. A linear-regression analysis to evaluate the relationship betweenthe BPD/OFD ratio and GA

3. A compared mean test of the BPD/OFD ratio between normaland trisomic 21 fetuses

4. A Receiver Operating Characteristic (ROC) curve for the BPD/OFD ratio in the normal group to define brachycephaly

5. Chi-square tests to evaluate the relationship between trisomy 21and absence of nasal-bone ossification and P2 hypoplasia

Results

Limb long-bone length

There was good correlation in both groups between thegrowth of the limb long bones and GA (Table 1). Inboth groups, we observed a slowing of growth of alllimb bones during the last trimester, but this phenome-non was more pronounced in the trisomic group, exceptfor the fibula. The growth curves for the femur and thetibia are given as examples (Figs. 1, 2). The small sizeof our series for each age did not allow the establish-ment of standard values for normal or pathologicalcases according to GA.

BPD/OFD ratio

There was no significant variation of the BPD/OFD ra-tio with GA in either group (Fig.3). This ratio was near-ly constant between 15 and 40 WG. The means of thetwo groups was greater in trisomic fetuses: 0.879 (d:0.054) vs 0.797 (d: 0.066), P < 0.001, whatever the GA.We established a ROC curve for this ratio in the normalgroup with a confidence interval of 5% (Fig. 4). Withthis curve, it was possible to choose arbitrarily a thresh-old value that defined brachycephaly (0.872) accordingto the choice of sensitivity (0.54) and specificity (0.91).Figure 5 illustrates the cranial morphology in a controland a trisomic fetus of the same GA (20 WG).

683

Table 1 Correlation study results between the limb long bonesand gestational age in both groups

Control group Trisomic 21 group

P P

Humerus 0.969 0.0001 0.916 0.0001Cubitus 0.958 0.0001 0.872 0.0001Radius 0.960 0.0001 0.893 0.0001Femur 0.980 0.0001 0.908 0.0001Tibia 0.964 0.0001 0.898 0.0001Fibula 0.893 0.0001 0.907 0.0001

Nasal bones

In normal fetuses, ossification of the nasal bones was al-ways present from 15 to 40 WG. By contrast, there wasno bone ossification in 14 of 60 (23 %) trisomic fetuses,whatever the age (Fig.5). The chi-square test showed astatistically significant relationship between the absenceof nasal-bone ossification and trisomy 21 (P < 0.0001,sensitivity 0.28, confidence interval 0.11, specificity 1).The distribution of nasal-bone length according to GA

in the two groups is given in Fig 6. The small size of ourseries did not permit the establishment of a standardgrowth curve.

Middle phalanx of the fifth digit (P2)

In normal fetuses, the middle phalanx of the fifth digit(P2) exhibited morphological changes according to GAand was always the last of the hand phalanges to be ossi-fied. Moreover, it ossified after the terminal phalanx ofthe same digit (P3). In addition, its length and shape var-ied from case to case between 16 and 23 WG. Its lengthwas either less than or equal to the P3 length, and itsshape was either quadrangular or tubular. When P2was longer than P3, it was tubular in shape. From 24WG, P2 length was always greater than P3 length andits shape was tubular (Fig.7 a, b).

In trisomic fetuses, both ossification and shape wereabnormal. Before 23 WG, P2 was not ossified (`am-esophalangia') in 11 of 31 cases. From 24 WG, P2 waseither not ossified or presented abnormal criteria ofdevelopment, being shorter than P3 or quadrangular ortubular in shape in 17 of 29 cases (Fig.7 c). These mor-phological features defined P2 hypoplasia, whichwas never observed in the control group. Chi-squaretest to determine the relationship between P2 hypopla-sia and trisomy 21 gave a highly significant result(P < 0.0001, sensitivity 0.56, confidence interval 0.126,specificity 1).

From these observations, we identified as key signsthe following three features: brachycephaly (increased

684

Fig.1 Femoral growth curve in control (*) and trisomic 21 ( * )groups

Fig.2 Tibial growth curve in control (*) and trisomic 21 ( * )groups

Fig.3 BPD/OFD ratio curve in control (*) and trisomic 21 ( * )groups. Control group: r2 0.11, P 0.392; trisomic group: r2 0.025, P0.233

BPD/OFD ratio), absence of nasal-bone ossification,and P2 hypoplasia.

Association of skeletal and visceral abnormalities

Four trisomic fetuses (6 %) presented only visceral mal-formations. Both visceral and skeletal abnormalitieswere found in 29 cases (48 %). Twenty-one fetuses(35 %) presented only skeletal abnormalities, of which11 (52 %) had one key sign, 8 (38 %) had two signs, and2 (9 %) had all three signs. Among these key signs, themost frequently observed was brachycephaly (n = 35,58%) followed by P2 hypoplasia (n = 28, 46%). Ab-sence of nasal-bone ossification was the least frequentof our key skeletal signs (n = 14, 23 %). Absence of na-sal-bone ossification was always associated with otherskeletal signs or visceral abnormalities, except in onecase. Six fetuses (10 %) did not present any skeletal norvisceral anomaly.

Discussion

Attempts at prenatal detection of trisomic 21 fetuses inthe absence of karyotype have already been made byseveral authors, mostly based on skeletal abnormalities.Besides absence of the twelfth pair of ribs or prematureossification of the calcaneum, four main features havebeen selected by ultrasonographers: short limb longbones, brachycephaly, short nasal bones, and brachym-esophalangia of the fifth digit.

We have reassessed these signs and emphasise thespecial value of three of them, depending on GA.

Limb long-bone length

There is no agreement in the literature on the value ofthis sign. Benacerraf et al. [12±14] showed that short-ened femur or humerus were good predictive signs for

685

Fig.4 Sensitivity and specificity for the BPD/OFD ratio in the con-trol group: ROC curve (se-sensitivity, 1-sp 1-specificity). The mainthreshold values of the ratio BPD/OFD (tv) are indicated on thecurve (arrows)

Fig.5a,b Lateral skull radiographs of a control and b trisomic fe-tuses at 20 WG. Note brachycephaly and absent nasal bones in thetrisomic fetus

a

b

Down's syndrome when associated with a thickened nu-chal fold. Grist et al. [15] found that a diminished ratioof measured-to-expected femur length (MFL/EFL)may indicate a risk of chromosomal abnormalities, in-cluding trisomy 21, sufficient to warrant amniocentesis.FitzSimmons et al. [16] also noted shortened limb longbones with predominance for the upper limbs. Other au-thors [4, 17, 18] did not find any significant differencebetween normal and trisomic patients. The tendency ofslowed growth of the limb long bones during the last tri-mester of pregnancy was observed in both groups of ourstudy, but was more pronounced in the trisomic group.However, the size of our series did not allow us to estab-

lish pathological values by age group Therefore, shortlimb long bones, when an isolated finding, were not con-sidered usable criteria to detect trisomy 21 in this study.We are actually enlarging our series in order to deter-mine precise threshold values for each GA.

BPD/OFD ratio: brachycephaly

Brachycephaly was defined as an increased BPD/OFDratio and has been described by many authors in trisomy21 [17±20]. In our study, the mean value of the BPD/OFD ratio was statistically higher in trisomy 21 com-pared to the control group, whatever the GA. However,there was individual overlap of both groups. Therefore,we established a ROC curve to choose arbitrarily athreshold value that defined brachycephaly in the con-trol group, with predetermined sensitivity and specifici-ty. For example, with the chosen value of 0.872, wecould detect Down's syndrome with a sensitivity of 0.54and a specificity of 0.91. In addition, we must emphasizethat, in our series, brachycephaly was always associatedwith one or two other key skeletal signs.

Nasal-bone ossification

Guist et al. [21] established a nasal-bone growth curvein a series of 376 normal fetuses aged from 14 to 35WG. They observed an increase of nasal-bone lengthfrom 4 mm at 14 WG to 12 mm at 35 WG. These resultsare in agreement with our observations. Talmant [22]first described abnormal growth of the nasal bones intrisomic fetuses, but did not give precise pathologicalvalues. We have the same experience, but the small sizeof our series did not allow the establishment of a stan-dard growth curve. Nevertheless, since nasal-bone ossi-fication is normally present by 15 WG, the 14 cases that

686

Fig.6 Nasal-bone growth in control (*) and trisomic 21 ( * )groups

a b c

Fig.7a±c Hand radiographsshowing variation in the ap-pearance of the fifth-digit pha-langes. a Control fetus at 21WG. P2 quadrangular andshorter than P3. b Control fe-tus at 27 WG. P2 tubular andlonger than P3. c Trisomic fe-tus at 27 WG. P2 hypoplasia

were not ossified after this age should be consideredpathological. This sign, even when present in isolation,should alarm the clinician.

Length and morphology of P2

The variability of the length of P2 among normal infantsand adults is well known. In a study of 10 different pop-ulations, Garn et al. [23] found the incidence of P2 hy-poplasia varied from 0.6 % among southwest Ohioadults to 5 % among Hongkong/Peru adults. In our pop-ulation, none of the normal fetuses demonstrated thisfeature.

Fifth-digit P2 hypoplasia has already been describedby Benacerraf et al. [24, 25] in trisomy 21 by comparingfifth- and fourth-digit middle phalanges. We found thiscomparison unreliable using prenatal US because of thetechnical difficulty of simultaneously observing two dig-its throughout their lengths. It is for this reason that wedefined P2 hypoplasia by comparing P2 and P3 of thesame digit. In addition, because of the characteristic pat-terns of P2 normal development, we considered that pre-natal detection of P2 hypoplasia must be interpreted inrelation to GA. Between 16 and 23 WG, short P2 is notabnormal, whereas absence of P2 ossification, whichwas never observed in controls at this period, is patholog-ical. After 24 WG, absence or hypoplastic P2 as definedin this study (shorter than P3 of the same digit and eitherquadrangular or tubular shaped) must be considered ab-normal. These P2 abnormalities were found in 28 of 60trisomic cases (46 %) and represent a reliable screeningparameter for Down's syndrome. This parameter is par-

ticularly useful in countries where medical terminationof pregnancy is allowed after 24 WG.

Finally, in the trisomy group we compared the rela-tive frequency of the key skeletal signs and the visceralmalformations found at post mortem. Skeletal abnor-malities were observed in 50 of the 60 cases, whereasvisceral malformations were present in 33 cases. Of in-terest is the fact that of the 33 visceral malformationsfound at post mortem, 14 demonstrated signs (in isola-tion or in combination) at prenatal US (atrioventricularcanal [n = 6], duodenal atresia [n = 2], hydrocephalus[n = 6], Dandy-Walker [n = 1], hydrops fetalis [n = 1]).Five abnormalities should have been detected (atrio-ventricular canal [n = 4], duodenal stenosis [n = 1]) and14 probably not (small cardiac septal defects); these 19patients had not undergone fetal morphological screen-ing because amniocentesis had been undertaken on thebasis of increased maternal age.

In conclusion, three key skeletal signs were observedeither singly or in combination on post-mortem radio-graphs of fetuses with trisomy 21. These signs are suit-able for fetal US screening. Bearing in mind thatbiological screening tests or maternal age indicationsdetect only a minority of Down's syndrome, we considerthat these skeletal signs, when present at whatever GA,fully justify evaluation of the fetus by amniocentesisand cytogenetic analysis.

Acknowledgements The authors are very grateful to the Materni-ty Hospitals of Paris ± Robert DebrØ, Montmorency and Senlis, forentrusting them with fetal examinations, to the cytogenetics teamof the Department for determining fetal karyotypes, and to PascalBlain and Maxette Pierrin for their valuable technical contribu-tion.

687

References

1. Wald NJ, Kennard A, Hackshaw A,et al (1997) Antenatal screening forDown's syndrome. J Med Screen 4:141±246

2. Benacerraf BR, Barss VA, Laboda LA(1985) A sonographic sign for the de-tection in the second trimester of thefetus with Down's syndrome. Am J Ob-stet Gynecol 151: 1078±1079

3. Benacerraf BR, Frigoletto FD Jr,Greene MF (1986) Abnormal facialfeatures and extremities in human tri-somy syndromes: prenatal US appear-ance. Radiology 159: 243±246

4. Biagotti R, Periti E, Cariati E (1994)Humerus and femur length in fetuseswith Down syndrome. Prenatal Diagn14: 429±434

5. Dicke JM, Crane JP (1991) Sonograph-ic recognition of major malformationsand aberrant fetal growth in trisomicfetuses. J Ultrasound Med 10: 433±438

6. Nyberg DA, Resta RG, Luthy DA, et al(1993) Humerus and femur lengthshortening in the detection of Down'ssyndrome. Am J Obstet Gynecol 168:534±538

7. Rodis JF, Vintzileos AM, Fleming AD,et al (1991) Comparison of humeruslength with femur length in fetuses withDown syndrome. Am J Obstet Gynecol165: 1051±1056

8. Stoll C, Dott B, Alembik Y, et al (1993)Evaluation of routine prenatal ultra-sound examination in detecting fetalchromosomal abnormalities in a lowrisk population. Hum Genet 91: 37±41

9. Grandjean H, Larroque D, Levi S(1998) Detection of chromosomal ab-normalities, an outcome of ultrasoundscreening. The Eurofetus Team. Ann NY Acad Sci 847: 136±140

10. Vintzileos AM, Egan JF, Smulian JC,et al (1996) Adjusting the risk for tri-somy 21 by a simple ultrasound methodusing fetal long bone biometry. ObstetGynecol 96: 953±958

11. StempflØ N, Huten Y, Fondacci C, et al(1995) Fetal bone age revisited: pro-posal of a new radiographic score. Pe-diatr Radiol 25: 551±555

12. Benacerraf BR, Gelman R, FrigolettoFD Jr (1987) Sonographic identificationof second trimester fetuses with Down'ssyndrome. N Engl J Med 317:1371±1376

13. Benacerraf BR, Cnann A, Gelman R,et al (1989) Can sonographers reliablyidentify anatomic features associatedwith Down syndrome in fetuses? Radi-ology 173: 377±380

688

14. Benacerraf BR, Neuberg D, FrigolettoFD Jr (1991) Humeral shortening insecond-trimester fetuses with Downsyndrome. Obstet Gynecol 77: 223±227

15. Grist TM, Westwood Fuller R, AlbiezKL, et al (1990) Femur length in the USprediction of trisomy 21 and otherchromosomal abnormalities, Radiology174: 837±839

16. FitzSimmons J, Droste S, Shepard TH,et al (1989) Long bone growth in fetus-es with Down syndrome. Am J ObstetGynecol 161: 1174±1177

17. LaFolette L, Filly RA, Anderson R,et al (1989) Fetal femur length to detecttrisomy 21. A reappraisal. J UltrasoundMed 8: 657±660

18. Shah YG, Eckl CJ, Stinson SK, et al(1990) Biparietal diameter/femurlength ratio, cephalic index, and femurlength measurements: not reliablescreening techniques for Down syn-drome. Obstet Gynecol 75: 186±188

19. Buttery B (1979) Occipitofrontal-bipa-rietal diameter ratio: an ultrasonic pa-rameter for the antenatal evaluation ofDown's syndrome. Med J Aust 2:662±664

20. Lockwood C, Benacerraf BR, KrinskyA, et al (1987) A sonographic screeningmethod for Down syndrome. Am J Ob-stet Gynecol 157: 803±808

21. Guist F, Vincent Y, Doumerc S, et al(1995) Ultrasound evaluation of thelength of the fetal nasal bones through-out gestation. Ultrasound Obstet Gy-necol 5: 304±307

22. Talmant Cl (1990) Apport de l'Øchog-raphie au diagnostic antØnatal de latrisomie 21. Perspectives d'avenir. MedFoetale Echogr Gynecol 1: 9±20

23. Garn SM, Fels SL, Israel, H (1967)Brachymesophalangia of digit five inten populations. Am J Phys Anthropol27: 205±209

24. Benacerraf BR, Osathanondh R, Frigo-letto, FD (1988) Sonographic demon-stration of hypoplasia of the middlephalanx of the fifth digit: a finding as-sociated with Down syndrome. Am JObstet Gynecol 159: 181±183

25. Benacerraf BR, Harlow BL, FrigolettoFD Jr (1990) Hypoplasia of the middlephalanx of the fifth digit. A feature ofthe second trimester fetus with Down'ssyndrome. J Ultrasound Med 9:389±394