Pediatr Radiol (2002) 32: 667–673 DOI 10.1007/s00247-001-0627-x Øystein E. Olsen Rolv T. Lie Helga Maartmann-Moe Jouko Pirhonen Ralph S. Lachman Karen Rosendahl Received: 2 April 2001 Accepted: 24 August 2001 Published online: 26 July 2002 Ó Springer-Verlag 2002 Presented in part at the 37th Congress of the European Society of Paediatric Radiology, Lisbon, May 2000 Ø.E. Olsen (&) Æ K. Rosendahl Department of Radiology, Haukeland University Hospital, 5021 Bergen, Norway E-mail: oeol@start.no Tel.: +47-55-972400 Fax: +47-55-975140 R.T. Lie Section for Medical Statistics and Medical Birth Registry of Norway, University of Bergen, Bergen, Norway H. Maartmann-Moe Department of Pathology, Haukeland University Hospital, Bergen, Norway J. Pirhonen Department of Obstetrics and Gynaecology, Ullevaal Hospital, University of Oslo, Oslo, Norway R.S. Lachman International Skeletal Dysplasia Registry, Cedars-Sinai Medical Center, Los Angeles, California, USA ORIGINAL ARTICLE Skeletal measurements among infants who die during the perinatal period: new population-based reference Abstract Background: Reference data for roentgen skeletal measurements among infants who die during the perinatal period is not available, although it might prove helpful in the study of pre-autopsy radiographs. Objective: Our aim was to define new population-based reference data for skeletal measurements among infants who die during the perinatal period. Materials and methods: We routinely took standardised pre-autopsy radiographs of aborted and stillborn fetuses from 16 weeks gestational age to 7 days after delivery during a period of 11 years in our hospital. The data presented here represents nearly all perinatal deaths in a well-defined geographical area during the study period. We calculated detailed plots of estimated 10th–90th centiles and quartiles of different skeletal measurements by gestational age at death. Results: High correlations were seen between birth weight and the different skeletal measurements, including cranial width (r>0.9, P<0.001). We were not able to Introduction The recognition of skeletal malformations in post-mortem radiographic examinations of stillborns and infants who die just after birth plays a significant role in assessing risks of recurrence. Under these circumstances, two important aims of the radiological examination are identify any asymmetrical pattern of skeletal growth. Reference plots for femoral, tibial, humeral, radial and lumbar spine lengths, and for pelvic width are presented. Conclusions: We suggest that the current population-based reference data might be beneficial, and that skeletal radiographic measurements might contribute substantially in the assessment of fetal growth stage and in detection of skeletal abnormalities in infants who die during the perinatal period. Keywords Conventional radiography Æ Fetus Æ Skeletal– appendicular Æ Skeletal–axial Æ Skeletal growth and development first to detect possible overall growth anomalies, i.e. growth restriction, and second to assess the proportions between different bones in order to identify particular abnormal growth patterns. For these purposes, reference data for skeletal development is essential. A search of the literature identified only a few studies providing reference material, none of which had a well-defined 668 population base. The first studies on perinatal radiography primarily focused on the ability to estimate gestational age [1, 2, 3] and to evaluate fetal maturity in late pregnancy [4]. Specific gestational age-independent US parameters for assessment of fetal growth, e.g. femur length/abdominal circumference ratio [5] and transverse cerebellar diameter/abdominal circumference [6], have shown discouraging predictive test values. The aim of this study was to provide population-based references data for skeletal linear measurements among infants who die during the perinatal period. Materials and methods Autopsy was performed routinely and in accordance with a local standard procedure in all cases of perinatal death, i.e. from 16 weeks gestational age to 7 days post partum, in Haukeland University Hospital, Bergen, Norway, during the period 1988– 1998. One requirement was parental consent. Induced abortions on a non-medical basis were not examined. A total of 1,024 examinations were done. In all cases, full-body radiographs were obtained using a Faxitron, fine-grain film and low-kilovolt technique. Two anteroposterior (AP) radiographs were taken with the fetus lying flat on the film with the extremities extended: one to demonstrate the trunk and one, 10 kV less, for the extremities. Additionally, one lateral radiograph was taken with the fetus lying in the decubitus position. This technique was described by Seppänen [7]. Chromosome analyses were carried out when the pathologist suspected a chromosome anomaly, and in all cases of external malformations. In order to reduce bias due to referral from the secondary to tertiary health care level, we included only cases where the mother resided in the Bergen local hospital area at the time of abortion or delivery. Haukeland University Hospital is the only maternity hospital in this area. The total population in the selected area was 316,000 in January 2000. We retrospectively reviewed a total of 542 cases. Of the 542 pregnancies, 47 (9%) were twin pregnancies, and these were excluded from further analyses. Of the 495 cases included, 352 (71%) were prenatally dead fetuses, 65 (13%) procured abortions, 52 (11%) early neonatally dead fetuses, and 26 (5%) unknown. There were 184 (37%) female fetuses, 306 (62%) male and 5 (1%) of unknown or uncertain sex. Relevant information was obtained from the clinical records (maternal health, gestational age estimates based on routine US screening, pregnancy and birth history, and clinical findings in the fetus), and from the autopsy records (fetal/neonate and placental findings at autopsy, chromosomal findings, and final diagnosis). The Medical Birth Registry of Norway provided information on birth weight and gestational age based on maternal menstrual history. Radiographic measurements were done by one of us (Ø.E.O.), in part on digitised films, computer-assisted, using the free UTHSCSA ImageTool program (developed at the University of Texas Health Science Center at San Antonio, Texas, and available from the internet via anonymous FTP from ftp://maxrad6.uthscsa.edu). The following measurements were made: humeral, radial, femoral, tibial and lumbar spine lengths, and pelvic width. The length of a tubular bone was defined as the maximum distance between the ossified rims of the opposite metaphyses in the AP projection. For paired bones, the mean length was used for further calculations. Pelvic width was measured as maximum pelvic osseous width in the AP projection; lumbar length as the distance from the cranial margin of the body of the first lumbar vertebra to the caudal margin of the body of the fifth lumbar vertebra. Measurements were disregarded when the predefined structures were not clearly identified on the films, when extremity flexion was seen (for the extremity measurements), or when a trunk rotational malposition was noted (for pelvic measurements). For all analyses, gestational age categories were defined using menstrual history data. Quartiles and centiles for skeletal measurements by gestational age were calculated using calculated mean and standard deviation [8], which was justified by a near-normal data distribution. For statistical analysis, we used the statistical package SPSS for Windows, release 9. The paired t-test was used to analyse differences in mean gestational age based on US and menstrual history data. For bivariate correlation analysis, Pearson’s coefficient was calculated. Cohen’s kappa was used to test the prediction of low birth weight from skeletal measurements. All reported P values are two-tailed. Results A fairly linear increase in skeletal measurements was seen (Fig. 1). Although we think that it is reasonable to present measurements by gestational age as continuous curves, we do not imply that these curves in any way represent a longitudinal pattern of growth. These data are cross-sectional and represent only fetuses who died during the perinatal period. Centiles for femoral (n=346), tibial (n=329), humeral (n=342), radial (n=333) and lumbar spine (n=284) lengths and pelvic width (n=343) within gestational age groups were estimated and plotted separately. In a few gestational age categories, increased spread between the curves was seen. Removing one or two outliers in each respective category could eliminate these deviations, but such a procedure was not considered justifiable for the final presentation. There were no differences in birth weight between female and male fetuses within gestational age categories. Mean gestational age based on the last menstrual period was 24.4 weeks (SE±0.38) (see Fig. 2). Mean gestational age based on US measurements in the second trimester was 23.5 weeks (SE±0.34), a difference of nearly 1 week (P<0.001). The placenta and umbilical cord were normal in 105 (21%) of the fetuses, while abruptio placentae was found in 82 (17%), chorioamnionitis in 61 (12%) and placental infarction in 52 (11%) of the cases. Other abnormal findings were present in 115 (23%) of the cases, e.g. velamentous insertion, fibrosis, reactive and degenerative changes. Description of the placenta was missing in 80 (16%) of the cases. Of the fetuses, 209 (42%) were considered ‘normal’ at autopsy. External malformations were found in 94 (19%) and other abnormalities in 185 (37%). The autopsy record was missing in seven cases (1%). Table 1 lists the main autopsy diagnoses. The cases were categorised into 2-week gestational age groups for further analysis. For a study of overall differences in size ranges between these age groups, we computed a new variable, mean length, which was the mean length of the humerus, radius, femur, tibia and the lumbar spine. This was done in order to reduce the influence of inaccurate single measurements. The mean length variable had a narrower 95% confidence interval 669 Fig. 1. Mean, quartiles and 10th–90th centiles of femoral (n=346), tibial (n=329), humeral (n=342), radial (n=333) and lumbar spine lengths (n=284) and of pelvic width (n=343) by 2-week gestational age categories. The quartiles and centiles were calculated from each gestational-age category mean and standard deviation. Indicative lines were drawn between the calculated values in the gestational age groups 26–27 weeks and below (less than 8 mm), and a wider interval in groups 28– 29 weeks and above (more than 7 mm). There were considerable overlaps of the 95% confidence intervals between groups from 22 to 23 weeks and 28 to 29 weeks, and also between the groups from 30 to 31 weeks and above. Similarly configured confidence intervals for gestational age were seen for humeral, radial, femoral, tibial and lumbar spine lengths, and there were fairly high correlations between these parameters (r>0.9, P<0.001). Both pelvic width and cranial outer width showed a fairly high correlation with the mean length variable (r>0.9, P<0.001). Femoral cylinder index, defined as femoral length/ femoral midshaft width, showed a fairly low correlation with mean length, birth weight and gestational age (r<0.37, P<0.001). To investigate possible effects of growth restriction on the body proportions, we identified cases where growth restriction might be suspected. These were cases where placental abnormality, apart from abruptio placentae, was reported, with abnormal findings at autopsy, 670 Fig. 2. Distribution of gestational age calculated as weeks from the first day of the last menstrual period Table 1. Main autopsy diagnoses in 495 singletons Diagnosis No. of cases Percent No definite diagnosis Procured abortion Abruptio placentae Chorioamnionitis Placenta infarction Other placental pathology Asphyxia/hypoxia Pulmonary hyaline membranes Multiple malformations Dysmaturity/prematurity Cardiovascular malformations Maternal disease Other 107 82 69 40 27 21 12 12 9 9 8 7 92 22% 17% 14% 8% 5% 4% 2% 2% 2% 2% 2% 1% 19% Total 495 100% chromosomal anomalies, and where growth restriction had been suspected clinically. According to these criteria, 344 cases (69%) were suspected and 151 (31%) were not suspected of having suffered growth restriction. Figure 3 plots the relationships between mean length and birth weight for cases suspected and not suspected of growth restriction. No differences between the two categories could be seen. Overall, log10 birth weight and mean length showed a fairly high correlation (r=0.93, P<0.001). Finally, we grouped cases according to birth weight above or below the estimated 25th centile for their gestational age group. These groups were then cross-tabulated according to the mean length being above or below the 25th centile for gestational age group. The kappa value of this cross-tabulation was 0.75 (P<0.001). Figure 4 demonstrates the range of skeletal growth within the same gestational age (Fig. 4). Fig. 3. a The relationship between gestational age and mean length variable (the mean of humeral, radial, femoral, tibial and lumbar spine lengths). b The correspondence between the mean length variable and birth weight. The cases were grouped according to whether growth restriction was suspected (open circles) or not suspected (filled circles) Discussion Our descriptive data shows a surprisingly regular pattern of skeletal measurements among infants who die during the perinatal period. There seems to be a fairly linear relationship between gestational age and the size of particular bones. The sizes of individual bones also had a fairly high correlation. No differences in measurements were seen between male and female fetuses, suggesting that there is no need for sex-specific references for skeletal measurements. The relationship between birth weight and bone measurements did not differ between cases that were suspected of having suffered growth restriction and, other cases (Fig. 3). 671 Fig. 4. Post-mortem radiographs of two prenatally dead fetuses, both of 24-week gestational age, showing considerable differences in skeletal size. The fetus on the left (femur length 34 mm) died from abruption of the placenta and was not considered at high risk of growth restriction. The fetus on the right (femur length 16 mm) was terminated due to US findings (hydrops) and was considered to be at high risk of growth restriction The results deviated in part from those of earlier reports. Although our linear measurements tended to be shorter than those reported in several US studies [9, 10, 11], a direct comparison is probably not justified because of considerably different study populations. Likewise, we also found shorter linear measurements than previously reported in a non-population-based radiography study [12, 13]. In a study of post-mortem radiographs, Seppänen [7] defined references for cranial summation index, which we did not repeat, and for femoral cylinder index [14], which in our material had a fairly low correlation with birth weight and gestational age, as well as other skeletal measurements. Stempfle et al. [15] proposed a model for estimating gestational age based on qualitative radiographic observations in a post-mortem study, but their findings were not directly comparable to ours. Earlier studies have attempted to collect information on infants thought to represent ‘normal material’ [7,15]. We report a population-based reference material of fetuses who died during the perinatal period. This probably accounts for the somewhat shorter skeletal measurements in our study. Infants were included on a geographical rather than on a medical basis, and we thereby defined a representative sample of singleton perinatal death in a Norwegian population, which could act as a reference for similar samples. Some earlier studies applied regression analyses as if handling longitudinal data [13, 15, 16]. This is methodologically questionable, as post-mortem data are truly cross-sectional. For this reason, the use of descriptive data and centiles was considered more suitable for our purposes. The continuous lines of Fig. 1 represent an attempt to smooth the data rather than indicating longitudinal data sampling. We have chosen to estimate gestational age as time since the first day of the last menstrual period and present centiles for skeletal measurements in categories of gestational age. There is a possible risk of underestimation of gestational age using US in a group of abnormal fetuses [17]. This possibility was supported by a significant difference between the estimates of the US and menstrual data methods in our material. Besides, describing growth patterns as a function of US age would be a misnomer, since US age estimation itself is based on skeletal size parameters. Nevertheless, the use of gestational age as ascertained by the last menstrual period is not free from possible error, as a considerable underestimation of gestational age among low-birthweight individuals has been reported [18]. We suspect that erroneous estimation of gestational age possibly accounts for the wide ranges of size within age groups. It 672 also possibly, in part, explains why we did not detect differences in size between the infants suspected and not suspected of suffering growth restriction. This may, however, not diminish the practical validity of our data, as the same extent of misdating would be expected in similar samples. The considerable overlaps in bone lengths and widths between gestational age groups in our material could also be explained by constitutional factors and by variable presence of abnormal growth between gestational age categories. The wider inter-centile ranges of the upper gestational age groups could be explained by smaller group sizes, but also by increasing constitutional and pathological variation in size by increasing gestational age. The true prevalence of growth restriction in our material could not be established, but should be higher than that reported for fetal and neonatal populations in general. Asymmetrical growth restriction, i.e. a deviation between weight and length measurements, and a possible divergence between different growth parameters have been discussed in the literature [19, 20, 21, 22]. We found a fairly high correlation between all measured bone lengths and also between bone measurements and fetal weight. When comparing groups at higher and lower risk of growth restriction, no differences were seen in these associations. There was also a high correlation between low birth weight and lowcentile skeletal measurements, represented by a kappa value of 0.75. Therefore, in our material, no evidence of weight–skeletal size divergence or of bone dispro- portionality was found. It has been suggested that premature delivery and higher morbidity rates have a greater association with symmetrical than with asymmetrical growth restriction [23], which could in part explain our findings. Another explanatory factor could be the left-skewed age distribution, although the concept of late onset of a more asymmetrical form of growth restriction is controversial [23, 24]. The quite regular patterns of skeletal sizes in our material are useful when our references are used in a diagnostic setting to screen for particular syndromes or dysplasias with known skeletal asymmetry. We conclude that skeletal measurements could contribute substantially in the assessment of growth stage for gestational age among infants who die during the perinatal period. We also suggest that the added use of population-based reference tables may be valuable in the assessment of individual cases, attempting to identify specific syndromes or dysplasias by anomalies of proportions between skeletal parts. Validation of our reference data in independent material would be desirable in order to establish their diagnostic accuracy. The practical use of our results will primarily be in the workup of post-mortem radiographs of fetuses who die during the perinatal period. Acknowledgements Haakon and Sigrun Odegaard’s Foundation supported this work financially. We thank several contributors at the Department of Obstetrics and Gynaecology, Haukeland University Hospital, for collecting the birth record data, and we thank the Medical Birth Registry of Norway for supplying valuable registry data. References 1. Williamson MR, Edwards DK (1980) Prediction of gestational age of infants from the abdominal radiograph. Pediatr Radiol 9:229–231 2. Kuhns LR, Sherman MP, Poznanski AK (1972) Determination of neonatal maturation on the chest radiograph. Radiology 102:597–603 3. Kuhns LR, Finnstrom O (1976) New standards of ossification of the newborn. Radiology 119:655–660 4. Margolis AJ, Voss RG (1968) A method for radiologic detection of fetal maturity. Am J Obstet Gynecol 101:383–389 5. Shalev E, Romano S, Weiner E, et al (1991) Predictive value of the femur length to abdominal circumference ratio in the diagnosis of intrauterine growth retardation. Isr J Med Sci 27:131–133 6. Tongsong T, Wanapirak C, Thongpadungroj T (1999) Sonographic diagnosis of intrauterine growth restriction (IUGR) by fetal transverse cerebellar diameter (TCD)/abdominal circumference (AC) ratio. Int J Gynaecol Obstet 66:1–5 7. Seppanen U (1985) The value of perinatal post-mortem radiography. Experience of 514 cases. Ann Clin Res 17 Suppl 44:1–59 8. Skjaerven R, Gjessing HK, Bakketeig LS (2000) New standards for birth weight by gestational age using family data. Am J Obstet Gynecol 183:689–696 9. Warda AH, Deter RL, Rossavik IK, et al (1985) Fetal femur length: a critical reevaluation of the relationship to menstrual age. Obstet Gynecol 66:69–75 10. Elejalde BR, de Elejalde MM (1986) The prenatal growth of the human body determined by the measurement of bones and organs by ultrasonography. Am J Med Genet 24:575–598 11. Persson PH, Weldner BM (1986) Normal range growth curves for fetal biparietal diameter, occipito frontal diameter, mean abdominal diameters and femur length. Acta Obstet Gynecol Scand 65:759–761 12. van der Harten HJ, Brons JT, Schipper NW, et al (1990) The prenatal development of the normal human skeleton: a combined ultrasonographic and postmortem radiographic study. Pediatr Radiol 21:52–56 13. Bagnall KM, Harris PF, Jones PR (1982) A radiographic study of the longitudinal growth of primary ossification centers in limb long bones of the human fetus. Anat Rec 203:293–299 14. Whitley CB, Gorlin RJ (1983) Achondrogenesis: new nosology with evidence of genetic heterogeneity. Radiology 148:693–698 673 15. Stempfle N, Huten Y, Fondacci C, et al (1995) Fetal bone age revisited: proposal of a new radiographic score. Pediatr Radiol 25:551–555 16. Axelsson O, Hemmingsson A (1974) Roentgenologic determination of foetal maturity. Ups J Med Sci 79:94–96 17. Nguyen T, Larsen T, Engholm G, et al (2000) A discrepancy between gestational age estimated by last menstrual period and biparietal diameter may indicate an increased risk of fetal death and adverse pregnancy outcome. BJOG 107:1122–1129 18. Gjessing HK, Skjaerven R, Wilcox AJ (1999) Errors in gestational age: evidence of bleeding early in pregnancy. Am J Public Health 89:213–218 19. Sherwood RJ, Robinson HB, Meindl RS, et al (1997) Pattern and process of growth of the abnormal human fetus. Hum Biol 69:849–871 20. Campbell BA (1998) Utilizing sonography to follow fetal growth. Obstet Gynecol Clin North Am 25:597–607 21. Kjar I (1974) Skeletal maturation of the human fetus assessed radiographically on the basis of ossification sequences in the hand and foot. Am J Phys Anthropol 40:257–275 22. Lee W, Barton S, Comstock CH, et al (1991) Transverse cerebellar diameter: a useful predictor of gestational age for fetuses with asymmetric growth retardation. Am J Obstet Gynecol 165:1044– 1050 23. Lin CC, Su SJ, River LP (1991) Comparison of associated high-risk factors and perinatal outcome between symmetric and asymmetric fetal intrauterine growth retardation. Am J Obstet Gynecol 164:1535–1541 24. Vik T, Vatten L, Jacobsen G, et al (1997) Prenatal growth in symmetric and asymmetric small-for-gestationalage infants. Early Hum Dev 48:167–176