- Email: [email protected]
Accuracy of developing tooth length as an estimate of age in human skeletal remains: The deciduous dentition
Forensic Science International 172 (2007) 17–22 www.elsevier.com/locate/forsciint
Accuracy of developing tooth length as an estimate of age in human skeletal remains: The deciduous dentition Hugo F.V. Cardoso * Departamento de Zoologia e Antropologia (Museu Bocage), Museu Nacional de Histo´ria Natural, Rua da Escola Polite´cnica 58, 1269-102 Lisboa, Portugal Received 25 May 2006; received in revised form 12 November 2006; accepted 13 November 2006 Available online 14 December 2006
Abstract Dental age assessments are widely used to estimate age of immature skeletal remains. Most methods have relied on fractional stages of tooth emergence and formation, particularly of the permanent dentition, for predicting the age of infants and very young children. In this study, the accuracy of regression equations of developing deciduous tooth length for age estimation (Liversidge et al.) is tested on a sample of 30 Portuguese subadult skeletons of known age at death. Overall the method shows high accuracy and the average difference between estimated and chronological age is between 0.20 and 0.14 years when using single teeth, and 0.06 years, when using all available teeth. However, there is a tendency for the deciduous molars to provide overestimates of chronological age. Results show that age estimates can be obtained within 0.10 years with a 95% confidence interval when several teeth are used. Overall between-tooth agreement in age estimates decreases with increasing age but there is less variability of estimates with more teeth contributing to overall mean age. One seemingly limitation of this method may be the fact that it was developed by combining the maxillary and mandibular teeth. The other is related to the accuracy with which radiographic tooth length can be used as a valid surrogate for actual tooth length. Nevertheless, the advantages of this metric method surpass the limitations of chronologies based on stages of dental development. # 2006 Elsevier Ireland Ltd. All rights reserved. Keywords: Forensic odontology; Age determination by teeth; Subadults; Tooth length; Deciduous dentition
1. Introduction Dental age assessments have long been considered the best indicators of age at death in immature skeletal remains and particular attention has been devoted to the development and test of methods based on the permanent dentition. However, when dealing with the remains of very young children, the forensic anthropologist or the bioarchaeologist can utilize additional age markers provided by the deciduous dentition. The formation of deciduous teeth begins before birth and is complete by about the fourth postnatal year, thus providing a number of chronological landmarks during this period [1]. Deciduous tooth emergence has been used to estimate age at death of unknown immature skeletons in forensic investigations [2,3] but it relies on the timing of a single, brief event in the process of tooth eruption. In contrast, tooth formation
* Tel.: +351 213 921 823; fax: +351 213 969 784. E-mail address: [email protected]. 0379-0738/$ – see front matter # 2006 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.forsciint.2006.11.006
progresses in a continuous and cumulative way and ends only when the tooth has been completely formed. Although postnatal deciduous tooth formation data is available [4,5] it is somewhat cursory and particularly scarce for the anterior teeth. Additional difficulties with methods that rely on schedules of tooth formation include problems of defining formation stages and subjectivity in identifying stages such as fractions of the crown [6]. Radiographic assessments of deciduous teeth have an additional problem with difficulty in distinguishing crown-completed stages in radiographs, since deciduous teeth are less opaque than permanent teeth. Quantitative data on dental development offer additional information about deciduous tooth development, can provide more objective methods of age estimation and overcome some of the problems with the use of developmental stages. The quantification of incremental lines in enamel provides the highest accuracy in age estimation from teeth because it is an absolute method without reference to growth standards of a particular population. The method requires counting all shortperiod incremental lines (cross-striations) along an enamel
18
H.F.V. Cardoso / Forensic Science International 172 (2007) 17–22
prism which gives the time in days taken to grow that prism, since cross-striations represent 1 day’s growth of enamel [7,8]. The disadvantage of this technique is that it is destructive of dental material, requires laboratory facilities and experienced technicians and is both expensive and time consuming. In addition, enamel microstructures are more difficult to discern clearly in teeth from the deciduous dentition [9]. Other quantitative methods are based on the relationship between age and macroscopic measurements of teeth. At least two methods allow age to be predicted from regression formulae of deciduous developing tooth length [10,11]. Although both methods can be used for the entire postnatal period of the developing deciduous dentition, the method developed by Deutsch et al. [10] provides age estimates only from the length of the anterior teeth. Liversidge and co-workers’ method [11] requires a simple measurement of tooth length that can be used from birth to around 4 years of age and provides regression formulae for all deciduous teeth. The accuracy and reliability of methods based on deciduous tooth formation have been rarely reported and are far less common than on the permanent teeth. In fact, only two such studies [12,13] were found in the literature. Both of the studies test the accuracy of methods based on developmental stages on skeletal samples of known age at death and in only one is a method based on developing tooth length also tested. Some of the problems in obtaining information about deciduous dental chronology and testing the accuracy of these methods is the practical difficult of taking radiographs of infant and small children. Radiographs of very young children are not considered ethical except if treatment is needed. This difficulty highlights the importance of documented collections of immature remains, particularly if they include very young individuals. The goal of this study is to test the accuracy of Liversidge et al. [11] quantitative method of age estimation from developing deciduous teeth by using a sample of personally identified subadult skeletons from a 20th century Portuguese skeletal collection. Liversidge and co-workers’ method was the first of two published [11,14] that focused on developing tooth length as an estimate of age. 2. Materials and methods The material used in this study consists of the skeletal remains of 30 subadult individuals of known sex and age at death from the Lisbon identified skeletal collection [15]. Exact calendar age was obtained from birth and death civil records and was converted to decimal age. The age and sex distribution is depicted in Fig. 1. Age ranges from 1 day to 3.42 years and there are almost twice as much males as females (females = 11, males = 19). The sample derives from modern cemetery sources and individuals were buried between 1921 and 1974, but most died between 1930 and 1950. Years of birth overlap years of death, from 1920 to 1974, with a significant peak in the 1930s. Developing mandibular teeth were measured according to Liversidge et al. [11], as the distance from the cusp-tip to the developing edge of crown or root in the midline, parallel to the long axis of the tooth. In teeth with more than one cusp or root, the maximum length was measured. Measurements were taken on isolated teeth and on periapical radiographs with a standard digital Mitutoyo1 calliper, recorded to the nearest tenth of millimetre. Radiographs were taken in relation to the lingual–buccal plane with the aid of an extension cone paralleling film-holding instrument and with a 10 mm metal bar attached to the specimen to provide a gauge over image enlargement, thus eliminating parallax. No attempt
Fig. 1. Age and sex distribution of the sample (n = 30). was made to dissect unerupted teeth and the maxillary dentition was excluded due to problems with radiographic image distortion and superimposition. All deciduous mandibular teeth were measured and a total of 94 were included in the analysis. Intra-observer error paired t tests for isolated teeth (uniradicular: n = 20, t = 0.13, p = 0.8992; multiradicular: n = 20, t = 0.43, p = 0.6713) and for teeth measured on the radiographs (uniradicular: n = 20, t = 0.74, p = 0.4689; multiradicular: n = 20, t = 0.14, p = 0.8916) show that there are no statistically significant differences between repeated measurements of tooth length in either uniradicular or multiradicular teeth. Tooth length measured on radiographs is used as a surrogate for actual tooth length and, therefore, it is important to assess whether the two measurements differ significantly. Twenty isolated uniradicular teeth and six isolated multiradicular teeth were measured and then replaced on their sockets, where they were radiographed. The differences between actual tooth length and tooth length measured on the radiographs were tested by a paired t-test. Estimated ages were calculated from the regression formulae published by Liversidge et al. [11] for all deciduous teeth. These formulae were developed from an 18th century coffin-buried population from Christ Church, Spitalfields, London. Because regression equations are not sex-specific, sexes are combined and presented together. Age was estimated for each individual tooth and then, for each individual, an overall mean age estimate was calculated as the arithmetic mean of ages obtained from all available teeth. Simple subtractions between estimated and chronological age provided a measure of the accuracy of Liversidge et al. [11] regression equations for age estimation. A positive score represents an over-estimation, whereas a negative score represents an underestimation. Estimated age obtained from each tooth and overall mean age calculated for each individual were compared to chronological age using paired t-tests. Standard deviations from the overall mean age estimate were calculated to provide a measure of between-tooth agreement. Finally, a measure of variation in age estimates was obtained by calculating within-individual coefficients of variation based on the variation contributed by different teeth to the overall mean age.
3. Results Table 1 shows the results for comparisons between actual tooth length and tooth length measured on radiographs. For either uniradicular or multiradicular teeth the difference between the two measurements is not statistically significant. Raw data for individual dental age assessments of the Lisbon collection children based on developing tooth length are shown in Table 2 and summary results for accuracy tests of developing tooth length as estimates of age are shown in Table 3 by tooth type. The average difference between estimated and chronological age varies from 0.20 years when using the second molar, 0.14 years when using the first incisor and 0.06 years
H.F.V. Cardoso / Forensic Science International 172 (2007) 17–22
19
Table 1 Summary statistics and significance of difference between actual and radiographic length of teeth as measured by paired t-tests n
Uniradicular teeth Multiradicular teeth
Actual length
20 6
Radiographic length
Mean
S.D.
Mean
S.D.
15.34 11.63
1.6196 3.3902
15.32 11.74
1.5786 3.0183
t
p
0.91 0.31a
0.3722 0.7530
n, number of teeth; mean, mean tooth length; S.D., standard deviation of tooth length; t, paired t-test value; p, probability that the observed difference between actual and radiographic length is not zero due to chance alone. a Due to small sample size a non-parametric Wilcoxon paired-sample test was used instead.
between estimated and chronological age is plotted against chronological age, Fig. 3 shows that it tends to provide slight overestimates in younger individuals and underestimates in older ones. However, if the most positive and most negative observations are excluded such tendency disappears and the method seems to provide only slight overestimates across the age interval. Fig. 4 shows that variation in age estimates between teeth increases with increasing age of the individuals, when standard deviations from the overall mean age are plotted against mean age. The correlation coefficient between the two variables is highly significant (r = 0.66, p = 0.0000). Fig. 5 shows the within-individual coefficients of variation in age estimates plotted against the number of teeth contributing to
when age is estimated with all available teeth. At the 0.05 probability level, four teeth (m1, c, i2 and i1) provide age estimates that do not differ significantly from chronological age, whereas the remaining tooth (m2) provides overestimates of chronological age. Estimated and chronological age differ for the first molar only at a marginal statistical level. Fig. 2 illustrates the accuracy of each tooth by showing the distribution of differences between estimated and chronological age by tooth type. Table 3 and Fig. 2 show that the variation in mean difference is least for the incisors and first molar. Data also shows that dental age tends to progressively underestimate chronological age from the more posterior to the more anterior teeth. Although accuracy results for overall mean age show that it is a good estimate of chronological age, when the difference
Table 2 Dental age assessments of the Lisbon collection children based on developing tooth length and according to the method of Liversidge et al. [11] SP
1573 1140 1479 1560 369 347 1581 1160 1473 1531 466 392 359 375 1575 81 1471 661 1567 1499 1534-A 1532 930 1557 1558 1555 1576 900 1574 1482
CA
0.00 2.00 0.08 0.00 2.17 2.75 3.08 3.42 2.58 1.58 2.83 3.25 2.67 0.00 1.83 1.83 2.17 1.75 1.92 1.00 1.17 0.42 1.50 1.50 1.42 1.42 1.17 0.75 0.33 0.25
TL
EA
aat
m2
m1
c
i2
i1
m2
m1
c
i2
i1
– 10.3 4.2 3.4 13.1 14.5 13.7 13.6 13.4 8.5 13.5 14.1 12.7 – 11.1 10.9 11.7 11.0 11.2 8.0 6.5 5.1 9.1 9.3 9.2 7.5 6.8 6.2 4.4 3.8
3.9 – – – – – – – 13.2 11.9 – – 15.3 4.5 12.5 13.4 – 12.6 12.8 10.4 8.1 6.0 11.2 11.5 11.2 9.7 9.1 7.9 5.3 4.1
– – – – 16.3 16.3 15.5 16.1 – – 16.3 18.2 15.9 4.2 – 12.6 14.6 13.3 13.9 9.1 7.3 5.0 11.3 12.2 10.7 8.5 8.8 7.4 4.8 3.9
– – – – – – – – – – – – – 5.0 13.9 – – – 14.5 12.1 8.1 6.7 – 13.4 13.0 10.6 12.2 8.8 6.0 4.9
– – – – – – – – – – – – – 5.4 13.2 – – – – – – – – 13.0 12.6 12.2 12.2 11.1 6.5 5.9
– 2.10 0.32 0.09 2.93 3.34 3.11 3.07 3.02 1.59 3.03 3.21 2.79 – 2.34 2.29 2.52 2.30 2.37 1.44 1.00 0.57 1.74 1.80 1.77 1.27 1.08 0.90 0.37 0.21
0.05 – – – – – – – 2.11 1.83 – – 2.57 0.17 1.95 2.17 – 1.97 2.02 1.48 0.99 0.52 1.67 1.73 1.68 1.34 1.21 0.94 0.37 0.09
– – – – 2.77 2.76 2.61 2.73 – – 2.77 3.17 2.69 0.22 – 1.99 2.40 2.13 2.26 1.25 0.89 0.39 1.72 1.91 1.58 1.14 1.18 0.90 0.35 0.16
– – – – – – – – – – – – – 0.19 1.54 – – – 1.64 1.27 0.66 0.45 – 1.46 1.40 1.03 1.28 0.77 0.34 0.17
– – – – – – – – – – – – – 0.12 1.25 – – – – – – – – 1.22 1.16 1.10 1.10 0.94 0.28 0.20
0.05 2.10 0.32 0.09 2.85 3.05 2.86 2.90 2.56 1.71 2.63 2.70 2.68 0.17 1.77 2.06 2.19 2.03 2.07 1.36 0.88 0.48 1.64 1.63 1.52 1.18 1.17 0.89 0.34 0.16
SP, specimen number; CA, chronological age (in years); TL, tooth length (in mm); EA, estimated age (in years); aat, all available teeth (overall mean age in years).
20
H.F.V. Cardoso / Forensic Science International 172 (2007) 17–22
Table 3 Summary statistics for difference between estimated and chronological age and t-test values by tooth type Teeth
n
Mean
Min
Max
S.D.
t
p
m2 m1 c i2 i1 aat
28 20 23 13 9 30
0.20 0.09 0.05 0.07 0.14 0.06
0.35 0.48 0.69 0.51 0.58 0.73
0.77 0.48 0.61 0.27 0.19 0.69
0.2635 0.2122 0.2937 0.2282 0.2376 0.2695
4.11 1.89 0.87 1.18 1.82 1.19
0.0003 0.0740 0.3953 0.2608 0.1070 0.2431
n, number of teeth; mean, mean difference between estimated and chronological age; min, minimum difference between estimated and chronological age; max, maximum difference between dental and chronological age; S.D., standard deviation of difference between estimated and chronological age; t, paired t-test value; p, probability that the observed difference between estimated and chronological age is not zero due to chance alone; aat, all available teeth (overall mean age).
Fig. 4. Scatter plot of standard deviations of individual age estimates against overall mean age.
overall mean age. The coefficients range from 0% to 28.1%, but most individuals cluster below 20%. Variability of estimates seem to increase somewhat with increasing number of teeth contributing to overall mean age until four teeth, the number beyond which variability seems to decrease with increasing number of teeth contributing to overall mean age. Sample sizes are, however, small particularly for variation in estimates for one and three teeth. 4. Discussion
Fig. 2. Box plots of the distribution of the difference between dental and chronological age by tooth type. The median is represented by the line, the box includes 25–75% of the distribution, the whiskers the non-outlier range and the dots the outlier values.
Comparisons between estimated and chronological age show that Liversidge et al. [11] method based on length of developing deciduous teeth proved to be an accurate estimate of age. However, although the overall mean estimated age performed well, its accuracy seems to result from the divergence of estimates provided by the anterior and posterior teeth. Whereas the molars tend to provide overestimates of age, the incisors tend to provide underestimates, but the difference is only statistically significant at the 0.05 levels for the second molar. This means that the overestimates provided by the posterior dentition may tend to be cancelled off by the underestimates provided by the anterior teeth. Despite such discrepancies, Liversidge and co-workers methods seems to
Fig. 3. Scatter plot of the difference between estimated and chronological age against chronological age.
Fig. 5. Scatter plot of within-individual coefficients of variation for estimated age against the number of teeth utilized to calculate the estimate.
H.F.V. Cardoso / Forensic Science International 172 (2007) 17–22
outperform methods based on developmental stages. For example, in two studies where documented subadult samples were used to test the accuracy of deciduous tooth formation schedules for age estimation [12,13], the average difference in age estimation is approximately 1.5 to 1 year, whereas in this study average difference in age estimation is significantly below 1.5 year, between 0.05 and 0.20 years when using single teeth and 0.06 years when using all available teeth to estimate age. Only when age is determined using one atlas [16] and one diagram method [17] is accuracy of developmental stages similar to developing tooth length [13]. The differences between estimated and chronological age obtained in this study are also similar to those obtained by Liversidge [13] for the accuracy of another quantitative method based on deciduous tooth length [10], which averages differences between 0.02 and 0.15 years. No other studies provided accuracy tests of Liversidge et al. [11] equations or other similar quantitative method. Values in Table 3 can be used to calculate confidence intervals of mean differences as a measure of the reliability of the method or as a gauge of how closely one might expect to estimate chronological age from developing tooth length of a forensic or archaeological case. For example, the values of n and S.D. for aat p can ffiffiffi be used to calculate the standard error of the mean (S:D:= n), which multiplied by the appropriate critical value for the normal distribution (Z = 1.96) will provide the 95% confidence interval for the overall mean difference between estimated and chronological age (0.10 years). In addition to the method providing divergent estimates of the posterior and anterior dentition, its accuracy also seems to decrease with age. However, if the two outliers are eliminated the trend disappears. Whether this trend between accuracy and age is a true tendency or an artifact of the small number of older children, only a larger sample would reveal. The fact that in Fig. 3 the chronological age of younger individuals is overestimated seems contrary to the underestimated ages provide by the incisors, as these are the early forming teeth. In fact, the overestimates provided by the first and second molars balances the underestimates provide by the incisors in the younger individuals, while the only few individual underestimates provided by the second molar are those from the older children. Least accuracy in older children is consistent with the increase with age of sources of variability in dental development. However, standard deviations of individual age estimates using the deciduous dentition are considerably less than when using the permanent dentition [18]. Comparatively, it represents a three fold mean improvement in variation in age estimates. Within-individual variation in age estimates using developing deciduous teeth is also small, below 20%. The fact that the within-individual coefficients of variation does not go above 5% when only one tooth is contributing to the overall mean age originates from two very young individuals who show very accurate dental ages. As for the coefficient of variation not going below 5% when five teeth are contributing to the overall mean age it is related to the fact that these tend to be older children and, therefore, show increased variation in age estimates. Results
21
show, therefore, that age should be estimated from as many available teeth as possible. The fact that Liversidge and co-workers’ method [11] overestimates the age of younger children, may suggest that the study sample children are smaller than the Spitalfields children and may reflect worse environmental conditions of the Lisbon population. Despite the geographic and temporal distances that separate the study and Spitalfields samples, both of them can be considered to be drawn from populations under adverse environmental conditions [19]. However, other factors may help explain the similarities and/or differences in growth of deciduous teeth between the study and Spitalfields samples and, consequently, the accuracy of Liversidge and co-workers’ method. One factor is that tooth length was measured directly by Liversidge et al., whereas in this study tooth length was also measured on radiographs. Although the test of differences between radiographic and actual tooth length does not seem to show significant image amplification or measurement error, radiographic length of molars is slightly larger than the actual length and this discrepancy may be contributing to the difference between estimated age using posterior and anterior teeth. Another factor is that the samples used to derive the method include both upper and lower teeth. In fact, Liversidge et al. do not provide accurate information on the differences between length of upper and lower teeth and why data from both sets have been combined, except for the second incisor. If maxillary and mandibular deciduous teeth differ in growth rates, age estimates obtained from Liversidge et al. equations are likely to be biased if only applied to the maxillary or mandibular teeth. Although there are sex differences in the chronology of the deciduous dentition [6], the fact that in Liversidge and co-workers’ method the sexes are pooled does not seem to be affecting the results as both females and males in the study sample show slightly and similar overall overestimated ages. All these factors combined with a small test sample may, however, contribute to the observed accuracy results. Finally, tooth wear patterns in the deciduous dentition [20] may also be of some concern when applying this method. Although at present it is not possible to assess wear effects, because crown height may tend to decrease with increasing age, it is recommended that developing deciduous teeth with excessive wear should not be used. Age prediction using regression equations of tooth length seem to be particularly dependent on accurate radiographic measurements. Since the method devised by Liversidge and coworkers is based on measurements of actual teeth, it will only provide accurate estimates of age if radiographic enlargement is eliminated and length is properly measured. Careful control over this factor and appreciation of the limitations of the method will make it particularly advantageous over methods that rely on developmental stages. The method only requires a simple measurement of tooth length, thus avoiding subjective assessments fractional stages of dental development. As with Liversidge and Molleson’s [14] method for the permanent dentition, regression equations for predicting age from deciduous teeth are specifically designed to predict age, something that most methods are not designed for [21].
22
H.F.V. Cardoso / Forensic Science International 172 (2007) 17–22
Compared to the permanent dentition, age estimates derived from deciduous teeth show smaller differences between estimated and chronological age. Variability in estimates is small and between-tooth agreement is also high, which results from faster rates of tooth formation and less age variation in deciduous tooth length [11]. Finally, because the sex of immature skeletal remains cannot be easily determined, age estimates derived from Liversidge and co-workers technique are not affected by methodological considerations of sex. 5. Conclusion Documented collections of subadult skeletons are an important means to assess and evaluate the reliability of methods of age determination. This study shows that accuracy of developing deciduous tooth length for age estimation is higher than of developing permanent tooth length, because of the faster rate of change in the deciduous dentition. It is also more accurate than methods based on the chronology of stages of tooth formation and it provides a wide range of ages for all deciduous teeth. Because it is an expedient method it is an important alternative to microstructural methods when teeth cannot be destroyed. Although there are some concerns on how the accuracy of the method is affected by the reliability of radiographic tooth length measurements, it is a strongly recommended method for use in osteological research. References [1] H.M. Liversidge, T. Molleson, Variation in crown and root formation and eruption of human deciduous teeth, Am. J. Phys. Anthropol. 123 (2004) 172–180. [2] M. Nystro¨m, L. Peck, E. Kleemola-Kujala, M. Evalahti, M. Kataja, Age estimation in small children: reference values based on counts of deciduous teeth in Finns, Forensic Sci. Int. 110 (2000) 179–188. [3] M. Muller-Bolla, L. Lupi-Pe´gurier, G. Quatrehomme, A.M. Velly, M. Bolla, Age estimation from teeth in children and adolescents, J. Forensic Sci. 48 (2003) 140–148. [4] R. Kronfeld, I. Schour, Neonatal dental hypoplasia, J. Am. Dent. Assoc. 26 (1939) 18–32. [5] C.F.A. Moorrees, E.A. Fanning, E.E. Hunt, Formation and resorption of three deciduous teeth in children, Am. J. Phys. Anthropol. 21 (1963) 205–213.
[6] H.M. Liversidge, B. Herdeg, F.W. Ro¨sing, Dental age estimation of nonadults: a review of methods and principles, in: K.W. Alt, F.W. Ro¨sing, M. Teschler-Nicola (Eds.), Dental Anthropology—Fundamentals, Limits, and Prospects, Springer-Verlag, Wien, 1998, pp. 419–442. [7] A. Boyde, Estimation of age at death of young human skeletal remains from incremental lines in the dental enamel, Excerpta Med. Int. Congr. Ser. 80 (1963) 36–46. [8] C.M. Fitzgerald, J.C. Rose, Reading between the lines: dental development and subadult age assessment using the microstructural growth markers of teeth, in: M.A. Katzenberg, S.R. Saunders (Eds.), Biological Anthropology of the Human Skeleton, Wiley–Liss, New York, 2000, pp. 163–186. [9] C.M. Fitzgerald, S.R. Saunders, Test of histological methods of determining chronology of accentuated striae in deciduous teeth, Am. J. Phys. Anthropol. 127 (2004) 277–290. [10] D. Deutsch, O. Tam, M.V. Stack, Postnatal changes in size, morphology and weight of developing postnatal deciduous anterior teeth, Growth 49 (1985) 202–217. [11] H.M. Liversidge, M.C. Dean, T.I. Molleson, Increasing human tooth length between birth and 5.4 years, Am. J. Phys. Anthropol. 90 (1993) 307–313. [12] S.R. Saunders, C. DeVito, A. Herring, R. Southern, R. Hoppa, Accuracy tests of tooth formation age estimations for human skeletal remains, Am. J. Phys. Anthropol. 92 (1993) 173–188. [13] H.M. Liversidge, Accuracy of age estimation from developing teeth of a population of known age (0–5.4 years), Int. J. Osteoarchaeol. 4 (1994) 37–45. [14] H.M. Liversidge, T.I. Molleson, Developing permanent tooth length as an estimate of age, J. Forensic Sci. 44 (1999) 917–920. [15] H.F.V. Cardoso, The collection of identified human skeletons housed at the Bocage Museum (National Museum of Natural History) in Lisbon, Portugal, Am. J. Phys. Anthropol. 129 (2006) 173–176. [16] I. Schour, M. Massler, The development of the human dentition, J. Am. Dent. Assoc. 28 (1941) 1153–1160. [17] G. Gustafson, G. Koch, Age estimation up to 16 years of age based on dental development, Odontol. Rev. 25 (1974) 297–306. [18] H.M. Liversidge, F. Lyons, M.P. Hector, The accuracy of three methods of age estimation using radiographic measurements of developing teeth, Forensic Sci. Int. 131 (2003) 22–29. [19] H.F.V. Cardoso, Patterns of growth and development of the human skeleton and dentition in relation to environmental quality, PhD Thesis, Department of Anthropology, McMaster University, Hamilton, Ontario, 2005. [20] J.J. Warren, T. Yonezu, S.E. Bishara, Tooth wear patterns in the deciduous dentition, Am. J. Orthod. Dentofac. Orthop. 122 (2002) 614–618. [21] B.H. Smith, Standards of human tooth formation and dental age assessment, in: M. Kelley, C.S. Larsen (Eds.), Advances in Dental Anthropology, Wiley–Liss, New York, 1991, pp. 143–168.
Accuracy of developing tooth length as an estimate of age in human skeletal remains: The deciduous dentition Hugo F.V. Cardoso * Departamento de Zoologia e Antropologia (Museu Bocage), Museu Nacional de Histo´ria Natural, Rua da Escola Polite´cnica 58, 1269-102 Lisboa, Portugal Received 25 May 2006; received in revised form 12 November 2006; accepted 13 November 2006 Available online 14 December 2006
Abstract Dental age assessments are widely used to estimate age of immature skeletal remains. Most methods have relied on fractional stages of tooth emergence and formation, particularly of the permanent dentition, for predicting the age of infants and very young children. In this study, the accuracy of regression equations of developing deciduous tooth length for age estimation (Liversidge et al.) is tested on a sample of 30 Portuguese subadult skeletons of known age at death. Overall the method shows high accuracy and the average difference between estimated and chronological age is between 0.20 and 0.14 years when using single teeth, and 0.06 years, when using all available teeth. However, there is a tendency for the deciduous molars to provide overestimates of chronological age. Results show that age estimates can be obtained within 0.10 years with a 95% confidence interval when several teeth are used. Overall between-tooth agreement in age estimates decreases with increasing age but there is less variability of estimates with more teeth contributing to overall mean age. One seemingly limitation of this method may be the fact that it was developed by combining the maxillary and mandibular teeth. The other is related to the accuracy with which radiographic tooth length can be used as a valid surrogate for actual tooth length. Nevertheless, the advantages of this metric method surpass the limitations of chronologies based on stages of dental development. # 2006 Elsevier Ireland Ltd. All rights reserved. Keywords: Forensic odontology; Age determination by teeth; Subadults; Tooth length; Deciduous dentition
1. Introduction Dental age assessments have long been considered the best indicators of age at death in immature skeletal remains and particular attention has been devoted to the development and test of methods based on the permanent dentition. However, when dealing with the remains of very young children, the forensic anthropologist or the bioarchaeologist can utilize additional age markers provided by the deciduous dentition. The formation of deciduous teeth begins before birth and is complete by about the fourth postnatal year, thus providing a number of chronological landmarks during this period [1]. Deciduous tooth emergence has been used to estimate age at death of unknown immature skeletons in forensic investigations [2,3] but it relies on the timing of a single, brief event in the process of tooth eruption. In contrast, tooth formation
* Tel.: +351 213 921 823; fax: +351 213 969 784. E-mail address: [email protected]. 0379-0738/$ – see front matter # 2006 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.forsciint.2006.11.006
progresses in a continuous and cumulative way and ends only when the tooth has been completely formed. Although postnatal deciduous tooth formation data is available [4,5] it is somewhat cursory and particularly scarce for the anterior teeth. Additional difficulties with methods that rely on schedules of tooth formation include problems of defining formation stages and subjectivity in identifying stages such as fractions of the crown [6]. Radiographic assessments of deciduous teeth have an additional problem with difficulty in distinguishing crown-completed stages in radiographs, since deciduous teeth are less opaque than permanent teeth. Quantitative data on dental development offer additional information about deciduous tooth development, can provide more objective methods of age estimation and overcome some of the problems with the use of developmental stages. The quantification of incremental lines in enamel provides the highest accuracy in age estimation from teeth because it is an absolute method without reference to growth standards of a particular population. The method requires counting all shortperiod incremental lines (cross-striations) along an enamel
18
H.F.V. Cardoso / Forensic Science International 172 (2007) 17–22
prism which gives the time in days taken to grow that prism, since cross-striations represent 1 day’s growth of enamel [7,8]. The disadvantage of this technique is that it is destructive of dental material, requires laboratory facilities and experienced technicians and is both expensive and time consuming. In addition, enamel microstructures are more difficult to discern clearly in teeth from the deciduous dentition [9]. Other quantitative methods are based on the relationship between age and macroscopic measurements of teeth. At least two methods allow age to be predicted from regression formulae of deciduous developing tooth length [10,11]. Although both methods can be used for the entire postnatal period of the developing deciduous dentition, the method developed by Deutsch et al. [10] provides age estimates only from the length of the anterior teeth. Liversidge and co-workers’ method [11] requires a simple measurement of tooth length that can be used from birth to around 4 years of age and provides regression formulae for all deciduous teeth. The accuracy and reliability of methods based on deciduous tooth formation have been rarely reported and are far less common than on the permanent teeth. In fact, only two such studies [12,13] were found in the literature. Both of the studies test the accuracy of methods based on developmental stages on skeletal samples of known age at death and in only one is a method based on developing tooth length also tested. Some of the problems in obtaining information about deciduous dental chronology and testing the accuracy of these methods is the practical difficult of taking radiographs of infant and small children. Radiographs of very young children are not considered ethical except if treatment is needed. This difficulty highlights the importance of documented collections of immature remains, particularly if they include very young individuals. The goal of this study is to test the accuracy of Liversidge et al. [11] quantitative method of age estimation from developing deciduous teeth by using a sample of personally identified subadult skeletons from a 20th century Portuguese skeletal collection. Liversidge and co-workers’ method was the first of two published [11,14] that focused on developing tooth length as an estimate of age. 2. Materials and methods The material used in this study consists of the skeletal remains of 30 subadult individuals of known sex and age at death from the Lisbon identified skeletal collection [15]. Exact calendar age was obtained from birth and death civil records and was converted to decimal age. The age and sex distribution is depicted in Fig. 1. Age ranges from 1 day to 3.42 years and there are almost twice as much males as females (females = 11, males = 19). The sample derives from modern cemetery sources and individuals were buried between 1921 and 1974, but most died between 1930 and 1950. Years of birth overlap years of death, from 1920 to 1974, with a significant peak in the 1930s. Developing mandibular teeth were measured according to Liversidge et al. [11], as the distance from the cusp-tip to the developing edge of crown or root in the midline, parallel to the long axis of the tooth. In teeth with more than one cusp or root, the maximum length was measured. Measurements were taken on isolated teeth and on periapical radiographs with a standard digital Mitutoyo1 calliper, recorded to the nearest tenth of millimetre. Radiographs were taken in relation to the lingual–buccal plane with the aid of an extension cone paralleling film-holding instrument and with a 10 mm metal bar attached to the specimen to provide a gauge over image enlargement, thus eliminating parallax. No attempt
Fig. 1. Age and sex distribution of the sample (n = 30). was made to dissect unerupted teeth and the maxillary dentition was excluded due to problems with radiographic image distortion and superimposition. All deciduous mandibular teeth were measured and a total of 94 were included in the analysis. Intra-observer error paired t tests for isolated teeth (uniradicular: n = 20, t = 0.13, p = 0.8992; multiradicular: n = 20, t = 0.43, p = 0.6713) and for teeth measured on the radiographs (uniradicular: n = 20, t = 0.74, p = 0.4689; multiradicular: n = 20, t = 0.14, p = 0.8916) show that there are no statistically significant differences between repeated measurements of tooth length in either uniradicular or multiradicular teeth. Tooth length measured on radiographs is used as a surrogate for actual tooth length and, therefore, it is important to assess whether the two measurements differ significantly. Twenty isolated uniradicular teeth and six isolated multiradicular teeth were measured and then replaced on their sockets, where they were radiographed. The differences between actual tooth length and tooth length measured on the radiographs were tested by a paired t-test. Estimated ages were calculated from the regression formulae published by Liversidge et al. [11] for all deciduous teeth. These formulae were developed from an 18th century coffin-buried population from Christ Church, Spitalfields, London. Because regression equations are not sex-specific, sexes are combined and presented together. Age was estimated for each individual tooth and then, for each individual, an overall mean age estimate was calculated as the arithmetic mean of ages obtained from all available teeth. Simple subtractions between estimated and chronological age provided a measure of the accuracy of Liversidge et al. [11] regression equations for age estimation. A positive score represents an over-estimation, whereas a negative score represents an underestimation. Estimated age obtained from each tooth and overall mean age calculated for each individual were compared to chronological age using paired t-tests. Standard deviations from the overall mean age estimate were calculated to provide a measure of between-tooth agreement. Finally, a measure of variation in age estimates was obtained by calculating within-individual coefficients of variation based on the variation contributed by different teeth to the overall mean age.
3. Results Table 1 shows the results for comparisons between actual tooth length and tooth length measured on radiographs. For either uniradicular or multiradicular teeth the difference between the two measurements is not statistically significant. Raw data for individual dental age assessments of the Lisbon collection children based on developing tooth length are shown in Table 2 and summary results for accuracy tests of developing tooth length as estimates of age are shown in Table 3 by tooth type. The average difference between estimated and chronological age varies from 0.20 years when using the second molar, 0.14 years when using the first incisor and 0.06 years
H.F.V. Cardoso / Forensic Science International 172 (2007) 17–22
19
Table 1 Summary statistics and significance of difference between actual and radiographic length of teeth as measured by paired t-tests n
Uniradicular teeth Multiradicular teeth
Actual length
20 6
Radiographic length
Mean
S.D.
Mean
S.D.
15.34 11.63
1.6196 3.3902
15.32 11.74
1.5786 3.0183
t
p
0.91 0.31a
0.3722 0.7530
n, number of teeth; mean, mean tooth length; S.D., standard deviation of tooth length; t, paired t-test value; p, probability that the observed difference between actual and radiographic length is not zero due to chance alone. a Due to small sample size a non-parametric Wilcoxon paired-sample test was used instead.
between estimated and chronological age is plotted against chronological age, Fig. 3 shows that it tends to provide slight overestimates in younger individuals and underestimates in older ones. However, if the most positive and most negative observations are excluded such tendency disappears and the method seems to provide only slight overestimates across the age interval. Fig. 4 shows that variation in age estimates between teeth increases with increasing age of the individuals, when standard deviations from the overall mean age are plotted against mean age. The correlation coefficient between the two variables is highly significant (r = 0.66, p = 0.0000). Fig. 5 shows the within-individual coefficients of variation in age estimates plotted against the number of teeth contributing to
when age is estimated with all available teeth. At the 0.05 probability level, four teeth (m1, c, i2 and i1) provide age estimates that do not differ significantly from chronological age, whereas the remaining tooth (m2) provides overestimates of chronological age. Estimated and chronological age differ for the first molar only at a marginal statistical level. Fig. 2 illustrates the accuracy of each tooth by showing the distribution of differences between estimated and chronological age by tooth type. Table 3 and Fig. 2 show that the variation in mean difference is least for the incisors and first molar. Data also shows that dental age tends to progressively underestimate chronological age from the more posterior to the more anterior teeth. Although accuracy results for overall mean age show that it is a good estimate of chronological age, when the difference
Table 2 Dental age assessments of the Lisbon collection children based on developing tooth length and according to the method of Liversidge et al. [11] SP
1573 1140 1479 1560 369 347 1581 1160 1473 1531 466 392 359 375 1575 81 1471 661 1567 1499 1534-A 1532 930 1557 1558 1555 1576 900 1574 1482
CA
0.00 2.00 0.08 0.00 2.17 2.75 3.08 3.42 2.58 1.58 2.83 3.25 2.67 0.00 1.83 1.83 2.17 1.75 1.92 1.00 1.17 0.42 1.50 1.50 1.42 1.42 1.17 0.75 0.33 0.25
TL
EA
aat
m2
m1
c
i2
i1
m2
m1
c
i2
i1
– 10.3 4.2 3.4 13.1 14.5 13.7 13.6 13.4 8.5 13.5 14.1 12.7 – 11.1 10.9 11.7 11.0 11.2 8.0 6.5 5.1 9.1 9.3 9.2 7.5 6.8 6.2 4.4 3.8
3.9 – – – – – – – 13.2 11.9 – – 15.3 4.5 12.5 13.4 – 12.6 12.8 10.4 8.1 6.0 11.2 11.5 11.2 9.7 9.1 7.9 5.3 4.1
– – – – 16.3 16.3 15.5 16.1 – – 16.3 18.2 15.9 4.2 – 12.6 14.6 13.3 13.9 9.1 7.3 5.0 11.3 12.2 10.7 8.5 8.8 7.4 4.8 3.9
– – – – – – – – – – – – – 5.0 13.9 – – – 14.5 12.1 8.1 6.7 – 13.4 13.0 10.6 12.2 8.8 6.0 4.9
– – – – – – – – – – – – – 5.4 13.2 – – – – – – – – 13.0 12.6 12.2 12.2 11.1 6.5 5.9
– 2.10 0.32 0.09 2.93 3.34 3.11 3.07 3.02 1.59 3.03 3.21 2.79 – 2.34 2.29 2.52 2.30 2.37 1.44 1.00 0.57 1.74 1.80 1.77 1.27 1.08 0.90 0.37 0.21
0.05 – – – – – – – 2.11 1.83 – – 2.57 0.17 1.95 2.17 – 1.97 2.02 1.48 0.99 0.52 1.67 1.73 1.68 1.34 1.21 0.94 0.37 0.09
– – – – 2.77 2.76 2.61 2.73 – – 2.77 3.17 2.69 0.22 – 1.99 2.40 2.13 2.26 1.25 0.89 0.39 1.72 1.91 1.58 1.14 1.18 0.90 0.35 0.16
– – – – – – – – – – – – – 0.19 1.54 – – – 1.64 1.27 0.66 0.45 – 1.46 1.40 1.03 1.28 0.77 0.34 0.17
– – – – – – – – – – – – – 0.12 1.25 – – – – – – – – 1.22 1.16 1.10 1.10 0.94 0.28 0.20
0.05 2.10 0.32 0.09 2.85 3.05 2.86 2.90 2.56 1.71 2.63 2.70 2.68 0.17 1.77 2.06 2.19 2.03 2.07 1.36 0.88 0.48 1.64 1.63 1.52 1.18 1.17 0.89 0.34 0.16
SP, specimen number; CA, chronological age (in years); TL, tooth length (in mm); EA, estimated age (in years); aat, all available teeth (overall mean age in years).
20
H.F.V. Cardoso / Forensic Science International 172 (2007) 17–22
Table 3 Summary statistics for difference between estimated and chronological age and t-test values by tooth type Teeth
n
Mean
Min
Max
S.D.
t
p
m2 m1 c i2 i1 aat
28 20 23 13 9 30
0.20 0.09 0.05 0.07 0.14 0.06
0.35 0.48 0.69 0.51 0.58 0.73
0.77 0.48 0.61 0.27 0.19 0.69
0.2635 0.2122 0.2937 0.2282 0.2376 0.2695
4.11 1.89 0.87 1.18 1.82 1.19
0.0003 0.0740 0.3953 0.2608 0.1070 0.2431
n, number of teeth; mean, mean difference between estimated and chronological age; min, minimum difference between estimated and chronological age; max, maximum difference between dental and chronological age; S.D., standard deviation of difference between estimated and chronological age; t, paired t-test value; p, probability that the observed difference between estimated and chronological age is not zero due to chance alone; aat, all available teeth (overall mean age).
Fig. 4. Scatter plot of standard deviations of individual age estimates against overall mean age.
overall mean age. The coefficients range from 0% to 28.1%, but most individuals cluster below 20%. Variability of estimates seem to increase somewhat with increasing number of teeth contributing to overall mean age until four teeth, the number beyond which variability seems to decrease with increasing number of teeth contributing to overall mean age. Sample sizes are, however, small particularly for variation in estimates for one and three teeth. 4. Discussion
Fig. 2. Box plots of the distribution of the difference between dental and chronological age by tooth type. The median is represented by the line, the box includes 25–75% of the distribution, the whiskers the non-outlier range and the dots the outlier values.
Comparisons between estimated and chronological age show that Liversidge et al. [11] method based on length of developing deciduous teeth proved to be an accurate estimate of age. However, although the overall mean estimated age performed well, its accuracy seems to result from the divergence of estimates provided by the anterior and posterior teeth. Whereas the molars tend to provide overestimates of age, the incisors tend to provide underestimates, but the difference is only statistically significant at the 0.05 levels for the second molar. This means that the overestimates provided by the posterior dentition may tend to be cancelled off by the underestimates provided by the anterior teeth. Despite such discrepancies, Liversidge and co-workers methods seems to
Fig. 3. Scatter plot of the difference between estimated and chronological age against chronological age.
Fig. 5. Scatter plot of within-individual coefficients of variation for estimated age against the number of teeth utilized to calculate the estimate.
H.F.V. Cardoso / Forensic Science International 172 (2007) 17–22
outperform methods based on developmental stages. For example, in two studies where documented subadult samples were used to test the accuracy of deciduous tooth formation schedules for age estimation [12,13], the average difference in age estimation is approximately 1.5 to 1 year, whereas in this study average difference in age estimation is significantly below 1.5 year, between 0.05 and 0.20 years when using single teeth and 0.06 years when using all available teeth to estimate age. Only when age is determined using one atlas [16] and one diagram method [17] is accuracy of developmental stages similar to developing tooth length [13]. The differences between estimated and chronological age obtained in this study are also similar to those obtained by Liversidge [13] for the accuracy of another quantitative method based on deciduous tooth length [10], which averages differences between 0.02 and 0.15 years. No other studies provided accuracy tests of Liversidge et al. [11] equations or other similar quantitative method. Values in Table 3 can be used to calculate confidence intervals of mean differences as a measure of the reliability of the method or as a gauge of how closely one might expect to estimate chronological age from developing tooth length of a forensic or archaeological case. For example, the values of n and S.D. for aat p can ffiffiffi be used to calculate the standard error of the mean (S:D:= n), which multiplied by the appropriate critical value for the normal distribution (Z = 1.96) will provide the 95% confidence interval for the overall mean difference between estimated and chronological age (0.10 years). In addition to the method providing divergent estimates of the posterior and anterior dentition, its accuracy also seems to decrease with age. However, if the two outliers are eliminated the trend disappears. Whether this trend between accuracy and age is a true tendency or an artifact of the small number of older children, only a larger sample would reveal. The fact that in Fig. 3 the chronological age of younger individuals is overestimated seems contrary to the underestimated ages provide by the incisors, as these are the early forming teeth. In fact, the overestimates provided by the first and second molars balances the underestimates provide by the incisors in the younger individuals, while the only few individual underestimates provided by the second molar are those from the older children. Least accuracy in older children is consistent with the increase with age of sources of variability in dental development. However, standard deviations of individual age estimates using the deciduous dentition are considerably less than when using the permanent dentition [18]. Comparatively, it represents a three fold mean improvement in variation in age estimates. Within-individual variation in age estimates using developing deciduous teeth is also small, below 20%. The fact that the within-individual coefficients of variation does not go above 5% when only one tooth is contributing to the overall mean age originates from two very young individuals who show very accurate dental ages. As for the coefficient of variation not going below 5% when five teeth are contributing to the overall mean age it is related to the fact that these tend to be older children and, therefore, show increased variation in age estimates. Results
21
show, therefore, that age should be estimated from as many available teeth as possible. The fact that Liversidge and co-workers’ method [11] overestimates the age of younger children, may suggest that the study sample children are smaller than the Spitalfields children and may reflect worse environmental conditions of the Lisbon population. Despite the geographic and temporal distances that separate the study and Spitalfields samples, both of them can be considered to be drawn from populations under adverse environmental conditions [19]. However, other factors may help explain the similarities and/or differences in growth of deciduous teeth between the study and Spitalfields samples and, consequently, the accuracy of Liversidge and co-workers’ method. One factor is that tooth length was measured directly by Liversidge et al., whereas in this study tooth length was also measured on radiographs. Although the test of differences between radiographic and actual tooth length does not seem to show significant image amplification or measurement error, radiographic length of molars is slightly larger than the actual length and this discrepancy may be contributing to the difference between estimated age using posterior and anterior teeth. Another factor is that the samples used to derive the method include both upper and lower teeth. In fact, Liversidge et al. do not provide accurate information on the differences between length of upper and lower teeth and why data from both sets have been combined, except for the second incisor. If maxillary and mandibular deciduous teeth differ in growth rates, age estimates obtained from Liversidge et al. equations are likely to be biased if only applied to the maxillary or mandibular teeth. Although there are sex differences in the chronology of the deciduous dentition [6], the fact that in Liversidge and co-workers’ method the sexes are pooled does not seem to be affecting the results as both females and males in the study sample show slightly and similar overall overestimated ages. All these factors combined with a small test sample may, however, contribute to the observed accuracy results. Finally, tooth wear patterns in the deciduous dentition [20] may also be of some concern when applying this method. Although at present it is not possible to assess wear effects, because crown height may tend to decrease with increasing age, it is recommended that developing deciduous teeth with excessive wear should not be used. Age prediction using regression equations of tooth length seem to be particularly dependent on accurate radiographic measurements. Since the method devised by Liversidge and coworkers is based on measurements of actual teeth, it will only provide accurate estimates of age if radiographic enlargement is eliminated and length is properly measured. Careful control over this factor and appreciation of the limitations of the method will make it particularly advantageous over methods that rely on developmental stages. The method only requires a simple measurement of tooth length, thus avoiding subjective assessments fractional stages of dental development. As with Liversidge and Molleson’s [14] method for the permanent dentition, regression equations for predicting age from deciduous teeth are specifically designed to predict age, something that most methods are not designed for [21].
22
H.F.V. Cardoso / Forensic Science International 172 (2007) 17–22
Compared to the permanent dentition, age estimates derived from deciduous teeth show smaller differences between estimated and chronological age. Variability in estimates is small and between-tooth agreement is also high, which results from faster rates of tooth formation and less age variation in deciduous tooth length [11]. Finally, because the sex of immature skeletal remains cannot be easily determined, age estimates derived from Liversidge and co-workers technique are not affected by methodological considerations of sex. 5. Conclusion Documented collections of subadult skeletons are an important means to assess and evaluate the reliability of methods of age determination. This study shows that accuracy of developing deciduous tooth length for age estimation is higher than of developing permanent tooth length, because of the faster rate of change in the deciduous dentition. It is also more accurate than methods based on the chronology of stages of tooth formation and it provides a wide range of ages for all deciduous teeth. Because it is an expedient method it is an important alternative to microstructural methods when teeth cannot be destroyed. Although there are some concerns on how the accuracy of the method is affected by the reliability of radiographic tooth length measurements, it is a strongly recommended method for use in osteological research. References [1] H.M. Liversidge, T. Molleson, Variation in crown and root formation and eruption of human deciduous teeth, Am. J. Phys. Anthropol. 123 (2004) 172–180. [2] M. Nystro¨m, L. Peck, E. Kleemola-Kujala, M. Evalahti, M. Kataja, Age estimation in small children: reference values based on counts of deciduous teeth in Finns, Forensic Sci. Int. 110 (2000) 179–188. [3] M. Muller-Bolla, L. Lupi-Pe´gurier, G. Quatrehomme, A.M. Velly, M. Bolla, Age estimation from teeth in children and adolescents, J. Forensic Sci. 48 (2003) 140–148. [4] R. Kronfeld, I. Schour, Neonatal dental hypoplasia, J. Am. Dent. Assoc. 26 (1939) 18–32. [5] C.F.A. Moorrees, E.A. Fanning, E.E. Hunt, Formation and resorption of three deciduous teeth in children, Am. J. Phys. Anthropol. 21 (1963) 205–213.
[6] H.M. Liversidge, B. Herdeg, F.W. Ro¨sing, Dental age estimation of nonadults: a review of methods and principles, in: K.W. Alt, F.W. Ro¨sing, M. Teschler-Nicola (Eds.), Dental Anthropology—Fundamentals, Limits, and Prospects, Springer-Verlag, Wien, 1998, pp. 419–442. [7] A. Boyde, Estimation of age at death of young human skeletal remains from incremental lines in the dental enamel, Excerpta Med. Int. Congr. Ser. 80 (1963) 36–46. [8] C.M. Fitzgerald, J.C. Rose, Reading between the lines: dental development and subadult age assessment using the microstructural growth markers of teeth, in: M.A. Katzenberg, S.R. Saunders (Eds.), Biological Anthropology of the Human Skeleton, Wiley–Liss, New York, 2000, pp. 163–186. [9] C.M. Fitzgerald, S.R. Saunders, Test of histological methods of determining chronology of accentuated striae in deciduous teeth, Am. J. Phys. Anthropol. 127 (2004) 277–290. [10] D. Deutsch, O. Tam, M.V. Stack, Postnatal changes in size, morphology and weight of developing postnatal deciduous anterior teeth, Growth 49 (1985) 202–217. [11] H.M. Liversidge, M.C. Dean, T.I. Molleson, Increasing human tooth length between birth and 5.4 years, Am. J. Phys. Anthropol. 90 (1993) 307–313. [12] S.R. Saunders, C. DeVito, A. Herring, R. Southern, R. Hoppa, Accuracy tests of tooth formation age estimations for human skeletal remains, Am. J. Phys. Anthropol. 92 (1993) 173–188. [13] H.M. Liversidge, Accuracy of age estimation from developing teeth of a population of known age (0–5.4 years), Int. J. Osteoarchaeol. 4 (1994) 37–45. [14] H.M. Liversidge, T.I. Molleson, Developing permanent tooth length as an estimate of age, J. Forensic Sci. 44 (1999) 917–920. [15] H.F.V. Cardoso, The collection of identified human skeletons housed at the Bocage Museum (National Museum of Natural History) in Lisbon, Portugal, Am. J. Phys. Anthropol. 129 (2006) 173–176. [16] I. Schour, M. Massler, The development of the human dentition, J. Am. Dent. Assoc. 28 (1941) 1153–1160. [17] G. Gustafson, G. Koch, Age estimation up to 16 years of age based on dental development, Odontol. Rev. 25 (1974) 297–306. [18] H.M. Liversidge, F. Lyons, M.P. Hector, The accuracy of three methods of age estimation using radiographic measurements of developing teeth, Forensic Sci. Int. 131 (2003) 22–29. [19] H.F.V. Cardoso, Patterns of growth and development of the human skeleton and dentition in relation to environmental quality, PhD Thesis, Department of Anthropology, McMaster University, Hamilton, Ontario, 2005. [20] J.J. Warren, T. Yonezu, S.E. Bishara, Tooth wear patterns in the deciduous dentition, Am. J. Orthod. Dentofac. Orthop. 122 (2002) 614–618. [21] B.H. Smith, Standards of human tooth formation and dental age assessment, in: M. Kelley, C.S. Larsen (Eds.), Advances in Dental Anthropology, Wiley–Liss, New York, 1991, pp. 143–168.