Introduction

There has been a significant increase in illegal migration during the last 20 years. Many migrants do not have identification documents or a record of their date of birth. Age estimation (AE) is important for legal, social, forensic, and archaeological purposes.

Dental age estimation offers a biological state that mimics chronological age thus enabling the following:

  1. 1.

    Estimation of the chronological age (CA) of an individual with unknown date of birth or missing identification documents. In addition to the point estimate of age, it is important to indicate the uncertainty associated with the estimate. Age estimation of these individuals is necessary to establish eligibility for civil rights, education, or social benefits.

  2. 2.

    A further need for AE is in criminal proceedings where the judicial process is determined by the age of the accused.

  3. 3.

    To provide reference standards for different ethnic, ancestral, racial, and identifiable human groups.

The United Nations Convention on the Rights of the Child (UNCRC 1989) [1] is a widely recognised international treaty that has changed and improved the way children are perceived across the globe. Young people aged less than 18 years are cared for under regulations pertaining to minors. There are large numbers of refugees of uncertain age who do not have birth records. Age estimation is necessary to ensure that social care, healthcare, and education are appropriately provided. There is also a safeguarding issue because adult asylum seekers who are over 18 years but claiming to be under 18 years of age should not be domiciled in the same accommodation as minors. DAE may assist in cases of child trafficking. Underage girls have been forced into marriage, and in these cases, traffickers claim that the children are older than they appear.

Age estimation may be important in criminal proceedings as some criminals, to avoid being punished as adults, claim to be younger than they look. This is so these young adults serve a sentence under the juvenile penal code. In Qatar, the age of criminal responsibility for children is between 7 and 17 years of age. This is stipulated by the Supreme Judiciary Council in Qatar.

There have been many attempts to devise schemes to estimate the age of children [2, 3] In a paper published in 1973 [2] it was stated that “ … it should be remembered that the sample on which they are based is of entirely French-Canadian origin. The dental maturity scores for a given chronological age may well be greater or less in other populations …” This is an important and often overlooked statement.

The purpose of this study is to use extant dental panoramic tomographs (DPT) to prepare a reference data set (RDS) for Qatari children, adolescents, and young adults to create an ethnic-specific dataset to enable assessment of dental maturity in Qatari subjects and from this to estimate dental age (DA) with the associated range of uncertainty.

Materials and methods

Reference data set

The RDS comprised 1600 DPTs of female and male Qatari children, adolescents, and young adults. The DPTs were retrieved from the radiological archives of Rumailah Hospital and Al-Wakra Hospital outpatient departments as well as an orthodontic office. The chronological age of each subject was calculated using the date of birth and the date of radiographic exposure of the DPT. Digital copies of the DPTs were transferred from Qatar using a portable password protected storage device. This was to ensure and protect anonymity. Ethical approval was granted by the Hamad Medical Corporation Research Centre in Qatar [Research protocol #1518/15 “Dental Age Assessment” Ref No. MRC 1535/2016.].

Exclusion criteria

  • Non-Qatari subjects

  • Poor quality DPTs

  • Subjects with a medical condition that may affect dental maturation

Tooth development stages

These were assessed using the 8 stage system described by Demirjian, Goldstein, and Tanner in 1973 (Fig. 1 and Table 1) [2]. The date of birth and sex was blinded from the assessor at the time of the assessment.

Fig. 1
figure 1

The 8 stages of dental development described by the Anglo-Canadian research team. These descriptions should be used when viewing dental panoramic tomographs to determine the stage of development from A to H

Full size image
Table 1 Written guide of morphologically related descriptions of Demirjian’s TDS
Full size table

The TDS were recorded on a specially designed form and then entered into a Microsoft Access database. The assessment included all the teeth on the left side of both the maxilla and mandible. Teeth that were fully developed (stage H) were recorded but excluded from calculations for DAE (Fig. 2).

Fig. 2
figure 2

DPT for subject ID 47 female showing the TDS assessed for this subject. Teeth on the left side only are used. The tooth development stages are indicated in the upper case type on the image

Full size image

Observer agreement

The primary investigator (NAO) recorded the TDSs on the left side of 10 DPTs, comprising 160 assessments. These DPTs were separate from the main study. The assessments were on 2 occasions 2 weeks apart to determine within observer agreement (WOA). A second observer (VSL) assessed the same DPTs to determine between observer agreement (BOA). Both the WOA and the BOA were calculated using Cohen’s kappa [4], and the outcome index was assessed according to the categories suggested by Landis and Koch in 1977 [5].

Data processing

The major part of the study was to establish a RDS for the young Qatari population. The TDSs (A through H) of each tooth present on a sample radiograph were entered into a Microsoft Access database. To extract data from the database, queries were created which included the following information: ID number, ethnicity, sex, and age at assessment (AaA) of each TDS.

These queries were exported to Excel and a small worksheet was created for each TDS for both females and males, for example, UL8Gf [Upper Left Third molar, Stage G, female]. Each of these queries comprised the dataset for that specific stage from which the number (n-tds), the mean (\(\overline{\mathrm{x} }-\mathrm{tds}\)), standard deviation (sd-tds), 0th%-ile (minimum), 25th%-ile, 50th%-ile (median), 75th%-ile, and the 100th%-ile (maximum) were calculated.

Validation study

The second part of this work was to assess the accuracy of the method of DAE when utilizing the Qatari RDS. The validation sample (VS) comprised 100 DPTs of Qatari children and adolescents. There were 50 females and 50 males with known date of birth and date of radiographic exposure. These data were blinded from the assessor. The radiographs were separate from the RDS. The TDS from each tooth on the left side was entered into an Excel worksheet which semi-automatically looked up the n-tds, \(\overline{\mathrm{x} }-\mathrm{tds}, \mathrm{and\;sd}-\mathrm{tds}\), from the RDS to calculate the DA to enable comparison of the CA with the DA for each VS subject [6]. When the 50 female and 50 male subjects had been assessed, CA and DA were compared using Student’s t test.

The distribution of individual TDS for subject 47f, gives the estimated DA with the variation of the individual TDS (Fig. 3). In addition to this formal statistical test, Bland–Altman graphs were used to show the variation of the difference between CA and DA around the line of no difference (Figs. 4 and 5) [7].

Fig. 3
figure 3

Forest plot showing the variation around the median value for DA using the 25th to 50th percentiles

Full size image

Results

Sample distribution

The number of female and male subjects by 1 year age bands is shown in Table 2.

Table 2 Numbers of subjects contributing radiographs by 1 year age spans A total of 1,597 subjects spread fairly evenly between females and males comprised the DPT from subjects resident in Qatar
Full size table

Observer agreement

The WOA score for TDS was 0.9203 and the BOA score for TDS was 0.8401 demonstrating an almost perfect agreement [5].

Reference data set

Radiographs of 1597 Qatari children, adolescents, and young adults age between 5 and 25 years were assessed by the main investigator (NAO) to create the RDS. The data for females and males across the whole range of the age are shown in Table 3 comprising normal distribution summary statistics and percentile distribution summary statistics.

Table 3 Qatari RDS for 782 females and 815 males
Full size table

Worked example

The detailed procedure for the computer-assisted procedure to estimate individual DA is detailed in a paper published in 2018 [6].

Validation study

The data extracted for each subject are illustrated in the worked example (Table 4). The VS of 100 subjects comprise 50 females and 50 males age between 5 and 25 years with known date of birth and date of radiographic exposure. The results from the calculations for the whole of the VS are shown separately for females (Fig. 4) and males (Fig. 5).

Table 4 Worked example. Data extracted for subject 47. The Demirjian TDS are indicated as B, E, F, G, H. The data for these stages are extracted from the RDS (Table 3). AaA, age at assessment
Full size table
Fig. 4
figure 4

Bland–Altman plot demonstrating that the spread of DA values is evenly distributed around the mean value of the CA (for females)

Full size image
Fig. 5
figure 5

Bland–Altman plot demonstrating that the spread of DA values is evenly distributed around the mean value of the CA (for males)

Full size image

The mean CA for females was 10.8 years, sd 1.74 years, the lower confidence limit was 10.31 years, and the upper confidence limit was 11.29 years.

The mean difference between CA and estimated DA is 0.40 years or an overestimate of 4.84 months (Table 5).

Table 5 The validation sample comparison of CA and DA for females
Full size table

The mean CA for males was 10.89 years, sd 1.82 years, the lower confidence limit was 10.37 years, and the upper confidence limit was 11.41 years.

The mean difference between CA and DA for males is an overestimate of 0.38 years or 4.55 months (Table 6).

Table 6 The validation sample comparison of CA and DA for males
Full size table

Bland–Altman graphs

These graphs [7] explore the agreement between CA and DA for both female and male subjects. The graphs (Figs. 4 and 5) demonstrate the difference between the CA and DA within the female and male VS groups. The horizontal dotted line is located at the mean difference between the CA and DA whilst the continuous line represents the “no difference” value between CA and DA. In both the female and male groups, the mean difference was statistically significant.

Discussion

The aim of DAE is to estimate the chronological age of an individual with unknown date of birth. It may be used for clinical, legal, civil, criminal, and forensic purposes. An important consideration is the error of the method. The preferred term is the uncertainty associated with the estimate. This is easily and reliably achieved using Draft Quicksheets [6].

In the present study, only the tooth development stages, which comprise the first part of the Demirjian, Goldstein, and Tanner method [2], were employed. This is because the mathematical procedures leading to the maturity score cannot be unravelled [8]. In the SAM procedure, all the developing teeth on the left side including third molars were used. This has increased the number of assessed teeth and therefore likely to lead to an improvement in the accuracy or reliability of the estimated age as it uses the maximum information available. This is born out to a considerable extent by the relatively small CA minus DA differences shown in the results of the validation study. In addition, the inclusion of third molars up to and including stage G makes the method suitable for older subjects aged 16 years or more. Stage H is not used as it is unbounded in its upper border for each tooth. This will be the subject of a further communication because the age of apex closure cannot be identified without censoring [8]. Also, the mean value for stage H cannot be used in simple mathematical procedures such as averaging the AaA of TDSs present in a subject. Therefore stage H is excluded from any calculations in SAM.

Each ethnicity or ancestral group has its own characteristics including with DAE. Many studies investigate the validity of using the French-Canadian standards on other populations [9,10,11]. However, it was found that overestimation or underestimation of age was present in many studies. From those findings developed the concept of establishing ethnic specific RDSs which is important for each specific Identifiable Human Group or Ancestral Group.

In this study, a RDS with the mean AaA of different TDS for the Qatari population was developed. The mean age at attainment is the term often used in DAE research reports. This is logically inconsistent as the age at which a stage has been attained cannot be ascertained from a single “snapshot” radiograph of developing teeth. Age at assessment (AaA) is the logical and better understood term.

The data collection was limited to Qatari subjects and partitioned by sex. The large number of DPTs results in a sufficiently large number in each 1 year age span and consequently for the individual TDS to ensure a narrow range of the CI of the differences between females and males. In practice, over 73% of the AaA values for the individual TDS were earlier in females. This emphasizes the importance of sex-specific data in addition to the need for ethnic specific data.

This work is part of a large dental age estimation project looking at different ancestral or identifiable human groups [6, 12, 13]. The main purpose of the study was to establish a RDS for the Qatari population. This was achieved using the 8 stage system of TDSs. A strong justification for this is that it has been shown to return the highest WOA and BOA using Cohen’s kappa. The scheme is practical, reliable, and easy to use as has been shown by the German research team [14]. Other publications using SAM have shown consistent accuracy within 6 months of the CA [13, 15].

The second part of the Demirjian, Goldstein, and Tanner method [2] has not been used as it has been suggested that age estimates based on the French-Canadian population are not applicable to other populations [2]. The method used here has been named as the simple average method (SAM) as it uses all the tooth morphology types without any idiosyncratic weighting, and on logical grounds, it is intuitively more informative to use as much information by using all tooth morphology types available [6].

A simple assessment to emphasize the need for ancestral specific data is to compare two TDS at random: Kuwait LL7Gf cf Qatar LL7Gf. These two stages from different Middle Eastern populations provide data as follows – Kuwait LL7Gf n-tds = 95, x-tds = 13.56, sd-tds = 1.55; Qatar LL7Gf n-tds = 222, x-tds = 11.65, sd-tds = 0.98. The difference of 1.91 years is highly statistically different. Whether this applies to the other TDSs in these RDSs is a matter for future investigation and report. This raises the important question of the suitability of using apparently similar genetic backgrounds for DAE.

The robust nature of the present study has been supported by the following:

  1. 1.

    A single identifiable human group – Qatari subjects – has been used to create the reference data set.

  2. 2.

    Training and careful management of the WOA and BOA supports the use of the 8 stage TDS system.

  3. 3.

    A validation set was used to test the accuracy of the RDS when used to estimate the dental age of Qatari children, adolescents, and young adults.

Conclusion

A reference data set for the mean age at assessment of the different TDS for the Qatari population has been established. The average difference between CA and DA for female and male subjects was 4.8 months and 4.5 months, respectively.

Key points

  1. 1.

    A large reference data set has been established for Qatari subjects.

  2. 2.

    There is a clear difference between Age at Assessment for females and males as regards Tooth Development Stages.

  3. 3.

    The Simple Average Method provides age estimation within 5 months of the Chronological Age.