Journal Pre-proof

Root Dentin Translucency and International Dental Database: Forensic

methodology for estimation age-at-death in adults using Single-Rooted
Teeth

Roberto C. Parra, Douglas H. Ubelaker, Joe Adserias-Garriga,

Karen J. Escalante-Flórez, Lucio A. Condori, Jane E. Buisktra

PII: S0379-0738(20)30434-5
DOI: https://doi.org/10.1016/j.forsciint.2020.110572
Reference: FSI 110572

To appear in: Forensic Science International

Received Date: 18 July 2020

Revised Date: 10 September 2020
Accepted Date: 29 October 2020

Please cite this article as: Parra RC, Ubelaker DH, Adserias-Garriga J, Escalante-Flórez KJ,
Condori LA, Buisktra JE, Root Dentin Translucency and International Dental Database:
Forensic methodology for estimation age-at-death in adults using Single-Rooted Teeth,
Forensic Science International (2020), doi: https://doi.org/10.1016/j.forsciint.2020.110572

This is a PDF file of an article that has undergone enhancements after acceptance, such as
the addition of a cover page and metadata, and formatting for readability, but it is not yet the
definitive version of record. This version will undergo additional copyediting, typesetting and
review before it is published in its final form, but we are providing this version to give early
visibility of the article. Please note that, during the production process, errors may be
discovered which could affect the content, and all legal disclaimers that apply to the journal
pertain.

© 2020 Published by Elsevier.

Root Dentin Translucency and International Dental Database: Forensic methodology
for estimation age-at-death in adults using Single-Rooted Teeth

Roberto C. Parra1, Douglas H. Ubelaker2, Joe Adserias-Garriga3, Karen J. Escalante-Flórez4,

Lucio A. Condori5, Jane E. Buisktra6.

1
Specialized Forensic Team, Office of the High Commissioner for Human Rights, United
Nations Mission in The Democratic Republic of the Congo (MONUSCO).
2
Department of Anthropology, NMNH, Smithsonian Institution, Washington, DC, USA.
3
Department Applied Forensic Sciences, Mercyhurst University, Pennsylvania, USA
4
Especialidad de Odontología Forense, Facultad de Estomatología, Universidad Científica

of
del Sur, Lima, Perú.
5
Equipo Forense Especializado, Instituto de Medicina Legal y Ciencias Forenses, Ministerio

ro
Público, Lima, Perú
6
School of Human Evolution and Social Change, Arizona State University, Tempe Campus,
AZ, USA.

Highlights
-p
re
 Reviewer #1:

 I congratulate the author(s) fot the comprehensive study.

lP

 We thank you for the kind words

Reviewer #2:
na

 Despite very interesting and solid goals methodology and results

this article must be rewritten completely as it is presented as a thesis report : The text has been
adjusted
ur

 reduction of the lenght of the paper must me drastic (around 50%). The document has been
reduced.
Jo

 introduction must present the current knowledge on Lamendin and derived methods with the
problems to be solved. The introduction has been strengthened according to the
recommendations.

 discussion only limited to the interest of the results and what it brings to the extensive literature
analysis. The discussion has been adjusted according to the recommendations

 conclusion : a few lines about : can we use or not this sample and the bayesian approach in any
place whatever personnels and equipment are at which cost of money and time. Three lines

1
 have been added according to the requirement.in its present state of presentation this very
interesting article is unfortunately boring and almost unitelligible. Thank you for the words,
which have served to strengthen the document.

ABSTRACT

Estimation of the age-at-death in adults is essential when the identification of deceased

persons with unknown identity is required in both humanitarian and legal contexts. However,
the methodologies and the results obtained can be questioned. Various efforts have been
developed to adjust procedures to specific populations, always seeking the precision and

of
accuracy of the methodologies. It is known that the estimation of the age-at-death in adults
coexists with wide margins of error, due to several reasons, including but not limited to

ro
statistical problems, the size of the sample or the physiological process of aging.

This research focuses on a degenerative indicator of the dentin (Root Dentin Translucency)
-p
and its combination with Periodontal height (PH) following the Lamendin´s technique for
estimation of the age-at-death in adults. The main objective of this research was to
re
demonstrate the forensic applicability of a Bayesian model based on a International Dental
Database (FIDB) that include Root Translucency Height (RTH) and PH as a method to age-
lP

at-death in adults. The conclusion of this research was that the combined both indicators
become a generalizable age-at-death in adults model for all human populations, where the
Bayesian method would offer optimal results in any population. In this way, those populations
na

that do not have had the possibility of validating a specific procedure, now have the
opportunity to apply a valid method for estimating age-at-death in adults to global scope.
ur

Keywords:

Population data
Jo

Age estimation
Forensic Odontology
Forensic Anthropology
Methodology validation
Lamendin´s technique

Roberto C. Parra

2
[email protected]

INTRODUCTION

In bioarchaeological and forensic practice, estimating age in adults is a complex issue that
requires a holistic perspective in order to understand and quantify skeletal degenerative
processes [1]. In the forensic field, valid estimates require the ability to recognize the nature
and sources of estimate error [2,3,4]. Forensic age estimation is a key component in the
identification of the dead in both humanitarian and legal contexts [5]. However, the
methodologies applied and the data obtained can be subjected to questioning [1] regarding the
population of reference and the suitability of the method according to the age group of the

of
studied individual. For example, Komar [6] observed that the methods developed based on
population samples from the United States could not be properly applied to the population of

ro
Srebrenica (Bosnia). One possible interpretation is that the population of Eastern Europe is
different from the one formed by European descendants in the United States [7]. Further
-p
forensic studies in the Balkans supported this observation [8], and it was necessary to
recalibrate the procedures for this particular region [9].
re
Similar situations occur in the forensic context with different populations around the world.
Analytical precision and accuracy –particularly in adjusting the methods for specific
lP

subgroups– are always a goal, particularly given the wide margins of error seen in adult age
estimates. These errors can arise due to statistical problems, such as the sample size used on
the methodology, or to the physiological process of aging itself [4]. According to Nawrocki
na

[4: 85], the aging process biases emerge from “investigators underestimating or under-
appreciating the normal range of variation in skeletal indicators, constructing
inappropriately-narrow error intervals, misapplying basic statistical tests, and deriving
ur

overly-specific interpretations from small and poorly-balanced samples.” When new methods
of analysis are introduced into a population, they must be subjected to specific tests to ensure
Jo

its robusticity.

Since the 1980s, researchers have questioned the application and use of the age estimation
methods in adults [10,11,12,13,14,15,16]. Some of this studies have suggested that statistical
methods should be modified in order to address the inaccuracy of some of the age estimation
methods [2,17,18,19,20,21] and that robust samples combining different populational groups
should be used, in order to reveal the general trends of aging [4].

3
Komar and Buikstra [22] and Nawrocki [4] stated the need to identify strategies and
methodological criteria applicable to different human groups, with the purpose of making the
procedure widespread and viable in different forensic contexts like the Demirjian method,
which is based on a large sample of subadults of different ethnic origins [23]. Growth and
developmental process are very consistent, so the population differences on these processes
are minimal [23,24,25,26]. In forensic contexts, methods such as Demirjian can be applied to
different populations because the method is based on a global theoretical understanding of the
development and maturation. Dental age estimation methods estimate age using growth and
development processes for subadults, and degenerative changes in adults [1].

A dental physiological process that has shown a good correlation with chronological age in

of
adult is Root Dentine Translucency (RDT)
[27,28,29,30,31,32,33,34,35,36,37,39,40,41,42,43,44]. RDT is a dentinal degenerative

ro
process that, according to Bang and Ramm [28], appears around the third decade of age as a
result of the accumulation of hydroxyapatite crystals in the dentinal tubules.
-p
This physiological process is an index of aging in the pulp-dentin system that has been
described by various researchers [42,73,74,75,76]. Tang et al. [75] have eloquently
re
summarized this physiological process that “is due to the gradual mineral occlusion of
microscopic tubules that run through the primary dentine. Unoccluded tubules have differing
lP

intertubular and intratubular refractive indices that cause transmitted light to scatter and
cause dentine to appear opaque, whereas the occluding material has a refractive index
similar to the intertubular dentinal matrix, allowing light to pass unscattered and producing
na

the appearance of translucency” [75: 332].

However, some studies have reported evidence of RDT , even in individuals as young as 18
years old [44,45,46,47,48,49,50,51,52], although it seems not to be releated to a normal
ur

degenerative process. Some morphometric studies of RDT have been carried out on sectioned
teeth [28,30,41,42,69,71,77,78] and others on intact teeth [28,45,49,50,51,53,60]. RDT
Jo

assessment techniques quantifying RDT using a caliper [45,48,49,50,51,53]. Drusini et al.

[60] proved that RDT can be accurately measured using a caliper and the results are just as
optimal as the ones obtained using complex computerized methodologies. Regarding the light
quality required to observe RDT, Adserias-Garriga et al. [62] measured RDT applying three
different lights: 6500 lx (microscopic light), 3000 lx (negatoscopic light) and 1600 lx
(equivalent to daily sunlight). This study established that the 1600 lx lighting is the most

4
adequate setting to measuring RDT and it concluded that lighting should be taken into
account to obtain an accurate age estimate.

In 1970, Bang and Ramm [28] studied RDT in a sample of 926 teeth from 265 known age
individuals from Norway, and developed a lineal regression of RDT measurements in
millimeters [29,54]. Furthermore, they expanded the method to include intact dentition [28].
In 1978, H. Lamendin [55] analyzed the correlation between RDT and chronological age,
concluding that the estimation improves when root translucency is combined with total root
length. Later, the same author simplified Gustafson's classic study [56], reducing Gustafson’s
[29] seven variables to just two (RDT and Periodontosis). In 1992, Lamendin et al. [53],
simplified the method further by eliminating the need for histological sections and using

of
intact teeth instead. This streamlined procedure is known as the Lamendin´s technique.

In 2002, Prince and Ubelaker [49] applied the Lamendin´s technique [53] on a diverse

ro
skeletal sample from the Terry Collection at the Smithsonian National Museum of Natural
History in Washington, D.C. This study presented a modification of the original method with
-p
specific formula for sex and ancestry, and reported higher precision among individuals
between 30 and 69 years of age.
re
In 2006, Sarajlić et al., [50] compared the Lamendin´ method and the Prince and Ubelaker
method in 415 single-root teeth from 100 individuals of known age and sex. The human
lP

remains were exhumed from eight sites in Bosnia and Herzegovina. Both models generated
the lowest mean error in individuals 20 to 49 years of age, and the authors concluded the
Lamendin´s method was more accurate for the studied sample.
na

In 2007, González-Colmenares et al., [45] studied a Mediterranean sample from Spain and a
dental collection obtained during autopsies conducted in Colombia. The study demonstrated
ur

the usefulness and applicability of dental traits for estimating age in these sample. The Prince
and Ubelaker [49] method offered higher precision and accuracy than the original method
Jo

proposed by Lamendin et al. [53] in this sample.

In 2008, Ubelaker and Parra applied the procedures based on intact single-rooted teeth [28,53]
to a sample collected during autopsies conducted at the Institute of Legal Medicine and
Forensic Sciences of Peru [51]. The study found that all methods obtained good results, with a
mean error of 8.3 years for the model proposed by Lamendin et al. [53], 7.6 years for the
Prince and Ubelaker method [49] and 8.8 years for the Bang and Ramm method [28].
Ubelaker and Parra [51] confirmed that these methods could be applied in Andean

5
populations, and that the impact of population variation on these methods was minimal
applying the Prince and Ubelaker formula for white individuals.

In 2008, Prince and Konigsberg presented the results of two types of statistical models
(Inverse calibration and classical statistics) using RDT and periodontosis. The sample
included 401 individuals killed during hostilities in Kosovo [48]. They concluded that the
Bayesian method showed the lowest mean error (1.56 years) in comparison to the inverse
calibration models of 4.85 years for Lamendin et al. [53] and 5.27 years for Prince and
Ubelaker [49]. This study also confirmed that the methodologies were valid for Eastern
European populations, as initially suggested by Sarajlić et al. [50].

In 2010, Schmitt et al. [57] established a new method to assess age-at-death by evaluating

of
RDT and periodontosis using a single tooth. They used three statistical types: linear
regression, multinomial regression and the Bayesian approach. Their results showed that the

ro
correlation between these two parameters and chronological age was low, when linear
regression was applied, the mean of the standard error of individual prediction was 13.67
-p
years which is clearly higher than the results of Lamendin et al. [53], Prince and Ubelaker
[49], and Prince and Konigsberg [48].
re
The application of the Lamendin´s method has been validated in different population groups,
and the evidence collected to date suggests that the impact of population variation is minimal
lP

[58,59]. Sex dimorphism should be taken into account whenever possible, although the
exclusion of this data does not seem to be a limitation for the use of these methods
[45,48,51,53,58,60].
na

On the other hand, properties and visual assessment of RDT may be affected by taphonomical
forces as water, soil chemical composition, pH, temperature and humidity, and faunal,
ur

scavenger microorganisms such as fungal, or bacterial activity [38,106,107,108;111]. Micro-

environment must be an important determinant of post-depositional change [109]. The post-
Jo

mortem environment and post-mortem interval play an important role in dental tissue
diagenesis which may constitute a limitation to use RDT for age-at-death estimation of adult
individuals [21,38,76,95,106].

In 1995, Lucy et al. [21] noted changes in the internal microstructure that obliterated
translucency dentin of sectioned tooth roots from a mediaeval cemetery. They highlighted that
such changes were generated by the effect of diagenesis and that these effects could alter
RDT. Some of these post-mortem changes have been recorded as diffuse decay and eroded

6
surfaces [38]. Poole and Tratman [98] have pointed out that there may be colonization and
proliferation of microorganisms during the post-mortem interval. Although the exact fungal
group is not yet known, data suggest that it is an acidogenic microorganism highly harmful
for the dissolution of inorganic substances. Sillen [111] suggested that acidogenic
microorganisms attack collagen and excrete organic acids. Such mechanisms can cause
dissolution of the inorganic component and destruction of the histological structure.

Sengupta et al. [38] studied the RDT of the Spitalfields collection (London), consisting of
individuals who lived during the 19th century. They pointed out that most teeth had a clayish
or chalky appearance, and that these characteristics had a negative impact on age estimation;
although Tang et al. [76] noted that the observations developed by Sengupta and colleagues

of
[38] did not appear to have a relationship with the post-mortem interval and instead was
thought to be caused by the post-mortem environment. In 1999, Kvaal y During [106]

ro
conducted a study on remains that had been underwater on a sunken ship for over 300 years.
They noted that root translucency was affected by the aquatic environment. The nature of
-p
these biochemical mechanisms of action are not well known.

Lamendin and Bang and Ramm techniques are very useful when applied to samples from
re
recently deceased individuals [44,45,47,48,50,93], but their usefulness may be limited in
cases of longer PMI in which diagenesis has affected the remains. Dentinal diagenetic
lP

changes have been found by several researchers to significantly reduce our ability to predict
age through RTH in bioarchaeological context [21,38,76,94,95,96,97,106] but, although
translucency is observed, there are also findings of wide biases that would be impacting the
na

results, not for physiological reasons, but rather for external forces [95]. On the other hand, if
translucency is not observed in intact historical teeth with may therefore not always indicate a
young age but rather an advanced degree of postmortem change [76].
ur

Given this wide background on this topic, two questions arise: i) are the combination of RTH
and PH a forensic valid indicator of adult age regardless of geographical origin and ii) what is
Jo

the margin of error for each variable independently and together? To answer these questions,
an Forensic International Dental Database (FIDB) of RTH and PH values of various
population groups was created. The FIDB includes 693 individuals of different phenotypic
populations: African-American and European-American [49], European-Mediterranean [45],
Colombian mestizos [45] and Peruvian mestizos [51] and Peruvian Andean Quechuas.

7
The main goal of this study is to determinate if RTH and PH could be applied in forensic
contexts in individuals regardless their ancestral origin. To test this RTH and PH was first
calculated in relation to sex, age, ancestry, and sample provenience. The second goal of the
study was to calculate the precision and accuracy of five methods of age estimation using
RTH and PH, and one that uses only RTH. These results were compared with a Bayesian
model (INT Bayesian) based on the FIDB [58]. A third goal was to control these results with
a sample that was not previously included in the FIDB. For this, a sample of Colombian
mestizos has been examined.

MATERIALS AND METHODS

The sample included individuals of known age, sex and ancestry from four countries. The

of
sample ages ranged from 21 to 99, with an average age of 51.18 years and a standard
deviation of 15.50 years.

ro
The sample was drawn from studies conducted by Prince and Ubelaker [49] on African-
American/Black and European-American/White individuals, González-Colmenares et al. [45]
-p
for White Spanish Mediterranean individuals as well as Colombian Mestizos, and Ubelaker
and Parra [51] on Peruvian Mestizos. The measurements of these studies were taken following
re
the technique outlined in Lamendin et al [53]. Prince and Ubelaker [49] and Ubelaker and
Parra [51] used a light box (3000 lx) and in the case of González-Colmenares et al. [45] a
lP

negatoscopy light was used (3000 lx). Additionally, a sample that was collected from 36
corpses of Peruvian Quechua Andean ethnic affiliation was included. The sample
corresponded to 26 males (with an age range between 23 and 58 years, average age of 40.54
na

years and standard deviation of 10.82 years) and 10 females (with an age range between 23
and 63 years, average age of 46.83 years and standard deviation of 12.83 years). The
measurements of these studies were taken following the technique outlined in Lamendin et al
ur

[53] and using a light box (3000 lx).

Table 1 shows the distribution of the FIDB sample by sex (281 females and 412 males)
Jo

according to phenotype. Female individuals represent 40.5% of the sample, including 40

Peruvian mestizo individuals (5.8%); seven Colombian mestizo individuals (1%); 34 White
Spanish Mediterranean individuals (4.9%); 100 European-American/White individuals (14%),
and 100 African-American/Black individuals (14%). The 412 males included in the study
represent 59% of the total sample, including 96 (13%) Peruvian mestizos; 71 (10%)
Colombian mestizos; 45 (6.5%) White Spanish Mediterraneans; 100 (14.4%) European-
American/White individuals, and 100 (14.4%) African-American/Black individuals.

8
On the other hand, a control sample of 150 Colombian mestizos was considered. This sample
included 114 males (with an age range between 20 and 93 years old, an average age of 45.26
years and a standard deviation of 20.73 years) and 36 females (with an age range between 21
and 81 years old, an average age of 44.81 years and a standard deviation of 16.26 years). The
sample was drawn from studies conducted by Escobar and Sanabria [61].

Methodology

RH, PH and RTH were measured with a digital caliper (with the values recorded in
millimeters). A light-box was used to illuminate the RDT (Figures 2 and 3) as it has a
significant influence in measurement quality [62]. Figure 1 shows the variables defined and

of
recorded, according to the proposed of Lamendin et al. [53] and Bang and Ramm [28].

Descriptive data were obtained using SPSS v.25 and analysis was performed using MiniTAB

ro
v.18. Measurements were obtained by analysts with extensive experience in the technique and
support the studies by Prince and Ubelaker in 2002 (PU) [49], González et al. in 2007
-p
(GBMF) [45], and Ubelaker and Parra in 2008 (UP) [51]. These studies show that the impact
of intraobserver and interobserver error is minimal.
re
The methods proposed by Lamendin et al. [53] (LBHTNZ), Prince and Ubelaker [49] (PU),
González et al. [45] (GBMF), Ubelaker and Parra [51] (UP) and Sarajlić et al. [50] (SCKSBT)
lP

were applied to the FIDB in order to evaluate the accuracy and precision of the mentioned
methods. Additionally, the regression equations for intact teeth presented by Bang and Ramm
[28] (BR) were also tested. Finally, the RTH, RH, and PH variables were evaluated for
na

different age cohorts within the sample.

Age Estimation
ur

Bayes’ theorem was utilized to estimate age-at-death according to Lamendin’s technique. A

Jo

Bayesian approach is based on three important concepts: prior probability, the likelihood, and
posterior probability [48, 63]. A Mathematical formulation of a Bayesian regression model
has been developed:

In the following Multiple Linear Regression model:

(1)

9
The estimated model is:

(2)

Where:

(3)

The maximum likelihood estimators for the model parameters are also least squares
estimators. This statement is verified when the model errors are normally and independently
distributed.
Assuming that the model errors are distributed as follows:
∼ (4)

of
for the data analyzed in this study, the results indicate a normal distribution (see Figure 10),

ro
for both the parameter estimators and the response variable.
~ (5)

As the random variable depends on the parameter

-p , the following notation is used:
re
(6)
lP

Where

Then, the likelihood function of is

na

(7)
ur

To find a posteriori distribution an a priori distribution is required (although it is considered

an informative a priori.) Having previously concluded that ~ , the a
Jo

priori is:

(8)

Then:

10
 (9)

Where are known and are symmetrical matrices.

To calculate a posteriori distribution; Bayes' theorem:

(10)

is utilized, substituting (9) and (7) in (10)

 1   1 
P( / Y )  exp   (   )´V 1 (   )  exp   (Y  X )´C 1 (Y  X ) 
 2   2 

of
 1 1 
  ( ´V    ´V   ´V   ´V   Y ´C Y 
1 1 1 1

P( / Y )  exp  2 
  Y ´C 1 X   ´ X ´C 1Y   ´ X ´C 1 X 

ro
 
 1 
P( / Y )  exp   ´(V 1  X ´C 1 X )   ´(V 1   X ´C 1Y )  ( ´V 1  Y ´C 1 X ) 
 2    -p   
escalar escalar 

The second term is the transposition of the third term in the exponential, and it is verified in:
re
, then
lP

(11)
na

Operating:

(12)
ur

Substituting in (1):
Jo

(13)

Completing squares in (13), since is a quadratic form.

Allusion:

11
Shaping the exponential in (13):

(14)

of
The result is:

ro
(15)

In consequence,
-p
(16)
re
Substituting (12) in (16), a posteriori distribution of the parameter vector of the Bayesian
lP

regression is:

(17)
na
ur

The likelihood given by (6) represents the sample data. It is assumed that this sample was
Jo

generated by a random parameter vector from a multiple regression, but with a distribution
assumed to be multivariate normal, according to the results of the analysis of the parametric
regression residuals.

This function depends on the matrix , which is the group of dental variables

considered (RH, PH, and RTH). It also depends on , which is a vector whose elements are

the ages of the individuals corresponding to the rows of . Finally, it depends on , which

12
is the variance of the errors of the classical regression model and which can be estimated from
the sample as the residual variance.

The a priori distribution (8) represents the information we already have, which in our
case is comprised by all records (693 cases). This consideration might be justified by
recognizing the sample as historical information. This a priori distribution contains
information on the distribution of the parameter vector that generated the data, which was
assumed to be multivariate normal, and depends on which can be estimated from the total
,
sample by least squares such as . It also depends on , which

can be estimated by , where is the residual variance given by

of
ro
.

-p
In the following section, this model will be compared to others using its prediction
errors. A model can be said to fit and forecast better if its prediction errors are lower on
average compared to others. These mathematical functions were calculated using the “R
re
4.0.2” software (http://www.r-project.org).
lP

Before developing the regression model, influential observations were identified in the sample
(5.48% of the total sample), using the DFFITS method [64]. These cases were removed from
the analysis to improve regression model fitness. After removing the outliers, regression
na

coefficients did not change dramatically, suggesting the adjustment trend is stable. The
regression model was used to estimate the likelihood of the Bayesian model.
ur

To determine the error made by the INT Bayesian method compared to other methods
[28,45,49,50,51,53], the absolute residuals for each model were evaluated, also the average of
Jo

the residual and absolute residuals by age group. These data allow us to visualize the possible
biases of the methods for estimating age. This information can be compared with the reports
that were published [28,45,49,50,51,53]. Likewise, the aforementioned studies have reported
in their results that the impact of intra-observer and inter-observer error is minimal. In this
study these error factors were re-evaluated in a sample of 47 teeth that were randomly
selected from the control sample of Colombian mestizos. The test used to determine the
concordance between observers was the Intraclass Correlation Coefficient (ICC).

13
Measurements were taken at one-week intervals by three independent observers for
interobserver error, and at three hours intervals for each observer for intra-observer error.

The accuracy and precision of the INT Bayesian method will be evaluated in the
Colombian population (control sample) if there are differences between the absolute error of
INT Bayesian and the other models [45,49,53] by age groups. It will thus be possible to verify
the accuracy, precision and generalizability of the Bayesian model.

RESULTS

Table 2 shows the descriptive statistical data of root height (RH), periodontal height (PH),
and root translucency height (RTH) according to age category.

of
An analysis of variance was conducted, indicating that phenotype and sex do not

ro
simultaneously influence RH, PH, and RTH in a significant way. RH is influenced by the
phenotype with a p-value of 0.011 and sex with a p-value of 0.000. No remarkable significant
difference is observed, but there is a relationship between the age and the height of the root,
-p
there is no notable difference in the interaction of sex phenotype. PH presents significant
differences due to the age covariate with a p-value of 0.000 and the phenotype factor with a p-
re
value of 0.00. It does not influence the sex of the studied individual; in the same way it does
not influence the phenotype-sex interaction. RTH is directly affected by the age covariate
lP

with a value of 0.00 and the phenotype factor with a p-value of 0.00, no differences have been
observed with sex and the interaction between phenotype and sex.
na

Results presented in Figures 4-6 demonstrate that phenotype significantly influences all dental
variables. Sex has a significant effect only on RH, which is the least effective variable for
estimating age with an of R2 0.04 being the lowest than the other variables since for PH is R2
ur

0.35 and RTH R2 0.72. As expected, RTH is the variable with the highest relationship with its
factors and covariant under study.
Jo

Figures 7-9 report the relationship between provenience and sex in RH, PH, and RTH. The
results show us that the RH is higher in the female groups, especially in the African-American
black phenotype, and lower in the Peruvian mestizos, both males and females, but more
marked in males. The PH is higher in almost all males except the Peruvian mestizo where it is
lower. The European whites is the group where the PH measurement is highest and the lowest
measurement is observed in Colombian mestizo, both women and men. In the RTH
measurement it is appreciated that in the group of Mediterranean whites it is the highest

14
measurement and on the other hand the lowest in female Colombian mestizos followed by
male Peruvian mestizos.

Several constants were identified within the regression analysis. For every millimeter of
increase in RH, age increases on average 0.665 years. For each unit of increase in PH, age
increases on average 0.312 years. Finally, for each unit of increase in RTH, age increases on
average 0.431 years, plus the regression constant of 13,028 years. Figure 10 shows that the
residuals behave as if they had been extracted from a population with a standard normal
distribution.

Table 3 summarizes the regression analyses, after removing outlying observations, model
fitness improved from R2 0.64 to R2 0.72. The standard deviation decreased from 9.32 to 7.73

of
years and the average residuals by age range were reduced. Likewise, the constants were
identified within the regression analysis for the model without outliers. Taking all these

ro
considerations into account, the INT Bayesian regression method can be expressed by the
following equation:
-p
AGE = 13.1279052 + 0.6620858(RH) + 0.3198992(PH) + 0.4297236(RTH)
re
Where the confidence intervals for the model coefficients:
lP

* 95% Confidence Interval for Constant: [9.90174 - 16.35407]

* 95% Confidence Interval for RH: [0.4539199 - 0.8702517]
* 95% Confidence Interval for PH: [0.2765968 - 0.3632015]
na

* 95% Confidence Interval for RTH: [0.4076154 - 0.4518318]

The confidence intervals for the Bayesian regression coefficients for PH and RTH are
ur

relatively small, which indicates that these estimated coefficients are very close to the
coefficients of a population regression. In the meantime, Figure 11 shows the comparison of
Jo

the mean absolute error between the LBHTNZ, PU, GBMF, UP, SCKSBT, BR and INT
Bayesian method by age range with influential observations (FIDB = 693 individuals). As
indicated in Figure 12 the UP model offers predictions with more homogeneous absolute
errors than the other models; however, the INT Bayesian method does the same but with
lower errors, except for the age range of 80 years and older (18.03 years).

Figure 13 compares mean absolute residuals where the INT Bayesian model offers the lowest
error (6.83) and that its predictive behavior is more homogeneous than the other methods.

15
This result indicates that when doing the prediction for any age, its estimate is more efficient
on average than the ones obtained through the other models. To determine if the Bayesian
(INT) model offers the best statistical fit, an analysis of variance was conducted. The
probability that the averages of the absolute residuals are equal is 0.000 (null); therefore, the
analysis of variance of a factor indicates that there is evidence that at least one average is
different from the others. Tukey's test indicates that the UP and BR models have statistically
equal absolute error averages and are also the highest, while the other models, including the
Bayesian one, have statistically equal absolute error averages and are the lowest.

Figure 14 and Table 4 indicate that models tend to overestimate slightly age for individuals
younger than 50 years and underestimate for those older than 70 years. The UP model offers

of
more homogeneous predictions with absolute errors than the other models; however, the INT
Bayesian method does the same, but with fewer errors, except for the age range of 80 years

ro
and older with an average error of 18.03 years. In all the models evaluated, greater error is
made in the ages over 80 years, being more prominent in the LBHTNZ (20.52) and GBMF
-p
models (19.13). Figure 15 shows the dispersion between real age and predicted age using INT
Bayesian, with a correlation of 0.85.
re
On the other hand, in the control sample of Colombian mestizos, the INT Bayesian regression
model obtained a correlation coefficient of real age and estimated age of 0.74. This result
lP

indicates that there is good precision when using INT Bayesian. Furthermore, as shown Table
5 the ICC obtained for both the same observer and the three observers are very high, which
guarantees the reliability of that study.
na

In table 6 shows the evaluation of the absolute residuals of the models that were applied in the
Colombian control population and it was found that there is no data normality. The model that
generates the least mean error with 11.43 years of average error is the INT Bayesian model,
ur

while the maximum error found is that given by the Prince and Ubelaker model with a
difference of 12.13 years average. However, among the age ranges of 30 to 59 years, all the
Jo

models analyzed offer homogeneous results, where INT Bayesian offers the lowest of all,
5.86 year of average error.

DISCUSSION

16
Dental age estimation methods using RDT have proven to be effective [45,48,49,50,51,65,66,
among others]. Such procedures have the potential for broader applications [4]. Numerous
research studies have pointed out that RDT is a physiological process that has an ascending
and linear correlation with adulthood [28,29,30,34,35,40,49,53,67,68,69]. Our study found
that RDT has a high correlation coefficient with chronological age (r = 0.90). Other
researchers have reported similar findings (r = 0.65–0.96) [28,30,37,42,45,69,70,71,72].

RDT has proven to be an important physiological marker of aging and it is only remotely
influenced by pathological disorders [39,42,79] or other factors, including biomechanical
[52], environmental, cultural, and genetic factors [33,52], contrary to what happens with
periodontosis [80,81,82]. Ubelaker and Parra [51] noted that this indicator of periodontal

of
disease must be appropriately considered. Since that the periodontosis could indeed affect the
accuracy of the method as it is conditioned by various aspects, including oral hygiene, cultural

ro
influences and dental pathology [51, 83, 84]. However, Tadjoedin et al. [85] have argued that
periodontal disease tends to be related to age. They also noted that gingival disease and
-p
chronic periodontal disease are observed among late youth and older adults, while an
aggressive modality of periodontitis has been observed in older adults. This study coincides
with the approach taken by Foti et al. [86] and Ubelaker and Parra [51].
re
Despite some clear problems detected, different studies have shown that PH offers similar
lP

margins of error within the samples studied [45,49,51,57]. Other studies have shown that
periodontosis is better correlated with age than RDT [65]. Our research confirms that PH
plays an important role in estimating adult age (r = 0.77), and the impact of periodontal
na

disease on the statistical criteria for the estimation is minimal. An interesting debate on the
factors that influence RTH and PH is presented in Prince and Crowley [52].

Furthermore, some researchers have suggested that one of the main problems in the
ur

application of this technique may be the impact of interobserver error [22,87]. Kimmerle et al.
[9] demonstrated that despite the significant discrepancy among observers who applied the
Jo

Lamendin´s technique, they suggest that the impact of intra and inter-observer biases in the
application of this method may be minimal [45,49,51,62,88]. This research strengthens that
perspective, even that the differences may be insignificant (Table 5).

Ancestry influence
Whittaker and Bakri [89] found differences in the process of RDT formation between Asian
and European populations. These findings suggest that phenotype and geographic

17
provenience may influence the development of RDT. Whittaker and Bakri [89] noted that the
reasons for these differences might be the result of a combination of genetic and
environmental factors. Figure 6 shows that phenotype significantly influences the
development of RTH; however, it is interesting that the margin of error in the estimates does
not influence the result. Our findings suggest that RDT formation is slower in South
American mestizo individuals (Colombian and Peruvian mestizos) than in black African
American groups and white populations of European descent. However, the value of
estimating adulthood using these criteria does not seem to be influenced by population
variation [45,48,49,50,51]. When Ubelaker and Parra [51] studied RDT and periodontosis in
Peruvian individuals, different from European and North American populations, they found

of
that the impact of population variation on age estimation was minimal; even when only one of
the variables (RTH) was used, the results were adequate [51]. The results obtained in our
study support the concept that the impact of population variation is minimal and does not

ro
influence the final calculation. The statistical behavior of the Bang and Ramm model [28] as
shown in Figure 12 reinforces this idea as well; similar observations occurred when RTH and
-p
PH were combined in Colombian, Spanish [45], and Eastern European populations [48,50].

Prince and Konigsberg [48] applied these markers to dental samples from Bosnian individuals
re
and showed that they obtained better results when using Bayesian statistics. Our study offers
similar results to those obtained by Prince and Konigsberg [48], where the Bayesian INT
lP

model showed an absolute mean error of 6.83 years, similar to what is reported for other
models (7.08 years for LBHTNZ, 7.02 years for PU, 6.92 years for GBMF, 9.78 years for UP,
7.48 years for SCKSBT and 9.82 years for BR). A correlation coefficient of 0.85 was
na

produced between the predicted ages and the actual ages using a Bayesian approach to
estimate age-at-death. Similar results were found by Prince and Konigsberg [48].
ur

When INT Bayesian was applied to the control sample of Colombian mestizos, compared to a
specific methodology for that population [45], the analysis showed that both methods offer
Jo

similar results. As expected, the best results were achieved in ages between 30 and 59 years,
with a mean error of 5.86 years for INT Bayesian and 5.96 years. The statistical trend
demonstrates the homogeneity of the method, which can be observed in Figure 13. These
results confirm the usefulness of the technique that combines both markers (RTH and PH)
and its methodology applicability at an international level. The combination of both indicators
(RTH and PH) is equally applicable across ethnicity and sex; additionally, the rate of
degenerative change is following a constant line, but it can reach a plateau (80<) where the

18
behavior of degeneration is too slow to be used as an indicator of estimate age-at-death
(Figure 16 and figure 12) and with a known error range (see Figure 14 and Table 4). This
procedure can be used as a global forensic criterion for the estimate age-at-death and could
also be interpreted as a procedure that involves what Howell [48] highlights as a
“uniformitarian assumption”.

Trajectory effect

Schmitt et al. [57] showed that the application of Bayes' theorem tended to minimize
underestimation of age in older individuals and obtained a mean standard error of 13.67 years.
We obtained similar findings applying a Bayesian model: for individuals between 60 and 69
years old, the standard error was 3.67 years; for those between 70 and 79 years old, the

of
standard error was 7.89 years, and for those individuals of 80 years old or older, the standard
error was 18.03 years. Prince and Konigsberg [48] observed the same trend.

ro
At older ages, the development of translucency is slower because the physiological process is
probably delayed when approaching the coronal region [28]. The dental area in this region is
-p
much wider than in the apical portion of the tooth root, and the physiological mechanism of
inorganic salt deposits probably requires more time to generate translucency (see figure 16).
re
At the apex level, the effect is exactly the opposite, as there is less dentin area, the process of
inorganic salt deposits is much faster, and the translucency is noticeable more quickly.
lP

Actually, this process may be a cause of the overestimation of age in young adults when
applying methods that include RTH and PH [45], in addition to the statistical problems that
have been reported [63, 48]. This overestimation and underestimation have been called
na

“trajectory effect” [4] which could be reduced but never eliminated.

According to the trajectory effect, the error observed in early ages is usually always less than
ur

in adult ages, both at the dental level and in the skeletal structure, due to the constancy of
biomechanical and physiological forces [4]. Therefore, various investigations have
Jo

highlighted the importance of analyzing the whole body in case it is available [48, 49, 53, 90].
Nawrocki [4], maintains that each degenerative indicator of skeletal and dental age represents
a significant contribution to the estimation of age. Thus, the gathering of various indicators
and criteria contributes positively to the verification of the constant degenerative variation to
which an individual is exposed.

Sex influence

19
While there are reports that RDT can be affected by biological sex [34,49,91], other studies
have shown that the impact of sex may be minimal (45,51,66,86,88], or even that it does not
represent a significant difference at all [57]. Our data confirm that RTH behavior does not
present statistically significant differences related to sex. Sex only significantly influences
RH, which is the variable that least explains the age of the individual (r = 0.21; see Figures 4-
6).

Conclusion

This study examined the forensic applicability of a Bayesian model using root translucency

of
height (RTH) and periodontal height (PH) as a means of estimate age-at-death in adults from
the Lamendin´s technique. RDT has shown a high correlation with chronological age (r =

ro
0.90), as does periodontosis (r = 0.77). The combination of both criteria provides an effective
model for estimate age-at-death (r = 0.85), particularly for individuals between 30 and 79
-p
years. The results also indicate that the degenerative changes in the RDT is not dependent
upon ancestral filiation or sex in any given age category. However, the rate of degenerative
change in the RDT is very slow in older adults as to be used as a single criterion for the
re
estimate age-at-death; the range of estimation error too wide in this age category for forensic
science purposes.
lP

Due to the “trajectory effect”, while the overestimation of age is slight in younger individuals
(20-29 years old/-7.97 years error) and underestimation of age in older adults persists (80
na

years old or more/18.03 years error) there are remarkable advantages since the procedure is
simple and accessible to any forensic context. This procedure not require intensive training or
specialized technologies to obtain satisfactory results which include high repeatability and
ur

replicability. However, this research adds to the recommendations that suggest combining
several procedures to estimate age-at-death in adults.
Jo

The model proposed here may be generalizable across all humans [4,22,92]. However, we
must specify that there is a difference between the method that can be used, with population-
specific algorithms, and the algorithms themselves that can be generalized as model. The
Bayesian model uses an FIDB through which specific methods can be operationalized for
each population-specific or used as a generalizable or global method for various populations.
We reaffirm that “Methods based on these robust samples are more likely to be applicable in

20
a wider array of situations than current methods that are (unnecessarily) tailored to specific
subgroups” [4: 99].

The results of the control sample from Colombia suggest that the method proposed here, in
comparison with a specially adjusted method for that population, had no substantial
differences. However, further verification is required to verify these findings. These results
demonstrate that this approach is “more likely to be applicable in a wider array of
situations”. More information a priori will be essential to consolidate this international
procedure.

The statistical model developed in this research, as well as the FIDB, could be used online in
the immediate future. The system will be free for use worldwide. FIDB will continue to be

of
constantly updated, revised and verified.

ro
Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal
-p
relationships that could have appeared to influence the work reported in this paper.
re
Acknowledgment
lP

The authors of this article want to acknowledge Humanitarian and Human Rights Resource
Center of the American Academic of Forensic Science and the National Institute of Justice
through their Forensic Technology Center of Excellence Program, RTI International of U.S.
Department of Justice by sponsorship provided. The authors wish to thank Debra Komar,
na

Dawnie W. Steadman and Daniel Suarez-Ponce for valuable comments on an earlier draft of
this chapter. Additionally, the authors want to acknowledge to Luz Dary Escobar, Debra
Prince, Gretel González-Colmenares, Miguel Botella-López, Johana Becerra-Alvarez for
providing us with the data that has been part of this research and to Christian Quispe por the
ur

assistance provided.
Jo

REFERENCES CITED

[1] N.R. Langley, B. Dudzik, The application of theory in skeletal age estimation, in: C.C. Boyd Jr.,
D.C. Boyd (Eds.), Forensic Anthropology: Theoretical Framework and Scientific Basis, John Wiley
& Sons Ltd. 2018, pp. 99–112.

[2] R.D. Hoppa, J.W. Vaupel, The Rostock Manifesto for paleodemography: the way from stage to
age, in: R.D. Hoppa, J.W. Vaupel (Eds.), Paleodemography: age distributions from skeletal samples,

21
Cambridge University Press, Cambridge, 2002, pp. 1–8.

[3] G.R. Milner, J.L. Boldsen, Skeletal Age Estimation: Where We Are and Where We Should Go,
in: D.C. Dirkmaat (Ed.), A Companion to Forensic Anthropology, Blackwell. 2012, pp. 224–238.

[4] S.P. Nawrocki, The nature and sources of error in the estimation of age at death from the skeleton,
in: K.E. Latham, M. Finnegan (Eds.), Age Estimation of the Human Skeleton, Charles C. Thomas
Publisher, LTD, Illinois, 2010, pp. 79–101.

[5] D.H. Ubelaker, H. Khosrowshahi, Estimation of age in forensic anthropology: historical

perspective and recent methodological advances, Forensic Sci Res. 4 (1) (2019) 1–9.

[6] D.A. Komar, Lessons from Srebrenica: the contributions and limitation of physical anthropology
in identifying victims of war crimes, J Foren Sci. 48 (2003) 713–716.

of
[7] D.H. Ubelaker, Issues in the global applications of methodology in forensic anthropology, Journal
of Forensic Sciences. 53 (3) (2008) 606–607. https://doi.org/10.1111/j.1556-4029.2008.00717.x

[8] A.H. Ross, L.W. Konigsberg, New formulae for estimating stature in the Balkans, Journal of

ro
forensic sciences. 47 (1) (2002) 165–167.

[9] E.H. Kimmerle, D.A. Prince, G.E. Berg, Inter-Observer Variation in Methodologies Involving the
-p
Pubic Symphysis, Sternal Ribs, and Teeth, J Forensic Sci. 53 (2008) 594–600.

[10] J.P. Bocquet-Appel, C. Masset, Farewell to paleodemography, J Hum Evol. 11 (1982) 321–333.
re
https://doi.org/10.1016/S0047-2484(82)80023-7

[11] J.P. Bocquet-Appel, C. Masset, Paleodemography: Resurrection or ghost?, J. Hum. Evol. 14

lP

(1985) 107–111. https://doi.org/10.1016/S0047-2484(85)80001-4

[12] J.P. Bocquet-Appel, C. Masset, Paleodemography: Expectancy and false hope, Am. J. Phys.
Anthrop. 99 (1996) 571–583. https://doi.org/10.1002/(SICI)1096-8644(199604)99:4<571::AID-
AJPA4>3.0.CO;2-X
na

[13] M. Cox, Ageing Adults from the Skeleton. Human Osteology, in: M. Cox, S. Mays (Eds.),
Archaeology and Forensic Science, Greenwich Medical Ltd., London, 2000, pp. 61–81.
ur

[14] J.E. Buikstra, L.W. Konigsberg, Paleodemography: Critiques and Controversies, American
Anthropologist. 87 (2) (1985) 316–333. https://doi.org/10.1525/aa.1985.87.2.02a00050
Jo

[15] R.P. Mensforth, C.O. Lovejoy, Anatomical, physiological, and epidemiological correlates of the
aging process: a confirmation of multifactorial age determination in the Libben skeletal population,
American Journal of Physical Anthropology. 68 (1985) 87–106.

[16] D.P. Van Gerven, G.J. Armelagos, Farewell to paleodemography? Rumors of its death have
been greatly exaggerated, Journal of Human Evolution. 12 (1983) 353–360.

[17] J.P. Bocquet-Appel, Explaining the Neolithic Demographic Transition, in: J.P. Bocquet-Appel,
O. Bar-Yosef (Eds.), The Neolithic Demographic Transition and its Consequences, Springer,
Dordrecht, 2008, pp. 35–55.

22
[18] G.R. Milner, J.W. Wood, J.L. Boldsen, Paleodemography, in: M.A. Katzenberg, S.R. Saunders
(Eds.), Biological Anthropology of the Human Skeleton, Wiley-Liss, New York, 2000, pp. 467–497.

[19] L.W. Konigsberg, S.R. Frankenberg, Estimation of age structure in anthropological

demography, Am J Phys Anthropol. 89 (1992) 235–256.

[20] L.W. Konigsberg, S.R. Frankenberg, Deconstructing death in paleodemography, Am J Phys

Anthropol. 117 (2002) 297–309.

[21] D. Lucy, A.M. Pollard, C.A. Roberts, A comparison of three dental techniques for estimating
age at death in humans, Journal of Archaeological Science. 22 (1995) 417–428.

[22] D.A. Komar, J.E. Buikstra, Forensic Anthropology: Contemporary Theory and Practice, Oxford
University Press, Oxford, 2008.

of
[23] N. Chaillet, M. Nystrom, A. Demirjian, Comparison of dental maturity in children of different
ethnic origins: international maturity curves for clinicians, Journal of Forensic Science. 50 (2005)
1164–1174.

ro
[24] H.M. Liversidge, Worldwide variation in human dental development, in: J.L. Thompson, A.
Nelson, G. Krovitz (Eds.), Growth and Development in the Genus Homo, Cambridge University
Press, Cambridge, 2003, pp. 53–88.
-p
[25] H.M. Liversidge, Dental age revisited, in: J.D. Irish, G.C. Nelson (Eds.), Technique and
re
Application in Dental Anthropology, Cambridge University Press, Cambridge, 2008, pp. 234–252.

[26] G. Willems, C. Moulin-Romsee, T. Solheim, Non-destructive dental-age calculation methods in

lP

adults: intra- and inter-observer effects, Forensic Sci Int. 126 (2002) 221–226.

[27] M. Ajmal, B. Mody, G. Kumar, Age estimation using three established methods. A study on
Indian population, Forensic Sci Int. 122 (2–3) (2001) 150–154. https://doi.org/10.1016/s0379-
0738(01)00501-1
na

[28] G. Bang, E. Ramm, Determination of age in humans from root dentin transparency, Acta
Odontol Scand. 28 (1970) 3–35.
ur

[29] G. Gustafson, Age determinations on teeth, J Am Dent Assoc. 41 (1950) 45–54.

[30] G. Johanson, Age Determination from Human Teeth: A Critical Evaluation with Special
Jo

Consideration of Changes After Fourteen Years of Age, 1971.

[31] M. López-Nicolás, M. Canteras, A. Luna, Age estimation by IBAS image analysis of teeth,
Forensic Sci Int. 45 (1–2) (1990) 143–150. https://doi.org/10.1016/0379-0738(90)90231-m

[32] M. Lopez-Nicolás, A. Morales, A. Luna, Morphometric study of teeth in age calculation, J

Forensic Odontostomatol. 11(1) (1993) 1–8.

[33] M. López-Nicolás, A. Morales, A. Luna, Application of dimorphism in teeth to age calculation,

J Forensic Odontostomatol. 14 (1) (1996) 9–12.

23
[34] M. Lorentsen, T. Solheim, Age assessment based on translucent dentine, J Forensic
Odontostomatol. 7 (1989) 3–9.

[35] W.R. Maples, An improved technique using dental histology for estimation of adult age, J
Forensic Sci. 23 (1978) 764–770.

[36] Z. Metzger, A. Buchner, M. Gorsky, Gustafson's method for age determination from teeth--a
modification for the use of dentists in identification teams, J Forensic Sci. 25 (4) (1980) 742–749.

[37] A.E. Miles, Dentition in the estimation of age, Journal of Dental Research. 42 (1963) 255–263.

[38] A. Sengupta, D.K. Whittaker, R.P. Shellis, Difficulties in estimating age using root dentine
translucency in human teeth of varying antiquity, Arch Oral Biol. 44 (1999) 889–899.

of
[39] R.F. Sognnaes, B.M. Grat, P.J. Papin, Biomedical image processing for age measurements of
intact teeth, Journal of Forensic Science. 30 (1985) 1082–1089.

[40] T. Solheim, Dental root translucency as an indicator age, Scand J Dent Res. 97 (1989) 189–197.

ro
[41] T. Solheim, P.K. Sundnes, Dental age estimation of Norwegian adults--a comparison of different
methods, Forensic Science International. 16 (1) (1980) 7–17.
-p
[42] L. Vasiliadis, A.I. Darling, B.G. Levers, The histology of sclerotic human root dentine, Arch
Oral Biol. 28 (1983) 693–700.
re
[43] D.K. Whittaker, Ageing from the dentition, in: M. Cox, S. Mays (Eds.), Human Osteology in
Archaeology and Forensic Science, Greenwich Medical Media, London, 2000, pp. 83–99.
lP

[44] Sengupta A, Shellis RP, Whittaker DK. Measuring root dentine translucency in human
teeth of varying antiquity. J Archaeol Sci. 25 (1998) 1221–1229.

[45] G. González-Colmenares, M.C. Botella-López, G. Moreno-Rueda, J.R. Fernández-Cardenete,

na

Age estimation by a dental method: a comparison of Lamendin's and Prince & Ubelaker's technique,
J Forensic Sci. 52 (2007) 1156–1160.

[46] L. Martrille, D.H. Ubelaker, C. Cattaneo, F. Seguret, M. Tremblay, E. Baccino, Comparison of

ur

four skeletal methods for the estimation of age at death on white and black adults, J Forensic Sci. 52
(2) (2007) 302–307. https://doi.org/10.1111/j.1556-4029.2006.00367.x
Jo

[47] A. Meinl, C.D. Huber, S. Tangl, G.M. Gruber, M. Teschler-Nicola, G. Watzek, Comparison of
the validity of three dental methods for the estimation of age at death, Forensic Sci Int. 178 (2008)
96–105.

[48] D.A. Prince, L.W. Konigsberg, New Formulae for Estimating Age-at-Death in the Balkans
Utilizing Lamendin's Dental Technique and Bayesian Analysis, J Forensic Sci. 53 (2008) 578–587.

[49] D.A. Prince, D.H. Ubelaker, Application of Lamendin’s adult dental aging technique to a diverse
skeletal sample, J Forensic Sci. 47 (2002) 107–116.

24
[50] N. Sarajlić, Z. Cihlarz, E.-E. Klonowski, I. Selak, H. Brkić, B. Topić, Two-criteria dental aging
method applied to a Bosnian population: comparison of formulae for each tooth group versus one
formula for all teeth, Bosn J Basic Med Sci. 6 (3) (2006) 78–83.

[51] D.H. Ubelaker, R. Parra C., Application of Three Dental Methods of Adult Age Estimation from
Intact Single Rooted Teeth to a Peruvian Sample, J Forensic Sci. 53 (2008) 608–611.

[52] D.A. Prince, K.M. Crowley, Human Odontology and Dentition in Forensic Anthropology, in:
N.R. Langley, M.T.A. Tersigni-Tarrant (Eds.), Forensic Anthropology. A Comprehensive
Introduction, Routledge, New York, 2017, pp. 111–124.

[53] H. Lamendin, E. Baccino, J.F. Humbert, J.C. Tavernier, R.M. Nossintchouk, A. Zerilli, A simple
technique for age estimation in adult corpses: The two criteria dental method, J Forensic Sci. 37
(1992) 1373–1379.

of
[54] G. Gustafson, Microscopic examination of teeth as a means of identification in forensic
medicine, J Am Dent Assoc. 35 (10) (1947) 720–724.
https://doi.org/10.14219/jada.archive.1947.0323

ro
[55] H. Lamendin, Critères dentinaires pour appréciation d'âge: Études de la translucidité et des
canalicules. Intérêt en odonto-stomatologie légale [Dentin criteria for estimation of age: Studies on
translucidity and on canals. Significance in forensic odontostomatology], Revue d'odonto-
stomatologie. 7 (2) (1978) 111–119.
-p
[56] H. Lamendin, Age determination with the “simplified” Gustafson method, Chir Dent Fr. 58
re
(1988) 43–47.

[57] A. Schmitt, B. Saliba-Serre, M. Tremblay, L. Martrille, An evaluation of statistical methods for

lP

the determination of age of death using dental root translucency and periodontosis, Journal of
Forensic Sciences. 55(3) (2010) 590–596. https://doi.org/10.1016/0379-0738(80)90174-7.

[58] R.C. Parra, J.E. Buikstra, D.H. Ubelaker, Dental Aging Methods and Population Variation:
Methodology for International Applications on Human Remains from Forensic Contexts, Proceeding
na

of the International Association of Forensic Science, Funcha-Madeira (Portugal), 2011.

[59] D.H. Ubelaker, R. Parra C., Dental Aging Methods and Population Variation as Demonstrated in
a Peruvian Sample, AAFS 60th Anniversary Scientific Meeting, Washington, D.C., 2008.
ur

[60] A. Drusini, I. Calliari, A. Volpe, Root dentine transparency: age determination of human teeth
using computerized densitometric analysis, Am J Phys Anthropol. 85 (1991) 25–30.
Jo

[61] L.D, Escobar, I. Sanabria, Estimación de edad en una población adulta colombiana del
cementerio Central de Bogotá D.C., mediante la aplicación del método Lamendin. Universidad
Nacional de Colombia. Bogotá D.C. Tesis de postgrado en Antropología Forense. 2002.

[62] J. Adserias-Garriga, L. Nogué-Navarro, S.C. Zapico, D.H. Ubelaker, Setting the Light
Conditions for Measuring Root Transparency for Age-at-Death Estimation Methods, Int J Legal
Med. 132 (2017) 637–641.

[63] D. Lucy, RG. Aykroyd, AM. Pollard, T. Solheim. A Bayesian approach to adult human age

25
estimation from dental observations by Johanson’s age changes. J Forensic Sci. 41 (1996) 189–94.

[64] DC. Montogomery, EA. Peck, VG. Geoffrey Introduction to Linear Regression Analysis (5th
ed.). Wiley. (2012).

[65] A. Ackermann, M. Steyn, A test of the Lamendin method of age estimation in South African
canines, Forensic Sci Int. 236 (2014) 192.e1–192.e6. https://doi.org/10.1016/j.forsciint.2013.12.023

[66] E. Zorba, N. Goutas, C. Spiliopoulou, K. Moraitis, An evaluation of dental methods by

Lamendin and Prince and Ubelaker for estimation of adult age in a sample of modern Greeks,
HOMO - Journal of Comparative Human Biology (2018). https://doi.org/10.1016/j.jchb.2018.03.006

[67] W.R. Maples, P.M. Rice, Some difficulties in the Gustafson dental age estimations, J Forensic
Sci. 24 (1979) 168–172.

of
[68] J. Nalbandian, F. Gonzales, R.F. Sognnaes, Sclerotic age changes in root dentine of human teeth
as observed by optical, electron and X-ray microscopy, Journal of Dental Research. 39 (1960) 598–
601.

ro
[69] L. Vasiliadis, C. Stavrianos, P. Kafas, Assessment of dentin translucency in estimating age: a
research, Int J Orofac Sci. 1 (2008) 4.
-p
[70] B. Azaz, Y. Michaeli, D. Nitzan, Aging of tissues of the roots of non functional human teeth
(impacted canines), Oral Surg Oral Med Oral Pathol. 43 (4) (1977) 572–578.
re
[71] E. Metska, C. Stavrianos, L. Vasiliadis, Estimation of dental age using root dentine translucency,
Surgery Journal. 4 (2009) 21–28.
lP

[72] L.A. Regalado, A. Del Ángel, Estimación de la edad con base en la medición de la transparencia
de la dentina radicular en dientes permanentes, Revista ADM. 65 (2008) 190–195.

[73] A. Nanci, Dentine-pulp complex, in: P. Rudolph (Ed.), Ten Cate’s Oral Histology.
Development, Structure, and Function, Mosby, St. Louis, 2003, pp. 192–239.
na

[74] L. Saruwatari, S. Ban, N. Ohno, Characterization of apatite in the translucent dentine of single
root, Aichi Gakuin J Dent Sci. 13 (2000) 37–46.
ur

[75] N. Tang, D. Antoine, S. Hillson, Application of the Bang and Ramm age at death estimation
method to two known-age archaeological assemblages, Am J Phys Anthropol. 155 (3) (2014) 332–
351. https://doi.org/10.1002/ajpa.22566
Jo

[76] D.F. Weber, Human dentine sclerosis: a microradiographic survey, Arch Oral Biol. 19 (1974)
163–169.

[77] A. Acharya, S. Vimi, Effectiveness of Bang and Ramm's formulae in age assessment of Indians
from dentin translucency length, Int J Legal Med. 123 (6) (2009) 483–488.

[78] A. Olze, G. Geserick, A. Schmeling, Age estimation of unidentified corpses by measurement of

root translucency, J Forensic Odontostomatol. 22 (2) (2004) 28–33.

26
[79] V.K. Kashyap, R.N. Koterwara, A modified Gustafson Method of age stimation from teeth,
Forensic Science International. 47 (1990) 237–247.

[80] S. Hillson, Dental Pathology, in: M.A. Katzenberg, S.R. Saunders (Eds.), Biological
Anthropology of the Human Skeleton, Wiley-Liss, New York, 2000, pp. 249–286.

[81] S. Hillson, Dental Pathology, in: M.A. Katzenberg, S.R. Saunders (Eds.), Biological
Anthropology of the Human Skeleton, Wiley-Liss, New York, 2008, pp. 301–340.

[82] G.C. Nelson, A Host of Other Dental Diseases and Disorders, in: J.D. Irish, G.R. Scott (Eds.), A
Companion to Dental Anthropology, John Wiley & Sons, Inc. 2016, pp. 465–483.

[83] L.L. Montes Merma, Relación entre la enfermedad periodontal en pacientes puérperas y el peso
de los neonatos del departamento de ginecología y obstetricia del hospital regional del Cusco en el
mes de noviembre, año 2019, Facultad de Ciencias de la Salud, Escuela Profesional de Odontología,

of
Universidad Nacional de San Antonio Abad del Cusco, Cusco, 2020.

[84] R.B. Condori Canaza, Características clínicas del periodonto por el consumo de sustancias
psicoactivas en pacientes de los centros de rehabilitación de la ciudad del Cusco, Facultad de

ro
Ciencias de la Salud, Escuela Profesional de Odontología, Universidad Nacional de San Antonio
Abad del Cusco, Cusco, 2019.
-p
[85] F.M. Tadjoedin, A.H. Fitri, S.O. Kuswandani, B. Sulijaya, Y. Soeroso, The correlation between
age and periodontal diseases, Journal of International Dental and Medical Research. 10 (2) (2017)
327–332.
re
[86] P. Foti, M. Adalian, Y. Signoli, O. Ardagna, G. Dutour, Leonetti Limits of the Lamendin
method in age determination, Forensic Sci. Int. 122 (2001) 101–106. https://doi.org/10.1016/S0379-
lP

0738(01)00472-8

[87] L. Klepinger, Fundamentals of Forensic Anthropology, Wiley-Liss, Hoboken, N.J., 2006.

[88] G. Garizoain, S. Petrone, M. Plischuk, A.M. Inda, M.N. Garcia, Evaluation of Lamendin’s age-
na

at-death estimation method in a documented osteological collection (La Plata, Argentina), Forensic
Science International: Reports 2 (2020) 100060. https://doi.org/10.1016/j.fsir.2020.100060

[89] D.K. Whittaker, M.M. Bakri, Racial variations in the extent of tooth root translucency in ageing
ur

individuals, Arch Oral Biol. 41 (1) (1996) 15–19.

[90] E. Baccino, DH. Ubelaker, LC Hayek, A. Zerilli. Evaluation of seven methods of

Jo

estimating age at death from mature human skeletal remains. J Forensic Sci. 44 (1999) 931-
36.

[91] D.A. Prince, Estimation of Skeletal Age-at-Death from Dental Root Translucency, University of
Tennessee, 2004.

[92] N. Howell, Toward a uniformitarian theory of hu- man paleodemography. In RH Ward

and KM Weiss (eds.):The Demographic Evolution of Human Populations. 1976. New York:
Academic, pp. 25-40.

27
 of
ro
TABLES AND FIGURES

-p
Table 1. Distribution of the FIDB sample by sex and phenotype.

Sex
re
Female Male Total
Phenotype n % n % n %
Peruvian Mestizo 40 5.8 96 13.9 136 19.6
Colombian Mestizo 7 1 71 10.2 78 11.4
lP

Mediterranean White 34 4.9 45 6.5 79 11.4

European American 100 14.4 100 14.4 200 28.9
White
African American Black 100 14.4 100 14.4 200 28.9
na

Total 281 40.5 412 59.5 693 100

ur

Table 2. Descriptive statistics of the behavior in millimeters (mm) root height (RH),
periodontal height (PH), and root translucency height (RTH) by age category.
Jo

RH in mm PH in mm RTH in mm
Edad N Mean SD Mean SD Mean SD
20-29 46 13.996 2.25 1.913 1.028 2.4 2.163
30-39 131 13.769 2.122 2.416 1.069 3.618 1.849
40-49 151 13.858 2.081 3.263 1.272 5.057 2.038
50-59 148 14.367 2.465 3.918 1.508 7.573 2.283
60-69 109 14.014 2.084 4.338 1.782 9.014 2.405
70-79 82 14.143 2.205 5.182 1.945 10.00 2.826

28
 80 < 26 13.547 2.038 4.954 1.872 10.592 2.61
Total 693 14.006 2.203 3.613 1.77 6.561 3.391

Table 3. Summary of regression analyses.

Total sample Sample w/o influencial

n = 693 observations n = 655
R-Sq 64.0% 72.1%
Adjusted R-Sq 63.9% 72.0%
R Multiple Correlation
Coefficient 0.80 0.85
Residual Standard Deviation 9.32 7.73
Regression Coefficients
 Constant 13.270 13.028

of
 RH 0.6585 0.6655
 PH 0.3292 0.3125
 RTH 0.4280 0.4310

ro
-p
Table 4. Mean residuals analysis in the seven estimation methods using the FIDB = 693
individuals.
re
LBHTNZ PU GBMF UP SCKSBT BR INT

[20 - 29] -8.95 -10.55 -7.69 -7.43 -11.39 -11.20 -7.97

lP

[30 - 39] -5.70 -6.90 -4.70 -6.63 -8.52 -10.38 -5.15

[40 - 49] -1.06 -2.88 -0.64 -6.03 -4.63 -7.17 -1.51

[50 - 59] 0.99 -1.29 0.13 -7.41 -3.11 -7.84 -0.49

na

[60 - 69] 5.51 3.16 4.40 -4.93 1.05 -2.26 3.67

[70 - 79] 10.70 8.13 9.16 -3.91 5.51 1.90 7.89

ur

[80 <} 20.52 17.07 19.13 7.85 15.43 13.41 18.03

Table 5. Intraobserver and interobserver evaluation

Jo

Observers Parameter n r P
1-2 RH 47 1.00 0.00
1-3 RH 47 1.00 0.00
2-3 RH 47 1.00 0.00
1-2 PH 47 0.97 0.00
1-3 PH 47 0.97 0.00
2-3 PH 47 0.97 0.00

29
 1-2 RTH 47 0.99 0.00
1-3 RTH 47 0.99 0.00
2-3 RTH 47 0.99 0.00
1 RH 47 1.00 0.00
2 RH 47 0.99 0.00
3 RH 47 1.00 0.00
1 PH 47 1.00 0.00
2 PH 47 0.99 0.00
3 PH 47 1.00 0.00
1 RTH 47 0.99 0.00
2 RTH 47 1.00 0.00
3 RTH 47 1.00 0.00

of
Table 6. Absolute mean error of the INT Bayesian and other three methods applied to a

ro
sample of the Colombian mestizo population.

Age category n INT PU LBHTNZ GBMF

20-29 45 11.93 13.06

-p
11.61 12.69
30-39 24 6.98 8.54 6.39 7.68
re
40-49 25 3.68 3.88 3.63 3.73
50-59 19 6.92 6.36 7.88 6.49
lP

60-69 9 17.25 17.31 18.83 16.96

70-79 19 21.37 21.67 23.89 21.42
80 < 9 26.2 27.8 29.98 25.83
Total 150 11.43 12.13 11.94 11.71
na

P-value 0.84 0.41 0.50 0.53

ur
Jo

Figure 1. Dental features used for age estimation. Adapted from Ubelaker and Parra
[50].

30
 of
ro
Figure 2. Root Dentin Translucency.
-p
re
lP
na
ur
Jo

Figure 3. Measurement of root translucency height (RTH).

31
 of
ro
-p
Figure 4. Interaction between Phenotype and Sex in RH.
re
lP
na
ur
Jo

Figure 5. Interaction between Phenotype and Sex in PH.

32
 of
ro
-p
re
Figure 6. Interaction between Phenotype and Sex in RTH.
lP
na
ur
Jo

Figure 7. Interaction between Provenience and Sex in RH.

33
 of
ro
-p
re
Figure 8. Interaction between Provenience and Sex in PH.
lP
na
ur
Jo

Figure 9. Interaction between Provenience and Sex in RTH.

34
 of
ro
-p
re
lP

Figure 10. Analysis of regression residuals without influential observations.

na
ur
Jo

35
Figure 11. Comparison of the mean absolute error between the LBHTNZ, PU, GBMF,
UP, SCKSBT, BR, and INT Bayesian methods by age range with influential
observations (FIDB = 693 individuals).

of
ro
-p
Figure 12. Comparison of the mean absolute error between the LBHTNZ, PU, GBMF,
UP, SCKSBT, BR and INT Bayesian methods by age range.
re
lP
na
ur
Jo

Figure 13. Comparison of the mean absolute error between the LBHTNZ, PU, GBMF,
UP, SCKSBT, BR and INT Bayesian methods.

36
 of
ro
-p
Figure 14. Comparison of the mean error among the seven methods using FIDB = 693
individuals.
re
lP
na
ur
Jo

Figure 15. Dispersion Diagram of Real Age vs. Predicted Age using FIDB = 693
individuals.

37
 of
ro
Figure 16. Root translucency height in millimeters by Age grouped (FIDB 693).
-p
re
lP
na
ur
Jo

38