How old are you?
Age assessments based on dental radiographs of third molars
around the 18-year-threshold in different populations, a systematic
review
Loka Holm Kvist
Stella Sabo
Supervisor: Gunilla Klingberg
Master’s thesis in Odontology (30 ECTS)
Malmö University
Dentistry program
Faculty of Odontology
1
Abstract
The mineralization and development of the third molars can be used as an indicator of age in the upper teens and is used in purpose of assessing age in asylum seekers without identification or birth certificates. The evaluation is determined with dental radiographs and different methods of
interpretation.
The aim of the study was to investigate the accuracy and precision of different age estimation models of radiographs of third molars, focusing on population groups and ethnicity. This is of great importance since the results of an age assessment constitute the basis in a juridical outcome,
deciding whether an individual is an adult or a child (over or under 18 years of age).
A systematic search of PubMed, Embase and Cochrane was made according to the PRISMA-statement, and 23 articles which met the inclusion criteria were assessed using a modified QUADAS 2 by two observers, ten of these were graded as having low risk of bias.
Population groups were poorly described, leading to difficulties in evaluating results and precision, thereby also comparisons between different studies. The problem both lay at a general level where definitions are ambivalent, but also in the separate articles where definitions are missing.
It was hard to compare the included papers since their aims and methods varied. The main finding was that most of the articles concluded that the use of the third molar as an age indicator in the upper teens is an insufficient system. The use and impact of the results should thereby be questioned.
Sammanfattning
Mineralisering och utveckling av visdomständer kan användas som åldersindikator i övre tonåren och används för asylsökande utan identifikation eller födelseattest. Utvärderingen görs med hjälp av dentala röntgenbilder och olika tolkningsmodeller.
Målet med den aktuella studien var att undersöka åldersbedömningsmetoder på röntgenbilder av visdomständers tillförlitlighet och överensstämmelse med fokus på populationsgrupper och etnicitet. Detta är av vikt då resultat från en sådan bedömning ligger till grund för juridiska beslut huruvida en individ är vuxen eller ett barn (över eller under 18 år).
En systematisk sökning genomfördes i PubMed, Embase och Cochrane enligt PRISMA-modellen, och 23 artiklar som uppfyllde inklusionskriterierna granskades av två observatörer enligt en modifierad QUADAS 2 och tio artiklar bedömdes ha låg risk för bias.
Populationsbeskrivningarna var bristande. Detta leder till svårigheter att utföra bedömningar av artiklars resultat och precision, därmed också jämförelser mellan olika studier. Problemet ligger på en generell nivå då definitioner är osäkra, men samtidigt i de separata artiklarna då de inte redovisat någon definition.
Det var svårt att jämföra de inkluderade artiklarna då deras syften och metoder varierade.
Huvudfyndet var att de flesta artiklarna konkluderade att användningen av visdomstanden som en åldersindikator i de övre tonåren är en otillräcklig metod. Användning och betydelse av dess resultat bör därför ifrågasättas.
2
Introduction ... 3
The importance of an ID ... 3
Race and ethnicity ... 4
Complementary medical assessment ... 5
Dental radiology ... 5
Methods for dental age estimations ... 6
Aim ... 9
Material and method ... 10
Results ... 12
Discussion ... 23
Definitions of population groups from an ethnic point of view ... 24
Methods for age determination... 25
Possible improvements of the current study ... 26
Improvements in general ... 27
Conclusions ... 27
References ... 28
Appendix ... 33
Introducti At present, certificate it is very im the correct The import In recent y unaccompa Norway. In notification seekers of one third o documents Figure 1. Stati According age with so seeker’s sta those, if an may includ pedodontis examinatio 95% is app assessment stated exam Unaccomp and to a leg and actions ion , the method are insuffic mportant to t chronologi tance of an years, an inc anied childr n 2014, 7 04 n over the la different eth of all childre s is also ofte istics of unacco to the Swe ome kind of atements. T ny doubt of de radiograp st/radiologis on by a paed propriate (6) t (3). Accor minations to panied child gal guardian s concernin ds of determ cient. This le investigate ical age of t ID crease in im ren sought a 49 unaccom ast ten year hnic origins en, globally en questione ompanied childr dish Migrat f ID (if it ex This is usual the true age phy of the h st (today the diatrician (4 ), but a dent rding to the o determine dren have th n (2,6). Acc ng the child mining the a eads to lega e and evalua the individu migration h asylum in E mpanied chil period, see s lack forma y, are not reg
ed (2).
ren seeking asy
tion Board i xists), the ca lly performe e remains, th hand and wr e methods o 4,5). For the tal assessme asylum pro the chrono e right to he cording to th shall be ma age of young al as well as ate different ual.
has been see Europe; one ldren sough e Figure 1 (1 al birth cert gistered unt ylum in Sweden it is up to th ase officer t ed by a num he asylum s rist and/or d of Mincer an e dental radi ent cannot b ocedures art ological age ealth care, e he UN conv ade in the be g asylum se s ethical dile t methods an en in Europe third of the ht asylum in 1). Many yo tificates. A U til the age o
n, graph modifie he asylum se then determ mber of age seeker is off dental radiog nd/or Kullm iographic as be the only ticle 17.5 th of the asylu education, to vention on t est interest o eekers that l emmas in th nd their suc e. In 2009, ese applied t n Sweden; th oung unacco UNICEF ev f five. The v
ed from the Swe
eeker to stat ines the cre oriented co fered a med graphs asses man is recom ssessment a material use e state may um seeker ( o be directe he Rights o of the child lack a prope he asylum p ccess in esta 15 100 to either Sw his was the ompanied a valuation sta validity of e edish Migration ate his/her id edibility of t onversations dical assessm ssed by a mmended) ( a probability ed for an ag y carry out th (6). ed to a muni of Children, (this is also 3 er birth process and ablishing weden or top asylum ates that existing n Board (1). dentity and the asylum s; after ment. This (3), or an y level of ge he above icipality decisions o is stated 3
4
for the child (7). A child is defined as a person younger than 18 years of age.
The definition of age differs in regard to the chronological age, which is determined by the person´s date of birth, and the biological age, which is individual. People develop biologically at different rates (8) and possibly different in diverse population groups (9,10). There is a correlation between the biological and chronological age, but the individual variation is wide (11) and the variance is greater within a population group than between different groups (12). It is very important that the persons evaluating test results are familiar with these differences, since the methods used for age estimations may be more accurate in determining the biological age rather than the chronological. It is also very important that the methods used meet necessary scientific criteria (6).
Race and ethnicity
The term ethnicity is currently defined as a group of people that self-identify to the particular group and are in some ways distinct from other sets of people, for example by common ancestry. Shared cultural symbols, diet, language, religion, norms and values are other components that might be significant (13-15). Ethnicity is a dynamic term which means that it is changeable due to political, historical, economical and social processes for the individual (14).
The suggested difference between ethnicity and race is that ethnicity is something you achieve while race is ascribed to you based on your physical appearance (16,17). The concept of race is wide and may or may not be considered dependent on biological and genetic factors. Race is believed to be made of historical, social, economic and cultural processes or to have a biological base. The opinion about biology is not absolute, some rely that, eventually, a full understanding of genetics will be obtained and believe that race, as defined presently, is a reasonable estimation of the genetic difference between groups. Some do not think race is representable for this, though there is an agreement that some groups are identifiable as distinctly different from others (13). When genetics revealed the great similarity between people, the term race was questioned and is less used nowadays. The association to ideas of a race superiority (from among others Darwinism) that in history has led to attitudes and acts like racism, discrimination, slavery and extermination (17) are also contributing factors to why this term may be inappropriate. In Sweden the term is mainly used when describing pet animals (16).
The use and the concept of the terms race and ethnicity in scientific studies are not very different from each other, and it is still up for discussion which should be used or if they should be used at all. That is why it is important to clarify the definitions and illuminate what they bring to the study (13).
Since there are differences described between groups, the question automatically arises concerning if these differences affect the dentition and its development. Some studies have found statistical differences between population groups (18), others claim that differences exist, but are negligible (19). Regression analysis is a statistical model used to describe relations and is often applied when describing age correlations. A regression model based on a Belgian population was investigated by Gunst et al. (20), and this Belgian formula has been used in other studies investigating its
applicability in different population groups. In some studies it is applicable, though they say it should be used cautiously (19,21).
5
Complementary medical assessment
As mentioned earlier, the method of using skeletal radiographs of hand and wrist is commonly used. The analysis is usually done in relation to Greulich and Pyle’s reference atlas by studying the size, shape and degree of fusion of the different bone elements (2). The method gives a standard
deviation of 0.6-1.1 years. The bones are fully developed at a mean age of 18 years in boys and 17 years in girls. Therefore, the method may not be suitable to accurately estimate the age at the requested interval (2,8).
Radiographs or CT scans of the clavicles can also be used for determining chronicle age. The method analyses the development of ossification of the epiphyseal plate in different stages (8). When the person in question is assumed to be older than 18 years this method is suitable as the development continues into the 20s, though the variation of complete closure ranges between 14.7-30 years for girls and 16.6 -14.7-30 years for boys (2). If the fusion is completed and a scar is visible, it is believable that a woman is at least 20 years old, and a man is at least 21 years old (8). This method has, however, been criticized because of the radiation of internal organs (2).
Age determination by dental radiographs in an individual around 18 years old can only be done on third molars (M3) since they are the only teeth still developing in the upper teens (2), and they are probably less affected by malnutrition etc. than the skeleton (10,22-25). They also show the greatest biological variation (26-28).The mineralization of the third molar roots begins approximately at the age of 15 and is finished at about 20 years of age (27). Many third molars are malformed, irregular or mal-positioned. Pathology and/or loss of the tooth are not uncommon. It is also important to note that agenesis of the third molar is common (28).
Age estimation by third molars can sometimes be hard to conduct because of biological variations of the tooth. In an article by Bolaños et al. (29) a sample of 776 individuals between 4 to 21 years of age were studied. 37.1% did not have any third molar while 38% had all of them. Only 6.7% had a single third molar, 10.6% had two and 7.6% had three molars. Consequently, the use of these methods to determine age by tooth development is often difficult.
In studies by Rozkovcová et al.; the results display a significant difference in development between groups with four existing wisdom teeth and groups with one to three. In the group with one to three wisdom teeth the development were directly delayed related to how many wisdom teeth that were missing. In these studies they found a deviation if a method was used on an individual with four wisdom teeth compared to one with one to three. Because of this the authors suggest that the
method should be modified to avoid age underestimation of persons with varied degrees of agenesis (30-32).
Dental radiology
There are different types of radiography and different ways to categorize them. There are the conventional and the digital forms, which describes how the image is produced and viewed and there are different ways to view the radiographs considering light and magnification etc. There are also ways to describe how the image is taken; intra- or extra-orally. The type of radiograph used depends on the clinical indication and what the radiograph should be used for (33).
6
usually used as an overview. Indications are, among others, assessment prior to surgical removal of wisdom teeth, evaluation of the dentition prior to orthodontic treatment or in the case of an
exceptionally neglected mouth. The radiograph gives an overall view of the fascial structures at a relative low radiation dose (0.004-0.03 mSv). The movement of the equipment and the varied distance between the focal trough and the receptor produces distortion and magnification. There are authors who claim that, because the magnification degree varies throughout the panoramic
radiograph (33), only measurements in the vertical dimension should be considered (25). This is because the magnification factor in the horizontal dimension varies non-linearly and this
subsequently produces inaccurate measurements. However, even measurements in the vertical dimension ought to be cautiously made, should no reference point be at hand (34).
Periapical radiographs are taken to display the teeth and the surrounding bone tissue. The technique is commonly used to view apical changes, during endodontic treatment and to assess non-erupted teeth. Using the paralleling technique, geometrical precise images may be obtained with little magnification and minimal foreshortening/elongation. Since the method is intraoral, it is sometimes difficult to achieve an accurate positioning of the receptor when taking radiographs of third molars. It takes about 15 radiographs to complete a full mouth survey, compared to a single panoramic radiograph, producing a large difference in radiation dose (one periapical film ca. 0.001-0.008 mSv) (33). The type of image used is chosen according to diagnostic purpose.
Computerized tomography (CT) is a technique used since 1972. The fan-shaped x-ray beam is transmitted from a radiographic tube, and as the machine moves around the patient, the detectors record the patient’s absorption characteristics. A computer reconstructs the information from
multiple projections to a coherent view. CT has several advantages compared to conventional films, such as eliminating superimposition of irrelevant structures and it administers high-contrast
resolution of the images. It can also be viewed in various planes. It is useful in several scientific fields, in dentistry for example, prior to dental implant surgery (35). It can also be used in age determination (36). Since the 1990s, CT has in dentistry often been replaced with cone beam computed tomography – CBCT, a technique that gives much lower radiation dose compared to conventional CT (37).
Methods for dental age estimations
One method to estimate chronological age using dental radiographs was described by Gleiser and Hunt in 1955. This method was originally developed by taking radiographs at a set time interval during the mineralization of the permanent first molar; in those radiographs 15 different stages were identified. In the same paper, the authors concluded that calcification was a more valid method to assess the somatic maturation than the tooth’s time of eruption (38). Subsequently, this method was modified by Köhler et al. (1994) into ten developmental stages; seven stages related to the
formation of the root and three related to the formation of the crown. Mesotten et al. used a modified system to estimate chronological age on a single third molar (28), see Figure 2.
Figure 2. Dev Demirjian tooth, A-H of the first chronologi quadrants ( (40). Minc third molar Figure 3. Dem Moorrees e stages of ro morpholog sometimes the radiogr and maxill elopment stage presented a H (40). This (together w ical age, ins
(seven or fo cer et al. use rs (26), see mirijans’ classifi et al. reporte oot formatio gical differe s a crypt-sta raphs of 345 ary incisors es according to G another meth method is i with Gleiser stead of asse ourteen teeth ed a modifie Figure 3. fications and de ed a method on (see Figu ences and ar age has been 5 North Am s. The stage Gleisner and Hu hod in the 1 ntended to b and Hunt ( essing the ti h) in the ma ed version o
finitions for the
d in 1963 (4 ure 4) and s re thereby sp n added (10, merican whit es are drawn Hunt modified by 1970’s whic be used in y (38)) to use ime of erup andible, an of Demirjian e development o 42). This me six for the cr paced unev ,43). Moorr te children, n for single-y Köhler et al. ( ch comprise younger chi dental radio tion. By inv average sco n’s scoring of a mandibular ethod descr rown (43). T enly across rees et al. de looking at a - and multi-(39). es eight stag ildren (3-17 ographs to e vestigating t ore of matur system A-H r molar modifie ibes eight d The stages a time (10). I efined these all permane rooted teeth ges of the de 7 years old) establish teeth in one rity can be d H, only eval ed by Mincer at different ma are defined In later year e stages afte ent mandibu h and the av 7 eveloping and is one e or two determined luating t al. (41). aturating by rs, er analysing ular teeth verage time 7 g
facilitating used in age Figure 4. Moo Kullman et age. It uses root length Figure 5. Kull All method methods. A tooth to be Subjective reliability. assessment of subjectiv In the work the stages a root length length of th Furthermor standardize judgment a g compariso e determina orrees’ classific t al. made a s seven stag h and by ass lmans classifica ds mentione An observer e at a certain approaches Garn et al. t of a radiog ve approxim ks of Kullm are focused h. Demirijan he develope re, Kullman ed compute and increase on of differe ations on thi cation of a multi a stage class ges determin sessing apex
ation of the thir
ed above (G r examines a n stage of de s are always (44) states graph and a mations, pro man et al. (27 d on the dev n focused on ed crown (4 n suggests y er digitizer. e standardiz ent teeth in t ird molars ( i-rooted tooth ( sification sy ned by the i x closure (27 d molar develop Gleiser and H a radiograph evelopment s subjects to that problem comparison oduces obvi 7), Moorree eloped part n the length 40). yet another m The aim is t zation. In six the same ch 10). (10). ystem of the initiation of 7), see Figu opment stages (2 Hunt, Moor h and, depe t and conseq o argument ms may also n to a regist ious difficul es et al. (42) of the root h of the deve method. Th to reduce ob x papers he hild (42). Th e third molar f different st ure 5. 27). rrees, Demir ending on th quently the due to the r o lie within ter of mean lties with de )and Köhle in compari eloped part his method u bserver vari examines d he multi-roo r for determ tructures in rjian, Kullm he chosen m person's est risk of low s the method values, whi ecreasing st er et al. of G son to an es of the root uses a more iations caus different sub oted model mining chron tooth devel
man) are sub method, dete timated age standardiza d itself. A su hich are also
tandardizati Gleiser & Hu stimation of in comparis objective a sed by subje bjects, inclu 8 is the one nological lopment, bjective ermines the e. tion and ubjective o made out ion. unt(28) f the final son to the approach, a ective uding the 8
9
digitizer’s precision and accuracy, its usefulness for age-estimating young adults, precision in other used methods and difference in inter- and intra-reliability comparing a subjective and an objective method (45). With Kullman’s method a number of reference points in a radiograph are registered and a software program calculates the distance between the coordinates. The data are then
compared to six pre-defined root stages, each stage covering about 12% of the root length (25).
Systematic Reviews
The PRISMA-statement (Preferred Reporting Items for Systematic Reviews and Meta-analysis) is the preferred process to analyze and summarize scientific results (46). In this process, articles are to be graded using a template, a standardized chart which focuses on assessing the articles risk of bias and overall quality. QUADAS 2 (Quality Assessment of Diagnostic Accuracy Studies) is the one used when analyzing studies of diagnostic accuracy (47,48).
Aim
How accurate and precise are age determination based on dental radiographs of the third molar in adults and/or adolescents of different population groups?
10
A systematic search was conducted in PubMed (20th February), Cochrane (17th March) and Embase (17th March) in year 2014 (search strategies in Appendix II) by the authors with help from librarians at Malmö University, Faculty of Odontology and SBU, Stockholm. The search was made according to the PRISMA-statement. In the primary search, 185 potential papers were found and abstracts were individually assessed by the two authors (LHK, SS) in relation to the following primary criteria:
Primary filter
Inclusion criteria Exclusion criteria
Age determination by radiographs on third molars (M3)
Humans, both sexes, adults and/or adolescents
Ages 12-25 years
Case reports Eruption studies DMFT/DMFS studies
Languages other than English and Scandinavian languages
Extracted teeth
After the individual assessment, all abstracts were discussed by LHK and SS, and 104 abstracts were agreed upon to be read in full text with the following secondary criteria added:
Secondary filter
Inclusion criteria Exclusion criteria
Defined and verified population through race/ethnicity/origin and geographic specification
Must include age 18
Must include identifiable results for M3
Living individuals
Reviews
After the primary search, 20 articles were finally evaluated by a modified QUADAS 2 which is a standardized method of measuring risk of bias and applicability in systematic reviews of studies of diagnostic accuracy (47,48) (Appendix III and IV).
A complementary search by the two authors was conducted in the beginning of year 2015 in the same databases. 15 new articles had been published in PubMed (January 18th), none in Embase (January 26th) or in Cochrane (February 5th). These 15 abstracts were individually assessed by the two authors with the same inclusion- and exclusion criteria as before and eleven of these were read in full text individually. One paper from PubMed was chosen to be assessed using QUADAS 2. A hand search of the reference lists of the 21 articles was conducted by the two authors. 60 titles were selected; four of these had no abstract available. The remaining 56 abstracts were individually assessed by the two authors, 22 articles were read in full text individually and two of them fitted the secondary criteria and were assessed using QUADAS 2.
In total, 23 articles were assessed using QUADAS 2 by the two authors (LHK, SS) and one additional observer (GK) individually.
To clarify: discussed a added. The observers. assessment The proces Figure 6. Flow The article All stages and agreed u e quality ass After this a ts. ss is shown w chart. es excluded in the proce upon togeth sessment us all 23 papers in Figure 6 by full text
ess were con her. In the la sing QUAD s were discu . are summe nducted sep ast stage, us DAS 2 was f ussed by the ed in Appen parately by t sing QUAD first perform e three obse ndix V.
the two auth DAS 2, a thir med individu ervers to rea hors (LHK, rd observer ually by the ach a consen 11 SS), and was e three nsus in the 1
12
The results from the QUADAS 2-assessments are summed in Table 1 and Figure 7.
Sections 2B and 3B were excluded as they were not applicable for the aim of the current study (Appendix IV), thus five of the seven sections in QUADAS 2 were used for assessing quality. Further, in order to grade the overall quality of the papers as low risk of bias, it was decided that the papers should receive low in at least three of the five used sections. This resulted in ten out of 23 papers with low risk of bias. The populations varied in the different articles, covering Indian (49), Hispanic (50), Swedish (25,27), Spanish/African (51), Swiss (52), Austrian (53), Chinese (54) and Italian (55) populations. One article compared several different populations (9). Nine out of ten studies used names/surnames as a verification of ethnic belonging. All included papers used panoramic radiography for the age assessment and six out of ten used all M3s when possible. The remaining four focused of the lower M3s. Five used Demirjian/Mincer’s method, two used Köhler et al., two used Kullman et al. and one used Cameriere et al. Detailed data for these studies are shown in Table 2.
The remaining 13 papers were assessed as having unclear risk of bias. Detailed data for the included studies with unclear risk of bias is shown in Table 3.
Table 1 – Risk of bias
Author 1A 1B 2A 3A 4 Risk of bias (summed) Bhowmik et al. (49) U L L L U 3 Blankenship et al. (56) U U U U U 0 Boopitaksathit et al. (22) U L U U U 1 Corradi et al. (23) U L U U U 1 De Salvia et al. (57) U U U U U 0 Garamendi et al. (24) U L U L U 2 Harris (10) U L U U U 1 Johan et al. (58) U L U U U 1 Kasper et al. (50) U L L L U 3 Knell et al. (59) U L U L U 2 Kullman et al. (25) U L U L L 3 Kullman et al. (27) U L L L L 4 de Las Heras et al. (51) U L L L U 3 De Luca et al. (55) U L U L L 3 Meinl et al. (53) L L U U L 3 Olze et al. (60) U L U L U 2 Scheurer et al. (52) U L L L U 3 Solari et al. (61) U L L U U 2 Thevissen et al. (19) U L U L U 2 Thevissen et al. (9) U L U L L 3 Thevissen et al. (62) U L U L U 2 Willershausen et al. (63) U L U L U 2 Zeng et al. (54) L L U L U 3
13
Figure 7. Risk of bias according to section.
5 15 6 21 2 18 8 17 2 21 4 3A 2A 1B 1A
Risk of bias according to section
14
Table 2 – Included studies grad
ed as low risk of bias
First author, yea
r, cou n try w h er e th e study is p erformed . P o pulation characteristics, setting Aim Radio-graphic method Inde x te st
(age deter- mination method)
Tooth Ob servers an d in terp retat ion Referen ce t es t (chronological age) Statistical analy sis Results
Risk of bias accord
in g to QUADAS 2 B. Bhowmik et al., 2013 , India. (49) 268 Indian subjects v isiting Department o f
Oral diagnosis and orthodon- tics.
Nation al it y verified b y addres s in o ral heal th re cords and giv en n ame and fa mi ly name. Age 14- 23 y ears . In 254 cas es of the OPGs (a ty pe of panoramic radiograph) ag e could b e es tim ated . To de term ine whether age es ti -mation of Indian s using the Belg ia
n-specific formulae of Gunst et al. produced r
esults that are com p ar a-ble of those re-ported in th e Belgian stud y, and ascertain ho w accur at el y the
same formula predic
ted
juvenile/adult status in
Indians
.
OPG (Orthopan- tomograph
). No document- tation of indic ation . Köhler e t al. 10 s tages . S p ecifi c chart b y Guns t et al .
All M3 when possible.
One m ain ex-am iner. 25
randomly selected OPGs where r
e-eva lu-ated b y this exam iner and one oth er after one to two months. Blinded . Oral he alth records .
Wilcoxon matched-pairs signed ranks test. Regression anal
ys is (Belgi an formula). Formula missing for cert ain tooth co m b ina-tions, h ence only 254 anal ys es . 74% of the Indian s ubj ects ’ ages fe ll within th e 68% of the Belgian conf ide n ce interv al of th e fo rm ula. 78% were corr ec tl y cat e-gorized b y age g roup (</> 18). Th e authors conclud ed th at the Belgian form ula e is appli -cable in an Indian popu-lation is v alid bu t should be us ed wi th cau tion s in ce it is not h ighl y re li able . Low. K. A Kasper et al., 2009, USA. (50)
950 Hispanics from Dallas
and Cameron County ag ed 12-22 y ears . Eth-nicity ver ified b y s u rnam e or ma id en na me from de mo-graphic section in th e d enta l records or parental v erif i-cation. To ev alua te ag e es tim ation ac cu-racy in a North and South Tex as Hispanic popu la-tion. OPG. No
documenta- tion of indic
ation . Dem irjian et al. Eigh t st ages .
All M3 when possible.
Ten observ ers. Ea ch ca se wa s ev al ua te d by more than one observer. 20 cas es wer e r e-ev al ua te d by t h e
same observer. 10 cas
es were exam ined b y a ll ten observ ers. Blinded . Birth cert ifi cat e, Medical docu -mentation or birth d ate listed in th e subje ct’s dental record.
Inter and intra weighted
kapp a statisti cs. T-t est. Male ahead of female and His p anics ah ead of Am erican whi tes (rel ativ e the r es u lts b y M incer et al. 1993). If at stage B or C, a Hi spa n ic fema le i s probably less th an 19 ye ars old and a m ale assumingly und er 18. The m axillar y M3s g enera ll y ahead of the mandibular . Low.
15 L . Kull ma n et al., 1995, S w eden. (25) 483 of th e sam p le te eth .
Same popula- tion as
in Kullm an et a l. 1992. To t est whe ther age es tim ations based on the
development of wisdom teeth could b
e im-proved b y using com puter techn iques to
measure root length
instead of subjectiv el y es ti-mating dev elop-m ent s tages . OPG. No
documenta- tion of indic
ation
.
Kullm
an et
al. (comput- eriz
ed). S ix st ages .
Lower left (LL) and lower right (
L
R)
M3.
Two observers. Not cl
ear
if
blinded
.
Known date of birth.
Multivari at e regression anal ys is . Student´ s t-test. The computerized method showed to b e inf erior com p ared to th e subjec-tive standard method. Ma le s a h ea d of fe ma le s. Low. L . Kull ma n et al., 1992, S w eden. (27) 677 y oung
adults <15-25< years visiting the de
p ar tme nt of oral Radiolog y in
Huddinge. Ethnicity verified b
y Scandinav ian nam e and s o cia l security number exclud ing adopted indi-viduals. To ex am ine wi th which pre cision th e r o ot de v el o p-ment of th e lower third molar can be us ed to es tim at e chronological ag e in an indiv idual. OPG. No
documenta- tion of indic
ation . Kullm an et al. S even st ages . LL and/or LR M3. Two observers. Blinded
.
Known date of birth.
Student´ s t-test. Minor differ ences between sexes in the tim e sequence of the minerali-zat ion, no diff er ence between right /l ef t were
shown. The method gives poor precision
and has in general a SD of 1-2 y ears around th e mean ages. A significant and sy stematic differen ce betw een observers. Low. S. Martin-d e La s He ra s e t al., 2007 , S p ain and Africa. (51) 344 patients from a priv ate denta l c lini c in Galic ia in Spa in and 228 (95 Ma gre b ia n origin and 133
Spanish origin) visiting pub
lic
oral h
eal
th
in
Ceuta, Africa. Ethnicity
veri-fi ed by b o th
parents’ sur- names. 14-22 ye
ar s. To id entif y diff er -ences in the pattern of devel-opment of th ird m o lars betwe en populations of differen t g eo-graphic and ethn ic origin. OPG. No
documenta- tion of indic
ation . Dem irjian et al m odifi ed b y M incer et al . E ight st ages .
All M3 when possible. One observer exam
ined al l. 20% of th e radiographs were r e-evalu-ated b y the
same observer and one add
i-tional. B
linded
.
Known age from dental records
. ROC analy sis. Weighted k appa statisti cs. Mul ti-varia te ana ly sis with logist ic regression. Maxilla ry M3s a h ead of mandibular. No differ-ences b etwe en s ides . Differences wer e obtained in both sexes (m ales ahead of f em ales ) and
origin, but not b
etween
the
individu
als living in
Ceuta. The indiv
iduals
from Ceuta bein
g ahead of those f rom Galicia. Th e authors th ereb y conclud ed that differ ences depends more on socio-g eograph-ica l fa ctors than ances tr y. Low.
16 A. Meinl et al., 2007 , Austria. (53) 610 patients visiting Bernhard Gottlieb University Dental
Clin
ic
in
Vienna. Individuals between
12-24
y
ears, excluding
foreign sur- names.
To est ablish Austrian ref eren ts data on third mol ar mi ne ral iza-tion evalu ated accord ing to th e eight s tages proposed b y Dem irjian et al . OPG. No
documenta- tion of indic
ation . Dem irjian et al. Eigh t st ages . LL and LR M 3 .
One main observer. 70 cas
es wer e r e-evalu ated af ter several months by t w o observers. Blinded . No record ed verification of age. P aired and unpaired t-test. Ma nn-Whi tn ey U-test. M ales s ignif ican tl y ahead of females in stages E and F. Wi th c o mpl ete mi ne r-ali zat ion of lowe r M 3 at stage H, 90% of the indi-viduals are 18 y ears or older. Low. E. Sch eurer et al., 2011, Switzerland. (52) 307 Centr al European indi-viduals living in Switzerland. 17.5-18.5 y ears . Ethnicity veri-fi ed by na me ,
missing data or unclear n
ame where exclud ed. T o val id at e t h e referen ce values
of wisdom tooth mineralization and erup
tion
of
a
German popula- tion in
a
sam
p
le
of
persons whit known chrono- logic
al age , to an al y se t h e di ffer-ences b etwe en es tim ated dent al
age and chrono- logi
ca l ag e a n d to
propose adjust- ments
if n eeded . A retro- spectiv e cr oss-section al evalu ation of OPG taken for m edical care . Dem irjian et al. m odified b y M incer et al . E ight st ages .
All M3 when possible. Two observers exam
ined al l OPGs. Blinded. Recorded da te of birth . Kappa an al ys is . The h igher the s tage, th e higher the risk o f overes-tim ation . Me an overesti-mation of two y ears in 76% of male cas es and 82% of f emale cases. Even s o , be low s tage E an
underestimation was made. A mean v
alue of the four M3s give s a sl ight ly be tt er SD. T o m inim ize er rors , the authors suggest the ref er-ence va lues to b e an al ys ed in a prospective stud y becaus e of its la ck of precision . Low. D. L. Ze ng et al., 2009, China. (54)
3100 Han in southern Chin
a visiting R adiol-og y Departmen t of the Affil iat ed Hospitals of Stomatolog y of
Sun Yet-sen University
. 4.1-26.9 y ears old . Ethnicity verified b y ethnic name. To stud y the chronological ag e based on third
molar develop- ment of Han
in southern Chin a and com p ar e th es e with re sults of
previous studies on other
ethnic
populations
.
OPG.
No
documenta- tion of indic
ation . Dem irjian et al. Eigh t st ages .
All M3 when possible. Two observers (radiolog
ists),
one viewing 1600 cases
and
the o
ther
1500
cases. 300 radiographs were evalu
ated b y both two tim es with a tim e in terva l of
four weeks. Not clear if blinded
. Recorded da te of birth . Student´ s t-test.
Wilcoxon matched-pairs signed rank tests.
No significant differences in gen
eral between th e diffe re nt M3s. Ma le s ahead of f em ales . Diff er-ences com p ared to oth er ethnic groups ar e discussed. Low.
17 S. De L u ca et al., 2013, Italy . (55) 397 Italian subjects , ag e 13-22 y ears , visiting the Unit
of Orthodontics and Paed
iatr ic dentistr y, University of Milan. Ethn ic it y verified b y birthplace of child and parents togeth er with names and su rnam es . To t est th e v alid-it y of Cam eri ere ’s cut-off v alu e on a sam p le of living subjects. OPG. No
documenta- tion of indic
ation
.
Cameriere et al
.
LL and LR M3. No documented number of observers. Not clear if blinded
. Recorded da te of birth from dental records. Bay es’ th eorem post-test prob a-bilit y. 91.4% co rrec t cl as si fied individuals seen to </>18 ye ars of age wit h the current method. 95% probability that an ind i-viduals in 18 y ears or older if 0 ,0 8. Sensitivit y in det erm ining over/under 18 y ears of age with this method 95.7%, specif icity is 86.6%. Low. P. M Thevis se n et al., 2010 , Belgium . (9)
Same popula- tion as
in Thevis se n et al ., 2010 (19). To com p ar e th e degree of third
molar develop- ment on r
adio-graphs between differen
t countries based on a stand -ardiz ed coll ect io n and an al ys is .
OPG taken for diagnos- tics and trea
tm
ents.
Same data as from Thevis
se
n et
al. 2010 (19). Gleis
er and
Hunt et al. modified b
y Köhler e t al. Ten s tages .
All M3 when possible. Nine observer viewing countr
y sp ecifi c cas es . A tenth observer evalu ated 10% of t h e ca se s i n each coun tr y. All ca se s we re re-evaluated after one month. Not cl ear if blinded . Birth cert ifi cat e
and/or identity card.
Em pirica l Bay es’ e sti ma -tion.
Comparison between countries shows differ- ences
in d evelop m ent. Significan t d ifferences were shown, tho ugh they were s m all and n o t con-sta n tly spre ad . Ma le s ahead of f em ales . Low.
18
Table 3 – Included studies graded
as unclear risk of bias
First author, yea
r, cou n try w h er e th e study is p erformed . P o pulation characteristics, setting Aim Radio-graphic method Inde x te st
(age deter- mination method)
Tooth Ob servers an d in terp reta tion Referen ce t es t (chronological age) Statistical analy sis Results
Risk of bias accord
in g to QUADAS 2 J. A Blankenship et al., 2007, USA. (56) African Am erican
whites (n=563) from USA and Canada and blacks
(Afri
can
Am
ericans
,
n=637) from Tennessee and Arkansas veri- fied b
y
demo-graphic infor- mation, aged between
14.0-24.9 y ears. Investiga te the chronolog y of M 3
development in the Afri
can Am erican population and com p are i t to Am erican whi tes .
OPG and periap
ical views when avai labl e. No document- tation of indication. Dem irjian et al. Eigh t st ages .
All M3 when possible. No documented number of observers. Not clear if blinded
. No record ed verification of age. Param etric surviva l anal ys is . Significan t d ifferences between Am eric an bl acks and whites, b lacks being approximately a y ear ahead of whi tes . W ith full y d evelop ed M3s the likelihood of an indi-vidual being at least 18 ye ars old is : Afri can Am erican m al e 93%, fe ma le 84%, whit e ma le 90%, f emale 93 %. S poradica ll y s p read significant d iffer ences between sexes. Left-r ight as y mmetr y occu rred in 13.2% of the mandibular M3s and 9% of the m axillar y . Unclear . T Boonpitaksa thit et al ., 2010, UK. (22) Caucas ia n adolescents aged 12 .6-24.9 y ears r esident in
London visiting Eastman Dental Hospital between
M
arch
2005 and July 2006 and f
rom the archiv es . Ethnicity veri-fied b y c lini cal records , dem o
-graphic data she
et s or by exam inat ion and questioning . T o e x pl ore t h e possibilit y of exclud ing data above the age at which ap ic al closure is com p lete in al l subjects in a population sample . Routine DPTs (denta l panoramic tomograph y ) Dem irjian and Goldstein. All M3 when possible. One investi- gator. 30 randomly chosen DTPs were evalu
ated
twice
b
y
two
separate observers with
a tim e in terva l of one week . Not clear if blinded .
Known date of birth.
Kappa sta tis tics. Regression anal ys is . Two sam p le t-test. The r esults show tha t th e averag e age is e v enl y distributed in an appropriately censored data . The m
ean age in this
sa mple wa s alway s low er than in an in ap propriat el y censored d ata, which shows an overes timated m ean ag e in s tag e H. Maxilla ry M3s a h ead of mandibular in stages E, F, G. No consistent pattern in th e g ender co mparison. Unclear .
19 F. Corradi et al., 2013 , Italy . (23) Itali ans ag ed between 16-22 y ears visiting privat e den tists ma in ly from t h e north-cen tre of Italy (approxi-mately 2 /3) and
from the south of Italy
. Ethnicity verified b y su rnam e. T o e x pl ore a differen t alt ernat ive of clas si fy ing individuals usin g a s uper v is ed Ba yes ian Naïv e clas si fi er and so ft eviden ce . OPG. No document- tation of indic ation . Dem irjian et al. Eigh t st ages .
All M3 when possible.
Two exper ts evalu ated a ll OPGs sepa ra te ly . Not clear if blinded . No record ed verification of age. Modified Naïve Bay
es.
Inter and intra observer reproducib
ility
index.
T
h
e use of soft evi
d en ce re su lt ed i n a hi gh er portion of corr ectly clas si fi ed ag es a nd decreased the nu mbers of false adults and f alse minors. Unclear . A. B. De Salvia et a l. , 2004, Italy . (57) Stud y sample of
400 OPT from the Den
tal Department o f the Univ ersit y of Valen cia, aged b etwe en 14.5-25 y ears
with Spanish nation
alit y, avoiding foreign su rnam es and ambiguous cas es . To a sse ss the
chronological development of the
third m o la r i n a group o f y oun g Caucasian in order to establis h a possible utilisa tion of d at a
for the diagnosis of age.
OPT. No document- tation of indic ation . Dem irjian et al m odifi ed by So la ri e t al. T en st ages (F 1 and G 1 added).
LR M3 when possible, otherwise LL.
No docu
m
ented
number of observers. Not clear if blinded
. No record ed verification of age. No docu m ented statis tical anal ys is . SD mean 2.47 years for the eight s tag es i n general. The added
stages did not
enhance the accuracy
of the method . The likelihood o f men and women being ov er 18 ye ars old in sign ifi cant in st ages F and G. Unclear . P. M Garamendi et al., 2004, S p ain and M o rocco. (24) 114 male Mo roccan,
supposedly minors, illegal im
m igrants in S p ain ag ed between 13-25 y ears. Nation-ality probab ly verified b y the M o roccan em ba ssy in Spain. To an al ys e the effi ca cy of t h e tes ts av ail abl e to es tim ate approximate chronological ag e and, p articularly , the r eli abil it y of an ag e es tim atio n of 18 y ears or more .
OPT taken in purpose of age determ
i-nation . Dem irjian et al. m odified b y M incer . Eight stages . LL and LR M3. Two observ ers exam ined al l OPTs. Not clear if blinded. Date of bir th confirmed b y the M o roc can em ba ssy in Spain. Weighted k appa
index. Area under ROC- curve.
No significant differ ences between th e an al ys is t ee th or wrists, though the com b ination of th e two
methods showed a significant improvement in chrono
logical age prediction. Th e likelihood of being 18 y ear s of ag e or older is pr esented. Unclear .
20 E. F H arris et al., 2007, USA. (10) Am erican-born
whites (n=1953) and blacks (n=2057) visiting Co
lleg
e
of dentistr
y
,
University Tennessee from western Tennessee, northern Mississippi a
nd eas tern Arkans as , ag ed 3-25 y ears . Ethnicity verified b y s elf-report. To descr ibe
normative development of the m
an d ibular third m o lar in Am erican bl acks and whit es and t o te st for se x differen ces in th e form ative s tages of this tooth . Routine OPG. M oorrees et al. 15 stages . LL or LR
M3 randomly chosen in cas
e of good quality on both sides. One observer . Not cl ear if blinded . No record ed verification of age. Intra-ind ivid u al
kappa statistics. Param
etric surviva l anal ys is . Am erican bl acks were significantly ah ead of Am erican whi tes in all stages, bu t no t in a
uniform manner. The results showe
d tha t M3s provides a rough age estim ation , sin ce varia tions ar e gr eat and the method not p recise for adoles cenc e. M ales ahe ad of fe ma le s. Unclear .
N.A Johan et al.,
2012, Mala y sia . (58) 1080 Northeast Mala y sian
residents aged between
14-25 y ears. Non loca ls exc luded b y viewing of registration identif ic ation card number . To a sse ss the variab ility of the lower th ird m o la r development in Northeas t Mala y sian population in bo th sexes, and to gener ate an age prediction model. OPG. No document- tation of indic ation . Dem irjian et al. Eigh t st ages . LL and LR M3. No documented number of observers. Not clear if blinded
. No record ed verification of age. Kappa sta tisti cs. Independen t t-tes t. S p earm an correlation coeffi cien t. Ste p wise regression and mu ltiple lin ear regression anal ys is . Significan t v ariations between sexes and LL and LR M3. M ale s ahe ad of fe ma le s. T h e a g e
prediction model showed a m
ean predi cti o n error betw een -0.17 to +3.14 ye ars and an av er age S D of 1.82 for f emales and 1.90 for males. Unclear . B. Knel l et al., 2009 , Switzerland. (59) 1137 patients aged b etwe en 15-22 y ears
with Swiss citi
zens
h
ip
and
123 from other European countries verified b
y clini ca l r egis-tration form s. To inv estiga te th e reli abil it y of roo t
growth of mandibular wisdom teeth
in determin ing 18 ye ars of age and the effe ct o f gender and or igin. OPG. No document- tation of indic ation . Dem irjian et al. Eigh t st ages . LL or LR M3.
No documented number of observers. Not clear if blinded
.
Clinic
al
registration forms. Logistic regression anal
ys is . S tage H w as obt ained in
individuals less than 18 years old
, and b ecause of this m ean v alu es are thereb y not rel ia b le in determ ina tion o f adult/juveni le s tatus . Ma le s a h ea d of fe ma le s b y about a y ear. 87% of the sam p le had b o th m an-dibular molars, 1 5 % of these h ad d iffer ent
development stages comparing righ
t side and left. M3s should not solemnly be us ed to asse ss 18 y ea rs of a g e. Unclear .
21 A. Olze et al., 2010 , Canada. (60) 605 individuals 11-29 y ears of age b elonging to the Oj ibwa tribe in Can ada . To pres en t statisti cal m eas urem ent on
the time course of third m
o
lar
mineralization in First Nation people in
Canad a. OPG, retrospec tiv el y v iewed . Indicated b y tooth extra ction ,
TMD diagnostics and oral lesions.
Dem irjian et al. Eigh t st ages .
All M3 when possible. No documented number of observers. Not clear if blinded
.
Patient identif
ic
ation
number.
Logistic regression anal
ys is . M ean ag e, m ini m u m and ma xi mum a g e and SD for st ages C-H are s hown in tabl es. The wr ite rs conclud e th at M 3 can be used for age determ ina tion. Unclear . A. C Solari et al., 2001, USA. (61) 679 Hispanic individuals. 14- 25 y ears old. Ethnicity veri-fied b y family
name and/or parents’ f
amily na me i n pa ti en t charts . To ev alua te the chronolog y of third m o lar dev el-opm ent in Hispa n ic s
following the stages
of formation as described by De mi rj ia n et al.
OPG (95%) and periap
ical
films (5%). No documen- tation
of indic ation . Dem irjian et al. m odified by So la ri e t al. T en st ages (G 1 and F 1 added).
All M3 when possible.
Eight observ ers viewing 100 film s each . Blinded . Known age. M ales ah ead of f em ales . The m ean diff er ence in chronological an d es tim ated ag e w as ±3 y ears for females and ±2.6 y ears for m ales. Th e probability of being under 18 years of age i n the st ages D-H is pr es ented . The m axi ll ar y M 3 s ahead of mandibular . Unclear . P. W. Thevis se n et al., 2010 , Belgium . (19) Ca se s from
Belgium (n=2203), China (n=543), Saudi- Arabia (n
=650),
Thailand (n=821), Turkey (n=597), Korea (n=660), Poland (n=552), Japan (n=526), South- India (n
=430). Nationa lit y verified b y birth ce rti fi ca te
and/or identity card. Subjects were check
ed to have lived a whol e li fe ti me in th eir n at ive countr y . 16-22 ye ar s. To co llect countr y spe ci fic da taba ses of third m o lar development to establish dental age pr edi ctions
for the sub-adult group. OPG taken for diag- nostics and trea
tme nt s. Gleis er and
Hunt et al. modified b
y Köhler e t al. Ten s tages .
All M3 when possible. Nine observer viewing countr
y sp ecifi c cas es . A tenth observer evalu ated 10% of t h e ca se s i n each coun tr y. All ca se s we re re-evaluated after one month. Not cl ear if blinded . Birth cert ifi cat e
and/or identity card.
Bay es´ statisti cs. Mc Ne ma r te sts. Kappa sta tisti cs. Using Belgium r efer ences instead of countr y sp ecif ic inform ation incr eas ed the
correct identified juveniles, bu
t d ecreased in identif ication of adults. M ean abs o lu te d ifferen ces
(MAD) and mean square error (MSE)
incr eased on averag e wi th 0 .5 and 2 .5
months with maximal incre
as es of, res p ec tivel y 1.6 and 7.3 months when using Belg ian r esults, though diff eren ces are quite sm all. No overal l
gender-specific con- clusions cou
ld b
e made.
Unclear
22 P. M Thevis se n et al., 2013 , India. (62) 1199 Indian subjects , ag e 4-34 y ears , coll ect ed in three cl ini cs in
New Delhi. Nationa
lit y verified b y birth ce rti fi ca te . T h e subje ct s must have lived th eir
whole lives in New Delhi or Hary
ana. To ev alua te whi ch development reg -istration t echniq u e of third m o lars i s
most promising for sub-adult ag
e estim ation . OPG. No documen- tation of indic ation . Gleis er and Hunt modified b y Köhler e t al. (GH), Haavikko, Dem irjian, Raungpaka, Gusta fson and Koch ,
Harris and Nortje, Kullm
an,
M
oorrees
(MO), Cameriere (CA). LL M3, LR if LL was missing. No docu m ented
number of observers. Not clear if blinded
. Birth cert ifi cat e. S p earm an correlations. Lowest RMS E f o r M oorrees and th e high es t . In gen eral, the differ-ences b etwe en th e methods were s m all. Regres si on an al ys es wer e made
for three methods,
and with th ese th e correc tl y iden tifi ed adu lts were 80% (GH), 80% (M O ), 66% (CA ). Probabilit y for b eing older than 18 y ears old range fo r females between 84-98% and for males between 85-96% at cert ain s tages . G enera l underestimation of the true age in all methods. Unclear . B. W illershaus en e t al. , 2001, Germany . (63) 1202 patients trea ted at cl ini c
and policlinic for dental, or
al
and jaw pathol- ogies at Johannes Gutenberg University Ma
inz , German y . 15-24 ye ar s. Nationa lit y w as derived from patient journals or b y name. To an al ys e and asse ss the correlation of third m o lar root
development with chronological ag
e. OPG. No documen- tation of indic ation . Kullm an et al. S even st ages .
All M3 when possible. Three or four, otherwise exclud
ed.
No documented number of observers. Not clear if blinded
. From OP G documentation. No documented statisti cal anal ys is . Margin erro r ±2-4 y ears . M o re advan ced i n m ales than f emales. No appar ent differ ences in gr owth pattern between differen t nation alit ies/ eth n icit ies or between right/lef t sides. 2.5% of the 18 years old had a ll the M3:s full y develop ed. The method is not fully r eliable, but could b e a good co mpl eme nt t o ot he r methods. Unclear .
23
Discussion
This study is an evaluation of current age estimation methods on third molars, focusing on different population groups and the 18 year threshold. After exclusions, 23 articles remained and were judged according to QUADAS 2 (modified). Ten articles were graded to have low risk of bias, and the remaining 13 as unclear risk of bias. In general, the authors of the different articles of low risk of bias state that the methods used are incomplete or inaccurate (21,52,54,55). They request further research for methodological improvement, and some suggest that the combination of different age determination methods (dental, skeletal etc.) should be studied further (27,55). But even if the methods are considered deficient, some authors conclude that since no other available method is at hand, maybe the third molar should be used for the time being (25,51,53). Only two of the articles considered the method being useful for the purpose of age determination alone (50,51).
In the current study, the exclusion criteria were rather radical (even if the search was broad) and a lot of material was discarded. To get high standard in the current study, while handling this
important subject motivated this approach. Books are generally not included in systematic reviews since the quality is uncertain and often not peer viewed assessed. Grey literature is not a
standardized term and much of it is not publicized, thereby it is not a reliable source. Reviews were excluded since the qualifications for a good review are generally high. It is important to be able to follow the process through the article and to confirm that the results are not influenced by the authors’ opinions. It is also hard to know if a review is systematic or not. Since this is hard to evaluate, all reviews were excluded. In the reference search, articles without available abstracts were excluded. Studies on corpses were excluded since cause of death, decomposition process and handling/storage might affect the body and thereby give results that are non-comparable with living subjects (64). In the case of the current study, it also might be hard to verify the ethnic belonging and “true” chronological age of the individual.
The ten articles graded as low risk of bias have different aims and perspective; therefore it is hard to compare them outright. Some authors argue that differences are considerable between population groups and that population specific data should be used (50), others that differences exist but are negligible (9,21) and therefore the Belgian formulae can generally be used (21). Results from the articles graded as unclear risk of bias displayed the same variations in their conclusions
(10,19,63,65).
Many studies approached the problem of determining the attainment 18 years of age. This can be studied in different ways, though several articles express it as a probability of being 18 years old at a certain developmental stage (three of the articles graded as low risk of bias (50,53,55)). As stated above, it is hard to compare different studies because of their different approaches and methods. One aspect that may influence the results is the age-range used; this influences for example the prevalence of a developmental stage in a sample. Other factors are the amount of stages, one method uses eight stages, another 14 etc. for the same development process, and the stages are not evenly distributed across time.
To test and report a method’s accuracy, the correct amount of classified individuals is shown, usually displaying frequencies between 75-80% (21). Another way is to evaluate the standard
24
solemnly compare these. Even so, the standard deviations were generally high and therefore not as precise as desired at the mean age of 18 years old, varying between ±0.6-3.45 years (27,52-55). It can be debated whether the results in the articles are good or not, some studies are well carried out and may therefore not be that deficient. But the accuracy and precision of the methods may not be good enough for the intended purpose. 18 years of age is an important cut off point, deciding if an individual is officially a child or an adult. An age determination process is intended to help identifying children and thereby give them the support and protection they need (7). If an individual is wrongly judged, it may lead to critical consequences since you automatically force a child to grow up and take away its right to a childhood. One study discussed whether it is better to correctly classify true adults or true minors (19). This question has many answers depending on the
perspective, for example; to falsely classify a child as an adult is worse for the individual than for an adult to be classified as a child. This is why the “benefit of the doubt” is practiced.
The “true” age is of importance throughout the whole asylum process. Correct handling of the case from a legal point of view by all involved authorities is important for both the individual and the society.
Definitions of population groups from an ethnic point of view
The human need to classify or divide people into different groups is a basic and natural behaviour, a survival instinct to recognise new subjects and situations and to be able to put them into a context. It might also come from a strong desire of belonging to a group (66).
Definitions for a population group vary in different disciplines, and several descriptions are available and presently used. Thomas Kuhn (1922-1996) used the term “incommensurable” when describing different paradigms, the terminology used and the problems that occur when one try to make comparisons between them. He stated that the lack of a common baseline makes it impossible to make an objective judgement regarding which one is more correct (67). Therefore, in the present systematic review, the comparison between different studies is problematic and possibly not even realistic, seeing to the lack of standardized terms.
Even if the same definitions were used in studies and disciplines, the ruling paradigm is never eternal (68). It varies with the transformation of the society, and the concept of ethnicity and race changes with it as the migration pathways spreads over the world, leading to an increasing and heterogenic mixture of people. As the paradigm shifts, so does the terminology and the definition of groups.
Three included studies (low risk of bias) have shown differences between population groups
(9,50,51), though one of these claims that they are negligible (9). This is of great importance for the age determination of individuals since according to this, there is no specific model for age
determination that is suitable for all. There need to be some kind of proof of a “true” ethnical belonging and this is not possible without a clear definition of the particular group.
The majority of the studies use name, surname or place of birth to verify ethnicity. Nine out of ten articles evaluated as low risk of bias verified ethnicity by name. These methods are not absolute and can per se not be verified without a birth certificate or comparable documents. Still, if these were available, it has to be questioned whether they provide ethic belonging. To conclude, it is a catch 22. It is an attempt to categorise the reality which probably never will succeed since definitions always fluctuates.
25
Some studies tried to make, for example, population specific regression models. However, the population definition and description were often inadequate in these studies. In case of age
determinations, this leads to difficulties in the actual use of the regression model and one can ask if the evaluated child fits into the population description.
Retrospective studies always inflict a part of contingency, especially concerning radiological studies when looking at certain populations. Since radiographs should not be taken without indication, it is not clear whether the sample reflects the population as a whole. For instance, studies have
investigated the difference in mineralization rate at impacted teeth and those who have fully emerged and found some variances (10,69). Is it then suitable to make a regression analysis on population level if all radiographs, for example, are of patients with impacted teeth prior to surgery, or of children all undergoing orthodontic treatment? This would be an ideal section to assess with the QUADAS 2-instrument, but unfortunately the majority of the included papers lacks proper description of the patient selection seeing to radiological indices. Out of the ten articles graded as low risk of bias, only two mentioned an indication, see Figure 7 (9,52).
Many studies are adjusted to minimise the influence of biological variance through exclusions of inappropriate subjects. The influence of these exclusions on the results should be evaluated since it might affect in targeting the aimed population. However, exclusions may have to be made to get useful results out of the methods, for example excluding obvious pathology or those with only maxillary third molars. In ten out of 23 articles, only mandibular third molars were used. Maxillary is often rejected because of difficulties in the analysis (58). In QUADAS 2, section 4, the sample is investigated, analysing at drop outs and exclusions. Here the sample is to be specified. However, the majority of the selected studies do not report if any patients were excluded (see Figure 7) and if they do, they do not declare why. This is a problem as it is not possible to assess study design or risk of selection bias.
The biological variance regarding the wisdom tooth is greater than that of other teeth. Among other, agenesis, angulation and morphological differences are more common in wisdom teeth (26-28). To use a method of a tooth with such a scattered variance add more complications and dubiety to the reading and results. Rozkovcova’s studies (30,32,70) show that the number of missing wisdom teeth effected the development of the remaining ones. However, none of the papers in this current study aimed to show this relationship and further studies are thus desired.
The effect of malnutrition and similar factors are mentioned in several studies, but none of the current articles investigate it. The tooth development seems to be less effected of these factors than the bone (10,22-25) but the extent of effect on the tooth is not reported.
Development of the third molars is shown to be earlier in males than females, six out of ten articles graded low risk of bias reported this finding (9,19,25,51,53,54). Left/right asymmetry was reported in one article graded as unclear risk of bias (59), two could not find any differences (27,54). The rest do not comment these findings. The maxillary third molars tend to develop prior to their mandibular counterparts, but are not often investigated, because of the difficulties to measure panoramic radiographs (22,54,61,65).
Methods for age determination
The Demirjian model modified by Mincer et al. was the most common method among the articles graded with “low risk of bias”. This method has rather few steps. This is sometimes appreciated by authors because it is easier to classify a stage, leading to lower inter-observation variation and, possibly, higher precision (10,54,57). Though, this might result in lower accuracy. A method with
26
theoretically higher accuracy, but more steps may result in lower precision.
Mincer’s method was the one most commonly used in the articles graded as low risk of bias, but the original study by Mincer describing the method (26) did not meet the inclusion criteria (not a verified/specified population group), and was not evaluated by QUADAS 2. This is interesting since the method is one of the recommended methods used in Sweden today.
The root mineralisation and development process are viewed in different ways in the mentioned methods. Moorrees et al., Kullman et al. and Gleiser and Hunt evaluate the formed root in relation to the expected root length while Demirjian modified by Mincer et al. estimate the developed root in relation to the formed crown (54). Cameriere et al. (55) use an index comparing the width of the apex compared to the length of the root. The expected length of the root might be hard to evaluate regarding the third molar, due to its biological variance, this could influence the reading as well as the results.
In the majority of the included articles, the radiological method used is panoramic radiography. This is a two-dimensional picture of a three-dimensional subject and gives a good overview, but it also produces distortions, magnifications and overlaps. The image is evaluated by one or more observers that exhibit various degrees of experience and there is an abundance of different methods to
evaluate them with. The evaluation is always a subjective process, and the results are never better than the interpretation of the radiograph. Because of this, it is of uttermost importance to evaluate inter- and intra-observation variations to get an apprehension of the methods’ precision. In eight out of 23 articles, the numbers of observers is not accounted for (see Figure 7), and some do not show any statistical analysis of observation variations, making them hard to appraise.
To test a methods’ accuracy, it is desirable that all observers have been blinded to the reference test and to the other observers’ results. This is appropriate since the person may otherwise risk corrupted readings wanting to perform well. However, once again many articles lack to retail if the studies have been blinded or not. 17 of the analysed 23 articles were unclear if blinded; four of these were in the low risk of bias-group.
The studies have different aims and therefore different results and ways of obtaining them. This makes it hard to compare the results – root mean square error (RMSE) to a regression analysis, standard deviation to Kappa statistics or a ROC-analysis. But even if the studies had had similar aims, they would be hard to compare because of the lack of standardized methods.
Some studies aimed to investigate how different statistical methods changed the interpretation of the results (23). This is interesting seeing, for example, how adding or excluding extreme values can change the mean. How may Bayes’ statistics change the analysis compared to a linear regression analysis? Since this is not the current aim it will not be discussed further, but it is interesting for further investigation.
Possible improvements of the current study
The purpose of the current study was to investigate the methods of age determination using third molars, focusing on population criteria. The results are all pointing in the same direction, but would they differ if the inclusion- and exclusion-criteria had been less strict? It is always hard to know if the criteria somehow have managed to reduce the search field and thereby the results. However, the search strategies were broad and the papers were not excluded on publication date. If the criteria had been less specific, more studies may have been assessed using QUADAS 2, but they would probably be graded as unclear or high risk of bias.
