Tracing the biological profile of an individual from skeletal remains is a major issue in forensic science, as well as in the study of ancient human populations. With particular reference to the forensic context, the anthropological analysis will allow for personal identification even when genetic analysis is not possible. While some analyses may lead to a fixed discrete classification (e.g., male or female, when a sex determination can be achieved), others, such as the methods traditionally used to estimate the age at death in adults [1, 2, 3], give a fairly wide range within which the age of the individual examined is presumably included. The difficulty in narrowing this range in adults depends on the fact that the methods used are generally based on the degree of bone remodeling or degenerative processes, which are strongly affected by biological variability or environmental factors. As a rule, the greater the age of the individual examined, the wider the age range; therefore, in some cases, only generic indications can be provided (e.g., age $>$50 years using dental wear [4], maxillary suture obliteration [5], pubic symphysial surface [6] and sternal rib end [7]). Given this, the possibility of achieving a more defined age of death is an important methodological objective in the identification process.

Analyzing tooth cementum annulation (TCA) is a histological dental method routinely used on non-human mammals [8] that has been demonstrated to have important applications on humans. The method is based on the principle that dental root cementum is subjected to continuous growth throughout life, resulting in the formation of annual incremental lines. These lines appear under the microscope as pairs of light and dark layers: each pair would represent annual growth. Adding the incremental line count to the age of eruption of the tooth yields an age estimation of the individual [9].

Although the prospect of determining the age at death in a more precise manner is very tempting, TCA analysis can hardly supplant traditional anthropological methods because it is not yet sufficiently standardized and tested in humans. Despite previous research in humans (see among others [10, 11, 12, 13]), the results shown in the literature studies reported heterogeneous accuracy with correlation coefficients ranging from 0.42 to 0.97 between chronological and biological age [14]. A progressive decrease in accuracy and bias has been shown with aging, leading to an underestimation of age especially in individuals over 40 years of age [11]. This has certainly increased the uncertainty in the application of the method by researchers [15]. From a technical point of view, despite numerous published protocol proposals (among others [12, 14, 15]) starting with Stott et al. [10], there are still difficulties in reading incremental bands [16, 17]. In addition to the destructive nature of this analysis, the current low diffusion of this method depends on several technical and theoretical reasons, starting from which is the most appropriate technique up to the interpretation of the biological nature of annuli [18]. Another highly debated issue concerns the use of tooth types. While some studies claim that only mono-rooted teeth are appropriate for TCA estimates [15], others consider all tooth types to be adequate [19].

Ideally, TCA analysis could be useful in supporting identification also in poorly preserved dead bodies if the protocols were better defined, validated, relatively simple, and using tools commonly found in a forensic histology laboratory. The main aim of this study is, therefore, to verify the applicability and reliability of TCA-estimated age in a sample of the Italian population with known age and sex by evaluating inter- and intra-observer effect on annulation counts, looking for procedures that do not require specific equipment (low-speed diamond saw) and verifying whether all tooth types are suitable. In addition to the validation of the annuli counting, this study aims to test the method’s effectiveness and limitations in estimating age with regard to chronological age, sex, root type, and dental arch.

The analysis of teeth can give useful information to establish the age of the individual. This can be of extreme importance in the forensic context concerning both the living and the dead person. In the latter case, the estimation of the age at death is a fundamental element in the identification process. Age estimation can be easily undertaken through dental development and eruption methods in children and adolescents (sub-adults) [25], while these methods are not applicable over the age of 21 [26]. When the third molar is absent, this threshold is set at 14–15 years. Thereafter, other traditional anthropological methods for age determination can be used that are based on degenerative changes, but these show great variability and often do not allow distinguishing between individuals over 50 years. Turning to dental methods again, the Lamendin method yielded very accurate results for middle adults [27], even if it applies only to individuals over 25.5 years of age. In this study, we examined, as a possible alternative method in determining age from teeth in adults, the cementochronology, which was already proposed about 40 years ago to estimate age in humans [10], as in other mammals. The full sequence of cementum annulations provides a record of the individual’s life history from the time of tooth eruption to the time of death or to the time of tooth extraction/loss. Being an invasive technique, this is not appropriate for living adults.

Since this promising method for age estimation is currently scarcely adopted in forensic and archaeoanthropological fields, we decided to test the methodological reliability on a sample of teeth extracted from individuals with known sex and age.

The reasons for its low diffusion may lie primarily because, besides being a destructive technique, the need for special and costly equipment restricts its applicability [28]. Consequently, we decided to apply this method by using tools and instruments commonly available in a well-equipped forensic histology laboratory. Starting therefore from the main distinction of the techniques executable on undecalcified sections [10] or decalcified sections [29], we chose to exclude the first ones as they require the use of very expensive and not widely spread devices (low-speed diamond saw), although the other procedures require longer processing times (decalcifying and embedding procedure).

In addition to technical problems that may be an obstacle in the choice of this method, it should be remembered that one of the main reasons for researchers to apply this method with some hesitation is the heterogeneous accuracy of the results [14, 17]. In this regard, we first investigated the precision in incremental line counts, and then evaluated the accuracy in age estimation that these counts allowed. The intra-observer and inter-observer reliability of annuli counts made on the same images allowed us to ascertain the precision of the cementum lines counting, achieving excellent results in both cases.

In the whole sample, we observed a moderate correlation between known and estimated age (0.56). Although statistically significant, this value was lower than the average correlation coefficient (0.87) reported in the review by Naji et al. [14], and the determination coefficient showed that the annuli counts do not accurately predict chronological age, as only 32% of the variability in chronological age is explained by the estimated age. It is possible that the methodology applied in this study (analysis of TCA on decalcified semi-thin sections of all tooth types) has conditioned this result and that a higher correlation between the variables and, consequently a higher accuracy in estimating age, could have been achieved in this sample if other validated protocols had been used (among others [12, 28, 30, 31]). Sultana et al. [32], for example, recently obtained an average difference of just one year between chronological and estimated ages in a sample of 60 teeth extracted from individuals in a narrow age range (20–50 years). Moreover, a strong correlation (r = 0.93) was found in a recent study carried out on 200 teeth of the same type (canines) [15].

The univariate and multivariate analyses conducted revealed that there is no significant difference in the accuracy of age estimation compared to chronological age between single-rooted and multi-rooted teeth. This allows an important possibility of extension of this methodology to dental specimens of any type, in contrast to the technique still prevailing that involves the use of single-rooted teeth only (see among others [11, 15]). Our findings have been confirmed by other studies [19, 30] but contradict, for example, the study by Sultana et al. [32], who observed a greater difference between chronological and estimated age in specimens consisting of third molars -only three third molars were included in their sample- than in the other types of teeth examined.

The different trends of sex comparisons in univariate versus multivariate analyses can be interpreted regarding the distribution of the sample with a reduced presence of males in younger age groups with good accuracy and their increased presence in older age groups characterized by lower accuracy in comparison to females (chi-square = 7.0, df = 2, p 0.05). Indeed, the results of multiple regression analysis showed that there are no differences in accuracy between the sexes, as shown in other studies [15, 33].

We also observed a different trend in the accuracy when comparing upper and lower teeth by univariate and multivariate analyses. In this case, we verified that there was a reduced presence of teeth from the upper arch in the younger age group with good accuracy and their increased presence in middle/old age groups, unlike the teeth of the lower arch (chi-square = 7.7, df = 2, p 0.025). Based then on the results of the multiple regression analysis, a significantly greater accuracy resulted for the upper dental arch than for the lower dental arch. This particular aspect, almost completely neglected by studies in the literature, deserves to be better explored and understood. According to Wittwer-Backofen et al. [12], comparable reliability resulted for teeth from maxilla or mandible (see also [13, 34]). We believe that the comparison between upper and lower teeth from the same individuals could certainly help to verify our findings. Karunakaran et al. [35] performed a comparative study between teeth from the two arches of the same individuals (n = 50). Using the database reported in this study, we calculated and compared the %$\mathrm{\Delta }$ (SD) of maxillary (5.63 $\pm$ 2.43%) and mandibular teeth (7.47 $\pm$ 4.15%) of the same individuals by paired t-test with statistically significant results (p = 0.001). This data from the literature, therefore, seems to confirm the different trend in accuracy between dental arches, with better performance in age estimation by maxillary teeth than mandibular teeth.

The comparison between the three different age groups of the sample showed that the age underestimation increases with the increase of chronological age. This result may be due to the greater difficulty in line counting in elderly subjects as the cementum becomes both thicker and denser with thinner lines [36]. We found an accuracy of 85% with a reduced bias (1.5 years on average) up to 30 years of age, while in the later age groups the age estimation worsens with an accuracy that drops to about 75% in the 31–60 years group and about 60% over 60 years of age. As in the forensic field, it is generally believed that an acceptable method of age estimation should not exceed a threshold of $\pm$10 years compared to the chronological age [37], we verified employing the linear regression equations on the entire sample (Fig. 4a) that this threshold falls at 43.4 years of age in our sample. Above age 43, there are underestimates of chronological age greater, on average, than 10 years and increases in %$\mathrm{\Delta }$ for age prediction that make age estimation problematic using only the TCA analysis according to our protocol. Conversely, Colard et al. [28] found the best results in three older individuals compared to younger ones by examining nine anthropological cases by a validated protocol [31]. Consistently with our results, Obertová and Francken [11] found considerable underestimation of age in individuals older than 40 years. Irregularities and a slowdown in the rate of cementum deposition are likely to occur with aging. In addition to the increased risk of error in the elderly because of the greater difficulty in counting thinner and more compressed annuli (less distinctness), the decreased accuracy of TCA-estimated age in the elderly would depend at least in part on changes in the physical properties of cementum with aging due to the reduction in mechanical stress (masticatory forces) with age [38]. Age-related decline in age estimation is a general characteristic of skeletal indicators of age, so much so that it has been suggested that there is no single indicator that can accurately reflect the age at death [39]. According to Martrille et al. [27], the best strategy is to use many dental and skeletal indicators to achieve reliable results. In particular, starting from the assumption that no single method of age assessment is adequate for the whole life span since they depend on several factors, such as sex, population, size, and individual characteristics [40, 41], it is suggested that the membership of the individual under consideration in a generic young to the old group be initially determined by reliable methods, such as that of the pubic symphysis [42, 43] or that of the sternal end of the fourth rib [44]. Thereafter, the age can be estimated using the methods that are more appropriate for that age group. Although these macroscopic methods have some limitations, as they are population-dependent and are based on inverse regression [45], this preliminary age assessment of the skeleton before selecting a specific method of age estimation is also supported by Baccino et al. [46] and other authors [47, 48], and is referred to as the two-step procedure. Our study indicated that the TCA age estimation can be a reliable method for young adults, but it should be combined with multiple age indicators of age estimation for older adults (age $>$40 years). Without a doubt, a forensic anthropologist should employ multiple regions of the skeleton and different methodologies of analysis to provide a final age estimation in the process of identifying the deceased.

Another important aspect is cementum composition, which was analyzed by Edinborough et al. [49]. They underlined that life-history parameters (e.g., pregnancies, skeletal trauma, and renal disease) for putative cementum deposition periodicity cannot be rigorously calculated, using optical microscopy. Their results reject the possibility of accurate estimation of the distribution of mineralization of tooth cementum using light and scanning electron microscopy alone and conclude that the best analysis for knowledge cementum composition is ToF-SIMS (Time-of-Flight Secondary Ion Mass Spectrometry). Finally, new and important steps have recently been taken in the non-destructive estimation of age at death in adults. In particular, Le Cabec et al. [50] have used synchrotron imaging to non-destructively display cementum annulations in known age teeth from 18–19th centuries collection, while Newham et al. [51] developed a new method for semi-automatic increment counting using three-dimensional images provided by X-ray propagation-based phase-contrast imaging on fresh dental tissue.

This study has several strengths, which include applying a technique that can be implemented easily without requiring expensive laboratory instruments. Furthermore, we excluded teeth from the same individual, as it may affect the findings [37]. However, the major strength of this study is the analysis of both single- and multi-rooted teeth. This made it possible to exclude that the different tooth types used may condition the inaccuracy of age estimate.

Among the limitations of this study, in addition to the small sample size, there is the use of a sample that consisted exclusively of therapeutically extracted teeth, despite the ongoing debate on this issue [15, 52]. Although we cannot exclude that periodontal disease may have affected the apposition of the cementum, we sectioned the middle third of the root according to Broucker et al. [53], as only limited effects on the cementum annuli count have been reported in this part of the root. Apart from cases of profound periodontal pathologies, Kagerer and Grupe [19] showed that teeth with only marginal periodontal disease yielded very satisfactory outcomes if the level of section was carefully selected. Other studies also support the slight impact of periodontal diseases on TCA counting [32, 54]. No influence was observed by Wittwer-Backofen et al. [12] who stated: “the accuracy of the TCA age estimation is independent of periodontal disease”.

Our results confirm that annuli counting is precise, but the TCA age estimates are poorly accurate when using decalcified, stained, thin sections of all tooth types. Moderate correlations between chronological and estimated ages compared with those in other studies could be explained by methodological differences related to the protocol adopted. In our study, accuracy in the age estimation by TCA is independent of root type and sex, while it mainly depends on the chronological age. Although the influence of dental arch on the accuracy in age estimation still requires further investigation, its relevance as an accuracy predictor emerged from this study. In particular, the analysis of TCA proves to be a very convenient method of estimating age for young adults, but not in older adults (over 40 years old) as cementum lines become increasingly difficult to differentiate with aging. In a forensic context, different approaches to the age at death estimation are always advisable to support the age estimation by TCA, especially in older adults.

TCA, tooth cementum annulations; ICC, Interclass Correlation Coefficient; VIF, variance inflation factor.

EG-R, NR and MN designed the research study. IS, PF, NR and JM performed the research. EG-R and NR analyzed the data. EG-R was a major contributor in writing the manuscript. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript.

Data processing complies with the general authorization for scientific research purposes granted by the Italian Data Protection Authority (1 March 2012 as published in Italy’s Official Journal no. 72 dated 26 March 2012) since the data do not entail any significant personalized impact on data subjects. Approval by an institutional and/or licensing committee is not required since experimental protocols are not applied in the study.The patients’ data were completely anonymized, according to Regulation (EU) 2016/679 of the European Parliament and the Council: “The data protection principles should therefore not apply to anonymous information, i.e., information that does not relate to an identified or identifiable natural person or to personal data made sufficiently anonymous to prevent or no longer allow the identification of the data subject. Therefore, this regulation does not apply to the processing of such anonymous information, including for statistical or research purposes”.

We would like to thank all the dentists who contributed to this research. Special thanks to the Centro Odontoiatrico Sforza and to Studio Odontoiatrico Placanica (Bologna). Special thanks to Daniele Sorgato for his assistance during the data collection process.

This research received no external funding.

The authors declare no conflict of interest.