Age estimation in adults using dental techniques is a complex subject. Only a few papers on this subject exist in the Portuguese population, and most address the pulp/tooth ratio in the lower second premolar [37], the upper and lower canines [38], and the central incisors [22]. The dental pulp's and periodontal ligaments'third molar radiographic visibility were also assessed [20, 21]. As for studies regarding root translucency and periodontosis, we could only find one from Fialho [50], making the need for more investigation in this field evident.

Our research highlights significant discrepancies between forensic models’ chronological and estimated ages, underscoring the need for methodological refinement and population-specific adaptation.

The Lamendin method, which combines root translucency, periodontosis, and root length for age estimation, demonstrated limitations, especially in male subjects in older subjects, where the mean underestimation of age exceeded 50 years (MEA = 37.73 vs. mean chronological age = 80.00; p < 0.001), and in females, in younger groups (50–59 years: MEA = 38.62; p < 0.001). Moreover, regardless of sex, this method consistently underestimated the MEA in all age groups, except for males, in the age group 30–39 years (p = 0.646), suggesting that the Lamendin method may not be reliable for older individuals.

This outcome aligns with the findings of other researchers, such as Foti et al. [42], who noted that population-specific variables, like genetic and environmental factors, may limit the Lamendin method’s accuracy.

However, our results don’t agree with recent studies that have refined Lamendin's method [54, 55]. In these studies, the authors use the same variables proposed by Lamendin et al., i.e., the maximum root length, the periodontal recession, and the root dentine translucency, and incorporate Bayesian algorithms and additional statistical techniques. This enhances the method's applicability across diverse populations while maintaining accuracy within 10 years, especially for individuals aged 30–60. The differences between these methodologies may explain the differences.

Moreover, the method's accuracy was corroborated in South African populations, where biological variations further contributed to its reduced effectiveness [36]. Multiple studies have consistently supported the centrality of root translucency as a robust indicator of age. Lamendin et al. [48] initially established root translucency as a reliable dental trait for age estimation, with the translucency increasing with age due to the deposition of hydroxyapatite crystals within the dentin tubules [26]. This natural process progresses from the root apex towards the cervix, and its correlation with age has been confirmed across various populations [56]. However, as pointed out by Gibelli et al. [19], environmental exposure, particularly to high temperatures or specific soil compositions, can alter root translucency, complicating its application in archaeological contexts. Despite these environmental factors, root translucency remains unaffected mainly by individual lifestyle choices like diet and oral hygiene [5], making it a reliable marker in controlled forensic contexts.

On the other hand, periodontosis presents more variability, as it is influenced by extrinsic factors such as diet, dental hygiene, and physical or mechanical irritation [27, 42]. Prince and Konigsberg [56] further support this, emphasizing that individuals'lifestyle and health status can lead to inconsistencies when using periodontosis as an age estimator. This variability was reflected in our study, where periodontosis coefficients showed a lower correlation with age than root translucency, suggesting that periodontosis should be applied cautiously, especially in heterogeneous populations.​

The modified Fialho method focuses solely on root translucency. Roberts et al. [11] referred to the Bang and Ramm method [29] as the gold standard in forensic odontology for age estimation, utilizing root translucency as an age proxy. This method is based on the premise that the dentinal tubule lumens gradually become occluded with mineral deposits over time, increasing light scatter within the root [23]. This process starts at the root apex and progresses coronally as the individual ages [7]. Several studies have presented results, with Tang's et al. research [25] reporting an average absolute difference between age and estimated age of 10.7 years and 8.4 years using the Bang and Ramm method. These authors also noted that age was overestimated in younger and underestimated in older individuals.

This method demonstrated higher accuracy in our sample, yet the limitations of single-factor models remain evident. The exclusion of multi-factorial variables, while simplifying the process, potentially reduces the precision of age estimation. This reflects broader critiques in forensic anthropology, where demographic and lifestyle variables can introduce considerable error if not adequately controlled [49]. The high correlation between tooth wear and age supports the inclusion of tooth length as a dynamic indicator of aging, complementing static measurements like root translucency and root length [11]​. The findings from our study further underscore the need for continuous calibration of age estimation models. As proposed by Lamendin et al. [48], incorporating multiple dental traits improves the reliability of age estimates, but the influence of population-specific factors remains a significant challenge. For future studies, adjustments to the Lamendin formula could involve greater localization, as indicated by Prince and Ubelaker [27], who successfully adapted the method for diverse skeletal samples, reducing the mean error when accounting for ancestry and sex​.

We divided our sample using a 10-year interval gap because it represents a balance between precision and inclusiveness in age estimation. A range of 10 years is generally reliable for forensic purposes, and very narrow ranges may exclude potential candidates from missing persons lists, reducing the chances of successful identification. Excessively wide ranges may include unrelated individuals, complicating and delaying the identification process [54]. Using a 10-year interval, we aim to provide an age range that is practical and effective for narrowing down potential matches without compromising accuracy.

Current literature [36, 54, 55] suggests a trend toward underestimation in the older age group, indicating that root translucency would cease to be helpful as a good estimator for individuals older than 60. In our investigation, in males, the Lamendin method performed exceptionally well in the 30–39 age group, estimating a mean age (MEA) of 34.44 years, which is remarkably close to the chronological mean age of 34.89 years. This demonstrates the method's strong reliability and accuracy for this age group. A similar level of performance was observed only in the 50–59 age group for males, but with the Fialho and Modified Fialho methods, which estimated MEAs of 53.33 years and 55.35 years, respectively. These estimations closely align with this group's chronological mean age of 53.88 years. Such accurate predictions highlight the potential of these methods when applied to specific age groups, although their performance is inconsistent across all groups. This reinforces the importance of tailoring the selection of methods to the demographic characteristics of the individuals being evaluated.

The absence of statistically significant differences in the 80–89 age group for the Fialho and Modified Fialho methods should be interpreted cautiously. This group likely had a small sample size, which may introduce bias or reduce the statistical power to detect significant differences. Consequently, these results may not fully represent the performance of these methods in estimating age for this group.

Generally, all methodologies presented fewer discrepancies between EA and CA in males, possibly due to the difference in sample size. The literature shows diverging results as to the technique's correlation with sex; some authors found better results for males [42, 57], others observed better estimates for females [58], while other studies reported no difference [27].

Our model relies on root translucency, a variable that, as explained, presents some advantages over periodontosis. We’ve also incorporated tooth length over root length. In forensic age estimation, tooth length and root length have been explored as key variables, each offering unique advantages and disadvantages. In our study, tooth length emerged as a more reliable indicator of age than root length, mainly when used with root translucency.

There are several advantages of using tooth length. Tooth length is dynamic and changes over time due to wear, making it a progressively responsive variable in age estimation. As individuals age, natural wear and tear on the teeth can be consistently measured [8, 31, 32, 59, 60]. This predictable wear makes tooth length a practical and reliable indicator of age progression in forensic contexts. Berbesque et al. [6] noted that dietary habits significantly influence tooth wear, which correlates predictably with age, providing a consistent basis for estimating age​.

Additionally, tooth length measurement is relatively straightforward and less susceptible to preservation issues than root length, especially in archaeological or forensic cases involving older remains [61]. As teeth are exposed and more straightforward to measure without significant destruction, this allows for more accurate assessments even when skeletal remains are damaged.

However, tooth length also has some limitations. It is highly influenced by external factors such as dietary habits, dental hygiene practices, and individual or population-specific mastication patterns [62]. This variability can introduce errors when applying general age estimation models across different populations. For instance, populations with diets that lead to accelerated tooth wear may have shorter tooth lengths at younger ages, complicating age estimations if these variables are not adequately accounted for. As noted by Foti et al., this variability requires adjustments or the development of population-specific models [42].

Moreover, tooth length may be significantly reduced in individuals with severe dental attrition or those who have undergone dental treatments, such as restorations or crowns. These interventions can artificially affect tooth length, reducing its effectiveness as an age estimator unless such factors are accounted for during the evaluation.

Root length, by contrast, tends to remain more stable after the tooth fully erupts and develops, making it less susceptible to external environmental or lifestyle factors than tooth length. This stability can offer a more consistent marker for age estimation, especially when dental wear is excessive or uneven. Root length is less affected by individual or population-specific factors, making it a potentially more neutral trait in forensic assessments [63]. Despite these advantages, root length has its limitations. Once fully developed, root length remains relatively static, which can diminish its utility in forensic cases where more dynamic indicators of aging are needed. Root translucency and other root characteristics correlate with age, and due to the static nature of root length post-eruption, they are less reliable as a standalone age predictor [64]​.

It is essential to address the composition differences in our sample, particularly regarding sex distribution and age structure. As noted, the female subsample was considerably older than the male subsample, with differences of 11 years in median age and 8.94 years in mean age. This age disparity likely contributed to the higher mean estimation errors observed in females across all methods. This outcome is consistent with the well-documented tendency for forensic age estimation methods — including Lamendin’s and its derivatives — to show increased error margins in older age cohorts [22, 41, 65]. Table 3 clearly illustrates that from age 60 onwards, all methods systematically underestimate chronological age. Given that half of the female sample is over 61 years old, it is reasonable to attribute part of the higher error observed in female individuals to the age composition of this subsample. Therefore, the observed discrepancy does not necessarily reflect sex-related methodological bias but rather highlights the impact of age distribution in the sample. This bias is an inherent limitation of working with heterogeneous and unbalanced samples, which we acknowledge and discuss. Future studies should aim for more balanced sample structures and may benefit from age-stratified formula adjustments to enhance accuracy in older age groups. Nevertheless, these findings reinforce the need for caution when applying standard models to older individuals and underline the importance of considering both sex and age distribution when interpreting estimation errors.