Relationships between Dental Topography, Gross Wear, and Bang and Ramm/Liversidge and Molleson Age Estimates for a Sample of Human Premolar Teeth

Abstract

Objectives

Molar crown wear is often used in bioarchaeological research as a proxy for age at death. However, a small number of researchers have used premolars or compared the application of different methods of relative age estimation.

Material and methods

Using a sample of 197 previously extracted maxillary first premolars from US dental patients, we considered three protocols for estimating age: the Bang and Ramm/Liversidge and Molleson (BRLM) age estimate method, occlusal topographic analysis, and the Smith system of macrowear scoring. A previous study utilizing the Bang and Ramm method yielded an age estimate range of 9.4 to 10.8 years for the sample.

Results

Our analyses showed no associations between occlusal topography parameters (occlusal slope, relief, or faceting) and BRLM age estimates, but some concordance was found between Smith scoring and BRLM ages estimates and between Smith scoring and occlusal topography parameters.

Conclusion

The results of the present study suggest that relationships between gross tooth wear, tooth shape, and dental age estimates are complex, and available methods should be considered together to gain a more comprehensive understanding of how teeth change their shape with wear throughout the lifecourse.

Introduction

Teeth are used to estimate age at death in a variety of disciplines, including forensics, bioarchaeology, dental ecology, and paleontology (1). The method used for age estimation varies largely by discipline (2-5).

The gold standard for forensic odontology is the Bang and Ramm method which uses root translucency as a proxy for age (4). The premise of this approach was that the lumen of dentinal tubules becomes occluded with mineral, thus augmenting light scatter in the root (6). This process begins at the root apex and advances coronally with age (7). This approach has proven to be accurate and precise within the past 5-10 years (8, 9). The Liversidge and Molleson method is often used alongside the Bang and Ramm method for teeth that are not fully developed (5, 10). The former involves the length measurement of the crown and developing root. These approaches are limited though, due to variation across populations and the fact that they require access to the root apex – something that is not feasible for in situ teeth in bioarchaeological and paleontological samples given the risk of damage to fragile and irreplaceable specimens.

The Smith system, on the other hand, outlines standards for quantifying occlusal macrowear, with potential application to age at death estimation in bioarchaeological contexts (2). This scientific method is a prescribed method for data recording of wear of the anterior dentition and premolars as outlined in the Standards for Data Collection from Human Skeletal Remains (11). Smith scores vary by degree and pattern of gross wear as measured by discrete crown attributes, such as presence and extent of faceting and dentin exposure. While patterns can vary independent of age, depending on the focal population and its exposure to dietary abrasives, the strong association between crown macrowear and chronological age within populations of disparate bioregional origin and dietary practices (subsistence, food preparation) (12-14) suggests that wear variables such as Smith scores can be applied meaningfully to age-at-death estimation. Most aging methods based on crown wear of dentally-mature individuals focus on molars (15). Nevertheless, incisor, canine, and premolar Smith scores can be useful for classifying individuals of unknown age into cohorts for forensic and bioarchaeological application (16).

Occlusal topographic analysis is another method for studying how tooth crowns wear with age (17-19). This approach focuses on shape changes related to wear. It involves modeling occlusal tables in three dimensions and characterizing topography with measurements of continuous attributes of crown form. Experimental studies have shown that topographic parameters such as average occlusal surface slope, relief (ratio of 3D to 2D surface area), and number of contiguous patches of a given orientation on the surface (OPCr) can all mirror both natural and simulated occlusal wear (20).

Study design and sample selection

The aforementioned methods have their strengths and limitations. Yet we are aware of the fact that no direct comparisons can be made between Bang and Ramm/Liversidge and Molleson (BRLM) age estimates, Smith scores, and dental topographic analysis parameters within a single population. In the current study, we compared all of the aforementioned in a sample of unidentified premolars extracted from dental patients in the United States. Extracted premolars were chosen for our study because of their availability in our tooth bank, as well as their potential to supplement molars for wear-based age estimation. In addition to the common tooth extraction reasons in adults, such as periodontal disease, first premolars have traditionally been the tooth of choice for orthodontic extraction in children and adults alike. Therefore, extracted first premolars are widely available across different age groups, unlike other types of teeth, thus making these teeth especially appropriate for this study.

Materials and methods

Dental specimens

Three-dimensional dental scans of 197 right maxillary first premolars (RP³s in anthropological/paleontological parlance and tooth number 14 according to FDI notation) were used in this study. These specimens were sampled from a collection of unidentified premolars that had been extracted from a random cohort of U.S. dental patients and donated to the Indiana University School of Dentistry’s Oral Health Research Institute (Indianapolis, IN). The tooth collection protocol ensured that all extracted teeth were unidentified and not associated with any patient data. The protocol was reviewed and approved by the local Institutional Review Board (IRB #NS0911-07).

Bang and Ramm / Liversidge and Molleson methodology

Dental age estimation methods were originally described in a publication by Algarni and colleagues (21). The Bang and Ramm method was used on fully formed teeth by measuring root dentin translucency (4, 5, 21). The average of the lowest and highest translucency length values from the apex of the root to the enamel-dentin junction (Figure 1) were used to estimate age based on published coefficients (21). For those few specimens without fully developed roots, the Liversidge and Molleson method was used. In these cases, age was estimated using published coefficients (5) for the anterior maxillary premolar associated with the distance between the buccal cusp tip and the edge of the developing root at midline. All measurements were performed using a digital sliding caliper (Fisher Scientific, Waltham, MA, USA) directly on the (unsectioned) extracted teeth, using a standard source of white light – as needed for translucency assessment – by a single trained examiner. Estimated individual ages reported here, ranging from 9.4 to 100.8, were taken from a previously published study of the same clinical sample (21).

Figure 1.

The Bang and Ramm method of age estimation involves measurement of translucency length values from the apex of the root to the enamel-dentin junction.

Occlusal topography methodology

Teeth with excessive chipping or breakage were excluded from this study as their occlusal surfaces could not be properly analyzed. A 3M True Definition clinical scanner (3M Oral Care, Monrovia, CA, USA) was utilized to scan the premolars. This scanner was designed to generate three dimensional models of teeth in a clinical setting and is readily available for use in many dental practices.

MeshLab and Geomagic Wrap were used to standardize the axes of each scan and to isolate the functional occlusal surface of the premolars (22, 23). Topographic analyses were performed on the cropped occlusal surfaces of each scan using the molaR software package in RStudio (24, 25). The molaR package was used to generate three parameters characterizing crown surface topography: Slope, Relief Index (RFI), and Orientation Patch Count rotated (OPCr).

Slope is defined as the average change in elevation across an occlusal surface (Figure 2). Higher values for Slope are indicative of the presence of higher premolar cusps, which typically corresponds to lower levels of gross wear (17, 20). RFI is a measure of the ratio of a tooth’s three-dimensional surface area to its two-dimensional planimetric view. OPCr represents the average orientation patch count of a surface measured at various orientations and indicates the complexity of a tooth’s occlusal surface based on the number of faces sorted into eight ordinal directions (19) (Figure 2). RFI typically decreases as teeth flatten with higher levels of wear and their surface areas decrease. On the other hand, OPCr has been shown to be relatively insensitive to variation in gross wear based on comparisons of different primate species (26), although this attribute tends to increase with simulated wear as occlusal surfaces become more multi-faceted (20).

Figure 2.

Dental topographic analysis of human upper third premolars. Left: molaR-generated images for Slope values as indicated, right: molaR-generated OPCr example.

Smith score methodology

Macrowear was scored from the premolar scans using the Smith system (2). The Smith system was originally designed for use in bioarchaeological samples. It utilizes an ordinal scale from 1 to 8, ranging from low (“unworn to polished or small facets”) to high (“complete loss of crown, no enamel remaining”) levels of occlusal wear, respectively. Each macrowear score is distinguished by standardized, wear-based morphological characteristics, such as facet development, enamel loss, and dentin exposure. Smith scores for this sample were restricted in range from 1 to 4 since the most worn specimens had exposed patches of dentin. However, occlusal enamel was still present.

Statistical analyses

All statistical analyses were conducted using the Systat 12 (27). Data sets were compared two-by-two. First, BRLM age estimates were compared with each topographic attribute (slope, RFI, OPCr) using Spearman’s non-parametric correlation coefficients. Smith scores were compared with BRLM age estimates and topographic data using general linear models. An ANOVA model was used with Smith score as the independent variable, while BRLM age estimate was used as the dependent variable. Age estimates were rank transformed to mitigate violation of assumptions inherent in parametric statistics (28). Tukey’s HSD and Fisher’s LSD pairwise comparisons tests were used to balance risks of Type I and Type II error while determining sources of significant variation (29). A MANOVA model was used to assess the effects of Smith score on rank-transformed topographic data. This allowed comparisons of all individuals in each group. Separate ANOVAs were performed for Slope, RFI, and OPCr, and Tukey’s HSD and Fisher’s LSD tests were again used to assess sources of significant variation.

Results

Summary statistics are given in Table 1, and results are presented in Figures 3-4 and Tables 2-4. Raw data are available in the supplemental file. Spearman’s correlation coefficient values were not significant for comparisons between BRLM age estimates and slope (r_s = 0.122, df = 197, p = 0.088), RFI (r_s = 0.092, df = 197, p = 0.199), or OPCr (r_s = 0.042, df = 197, p = 0.558) (Table 2). In other words, premolar crown shape, as reflected in dental topographic attributes measured in this study, did not vary with BRLM age estimates (Figure 3).

Table 1. Summary statistics. Median, mean, and SD values for Bang and Ramm/Liversidge and Molleson (BRLM) age estimates and dental topographic attributes (Slope, RFI, and OPCr) by Smith score.

Smith Score 1, n = 114				Smith Score 3, n = 23
	Median	Mean	SD		Median	Mean	SD
BRLM	26.100	36.678	21.135	BRLM	51.300	56.548	17.861
Slope	54.046	53.988	2.384	Slope	52.152	51.883	3.020
RFI	0.291	0.291	0.029	RFI	0.279	0.274	0.030
OPCr	96.685	97.794	22.57	OPCr	103.880	111.994	33.021
Smith Score 2, n = 48				Smith Score 4, n = 12
	Median	Mean	SD		Median	Mean	SD
BRLM	46.850	50.373	18.391	BRLM	46.000	50.708	14.606
Slope	51.833	51.997	2.675	Slope	49.207	48.71	4.317
RFI	0.271	0.274	0.030	RFI	0.253	0.244	0.040
OPCr	110.435	113.974	28.085	OPCr	118.250	127.989	30.095

Figure 3.

Comparisons of Bang and Ramm/Liversidge and Molleson (BRLM) age estimates with dental topography attributes and Smith wear scores for the clinic sample. Bivariate plots of dental topography attributes against BRLM values on top and lower left. Box-and-whiskers plot of BRLM values for different Smith scores on the bottom right. The hinges mark the first and third quantiles, the vertical lines between them are medians, each whisker represents a value 1.5 times the interquartile range, and the asterisks indicate outliers.

Figure 4.

Box-and-whiskers plots comparing topographic attribute values for each Smith score. The hinges mark the first and third quantiles, the vertical lines between them are medians, each whisker represents a value 1.5 times the interquartile range, and the asterisks and circles indicate outliers and far outliers, respectively.

Table 2. Spearman’s correlation coefficients for comparisons of Bang and Ramm/Liversidge and Molleson (BRLM) age estimates with dental topography attributes (Slope, RFI, and OPCr). Sample size, n = 197.

	Slope	RFI	OPCr
rs	0.122	0.092	0.042
p	0.088	0.199	0.558

Table 3. ANOVA with Smith Score as the independent variable and Bang and Ramm/Liversidge and Molleson (BRLM) age estimates as the dependent variable for the clinical sample (top). Pairwise comparisons of results for different Smith scores (bottom). All data rank-transformed prior to analysis.

	F	df	p
ANOVA	13.150	3, 193	<0.001
	2	3	4
1	-41.818**	-57.157**	-43.735**
2		-15.339	-1.917
3			13.422

**p < 0.05 for Tukey’s HSD and Fisher’s LSD tests.

Table 4. Test results for general linear model tests comparisons of Smith score categories for topographic attributes (Slope, RFI, and OPCr). MANOVA (Wilks λ) and individual ANOVA values (top) followed by pairwise comparisons of variables showing significant variation. All data rank-transformed prior to analysis.

	F	df	p
Wilks λ	10.081	9, 464	<0.001
Slope	15.830	3, 193	<0.001
RFI	10.299	3, 193	<0.001
OPCr	7.754	3, 193	<0.001
Slope	2	3	4
1	43.184**	42.655**	79.601**
2		-0.529	36.417*
3			36.946*
RFI	2	3	4
1	35.072**	32.571**	70.031**
2		-2.502	34.958*
3			37.460*
OPCr	2	3	4
1	-33.811**	-24.511*	-59.561**
2		9.301	-25.750
3			-35.051

**p < 0.05 for Tukey’s HSD and Fisher’s LSD tests.

*p ≤ 0.05 for Fisher’s LSD but not Tukey’s HSD tests.

ANOVA test results (Table 3) did, however, show significant variation between wear scores and BRLM age estimates (F = 13.15, df = 3, 193, p < 0.001). Median BRLM values tended to be higher with higher Smith scores (Figure 3). This means that elderly individuals tended to show greater premolar wear. Tukey’s HSD test results show that Smith score 1 had significantly younger BRLM age estimates than Smith scores 2-4.

The MANOVAs (Table 4) also indicated significant variation between wear scores and topographic attributes (Wilks λ F = 10.081, df = 9, 464, p < 0.001). ANOVA results indicated that Slope (F =15.83, df = 3, 193, p < 0.001), RFI (F = 10.299, df = 3, 193, F = < 0.001, and OPCr (F = 7.754, df = 3, 193, p < 0.001) all varied significantly by Smith score. There is a tendency for decreasing Slope and RFI and increasing OPCr with increasing wear (Figure 4). In other words, the crown surfaces get flatter but more faceted with wear. In nearly all cases, Smith score 1 differed significantly from scores 2-4 based on Tukey’s HSD test results, except for OPCr comparison of Smith scores 1 and 3, which differed by Fisher’s LSD test only. Smith scores 2 and 3 also differed by Fisher’s LSD test from score 4 for slope and RFI.

Discussion

Results presented here indicate that there is no association between BRLM age estimate and dental topography attributes (Slope, RFI, or OPCr). On the other hand, there is rough correspondence between age as estimated by BRLM and level of gross tooth wear measured by Smith score. Less worn teeth tend to evince younger age estimates with the BRLM method, especially when comparing individuals with premolar crowns characterized as Smith score 1 (mean age: 37) with those characterized as Smith scores 2-4 (mean ages: 50-57). Still, age estimates vary considerably within each Smith score sample, and there are no significant differences in BRLM estimates between higher Smith score samples. This suggests a fairly limited application of premolar wear scores to biological profiling in bioarchaeological and, especially, forensic contexts, with the extent of their utility being to define broad age cohorts of “younger adults” (<50 years) and “older adults.” We note, however, that these results also reflect the challenges of applying a method originally designed for archaeological samples to modern populations characterized by limited macrowear. We discuss these challenges below.

Furthermore, there is a rough correspondence between individual topographic attributes (Slope, RFI, OPCr) and level of gross tooth wear measured by Smith score. Less worn teeth tend to have steeper occlusal slopes, more relief, and less faceting (lower OPCr values) than more worn specimens. But again, there is a lot of variation in topographic attributes within each Smith score sample, and the significant differences are between the extremes (score 4 and especially score 1) and between them and intermediate Smith scores.

The lack of association between BRLM age estimates and occlusal topography parameters is especially unexpected given that BRLM age estimates and occlusal topography parameters both track gross wear as measured by the Smith system, albeit roughly. To put it simply, occlusal Slope, RFI, and OPCr values cannot be used to predict the chronological age of an individual for our clinical sample, at least as estimated by BRLM procedures. The biological variation in tooth shape between individuals appears to be too great, at least for this sample, which represents a contemporary and genetically-variable population of U.S. dental patients. On the other hand, the rough correspondence between Smith scores and BRLM estimates is consistent with increasing tooth wear with age, at least when comparing individuals with lowest degree of premolar wear to other individuals. Furthermore, differences between occlusal topography attributes by Smith score suggest that crown shape changes in a fairly consistent manner as wear progresses. Again, this is especially the case for comparisons of individuals with unworn or minimally worn teeth (Smith score 1) to other individuals.

One possible explanation for the lack of association between age estimates and occlusal topography attributes is that genetically mediated variation in occlusal form or variation in diet among individuals adds noise to the system, obscuring any relationship between the two variables for this clinical sample. Analyses of samples consisting of more homogeneous populations (e.g., uniform bioregional affiliation, similar dietary behaviors) than that represented by the Indiana University School of Dentistry’s Oral Health Research Institute tooth bank is needed to test these hypotheses. In addition, this sample is restricted in its macrowear profile (i.e., no individuals characterized by Smith scores 5-8), likely due, in part, to the relatively soft, highly processed diet typical for US populations today. The lack of highly worn teeth in this study may have obscured relationships between macrowear, age, and occlusal topography, thus impacting our results.

Our results are also limited because there were no definitive age values for the study. In this way, our findings are more directly transferrable to bioarchaeological contexts, where unknown chronological age is reconstructed from any available skeletal or dental indicator. In contexts where teeth are the only elements preserved, it is useful to understand the complex relationship between progressive changes across distinct hard tissues (i.e., root translucency versus crown wear). Importantly, our results indicate that these relationships are not always one-to-one. While the lack of age and other identifying information for this sample allowed the teeth to be anonymized and dissociated from any patient data in accordance with IRB-approved protocols, our results could, ideally, have compared known ages to each of these three age estimation methods. Further studies are needed to evaluate individual age and sex for these purposes.

In addition, further research is needed to consider other tooth types and to devise more specific applications and combinations of these three methods to best serve bioarchaeological, functional morphological and forensic purposes. Clement and Hillson found that a divergence of wear patterns occurred based on cultural practices and a gendered division of labor within a genetically isolated Inuit community, for example, which indicates a need to better understand the confluence of both biological and social factors that can impact patterns of tooth wear beyond age (30).

Conclusions

Smith scoring, occlusal topographic analyses and BRLM age estimation have roles to play in the study of tooth wear, shape, and age estimation. That said, none of these attributes are an especially good proxy for other attributes, at least for maxillary first premolars, and none of these attributes can replace the other attribute, given the evidence for within-sample variation. Thus, there is value in combining approaches to gain the most complete picture possible of how teeth in different populations change shape with wear and age, especially in bioarchaeological contexts (31, 32).