The impact of activity on pelvic age-at-death estimation
Abstract
This study assesses whether greater levels of physical activity result in increased rates of degenerative change in the pubic symphysis and the iliac auricular surface. The sample comprises 131 skeletons from the Athens Collection. Skeletal age-at-death was estimated from the pubic symphysis and the iliac auricular surface. Skeletal activity was assessed using femoral cross-sectional geometric (CSG) properties and fibrocartilaginous entheses. The association between skeletal age stages, CSG properties and entheseal changes (EC) was tested using Spearman correlation followed with partial Spearman correlation controlling for the effect of documented age-at-death and estimated body mass, as well as generalised linear models. Moreover, Kruskal–Wallis tests were used to compare the EC and CSG values among individuals who were underaged, correctly aged and overaged using the pubic symphysis and the iliac auricular surface. Our results show a negative correlation of skeletal age stages with particular CSG properties, implying a decrease in skeletal rigidity as age increases, and a positive correlation with EC scores, suggesting that older individuals exhibit more pronounced EC. Controlling for documented age-at-death and body mass produced low correlation coefficients and very few statistically significant results. The difference in EC and CSG values among underaged, correctly aged and overaged individuals was significant only for ECs in individuals over 50 years old. The present study highlights that the effect of activity on pelvic age markers is not pronounced, because a limited association between activity and skeletal degeneration in the pubic symphysis and iliac auricular surface was found. Considering the difficulty in identifying past occupations skeletally, our study supports that this often missing parameter in past osteobiographies does not affect skeletal age estimation. Moreover, our results suggest that taking into account the EC scores when evaluating skeletal age can provide further insights in age-at-death estimation in older individuals.
1 INTRODUCTION
Estimation of age-at-death holds a central role in skeletal analysis, whether the remains come from archaeological or modern contexts. In the case of bioarchaeological assemblages, age-at-death estimation is key to palaeodemographic reconstruction, and age is also an important factor to consider when examining palaeopathology, activity and other osteobiographic markers (Larsen, 2015). In forensic anthropology, age-at-death is one of the main elements of the biological profile of an individual, along with sex, stature and ancestry, contributing greatly to the individuation process (Christensen, Passalacqua, & Bartelink, 2014).
Age-at-death estimation in adult remains focuses on different parts of the skeleton, such as the pelvis, thorax and cranium. Some of the most frequently employed methods utilise the morphology of the pubic symphysis and the iliac auricular surface, and these are the anatomical areas that the current paper will also examine (Berg, 2008; Brooks & Suchey, 1990; Buckberry & Chamberlain, 2002; Chen, Zhang, & Tao, 2008; Gilbert & McKern, 1973; Hanihara & Suzuki, 1978; Hartnett, 2010; Igarashi, Uesu, Wakebe, & Kanazawa, 2005; Katz & Suchey, 1986; Kimmerle, Konigsberg, Jantz, & Baraybar, 2008; Lottering, Reynolds, MacGregor, Meredith, & Gregory, 2014; Lovejoy, Meindl, Pryzbeck, & Mensforth, 1985; McKern & Stewart, 1957; Meindl, Lovejoy, Mensforth, & Walker, 1985; Osborne, Simmons, & Nawrocki, 2004; Todd, 1920, 1921a, 1921b). It should be noted that the above-mentioned methods are macroscopic and nondestructive, therefore, easily applicable at a very low cost, and not imposing any of the legal and ethical issues linked with destructive analysis.
Despite the importance of accurate age-at-death estimation and the multiplicity of available methods, the correlation between skeletal age indicators and chronological age is imperfect, an issue that tends to be aggravated as an individual's chronological age increases. This phenomenon is due to the fact that in adults, skeletal age is largely assessed based on the degeneration of the skeleton, and this degeneration is controlled by a number of factors beyond age (Mays, 2015; Nawrocki, 2010). As a result, there is great interindividual and interpopulation variability in the rate of skeletal degeneration and associated skeletal ageing, which renders current skeletal age estimation methods imperfect and limits the applicability of ageing standards in different populations (e.g. Bocquet-Appel & Masset, 1982; Cowie, Bekvalac, & Kausmally, 2008; Falys, Schutkowski, & Weston, 2006; Rissech, Wilson, Winburn, Turbón, & Steadman, 2012). The latter issue has given rise to Bayesian applications in age-at-death estimation during the past years (Boldsen, Milner, Konigsberg, & Wood, 2002; Konigsberg & Frankenberg, 1992; Nikita & Nikitas, 2019; Prince, Kimmerle, & Konigsberg, 2008).
In his review of factors that may affect the expression of skeletal age markers besides age itself, Mays (2015) highlighted that the morphological changes seen on the pubic symphysis can be linked to two primary factors: hormones and activity. Increased oestrogen levels, such as during pregnancy, lead to bony resorption at the pubis (Gardner, 1936; Hall, 1950; Tague, 1988; Vix & Ryu, 1971). Relaxin, a hormone secreted during pregnancy, increases joint laxity in order to allow the joint space at the pubic symphysis to widen sufficiently during childbirth (Dehghan et al., 2014; Keriakos, Bhatta, Morris, Mason, & Buckley, 2011; Marnach et al., 2003). Although joint laxity is greatly reduced after parturition, there remains greater mobility in the symphysis in parous women, which accelerates degenerative changes in this region (Tague, 1988). With regard to the role of activity, stress injuries are often found at the pubic symphysis (osteitis pubis) of individuals regularly involved in strenuous activities and in athletes (Cunningham et al., 2007). Although such injuries may be of an acute nature, most often they result from long-term mechanical stresses applied upon the pelvis (Budak & Oliver, 2013). Bone alterations include erosion of the subchondral bone, sclerosis, cyst formation and, less often, periarticular osteophytes (Budak & Oliver, 2013; Cunningham et al., 2007; Lentz, 1995).
Mays (2015) identified sex and mechanical strain as the key factors affecting also iliac auricular surface morphology. In specific, the increased laxity that characterises the sacroiliac articulation during pregnancy and the greater range of motion of the sacroiliac joints in females compared with males (Damen et al., 2001) must be linked to an earlier occurrence and more rapid progress of degenerative changes in females (Shibata, Shirai, & Miyamoto, 2002). In addition, physically demanding tasks generally increase the risk for osteoarthritis and acute joint injury (Richmond et al., 2013). In the latter case, a notable example is osteitis condensans ilii, a stress injury of the sacro-iliac articulation often seen in athletes and occasionally in other groups (Mitra, 2010). In the case of the iliac auricular surface, Mays (2015) identified two additional factors that may affect the rate of degeneration, that is, genetic predisposition and obesity. With regard to the latter, obesity has been linked to the development of osteoarthritis as it imposes increased mechanical loads and biochemical alterations (Pottie et al., 2006; Sowers, 2001), and it also affects the posture of individuals and their motion patterns in sitting and standing, which ultimately also affects the patterns of joint degeneration (Gilleard & Smith, 2007). Hence, body mass may also play a role in the morphology of the iliac auricular surface and by extent affect the relevant skeletal aging markers.
The aim of the current study is to assess whether greater levels of physical activity result in an increase of the rate of degenerative change in the pubic symphysis and the iliac auricular surface. The pelvic joints are associated with locomotion and weight bearing. Thus, a more physically demanding lifestyle is anticipated to lead to an increase in pressure that can affect the rate of degenerative change of the pubic symphysis and iliac auricular surface. Consequently, individuals with more physically active lifestyles should appear older than they are. At the same time, the pubic symphysis and sacroiliac joints are not experiencing identical stress and subsequent joint degeneration; hence, it is interesting to explore how age-at-death estimates based on each of them may be differentially affected. Our study builds on earlier work by Campanacho, Santos and Cardoso (2012), who examined the effect of physical activity on the pubic symphysis, but extends this further by examining also the iliac auricular surface, employing a range of skeletal markers of robusticity and including among the predictors sex and body mass.
2 MATERIALS AND METHODS
2.1 The Athens Collection
The dataset used to test the effect of activity on pelvic skeletal age included 131 skeletons from the University of Athens Modern Human Skeletal Reference Collection (hereafter Athens Collection), housed at the Department of Animal and Human Physiology at the National and Kapodistrian University of Athens, Greece. The Athens Collection comprises the skeletal remains of individuals who lived mostly in the 20th century and were buried in cemeteries in the area of Athens. The age-at-death, sex, occupation, place and year of birth, as well as the cause of death for most of these individuals are documented in their death certificates (Eliopoulos, Lagia, & Manolis, 2007; Nikita, 2020). The distribution of the material per sex and age group is given in Table 1. The selection of the skeletons was made on the ground of good preservation of (a) the pubic symphysis and iliac auricular surface, (b) full length of the femur, (c) femoral head diameter and (d) bone surfaces so that the presence or absence of femoral entheseal changes (EC) could be recorded. All selected individuals lacked any evidence of genetic disorders that could potentially act as confounding variables to this sample. In addition, we tried to keep the sample as balanced as possible with regard to the representation of males and females as well as different age groups.
| Age group | Males | Females |
|---|---|---|
| 20–30 | 13 | 4 |
| 31–40 | 7 | 6 |
| 41–50 | 13 | 10 |
| 51–60 | 16 | 9 |
| 61–70 | 14 | 6 |
| 71–80 | 7 | 11 |
| >81 | 6 | 9 |
| Total | 76 | 55 |
2.2 Skeletal age-at-death estimation methods
The skeletal age-at-death estimation methods used included the Brooks and Suchey (1990) method for the pubic symphysis and the Lovejoy et al. (1985) as well as the Buckberry and Chamberlain (2002) methods for the iliac auricular surface. These methods will be denoted as SB, L and BC, respectively. Data collection took place in the context of an earlier study (Xanthopoulou, Valakos, Youlatos, & Nikita, 2018) where intrarater and interrater agreement were also examined and found to be statistically significant (Cohen's κ > 0.7, p-value < 0.05). In the context of this earlier study, the correlation between documented age-at-death (chronological age) and the skeletal stages used in the above methods was tested, and it was found to be statistically significant for both sexes and for all three methods (rs = 0.32–0.79, p-value < 0.01; Table 6 in Xanthopoulou et al., 2018). Similarly, the percentage of individuals assigned to the correct age class using these methods was around 70% and over 80% for males and females, respectively, when using SB, at least 78% for both sexes when using L, but merely 56% when adopting BC (Table 2 in Xanthopoulou et al., 2018).
2.3 Activity reconstruction
As mentioned above, the profession/occupation of many individuals that form the Athens Collection is also documented. However, we decided to omit this information and assess the physical activity levels skeletally because documented occupations in modern reference collections are often misleading (Cardoso & Henderson, 2013). In particular, in the Athens Collection, many occupations were vague, such as ‘civil servant’ or ‘housekeeper’. Furthermore, death certificates usually state only the last known occupation. These limitations do not allow the assessment of how mechanically strenuous daily life was for these individuals. Hence, the activity levels of each individual were assessed skeletally using the cross-sectional geometric (CSG) properties of the femur, as well as femoral fibrocartilaginous EC.
The CSG properties used in the analyses included the total subperiosteal area (TA), which is a measure of bone robusticity particularly with respect to tension, compression or shear, as well as ratios of second moments of area Ix/Iy and Imax/Imin, whereby Ix, Iy, Imax and Imin express resistance to bending loads, and their ratios are indicative of cross-sectional shape as they represent the distribution of cortical bone within the section (Ruff, 1987, 2008). These properties were calculated from virtual 3D bone models of the Athens Collection with the long-bone-diaphyseal-CSG-Toolkit (Bertsatos & Chovalopoulou, 2019). The CSG Toolkit is a free computer software that automatically analyses 3D models of human long bones and provides highly accurate measurements of their CSG properties along the diaphyseal shaft at regular intervals (20%, 35%, 50%, 65% and 80% of the maximum bone length). These properties are estimated based on the subperiosteal contours of the femur, that is, they exclude the medullary cavity. This approach has been found to exhibit a very high correlation with true cross-sectional geometry (e.g. r2 = 0.998 in Stock & Shaw, 2007). It must be noted that with increasing age, there is endosteal resorption and cortical thinning, which might affect the estimation of CSG (Ruff & Hayes, 1983), especially when using a method that excludes the medullary cavity. Nonetheless, Sparacello and Pearson (2010) showed that cortical thickness has minimal impact on average biomechanical properties; thus, the data obtained from subperiosteal contours may accurately reflect bending rigidity.
To avoid including too many intercorrelated parameters in our statistical tests, we only used the properties estimated at the midshaft (50% of maximum length). The practice of using the midshaft value as a sufficient measure to express diaphyseal rigidity has been adopted by numerous scholars (e.g. Sparacello & Pearson, 2010; Stock & Shaw, 2007). The femoral 3D models used in the present study have been reconstructed from their respective dry bone counterparts by means of 3D photogrammetry with a rated accuracy of ~0.2 mm (Bertsatos, Gkaniatsou, Papageorgopoulou, & Chovalopoulou, 2019). The rated accuracy corresponds to the maximum femoral length measurement, which implies that the expected measurement error in CSG calculated values is much less.
The femoral fibrocartilaginous entheses examined included gluteus minimus, psoas major, iliacus and gluteus medius. All entheses were preferably recorded on the left side limbs, unless these were damaged, in which case the right side was preferred. EC were recorded using the New Coimbra Method (Henderson, Mariotti, Pany-Kucera, Villotte, & Wilczak, 2016). An estimate of repeatability of the recorded values was established by interval-by-interval intraobserver agreement on a subsample of 22 femurs, which were recorded twice with a gap of 3 weeks between each recording session. Intraobserver agreement was calculated for each individual variable based on the scored difference and an average of 92.02% was established over all entheses examined. Given that this recording scheme involves multiple variables, such as bone formation and erosion, microporosity and macroporosity, cavitations, and textural change, in most statistical analyses, we summed the values of these variables so that each enthesis would be represented by a single composite value. We acknowledge that this approach has the limitation that certain variables (EC) may be more strongly linked to activity levels and this effect may be lost within a composite score. For this reason, we also performed selected analyses using every single EC variable, as outlined below, in order to make sure that we have avoided missing any significant activity effect. Note that the effectivity of EC as skeletal activity markers has been criticised in a number of studies (e.g. Cardoso & Henderson, 2010; Michopoulou, Nikita, & Valakos, 2015; Milella, Belcastro, Zollikofer, & Mariotti, 2012; Nikita, Xanthopoulou, Bertsatos, Chovalopoulou, & Hafez, 2019; Weiss, 2003, 2004), many of which have stressed that the primary factors affecting EC expression are age and body size/mass rather than mechanical stress (Cardoso & Henderson, 2010; Milella et al., 2012; Niinimäki, 2011; Weiss, 2003, 2004, 2007; Weiss, Corona, & Schultz, 2012; Wilczak, 1998). Nonetheless, recent research has experimentally reaffirmed the link between activity levels and EC (Karakostis, Jeffery, & Harvati, 2019Karakostis, Wallace, Konow, & Harvati, 2019). On this ground, we have included EC as a skeletal activity marker, bearing in mind its inherent limitations. Note that individuals exhibiting seronegative spondyloarthropathies, DISH or other diseases that may accentuate the expression of EC were not included in our sample (Villotte & Knüsel, 2013). Additional to these pathological categories, any ‘bone-formers’, that is, individuals with an inherent tendency toward bone formation, were similarly excluded from the sample (Mays, 2016; Rogers, Shepstone, & Dieppe, 1997).
Tables S1 and S2 provide descriptive statistics for the CSG properties and EC per sex and age group.
2.4 Body mass estimation
The body mass of the individuals comprising the Athens Collection is not known; hence, it had to be skeletally estimated. The estimation of body mass from skeletal remains may adopt ‘mechanical’ or ‘morphometric’ approaches (Auerbach & Ruff, 2004). Mechanical approaches are based on the high correlation that exists between body mass and the dimensions of the skeletal elements that support the weight of the body, mostly the femoral head diameter (Auerbach & Ruff, 2004). Morphometric approaches view the human body as a cylinder, with stature equal to the cylinder's height and biiliac breadth equal to its diameter. As such, these methods use the stature and biiliac breadth to estimate body mass (Ruff, 2010). To produce population-specific equations for body mass estimation, a so-called ‘hybrid’ approach may be employed, whereby the morphometric method is used to calculate the body mass, and then regression equations are obtained using femoral head diameter (e.g. Ruff et al., 2012). This approach was adopted by Nikita and Chovalopoulou (2017) based on the Athens Collection using 75 individuals whose skeletal remains were sufficiently well preserved to allow the estimation of stature and the measurement of biiliac breadth so that the morphometric method could be applied. The regression equations generated from that study for the estimation of body mass based on femoral head diameter exhibited low random and directional error and were adopted in the current paper to estimate body mass in the 131 skeletons under examination. At this point, we must stress that a rather recent study found that the estimation of body mass from skeletal remains is only accurate and reliable for estimating the population average body mass, not that of individuals (Jeanson et al., 2017). Despite this potential limitation, body mass here is used only as a proxy of body size to examine if larger body size affects the rate of degeneration of the auricular surface and pubic symphysis and not as an absolute measure of the weight of the individuals under study.
2.5 Statistical analysis
First, we assessed whether there is a significant bilateral asymmetry in the skeletal age marker scores in order to test if we need to examine the association between activity and skeletal ageing in each side separately or not. Wilcoxon signed ranks tests were used to test for the aforementioned bilateral asymmetry. Subsequently, the correlation between skeletal age stages and CSG properties was tested using Spearman correlations. Additionally, we adopted Monte Carlo permutations for the p-value with 1000 iterations. This analysis was followed up with partial Spearman correlations to test again the correlation between skeletal age stages and CSG properties but this time controlling for the effect of chronological age and body mass. Spearman correlations were also used to test the correlation between skeletal age stages and summed EC scores, whereas partial Spearman correlations tested this association controlling for the effect of chronological age and body mass. All (partial) correlations were ran for pooled sexes and separately for males and females to test for any sex bias. Partial Spearman correlations with chronological age and body mass as covariates were rerun for the EC, this time separately for each individual variable, rather than for the summed scores, in order to assess whether there is a significant pattern missed due to the summing process. In addition, generalised linear models (GLM) were used to examine simultaneously the effect of sex, chronological age, body mass, EC sums and CSG properties on the skeletal age stages. In this case, the tests were repeated twice, treating EC sums both as ordinal and as continuous/scale data. Note that EC scores are ordinal variables; however, there are instances in bioarchaeology and forensic anthropology where ordinal variables are treated as continuous when modelling data (e.g. Klales, Ousley, & Vollner, 2012; Walker, 2008). Such an approach is based on the fact that models based on continuous data are much simpler, whereas when using ordinal variables, all levels of each factor should be present in the dataset, a precondition often not satisfied. For these reasons, in the current study, we treated EC scores in the GLM models both as ordinal and as continuous variables (for more details, see Milella, Belcastro, Mariotti, & Nikita, 2020).
Considering that the accuracy of skeletal ageing methods tends to decrease in older individuals, as an additional means of exploring the effect of activity on skeletal ageing, we divided our sample in individuals younger than or equal to 50 years and older than 50 years. Subsequently, we classified each individual in one of three categories based on the performance of each ageing method: underaged, correctly aged and overaged. This classification was based on whether an individual's chronological age fell within the interval mean age ± standard deviation provided in the publication of the SB and BC methods and within the age intervals suggested for the L method. The EC scores and the CSG values were then compared among the underaged, correctly aged and overaged individuals using Kruskal–Wallis tests in order to assess if these skeletal activity markers are significantly differentiated in individuals for whom skeletal ageing exhibits differential success.
All correlations were ran using Excel macros developed by Nikita (2017), whereas Wilcoxon signed ranks tests, GLM and Kruskal–Wallis tests were ran in SPSS v. 21.0. The data that support the findings of this study are available from the corresponding author.
3 RESULTS AND DISCUSSION
No statistically significant difference was found in the bilateral expression of any of the skeletal age stages (SB: Wilcoxon Z = −0.218, p-value = 0.827; L: Wilcoxon Z = −1.753, p-value = 0.080; BC: Wilcoxon Z = −0.865, p-value = 0.387). Based on these results, we decided to analyse the impact of activity only on the left os coxae.
First, we tested the Spearman correlation between skeletal age stages and CSG properties, additionally adopting Monte Carlo permutations for the p-value with 1000 iterations. The results are given in Table 2. For pooled sexes, it can be seen that this correlation is very small (rho values ranging from −0.225 to 0.149). Nonetheless, it is statistically significant for Ix/Iy (p-values < 0.05) for all three skeletal age methods. The negative correlation coefficients for Ix/Iy suggest that as skeletal degeneration increases, the rigidity of the skeleton to bending forces expressed as the distribution of cortical bone along the antero-posterior versus medio-lateral axes of the cross-sectional plane overall decreases, as expected. For TA, rho is also negative, suggesting a decrease in the rigidity of the skeleton to compressive, tensile and shear forces as skeletal degeneration increases; however, this was only significant for the pubic symphysis (p-value = 0.03). In the case of Imax/Imin, we obtain positive rho values, which might appear contradictory to the Ix/Iy results as they suggest an increase in bending rigidity with increasing skeletal degeneration. Nonetheless, Imax and Imin represent orthogonal maximum and minimum bending rigidity along an arbitrary axis on the cross-sectional plane. Taking into account that TA decreases with skeletal age, this merely reflects that the cross-sectional area becomes more elongated suggesting that the femoral diaphysis weakens even further along some particular orientation of bending force. In any case, this pattern was only significant for BC (p-value = 0.04).
| Pooled sexes | Males | Females | |||||
|---|---|---|---|---|---|---|---|
| Skeletal ageing method | CSG property | Spearman's rho | Partial Spearman's rho | Spearman's rho | Partial Spearman's rho | Spearman's rho | Partial Spearman's rho |
| SB | TA | −0.165* | −0.075 | 0.047 | −0.136 | 0.233* | 0.188 |
| L | TA | 0 | −0.079 | 0.094 | −0.097 | 0.164 | 0.119 |
| BC | TA | −0.072 | −0.188* | 0.069 | −0.173 | −0.089 | −0.239* |
| SB | Ix/Iy | −0.225** | −0.042 | −0.190* | −0.072 | −0.128 | −0.010 |
| L | Ix/Iy | −0.154* | 0.047 | −0.099 | 0.069 | −0.124 | 0.026 |
| BC | Ix/Iy | −0.137* | 0.041 | −0.118 | 0.033 | −0.062 | 0.070 |
| SB | Imax/Imin | 0.076 | 0.011 | 0.074 | −0.043 | 0.006 | 0.023 |
| L | Imax/Imin | 0.081 | 0.011 | 0.195* | 0.131 | −0.119 | −0.158 |
| BC | Imax/Imin | 0.149* | 0.115 | 0.306** | 0.301** | −0.115 | −0.137 |
The next step was to examine how the above correlations are affected by the chronological age-at-death and body mass of the individuals. For this reason, we repeated the analysis using partial Spearman's correlations, controlling for the effect of chronological age and body mass. The results are also given in Table 2, and it can be seen that for pooled sexes, the only statistically significant correlation is between TA and BC, whereas rho values cluster even closer to zero. This result supports a very limited association between activity and skeletal degeneration in the pubic symphysis and iliac auricular surface.
When repeating all analyses separately for males and females to test for any sex bias (Table 2), the number of significant correlations between CSG and skeletal age stages decreased notably, whereas when chronological age and body mass were controlled for, merely two significant results were obtained (males BC-Imax/Imin rho = 0.301, p-value = 0.004; females BC-TA rho = −0.239, p-value = 0.014). The smaller number of significant results may be attributed to the smaller sample size as this is now broken down in males and females. In any case, there does not appear to be any clear sex-related pattern in the effect of activity on skeletal age stages, whereas this effect remains minimal, as was the case for pooled data.
Subsequently, we tested the Spearman correlation between the skeletal age stages and summed EC scores. The results are given in Table 3. For pooled sexes, it is seen that the correlation between skeletal degeneration and EC is always positive (rho ranging from 0.156 to 0.642) and statistically significant. This result agrees with earlier studies that have identified a key role of age on EC expression, that is, an increase in EC scores with increasing age (see Section 1). As our study aims at exploring whether heightened activity may accelerate skeletal degeneration in the pubic symphysis and iliac auricular surface, we repeated our analysis using partial Spearman's correlations, controlling for the effect of chronological age and body mass. The results are again given in Table 3. It can be seen that the correlation between EC and skeletal degeneration in the pubic symphysis and iliac auricular surface ceases being significant and the correlation coefficients drop substantially (−0.145 to 0.165). The only significant results are found for the L method for iliacus and psoas major, but they are conflicting as psoas major shows a positive correlation whereas iliacus exhibits a negative correlation with auricular surface degeneration. These results suggest that even though there is a significant correlation between ECs and skeletal age stages, this association is due to the effect of chronological age and to some extent body mass, which simultaneously affect the expression of both variables. Once the effect of age and body mass is removed, there is a lack of association between activity, as captured by means of femoral ECs, and skeletal age stages. In other words, increased activity (greater EC scores) does not appear to accelerate the degeneration of the pubic symphysis or iliac auricular surface.
| Pooled sexes | Males | Females | |||||
|---|---|---|---|---|---|---|---|
| Skeletal ageing method | EC | Spearman's rho | Partial Spearman's rho | Spearman's rho | Partial Spearman's rho | Spearman's rho | Partial Spearman's rho |
| SB | Gluteus minimus | 0.426** | −0.108 | 0.426** | −0.112 | 0.367** | −0.114 |
| L | Gluteus minimus | 0.509** | −0.064 | 0.516** | 0.019 | 0.464** | −0.152 |
| BC | Gluteus minimus | 0.461** | −0.065 | 0.488** | −0.047 | 0.394** | −0.117 |
| SB | Psoas major | 0.511** | 0.018 | 0.523** | 0.010 | 0.431** | −0.027 |
| L | Psoas major | 0.642** | 0.165* | 0.634** | 0.194* | 0.607** | 0.123 |
| BC | Psoas major | 0.502** | −0.049 | 0.534** | −0.036 | 0.420** | −0.109 |
| SB | Iliacus | 0.156* | −0.090 | 0.153 | −0.147 | 0.161 | −0.033 |
| L | Iliacus | 0.174* | −0.145* | 0.138 | −0.201* | 0.232* | 0.021 |
| BC | Iliacus | 0.176* | −0.098 | 0.175 | −0.148 | 0.179 | 0.003 |
| SB | Gluteus medius | 0.367** | −0.050 | 0.395** | −0.057 | 0.323** | −0.031 |
| L | Gluteus medius | 0.410** | −0.097 | 0.415** | −0.060 | 0.382** | −0.071 |
| BC | Gluteus medius | 0.345** | −0.137 | 0.383** | −0.125 | 0.261* | −0.151 |
The examination of an association between ECs and skeletal age stages per sex (Table 3) gave significant results in all cases, except for the iliacus, which was significantly correlated only with L for females (rho = 0.232, p-value = 0.04). When chronological age-at-death and body mass were included as covariates, only two significant patterns were observed and only among the male sample (psoas major—L: rho 0.194, p-value = 0.049; iliacus—L: rho = −0.201, rho = 0.043). These results agree with those for pooled sexes that activity, as expressed by means of femoral EC, does not appear to affect the rate of skeletal degeneration in the pubic symphysis and iliac auricular surface, and this applies to both sexes. Instead, the strong correlation between EC scores and skeletal age stages is due to the chronological age of the individuals, which affects both simultaneously.
As stated above, the fact that we have summed the ordinal scores of the different variables per enthesis may have ‘masked’ some significant effects. Many of the Coimbra method variables systematically exhibited zero values in our dataset, whereas others exhibited diversity in their scores; thus, by summing all scores per enthesis, we may be losing the effect of the variables that are driving variation in EC expression in our sample. To examine whether this is the case, we repeated the partial correlations with chronological age and body mass as covariates, separately for each of the EC variables. The results are given in Table 4, where it can be seen that 31 correlations were significant out of the 288 (~11%). Most of the significant correlations (21 out of the 31) were identified in Zone 1, but no other pattern emerged with regard to the enthesis, the specific EC or the skeletal ageing markers most commonly exhibiting a significant effect.
| SB | L | BC | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| EC | Variable | Pooled sexes | Males | Females | Pooled sexes | Males | Females | Pooled sexes | Males | Females |
| Gluteus minimus | BF Z1 | −0.129 | −0.282** | −0.033 | −0.078 | −0.130 | −0.020 | −0.152* | −0.218* | −0.152 |
| ER Z1 | −0.132 | −0.236* | 0.005 | −0.218** | −0.189* | −0.265* | −0.115 | −0.137 | −0.113 | |
| BF Z2 | −0.087 | −0.114 | −0.082 | −0.078 | 0.022 | −0.130 | −0.059 | 0.014 | −0.098 | |
| ER Z2 | 0.113 | 0.154 | 0.068 | −0.059 | 0.110 | −0.234* | −0.064 | 0.006 | −0.138 | |
| TC | −0.035 | −0.105 | −0.021 | −0.134 | −0.154 | −0.144 | 0.046 | −0.018 | 0.061 | |
| FPO | −0.170* | −0.174 | −0.181 | −0.110 | −0.064 | −0.130 | −0.005 | −0.030 | 0.054 | |
| MPO | −0.060 | 0.018 | −0.133 | 0.056 | 0.161 | −0.090 | −0.064 | 0.060 | −0.227 | |
| CA | 0.032 | 0.094 | 0.035 | −0.013 | 0.009 | −0.005 | −0.075 | −0.065 | 0.008 | |
| Psoas major | BF Z1 | −0.086 | 0.095 | −0.218 | 0.158* | 0.207* | 0.216 | −0.017 | −0.084 | 0.165 |
| ER Z1 | −0.088 | −0.024 | −0.205 | −0.156* | −0.251* | 0.097 | −0.156* | −0.195* | −0.060 | |
| BF Z2 | 0.109 | 0.054 | 0.104 | 0.112 | 0.175 | 0.059 | −0.039 | −0.024 | −0.077 | |
| ER Z2 | −0.009 | −0.103 | 0.061 | −0.037 | −0.052 | 0.005 | −0.105 | −0.098 | −0.165 | |
| TC | 0.073 | 0.130 | −0.010 | 0.101 | 0.140 | 0.017 | 0.072 | 0.122 | −0.025 | |
| FPO | 0.026 | −0.101 | 0.129 | 0.139 | 0.299** | −0.129 | 0.134 | 0.139 | 0.064 | |
| MPO | 0.026 | 0.093 | −0.083 | −0.099 | 0.074 | −0.339** | −0.050 | −0.024 | −0.145 | |
| CA | 0.006 | −0.077 | 0.088 | 0.173* | 0.142 | 0.164 | −0.141 | −0.067 | −0.303* | |
| Iliacus | BF Z1 | −0.244** | −0.266* | −0.152 | −0.217** | −0.212* | −0.146 | −0.188* | −0.319** | 0.010 |
| ER Z1 | −0.103 | −0.105 | 0.035 | −0.035 | −0.023 | −0.005 | −0.072 | −0.068 | 0.009 | |
| BF Z2 | 0.090 | 0.060 | 0.129 | 0.069 | 0.053 | 0.103 | 0.032 | 0.133 | −0.123 | |
| ER Z2 | −0.111 | −0.200* | 0.066 | 0.046 | 0.022 | 0.141 | 0.049 | −0.051 | 0.179 | |
| TC | −0.009 | 0.022 | 0.035 | −0.007 | 0.008 | −0.005 | −0.021 | 0.027 | 0.009 | |
| FPO | 0.020 | −0.021 | −0.028 | −0.061 | −0.154 | 0.117 | −0.008 | 0.009 | 0.005 | |
| MPO | 0.055 | −0.044 | 0.161 | −0.100 | −0.114 | −0.080 | −0.025 | 0.008 | −0.090 | |
| CA | −0.009 | 0.022 | 0.035 | −0.007 | 0.008 | −0.005 | −0.021 | 0.027 | 0.008 | |
| Gluteus medius | BF Z1 | 0.033 | −0.012 | 0.067 | −0.049 | −0.066 | 0.045 | −0.059 | −0.016 | −0.109 |
| ER Z1 | −0.039 | −0.075 | 0.023 | −0.129 | −0.044 | −0.182 | −0.195* | −0.088 | −0.315* | |
| BF Z2 | −0.008 | 0.072 | −0.083 | −0.071 | −0.070 | 0.032 | −0.099 | −0.081 | −0.053 | |
| ER Z2 | −0.067 | −0.097 | −0.024 | −0.166* | −0.113 | −0.204 | −0.176* | −0.204* | −0.135 | |
| TC | −0.024 | 0.022 | −0.009 | 0.010 | 0.008 | 0.017 | −0.024 | 0.027 | −0.025 | |
| FPO | −0.134 | −0.148 | −0.177 | −0.083 | −0.036 | −0.166 | −0.099 | −0.142 | −0.098 | |
| MPO | −0.001 | −0.011 | 0.048 | 0.002 | 0.130 | −0.216 | −0.051 | −0.001 | −0.089 | |
| CA | 0.050 | 0.051 | 0.021 | 0.049 | 0.054 | −0.012 | 0.005 | −0.023 | −0.017 | |
- Abbreviations: BC, Buckberry and Chamberlain (2002); BF Z1, bone formation in zone 1; BF Z2, bone formation in zone 2; CA, cavitation; EC, entheseal changes; ER Z1, erosion in Z1; ER Z2, erosion in Z2; FPO, fine porosity; L, Lovejoy et al. (1985); MPO, macro-porosity; SB, Brooks and Suchey (1990); TC, textural change.
- * p-value < 0.05.
- ** p-value < 0.01.
In order to examine simultaneously the impact of sex, body mass, chronological age, ECs and CSG on skeletal age stages, we employed GLM. Because the analysis of each individual EC variable (Table 4) did not reveal any particular pattern, and rather the general conclusion that there is no systematic effect of EC on skeletal ageing agreed with the results obtained from the summed scores, we decided to use summed EC scores in the GLM. An additional reason for our decision was the fact that the inclusion of too many variables in GLM would have resulted in unnecessary complex models. The results given in Table 5 show that chronological age is the only factor that systematically has a significant effect on the expression of skeletal age stages, as expected. Sex is significant in the pubic symphysis, suggesting that there are differences between males and females in the rate of degeneration of this joint, which are not linked to activity, body mass or chronological ageing and must be attributed to other factors, such as hormonal differences (see Mays, 2015). CSG properties never exhibit a significant effect on skeletal degeneration in the pelvic joints. Finally, among the ECs, only the gluteus minimus and iliacus exhibit some significant effect on the iliac auricular surface morphology. Overall, the GLM results support a strong, as expected, effect of chronological age, significant sexual dimorphism for the pubic symphysis, and limited effect of activity on the iliac auricular surface, as captured by a few of the femoral ECs.
| Dependent variable | Predictors | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Sex | BM | Gluteus minimus | Psoas major | Iliacus | Gluteus medius | Age | TA | Ix/Iy | Imax/Imin | |
| SB | 0.017 | 0.973 | 1.000 | 0.955 | 0.372 | 1.000 | 0.000 | 0.438 | 0.254 | 0.929 |
| L | 0.180 | 0.824 | 0.937 | — | 0.002 | 0.659 | 0.000 | 0.163 | 0.332 | 0.942 |
| BC | 0.053 | 0.132 | 0.009 | 0.097 | 0.107 | 0.511 | 0.000 | 0.351 | 0.176 | 0.635 |
The comparison of EC scores and CSG values among underaged, correctly aged and overaged individuals (Tables 6 and 7) largely corroborates the above findings and offers additional interesting insights. It can be seen that among the CSG properties (Table 6), only TA is very rarely significantly different among these groups, supporting a lack of impact of activity, as deduced through CSG properties, on the degeneration of the pubic symphysis and the iliac auricular surface. With regard to EC (Table 7), when there is no separation of the dataset in younger and older individuals, these are almost always significantly different among underaged, correctly aged and overaged individuals. However, upon closer inspection of the results, it is observed that the significant difference is seen almost exclusively in individuals over 50 years old; hence, it must be linked to the well-documented effect that chronological age has on the expression of EC rather than on the effect that the latter have on skeletal degeneration in the pelvic joints. It is interesting to note that the above pattern was not visible in the L method, where no significant results were obtained when dividing the samples in age groups, contrary to the BC one, where it is very clear. Even though both methods focus on the same anatomical area, the iliac auricular surface, they adopt distinct ways of recording morphological changes in this area, and they associate different age intervals with different morphological changes (notably, the L method places all individuals over 60 years in a single age group). It is also interesting to observe that the significant difference in EC expression between underaged, correctly aged and overaged individuals when employing the SB method is only seen in females over 50 years old but not in males. This sex dimorphic pattern is most likely linked to hormonal changes related to menopause, which affects skeletal degeneration in older females differently than males (Ji & Yu, 2015). However, other sex-related physiological and functional differences such as parturition might explain why this differentiation between males and females is only observed in the pubic symphysis and not in the auricular surface (Becker, Woodley, & Stringer, 2010).
| Ageing method | Groups | TA | Ix/Iy | Imax/Imin | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean ranks | Mean ranks | Mean ranks | |||||||||||
| 1 | 2 | 3 | p-value | 1 | 2 | 3 | p-value | 1 | 2 | 3 | p-value | ||
| SB | All | 73.7 | 66.5 | 40.6 | 0.019 | 60.0 | 67.33 | 76.5 | 0.332 | 66.1 | 68.0 | 54.7 | 0.483 |
| Males | 45.5 | 35.6 | 32.2 | 0.166 | 34.4 | 40.1 | 43.4 | 0.521 | 41.2 | 37.9 | 31.0 | 0.615 | |
| Females | 29.4 | 31.0 | 15.6 | 0.037 | 25.3 | 27.2 | 35.7 | 0.270 | 25.4 | 31.2 | 22.6 | 0.260 | |
| ≤50 years | 29.7 | 29.8 | 19.3 | 0.093 | 14.7 | 28.6 | 25.6 | 0.299 | 12.7 | 29.7 | 23.2 | 0.105 | |
| >50 years | 42.5 | 36.7 | — | 0.254 | 41.1 | 38.0 | — | 0.549 | 40.2 | 38.8 | — | 0.780 | |
| M ≤ 50 years | 18 | 17.1 | 16.2 | 0.972 | 10.5 | 17.7 | 16 | 0.580 | 8.5 | 18.2 | 14.4 | 0.320 | |
| M > 50 years | 24.5 | 19.4 | — | 0.181 | 21.7 | 22.3 | — | 0.884 | 23.7 | 20.2 | — | 0.356 | |
| F ≤ 50 years | 13.0 | 11.3 | 9.3 | 0.701 | 2.0 | 11.3 | 10.6 | 0.325 | 5.0 | 12.8 | 8.6 | 0.188 | |
| F > 50 years | 16.8 | 19.0 | — | 0.529 | 19.2 | 17.0 | — | 0.529 | 16.8 | 19.1 | — | 0.508 | |
| L | All | 73.7 | 61.8 | 66.6 | 0.340 | 59.6 | 63.8 | 76.4 | 0.158 | 57.1 | 71.9 | 63.6 | 0.174 |
| Males | 40.4 | 36.9 | 39.0 | 0.843 | 31.3 | 39.4 | 44.4 | 0.136 | 30.9 | 44.4 | 37.5 | 0.084 | |
| Females | 29.2 | 29.7 | 22.5 | 0.403 | 26.0 | 27.1 | 32.2 | 0.583 | 28.5 | 28.4 | 26.6 | 0.942 | |
| ≤50 years | 26.5 | 24.0 | 28.5 | 0.698 | 22.0 | 28.3 | 28.6 | 0.443 | 18.7 | 30.1 | 29.2 | 0.103 | |
| >50 years | 47.8 | 35.7 | 44.0 | 0.107 | 38.9 | 40.2 | 34.4 | 0.851 | 37.9 | 42.1 | 19.6 | 0.099 | |
| M ≤ 50 years | 18.3 | 12.0 | 18.7 | 0.243 | 11.8 | 14.6 | 19.6 | 0.165 | 11.5 | 17.6 | 18.5 | 0.299 | |
| M > 50 years | 22.0 | 21.8 | 24.0 | 0.958 | 20.9 | 23.9 | 12.7 | 0.310 | 18.2 | 25.6 | 13.3 | 0.086 | |
| F ≤ 50 years | 11.7 | 11.3 | 9.5 | 0.747 | 10.3 | 13.5 | 9.4 | 0.502 | 7.2 | 13.8 | 11.2 | 0.197 | |
| F > 50 years | 20.4 | 17.2 | 23.5 | 0.597 | 12.4 | 18.7 | 22.5 | 0.368 | 25.8 | 17.5 | 5.5 | 0.051 | |
| BC | All | 85.5 | 69.5 | 59.5 | 0.149 | 59.2 | 61.0 | 73.2 | 0.185 | 86.2 | 63.1 | 67.5 | 0.337 |
| Males | 67.3 | 39.2 | 34.8 | 0.049 | 34.0 | 35.2 | 43.5 | 0.267 | 62.0 | 35.6 | 40.2 | 0.115 | |
| Females | 37.0 | 30.8 | 23.6 | 0.162 | 27.0 | 24.9 | 31.7 | 0.312 | 28.0 | 28.4 | 27.6 | 0.985 | |
| ≤50 years | — | 34.1 | 25.4 | 0.107 | — | 25.6 | 27.3 | 0.750 | — | 20.4 | 28.5 | 0.134 | |
| >50 years | 49.8 | 38.4 | 40.1 | 0.494 | 41.2 | 39.7 | 37.5 | 0.936 | 49.7 | 38.5 | 39.6 | 0.514 | |
| M ≤ 50 years | — | 18.0 | 16.6 | 0.716 | — | 14.0 | 18.1 | 0.275 | — | 12.9 | 18.5 | 0.135 | |
| M > 50 years | 37.0 | 20.8 | 21.3 | 0.100 | 21.0 | 22.1 | 22.1 | 0.990 | 33.0 | 20.8 | 23.0 | 0.265 | |
| F ≤ 50 years | — | 15.0 | 10.3 | 0.435 | — | 15.0 | 10.3 | 0.435 | — | 10.0 | 10.5 | 0.931 | |
| F > 50 years | 22.3 | 17.4 | 18.4 | 0.732 | 20.7 | 17.9 | 16.8 | 0.872 | 18.0 | 18.3 | 16.2 | 0.913 | |
| Ageing method | Groups | Gluteus minimus | Psoas major | Iliacus | Gluteus medius | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean ranks | Mean ranks | Mean ranks | Mean ranks | ||||||||||||||
| 1 | 2 | 3 | p-value | 1 | 2 | 3 | p-value | 1 | 2 | 3 | p-value | 1 | 2 | 3 | p-value | ||
| SB | All | 91.8 | 57.4 | 37.0 | 0.000 | 87.7 | 57.7 | 37.0 | 0.000 | 80.2 | 59.2 | 56.4 | 0.006 | 85.8 | 59.1 | 35.5 | 0.000 |
| Males | 53.4 | 32.4 | 24.5 | 0.000 | 49.5 | 32.6 | 24.3 | 0.002 | 46.1 | 33.7 | 38.4 | 0.059 | 49.5 | 33.9 | 21.1 | 0.002 | |
| Females | 40.5 | 25.7 | 11.9 | 0.000 | 40.0 | 26.1 | 11.6 | 0.000 | 34.7 | 26.2 | 21.4 | 0.061 | 37.1 | 25.9 | 14.3 | 0.001 | |
| ≤50 years | 21.5 | 27.4 | 27.3 | 0.780 | 22.7 | 25.9 | 28.8 | 0.722 | 32.5 | 25.9 | 28.6 | 0.649 | 22.7 | 28.2 | 23.1 | 0.438 | |
| >50 years | 46.9 | 32.5 | — | 0.004 | 44.3 | 33.8 | — | 0.034 | 43.6 | 34.5 | — | 0.060 | 44.6 | 33.5 | — | 0.025 | |
| M ≤ 50 years | 17.8 | 16.9 | 17.5 | 0.979 | 14.3 | 16.4 | 17.7 | 0.890 | 19.3 | 16.3 | 19.8 | 0.673 | 17.3 | 17.5 | 14.1 | 0.710 | |
| M > 50 years | 25.5 | 18.4 | — | 0.057 | 22.9 | 19.9 | — | 0.417 | 23.9 | 18.9 | — | 0.164 | 23.2 | 19.6 | — | 0.321 | |
| F ≤ 50 years | 4.0 | 12.2 | 9.4 | 0.251 | 9.0 | 11.2 | 9.9 | 0.833 | 14.0 | 10.1 | 10.6 | 0.778 | 6.0 | 11.6 | 8.9 | 0.379 | |
| F > 50 years | 23.0 | 13.8 | — | 0.007 | 22.9 | 13.8 | — | 0.007 | 20.3 | 16.1 | — | 0.195 | 22.2 | 14.5 | — | 0.024 | |
| L | All | 71.7 | 75.7 | 42.3 | 0.000 | 66.2 | 79.4 | 37.6 | 0.000 | 71.5 | 72.5 | 46.7 | 0.001 | 67.9 | 74.4 | 44.9 | 0.001 |
| Males | 46.8 | 40.4 | 27.4 | 0.009 | 41.4 | 44.8 | 23.6 | 0.001 | 45.8 | 39.9 | 27.6 | 0.011 | 42.5 | 43.3 | 26.0 | 0.006 | |
| Females | 26.4 | 33.3 | 15.4 | 0.003 | 25.5 | 33.8 | 14.9 | 0.001 | 24.3 | 32.4 | 19.8 | 0.029 | 26.4 | 30.8 | 19.6 | 0.092 | |
| ≤50 years | 28.5 | 21.8 | 28.5 | 0.337 | 32.1 | 28.1 | 23.6 | 0.193 | 26.3 | 33.5 | 24.6 | 0.177 | 29.8 | 25.5 | 25.7 | 0.655 | |
| >50 years | 43.8 | 39.4 | 22.8 | 0.169 | 33.5 | 41.7 | 34.5 | 0.315 | 44.3 | 38.4 | 23.2 | 0.124 | 37.1 | 40.1 | 36.2 | 0.828 | |
| M ≤ 50 years | 17.8 | 11.9 | 18.9 | 0.143 | 18.3 | 17.5 | 15.6 | 0.758 | 20.2 | 19.2 | 15.1 | 0.344 | 21.1 | 15.3 | 16.4 | 0.411 | |
| M > 50 years | 25.1 | 21.4 | 10.0 | 0.139 | 19.3 | 22.9 | 22.0 | 0.654 | 24.8 | 20.2 | 14.2 | 0.248 | 20.0 | 23.0 | 18.0 | 0.648 | |
| F ≤ 50 years | 10.7 | 11.6 | 10.0 | 0.868 | 13.2 | 10.9 | 8.8 | 0.278 | 8.0 | 15.4 | 10.1 | 0.096 | 10.1 | 10.6 | 9.7 | 0.951 | |
| F > 50 years | 24.8 | 17.2 | 12.8 | 0.225 | 16.7 | 18.7 | 11.5 | 0.583 | 20.0 | 18.3 | 9.3 | 0.383 | 19.6 | 17.6 | 19.0 | 0.913 | |
| BC | All | 112.8 | 78.4 | 45.1 | 0.000 | 113.3 | 79.4 | 40.5 | 0.000 | 98.0 | 70.1 | 55.9 | 0.006 | 118.3 | 76.4 | 44.6 | 0.000 |
| Males | 63.5 | 44.5 | 28.0 | 0.001 | 64.5 | 44.9 | 24.1 | 0.000 | 66.3 | 39.8 | 32.6 | 0.018 | 70.0 | 44.7 | 26.1 | 0.000 | |
| Females | 50.8 | 35.1 | 16.8 | 0.000 | 49.7 | 36.0 | 15.9 | 0.000 | 34.5 | 30.9 | 23.8 | 0.173 | 49.0 | 32.4 | 18.7 | 0.000 | |
| ≤50 years | — | 18.3 | 29.0 | 0.028 | — | 24.9 | 26.9 | 0.689 | — | 24.5 | 27.6 | 0.534 | — | 29.9 | 25.7 | 0.375 | |
| >50 years | 60.6 | 40.6 | 23.3 | 0.003 | 61.8 | 40.3 | 19.6 | 0.000 | 53.1 | 38.6 | 33.7 | 0.184 | 66.6 | 38.7 | 26.6 | 0.001 | |
| M ≤ 50 years | — | 13.4 | 18.3 | 0.138 | — | 16.5 | 16.5 | 1.000 | — | 16.1 | 17.3 | 0.725 | — | 18.9 | 16.3 | 0.422 | |
| M > 50 years | 31.2 | 23.0 | 13.2 | 0.065 | 32.8 | 22.5 | 10.3 | 0.017 | 35.5 | 21.1 | 16.6 | 0.066 | 37.0 | 21.4 | 15.3 | 0.029 | |
| F ≤ 50 years | — | 4.0 | 10.8 | 0.217 | — | 16.5 | 10.2 | 0.252 | — | 5.0 | 10.8 | 0.292 | — | 13.5 | 9.8 | 0.473 | |
| F > 50 years | 30.8 | 18.0 | 10.4 | 0.022 | 29.8 | 18.3 | 9.4 | 0.018 | 19.2 | 17.9 | 17.7 | 0.975 | 30.2 | 17.8 | 11.9 | 0.043 | |
Our results overall suggest that there is often a significant association between skeletal activity markers and skeletal age stages. This association is identified particularly in Ix/Iy and TA among the CSG properties, as well as in all EC. Nonetheless, once the effect of chronological age is controlled for, this association diminishes, suggesting that it is directly attributed to the chronological age of the individuals, which simultaneously affects the expression of skeletal activity markers and skeletal age stages. Similarly, differences in CSG scores are very rarely present among underaged, correctly aged and overaged individuals, while such differences are found rather systematically only in individuals older than 50 years for ECs, supporting the key role of chronological age in the expression of morphological changes both on the pelvic joints and the entheses. Taken together, the results of our analyses support that in our dataset, there does not appear to be a direct effect of activity on skeletal degeneration in the pubic symphysis and iliac auricular surface. These results agree with those of Campanacho et al. (2012), who investigated pubic symphyseal morphology in males in a recent skeletal collection from Portugal and did not find an association with physically demanding occupations. A similar study, although focused on the acetabulum, employing material from Christ Church Spitalfields, London, was conducted by Mays (2012) and also did not find any association between manual occupations and more rapid age changes.
A point worth highlighting is the possible impact of body mass. One could argue that heavier individuals would be expected to exhibit higher skeletal age marker scores at an earlier age than normal in the sense that the increased mechanical loading of the body mass on the pelvis should have resulted in a faster degeneration of the pelvic joints in bigger individuals. Nonetheless, no significant body mass effect was found in our data in the GLM models. This finding is also in agreement with earlier studies. In modern US collections, Wescott and Drew (2015) found that obese individuals had aged more rapidly in the auricular surface but not in the pubic symphysis, whereas Merritt (2015) found that individuals with low body size actually tended to age earlier according to the degeneration of the pubic symphysis, auricular surface and sternal rib end.
There are many factors that may explain the overall lack of significant association between activity and earlier onset of skeletal degeneration in the pelvis. First of all, the skeletal activity markers we employed have inherent limitations. As mentioned in Section 1, numerous studies have criticised the role of EC as skeletal activity markers, stressing that the expression of these changes is largely controlled by the chronological age of the individuals and their body mass (Benjamin, Toumi, Suzuki, Hayashi, & McGonagle, 2009; Cardoso & Henderson, 2010; Mariotti, Facchini, & Belcastro, 2004, 2007; Michopoulou et al., 2015; Milella et al., 2012; Niinimäki, 2011; Robb, 1998; Villotte, 2009; Weiss, 2003, 2004, 2007; Wilczak, 1998). In addition, even though CSG properties are considered accurate skeletal markers of activity, they express resistance to mechanical loads applied on the long bones along specific planes, and these may not correspond to the same mechanical loads applied on the pubic symphysis or the iliac auricular surface.
A final factor is the possible impact of parameters that could not be taken into account in our study, such as diet and disease. The quality of diet affects bone formation and loss, bone mass and mineral density, as well as bone remodelling rates (Sahni, Zoltick, McLean, & Hannan, 2010). In our material, just like in most archaeological and modern skeletal collections, it is not possible to know the dietary quality of the individuals under study. With regard to pathological conditions, we excluded from the dataset individuals with visible signs of conditions that may affect the rate of degeneration of the pelvic joints, such as ankylosing spondylitis, Reiter's syndrome, rheumatoid arthritis, tuberculosis and trauma (see Campanacho et al., 2012, for similar approach). Nonetheless, we cannot exclude the possible effect of other, less easily identifiable pathological conditions on the os coxae.
Despite the above possible limitations, the current study highlights that the effect of activity on the degeneration of the pelvic joints is not as pronounced as we would anticipate, which is in agreement to a number of earlier studies, even though our materials and methods differ from those of these earlier studies. Considering the difficulty in identifying past occupations skeletally (Jurmain, Alves Cardoso, Henderson, & Villotte, 2012), our results are important as they support that this often missing parameter in past osteobiographies does not significantly affect skeletal age. However, it should be stressed that these results are specific to the Athens Collection, and there could be differences in other populations making it worthwhile duplicating this work on different samples. Moreover, our results suggest that taking into account the EC scores when evaluating skeletal age may provide further insights in age-at-death estimation in older individuals in the sense that older individuals are often underaged using skeletal age markers. In such cases, an observation of pronounced EC may point to a higher than estimated age-at-death for the individual in question, and there have in fact been recent attempts to use EC as age-at-death estimators by means of regression analysis (Milella et al., 2020). Of course, this requires additional work on a larger and broader sample to test its validity and, most importantly, to calculate a statistically appropriate bias correction model.
ACKNOWLEDGEMENTS
The authors would like to thank two anonymous reviewers for their valuable comments, which greatly improved this manuscript. Efthymia Nikita's contribution to this research was supported by the European Union’s Horizon 2020 research and innovation programme under grant agreement No 811068 and the European Regional Development Fund and the Republic of Cyprus through the Research and Innovation Foundation (People in Motion project: EXCELLENCE/1216/0023).
CONFLICT OF INTEREST
The authors declare no conflict of interest.
