The effect of factors other than age upon skeletal age indicators in the adult.

CONTEXT
Estimation of adult age from skeletal remains is problematic due to the weak and variable relationship between age indicators and age.


OBJECTIVES
To assess the proportion of variation in age indicators that is associated with factors other than age and to attempt to identify what those factors might be.


METHODS
The paper focuses on frequently used adult bony age markers. A literature search (principally using Web of Science) is conducted to assess the proportion of variation in age indicators associated with factors other than age. The biology of these age markers is discussed, as are factors other than age that might affect their expression.


RESULTS
Typically, ∼60% of variation in bony age indicators is associated with factors other than age. Factors including inherent metabolic propensity to form bone in soft tissue, vitamin D status, hormonal and reproductive factors, energy balance, biomechanical variables and genetic factors may be responsible for this variation, but empirical studies are few.


CONCLUSION
Most variation in adult skeletal age markers is due to factors other than age; dry bone study of historic documented skeletal collections and high resolution CT scanning in modern cadavers or living individuals is needed to identify these factors.


Introduction
Estimation of age at death is fundamental in the study of archaeological and forensic human skeletal remains. In archaeological population studies it plays a crucial role in palaeodemography and also provides a framework for interpreting other skeletal data. In forensic osteology, age at death is a major element in the reconstruction of the biological profile of an individual. However, in adult skeletal remains, estimating age at death is fraught with difficulties and has been characterized as one of the thorniest methodological problems in archaeological (Mays, 2010a: 59) and forensic (Cunha et al., 2009) osteology.
Skeletal age indicators scored from gross inspection of skeletal remains are less costly and time-consuming to apply than those that require medical imaging or microscopy. This means that the former are more frequently used to estimate age at death. The most frequently applied techniques for estimating adult age are the morphology at the pubic symphysis, the auricular surface of the ilium and the sternal end of the rib; cranial suture closure; and, in archaeological populations, dental wear (Falys & Lewis, 2011;Garvin & Passalacqua, 2012).
Skeletal age indicators in the adult are only imperfectly correlated with chronological age. There is characteristically a wide range of variation in morphology of an age indicator among individuals of a given age, meaning that factors other than age contribute substantially to variation in age indicators (Jackes, 2000). Within a population, this leads to a great deal of ''noise'' in the relationship between indicator and age. It is also responsible for the well-established fact that relationships between age indicators and age differ in different populations (e.g. Hoppa, 2000;Rissech et al., 2012;Schmitt et al., 2002). For most age indicators, estimation of age involves the assumption that the relationship between age indicator and age is similar in the target population to that existing in the reference population used to develop the ageing standard. Therefore, the existence of substantial inter-population differences is a major stumbling block to estimating age at death in skeletal remains. The aim of this article is to assess the proportion of variation in age indicators that is typically associated with factors other than age and to attempt to identify some of the more important of those factors. age indicator and age is a measure of the proportion of variation in the age indicator associated with variation in age. A literature search, principally using the Web of Science, located 41 investigations of gross bony age indicators and 10 studies of dental wear that report r 2 values or present data from which they can be calculated. The studies are tabulated in Supplementary Table 1 and plotted in Figure 1.
There are difficulties in interpreting the data in Figure 1. In some studies, the coefficient of determination is based on Pearson's r, in others on Spearman's rank correlation coefficient. In addition, some studies report correlations between estimated age and true age, some between true age and indicator stage. Nevertheless, some cautious observations can be made. For the bony age indicators, r 2 values range from 0.03-0.90, with a median of 0.38 and 70% of studies report r 2 values of 50.5. This suggests that, in general, the majority of variation reflects factors other than age. This is consistent with earlier commentators who have variously generalized that, in adults, less than half (Nawrocki, 2010) or 30% (Jackes, 2000) of the variability in age indicators is associated with chronological age.
For dental wear, the median r 2 is 0.52, but eliminating values for low dental wear 19th and 20th century European populations gives a median of 0.90. As well as being more strongly correlated with age (at least in high wear groups), dental wear also differs from other ageing methods in other ways. In population studies of high dental wear groups, it does not require the application of ageing standards from a reference population because the rate of wear can be estimated in the group under study. This can be done by examining wear on erupted permanent molars of juvenile individuals whose ages can be accurately assessed from dental development, using methodology developed by Miles (1963). Adult age estimation using the Miles method has proven reliable when tested in a high dental wear population of known age (Kieser et al., 1983). A reasonable number (20 or more; Nowell, 1978) of juvenile dentitions with erupted permanent molars is needed for this calibration procedure. If insufficient juveniles are available, a measure of wear rates may instead be gained from comparing wear on the different molars in adult skeletons (e.g. Benfer & Edwards, 1991) and a recent study (Gilmore & Grote, 2012) suggests a way in which this principle might be operationalized into a calibrated dental wear ageing method for populations lacking sufficient juveniles to support the classic Miles calibration. In addition, the biology of dental wear and the extraneous factors that may affect its rate (mainly the coarseness and toughness of diet) are fairly well understood (Mays, 2010a: 71-76). Because of their more problematic nature, the emphasis in this paper is on bony age indicators. In the next section, the biological basis of the four most commonly used age indicators is discussed and, in this light, some factors that may potentially affect their expression are considered.

The biology of bony age indicators
The pubic symphysis The pubic symphysis is an amphiarthrodial joint made up of the hyaline cartilage covered medial surfaces of left and right pubic bones separated by a fibrocartilagenous disc which cushions and dissipates mechanical forces. It is capable of a small amount of movement under physiological Coefficients of determination reported in the literature between age and age indicator states or between age and age estimated using skeletal indicators. Based on Pearson's r from linear regression or Spearman's rank correlation coefficient. For studies reporting r 2 for the sexes separately, the mid-point between the two values is used in the plot. For data sources see Supplementary Table 1. conditions -2 mm vertical shift and 1 rotation (Becker et al., 2010). In the adolescent, the bony symphyseal face bears an approximately antero-posteriorly orientated ridge and furrow pattern. During the earlier part of adult life, bone is deposited upon this surface, thickening the cortical bone and eventually obliterating the ridge and furrow morphology. The resulting fairly smooth surface is characteristically surrounded by a slightly raised rim of bone. Subsequently, from about middle age onward, the surface of the symphysis, including the rim, undergoes irregular erosion, thinning the cortical bone, producing a pitted, porous, roughened surface (Berg, 2010;Brooks & Suchey, 1990;Lottering et al., 2014;Scheuer & Black, 2000: 370-372). The changes occurring from middle age onward are often referred to as degenerative (e.g. Scheuer & Black, 2000: 371), and dissection study (Putschar, 1976) appears to confirm that they are associated with degeneration of the hyaline cartilage that caps the bony surface. These observations have been systematized into separate ageing standards for males and females (Brooks & Suchey, 1990;Gilbert & McKern, 1973;McKern & Stewart, 1957).
Studies on laboratory animals (Crelin, 1969;Lovejoy et al., 1997;Pinheiro et al., 2004;Tague, 1988;Zhao et al., 2000), principally mice and guinea pigs, have demonstrated the sensitivity of the pubic bone to oestrogens. Oestrogen administration leads to bony resorption at the pubis. This also occurs during normal pregnancy (presumably reflecting the rise in oestrogen production) and multiparous females show particularly marked changes. The effect of oestrogens on the pubic bone stands in contrast to their effects on other parts of the skeleton, where they enhance bone formation. Bony resorption at the pubic symphysis has been confirmed during pregnancy in humans (Abramson et al., 1934;Putschar, 1976;Vix & Ryu, 1971).
Relaxin is a hormone secreted during pregnancy. It has a systemic effect on ligamentous tissue, leading to increased joint laxity (Marnach et al., 2003). The ligamentous laxity that accompanies pregnancy widens the joint space at the symphysis. In humans, this process peaks at full term (Bahlmann et al., 1993) when the joint space may be 3 mm greater than the 4-6 mm characteristic of female nullipares (Garagiola et al., 1989;Keriakos et al., 2011;Vix & Ryu, 1971). The joint space in most cases widens further during labour (Rustamova et al., 2009). These alterations facilitate the passage of the foetus through the birth canal. Dissection study (Putschar, 1976) indicates that some damage to the soft tissue of the symphysis is inevitable during parturition; tearing of the disc and the hyaline cartilage and avulsion of the bony endplate are observed. MRI in living subjects shows that oedematous alteration to bone and cartilage following parturition is generally symptomless and could be considered normal (Wurdinger et al., 2002). The trauma suffered by the symphysis during parturition has long been thought a potential cause of alterations to its bony surface and prompted the recognition that separate ageing standards would be needed for males and females (Stewart, 1957).
The pelvic ligaments tighten again after parturition, but there remains greater mobility in the symphysis in parous women and the degree of mobility is correlated with the number of pregnancies (Garras et al., 2008). Joint mobility generally accelerates degenerative changes (Gamble et al., 1986). The biochemical and biomechanical consequences of pregnancy and parturition suggest that the degenerative changes at the symphysis that feature in ageing schemes might be accelerated in women of high parity.
Trauma from other causes may lead to similar symphysial changes to those observed as sequelae of parturition trauma. Pubic symphysis stress injury (also known as osteitis pubis) is a condition found in individuals who regularly undertake strenuous activity and is today a common cause of pelvic pain in athletes. Although it may follow acute injury, it is more usually insidious in onset and is thought to be a response to long-term mechanical stresses placed upon the pelvis (Budak & Oliver, 2013). Soft tissue lesions include posterior herniation of the symphysial disc, periarticular oedema and tearing of the disc and periarticular entheses. Bone alterations include erosion of the subchondral bone, leaving an irregular symphysial surface, sclerosis, cyst formation and, less often, periarticular osteophytes (Besjakov et al., 2003;Budak & Oliver, 2013;Cunningham et al., 2007;Major & Helms, 1997;Mehin et al., 2006). The bony alterations resemble both sequelae from parturition trauma (Putschar, 1976;Vix & Ryu, 1971;Wurdinger et al., 2002) and gross features of later phases described in ageing schemata. Pubic symphysis stress injury is common in athletes -an incidence of 14-28% has been reported in soccer players (Hackney, 1993) and the frequency of bony alterations may be greater than this given that those with no or minor symptoms are unlikely to come to the attention of physicians. The association of the condition with athletes provides support for the idea that alterations of the type that culminate in pelvic stress injury may be more common in physically active groups. There is, thus, the potential for earlier onset of degenerative changes at the symphysis in archaeological than in most recent populations, where the proliferation of mechanical labour-saving devices promotes a more sedentary lifestyle.

The auricular surface
Unlike the pubic symphysis, the sacroiliac articulations are true synovial joints. However, as with the symphysis, joint movement under physiological conditions is normally very limited: 2-3 dorso-ventral sacral tilting and 0.7 mm translation (Stureson et al., 1989). This small movement helps dissipate mechanical forces. Synovial joints normally have opposing surfaces of hyaline cartilage, but, at the sacroiliac joint, the iliac surface is fibrocartilagenous (Bowen & Cassidy, 1981). It is much thinner than the hyaline cartilage on the sacral side (Brunner et al., 1991;Sashin, 1930). It undergoes more severe and earlier degenerative alterations (Resnick et al., 1975;Walker, 1986) and it was in part for this reason that the iliac rather than the sacral surface was selected as a potential age indicator .
In the young adult, the bony auricular surface is fine-grained in texture and bears transversely orientated grooves and/or striations. These transverse features are gradually obliterated, the surface becomes denser and more coarsely textured and pores appear in increasing size and number. Finally, the surface becomes more irregular in profile and osteophytes may form at its margins (Buckberry & Chamberlain, 2002;Igarashi et al., 2005;Lovejoy et al., 1985). Although detailed studies of cortical thickness changes are lacking, it may be that, analogous to the changes at the symphysis, the early alterations (densification, filling in of the transverse striations and grooves) are a result of net bone deposition. The later changes appear to be bony accompaniments to cartilage degeneration. The cartilage at the sacro-iliac joint undergoes the degenerative changes typical of osteoarthritis at a synovial jointroughening, fibrillation and erosion (Resnick et al., 1975;Shibata et al., 2002;Walker, 1986), which result in joint space narrowing and associated subchondral sclerosis and osteophyte formation (Bowen & Cassidy, 1981). Eburnation is rarely seen, perhaps because in older individuals fibrous ankylosis of the joint is common (Walker, 1986). Degenerative alterations are often visible on medical imaging modalities from the third decade of life, markedly earlier than at other synovial joints (Faflia et al., 1998;Shibata et al., 2002).
Unlike the pubis, bone at the sacroiliac joint does not appear to be resorbed in response to oestrogens (Tague, 1990), but, in common with other joints, the sacroiliac articulations show increased laxity during pregnancy (Brooke, 1924). Although the ligaments recover following parturition, the range of motion of the sacroiliac joints in females is characteristically greater than in males (Brooke, 1924;Bussey et al., 2009;Stureson et al., 1989). Degenerative changes may occur earlier and progress more rapidly in females, and may be related to childbearing (Shibata et al., 2002). Medical imaging studies indicate subchondral sclerosis at the sacroiliac joint is characteristically greater in females and is more common in multipares (Faflia et al., 1998), but marginal osteophyte formation appears more common in males, as does bony ankylosis in old age (Dar et al., 2005;Sashin, 1930;Stewart, 1984). When they occur in females, marginal osteophytes are more frequent in multipares (Faflia et al., 1998). Despite these observations, separate ageing schemes have not been developed for male and female auricular surfaces and generally tests of auricular surface ageing have not detected sex differences, although there are exceptions (e.g. Igarashi et al., 2005).
Evidence that exposure to physically demanding tasks forms a risk factor for osteoarthritis is substantial (Richmond et al., 2013); moderate mechanical forces are necessary to maintain healthy cartilage, but loading becomes catabolic if critical limits are exceeded. Physically arduous lifestyles also increase the risk of acute joint injury, another factor that predisposes to development of osteoarthritis (Richmond et al., 2013). Development of osteoarthritis is affected by other risk factors. It appears to be associated with obesity, via increased mechanical forces and biochemical alterations produced by adipose tissue deleterious to cartilage integrity (Sowers, 2001). Genetic factors are also important (Bateman, 2005;Spector & MacGregor, 2004). Although there seems to be a paucity of studies specifically of osteoarthritis at the sacroiliac joint, it would be surprising if these risk factors did not also apply there.
Osteitis condensans ilii is a stress injury of the sacro-iliac articulations. It is seen in athletes and occasionally in other groups (Major & Helms, 1997). It may co-occur with osteitis pubis, but appears less common: it is said to have an incidence of 52.5% in the general population (Mitra, 2010) but, as most cases appear symptomless, this may be an under-estimate. As with osteoarthritis, there is sclerosis of the subchondral bone (Olivieri et al., 1996;Mitra, 2010). The extent to which osteitis condensans ilii affects morphology of the auricular surface is unclear, but, in view of its potentially frequent occurrence and association with physically strenuous lifestyles, this may be worth investigating.

The sternal rib end
The costal cartilages are bars of hyaline cartilage attached to the sternal extremities of the ribs. The costochondral junction is not a joint per se, but rather the bone and cartilage merge together and the periosteum and perichondrium are continuous across the junction (Scheuer & Black, 2000: 235-236). In the adolescent and young adult, the flattish sternal rib end becomes more indented, forming a pit. With age, this pit deepens, reflecting increased bone deposition at its margins, and assumes a more U-shaped cross-section. In females, the increase in pit depth is less consistent than in males and by middle age there is normally bone deposition in the base of the pit which eventually forms a protruding tongue of bone. These changes were formulated into separate ageing schemes for males and females ( _ Işcan et al., 1984( _ Işcan et al., , 1985_ Işcan, 1991). They were originally devised for ageing the fourth rib, but subsequent work (Dudar, 1993;Yoder et al., 2001) shows no great differences between ribs 2-9.
To a great extent, the alterations in the sternal rib end ascribed to age by _ Işcan and co-workers represent a process of bone deposition into the costal cartilage at the rib end. In both sexes, ossification in the costal cartilages begins in the second decade, earlier than in other hyaline cartilages (Rejtarová et al., 2009). It has long been recognized (e.g. Fischer, 1955) that, other than at the first rib (McCormick & Stewart, 1988), the pattern of ossification of the costal cartilages differs between the sexes and this was why separate ageing standards were devised for men and women. At the costochondral junction, males tend to begin ossification at the superior and inferior margins, whereas females tend to start ossification by bone projecting centrally into the cartilage (Navani et al., 1970;Rejtarová et al., 2009). The sex difference may be mediated by hormonal factors. In pre-menopausal women, ossification of costal cartilages appears to be increased in individuals with amenorrhea or menstrual cycle irregularity (Horner, 1949) and there is a hint that women who have undergone oophorectomy or hysterectomy may tend to show the male pattern of costal cartilage ossification (Sanders, 1966(Sanders, , 1971. Vitamin D deficiency appears to inhibit costal cartilage ossification (Vieira Cesar et al., 1966); in contrast chest trauma may initiate or enhance it (Gleize et al., 2007;Malghem et al., 2001). Other factors, including genetic influences (Vastine et al., 1948), may also affect ossification patterns.

Cranial suture closure
Cranial sutures are synarthroses. The edges of the bones are in close apposition, separated by a thin layer of collagenous fibrous tissue. In adult life, the fibrous tissue gradually ossifies and, in time, this process may result in complete obliteration of the suture. The sutures of the cranial vault appear to begin ossification in early adult life, whereas those of the facial skeleton often remain patent into old age, so most ageing studies have concentrated on the former. In humans, vault suture closure generally occurs initially on the endocranial surface (Masih et al., 2014;. There is early fusion in the sagittal suture, subsequently extending to the coronal and lambdoid sutures (Cray et al., 2014). Studies which have attempted to craft methods of age estimation from the patency/degree of obliteration of the sutures have generally used recording methods based on gross examination of the endo-and/or ectocranial surface (e.g. Acsádi & Nemeskéri, 1970;Perizonius, 1984;Todd & Lyon, 1924, 1925. A large literature has accumulated on the biology of cranial suture closure. The great majority focuses on craniosynostosis, the premature fusion of cranial sutures early in ontogeny. Craniosynostosis is one of the most common developmental anomalies and, unlike normal suture closure, frequently has major clinical implications (dysmorphism and functional impairments) (Warren et al., 2001). Biomedical research prompted by a need to understand craniosynostosis has also shed some important light on the biology of normal suture closure. Work on laboratory rats suggests that the dura mater regulates fusion in overlying sutures. This occurs through a complex array of pathways, including via transforming growth factor-and bone morphogenetic proteins. Transforming growth factor-(TGF-) is a group of potent growth regulatory molecules that, inter alia, are involved in regulating the proliferation of osteoblasts and osteoclasts and play an important role in suture fusion (Opperman et al., 1997;Roth et al., 1997). Bone morphogenetic proteins (BMPs) promote bone formation, but their action is inhibited by noggin, a BMP antagonist. Noggin is present only in patent sutures and is regulated by fibroblast growth factor (FGF) signalling (Heller et al., 2007;Warren et al., 2003).
Work with mouse cranial bone suggests that sex steroid hormone signalling plays a role in sutural fusion (James et al., 2009;Lin et al., 2007). 5-dihydrotestosterone induces suture closure in vitro (Lin et al., 2007) and, in humans, male infants show a higher rate of sagittal and metopic premature synostosis, perhaps due to higher circulating levels of androgens. Although many studies of normal cranial suture closure in human adults fail to find sex differences, those that do generally report earlier fusion in males (e.g. Brooks, 1955;Hershkovitz et al., 1997;Key et al., 1994;Masset, 1989;Sahni et al., 2005). It is tempting to speculate that this might reflect the influence of androgens, but the extent to which hormonal factors affect the rate of adult suture closure and, hence, play a part in sex differences is unclear.
In the adult cranium, the most important source of biomechanical forces is cyclical loading due to mastication. Although it seems that increased masticatory forces result in increased sutural complexity (Byron et al., 2004;Byron, 2009), the evidence that they influence suture closure is less clear. In vitro study (Heller et al., 2007) of rat cranial bone showed that, compared to controls, mechanical oscillation led to an elimination of noggin expression, increased osteoblast differentiation and sutural fusion, but in vivo work involving stimulation of rat masticatory muscles failed to find an association between fusing sutures and mechanical forces (Shibazaki et al., 2007).
Artificial cranial deformation by binding the heads of infants and young children was practiced by some populations in the past, particularly in the Americas (Gerszten & Gerszten, 1995). There is evidence that this practice may alter the pattern of suture closure. For example, White (1996) reports premature closure of the sagittal suture in artificially antero-posteriorly deformed crania. Artificial cranial deformation, therefore, has the potential to affect the reliability of adult age estimation using cranial suture closure (O'Brien & Sensor, 2008).
Craniosynostosis in children may occur as a result of more than 150 genetic syndromes (Warren et al., 2001). Given the importance of genetic background in abnormal cranial suture fusion, it would be surprising if there were not important genetic effects on normal sutural fusion in adults. This has been little studied, but recent work (Wolff et al., 2013) did find an association between extent of suture obliteration in adults and genetic polymorphisms that may be associated with BMP signalling pathways.

Discussion
The biology of the age indicators, outlined above, leads one to offer the following observations: Age-related changes in the sternal rib ends and cranial sutures are predominantly bone-forming, as are the early alterations in the pubic symphysis and (probably) at the auricular surface. Sex hormones influence the pubic bone, the soft tissue of the symphysial and sacro-iliac joints, the pattern of ossification into the cartilage at the sternal rib ends and (possibly) the timing of cranial suture closure. The later changes at the auricular surface appear to be osteoarthritic in nature; the late changes at the pubic symphysis also appear to be degenerative. Biomechanical forces may, therefore, potentially influence the timing of the later phases of alteration at these two articulations. Biomechanical forces transmitted to the cranial bones during mastication may potentially influence cranial suture closure. Genetic factors exert an important influence on joint degeneration, cranial suture closure and ossification of costal cartilages. The following discussion considers the relevance of these observations for age indicators.
Some people, perhaps for genetic reasons, appear to show an enhanced tendency toward bone formation in periarticular soft tissue and elsewhere (Waldron, 2009: 72-73). For example, there appears to be a positive association between the formation of enthesophytes and periarticular osteophytes, indicating in affected individuals an enhanced propensity toward ossification (Hardcastle et al., 2014;Rogers et al., 1997). There is evidence that osteoarthritis (net gain of periarticular bone) and osteoporosis (net loss of bone) are inversely associated (Im & Kim, 2014), as are enthesophyte formation and osteoporosis (Hardcastle et al., 2014). There appears to be a continuum between 'bone-losers' and 'bone-formers', with diffuse idiopathic skeletal hyperostosis (DISH) and osteoporosis lying at opposite ends (Greenfield & Goldberg, 1997;Waldron, 2009: 73). The propensity to form bone shows population differences. For example, populations differ in their prevalence of osteoporosis (Kanis et al., 2002;Villa, 1994) and DISH (Julkunen et al., 1971;Weinfeld et al., 1997;Westerveld et al., 2008). Given that many of the agerelated changes in bony indicators result from a localized net deposition of bone, the relationship between bony age indicators and age might potentially differ in individuals or populations lying at different points on the bone-former/boneloser continuum (Schmitt et al., 2007). DISH and osteoporosis can be identified in skeletal remains and, hence, could be used as markers of a relative tendency toward bone formation. One might hypothesize that there may be a link between bone deposition at age indicator sites and a more general tendency toward bone formation. Some work has begun to investigate this possibility. Rissech et al. (2006) described an ageing method using the morphology of the acetabulum. Most of the traits identified involved periarticular bone proliferation. It was hypothesized (Mays, 2012) that such age changes should be more advanced in individuals showing bony signs of DISH, but study of an archaeological group of documented age failed to support this.
Vitamin D deficiency leads to inadequately mineralized bone. As discussed above, there is empirical evidence for reduced costal cartilage ossification in vitamin D deficiency, but one might advance the more general hypothesis that alterations resulting from net bone formation at age indicator sites might be inhibited in individuals with inadequate vitamin D (and potentially in other conditions that result in inadequately mineralized bone). To my knowledge, the effect of vitamin D status on age indicators has yet to be investigated empirically, but this could readily be done given that vitamin D deficiency can be identified in adult remains grossly and (especially) histologically (Mays, 2008). A potential link with vitamin D is important because 19th century European collections of known age at death, often used to generate and test ageing standards (e.g. Molleson & Cox, 1993), are predominantly from urban populations where, due mainly to attenuation of sunlight by particulate atmospheric pollution, vitamin D status was poor (Mays, 2003). In contrast, in target populations from more ancient times, vitamin D deficiency is unusual, as it is in most modern Western groups (Mays, 2003).
There may potentially be differences in hormonal status between ancient and more recent populations. This might affect both the relationships between age indicators and age and the existence and degree of any sex difference in age indicator expression. Perhaps the most obvious difference is the widespread use of artificial hormonal contraceptives in modern female populations, but there are also others. There has been a secular trend toward earlier puberty, connected with improved nutrition (Bogin, 1988;Lee & Styne, 2013), so that the timing of the rise in sex hormones in adolescence was probably later in the past. In both sexes, energetic status is a determinant of reproductive function. In males, long-term or extreme caloric deficiencies may lead to low sex hormone levels; in addition, patterns of age-related decline in male sex hormones differ in different world populations, perhaps reflecting differences in energetic status (Bribiescas, 2001). In women, reproductive function is more sensitive to energy balance than it is in males. High levels of physical activity and inadequate diet may cause suppression of oestrogen levels (Ellison et al., 1993;Ellison, 1994). I have posited this scenario in women from a Mediaeval peasant population (Mays, 2010b) and it may have wider applicability, especially to populations of low socioeconomic status and/or living in harsh environments. Although, in many ancient populations, poor energy balance may have been endemic, this is unlikely to be so for many modern or recent historic reference groups used to formulate ageing standards. Known age 19th century skeletal series from Europe, for example Christ Church, Spitalfields, London (Molleson & Cox, 1993), St Bride's Church, London (Bekvalac, 2012) and Zwolle, The Netherlands (Clevis and Constandse-Westermann, 1992), often come from prosperous urban populations. Among these groups, maternal breastfeeding was often brief or nonexistent, with infants being farmed out to wet-nurses, and effective birth-control methods were not yet available. Nutrition was better and physical labour typically less than in earlier or lower status groups. Infant mortality was high. These factors combined to favour large family size with repeated closely spaced pregnancies (Fildes, 1986;Hart, 1998;van Poppel, 1992). Given the elevation in oestrogen levels that occurs during pregnancy, the hormonal environment experienced by these women during their fertile years may have been very different from that customary in many archaeological populations.
Biomechanical forces on the cranial and post-cranial skeleton likely differ between recent reference populations and more ancient human groups. The advent of mechanized labour saving devices has led to a dramatic increase in sedentism in recent times. Some workers have attempted to assess the extent to which activity patterns, as measured by occupation, might be responsible for variation in age indicators. Campanacho et al. (2012) investigated pubic symphysial morphology in males in a recent skeletal collection from Portugal, but failed to find an association with physically demanding occupations. A similar study was done on acetabular age changes (some of which at least might be considered aspects of articular degeneration). It was hypothesized that those in manual occupations would show more rapid age-changes. In fact the opposite was the case (Mays, 2012). Using 20th century US skeletal material, Wescott and Drew (2015) found individuals who were obese at death had aged more rapidly at the auricular surface (although the effect was weak). There was no difference at the pubic symphysis. This was interpreted as due to the greater mechanical forces across the sacro-iliac joints in obese individuals, whereas the accentuation of mechanical forces across the pubic symphysis with increased body weight would be less marked as it is less weight-bearing. In contrast Merritt (2015) reports that, in another 20th century US skeletal collection, it was individuals with low body size who tended to attain a given age-phase at the pubic symphysis, auricular surface and sternal rib end earlier in life. She argued that differences in biomechanical forces and lean muscle mass may be implicated.
From the 17th century onward in Europe, there was a marked decline in masticatory forces, at least in urban groups, due to alterations in physical properties of foods consumed (Moore et al., 1968;Moore & Corbett, 1975). This would have led to lesser mechanical stress upon cranial bones. The likely greater parity among women in some 19th century reference populations than in more ancient times would mean that there were differences in prevalences of parturition trauma to the pelvic ring. These biomechanical factors might potentially affect age-changes at the cranial sutures and pelvic joints, but this possibility has yet to be investigated empirically.

Conclusions
Over the past 100 years, a great many studies have been published investigating the relationship between age indicators and age. This literature shows that the majority of variation seen in age indicators is unassociated with age. An exception may be dental wear. In high attrition populations, dental wear contains more information on age than do bony age indicators, so it is clearly the method of choice. However, in low dental wear populations, such as modern industrialized groups and post-17th century urban European archaeological populations, it performs less well.
For all bony ageing methods there is marked interpopulation variability in relationships between indicator and age. No ageing standard that has general applicability has yet been found, and there is no biological reason why it ever should be. Rather than standardize our methods, we need to select ageing techniques and standards that are the most appropriate for the particular material under study. This means choosing standards generated on reference populations which, as far as possible, resemble our target population in terms of the environmental and genetic factors that affect the relationship between an age indicator and age. In order to make more informed choices regarding what ageing standards to use for a given target population, we need to know much more about what those factors might be. This paper has discussed some that might potentially be important, but with a paucity of empirical studies, at present one is largely restricted to suggesting hypotheses rather than reviewing research results.
In this paper, I have made the case that the impact of inherent tendency toward bone formation ('bone formers' vs 'bone losers'), vitamin D status, energy balance, reproductive factors, biomechanical factors and genetic influences on bony age indicators may be worth investigating. To do this, knownage study samples are needed. In historic known-age skeletal collections, some factors, such as reproductive history or occupation, can be identified for individuals from documentary sources (e.g. Molleson & Cox, 1993). Other factors, such as 'bone-former' status, resistance of diet to mastication or vitamin D status, can be identified from markers on the skeletons themselves. However, studies on historic skeletal populations have their limitations. For example, data on number of births in females may be of uncertain reliability due to under-reporting of illegitimate births and stillbirths. Occupation is an imperfect measure of physical activity patterns and is generally only available for males. It is also very difficult to control for the effects of extraneous factors. For example, individuals who differ in occupation are likely also to differ in myriad other ways, e.g. in diet, lifestyle and nutrition. This may be a reason why, in the few studies that have been done, links that have been hypothesized between age indicators and occupation have not been supported (Campanacho et al., 2012;Mays, 2012).
An alternative to using historic skeletal collections is to use modern study samples -willed cadavers or living individuals. Background documentation on relevant factors for individuals is likely more plentiful and more reliable, facilitating better controlled studies. High resolution CT scanning offers sufficient resolution for application of cranial suture closure (Harth et al., 2010), sternal rib end (Dedouit et al., 2008) and pubic symphysial (Lottering et al., 2014;Telmon et al., 2005) methods, although visualizing the more subtle alterations at the auricular surface has, thus far, proven problematic (Barrier et al., 2009;Villa et al., 2013). High resolution CT would potentially enable study of age indicators in autopsy specimens and in living individuals. Such an approach might be of value as we seek to understand the factors that affect the relationships between age indicators and age.

Declaration of interest
The author reports no conflicts of interest. The author alone is responsible for the content and writing of this article.