Sex differences have been noted in the prevalence, incidence, and severity of osteoarthritis (OA) for many years [1–3]. The incidence of knee, hip, and hand OA is higher in women than men and in women increases dramatically around the time of menopause [3, 4], prompting many investigations into the roles of estrogen and other hormones as possible explanatory factors. Results of clinical and epidemiologic studies have been conflicting [5–7], with some showing a protective effect for estrogen or hormone replacement therapy (HRT) on radiographic knee and hip OA [8, 9] or progression to joint replacement [10] but no effect on joint symptoms [11]. Differences in study results can be ascribed to differences in (a) study populations and study designs; (b) distribution of, or confounding by, other risk factors such as age, race/ethnicity, body mass index (BMI), and smoking; and (c) joints affected by OA and definitions used to define OA [3] and statistical methodology. A recent systematic review of 17 studies illustrates this point. There was no clear association between sex hormones and hand, knee, or hip OA in women, but study heterogeneity precluded combining them into a single analysis [6]. For example, radiographic OA can be defined by overall radiographic scoring, such as the Kellgren-Lawrence system (grades 0 to 4) [12], or by individual radiographic features, such as osteophytes or joint space narrowing [13]. Other definitions may incorporate joint-specific symptoms with or without accompanying radiographic OA. With these multiple methods of examining sex differences in OA, it is no wonder that the issue is far from settled.

One way to examine the basis of sex differences in OA is to examine sex differences in the individual components of the joint through magnetic resonance imaging (MRI) and joint metabolism biomarkers. Structural features of OA can be examined using overall grading [14] or, more commonly, by individual features, such as cartilage thickness or volume, synovial hypertrophy and effusion, bone marrow lesions, or meniscal pathology. Additionally, rather than studying individuals who already have OA (that is, prevalent disease) as is commonly done, studying unaffected individuals followed prospectively may inform this question better since this approach avoids the question of temporality inherent in cross-sectional studies of prevalent disease, allows an assessment of the predictive value of specific findings for the development of OA, and potentially provides an opportunity for primary prevention and early intervention. This entails the examination of intrinsic characteristics of younger healthy men and women before they develop OA or even the examination of normal children. With increasing attention to risk factors for multiple diseases across the life course [15, 16], this approach could expose potential mechanisms behind sex differences in OA and identify high-risk people in advance of disease.

Reviews of the epidemiology of sex differences in OA have recently been published [5–7, 17]. This review will instead highlight novel study designs or studies of unaffected and younger individuals, including children, to focus on sex differences in (a) structural joint components using MRI and (b) biomarkers of joint metabolism.

Insight into sex differences in structural joint components using magnetic resonance imaging

Cartilage thickness, volume, and defects

Since men and women vary in body size, one might assume that men have greater cartilage volume. Cicuttini and colleagues [18] were among the first teams to test this hypothesis in knee cartilage volume using MRI. In 17 Australian men and 11 women who had normal knee radiographs and who were having knee MRI because of knee pain of less than 3 months in duration, men had larger femoral and patellar cartilage volumes than women, independently of age, height, weight, and bone volume. Sex differences in patellar cartilage volume were magnified with increasing age [18]. In a study of nine healthy German men and nine women in their early 20 s without a history of athletic or heavy physical activity, Faber and colleagues [19] confirmed lower cartilage volumes in women than men and showed that this sex difference was related primarily to differences in joint surface area or bone size rather than cartilage thickness, where differences were less pronounced and not statistically significant.

Otterness and Eckstein [20] hypothesized that smaller joint surfaces in women might explain sex differences in knee OA because of higher articular pressures with smaller surface area. Using healthy men and women, the authors confirmed that men have greater knee subchondral bone area, cartilage thickness, and cartilage volume compared with women, after adjustment for height and weight [20]. Estimated tibial or patellar pressures, using the metric of body weight/joint surface area, however, were equivalent in men and women, suggesting that smaller joint surfaces in women were not a likely explanation for sex differences in knee OA [20].

These authors found that total subchondral bone area and cartilage volume were strongly associated in young healthy men and women. However, while cartilage volume and bone area were strongly related to height in women, their associations with height in men were weak and inconsistent, leading the authors to suggest the possibility that different factors are responsible for bone and cartilage growth in men and women [21].

Longitudinally, Australian women have been shown to have a higher rate of cartilage loss than men [22, 23], whereas men may have a higher rate of cartilage loss than women in studies in the US [24, 25]. In 135 Australian men and 190 women from 26 to 61 years old (mean age of 45 years), Ding and colleagues [22] reported that, over an average of 2.3 years, women had a higher annual rate of cartilage volume loss than men in all knee compartments, although only tibial cartilage loss was statistically significantly different by sex. These sex differences first appeared at age 40 and increased with age [22]. Importantly, there were no significant sex differences in the crude annual percentage change or in the annual percentage change adjusted for age, BMI, and offspring/control status in cartilage volume in any plates; sex differences were evident only after further adjustment for baseline cartilage volume and bone size, and this could have inflated the difference. The composition of this convenience sample was intriguing; the sample consisted of offspring of people who had undergone knee arthroplasty for knee OA and the rest were from the general population. Interestingly, the magnitude of cartilage loss was higher in off spring than the general population, suggesting a high risk for the development of cartilage loss and presumably, later, for the development of knee OA [22]. Women were also three times more likely than men to have increases in tibial cartilage defects over time [22, 23].

A different result was obtained from the Osteoarthritis Initiative (OAI) [24, 25], an ongoing multi-center study in which a 3-Tesla MRI of the knee is obtained annually in approximately 4,800 individuals from 45 to 79 years old at baseline either with symptomatic radiographic knee OA (progression cohort) or with risk factors to develop knee OA (incidence cohort) [26]. An early study of the progression subcohort evaluated individuals (79 women and 77 men, mean age of 61 years) with frequent knee symptoms and radiographic knee OA in at least one knee. After 1 year, modest cartilage thickness loss occurred, more in the medial compartment than in the lateral, more in the medial femur than in the medial tibia, and more in the lateral tibia than in the lateral femur. There were no statistically significant differences in the rate of change of cartilage volume or thickness by age, sex, BMI, frequent symptoms, or radiographic Kellgren-Lawrence grade [24]. In the OAI, in contrast to the Australian studies, there was a non-statistically significant trend for men to have a greater rate of change in cartilage volume and thickness than women. The authors conceded that statistical power was limited, the period of observation was short, and only one knee (which may not have been the symptomatic knee) for imaging with Coronal FLASHwe (fast low angle shot with water excitation) was studied [24, 25]. When only some plates demonstrate differences, it is unclear whether this illuminates potential mechanisms, perhaps biomechanical, or represents a chance occurrence. None of these studies accounted for multiple comparisons inherent in the analysis of detailed MRI data, and it remains to be seen whether differences will be replicated in larger samples in which such multiple comparison testing is considered.

Hormonal associations with cartilage metrics by magnetic resonance imaging

Sex hormones and HRT after menopause have received considerable attention in the assessment of radiographic knee and hip OA, symptomatic OA, and joint replacement but with conflicting results [8–11, 27–29]. Using MRI outcomes, Wluka and colleagues [30] reported that healthy women who had no knee pain and who were taking HRT had greater knee cartilage volume than women not on HRT, suggesting a chondroprotective role for HRT. Such promising cross-sectional results were not borne out on longitudinal assessment [31]. Serum testosterone levels in such healthy women were not associated with cartilage thickness, cartilage defects, bone surface area, or large bone marrow lesions [32].

In healthy men without knee pain, on the other hand, cartilage volume was directly related to serum testosterone levels, but testosterone was not associated with change in cartilage volume [33]. The authors of these studies acknowledged that serum measures of androgens may not accurately reflect levels and activity at the site of interest, and so the implication of these findings is unclear [32, 33].

Leptin, a 16-kDa non-glycosylated hormone encoded by the obese gene and secreted by adipocytes, osteoblasts, and chondrocytes, has received some attention in OA and may be related to sex differences in OA [34]. Leptin is elevated in OA cartilage and osteophytes [35, 36], and higher concentrations have been found in synovial fluid in OA [37]. Women and those with higher BMI have higher leptin levels as well [38]. Ding and colleagues [39] evaluated cross-sectional associations between serum leptin levels and cartilage volume and cartilage defects in a subsample (n = 190, 48% were women, and mean age was 63 years) of the Tasmanian Older Adult Cohort, a population-based cohort of incidence and progression of OA and osteoporosis. In multi-variable analyses controlling for sex, age, BMI, smoking, radiographic knee OA, bone size, and other diseases (rheumatoid arthritis, cardiovascular disease, asthma, and diabetes), log-transformed leptin levels were significantly associated with knee cartilage volume but not with cartilage defects. Importantly, for the purposes of this review on sex differences in cartilage volume, leptin levels partially mediated the relationship between sex and cartilage volume, with a decrease in the R2 of the multi-variable model from 51% to 30% with additional adjustment for leptin levels [39]. Leptin did not mediate sex associations with cartilage defects. These studies suggest that leptin may mediate some sex differences in OA.

Imaging of cartilage in children

In keeping with the premise that sex differences in cartilage may be intrinsic or present many years in advance of OA onset, Jones and colleagues [40] performed a cross-sectional study of knee cartilage thickness and volume and bone surface area in 49 boys and 43 girls from 9 to 18 years old. One might expect boys to have larger cartilage volume than girls, and the authors controlled for multiple factors - such as age, BMI, bone area, number and type of sports participated in, vigorous physical activity, and lower limb muscle strength - that could influence these relationships. After adjustment, boys had greater cartilage thickness and volume than girls in all Tanner stages, with sex accounting for 20% of patellar volume, 26% of medial tibial volume, and 8% of lateral tibial cartilage volume. There was no difference in cartilage volume between pre- and post-menarchical girls.

Although many factors are related to the sex difference in cartilage parameters, these results imply that sex differences exist from early stages in the life course and that OA is likely determined or at least influenced by events in early life, even in the absence of joint injury. This principle is consistent with murine mesenchymal stem cell studies in which cells from male animals produced a 'richer extra-cellular matrix' [41] and larger culture pellet than cells from female animals. Furthermore, the regenerative potential of male cells was superior to that of female cells, with male cells providing better cartilage repair in nude mice than female cells did [41]. Koelling and Miosge [42] recently described sex differences in chondrogenic progenitor cells in cartilage from men and women undergoing knee joint replacement for OA. The authors observed that gene expression patterns differed by sex for ESR-1 and -2 genes, the transcription factor Sox9, and types I and II collagen [42]. There were also sex differences in the effect of sex hormones upon collagen II gene expression and in regulatory effects independently of Sox9 and Runx2. These studies showed that sex differences in cartilage occur from early development and persist through end-stage OA, suggesting that therapies might need to be tailored to men and women.

Meniscus and ligaments

Girls and women have different biomechanics, gait, and structural and morphometric properties of tendons and ligaments than boys and men [43–47]. It is likely that these intrinsic differences contribute to the increased frequency of ligamentous injury in young female athletes [43–47]. Whether these issues potentially contribute to the sex disparity in OA in later life is unclear, and surprisingly little attention has been devoted to sex differences in these structures using MRI in adults [48]. Fayad and colleagues [48] described differences in anterior cruciate ligament bundle volumes in 33 men and 30 women (mean age of 43, range of 15 to 70 years) referred for clinical MRI. Although there was a sex difference in anterior cruciate ligament volume, this was explained entirely by differences in height [48].

In a study that was mostly of asymptomatic individuals in Australia, women were over four times more likely than men to have meniscal tears by 1.5-Tesla MRI [49]. In contrast, 32% of the men and 19% of the women in the Framingham cohort (n = 991, 57% were women, and mean age was 62.3 years) had meniscal tears or destruction by 1.5-Tesla MRI, with the prevalence in both sexes increasing with age [50]. In a study of women who were older than 40 years of age, had knee OA, and had been screened into a clinical trial for OA, meniscal tears were present in 73%, and not surprisingly, these were associated with significant impairment in walking endurance and balance after OA duration, symptoms, disability, body composition, and relevant clinical characteristics were controlled for [51].

Finally, Stehling and colleagues [52] recently reported associations between various knee lesions on 3-Tesla MRI and physical activity in 236 individuals from 45 to 55 years old in the asymptomatic incidence subcohort of the OAI. Although assessment of sex differences in MRI features was not the purpose of the paper, ligamentous abnormalities were more likely in men than women (23% versus 12.5%) and meniscal lesions were more common and more likely to be severe in men (54% versus 42%). In contrast, cartilage abnormalities were slightly more common in women (76.5% versus 72%) but full-thickness cartilage defects were more common in men (24% versus 14.7%). The prevalence of some features was higher than in previous studies, especially since these were asymptomatic individuals, and this is likely related to increased sensitivity from the stronger magnetic field used in this study.

These few studies show widely disparate results, likely because of different study designs, statistical power, MRI protocols, and study populations varying by geographic location, age distribution, clinical characteristics, and source of participants. These factors make interpretation impossible and mandate further research to determine whether sex differences exist in these parameters and to understand the mechanisms behind such differences.

Bone marrow lesions and bone cysts

In OA, bone marrow lesions are common and are associated with knee OA progression and pain [53–55]. In the first description of these in healthy men and women free of knee pain, Davies-Tuck and colleagues [56] reported that sex was not associated with the presence, development, or persistence of bone marrow lesions on knee 1.5-Tesla MRI over 2 years in the Melbourne Collaborative Cohort Study. Studies evaluating the coexistence of these lesions with cartilage loss, meniscal abnormalities, and bone cysts have for the most part not focused on sex differences in these relationships [54]. Tanamas and colleagues [57] reported that bone cysts were more common in men than women in a study of the relationship between bone cysts and subsequent knee replacement 4 years later, but further investigation into the role of sex differences in these lesions was not conducted.

Sex differences in biomarkers of joint metabolism

Another way to understand the etiopathogenesis of sex differences in OA is to examine factors representative of joint metabolism. In order for synovial and systemically measured biomarkers to be used to identify high-risk individuals before OA occurs or before it becomes clinically manifest, normative data in various populations, including subgroups by sex, are required. For some markers, such as type I collagen N-telopeptide (NTX-I) and osteocalcin (which are markers of bone resorption and synthesis, respectively), much is known about sex differences and (within women) the effects of menopause and HRT, but for other markers, much less is known. Reports using markers for OA frequently control for sex but do not describe sex differences specifically [58, 59]. Critically important is knowing what other factors - such as the body burden of OA, BMI, hormonal status, or other medical conditions - might confound a sex difference in a specific marker, especially when considering a biomarker that is ubiquitous in connective tissue, such as hyaluronan (HA) [60]. This report will examine sex differences in several of the more frequently used markers in OA, targeting presumably different processes.

Type II collagen degradation

Mouritzen and colleagues [61] described a marker of type II collagen turnover, cartilage-derived urinary collagen type II C-telopeptide degradation products (CTX-II), in 615 healthy men and women from 20 to 87 years old. Levels were similar in men and women from 30 to 45 years old and then increased in both men and women, with the levels of women being slightly higher than those of men (Figure 1 from [61]). Levels were also higher in post-menopausal women compared with premenopausal women; and in post-menopausal women, those taking HRT had lower levels than those not taking HRT. Furthermore, those taking HRT for a longer time had lower levels than those taking HRT for a shorter duration [61].

Kojima and colleagues [62] described serum levels of C2C, a marker of intra-helical type II collagen cleavage, in 69 Japanese men (mean age of 43 years) and 71 Japanese women (mean age of 44 years; 34% of the women were post-menopausal) who did not have joint or spinal pain or major medical conditions and who were not taking medications affecting bone metabolism. In individuals younger than 50 years, C2C was higher in women than men, and the reverse was the case in those older than 50 years. C2C levels were unrelated to menopausal status. Since CTX-II and C2C are both markers of type II collagen cleavage, why would results be different for each marker? The authors propose that differences might be because the markers are the product of different areas in the type II collagen molecule that get degraded, and the markers' different locations in the joint [62].

Matrix protein degradation

Cartilage oligomeric matrix protein (COMP) is a 64-kDa pentameric matrix protein found in most joint tissues, including cartilage, bone, tendon, ligament, synovium, and vascular smooth muscle. It is elevated in OA [63–67], predicts incidence of radiographic hip OA [65, 66], and is higher with increasing body burden of OA-affected large joints [63, 64, 67]. Clark and colleagues [63] described the first and largest population-based assessment of serum COMP using competitive enzyme-linked immunosorbent assay (ELISA) with monoclonal antibody 17-C10 in Caucasians in the Johnston County Osteoarthritis Project (JoCo OA). COMP increased with age and was higher in OA than controls but did not significantly vary by sex. Later, in the same study population, Jordan and colleagues [64] used a sandwich ELISA with monoclonal antibodies 16-F12 and 17-C10 and reported that serum COMP levels among Caucasians but not African-Americans were higher in men than women (Figure 1, page 679 from [64]). Serum COMP levels were associated with a 30% increased risk of hip OA development in older Caucasian women in the Study of Osteoporotic Fractures [65]; those in the highest three quartiles of change in this marker had a fivefold increase in the risk of incident hip OA compared with those in the lowest quartile of change [66]. No direct comparison of these results to men was possible in this study. No further specific evaluations of sex differences in COMP in relation to OA have been conducted.

Synovial inflammation

HA is a ubiquitous glycosaminoglycan formed from alternating units of glucosamine and glucuronic acid. It is a constituent of synovium and cartilage and is indicative of synovial inflammation and has been shown to be an important marker of systemic burden of OA in women [60]. Elliott and colleagues [60] reported that men had higher serum HA levels than women did in the JoCo OA; importantly, this effect was independent of differences in age, race/ethnicity, OA burden in knees or hips, BMI, or comorbidities. This was confirmed in a recent study of mitochondrial DNA haplogroups and their effect upon serum levels of multiple biomarkers in Spanish patients with knee and hip OA [68].

Systemic and synovial cytokines

Pagura and colleagues [69] examined systemic and synovial measures of cytokines (interleukin [IL]-1-alpha/beta, tumor necrosis factor-alpha, and IL-6) and growth factors (insulin-like growth factor-1 [IGF-1], transforming growth factor-beta [TGF-β], and interleukin 1 receptor antagonist [IRAP]) in a small study of 9 Canadian men and 8 women awaiting knee replacement and compared them with 21 age- and sex-matched controls recruited from the local community. Men had higher levels of serum and synovial IGF-1, but there were no sex differences in any of the other markers. However, the very small sample size and undetectable levels of cytokines, except for IL-6, render the impact of this study questionable [69].

Growth factor

Lastly, in the largest study of serum TGF-β to date, Nelson and colleagues [70] reported that this marker was higher in women than men in the JoCo OA and that the associations between this marker and prevalent radiographic knee and hip OA, osteophytes, and joint space narrowing were similar in men and women. A single measure of this marker was unable to predict incidence or progression of radiographic knee or hip OA, osteophytes, or joint space narrowing in either men or women [71].

Summary and suggestions for future research

Sex differences in prevalence, incidence, and severity of radiographic and clinical OA have been described, but specific examination of sex differences in MRI biomarkers and in biomarkers of joint metabolism are few, and results vary considerably. Despite some variation across study designs and study populations, women appear to have a thinner and more reduced volume of cartilage in the knee than men, and this may occur from early childhood. The relationship between cartilage volume and bone area cannot be ignored in analyses of these issues. Whether women have a more accelerated rate of cartilage volume loss than men remains unsettled. Few data exist on sex differences in other tissues of the knee by MRI and in systemic biomarkers of joint metabolism, and those that do exist frequently vary in the assessment of potential mediators of sex differences.

Most studies of OA have been limited to Caucasians. Future studies should examine these relationships in other race/ethnic groups and, perhaps more importantly, delve into aspects that are likely to shed light upon mechanisms behind sex differences. In particular, studies of imaging and biochemical biomarkers, rather than merely control for sex, should specifically examine whether sex differences exist in that biomarker. Additionally, studies should establish whether risk factors act similarly or differently in men and women, with an eye to determining whether sex-specific therapies make sense. Studies of mesenchymal stem cells, outlined above [41], suggest that this may not be such a far-fetched idea.

Another question deserving attention is whether height, weight, and bone area can adequately serve as proxies for body size, a crucial issue in understanding sex differences in imaging biomarkers. Future studies might evaluate other potential proxies, such as height × weight and others [19], in relation to sex differences in joint structures by MRI. Finally, sex differences in the genetics of OA have been noted for multiple genes and joint sites [72–74], and studies of sex differences in these genetic effects should dovetail with those using advanced imaging and biochemical biomarkers for maximal mechanistic insight.

Note

This article is part of a review series on New developments in osteoarthritis, edited by Martin Lotz and Stefan Lohmander. Other articles in the series can be found online at http://arthritis-research.com/series/osteoarthritis