Proton MRS is the most widely used method to quantitatively assess marrow fat. MRS uses the fat: water ratio to determine the fat content (Figs. 4, 5). An obvious limitation is that a constant water content (%) is assumed. In other words, fat: water ratios may change due to a change in water content rather than fat content. MRS requires a minimal volume of approximately 1 cm² to acquire a sufficient signal to noise ratio. Other non-spectroscopic yet precise methods of quantifying fat fraction are available such as the two-point Dixon method which involves sequential suppression or fat and water, the three-point Iterative Decomposition of water and fat with Echo Asymmetry and Least-squares estimation (IDEAL) (Gerdes et al. 2007), or the analogous Gradient-Echo Sampling of the Free Induction Decay and Echo method (Wehrli et al. 2000). The accuracy of MRI spectroscopic and non-spectroscopic methods in detecting the relative amounts of water and fat has been tested against 11 different emulsions of increasing fat content. This study confirmed a high correlation (r ² > 0.92) between MR methods of fat quantification and the % fat volume fraction within test bottles (Bernard et al. 2008). Also, reproducibility of proton MRS in a clinical setting is high, ranging from 0.78 to 0.85, with the highest reproducibility being in those areas with the highest inherent fat content, i.e. the femoral head and lowest in the femoral neck (Griffith et al. 2009).

An inverse relationship between increasing marrow fat and trabecular bone loss in senile osteoporosis has been recognised histologically for 40 years (Dunnill et al. 1967). However, it is only recently, though MRS and other MR-based techniques that marrow fat content can be quantified non-invasively on a large scale (De Bisschop et al. 1993; Schellinger et al. 2000; Kugel et al. 2001; Jung et al. 2000; Wehrli et al. 2000; Shih et al. 2004; Chen and Shih 2006; Liney et al. 2007) and at different anatomical sites (Duda et al. 1995) There is a gradual physiological increase in percentage marrow fat content with advancing years (Kugel et al. 2001; Griffith et al. 2012). An easy approximation to remember is that vertebral body marrow fat content is 25 % at 25 years and 65 % at 65 years of age (Kugel et al. 2001; Griffith et al. 2012).

There is also a distinct sex difference does exist in marrow fat content (Kugel et al. 2001; Griffith et al. 2012). Young males have about 10 % more fat in their marrow than females of equivalent age up to about 50 years of age (Kugel et al. 2001). Males show a gradual steady increase in marrow fat content of 7 % per decade throughout life from young to old (Kugel et al. 2001; Griffith et al. 2012) (Table 1, Fig. 6). Females, in contrast, show a less steep increase in marrow fat of about 2–7 % up to 55 years and then a dramatic increase between the ages of 55 and 65 years (Kugel et al. 2001; Griffith et al. 2012) (Table 1, Fig. 6). By 60 years of age, healthy females tend to possess about 10 % more marrow fat in their vertebrae than males (Griffith et al. 2012) (Table 1, Fig. 6).

The sharp rise in marrow fat content with the menopause may be due to a reduced haematopoietic requirement with cessation of menstruation. This may not, however,be the only cause given that menstrual blood loss is generally quite low (median of about 43 ml per menstrual cycle) (Gao et al. 1987). The sharp increase in marrow fat content in early post-menopausal females may be a more direct effect of estrogen deficiency influencing fat deposition (or stem-cell differentiation) both inside and outside the skeleton. In this respect, the increase in marrow fat content does tally with changes in female extra-skeletal fat distribution recognised to occur at this time.

Androgen and estrogen levels both decline in later years, though estrogen levels fall more sharply in menopausal females leading to a higher circulating androgen: estrogen ratio. This, and other factors, promotes greater intra-abdominal or visceral fat, i.e. an ‘android’ pattern of fat deposition in post-menopausal females (Toth et al. 2000; Blouin et al. 2008). This is different to the gynoid-pattern of fat distribution seen in pre-menopausal women when fat accumulates in the gluteal and thigh areas (Toth et al. 2000; Blouin et al. 2008). Whilst there is no specific literature available on the relationship between estrogen and marrow fat content, it is known that visceral fat content (i.e. an android pattern of fat distribution) does correlate positively with marrow fat content (Bredella et al. 2011). It is feasible, therefore, that the increased marrow fat content seen in females in the post-menopausal era may be the bone-equivalent of android fat deposition. Android fat deposition is also associated with increased risk of cardiovascular disease and metabolic syndrome (Bredella et al. 2011).

Similar findings are found using multivoxel chemical shift registration MR imaging to measure variation in the water fraction of the lumbar vertebral bone marrow with age and sex (Ishijima et al. 1996). The water fraction for males was 75 % for young males, decreased to about 50 % for middle-aged males and remained almost constant for later years (Table 2, Fig. 7) (Ishijima et al. 1996). Conversely, in females, the water fraction for young females remained fairly constant at around 70 % but decreases quite rapidly around the time of menopause such that it is lower than in males during later years (Table 2, Fig. 7) (Ishijima et al. 1996). This tallies with the previously noted lifelong changes in % fat content since fatty marrow contains much less water (~5 %) than red marrow (~35 %) (Hwang and Panicek 2007).

Overall, there is at least a 40–50 % increase in fat cell content with increasing age. This increase in fat cell volume will occur at the expense of functioning marrow volume. Trabecular volume decreases by about one-third to one-half with increasing age, though the relative percentage of the marrow space occupied by trabecular bone is small. Since, the marrow cavity is a defined space and vascular sinusoids do not seem to expand with age, one can infer that an increase in marrow fat content is really a marker for a decrease in the amount of functioning marrow, i.e. a decrease in red marrow volume.

Over and above the physiological increase in marrow fat content with age, osteoporosis is associated with an even greater increase in marrow fat content. In the third lumbar vertebral body, for example, post-menopausal subjects with normal bone mineral density (BMD) have less marrow fat content than subjects with osteopenia. Similarly, subjects with osteopenia have less marrow fat content than this with osteoporosis (De Bisschop et al. 1993; Schellinger et al. 2000; Kugel et al. 2001; Jung et al. 2000; Wehrli et al. 2000; Shih et al. 2004; Chen and Shih 2006; Liney et al. 2007; Griffith et al. 2005, 2006; Shen et al. 2007; Tang et al. 2010; Liu et al. 2010) (Table 3). The proximal femur, which has a higher fat content than the vertebral body, also shows similar changes in increasing marrow fat content as the bone becomes more osteoporotic (Griffith et al. 2008) (Table 3). Even the femoral head, which has a very high intrinsic fat content, also shows an increase in marrow fat content with decreasing BMD though this increase is not as pronounced as in other areas.

It is possible that the aforementioned findings of increasing marrow fat content with decreasing BMD as measured by dual X-ray absorptiometry (DXA) may be spurious due to the effect of increasing marrow fat on BMD estimation by DXA. Increase in marrow fat content may cause an erroneous reduction in BMD measurements made by DXA (Sorenson 1990; Bolotin 1998; Bolotin et al. 2001; Bolotin 2007). This is because DXA evaluates BMD by measuring the transmission of X-rays at two different photon energies (Blake et al. 2009). The mathematical theory of DXA (basis set decomposition) holds that across a broad range of photon energies, the X-ray transmission factor through any physical object can be decomposed into the equivalent areal densities (g/cm²) of any two designated materials (Blake et al. 2009). For DXA scans, the two materials chosen are bone mineral (hydroxyapatite) and lean tissue. As a result, DXA measurements will only accurately reflect true BMD if the object being examined is composed entirely of hydroxyapatite and lean tissue. In practice, the human body is made up of not two but three main types of tissue, namely bone, lean tissue and fat. Neglecting the difference between lean and fat may lead to a spurious reduction in DXA–BMD measurement. When marrow fat content is known, DXA estimation of BMD needs to be corrected by 0.0014 g/cm² in women and 0.0016 g/cm² in men for every 1  % increase in marrow fat about zero (Blake et al. 2009). Applying this correction, the aforementioned results of increasing marrow fat with decreasing BMD still hold true.

Since changes in marrow fat composition can affect bone metabolism in vivo, and diets rich in polyunsaturated fats can affect BMD, it is conceivable that changes in marrow fat composition can affect bone metabolism (Yeung et al. 2005). To address, this question samples of marrow fat and subcutaneous fat from 126 subjects (98 females, 34 males, mean age 69.7 ± 10.5 years) undergoing orthopaedic surgery were analysed for fatty acid composition using gas chromatography and results correlated with BMD–DXA (Griffith et al. 2009; Yeung et al. 2008) (Fig. 8a, b). A total of 22 fatty acids were identified in marrow and subcutaneous fat. Significant differences existed between marrow and subcutaneous fat fatty acid composition as well as between marrow fat samples obtained from the relatively haematopoietic proximal femur and relatively fatty proximal tibia. Other than cis-7-hexadecenoic acid [C16:1 (n = 9)] and docosanoic acid [C22:0], no difference in marrow fatty acid composition was evident between subject groups of varying BMD (normal, low bone mass and osteoporosis). In particular, the overall polyunsaturated fatty acid content, the n − 6/n − 3 ratio and the percentage composition of those fatty acids most frequently implicated in bone remodelling, namely docosahexaenoic acid, arachidonic acid, γ-linolenic acid and eicosapentaenoic acid, were unchanged in subjects with normal BMD, low bone mass or osteoporosis (Griffith et al. 2009). Overall, it seems less likely that a change in marrow fat composition is directly affecting bone metabolism. The two associations found between fatty acid composition and BMD may be inconsequential given that they account for <1 % (for C16:1(n − 9)) and <0.1 % (for C22:0) of the total marrow fatty acid composition and they do not have any known effect on bone metabolism (Griffith et al. 2009).

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), also known as MR perfusion imaging, measures bone marrow perfusion as opposed to bone marrow blood flow (Griffith and Genant 2011). DCE-MRI is a robust technique that yields empirical indices of perfusion such as maximal signal intensity enhancement (E ^max) and enhancement slope (E ^slope) (Figs. 9, 10). E ^slope and E ^max are derived from the first-pass phase of signal intensity enhancement and have been shown to be strongly predictive of tissue vascularity, microvessel density and tissue necrosis. In simple terms, E ^slope can be thought of as gadolinium delivery to the bone marrow and is a feature of blood supply, vascular sinusoidal size and permeability. E ^max is dependent on these factors though also on the perfusion requirements (i.e. metabolic activity) of the bone marrow. Reproducibility of bone marrow DCE-MRI is moderate to high ranging from 0.59 to 0.98 with best reproducibility in those areas with the highest inherent bone marrow perfusion (Griffith et al. 2009).

Perfusion data acquired from dynamic contrast-enhanced MR imaging is also amenable to two-compartment pharmacokinetic modelling using models such as the Tufts or Brix model (Fig. 11). The Tufts model uses a combination of arterial input function (AIF), and rate constants K ^trans, K ^ex and K ^el. AIF is assessed by analyzing the first pass intensity profile of the feeding artery. K ^trans refers to the transport constant and is influenced primarily by blood flow. K ^ex refers to capillary exchange and is influenced by capillary space, permeability, interstitial pressure and extracellular space. K ^el refers to elimination or wash-out and is influenced by venous return. The Brix model does not rely on AIF or K ^trans but still considers K ^ex and K ^el. It assumes a linear relationship between MR signal enhancement and tissue contrast concentration or, in other words, it assumes that tissue contrast concentration directly correlates with perfusion. No specific pharmacokinetic model to reflect the unique characteristics of marrow perfusion has been developed. Measurement of bone marrow perfusion can also be undertaken by PET-CT imaging undertaken using ¹⁸F-fluoride which has a half-life of 112 min. Since this tracer is metabolised in bone, ¹⁸F-fluoride imaging is a combined measure of both bone perfusion and bone metabolism as compared to MR perfusion imaging which only measures bone perfusion. Pure bone perfusion can be evaluated by PET using the freely diffusible tracer ¹⁵OH₂0. However, these studies are difficult to perform as ¹⁵OH₂0 has a half-life of only 122 s and thus requires an on-site cyclotron. Nevertheless, a highly significant correlation between blood perfusion measured using ¹⁸F-fluoride and true bone perfusion using ¹⁵OH₂0 has been reported (Piert et al. 2002).

Bone marrow perfusion deceases with increasing age (Chen et al. 2001; Montazel et al. 2003; Baur et al. 1997). Subjects aged more than 50 years have a 62 % lower E ^max (21.88 ± 14.77) that those aged less than 50 years (58.21 ± 44.65, P < 0.005) (Chen et al. 2001). When this is further analysed according to sex, a greater discrepancy is observed. In women, E ^slope decreased by 80 % (from 87.17 ± 54.13 to 17.98 ± 13.80) in those older than age 50 years (P < 0.005). A similar trend is seen in men with E ^slope decreases by 33 % from 38.16 ± 21.69 to 25.38 ± 15.43 in subjects more than 50 years though this change did not reach statistical significance (P > 0.05) (Chen et al. 2001). Overall, vertebral bone marrow perfusion is higher in young females than young males (Chen et al. 2001). However, the rate of decrease of perfusion is less in males, which leads to vertebral bone marrow perfusion being higher in elderly males than elderly females (Chen et al. 2001). Similar findings were shown by Montazel JL et al. E ^max values being significantly higher in patients younger than 40 years than in those aged more than 40 years (P < 0.001). Perfusion parameters decreased with increasing age in a logarithmic relationship (r = 0.71) and correlated with increase in marrow fat content (Montazel et al. 2003). Savvopoulou et al. (2008) showed how the upper (L1, L2) lumbar vertebral bodies were better perfused than the lower (L3, L4, L5) vertebral bodies. In elderly subjects with normal BMD, E ^max was lower in females (32.3 ± 8.5 %) than males (34.5 ± 13 %) while E ^slope was higher in females (1.70 ± 5.2 %/s) than males (1.48 ± 0.7 %/s) (Griffith et al. 2005, 2006). To summarise, vertebral marrow perfusion is higher in young females than young males. However, perfusion decreases to a greater degree in females than males. Elderly females have reduced E ^max but not E ^slope compared to elderly males.