Proton MR spectroscopy of the femoral head—Evaluation of patients at risk for avascular necrosis

Authors


Abstract

Purpose

To detect alteration of the fatty component by measuring the in vivo lipid and water content of normal-looking femoral heads of patients with and without risk for avascular necrosis (AVN) by using proton MR spectroscopy (MRS).

Materials and Methods

Marrow composition was measured by proton MRS (TR/TE = 5000/20 msec) in a sample volume placed in the epiphysis of the intact femoral heads of patients with unilateral osteonecrosis of the hip (group 1, N = 61, then excluding the post-traumatic or steroid user, final N = 45) and age-matched controls (group 2, N = 49). Three response variables were derived from MRS: the lipid linewidth (LW), water LW, and lipid/water ratio.

Results

Of the three variables, the lipid and water LWs differed significantly between groups (P < 0.001 and P = 0.05, respectively; t-test). The lipid/water ratio had borderline significance (P = 0.06). The three variables differed significantly between groups when multivariate regression (P < 0.0001) was analyzed; and age and sex had no significant effect on the three dependent variables.

Conclusion

Proton MRS can depict alteration in the lipid and water composition of normal-looking femoral heads with and without AVN on the contralateral hip. Proton MRS may be a potential tool for investigating of the femoral head component in vivo and predicting the risk for development of AVN. J. Magn. Reson. Imaging 2006. © 2006 Wiley-Liss, Inc.

IMPAIRED BLOOD PERFUSION and/or increased intraosseous pressure have been postulated as causes of avascular necrosis (AVN) of the femoral head (1). Within the closed compartment of the femoral head, small, thin-walled and collapsible vessels provide the blood supply and drain the marrow spaces. Any elevation in intraosseous pressure will be transmitted to these small, collapsible vessels within the bone and thus cause a decrease in blood flow to bone and bone marrow. Rapid or uncompensated increases in intraosseous pressure result in irreversible ischemia and subsequent tissue damage, which in turn induces edema. As a consequence, pressures within the closed marrow compartment are further elevated (2).

Lipid and water are two important components of bone marrow (3). Therefore, any metabolic change in the fat component readily alters the internal environment. As suggested in the literature, corticosteroid use is an important etiologic factor that leads to AVN (4). Treatment with corticosteroids can elicit derangements in fat metabolism and thus increase fat deposition in peripheral tissues, such as the marrow spaces of bone (5, 6). Furthermore, lipocyte hypertrophy may result from corticosteroid treatment. Within the closed space of the femoral head, lipocyte hypertrophy impedes perfusion of the marrow (7, 8). Thus, alterations in intramedullary fat cell metabolism can change the perfusion of bone marrow in the femoral head and ultimately result in AVN (9).

MRI has been demonstrated to disclose marrow changes resulting from AVN of the femoral head (10–12). MR signal patterns for AVN have been described as varying widely. For example, Bassett et al (10) performed MRI on an ischemic necrotic femoral head, and disclosed a region of normal, high signal intensity representing “mummified” fat. Mitchell et al (11, 12) demonstrated a central region within the necrotic femoral head that was isointense for marrow fat on both T1- and T2-weighted images.

The above data suggest that MR signal patterns deriving from the necrotic and adjacent area of the femoral head largely reflect the signal changes corresponding to the fat cell component, since fat cells constitute the dominant portion of the femoral head. On the other hand, the signal pattern from viable or dead fat cells contained within the bone strongly reflect the MR signals deriving from AVN of the femoral head. Thus, further investigation of the fat and lipid content of the femoral head using different techniques, such as MR spectroscopy (MRS), is considered a potentially valuable adjunct to MRI in AVN research.

Proton MRS techniques have been developed and used to measure the musculoskeletal and fat tissue in vivo (13–18). In 1995, Bluemke et al (13) measured fat content (as a percentage) in the femoral head by using MRS on patients with systemic lupus erythematosus, long-term corticosteroid treatment patients, and healthy subjects. Ballon et al (14) used a volume-localized MRS technique to estimate bone marrow cellularity and found that the results correlated with the results from marrow-core biopsies. Shick et al (15, 18) measured MRS from the vertebral bodies and compared the water signal distributions from healthy people and patients after bone marrow transplantation, and found a relative difference between the two. In 1997 Amano and Kumazaki (16) performed MRS using a stimulated echo acquisition mode (STEAM) technique to demonstrate a significant difference in the marrow water to fat ratio between patients with aplastic anemia and healthy subjects. Schellinger et al (19) used a similar technique to measure the lipid/water ratio and LW in 57 normal subjects, and their results appeared to suggest an age-dependent linear increase in lipid content. In our previous research, we also demonstrated a significant correlation between the lipid/water content and bone mineral density on the lumbar spine (20). The above-mentioned MRS techniques have mainly been applied to the red, hematopoietic marrow of the vertebral body (14–20); only Bluemke et al (13) applied it to the yellow, fatty marrow of the femoral head. The alteration of the fat and water content of the yellow marrow may result in a change of the internal environment in the femoral head, which is important in the subclinical or early stages of AVN.

In this study we performed proton MRS to evaluate the healthy hips of patients with nontraumatic and non-steroid-induced AVN in the contralateral hip, as well as in the hips of age- and sex-matched control subjects without AVN of either femoral head. We hypothesized that proton MRS would be able to detect alterations in the lipid and water content of the femoral head in patients at risk for AVN before morphologic evidence was apparent.

MATERIALS AND METHODS

Patients and Subjects

This study was approved by the Institutional Review Board of the National Taiwan University Hospital, Taipei. All subjects provided written informed consent to undergo MRI and proton MRS of both hips.

We enrolled 110 patients and control subjects between August 1999 and March 2002. Sixty-one patients were known to have AVN of one femoral head, as proved by characteristic MRI findings and their clinical presentation. Twelve had experienced a traumatic episode involving the diseased hip, according to their clinical history and radiologic manifestations; therefore, these 12 patients with traumatic AVN were excluded. Another four patients with AVN of a unilateral femoral head had a history of corticosteroid use and were also excluded. In total, 45 patients with nontraumatic, non-steroid-induced AVN of a femoral head were included in group 2. This group included 36 men and nine women with an age range of 21–68 years (mean age ± standard deviation (SD), 44 ± 13 years). The average interval between the diagnosis of AVN and the patients' inclusion in our study was 3.6 years (range = 0.8–5 years).

Forty-nine subjects with no evidence of AVN of the femoral heads were included as an age- and sex-matched control group (group 1). This group included 30 men and 19 women with an age range of 22–76 years (mean age ± SD, 49 ± 15 years).

Both groups underwent plain radiography and MRI during follow-up for at least 18 months in the outpatient clinic.

MRI and MRS

The examination of the hips included MRI and MRS measurements. All studies were performed with a 1.5-T whole-body imaging system (Magnetom Vision Plus or Magnetom Sonata; Siemens, Erlangen, Germany). To assess the condition of the femoral head, the imaging protocol included a coronal T1-weighted spin-echo sequence (TR/TE = 500/12 msec, matrix size = 512 × 384, FOV = 320–360 mm) with a 4-mm section thickness and 0.8-mm section gap. Short-tau inversion recovery (STIR) images were acquired in the coronal orientation (TR/TE = 4300/30 msec, inversion time = 170 msec, turbo factor = 7, matrix = 192 × 256, FOV = 320 × 360 mm) with the same section thickness and gap. Fast spin-echo T2-weighted images (TR/TE = 6000/90 msec, turbo factor = 11, matrix = 384 × 256, FOV = 300 × 340 mm) were acquired in the axial plane using the same dimension and spatial relationship as before. Proton density images with fat suppression (matrix = 384 × 512, FOV = 200 mm) were acquired in an oblique coronal view and performed separately for each hip. A set of postcontrast coronal T1-weighted images (matrix = 512 × 384 mm, FOV = 320 × 360 mm) with fat suppression were acquired. For the above protocol, the signal intensity and image resolution were considered comprehensive for the detection of early AVN lesions.

The MRI images were used to evaluate the patients and subjects for any disease or abnormality of the femoral head, such as osteonecrosis with or without deformity, infection, or neoplasm. In femoral heads with normal signal intensity and contours, proton MRS was then performed (i.e., MRS was conducted on the asymptomatic, normal-looking side in patients with unilateral AVN (group 2), and on both sides in the control subjects (group 1)).

The proton MRS protocol acquired data from a sample volume from the femoral head (Fig. 1b) using single voxel spectroscopy (SVS) with a stimulated echo-acquisition mode (STEAM) sequence for three-dimensional localization of the sample volume (the proton MRS STEAM sequence has been described elsewhere (15, 19)). Automatic global shimming was performed prior to acquisition by an automated algorithm. Unlike cerebral proton MRS, MRS of the femoral head does not include water suppression; the water signal is fully exploited by manually setting the water suppression pulse voltage to zero before the data are acquired. Parameters for the SVS measurements included a TR of 5000 msec, TE of 20 msec, a cubic sample volume with dimensions of 1.25 × 1.25 × 1.25 cm3, acquisition of 32 signals each 1024 data points, and a spectral bandwidth of 1000 Hz. Each sample volume measurement was placed within the femoral head (Fig. 1), and the same investigator (Y.D.L.) performed the positioning in all patients and subjects. The 5000-msec TR was chosen to allow us to observe fully T1-relaxed signals. The TE was kept at 20 msec to minimize signal reduction due to T2-relaxation effects.

Figure 1.

a: T1-weighted image (coronal view of both hips in a 49-year-old male patient) reveals a crescent-shaped low signal lesion on the right femoral head, consistent with AVN. The left femoral head is intact. b: Proton MRS was performed on the left femoral head. A cubic voxel (1.25 cm)3 was positioned at the femoral head with three-dimensional localization. A STEAM sequence with TR = 5000 msec, TE = 20 msec was used.

Proton MRS spectra have a dominant water peak and a series of lipid peaks (13, 15). The lipid signal is composed of at least eight fractions (15), with the methylene group at 1.6 ppm contributing the largest signals. Two major spectra were identified (Fig. 2): a larger, steeper spectrum on the right side, which corresponded to lipid spectrum with the main peak at 1.6–1.8 ppm (the methylene group), and a smaller, wider spectrum on the left side, which corresponded to the water spectrum. Lipid and water spectra were separated by 3.1 ppm (220 Hz).

Figure 2.

Proton MRS of the femoral head revealed two major spectra: a larger, steeper spectrum on the right side corresponding to lipid spectrum at 0.8–1.8 ppm, with the main peak as the methylene group, and a smaller, wider spectrum on the left side corresponding to the water spectrum. Lipid and water spectra were separated by 3.1 ppm (220 Hz).

Quantification of the lipid and water signals was performed by means of curve-fitting software provided by the scanner manufacturer (Magnetom Vision; Siemens, Erlangen, Germany). As an alternative to expressing the absolute signal peak for quantification, we derived a relative lipid/water ratio for each sample volume by integrating the area under the lipid and water peaks. An estimate of the relative fat content or fraction was thus obtained. The LWs of the lipid and water peaks were also evaluated and determined from the same curve-fitting software, expressed as the full width of the curve at its half height (FWHM). Therefore, three response variables were used to represent the proton MRS findings in the femoral head: the lipid LW, the water LW, and the lipid/water ratio (Table 1).

Table 1. Descriptive Statistics of the Variables of Proton MR Spectroscopy Among the 94 Patients and Its Subcategorization into Group 1 (N = 49) and Group 2 (N = 45)
VariableAll N = 94 Mean ± SDGroup 1aN = 49 Mean ± SDGroup 2bN = 45 Mean ± SDt-test P value*
  • *

    P value: Comparison between Groups 1 and 2.

  • a

    Group 1: Control, subjects with no evidence of avascular necrosis of their femoral heads.

  • b

    Group 2: Subjects with avascular necrosis of their unilateral femoral heads.

  • LW = line width (Hz) by measuring the signal's full width at its half height.

Lipid LW28.53 ± 3.5529.69 ± 3.9926.50 ± 2.48<0.001
Water LW45.41 ± 12.5047.45 ± 9.5643.20 ± 14.870.05
Lipid water ratio14.86 ± 10.6615.05 ± 13.1314.66 ± 6.820.06

Statistical Analysis

The age distributions of group 2 (patients) and group 1 (control subjects) were analyzed by using univariate tests of significance. Proton MRS variables were analyzed with univariate and multivariate tests of significance. Univariate tests were further performed for MRS variables after we adjusted for age as a covariate. Age was further analyzed with the three response variables by means of multivariate regression. To clarify the relationship between the lipid/water signals and the age factor, we performed a Pearson correlation coefficient analysis and scatter plots with regression analysis (STATA version 8.0; Stata Corp., College Station, TX, USA) for the subjects (N = 94), group 1 (N = 49), and group 2 (N = 45). A P-value of 0.05 indicated a statistically significant difference.

RESULTS

Groups 1 and 2 did not differ significantly in their age distribution (mean age ± SD, 49 ± 15 years for group 1, and 44 ± 13 years for group 2; P = 0.07). For control subjects (group 1), plain radiographs and MRI showed a normal appearance of both hips during follow-up over a period of at least 18 months in the outpatient clinic. Their follow-up images showed no evidence of abnormality or need for surgical intervention. In the patients in group 2, follow-up imaging showed no new-onset AVN of the intact femoral head. We also reviewed the plain radiographs of normal hips in the diseased group. These femoral heads showed normal contours, and the principle and tensile trabeculations were intact.

The three proton MRS variables differed between groups 1 and 2. When we analyzed the three variables individually by applying univariate tests of significance (Table 1), the lipid LW significantly differed between groups (P < 0.001, t-test). The water LW also showed a significant difference between groups (P = 0.05). However, the lipid/water ratio had only borderline significance (P = 0.06).

When the data were analyzed with multivariate tests of significance, the proton MRS variables differed significantly between groups 1 and 2 (P < 0.0001; Fig. 3). Figure 4 shows the box plots of these variables (lipid LW, water LW, and lipid/water ratio) for groups 1 and 2.

Figure 3.

The three variables of proton MRS of the femoral head (lipid LW, water LW, and lipid/water ratio) revealed a significant difference between the control (N = 49, group 1) and diseased (N = 45, group 2) groups analyzed with multivariate tests. Group 1 showed higher values for the three variables than group 2.

Figure 4.

Box plots of the three variables (lipid LW, water LW, and lipid/water ratio) for groups 1 and 2.

When we analyzed the three variables by age and sex in our multivariate regression analysis, we noted no significant influence of age or sex on the variables (Table 2).

Table 2. Multivariate Regression Analysis
Independent variablesDependent variables
Lipid line widthWater line widthLipid water ratio
  • *The age or sex had no significant effect on the three dependent variables. Numbers in each cell are multivariate regression coefficients and numbers in parentheses are the t values. Group 1: Control; subjects with no evidence of avascular necrosis of their femoral heads. Group 2: Diseased, subjects with avascular necrosis of their unilateral femoral heads.

  • a

    Significant at the 95% confidence level.

All subjects (N = 94)   
 Intercept35.39753a (17.06)57.47052a (7.64)15.90391a (2.4)
 Group−2.812667a (−3.9)−5.72004a (−2.19)−0.2072868 (−0.09)
 Sex−0.9662135 (−1.24)−2.060068 (−0.73)0.0415724 (0.02)
 Age−0.0306235 (−1.23)−0.0081528 (−0.09)−0.0199903 (−0.25)
Control group (N = 49)   
 Intercept32.16453a (13.28)50.20556a (9.71)16.83151a (2.14)
 Sex−1.112592 (−0.92)−4.416118 (−1.70)2.932467 (0.74)
 Age−0.0179734 (−0.46)0.0895336 (1.07)−0.1244169 (−0.98)
Diseased group (N = 45)   
 Intercept30.46243a (17.20)48.29455a (4.40)14.00026a (2.90)
 Sex−0.8760368 (−0.96)0.6544315 (0.12)−3.473199 (−1.40)
 Age−0.0487884 (−1.70)−0.1337684 (−0.75)0.1098409 (1.40)

To clarify the relationship between the lipid/water signals and the age factor, we performed two tests to confirm our data: a Pearson correlation coefficient analysis (Table 3), and a scatter plot and regression analysis model (as Fig. 5). For the three different groups (pooled, N = 94; diseased, N = 45; control, N = 49), the results show that the age factor and three variables (lipid LW, water LW, and lipid/water integration ratio) had no significant correlation (all P-values > 0.1).

Table 3. Univariate Analysis of Age and Lipid Line Width, Water Line Width, and Lipid-Water Integration Ratio*
VariablePearson correlation coefficient with age (P-value)
  • *

    For all subjects and subcategorization of control or disease group, the three variables (lipid line-width, water line-width, and lipid-water integration ratio) have no correlation with their age (P > 0.1).

  • Numbers in parentheses are P-values.

All Subjects (N = 94) 
 Lipid line-width−0.062 (0.555)
 Water line-width0.026 (0.802)
 Lipid-water integration ratio−0.025 (0.810)
Control group (N = 49) 
 Lipid line-width−0.089 (0.543)
 Water line-width0.116 (0.427)
 Lipid-water integration ratio−0.127 (0.384)
Diseased group (N = 45) 
 Lipid line-width−0.242 (0.110)
 Water line-width0.117 (0.443)
 Lipid-water integration ratio−0.224 (0.139)
Figure 5.

Scatter plots and regression analysis model between variables (lipid LW, water LW, and lipid/water ratio) and age. The Pearson correlation coefficients had no significance (all P-values > 0.1).

DISCUSSION

MRI is the most sensitive and well-established method for diagnosing AVN of the femoral head (10–12, 21–23). In AVN of the femoral head, signal intensity changes and MRI patterns principally reflect the signal intensity changes in fat due to necrotic bone marrow. Dynamic or static contrast-enhanced MRI can depict perfusion changes in the femoral head as early as conventional MRI, or possibly earlier, as observed in some studies (24, 25). According to the known mechanism of AVN (8, 9), the fat-cell changes as a mechanism may be the most important change in the femoral head at risk for AVN.

Proton MRS can be clinically applied to evaluate the fat content of bone marrow fat (13–16, 19–20). Several investigators (16, 18, 26) have used the MRS technique to evaluate the lipid and water spectra of bone marrow in the lumbar vertebra of patients with leukemia, aplastic anemia, or Gaucher's disease. Only Bluemke et al (13) used MRS to measure the fat content of the yellow marrow of the femoral head. Years later, Schellinger et al (19) used the method with the parameters TR/TE = 5000/20 msec to evaluate the status of the lumbar vertebrae in normal subjects, and confirmed that the lipid/water ratio progressed with age in a linear fashion for both genders. He also demonstrated that the fat fraction was significantly higher in subjects with weakened bone, and suggested it could serve as a measure of bone quality (27). In this study we demonstrated the spectra of the femoral head as a water peak on the left and a lipid peak on the right (Fig. 2), separated by 3.1 ppm. In the lipid signals, the main peak represented the methylene group, and another small peak at its left aspect that merged with the main peak represented the ester group. The water spectrum revealed a major peak, and another small peak on the left represented vinyl groups of saturated fatty acids at 5.6 ppm. By using a long TR and a short TE, we successfully demonstrated the lipid and water spectra of the yellow, fatty marrow of the femoral head.

AVN of the femoral head is usually bilateral. Patients with nontraumatic AVN in one hip have a 50–72% likelihood of developing AVN in the other hip (28–30). In our patients with a contralateral femoral head at risk for AVN (group 2), lipid and water content measured during in vivo proton MRS significantly differed from that of the control subjects (group 1). This finding suggests that the alteration of the lipid and water in the femoral head occurs before any morphologic changes develop. The lipid and water spectra had significantly different (narrower) LW in subjects at risk for AVN than in control subjects, and the lipid/water ratio showed borderline difference between the patient and control groups.

Regarding the reduced (narrower) LW in spectra of the femoral head in the disease group, several issues should be discussed. Many studies have been published concerning the trabecular density and the MR spectral LW, and most of these studies focused on the bone marrow and trabeculation of the spinal vertebrae (20, 31–35). These studies discussed MRS evaluation of hematopoietic bone marrow (the vertebral body) with different bone mineral densities (BMDs). The lipid LW had a significant correlation with BMD with controlling age factor, but the lipid LW had no significant correlation with age while controlling BMD (20). Higher bone density is expected to cause greater magnetic inhomogeneity and wider spectra. Schick (33) and Wehrli et al (34, 35) mentioned that MR bone densitometry and interferometry reflect the two principal chemically shifted components of bone marrow (fat and water), which are influenced as arising from a difference in the intertrabecular space. However, they did not mention the yellow fatty marrow of the epiphysis of the femoral head. Only Bluemke et al (13) performed MRS on the femoral head, and found that the fat content percentage was similar between patients with systemic lupus erythematosus (SLE) and long-term steroid treatment, and non-SLE patients.

Is there a possible relationship between reduced trabecular density (osteoporosis) and AVN? We examined the possible combination of these two disease entities and found only steroid-induced osteoporosis (36) and steroid-induced osteonecrosis (37). However, the combination appears to be rare and only case report was found (38).

In our study subjects (see Materials and Methods section), we excluded patients who used corticosteroids. We did that for two reasons: 1) the steroid user may have a combination of these two situations, and 2) the steroid user may have a major medical disease (such as SLE, renal transplantation, etc.) that would influence the bone marrow contents. In addition, the age distribution of our disease and control group had no significant difference (mean age: 44 ± 13 years and 49 ± 15 years). Since the BMD is age-dependent, the BMD in our study subjects could presumably no differ between these two groups; and our research target is the normal-looking femoral head in the patients with contralateral side AVN (diseased group) vs. the normal femoral heads in control group.

More importantly, the routine Dual Energy X-ray Absortiometry (DEXA) measurement of BMD of the hip shows several data, including the neck region, trochanter, intertrochanter, Ward's triangle, and total area. The femoral head, especially the epiphysis, cannot be measured from the anteroposterior view of the hip BMD study alone because of its overlapping with the acetabulum. Therefore, no data for the region of femoral head can be obtained separately. We do not believe the BMD data of the neck region can be used to represent the BMD of the femoral head. Another possible measurement is quantitative CT of the bone trabeculation; however, our institute does not have the phantom for this calibration. This is one of the limitations of our research, because it was difficult to analyze the correlation between the femoral head MRS and femoral head BMD.

Furthermore, trabeculation is not the only factor that can influence the MR spectra. For example, the MR spectra could change in the bone marrow of leukemia, aplastic anemia, or glycogen storage disease, and the marrow in the active stage or complete remission will also have different MR spectra (16, 18, 26, 39–41). The cellularity of bone marrow can be calculated from the most dominant peaks in the proton spectrum of marrow from methylene and water peaks (14). Proton MRS showed a good correlation between the cellularity and water signal intensity compared to bone marrow biopsy (14, 40, 41). These studies suggest that MRS can be used to evaluate the bone marrow cellularity or biochemical composition in vivo in different disease entities.

Our research focused mainly on MRS of normal-looking femoral heads. The adult femoral head (epiphysis) undergoes marrow conversion to fatty marrow before the age of 20 and remains stable. The mean ages of our patients in the two groups were 44 ± 13 years and 49 ± 15 years. Thus, their femoral heads are considered to be stable in the fatty component. This is also confirmed by our data from the Pearson correlation coefficient analysis and scatter plots with regression analysis. For the different groups (pooled, N = 94; diseased, N = 45; control, N = 49), the results show that age factor and three variables (lipid LW, water LW, and lipid/water integration ratio) had no significant correlation (all P-values > 0.1).

Therefore, there is another possible explanation for the difference in MRS of the femoral head: it may be due to biochemical alteration of the lipid or water contents of the bone marrow within the femoral head. In most cases mentioned by Schellinger et al (19), the lipid signals were with various lipid fragments generally not solved. Brix et al (42) mentioned that the lipid signal is composed of at least eight fractions; it is the methylene group that contributes the largest signals and another ester group at its left aspect. In our experience, it is easy to identify methylene or ester groups in the range of the lipid spectrum. Although we did not analyze the different fragments of the lipid spectra, we expect that more fragments would be found in a wider spectrum. Thus, a narrower spectrum may suggest fewer fragments of the lipid peak. When the bone marrow becomes more fatty, the lipid contents may become more homogenous and lose some chemical contents. This alteration in the internal biochemical contents of the femoral head may play a more important role than vascular or other factors in the mechanism of AVN.

Several possible mechanisms lead to AVN. For example, a change in the intraosseous compartment can increase intraosseous pressure and decrease blood flow to the medullar bone (1). An important mechanism is a change in fat cells in the femoral head, as typically observed in animals treated with corticosteroids (9). Under histological examination, fat cells in the marrow increase by more than 10 μm in diameter, and the epiphyseal fat-cell volume fraction increases by 25–28% after steroid treatment (8). The increased size of fat cells certainly decreases the only compressible space in the femoral head, the sinusoidal vascular bed, resulting in a variety of ischemic changes. Thus, steroid-induced ischemic necrosis is one of the well-known mechanisms. In addition, post-traumatic osteonecrosis is another well-recognized etiology due to mechanical disruption of the vascular supply of the affected femoral head. In our research, we only excluded patients with post-traumatic and steroid-induced AVN of femoral head. Other factors, such as alcoholism, hyperbaric diving, hemoglobinopathies, other drugs, pregnancy, collagen vascular disease, and pancreatitis, were not screened.

One of the uncertainties of this study is whether the differences in the MRS data are related to age and sex. De Bisschop et al (43) evaluated the fat fraction of bone marrow by using MRS methods and found that the measured percentage of fat increased by about 7% per decade of age. Schellinger et al (19) used proton MRS to evaluate vertebral fat content and observed an age-dependent, linear increase in fat content, with sex dependence. They also investigated the potential value of proton MRS for evaluating vertebral bone weakness, and found that the fat fraction increased by 45% in subjects with MR findings of bone weakness (27). These reports were related to the hematopoietic bone marrow of lumbar vertebrae. In our study the bone marrow of the femoral head epiphysis was fatty marrow, which completes its fatty marrow conversion before the age of 20. We also confirmed this by showing no difference in MRS parameters of the femoral head among different age and sex groups.

There are several limitations to our study. First, we need to figure out the relationship between the MR spectral LW and bony trabeculation, so we need the BMD for the femoral head. However, the DEXA of the hip cannot offer such data because the femoral head epiphysis is overlapped with the acetabulum. The separate BMD of the femoral head epiphysis cannot be obtained from DEXA. Quantitative CT scanning is one of the best ways to measure the bone density of the femoral head; however, this equipment or phantom for calibration is not available in some research institutes.

A sequential follow-up with MRI and MRS in our diseased and control groups will provide important information. We can detect early AVN from serial follow-ups. According to the new occurrence of AVN, the patients at risk for AVN can be further subdivided into low- and high-risk groups, and the cutoff point can be estimated according to their initial MRS parameters. We can also observe the sequential changes of lipid/water spectra in the diseased and control groups. Those data may be helpful for defining the cutoff point of the MRS index or the odds ratio to predict the risk for AVN. Therefore, studies with an increased number of MRS subjects and sequential follow-up are ongoing at our institute.

In conclusion, the lipid and water spectra of the femoral head, measured by proton MRS, proved to be potentially useful for evaluating alterations in the internal environment of the bone marrow. A change in the lipid/water spectra of the femoral head in people at risk for AVN compared with control subjects is detectable before any morphologic changes can be observed. Our data may help elucidate at least one of the mechanisms that cause idiopathic osteonecrosis of the human hip. Further follow-up studies with sequential MRS comparisons in individuals at risk for AVN may reveal more information about the possible mechanisms and predict the risk for developing AVN of the femoral heads.

Acknowledgements

The authors thank their colleagues at the MRI Division of the Department of Radiology, National Taiwan University Hospital, and the Siemens medical engineers for their support and assistance in this research.

Ancillary