Proton MR Spectroscopy in Normal Breasts Between Preand Postmenopausal Women: A Preliminary Study

Transcription

1 Women s Imaging Original Research MR Spectroscopy in Normal Breasts Women s Imaging Original Research Jane Wang 1,2 Pao-Ling Torng 3 Tsang-Pai Liu 4,5 Kuan-Lin Chen 1 Tiffany Ting-Fang Shih 1,2 Wang J, Torng PL, Liu TP, Chen KL, Shih TTF Keywords: breast MRI, MRI, MR spectroscopy, normal breast, postmenopause, premenopause, single-voxel proton MR spectroscopy ( 1 H-MRS) DOI:1214/AJR Received April 8, 2007; accepted after revision August 17, Supported by grant 94S43 from the National Taiwan University Hospital. 1 Department of Radiology, National Taiwan University College of Medicine, Taipei, Taiwan. 2 Department of Medical Imaging, National Taiwan University Hospital, 7 Chung-Shan S Rd., Taipei 100, Taiwan. Address correspondence to T. T. F. Shih. 3 Department of Obstetrics and Gynecology, National Taiwan University Hospital and National Taiwan University College of Medicine, Taipei, Taiwan. 4 Department of Surgery, Mackay Memorial Hospital, Taipei, Taiwan. 5 Department of Surgery, Mackay Medicine, Nursing, and Management College, Taipei, Taiwan. AJR 2008; 190: X/08/ American Roentgen Ray Society Proton MR Spectroscopy in Normal Breasts Between Preand Postmenopausal Women: A Preliminary Study OBJECTIVE. The purpose of our study was to assess the differences in the water and lipid fractions and lipid line widths in normal breasts between premenopausal and postmenopausal women using single-voxel proton MR spectroscopy ( 1 H-MRS). MATERIALS AND METHODS. Thirty-two premenopausal and 25 postmenopausal women were enrolled in the study. Single-voxel proton MR spectroscopy of the breast was performed using point-resolved spectroscopy (PRESS) with water suppression and stimulated echo acquisition mode (STEAM). On STEAM, water fraction 1 was the ratio of the integration of water to the sum of the integration of water and methylene resonances, and the lipid fraction 1 was the ratio of the integration of methylene to water and methylene resonances. Lipid fraction 2 was the ratio of the integration of allylic methylene to water and allylic methylene resonances. Lipid line width was measured on PRESS. RESULTS. The premenopausal group had a higher water fraction 1 and lower lipid fraction 1 than the postmenopausal group (p < 1, Student s t test). The breast density had a positive effect on water fraction 1 and a negative effect on lipid fraction 1 for premenopausal women (p = 18, multivariate regression) and for the total population (p = 19). The premenopausal women had a higher lipid fraction 2 than postmenopausal women without significance (Student s t test), but the premenopausal status had a positive effect on lipid fraction 2 (p = 24, multivariate regression). There was no significant correlation between all independent variables and lipid line width. CONCLUSION. Breast 1 H-MRS shows the differences of water and lipid compositions between pre- and postmenopausal women. Lipids containing methylene and allylic methylene protons had different implications in normal breasts. S ingle-voxel proton MR spectroscopy ( 1 H-MRS) of the breast is occasionally used to investigate the chemical composition of breast tumors [1 4]. It was reported that a choline peak at 3.25 ppm may be used as an indicator of malignancy, and the presence or absence of the choline peak increases the specificity of differentiating benign from malignant breast lesions, thus reducing unnecessary biopsies [1 7]. MRI is sometimes used as a screening tool for evaluation of breast tissue changes using various techniques for presumably normal breasts in asymptomatic women [8 11]. Reichenbach et al. [9] reported that 3D MRI using a segmentation algorithm and a histogram was able to detect and quantify breast tissue changes before and during hormonal replacement therapy. Single-voxel proton MR spectroscopy has been applied to ex vivo normal breast tissue to estimate the volumetric water content of the extracted tissue, from which the fraction of fibroglandular tissue within individual samples can be inferred [10]. However, changes of the water and lipid compositions of normal breast tissue due to longer hormonal effects, such as pre- or postmenopausal status, have not been studied. The water and lipid content evaluations between premenopausal and postmenopausal women using 1 H-MRS were conducted in bone marrow [12], and the results revealed that the lipid water ratio differed significantly between pre- and postmenopausal women [12]. Knowledge of the differences of water and lipid compositions in normal breast tissue between pre- and postmenopausal women is important when 1 H-MRS is used for determining the metabolic activity of breast diseases. The goal of our study was to measure the water and lipid fractions and lipid line width in normal breast tissues of pre- and AJR:190, February

2 postmenopausal women using 1 H-MRS and thus provide baseline information on normal breast metabolism in pre- and postmenopausal women. The correlations between the quantitative data and possible causative and correlated factors were also analyzed. Materials and Methods Study Population From October 2004 to April 2006, 57 women were enrolled in the study. Of them, 32 were premenopausal (age range, years; mean age (± SD), ± 7.56 years) and 25 were postmenopausal women (age range, years; mean age, ± 4.98 years). All subjects met the following inclusion criteria: all were asymptomatic without a history or present evidence of breast cancer. The criteria for enrollment into the postmenopausal group were: amenorrhea for more than 1 year, signs of hypoestrogenemia, and follicle-stimulating hormone (FSH) level greater than 40 IU/L [13]. None of the premenopausal participants received oral contraceptives. Hormone replacement therapy was not prescribed for the postmenopausal women within 6 months before commencement of the study. All study participants received either screening mammography or sonography within 3 months either before or after the 1 H-MRS examination. Screening mammography was performed in participants 50 years or older, mammography or screening sonography was performed for those 40 to 49 years old, and sonography was performed for participants under 40 years old. All of the screening imaging studies showed normal or typically benign findings. This study was approved by the institutional review board of our hospital, and all the study participants signed informed consent before the 1 H-MRS examination. MRI Protocol for 1 H-MRS Acquisition All study participants underwent MRI of the breast on a 1.5-T MR system (Magnetom Sonata, Siemens Medical Solutions). The imaging protocol was unilateral sagittal FLASH (TR/TE, 20/5; flip angle, 25 ; slice thickness/interslice gap, 2/ mm; matrix, ; field of view, 300 mm), bilateral coronal 2D FLASH (20/5; flip angle, 25 ; slice thickness/gap, 2/ mm; matrix ), bilateral axial FLASH (25/5; slice thickness/gap, 3/ mm; matrix, ; field of view, 300 mm), bilateral axial fast STIR (9,120/67, inversion time, 150 milliseconds; flip angle, 150 ; slice thickness/gap, 5/1; matrix, ; field of view, 300 mm). We performed unilateral right sagittal FLASH and 1 H-MRS for 51 participants. For the other six women, unilateral left breast images were obtained because they had a history of excisional biopsy for benign lesions on the right. A cubic volume of interest (VOI) with dimensions of 10 or 15 mm was sampled over the area with the most abundant fibroglandular tissue according to sagittal, coronal, and axial FLASH images (Fig. 1), avoiding areas of cysts according to axial STIR sequences. A cubic voxel of 10 mm was used for 47 participants (23 pre- and 24 postmenopausal women), and a cubic voxel of 15 mm was used for 10 participants (nine premenopausal women and one postmenopausal woman). Voxel positioning and measurements for all study participants were performed by the same investigator. Point-resolved spectroscopy (PRESS) and stimulated echo acquisition mode (STEAM) were performed subsequently. The protocol for PRESS was 1,500/30; flip angle, 90 with water suppression; bandwidth, 1,000 Hz; acquisition, 32 signals; 1,024 vector sizes. The STEAM protocol was 1,500/20; flip angle 90, acquisition, 64 signals; 1,024 vector sizes; without water suppression; bandwidth, 1,000 Hz. The area of integration under the lipid and water peaks and line width on each 1 H-MRS pulse sequence were measured by curve fitting using the 1 H-MRS software package for the Sonata system (Siemens Medical Solutions). Clinical and 1 H-MRS Data Measurements and Analysis The basic clinical data of age, body mass index (BMI), and breast composition for each participant were recorded. BMI was measured as body weight in kg / (body height in m) [2]. The breast composition (breast density) was scored according to the American College of Radiology (ACR) BI-RADS for MRI [14]: 1, the breast is almost entirely fat; 2, the breast is composed of scattered fibroglandular tissue; 3, the breast is composed of heterogeneous fibroglandular A Fig year-old woman with presumably normal right breast. A and B, Coronal (A) and sagittal (B) FLASH and axial images (not shown) show cubic voxel positioning at area with most abundant fibroglandular breast tissue. Single-voxel proton MR spectroscopy was then performed. tissue; and 4, the breast is composed mostly of fibroglandular tissue. The breast density pattern was scored by one of the study authors on the basis of the FLASH and STIR pulse sequences. We defined the lipid 1 (L1) as the methylene resonance (CH 2 ) at about 1.6 ppm on 1 H-MRS, whereas lipid 2 (L2) was defined as the allylic methylene resonance (CH 2 CH 2 CH=) at about 2.3 ppm, and water peak (H 2 O peak) at 4.8 ppm (Fig. 2). The H 2 O peak and L1 peak were separated by about 3.2 ppm (220 Hz) using a 1.5-T magnet. Water fraction and lipid fraction were measured on the basis of the STEAM sequence and were defined as Water fraction 1 = I H2O / (I H2O + I L1 ) and lipid fraction 1 = I L1 / (I H2O + I L1 ). Therefore, water fraction 1 + lipid fraction 1 = 1. Lipid fraction 2 = I L2 / (I H2O + I L2 ) where I H2O, I L1, and I L2 are the integrations for water, lipid 1, and lipid 2, respectively. According to the PRESS imaging for each participant, the lipid line width for lipid 1 was defined as the line width at the lipid 1 peak. For lipid 2, the line width was defined as the line width at the lipid 2 peak. The integrations and line width values for water and lipid peaks were recorded for only the signals that were clearly measurable by curve fitting on 1 H-MRS. However, there was another peak at about 0.9 ppm contributing to the lipid of methyl protons, which was clearly measurable in only one of 57 participants (who was postmenopausal) on STEAM and 26 of 57 participants (13 pre- and 13 postmenopausal) on PRESS. Therefore, we did not include the data analysis for this peak at 0.9 ppm. In addition, we did not analyze another peak at ppm in proximity to the H 2 O B 506 AJR:190, February 2008

3 MR Spectroscopy in Normal Breasts SI ppm Fig. 2 Major water and lipid peaks on single-voxel proton MR spectroscopy of breast in 45-year-old woman with presumably normal right breast. SI = signal intensity. A, Graph shows that on STEAM (stimulated echo acquisition mode) pulse sequence, there is major lipid peak contributing to methylene resonance located at approximately 1.6 ppm (lipid 1 [L1], arrow), and second most common lipid peak at about 2.3 ppm, contributing to allylic methylene protons (lipid 2 [L2], arrowhead). Water peak is located at 4.8 ppm (double thin arrows). B, Graph of PRESS (point-resolved spectroscopy) with water suppression shows lipid 1 (L1, arrow), lipid 2 (L2, arrowhead), and water peak (double thin arrows). peak because it was clearly measurable in only 10 participants on STEAM (seven premenopausal and three postmenopausal) and none on PRESS. Statistical Analysis The basic clinical data of age and BMI were compared between the pre- and postmenopausal groups using the Student s t test. The difference in breast density between the pre- and postmenopausal groups was compared using Fisher s exact test. There were five dependent variables derived from 1 H-MRS measurements: water fraction 1, lipid fraction 1, lipid fraction 2, lipid line width at lipid 1, and lipid line width at lipid 2. The differences of the five dependent variables from 1 H-MRS were compared between the pre- and postmenopausal groups using Student s t test. The presumed three independent factors of age, BMI, and breast density and the correlation with the five dependent factors in the premenopausal group were analyzed using multivariate regression; the same analytic method was also applied to the postmenopausal group. These are the intragroup analyses. When considering all of the study participants as a whole, we added a fourth independent factor: hormonal status (pre- or postmenopausal status, with premenopausal status coded as 1 and postmenopausal status coded as 0). Therefore, the correlations of the five dependent variables and the four independent factors of age, BMI, breast density (scored as 1, 2, 3, or 4), and hormonal status (coded as 1 or 0) were also tested by multivariate regression to further clarify the correlation of pre- or postmenopausal status to all the dependent factors. All the statistical analyses were performed using SAS, version 8.1 A SI (SAS Institute). A p value of less than 5 was considered statistically significant. Results On STEAM images, all the participants had an H 2 O peak and a major L1 peak, resulting in 57 sets of water fraction 1/lipid fraction 1 for analysis. The second most common lipid peak, L2, was found on the STEAM images of 48 participants but not on those of the other nine participants (eight premenopausal women and one postmenopausal woman), resulting in 48 sets of lipid fraction 2 (24 pre- and 24 postmenopausal) for data analysis. On the other hand, on the PRESS sequence images, all participants 0 2 ppm B showed a clearly measurable L1 peak, with 57 sets of lipid line width at lipid 1 for analysis. But there were a total of 53 sets of lipid line width at lipid 2 for data analysis (29 preand 24 postmenopausal women) because four (three premenopausal and one postmenopausal) women did not show a clearly measurable L2 peak on PRESS images. The variables and their analyses using Student s t test are summarized in Table 1. The BMI of the premenopausal group was significantly lower than that of the postmenopausal group. The water fraction 1 analysis revealed a significant difference between the premenopausal and postmenopausal groups, with water fraction 1 of the premenopausal group being greater than that of the postmenopausal group. Therefore, the lipid fraction 1 of the premenopausal group was significantly lower than that of the postmenopausal group. The lipid 2 of the premenopausal group was slightly higher than that of the postmenopausal group but without statistical significance. Lipid line widths at lipid 1 (contributing to the methylene peak at 1.6 ppm) on PRESS sequence images for the postmenopausal group were significantly lower than those of the premenopausal group. Lipid line width at lipid 2 did not differ significantly between the two groups. The breast composition (density) for the preand postmenopausal women was also evaluated (Table 2). The distributions of breast density pattern between the pre- and postmenopausal groups differed significantly, with the premenopausal group having a denser breast parenchymal pattern than the postmenopausal group (p < 001, Fisher s exact test). TABLE 1: Comparison of Continuous Variables Between Pre- and Postmenopausal Women Variable Premenopausal Postmenopausal Cases No. ± SD Cases No. ± SD Age (y) ± ± BMI ± ± H 2 OF ± ± b LF ± ± b LF ± ± 7 NS L1W ± ± L2W ± ± 14 NS Note BMI = body mass index, H 2 OF1 = water fraction 1, LF1 = lipid fraction 1, LF2 = lipid fraction 2, L1W = lipid 1 line width in PRESS (point-resolved spectroscopy), L2W = lipid 2 line width in PRESS, NS = not significant. a Student s t test. b The p values of H 2 OF1 and LF1 are the same because H 2 OF1 + LF1 = 1. p a AJR:190, February

4 TABLE 2: Distribution of Breast Pattern Between Pre- and Postmenopausal Women Premenopausal No. (%) Postmenopausal No. (%) 1 0 () 6 (24.0) 2 9 (28.12) 10 (4) 3 11 (34.38) 8 (32.0) 4 12 (37.50) 1 (4.0) Note Breast density scored as 1 4 in accordance with ACR BI-RADS [14]. Fisher s exact test, p = We further analyzed the correlations of the three independent and the five dependent variables using multivariate regression analyses for the pre- and postmenopausal groups separately. For the premenopausal group, after controlling for age and BMI effects, water fraction 1 had a significantly positive correlation with breast density. Thus lipid fraction 1 had a significantly negative correlation with breast density (Table 3 and Figs. 3A and 3B) that is, water fraction 1 tended to be higher and lipid fraction 1 tended to be lower in denser breasts than in more fatty breasts. Lipid fraction 2, lipid line width at lipid 1, and lipid line width at lipid 2 of the premenopausal group showed no significant correlation with breast density. BMI of the premenopausal group showed no significant effect on all of the dependent variables (Table 3). We further analyzed the correlation of the aforementioned independent and dependent variables for the postmenopausal group using multivariate regression analyses (Table 3). Lipid fraction 2 had a significantly positive correlation with age in postmenopausal women (Table 3 and Fig. 3C). Neither the water fraction nor the lipid line width data had a significant correlation with BMI or breast density in the postmenopausal group. When considering the pre- and postmenopausal groups as a whole, we estimated the correlation of the five dependent variables with the four independent variables of age, BMI, breast density, and hormonal status using multivariate regression analyses (Table 4). When we controlled for age, hormonal status, and BMI, water fraction 1 had a significantly positive correlation and lipid fraction 1 had a significantly negative correlation with breast TABLE 3: Correlation Between Independent and Dependent Variables in Premenopausal and Postmenopausal Groups Variable H 2 OF1 (p) LF1 (p) LF2 (p) L1W (p) L2W (p) Premenopausal Age (y) 01 (NS) 01 (NS) 04 (NS) (NS) (NS) BMI 004 (NS) 004 (NS) 13 (NS) (NS) (NS) 84 (18) 84 (18) 5 (NS) (NS) (NS) Postmenopausal Age (y) 06 (NS) 06 (NS) 06 (49) 85 (NS) 54 (NS) BMI 11 (NS) 11 (NS) 04 (NS) 76 (NS) 55 (NS) 28 (NS) 28 (NS) 15 (NS) (NS) (NS) Note Data are multivariate regression analysis with p value in parentheses. H 2 OF1 = water fraction 1, LF1 = lipid fraction 1, LF2 = lipid fraction 2, L1W = lipid 1 line width in PRESS (point-resolved spectroscopy), L2W = lipid 2 line width in PRESS, NS = not significant, BMI = body mass index. TABLE 4: Correlation Between Independent and Dependent Variables in Total Study Population Variable H 2 OF1 (p) LF1 (p) LF2 (p) L1W (p) L2W (p) Age (y) 02 (NS) 02 (NS) 04 (NS) 90 (NS) (NS) Hormonal 09 (NS) 09 (NS) 98 (24) (NS) (NS) status a BMI 07 (NS) 07 (NS) (NS) (NS) (NS) 59 (19) 59 (19) 30 (NS) (NS) (NS) Note Data are multivariate regression analysis with p value in parentheses. H 2 OF1 = water fraction 1, LF1 = lipid fraction 1, LF2 = lipid fraction 2, L1W = lipid 1 line width in PRESS (point-resolved spectroscopy), L2W = lipid 2 line width in PRESS, NS = not significant, BMI = body mass index. a Premenopausal status coded as 1 and postmenopausal status coded as 0. density (Table 4 and Figs. 4A and 4B). The breast density showed no significant effect on the two lipid line width variables. BMI showed no significant effect on any of the five dependent variables after controlling for age, breast density, and hormonal effects for the total study population. When we controlled for age, BMI, and breast density effects, the premenopausal status had a significantly positive effect on lipid fraction 2. This implied that the premenopausal women tended to have a higher lipid fraction 2 than the postmenopausal women (Table 4 and Fig. 4C). Hormonal status showed no significant effect on water fraction 1, lipid fraction 1, and the two lipid line width variables. Discussion We found that the premenopausal women tended to have a higher water fraction 1 and a lower lipid fraction 1 for methylene resonances than those of postmenopausal women. The breast density shows a significantly positive effect on water fraction 1 after controlling for other effects in premenopausal women and in the total population. The premenopausal women tended to have a higher lipid fraction 2 for allylic methylene resonances than postmenopausal women when controlling for other effects. Although the lipid 1 line width of the premenopausal group was significantly higher than that of the postmenopausal group, it did not have significant correlations with any independent variables on multivariate regression analysis. The fibroglandular volume fraction of the breast was reported to have a strong correlation with the relative volumetric water content [11]. The percentage of fibroglandular tissue in the breast decreases with age, but noticeable fatty replacement occurs around the time of menopause [15, 16]. Therefore, our results for water fraction 1 and lipid fraction 1 are compatible with those of the other series because we presumed that the water fraction had a positive correlation with the fibroglandular tissue fraction and, inversely, that the lipid fraction represented the intermixed fatty portion in a given voxel of the breast. However, there are also many other factors that could have contributed to our results. The premenopausal group had a younger mean age than that of the postmenopausal group. Accordingly, it is reasonable that premenopausal women tended to have denser breast composition, possibly not only caused by hormonal effects but also due to age. Therefore, we performed the multivariate 508 AJR:190, February 2008

5 MR Spectroscopy in Normal Breasts H2OF1 Fig. 3 Scatter plots from multivariate regression analysis in pre- and postmenopausal women. Only results with statistical significance (p < 5) are shown. A, For premenopausal group, water fraction 1 (H2OF1) is positively correlated with breast density. (β [regression coefficient] = 84; p = 18) B, For premenopausal group, lipid fraction 1 (LF1) is negatively correlated with breast density (β = 84; p = 18) C, For postmenopausal group, lipid fraction 2 (LF2) has positive correlation with age (β = 06; p = 49) H2OF1 LF1 regression analyses, which revealed that water fraction 1 was positively correlated with breast density in the premenopausal group and the total study population but not significant for the postmenopausal group. This was possibly because the postmenopausal group consisted of a smaller sampling size. The most obvious difference between lipid fraction 1 and lipid fraction 2 is the lipid components, and the hormonal effect was attributed more to the allylic methylene protons according to multivariate regression analysis (Table 4). We assumed that the allylic methylene protons had a different, even opposite, contribution from the methylene protons of the breast according to our results. Allylic methylene protons originate from fatty acid chains with multiple double bonds such as linoleic acid, linolenic acid, and arachidonic acid [8]. The intensity changes of allylic methylene protons during the menstrual cycle have a 14-day cyclic change over 28 days, and the cyclic change shows a reciprocal tendency compared with methylene protons the lipid 1 peak in our study [8]. Although our study was not longitudinal for lipid composition changes during the menstrual cycle and the method of estimating the lipid fractions is different from the study just discussed, we also observed somewhat different implications and changes between the methylene and allylic methylene protons. To our knowledge, there has never been any published data discussing the lipid line width in the normal breast. According to the studies of lipid line width from bone marrow, a higher bone density is expected to reveal greater magnetic field inhomogeneity and wider spectral peaks. Line width was therefore influenced by magnetic field inhomogeneity stemming from bone trabeculae as well as the internal composition of the bone marrow content [12, 17]. However, from our study, we presumed that the premenopausal women tended to have wider line width than did the postmenopausal women, possibly due to more overlapping, LF Age (y) A B C LF1 heterogeneous species of lipid spikes in premenopausal breasts. However, we could not find any statistically significant correlation between the lipid line width and all the independent variables (Tables 3 and 4). Although the premenopausal participants tended to have higher lipid fraction 2 values, the older women in the postmenopausal group also had higher lipid fraction 2. These two results were contradictory because most of the premenopausal women are younger than the postmenopausal women. We have four possible explanations for these results. First, age and hormonal status could be two distinctly independent factors that have a certain link but cannot be discussed altogether. Second, although there was a significant positive correlation (p = 49) between lipid fraction 2 and age in the postmenopausal group, this correlation was not significant in the premenopausal group and the total study population. Therefore, we cannot prove the definite correlation of lipid fraction 2 with the age fac- Fig. 4 Scatter plots from multivariate regression analysis in total study population. Only results with p < 5 are shown. A and B, Water fraction 1 (H2OF1) (A) is positively correlated with breast density (β [regression coefficient] = 59; p = 19), and lipid fraction 1 (B) has negative correlation with breast density (β = 59; p = 19). C, Lipid fraction 2 has positive correlation with hormonal status, with the premenopausal status coded as 1 and the postmenopausal status coded as 0. (β = 98; p = 24) LF2 0 1 Hormonal Status A B C AJR:190, February

6 tor in the premenopausal group and the total population. Third, in postmenopausal women, there was a marked decrease in serum estradiol levels, increased FSH and luteinizing hormone (LH), and mild reduction of testosterone when compared with premenopausal women [13], and these hormonal changes also progress with age. Although the most prominent change among these was estradiol level, changes of the other hormonal levels between the pre- and postmenopausal groups may also possibly play important roles in the correlation between hormonal status and lipid fraction 2. Therefore, we further speculated that estradiol level has a stronger positive effect on lipid fraction 2 and contributes more than do FSH and LH. However, the above presumptions were difficult to clarify because we did not obtain detailed serum hormone levels close to or on the day of 1 H-MRS examination for each participant. Fourth, we did not further stratify the premenopausal group into different phases, such as follicular or luteal phase of the menstrual cycle, because the sampling size for each phase would have been too small for meaningful statistical analysis. The result after stratification for different phases may likely be different from our current study because there are cyclic changes in estradiol, FSH, LH, progesterone [13], and methylene protons and allylic methylene protons during the menstrual cycle [8]. What are the potential clinical benefits from our study? According to the results that the water fraction 1 had a positive correlation with breast density and the lipid fraction 2 was related to hormonal status, the water and lipid fractions may be indicators to assess breast cancer risk for women with presumably normal breasts, given a larger sampling size and a longer follow-up period. There are some limitations that need to be further discussed. First, we did not include the peak at 5.4 to 6.0 ppm adjacent to major water proton peak (4.8 ppm) for water fraction measurements on STEAM because this peak may have been contaminated by the resonance from olefinic acid, which may be due to insufficient shimming. Moreover, only 10 participants showed this peak, and the sampling size was limited. Second, we did not analyze the lipid peak at 0.9 ppm (methyl protons) in this study because this peak was measurable in only one participant on STEAM and 26 participants on PRESS, with an inadequate sampling size for analysis. Third, the stratification of breast density according to ACR BI-RADS scales may be suboptimal because the categories may be subjectively scored by radiologists, and another alternative method is to quantify the volume of fibroglandular tissue by segmentation techniques in MRI [9]. Fourth, as mentioned previously, we didn t stratify the premenopausal group into different phases to measure water and lipid compositions. Because it was reported that there were really cyclic changes of water and lipid compositions in different phases [8, 11], other alternative methods are to perform all the 1 H-MRS examinations at the same phase or to perform serial 1 H-MRS examinations throughout the different phases for the same group of participants, thus ensuring the interval validity while still maintaining an acceptable sampling size [8, 11]. In conclusion, water fraction 1 showed a positive correlation and lipid fraction 1 for methylene protons showed a negative correlation with breast density in the premenopausal group and in the total population. The premenopausal women had a higher lipid fraction 2 for allylic methylene protons than that of the postmenopausal women. The methylene and allylic methylene protons were expected to represent different implications in the 1 H- MRS of the normal breast, and the composition changes of allylic methylene protons are more prone to be influenced by pre- or postmenopausal status. These results need to be validated by larger-scale studies. References 1. Gribbestad IS, Sitter B, Lundgren S, Krane J, Axelson D. Metabolic composition in breast tumors examined by proton nuclear magnetic resonance spectroscopy. Anticancer Res 1999; 19: Roebuck JR, Cecil KM, Schnall MD, Lenkinski RE. Human breast lesions: characterization with proton MR spectroscopy. Radiology 1998; 209: Mackinnon WB, Barry PA, Malycha PL, et al. Fine-needle biopsy specimens of benign breast lesions distinguished from invasive cancer ex vivo with proton MR spectroscopy. Radiology 1997; 204: Gribbstad IS, Singstad TE, Nilsen G, et al. In vivo 1 H MRS of normal breast and breast tumors using a dedicated double breast coil. J Magn Reson Imaging 1998; 8: Bartella L, Morris EA, Dershaw DD, et al. Proton MR spectroscopy with choline peak as malignancy marker improves positive predictive value for breast cancer diagnosis: preliminary study. Radiology 2006; 239: Kvistad KA, Bakken IJ, Gribbestad IS, et al. Characterization of neoplastic and normal human breast tissues with in vivo 1 H MR spectroscopy. J Magn Reson Imaging 1999; 10: Yeung DKW, Yang WT, Tse GMK. Breast cancer: in vivo proton MR spectroscopy in the characterization of histopathologic subtypes and preliminary observations in axillary node metastases. Radiology 2002; 225: Dzendrowskyj TE, Noyszewski EA, Beers J, Bolinger L. Lipid composition changes in normal breast throughout the menstrual cycle. MAGMA 1997; 5: Reichenbach JR, Przetak C, Klinger G, Kaiser WA. Assessment of breast tissue changes on hormonal replacement therapy using MRI: a pilot study. J Comput Assist Tomogr 1999; 23: Graham SJ, Ness S, Hamilton BS, Bronskill MJ. Magnetic resonance properties of ex vivo breast tissue at 1.5T. Magn Reson Med 1997; 38: Graham SJ, Stanchev PL, Lloyd-Smith JO, Bronskill MJ, Plewes DB. Changes in fibroglandular volume and water content of breast tissue during the menstrual cycle observed by MR imaging at 1.5 T. J Magn Reson Imaging 1995; 5: Shih TT, Chang CJ, Hsu CY, Wei SY, Su KC, Chung HW. Correlation of bone marrow lipid water content with bone mineral density on the lumbar spine. Spine 2004; 29: Hurd WH, Amesse LS, Randolph JF. Menopause. In: Berek JS, ed. Novak s gynecology, 13th ed. Philadelphia, PA: Lippincott Williams & Wilkins, 2002: American College of Radiology (ACR). Magnetic resonance imaging. In: ACR Breast Imaging Reporting and Data System, breast imaging atlas. Reston, VA: American College of Radiology, Kopans DB. Breast imaging, 2nd ed. Philadelphia, PA: Lippincott-Raven, 1998: Wolfe JN. Breast parenchymal patterns and their changes with age. Radiology 1976; 121(3 Pt 1): Schellinger D, Lin CS, Fertikh D, et al. Normal lumbar vertebrae: anatomic, age, and sex variance in subjects at proton MR spectroscopy initial experience. Radiology 2000; 215: AJR:190, February 2008