1991, The British Journal of Radiology, 64, 210-216
In vivo phosphorus-31 magnetic resonance spectroscopy of normal and pathological breast tissues By O. M. Redmond, PhD, J. P. Stack, MRCPI, FFR(RCSI), N, G. O'Connor, SRN, M. B. Codd, MB, MPH and J. T. Ennis, MD, FRCPI, FRCR Institute of Radiological Sciences, 52 Eccles Street, Dublin 7, Ireland {Received November 1989 and in revised form August 1990) Keywords: Phosphorus-31 MRS, Breast, Malignant, Benign, Tissue characterization Abstract. In vivo phosphorus-31 magnetic resonance spectroscopy, at 1.5 T, in addition to magnetic resonance imaging and mammography, was performed on the breast tissue of 59 subjects, using a 40 mm or 80 mm surface coil for spectral localization. The patients were divided into three groups: Group 1, 46 control subjects; Group 2, nine patients with breast carcinoma; Group 3, four patients with benign breast disease. The relationship of age, menopausal status, breast size and pattern, use of contraceptive pill and history of breast disease to spectral characteristics of breast tissue was examined for the control group. In multivariate analysis, only menopausal status and age were found to be significantly related to tissue biochemistry. Pre-menopausal women had reduced phosphocreatine (PCr) (%) {p = 0.02), and increased phosphomonoesters (PMEs) and /?-nucleotide triphosphate (/9-NTP) (%) (p = 0.05), while the fat-to-water ratio was higher in older women (p = 0.02). No significant differences were identified between the control subjects and the patients with benign breast disease. When spectra from patients with breast carcinoma were compared with an age-matched volunteer group, a- and y-NTP (%) were found to be higher in the cancer tissue (p < 0.01 and p = 0.01, respectively), while PCr (%) was reduced (p < 0.01). The ratio /?-NTP:PCr was higher in the carcinoma group of patients (p < 0.05). In vivo phosphorus-31 magnetic resonance spectroscopy is a non-invasive examination which may prove useful in the early differentiation of malignant breast disease from normal and benign conditions.
It is estimated that one in 14 women in western society can expect to develop breast carcinoma in their lives, but there is, as yet, no single biochemical test to identify women "at risk". The main risk factors associated with breast carcinoma include menstrual, reproductive and family histories (Kelsey, 1979; Ottman et al, 1983). Strong and/or prolonged oestrogen stimulus has also been associated with increased risk, although the mechanism of action is poorly understood (Vihko & Apter, 1986). This suggests that there may be an, as yet, unidentified high risk endocrine or biochemical profile. Phosphorus-31 magnetic resonance spectroscopy (MRS) is a non-invasive technique that has been used for many years to study metabolism in vitro and experimental animals in vivo (Moon & Richards, 1973; Cozzane & Jardetzky, 1976; Ackerman et al, 1980; Evanochko et al, 1982; Ng et al, 1982; Evanochko et al, 1983, 1984; Hollander, 1986; Irving et al, 1986; Mountford et al, 1986; Sijens et al, 1987). The phosphorus-containing biomolecules, which contribute to the spectrum, include the low energy phosphomonoesters (PMEs) and phosphodiesters (PDEs), thought to be important precursors in cell membrane formation and products of phospholipid catabolism respectively (Cozzane & Jardetzky, 1976). The relative concentration of the inorganic phosphate (Pi) peak and the high energy phosphates, phosphocreatine (PCr) and nucleotide triphosphates (NTP), provide valuable information with respect to tissue energetics, while the position of Pi
Address correspondence to: Dr O. M. Redmond, Institute of Radiological Sciences, 52 Eccles Street, Dublin 7, Ireland.
210
is an indicator of intracellular tissue pH. (This is calculated by measuring the chemical shift of Pi, which varies within the physiological pH range to a peak of fixed position within this range, such as PCr or /J-NTP.) Whole-body MR imagers, operating at 1.5 T or greater, have recently become available, enabling in vivo MRS of human tissues to be performed. Owing to the superficial nature of breast tissue, surface coils were used in this study to obtain phosphorus-31 spectra. This is a simple and quick method of spectral localization; however, the risk of contamination from adjacent muscle tissue is high. To date, phosphorus-31 MRS of human breast tissue has been attempted by a small number of groups (Cohen et al, 1986; Degani et al, 1986; Oberhaensli et al, 1986; Sijens et al, 1987; Merchant et al, 1988; Redmond et al, 1988; Sijens et al, 1988). Using perchloric acid extracts, Merchant et al (1988) demonstrated that malignant, benign and non-involved breast tissue could be differentiated in vitro on the basis of phosphorus-31 MR spectra. In vivo results have varied considerably, which is not surprising since patient numbers tended to be very small. A number of groups (Degani et al, 1986; Oberhaensli et al, 1986; Merchant et al, 1988; Redmond et al, 1988) have obtained phosphorus-3 1 spectra of breast carcinoma in situ and found an increase in the concentration of PME in breast carcinomas and a decrease in the fat-to-water ratio (FWR). To our knowledge, there has been no extensive study to investigate the variation in phosphorus-31 metabolites in normal breast tissue. We performed in vivo phosphorus-31 MRS on 46 normal volunteers and examined the relative concentraThe British Journal of Radiology, March 1991
In vivo 3IP MRS of breast tissue
MR images, the breast pattern was deemed to be (1) largely glandular (less than 30% fat), (2) mixed glandular and fatty (30-70% fat), or (3) fatty (greater than 70% fat). Breast size was determined from the maximum distance from the skin of the breast to the chest wall. While supine positioning of the patient increased the likelihood of muscle contamination, we attempted to minimize this by using the 40 mm diameter surface coil Patients and methods Phosphorus-31 MRS and magnetic resonance when the distance from the skin to the chest wall was imaging (MRI) were performed on a total of 59 women less than or equal to 5 cm. When this distance was who were divided into three groups. Group 1 consisted greater than 5 cm, the 80 mm coil was used. The transof 46 volunteers (mean, 45 years; range, 20-68 years). mitter voltage was selected (from calibration graphs The volunteers were asked to complete a questionnaire constructed from phantom studies) depending on coil which gave information with respect to menopausal size and the distance from the skin to the chest wall, in status, use of contraceptive pill, family history of breast order to obtain the maximum signal from the breast disease and history of breast feeding. Group 2 (n — 9) tissue with minimal contamination from muscle. had biopsy proven breast carcinoma (mean, 58 years; Hydrogen-1 images were acquired with the phosrange, 49-74 years). One patient (patient 13) had bila- phorus-31 surface coil acting as receiver. Although the teral breast carcinoma. Spectra from both breasts were volumes sampled at hydrogen and phosphorus frequenvery similar and only one set of peak areas (from the left cies do not correlate precisely, owing to differences in breast) was included for data analysis (see Table I). All coil sensitivity at the different frequencies, this proof these patients had MRI and MRS examinations prior cedure is useful to ensure accurate positioning of the to surgery and/or chemotherapy. Group 3 consisted of surface coil and to give a rough estimate of the sampled four patients with biopsy proven benign breast disease volume of the resulting spectral data. (mean, 52 years; range, 48-57 years), all of whom had Initially, equal numbers of spectra were obtained MRI and MRS performed. from the right and left breasts. It soon became apparent, Magnetic resonance imaging and MRS were carried however, that spectra from the left breast had inferior out with a whole-body 1.5 T superconducting magnet resolution due to cardiac motion. Consequently, spectra (Siemens Magnetom). Both breasts were imaged with were preferentially obtained from the right breast. the patient in a supine position, the body coil acting as Shimming of the magnetic field Bo to the sensitive the transmitter and receiver. A multislice T,-weighted volume of the surface coil was done at hydrogen sequence (repetition time (TR) = 0.5 s, echo time frequency (63 MHz) until the resolution (full width at (TE) = 17 ms) was used as a localization technique for half maximum (FWHM)) of the proton peak from spectroscopy as the phosphorus-31 surface coil was in tissue water was less than 0.5 ppm. This was not always position during the sequence. In addition, this sequence possible, however, owing to respiratory and cardiac gave useful information with respect to breast size and movement and in a small number of cases (n — 3) the pattern, and location and size of any lesions. From these resolution (FWHM) was as poor as 1.0 ppm. tions of phosphorus-31 metabolites in relation to the following factors: age, stage of menstrual cycle, breast size and pattern, use of contraceptive pill and history of breast disease. We also attempted to characterize benign and malignant breast lesions on the basis of their in vivo phosphorus-31 MR spectra.
Table I. Characteristics of patient group Patient
Age (years)
2
1
51 53
3
56
4
49
5 6
74 56
7 8 9 10 11 12 13
65 65 52 57 48 53 58
Vol. 64, No. 759
Histology
Remarks
2 x 1 cm3 masses 5 cm diameter, metastatic disease in 3 axillary Infiltrating, poorly differentiated ductal carcinoma nodes 5 cm diameter, metastatic disease in 7 axillary Infiltrating, poorly differentiated ductal carcinoma nodes 13 cm diameter Infiltrating, high-grade poorly differentiated 2 cm diameter carcinoma 8 cm diameter, metastatic disease in 7 axillary Infiltrating ductal carcinoma Diffusely infiltrative, poorly differentiated lobular nodes 3 cm diameter carcinoma Infiltrating intraductal and intralobular carcinoma 3 cm diameter Multiple small cysts ( < 1 cm) Ductal adenocarcinoma Fibrocystic disease/focal epitheliosis Benign dysplastic breast disease Fibrocystic disease Diameter of cyst, 3 cm aspirated Fibrocystic disease Left breast 10 cm diameter, right breast 8 cm Bilateral infiltrating ductal carcinoma diameter Infiltrating intraductal carcinoma
211
O. M. Redmond, J. P. Stack, N. G. O'Connor, M. B. Codd and J. T. Ennis
Phosphomonoester, Pi and PDE peaks were poorly resolved in the corresponding phosphorus-31 spectra and they were not included in subsequent analyses. The FWR was calculated from the relative areas of the fat and water peaks in the hydrogen spectra. One hundred and twenty-eight pulse excitations were used for the phosphorus-31 spectra with TR = 6 s, delay time De = 500 n&, and dwell time Dw = 400 /xs for 512 data points (12.48 min). Multiplication of the free induction decay (FID) by a decaying exponential was introduced in cases of poor signal-to-noise ratio. This results in a signal with enhanced signal-to-noise ratio, but also with increased linewidth. Following Fourier transformation of the FID data, from the time to the frequency domain, the spectra were phase corrected and baseline subtracted as required. The subtracted baseline consisted of polynomial sections, computer constructed, from up to 64 fitting points specified by the user. Peak areas were calculated by plotting the spectra on graph paper and extending the peak sides to the baseline. Where lines from different peaks overlapped, the angle of the uppermost vertex of the resulting triangle was bisected. The number of squares per peak was then counted and the peak areas were expressed as a percentage of the total spectral phosphorus-31 metabolites. Peak ratios were also calculated in an attempt to minimize noise differences between spectra. The pH was read off a standard curve of pH versus |PCr-Pi| chemical shift. Lateral views of both breasts were acquired with xeromammography in volunteers over 40 years old (or over 35 years old if in a high risk group). Both lateral and antero-posterior views were acquired in Groups 2 and 3 (i.e. in the patients with pathological lesions). Of the nine patients with breast carcinoma, four had immediate surgery, two had surgery following chemotherapy, and three had chemotherapy with no surgery. Patient details and phosphorus-31 MR spectral characteristics were entered into pre-coded data sheets to facilitate computer analysis. Statistical analysis was carried out, using SAS (statistical analysis system), on an IBM mainframe computer. Univariate analyses consisted of Student's /-tests or / 2 tests as appropriate. The data were also subjected to multivariate analysis using multiple regression. The relationships of PME, Pi, PDE, PCr, y-NTP, a-NTP and jS-NTP peak areas (as percentages) as well as the FWR and chemical shift of |PCr-Pi|, to age, family history, use of contraceptive pill, breast feeding, pre- or post-menopause, breast size and pattern, were examined. Results
Table II shows the results of the questionnaire obtained from the volunteer group. Twenty-eight of the women examined were pre-menopausal while 17 were post-menopausal. The mean age was 45.1 + 10.8 years, 21 volunteers being less than or equal to 45 years old and the remaining 25 being over 45 years old. Fifteen women had a maximum distance of less than 5 cm from the skin to the chest wall, as seen at MRI with the 212
Table II. Details of volunteer group Number Percentage Age (years)
45 Yes No Yes No Yes No Pre Post < 5 cm
21 25 8 38 2 44 16 30 28 17 15
46 54 17 83 4 96 35 65 62 38 33
Maximum anteroposterior breast size (from MRI) ^ 5 cm Breast pattern 70%
31 13 19 14
67 28 41 31
Family history of breast cancer Use of contraceptive pill Breastfed at least 1 child Menopause
patient in a supine position, while this was greater than or equal to 5 cm in the remaining 31 patients. Magnetic resonance spectroscopy was performed on the right breast of 40 volunteers and on the left breast of six. Examination of the right breast provided optimal spectral resolution. The spectral resolution was considered good in 15 volunteers (FWHM less than or equal to 0.5 ppm), fair in 23 (FWHM of 0.5-0.7 ppm), and poor in eight (FWHM greater than 0.7 ppm). Phosphomonoester, Pi and PDE peaks were the most dificult to resolve. An increased number of acquisitions (with a considerable increase in MRS examination time!) should increase the spectral resolution and signalto-noise ratio (SNR) of normal breast tissue spectra. Some degree of spectral contamination, evident from Table III. Mean phosphorus-31 metabolite peak areas ( + 1 SD) expressed as the percentage of total phosphorus-31 spectral area in the volunteer group (Group 1), patients with breast carcinoma (Group 2) and patients with benign breast disease (Group 3)
Age PME Pi PDE PCr y-NTP a-NTP jS-NTP pH F:W
Group 1 (n = 46)
Group 2 (n = 9)
Group 3 (« = 4)
Mean (SD)
Mean (SD)
Mean (SD)
45.1 (10.8) "10.3 (4.6) "13.7 (3.2) "12.7 (3.8) 30.3 (6.3) 10.3 (1.5) 12.4 (2.0) 10.1 (1.6) 7.08 (0.12) 1.03 (0.74)
58.3 (8.1) 9.9 (4.6) 10.8 (2.6) 18.3 (8.4) 24.3 (12.3) 13.2 (2.0) 14.8 (2.8) 9.8 (1.3) 7.23 (0.13) 0.98 (0.7)
52.6 10.3 11.6 10.6 32.4 10.9 13.5 10.9 7.01 1.07
(3.6) (4.0) (3.0) (6.0) (6.6) (1.9) (1.6) (2.0) (0.23) (1.00)
"PME, Pi and PDE peak areas from three volunteer subjects were not included owing to poor spectral resolution; n = 43 for these three values. The British Journal of Radiology, March 1991
In vivo 3IP MRS of breast tissue
I
PME
Pi APDE wTP A
%
10.
5.
0.
-5.
-10.
15.
-15.
Chemical shift / ppm
10.
5.
0.
- 5 . -10. -15. -20. -25.
Chemical s h i f t / ppm
(a) (b) Figure 1. Phosphorus-31 MR spectra of (a) normal breast tissue and (b) breast carcinoma prior to therapy.
the prominent PCr peaks, was evident in most volunteer spectra. Table III shows the mean (±SD) phosphorus-31 metabolic peak areas in the volunteer and two patient groups. Figure 1 shows a phosphorus-31 spectrum of normal breast tissue compared with a spectrum obtained from a breast carcinoma. Phosphodiesters are elevated and PCr is reduced in the breast carcinoma spectrum. When patient data and spectral characteristics were
subjected to univariate analysis, a number of significant relationships were observed (see Table IV). Phosphomonoester (%) and PCr (%) were significantly associated with both menopausal status and age; j?-NTP was significantly associated with both menopausal status and breast pattern, while the FWR was associated with menopausal status, age, breast pattern and size. No association was observed between any spectral characteristics and variables such as history of breast
Table IV. Univariate and multivariate analyses of PME (%), ^-NTP (%), PCr (%) and F:W ratio ( + 1 SD) versus menopausal status, age, breast pattern and breast size in the volunteer group
Menopausal status Pre-menopausal Post-menopausal Significance level Univariate Multivariate Age ^ 45 years > 45 years Significance level Univariate Multivariate Breast pattern < 30% fat > 70% fat Significance level Univariate Multivariate Breast size "Diameter ^ 5 cm "Diameter < 5 cm Significance level Univariate Multivariate
PME (%) Mean (SD)
jS-NTP (%) Mean (SD)
PCr (%) Mean (SD)
F:W Mean (SD)
11.6(4.8) 8.3 (3.7)
10.4(1.5) 9.3 (1.4)
28.2 (4.7) 33.5 (7.5)
0.74 (0.5) 1.45 (0.9)
p = 0.02 NS (p = 0.06)
p = 0.04 p = 0.05
p = 0.02 p = 0.02
p = 0.005 NS
11.8(4.8) 9.0 (4.1)
10.3 (1.5) 9.9 (1.7)
28.2 (4.8) 32.3 (7.0)
0.69 (0.4) 1.28(0.8)
p = 0.05 NS
NS NS
p = 0.04 NS
p = 0.005 p = 0.02
11.8 (5.7) 8.4 (3.6)
10.7 (1.6) 9.2 (1.5)
28.5 (4.9) 33.4 (8.0)
0.83 (0.7) 1.61 (0.8)
NS NS
p = 0.03 NS
NS NS
p = 0.02 NS
10.3 (4.1) 10.3 (4.9)
10.6(1.7) 9.8 (1.5)
29.9 (5.3) 30.4 (6.9)
1.19(0.7) 0.70 (0.7)
NS NS
NS NS
NS NS
p = 0.04 NS
"Diameter is the maximum antero-posterior diameter. Vol. 64, No. 759
213
O. M. Redmond, J. P. Stack, N. G. O'Connor, M. B. CoddandJ. T. Ennis
feeding, use of contraceptive pill and family history of breast cancer. Multivariate analysis of this data showed phosphorus metabolism of breast tissue to be most strongly associated with menopausal status of the volunteer. The high energy phosphates (/J-NTP and PCr) were both significantly related to menopausal status. When age and menopausal status were considered together in relation to PME (%), using a multivariate model, "age" was not found to be statistically significant, while "menopausal status" only tended towards significance {p = 0.06). This suggests that "age" may be a confounding factor for "menopausal status" in relation to PME (%). The statistical significance of PME (%) with respect to menopausal status should improve with increased numbers. Multivariate analysis of the FWR with age, menopausal status, breast pattern and size showed "age" to be the most significant factor (p = 0.02). The nine patients with breast carcinoma were compared with an age-matched group of volunteers (n = 12). Stronger signal and improved spectral resolution were readily obtained from malignant breast tissue compared with normal breast tissue, despite similar acquisition parameters. The mean peak areas (%) for each of the phosphorus-31 metabolites is shown in Table V. Mean Pi and PCr (%) and FWR were all reduced in the carcinoma group, while PDE, a-NTP and y-NTP (%) were increased. When subjected to statistical analysis, however (Student's /-test), only the differences in percentage o'f y-NTP (p = 0.01), a-NTP (p < 0.01), and PCr (p < 0.01) were statistically significant. When ratios of phosphorus-31 metabolites were calculated, Pi:PDE was lower in the carcinoma group of patients (p = 0.10) while j3-NTP:PCr was higher (p < 0.05). Mean PDE:PCr values were also considerably higher in the breast carcinoma tissues although the differences were not statistically significant. The mean Table V. Phosphorus-31 metabolite peak areas (%) ( ± 1SD) in patients with breast carcinoma versus an age-matched control group
PME Pi PDE PCr y-NTP a-NTP 0-NTP pH F:W Pi:PDE PDE:PCr j?-NTP:PCr
214
Volunteers (n = 12)
Breast carcinoma Significance level (« = 9)
Mean (SD)
Mean (SD)
8.30 (4.0) 9.9 (4.6) 13.1 (3.3) 10.8 (2.6) 18.3 (8.4) 12.9 (1.1) 32.9 (80) 24.3 (12.3) 10.8 (1.6) 13.2 (2.0) 12.4 (2.3) 14.8 (2.8) 9.4 (1.5) 9.8 (1.3) 7.08 (0.12) 7.23 (0.13) 1.64 (0.8) 0.98 (0.7) 1.10 (0.4) 0.67 (0.3) 0.44 (0.3) 1.13 (0.9) 0.52 (0.3) 0.30(0.1)
V
Br.Ca. Pis
Br.Ca. Pts
Vote
NTP/PCr
Br.Ca.Pts Vote PI/PDE
Figure 2. Pi:PDE, PDErPCr and j9-NTP:PCr ratios in breast carcinoma ( # ) versus normal breast tissue (O) from an agematched group of 12 volunteers.
pH value for the carcinoma group was pH 7.23 ±0.13. This is higher than that observed in the volunteer group (pH 7.08 + 0.12) although the difference is not statistically significant. Figure 2 is a graph of Pi:PDE, PDE:PCr and /?-NTP:PCr ratios in breast carcinoma compared with normal breast tissue from an age-matched volunteer group. Mean phosphorus-31 metabolite peak areas for patients with benign breast disease are given in Table VI. These results suggest an increase in PME levels and a reduction in Pi and PDE values and in Table VI. Phosphorus-31 metabolite peak areas (%) ( + 1 SD) for patients with benign breast disease versus control group
Mean NS NS NS p < 0.01 /> = 0.01 p < 0.01 NS NS NS /7 = 0.10 NS p < 0.05
Vols
PDE/PCr
PME Pi PDE PCr y-NTP a-NTP
0-NTP pH F:W
Volunteers (n = 12)
Benign breast disease (n = 4)
Mean (SD)
Mean (SD)
8.3 (4.0) 13.1 (3.3) 12.9 (1.1) 32.9 (8.0) 10.8 (1.6) 12.4 (2.3) 9.4 (1.5) 7.08 (0.12) 1.64 (0.8)
10.3 (4.0) 11.6 (3.0) 10.6 (6.0) 32.4 (6.6) 10.9 (1.9) 13.5 (1.6) 10.9 (2.0) 7.01 (0.2) 1.07 (1.0)
The British Journal of Radiology, March 1991
In vivo 3IP MRS of breast tissue
FWRs in the patient group, although the patient numbers are too small to allow reliable statistical analysis. Discussion
This study demonstrates that differences in tissue biochemistry can be detected in breasts of normal volunteers (n — 46) depending on menopausal status and age. High energy NTPs, such as adenosine triphosphate (ATP), were elevated in pre-menopausal women compared with post-menopausal women. Peak areas of PMEs, which include contributions from phosphorylcholine, phosphorylethanolamine, glucose and fructose phosphates, as well as other sugars, were also elevated in the pre-menopausal group. Phosphorus-31 MRS identified a doubling of the mean FWR in the breast tissue of volunteers over 45 years old, compared with those less than 45 years old. This is in agreement with autopsy studies and corresponds with fatty replacement of parenchyma in breast tissue with time. The increased fat content of older breast tissue may have a role in local oestrogen production which may be especially important in the case of post-menopausal women (Miller, 1986), as the supply of preformed active hormone reaching the breast declines after the menopause. Phosphocreatine (%) was reduced in the breast tissue of pre-menopausal women while NTP (%) was increased. This is contrary to the finding of Sijens et al (1988) who, in a small study of four normal controls (range 25-40 years), showed elevated PCr relative to five patients with malignant breast tumours. Their conclusion that "PCr is high in young women" may, in fact, relate more to reduced PCr in carcinomatous breast tissue as opposed to elevated PCr in young control subjects. Older, post-menopausal control subjects were not investigated. The pill has been suspected of being a cancer induction agent, particularly in women on long-term medication. Only two women in the group of volunteers had used the contraceptive pill and these showed no significant differences from the rest of the group, although the patient number is obviously too small to draw any conclusions. History of breast feeding or family history of breast carcinoma could not be shown to correlate with any specific spectral characteristics. A wide range of pH values was obtained in this group (pH 6.9-7.4), which did not appear to correlate with any of the variables investigated. Improved localization techniques, however, with the elimination of possible contamination by underlying muscle tissue is required to confirm these findings. We were unable to differentiate benign breast disease from normal breast tissue on the basis of in vivo phosphorus-31 spectra. This may have been because of the low numbers investigated in this group (n — 4), making statistical analysis with the other two groups both difficult and unreliable. It may also be due to partial voluming effect with contamination of benign breast tissue spectra with signal from surrounding normal Vol. 64, No. 759
breast tissue. This is also a problem with the malignant tumours studied. Some of these were smaller than the 4 cm surface coil used, and spectral contamination from surrounding normal breast tissue could reduce the differences observed between the malignant and control groups. Degani et al (1986) differentiated benign breast lesions (n = 5) from breast carcinoma (n = 9) in vitro, on the basis of reduced PME and NTP in the benign group relative to the carcinoma group. Neither group, however, was compared with normal breast tissue. Another group (Merchant et al, 1988) again performing in vitro phosphorus-31 MRS of human breast tissue extracts, examined the same three tissue groups as ourselves (malignant, benign and normal breast parenchymal specimens). They found that PME metabolism was altered in all neoplastic tissues relative to noninvolved tissues, that PCr was elevated in benign disease, and that benign breast disease showed an increase in an uncharacterized phosphate, resonating at 3.66(5, relative to the shift position of phosphoric acid. We did not observe any of these changes. Even if the number of patients in this group were increased in our study, it is still unlikely that in vivo phosphorus-31 MRS would have the potential to differentiate "normal" breast tissue from "benign breast disease", as the term "benign breast disease" is used to encompass a wide range of pathologies including mammary cysts, adenosis, fibroadenomas, apocrine changes and epithelial hyperplasia (Love et al, 1982). Even in the work of Merchant et al (1988) referred to above, the benign specimens (n = 8) included a very wide range of benign breast disease, i.e. fibrocystic disease, fibroadenoma, benign cystic changes and focal fibrosis. In a recent American study, in which 360 women with no history of breast problems were autopsied, 58% were shown to have "benign breast disease". In the group of patients with breast carcinoma (n = 9), PCr values (%) were lower than in either the control group or the patient group with benign breast disease (p < 0.01). This is in agreement with the work of many other groups (Evanochko et al, 1984; Oberhaensli et al, 1986; Merchant et al, 1988; Sijens et al, 1988). Mean PDE (%) was increased in the breast carcinoma group, although the range of values in each group was quite high. A similar increase in breast carcinoma PDE (%) was identified by Sijens et al (1988) and probably represents a rapid turnover in phospholipids with a simultaneous formation of new cells. Reduced Pi:PDE (p = 0.10) and increased PDE:PCr ratios were observed in the group of patients with breast carcinoma, possibly owing to increased PDE (%). The statistical significance of these findings should improve with increased patient numbers. Expressing phosphorus-31 metabolite peak areas as ratios minimizes the effect of different levels of spectral noise, and therefore enables a more accurate comparison between spectra. Greater numbers are required to determine if elevated PDE (%) is a useful indicator of malignant breast disease. y-NTP (%) was found to be elevated in breast carcinoma tissue relative to the other two groups (p = 0.01) 215
O. M. Redmond, J. P. Stack, N. G. O'Connor, M. B. Codd and J. T. Ennis
as was a-NTP (%) (p < 0.01). This suggests that levels of nucleotide diphosphates, most probably ADP, which contribute to the phosphorus-31 spectra in the same region as the y-NTP and a-NTP peaks, are elevated in breast carcinoma tissue. Mean pH values for the carcinoma group were higher than those observed in either the volunteer group or the benign group, although this difference was not statistically significant. This would agree, however, with other work demonstrating the alkaline nature of breast carcinomas (Oberhaensli et al, 1986). It suggests that these carcinomas are well vascularized and/or oxygenated and, while a small number of dispersed hypoxic cells could easily escape detection by in vivo MRS, that they were not the predominant population in the tumours studied. (The presence of even a small number of hypoxic cells, however, could be highly significant in determining the success or failure of treatment, particularly with radiotherapy.)
by in vivo 3I P-NMR in athymic mice. Biochemical and Biophysical Research Communications, 109, 1346-1352. EVANOCHKO, W. T., N G , T. C , LILLY, M. B., LAWSON, A. J., CORBETT, T. H., DURRANT, J. R., & GLICKSON, J. D., 1983.
In vivo 3I P study of the metabolism of murine mammary 16/C adenocarcinoma and its response to chemotherapy, x-radiation and hyperthermia. Proceedings of the National Academy of Sciences of the United States of America, 80, 334-338. EVANOCHKO, W. T., N G , T. C. & GLICKSON, J. D., 1984.
Applications of in vivo NMR spectroscopy to cancer. Magnetic Resonance in Medicine, 1, 508-534. HOLLANDER, J., 1986. In vivo tissue analysis by NMR spectroscopy. Diagnostic Imaging Clinical Medicine, 55, 9-19. IRVING, M. G., SIMPSON, S. J., FIELD, J. & DODDRELL, D. M.,
1985. Use of high resolution 3lP-labelled topical magnetic resonance spectroscopy to monitor in vivo tumour metabolism in rats. Cancer Research, 45, 481-486. KELSEY, J. L., 1979. A review of the epidemiology of human breast cancer. Epidemiologic Reviews, 1, 74-109. LOVE, S. M., GELLMAN, R. S. & SILEN, W., 1982. Fibrocystic
Conclusion
This study has shown that is is possible, using phosphorus-31 MRS, to detect differences in breast tissue biochemistry in a group of control subjects on the basis of age and menopausal status. Likewise, this non-invasive in vivo technique identified abnormalities in phosphorus metabolism in carcinomatous breast tissue. Magnetic resonance spectroscopy may, therefore, improve the differentiation of benign but symptomatic tumours, such as fibroadenomas, from malignant disease. Magnetic resonance spectroscopy also has potential in monitoring response to therapy. Of the nine patients with malignant breast disease evaluated in this study, five had follow-up MRS examinations post-chemotherapy (mean 14 per patient). Spectral changes were monitored and are currently being evaluated with respect to patient response. Acknowledgments We would like to thank the Irish Health Research Board for partial funding of this project.
"disease" of the breast—a non-disease? New England Journal of Medicine, 307, 1010-1014. MERCHANT, T. E., GIERKE, L. W., MENESES, P. & GLONEK, T.,
1988. 31P magnetic resonance spectroscopic profiles of neoplastic human breast tissues. Cancer Research, 48, 5112-5118. MILLER, W. R., 1986. Steroid metabolism in breast cancer. Reviews on Endocrine-Related Cancer, 23, 23-30. MOON, R. B. & RICHARDS, J. H., 1973. Determination of
intracellular pH by 3I P magnetic resonance. Journal of Biological Chemistry, 248, 7276-7278. MOUNTFORD, C , MAY, G., WILLIAMS, P., TATTERSALL, M., RUSSELL, P., SAUNDERS, J., HOLMES, K., FOX, R., BARR, J. &
SMITH, I., 1986. Classification of human tumours by highresolution magnetic resonance spectroscopy. Lancet, 1, 641-653. N G , T. C , EVANOCHKO, W. T., HIRAMOTO, R. N., GHANTA, V. K., LILLY, M. B., LAWSON, A. J., CORBETT, T. H.,
DURRANT, J. R. & GLICKSON, J. D., 1982. "P-NMR spectro-
scopy of in vivo tumours. Journal of Magnetic Resonance, 49, 271-296. OBERHAENSLI, R. D., HILTON-JONES, D., BORE, P. J., HANDS, L. J., RAMPLING, R. P. & RADDA, G. K., 1986. Biochemical
investigation of human tumours in vivo with phosphorus-31 magnetic resonance spectroscopy. Lancet, 11, 8-11. OTTMAN, R., PIKE, M. C , KING, M. & HENDERSON, B. E.,
References ACKERMAN, J., GROVE, T., WONG, G., GADIAN, D. & RADDA, 3I
G.K. 1980. Mapping of metabolites in whole animals by P NMR using surface coils. Nature, 283, 167-170. COHEN, J. S., LYON, R. C , CHEN, C , FAUSTINO, P. J., BATIST, G., SHOEMAKER, M., RUBALCABA, E. & COWAN, K. H., 1986.
Differences in phosphate metabolite levels in drug-sensitive and -resistant human breast cancer cell lines determined by P-31 magnetic resonance spectroscopy. Cancer Research, 46, 4087-4090. COZZANE, P. J. & JARDETZKY, O., 1976. Phosphorus-31 Fourier
transform nuclear magnetic resonance study of mononucleotides and dinucleotides: 1. Chemical shifts. Biochemistry, 15, 4853-4859. DEGANI, H., HOROWITZ, A. & ITZCHAK, Y.,
1986. Breast
tumours: evaluation with P-31 MR spectroscopy. Radiology, 161, 53-56. EVANOCHKO, W. T., N G , T. C , GLICKSON, J. D., DURRANT,
J. R. & CORBETT, T. H., 1982. Human tumours as examined
216
1983. Practical guide for estimating risk for familial breast cancer. Lancet, 2, 556-558. REDMOND, O., STACK, J. P., SCULLY, M., DERVAN, P., CARNEY,
D. & ENNIS, J. T., 1988. Tissue characterisation and spectral analysis of malignant tumours following chemotherapy. Tumor Diagnostik und Therapie, 9, 167-168. SlJENS, P. E., WlJRDEMAN, H. K., VERMEULEN, J. W. A H. & LUYTEN, P. R., 1987. NMR spectroscopy of human breast tumours before and after radiotherapy using a 1.5 T MRI system. In Proceedings of the Society for Magnetic Resonance in Medicine, Sixth Annual Meeting and Exhibition, New York, Vol. 2, p. 979. SlJENS, P. E., WlJRDEMAN, H. K., MOERLAND, M. A., BARKER, C. J. G., VERMEULEN, J. W. A. H. & LUYTEN, P.R., 1988. Human breast cancer in vivo: H-l and P-31 MR spectroscopy at 1.5 T. Radiology, 169, 615-620. VIHKO,
R. & APTER,
D., 1986. Hormonal
predisposition to breast cancer. Endocrine-Related Cancer, 23, 11-15.
profile and
Reviews
on
The British Journal of Radiology, March 1991
In vivo phosphorus-31 magnetic resonance spectroscopy of normal and pathological breast tissues By O. M. Redmond, PhD, J. P. Stack, MRCPI, FFR(RCSI), N, G. O'Connor, SRN, M. B. Codd, MB, MPH and J. T. Ennis, MD, FRCPI, FRCR Institute of Radiological Sciences, 52 Eccles Street, Dublin 7, Ireland {Received November 1989 and in revised form August 1990) Keywords: Phosphorus-31 MRS, Breast, Malignant, Benign, Tissue characterization Abstract. In vivo phosphorus-31 magnetic resonance spectroscopy, at 1.5 T, in addition to magnetic resonance imaging and mammography, was performed on the breast tissue of 59 subjects, using a 40 mm or 80 mm surface coil for spectral localization. The patients were divided into three groups: Group 1, 46 control subjects; Group 2, nine patients with breast carcinoma; Group 3, four patients with benign breast disease. The relationship of age, menopausal status, breast size and pattern, use of contraceptive pill and history of breast disease to spectral characteristics of breast tissue was examined for the control group. In multivariate analysis, only menopausal status and age were found to be significantly related to tissue biochemistry. Pre-menopausal women had reduced phosphocreatine (PCr) (%) {p = 0.02), and increased phosphomonoesters (PMEs) and /?-nucleotide triphosphate (/9-NTP) (%) (p = 0.05), while the fat-to-water ratio was higher in older women (p = 0.02). No significant differences were identified between the control subjects and the patients with benign breast disease. When spectra from patients with breast carcinoma were compared with an age-matched volunteer group, a- and y-NTP (%) were found to be higher in the cancer tissue (p < 0.01 and p = 0.01, respectively), while PCr (%) was reduced (p < 0.01). The ratio /?-NTP:PCr was higher in the carcinoma group of patients (p < 0.05). In vivo phosphorus-31 magnetic resonance spectroscopy is a non-invasive examination which may prove useful in the early differentiation of malignant breast disease from normal and benign conditions.
It is estimated that one in 14 women in western society can expect to develop breast carcinoma in their lives, but there is, as yet, no single biochemical test to identify women "at risk". The main risk factors associated with breast carcinoma include menstrual, reproductive and family histories (Kelsey, 1979; Ottman et al, 1983). Strong and/or prolonged oestrogen stimulus has also been associated with increased risk, although the mechanism of action is poorly understood (Vihko & Apter, 1986). This suggests that there may be an, as yet, unidentified high risk endocrine or biochemical profile. Phosphorus-31 magnetic resonance spectroscopy (MRS) is a non-invasive technique that has been used for many years to study metabolism in vitro and experimental animals in vivo (Moon & Richards, 1973; Cozzane & Jardetzky, 1976; Ackerman et al, 1980; Evanochko et al, 1982; Ng et al, 1982; Evanochko et al, 1983, 1984; Hollander, 1986; Irving et al, 1986; Mountford et al, 1986; Sijens et al, 1987). The phosphorus-containing biomolecules, which contribute to the spectrum, include the low energy phosphomonoesters (PMEs) and phosphodiesters (PDEs), thought to be important precursors in cell membrane formation and products of phospholipid catabolism respectively (Cozzane & Jardetzky, 1976). The relative concentration of the inorganic phosphate (Pi) peak and the high energy phosphates, phosphocreatine (PCr) and nucleotide triphosphates (NTP), provide valuable information with respect to tissue energetics, while the position of Pi
Address correspondence to: Dr O. M. Redmond, Institute of Radiological Sciences, 52 Eccles Street, Dublin 7, Ireland.
210
is an indicator of intracellular tissue pH. (This is calculated by measuring the chemical shift of Pi, which varies within the physiological pH range to a peak of fixed position within this range, such as PCr or /J-NTP.) Whole-body MR imagers, operating at 1.5 T or greater, have recently become available, enabling in vivo MRS of human tissues to be performed. Owing to the superficial nature of breast tissue, surface coils were used in this study to obtain phosphorus-31 spectra. This is a simple and quick method of spectral localization; however, the risk of contamination from adjacent muscle tissue is high. To date, phosphorus-31 MRS of human breast tissue has been attempted by a small number of groups (Cohen et al, 1986; Degani et al, 1986; Oberhaensli et al, 1986; Sijens et al, 1987; Merchant et al, 1988; Redmond et al, 1988; Sijens et al, 1988). Using perchloric acid extracts, Merchant et al (1988) demonstrated that malignant, benign and non-involved breast tissue could be differentiated in vitro on the basis of phosphorus-31 MR spectra. In vivo results have varied considerably, which is not surprising since patient numbers tended to be very small. A number of groups (Degani et al, 1986; Oberhaensli et al, 1986; Merchant et al, 1988; Redmond et al, 1988) have obtained phosphorus-3 1 spectra of breast carcinoma in situ and found an increase in the concentration of PME in breast carcinomas and a decrease in the fat-to-water ratio (FWR). To our knowledge, there has been no extensive study to investigate the variation in phosphorus-31 metabolites in normal breast tissue. We performed in vivo phosphorus-31 MRS on 46 normal volunteers and examined the relative concentraThe British Journal of Radiology, March 1991
In vivo 3IP MRS of breast tissue
MR images, the breast pattern was deemed to be (1) largely glandular (less than 30% fat), (2) mixed glandular and fatty (30-70% fat), or (3) fatty (greater than 70% fat). Breast size was determined from the maximum distance from the skin of the breast to the chest wall. While supine positioning of the patient increased the likelihood of muscle contamination, we attempted to minimize this by using the 40 mm diameter surface coil Patients and methods Phosphorus-31 MRS and magnetic resonance when the distance from the skin to the chest wall was imaging (MRI) were performed on a total of 59 women less than or equal to 5 cm. When this distance was who were divided into three groups. Group 1 consisted greater than 5 cm, the 80 mm coil was used. The transof 46 volunteers (mean, 45 years; range, 20-68 years). mitter voltage was selected (from calibration graphs The volunteers were asked to complete a questionnaire constructed from phantom studies) depending on coil which gave information with respect to menopausal size and the distance from the skin to the chest wall, in status, use of contraceptive pill, family history of breast order to obtain the maximum signal from the breast disease and history of breast feeding. Group 2 (n — 9) tissue with minimal contamination from muscle. had biopsy proven breast carcinoma (mean, 58 years; Hydrogen-1 images were acquired with the phosrange, 49-74 years). One patient (patient 13) had bila- phorus-31 surface coil acting as receiver. Although the teral breast carcinoma. Spectra from both breasts were volumes sampled at hydrogen and phosphorus frequenvery similar and only one set of peak areas (from the left cies do not correlate precisely, owing to differences in breast) was included for data analysis (see Table I). All coil sensitivity at the different frequencies, this proof these patients had MRI and MRS examinations prior cedure is useful to ensure accurate positioning of the to surgery and/or chemotherapy. Group 3 consisted of surface coil and to give a rough estimate of the sampled four patients with biopsy proven benign breast disease volume of the resulting spectral data. (mean, 52 years; range, 48-57 years), all of whom had Initially, equal numbers of spectra were obtained MRI and MRS performed. from the right and left breasts. It soon became apparent, Magnetic resonance imaging and MRS were carried however, that spectra from the left breast had inferior out with a whole-body 1.5 T superconducting magnet resolution due to cardiac motion. Consequently, spectra (Siemens Magnetom). Both breasts were imaged with were preferentially obtained from the right breast. the patient in a supine position, the body coil acting as Shimming of the magnetic field Bo to the sensitive the transmitter and receiver. A multislice T,-weighted volume of the surface coil was done at hydrogen sequence (repetition time (TR) = 0.5 s, echo time frequency (63 MHz) until the resolution (full width at (TE) = 17 ms) was used as a localization technique for half maximum (FWHM)) of the proton peak from spectroscopy as the phosphorus-31 surface coil was in tissue water was less than 0.5 ppm. This was not always position during the sequence. In addition, this sequence possible, however, owing to respiratory and cardiac gave useful information with respect to breast size and movement and in a small number of cases (n — 3) the pattern, and location and size of any lesions. From these resolution (FWHM) was as poor as 1.0 ppm. tions of phosphorus-31 metabolites in relation to the following factors: age, stage of menstrual cycle, breast size and pattern, use of contraceptive pill and history of breast disease. We also attempted to characterize benign and malignant breast lesions on the basis of their in vivo phosphorus-31 MR spectra.
Table I. Characteristics of patient group Patient
Age (years)
2
1
51 53
3
56
4
49
5 6
74 56
7 8 9 10 11 12 13
65 65 52 57 48 53 58
Vol. 64, No. 759
Histology
Remarks
2 x 1 cm3 masses 5 cm diameter, metastatic disease in 3 axillary Infiltrating, poorly differentiated ductal carcinoma nodes 5 cm diameter, metastatic disease in 7 axillary Infiltrating, poorly differentiated ductal carcinoma nodes 13 cm diameter Infiltrating, high-grade poorly differentiated 2 cm diameter carcinoma 8 cm diameter, metastatic disease in 7 axillary Infiltrating ductal carcinoma Diffusely infiltrative, poorly differentiated lobular nodes 3 cm diameter carcinoma Infiltrating intraductal and intralobular carcinoma 3 cm diameter Multiple small cysts ( < 1 cm) Ductal adenocarcinoma Fibrocystic disease/focal epitheliosis Benign dysplastic breast disease Fibrocystic disease Diameter of cyst, 3 cm aspirated Fibrocystic disease Left breast 10 cm diameter, right breast 8 cm Bilateral infiltrating ductal carcinoma diameter Infiltrating intraductal carcinoma
211
O. M. Redmond, J. P. Stack, N. G. O'Connor, M. B. Codd and J. T. Ennis
Phosphomonoester, Pi and PDE peaks were poorly resolved in the corresponding phosphorus-31 spectra and they were not included in subsequent analyses. The FWR was calculated from the relative areas of the fat and water peaks in the hydrogen spectra. One hundred and twenty-eight pulse excitations were used for the phosphorus-31 spectra with TR = 6 s, delay time De = 500 n&, and dwell time Dw = 400 /xs for 512 data points (12.48 min). Multiplication of the free induction decay (FID) by a decaying exponential was introduced in cases of poor signal-to-noise ratio. This results in a signal with enhanced signal-to-noise ratio, but also with increased linewidth. Following Fourier transformation of the FID data, from the time to the frequency domain, the spectra were phase corrected and baseline subtracted as required. The subtracted baseline consisted of polynomial sections, computer constructed, from up to 64 fitting points specified by the user. Peak areas were calculated by plotting the spectra on graph paper and extending the peak sides to the baseline. Where lines from different peaks overlapped, the angle of the uppermost vertex of the resulting triangle was bisected. The number of squares per peak was then counted and the peak areas were expressed as a percentage of the total spectral phosphorus-31 metabolites. Peak ratios were also calculated in an attempt to minimize noise differences between spectra. The pH was read off a standard curve of pH versus |PCr-Pi| chemical shift. Lateral views of both breasts were acquired with xeromammography in volunteers over 40 years old (or over 35 years old if in a high risk group). Both lateral and antero-posterior views were acquired in Groups 2 and 3 (i.e. in the patients with pathological lesions). Of the nine patients with breast carcinoma, four had immediate surgery, two had surgery following chemotherapy, and three had chemotherapy with no surgery. Patient details and phosphorus-31 MR spectral characteristics were entered into pre-coded data sheets to facilitate computer analysis. Statistical analysis was carried out, using SAS (statistical analysis system), on an IBM mainframe computer. Univariate analyses consisted of Student's /-tests or / 2 tests as appropriate. The data were also subjected to multivariate analysis using multiple regression. The relationships of PME, Pi, PDE, PCr, y-NTP, a-NTP and jS-NTP peak areas (as percentages) as well as the FWR and chemical shift of |PCr-Pi|, to age, family history, use of contraceptive pill, breast feeding, pre- or post-menopause, breast size and pattern, were examined. Results
Table II shows the results of the questionnaire obtained from the volunteer group. Twenty-eight of the women examined were pre-menopausal while 17 were post-menopausal. The mean age was 45.1 + 10.8 years, 21 volunteers being less than or equal to 45 years old and the remaining 25 being over 45 years old. Fifteen women had a maximum distance of less than 5 cm from the skin to the chest wall, as seen at MRI with the 212
Table II. Details of volunteer group Number Percentage Age (years)
45 Yes No Yes No Yes No Pre Post < 5 cm
21 25 8 38 2 44 16 30 28 17 15
46 54 17 83 4 96 35 65 62 38 33
Maximum anteroposterior breast size (from MRI) ^ 5 cm Breast pattern 70%
31 13 19 14
67 28 41 31
Family history of breast cancer Use of contraceptive pill Breastfed at least 1 child Menopause
patient in a supine position, while this was greater than or equal to 5 cm in the remaining 31 patients. Magnetic resonance spectroscopy was performed on the right breast of 40 volunteers and on the left breast of six. Examination of the right breast provided optimal spectral resolution. The spectral resolution was considered good in 15 volunteers (FWHM less than or equal to 0.5 ppm), fair in 23 (FWHM of 0.5-0.7 ppm), and poor in eight (FWHM greater than 0.7 ppm). Phosphomonoester, Pi and PDE peaks were the most dificult to resolve. An increased number of acquisitions (with a considerable increase in MRS examination time!) should increase the spectral resolution and signalto-noise ratio (SNR) of normal breast tissue spectra. Some degree of spectral contamination, evident from Table III. Mean phosphorus-31 metabolite peak areas ( + 1 SD) expressed as the percentage of total phosphorus-31 spectral area in the volunteer group (Group 1), patients with breast carcinoma (Group 2) and patients with benign breast disease (Group 3)
Age PME Pi PDE PCr y-NTP a-NTP jS-NTP pH F:W
Group 1 (n = 46)
Group 2 (n = 9)
Group 3 (« = 4)
Mean (SD)
Mean (SD)
Mean (SD)
45.1 (10.8) "10.3 (4.6) "13.7 (3.2) "12.7 (3.8) 30.3 (6.3) 10.3 (1.5) 12.4 (2.0) 10.1 (1.6) 7.08 (0.12) 1.03 (0.74)
58.3 (8.1) 9.9 (4.6) 10.8 (2.6) 18.3 (8.4) 24.3 (12.3) 13.2 (2.0) 14.8 (2.8) 9.8 (1.3) 7.23 (0.13) 0.98 (0.7)
52.6 10.3 11.6 10.6 32.4 10.9 13.5 10.9 7.01 1.07
(3.6) (4.0) (3.0) (6.0) (6.6) (1.9) (1.6) (2.0) (0.23) (1.00)
"PME, Pi and PDE peak areas from three volunteer subjects were not included owing to poor spectral resolution; n = 43 for these three values. The British Journal of Radiology, March 1991
In vivo 3IP MRS of breast tissue
I
PME
Pi APDE wTP A
%
10.
5.
0.
-5.
-10.
15.
-15.
Chemical shift / ppm
10.
5.
0.
- 5 . -10. -15. -20. -25.
Chemical s h i f t / ppm
(a) (b) Figure 1. Phosphorus-31 MR spectra of (a) normal breast tissue and (b) breast carcinoma prior to therapy.
the prominent PCr peaks, was evident in most volunteer spectra. Table III shows the mean (±SD) phosphorus-31 metabolic peak areas in the volunteer and two patient groups. Figure 1 shows a phosphorus-31 spectrum of normal breast tissue compared with a spectrum obtained from a breast carcinoma. Phosphodiesters are elevated and PCr is reduced in the breast carcinoma spectrum. When patient data and spectral characteristics were
subjected to univariate analysis, a number of significant relationships were observed (see Table IV). Phosphomonoester (%) and PCr (%) were significantly associated with both menopausal status and age; j?-NTP was significantly associated with both menopausal status and breast pattern, while the FWR was associated with menopausal status, age, breast pattern and size. No association was observed between any spectral characteristics and variables such as history of breast
Table IV. Univariate and multivariate analyses of PME (%), ^-NTP (%), PCr (%) and F:W ratio ( + 1 SD) versus menopausal status, age, breast pattern and breast size in the volunteer group
Menopausal status Pre-menopausal Post-menopausal Significance level Univariate Multivariate Age ^ 45 years > 45 years Significance level Univariate Multivariate Breast pattern < 30% fat > 70% fat Significance level Univariate Multivariate Breast size "Diameter ^ 5 cm "Diameter < 5 cm Significance level Univariate Multivariate
PME (%) Mean (SD)
jS-NTP (%) Mean (SD)
PCr (%) Mean (SD)
F:W Mean (SD)
11.6(4.8) 8.3 (3.7)
10.4(1.5) 9.3 (1.4)
28.2 (4.7) 33.5 (7.5)
0.74 (0.5) 1.45 (0.9)
p = 0.02 NS (p = 0.06)
p = 0.04 p = 0.05
p = 0.02 p = 0.02
p = 0.005 NS
11.8(4.8) 9.0 (4.1)
10.3 (1.5) 9.9 (1.7)
28.2 (4.8) 32.3 (7.0)
0.69 (0.4) 1.28(0.8)
p = 0.05 NS
NS NS
p = 0.04 NS
p = 0.005 p = 0.02
11.8 (5.7) 8.4 (3.6)
10.7 (1.6) 9.2 (1.5)
28.5 (4.9) 33.4 (8.0)
0.83 (0.7) 1.61 (0.8)
NS NS
p = 0.03 NS
NS NS
p = 0.02 NS
10.3 (4.1) 10.3 (4.9)
10.6(1.7) 9.8 (1.5)
29.9 (5.3) 30.4 (6.9)
1.19(0.7) 0.70 (0.7)
NS NS
NS NS
NS NS
p = 0.04 NS
"Diameter is the maximum antero-posterior diameter. Vol. 64, No. 759
213
O. M. Redmond, J. P. Stack, N. G. O'Connor, M. B. CoddandJ. T. Ennis
feeding, use of contraceptive pill and family history of breast cancer. Multivariate analysis of this data showed phosphorus metabolism of breast tissue to be most strongly associated with menopausal status of the volunteer. The high energy phosphates (/J-NTP and PCr) were both significantly related to menopausal status. When age and menopausal status were considered together in relation to PME (%), using a multivariate model, "age" was not found to be statistically significant, while "menopausal status" only tended towards significance {p = 0.06). This suggests that "age" may be a confounding factor for "menopausal status" in relation to PME (%). The statistical significance of PME (%) with respect to menopausal status should improve with increased numbers. Multivariate analysis of the FWR with age, menopausal status, breast pattern and size showed "age" to be the most significant factor (p = 0.02). The nine patients with breast carcinoma were compared with an age-matched group of volunteers (n = 12). Stronger signal and improved spectral resolution were readily obtained from malignant breast tissue compared with normal breast tissue, despite similar acquisition parameters. The mean peak areas (%) for each of the phosphorus-31 metabolites is shown in Table V. Mean Pi and PCr (%) and FWR were all reduced in the carcinoma group, while PDE, a-NTP and y-NTP (%) were increased. When subjected to statistical analysis, however (Student's /-test), only the differences in percentage o'f y-NTP (p = 0.01), a-NTP (p < 0.01), and PCr (p < 0.01) were statistically significant. When ratios of phosphorus-31 metabolites were calculated, Pi:PDE was lower in the carcinoma group of patients (p = 0.10) while j3-NTP:PCr was higher (p < 0.05). Mean PDE:PCr values were also considerably higher in the breast carcinoma tissues although the differences were not statistically significant. The mean Table V. Phosphorus-31 metabolite peak areas (%) ( ± 1SD) in patients with breast carcinoma versus an age-matched control group
PME Pi PDE PCr y-NTP a-NTP 0-NTP pH F:W Pi:PDE PDE:PCr j?-NTP:PCr
214
Volunteers (n = 12)
Breast carcinoma Significance level (« = 9)
Mean (SD)
Mean (SD)
8.30 (4.0) 9.9 (4.6) 13.1 (3.3) 10.8 (2.6) 18.3 (8.4) 12.9 (1.1) 32.9 (80) 24.3 (12.3) 10.8 (1.6) 13.2 (2.0) 12.4 (2.3) 14.8 (2.8) 9.4 (1.5) 9.8 (1.3) 7.08 (0.12) 7.23 (0.13) 1.64 (0.8) 0.98 (0.7) 1.10 (0.4) 0.67 (0.3) 0.44 (0.3) 1.13 (0.9) 0.52 (0.3) 0.30(0.1)
V
Br.Ca. Pis
Br.Ca. Pts
Vote
NTP/PCr
Br.Ca.Pts Vote PI/PDE
Figure 2. Pi:PDE, PDErPCr and j9-NTP:PCr ratios in breast carcinoma ( # ) versus normal breast tissue (O) from an agematched group of 12 volunteers.
pH value for the carcinoma group was pH 7.23 ±0.13. This is higher than that observed in the volunteer group (pH 7.08 + 0.12) although the difference is not statistically significant. Figure 2 is a graph of Pi:PDE, PDE:PCr and /?-NTP:PCr ratios in breast carcinoma compared with normal breast tissue from an age-matched volunteer group. Mean phosphorus-31 metabolite peak areas for patients with benign breast disease are given in Table VI. These results suggest an increase in PME levels and a reduction in Pi and PDE values and in Table VI. Phosphorus-31 metabolite peak areas (%) ( + 1 SD) for patients with benign breast disease versus control group
Mean NS NS NS p < 0.01 /> = 0.01 p < 0.01 NS NS NS /7 = 0.10 NS p < 0.05
Vols
PDE/PCr
PME Pi PDE PCr y-NTP a-NTP
0-NTP pH F:W
Volunteers (n = 12)
Benign breast disease (n = 4)
Mean (SD)
Mean (SD)
8.3 (4.0) 13.1 (3.3) 12.9 (1.1) 32.9 (8.0) 10.8 (1.6) 12.4 (2.3) 9.4 (1.5) 7.08 (0.12) 1.64 (0.8)
10.3 (4.0) 11.6 (3.0) 10.6 (6.0) 32.4 (6.6) 10.9 (1.9) 13.5 (1.6) 10.9 (2.0) 7.01 (0.2) 1.07 (1.0)
The British Journal of Radiology, March 1991
In vivo 3IP MRS of breast tissue
FWRs in the patient group, although the patient numbers are too small to allow reliable statistical analysis. Discussion
This study demonstrates that differences in tissue biochemistry can be detected in breasts of normal volunteers (n — 46) depending on menopausal status and age. High energy NTPs, such as adenosine triphosphate (ATP), were elevated in pre-menopausal women compared with post-menopausal women. Peak areas of PMEs, which include contributions from phosphorylcholine, phosphorylethanolamine, glucose and fructose phosphates, as well as other sugars, were also elevated in the pre-menopausal group. Phosphorus-31 MRS identified a doubling of the mean FWR in the breast tissue of volunteers over 45 years old, compared with those less than 45 years old. This is in agreement with autopsy studies and corresponds with fatty replacement of parenchyma in breast tissue with time. The increased fat content of older breast tissue may have a role in local oestrogen production which may be especially important in the case of post-menopausal women (Miller, 1986), as the supply of preformed active hormone reaching the breast declines after the menopause. Phosphocreatine (%) was reduced in the breast tissue of pre-menopausal women while NTP (%) was increased. This is contrary to the finding of Sijens et al (1988) who, in a small study of four normal controls (range 25-40 years), showed elevated PCr relative to five patients with malignant breast tumours. Their conclusion that "PCr is high in young women" may, in fact, relate more to reduced PCr in carcinomatous breast tissue as opposed to elevated PCr in young control subjects. Older, post-menopausal control subjects were not investigated. The pill has been suspected of being a cancer induction agent, particularly in women on long-term medication. Only two women in the group of volunteers had used the contraceptive pill and these showed no significant differences from the rest of the group, although the patient number is obviously too small to draw any conclusions. History of breast feeding or family history of breast carcinoma could not be shown to correlate with any specific spectral characteristics. A wide range of pH values was obtained in this group (pH 6.9-7.4), which did not appear to correlate with any of the variables investigated. Improved localization techniques, however, with the elimination of possible contamination by underlying muscle tissue is required to confirm these findings. We were unable to differentiate benign breast disease from normal breast tissue on the basis of in vivo phosphorus-31 spectra. This may have been because of the low numbers investigated in this group (n — 4), making statistical analysis with the other two groups both difficult and unreliable. It may also be due to partial voluming effect with contamination of benign breast tissue spectra with signal from surrounding normal Vol. 64, No. 759
breast tissue. This is also a problem with the malignant tumours studied. Some of these were smaller than the 4 cm surface coil used, and spectral contamination from surrounding normal breast tissue could reduce the differences observed between the malignant and control groups. Degani et al (1986) differentiated benign breast lesions (n = 5) from breast carcinoma (n = 9) in vitro, on the basis of reduced PME and NTP in the benign group relative to the carcinoma group. Neither group, however, was compared with normal breast tissue. Another group (Merchant et al, 1988) again performing in vitro phosphorus-31 MRS of human breast tissue extracts, examined the same three tissue groups as ourselves (malignant, benign and normal breast parenchymal specimens). They found that PME metabolism was altered in all neoplastic tissues relative to noninvolved tissues, that PCr was elevated in benign disease, and that benign breast disease showed an increase in an uncharacterized phosphate, resonating at 3.66(5, relative to the shift position of phosphoric acid. We did not observe any of these changes. Even if the number of patients in this group were increased in our study, it is still unlikely that in vivo phosphorus-31 MRS would have the potential to differentiate "normal" breast tissue from "benign breast disease", as the term "benign breast disease" is used to encompass a wide range of pathologies including mammary cysts, adenosis, fibroadenomas, apocrine changes and epithelial hyperplasia (Love et al, 1982). Even in the work of Merchant et al (1988) referred to above, the benign specimens (n = 8) included a very wide range of benign breast disease, i.e. fibrocystic disease, fibroadenoma, benign cystic changes and focal fibrosis. In a recent American study, in which 360 women with no history of breast problems were autopsied, 58% were shown to have "benign breast disease". In the group of patients with breast carcinoma (n = 9), PCr values (%) were lower than in either the control group or the patient group with benign breast disease (p < 0.01). This is in agreement with the work of many other groups (Evanochko et al, 1984; Oberhaensli et al, 1986; Merchant et al, 1988; Sijens et al, 1988). Mean PDE (%) was increased in the breast carcinoma group, although the range of values in each group was quite high. A similar increase in breast carcinoma PDE (%) was identified by Sijens et al (1988) and probably represents a rapid turnover in phospholipids with a simultaneous formation of new cells. Reduced Pi:PDE (p = 0.10) and increased PDE:PCr ratios were observed in the group of patients with breast carcinoma, possibly owing to increased PDE (%). The statistical significance of these findings should improve with increased patient numbers. Expressing phosphorus-31 metabolite peak areas as ratios minimizes the effect of different levels of spectral noise, and therefore enables a more accurate comparison between spectra. Greater numbers are required to determine if elevated PDE (%) is a useful indicator of malignant breast disease. y-NTP (%) was found to be elevated in breast carcinoma tissue relative to the other two groups (p = 0.01) 215
O. M. Redmond, J. P. Stack, N. G. O'Connor, M. B. Codd and J. T. Ennis
as was a-NTP (%) (p < 0.01). This suggests that levels of nucleotide diphosphates, most probably ADP, which contribute to the phosphorus-31 spectra in the same region as the y-NTP and a-NTP peaks, are elevated in breast carcinoma tissue. Mean pH values for the carcinoma group were higher than those observed in either the volunteer group or the benign group, although this difference was not statistically significant. This would agree, however, with other work demonstrating the alkaline nature of breast carcinomas (Oberhaensli et al, 1986). It suggests that these carcinomas are well vascularized and/or oxygenated and, while a small number of dispersed hypoxic cells could easily escape detection by in vivo MRS, that they were not the predominant population in the tumours studied. (The presence of even a small number of hypoxic cells, however, could be highly significant in determining the success or failure of treatment, particularly with radiotherapy.)
by in vivo 3I P-NMR in athymic mice. Biochemical and Biophysical Research Communications, 109, 1346-1352. EVANOCHKO, W. T., N G , T. C , LILLY, M. B., LAWSON, A. J., CORBETT, T. H., DURRANT, J. R., & GLICKSON, J. D., 1983.
In vivo 3I P study of the metabolism of murine mammary 16/C adenocarcinoma and its response to chemotherapy, x-radiation and hyperthermia. Proceedings of the National Academy of Sciences of the United States of America, 80, 334-338. EVANOCHKO, W. T., N G , T. C. & GLICKSON, J. D., 1984.
Applications of in vivo NMR spectroscopy to cancer. Magnetic Resonance in Medicine, 1, 508-534. HOLLANDER, J., 1986. In vivo tissue analysis by NMR spectroscopy. Diagnostic Imaging Clinical Medicine, 55, 9-19. IRVING, M. G., SIMPSON, S. J., FIELD, J. & DODDRELL, D. M.,
1985. Use of high resolution 3lP-labelled topical magnetic resonance spectroscopy to monitor in vivo tumour metabolism in rats. Cancer Research, 45, 481-486. KELSEY, J. L., 1979. A review of the epidemiology of human breast cancer. Epidemiologic Reviews, 1, 74-109. LOVE, S. M., GELLMAN, R. S. & SILEN, W., 1982. Fibrocystic
Conclusion
This study has shown that is is possible, using phosphorus-31 MRS, to detect differences in breast tissue biochemistry in a group of control subjects on the basis of age and menopausal status. Likewise, this non-invasive in vivo technique identified abnormalities in phosphorus metabolism in carcinomatous breast tissue. Magnetic resonance spectroscopy may, therefore, improve the differentiation of benign but symptomatic tumours, such as fibroadenomas, from malignant disease. Magnetic resonance spectroscopy also has potential in monitoring response to therapy. Of the nine patients with malignant breast disease evaluated in this study, five had follow-up MRS examinations post-chemotherapy (mean 14 per patient). Spectral changes were monitored and are currently being evaluated with respect to patient response. Acknowledgments We would like to thank the Irish Health Research Board for partial funding of this project.
"disease" of the breast—a non-disease? New England Journal of Medicine, 307, 1010-1014. MERCHANT, T. E., GIERKE, L. W., MENESES, P. & GLONEK, T.,
1988. 31P magnetic resonance spectroscopic profiles of neoplastic human breast tissues. Cancer Research, 48, 5112-5118. MILLER, W. R., 1986. Steroid metabolism in breast cancer. Reviews on Endocrine-Related Cancer, 23, 23-30. MOON, R. B. & RICHARDS, J. H., 1973. Determination of
intracellular pH by 3I P magnetic resonance. Journal of Biological Chemistry, 248, 7276-7278. MOUNTFORD, C , MAY, G., WILLIAMS, P., TATTERSALL, M., RUSSELL, P., SAUNDERS, J., HOLMES, K., FOX, R., BARR, J. &
SMITH, I., 1986. Classification of human tumours by highresolution magnetic resonance spectroscopy. Lancet, 1, 641-653. N G , T. C , EVANOCHKO, W. T., HIRAMOTO, R. N., GHANTA, V. K., LILLY, M. B., LAWSON, A. J., CORBETT, T. H.,
DURRANT, J. R. & GLICKSON, J. D., 1982. "P-NMR spectro-
scopy of in vivo tumours. Journal of Magnetic Resonance, 49, 271-296. OBERHAENSLI, R. D., HILTON-JONES, D., BORE, P. J., HANDS, L. J., RAMPLING, R. P. & RADDA, G. K., 1986. Biochemical
investigation of human tumours in vivo with phosphorus-31 magnetic resonance spectroscopy. Lancet, 11, 8-11. OTTMAN, R., PIKE, M. C , KING, M. & HENDERSON, B. E.,
References ACKERMAN, J., GROVE, T., WONG, G., GADIAN, D. & RADDA, 3I
G.K. 1980. Mapping of metabolites in whole animals by P NMR using surface coils. Nature, 283, 167-170. COHEN, J. S., LYON, R. C , CHEN, C , FAUSTINO, P. J., BATIST, G., SHOEMAKER, M., RUBALCABA, E. & COWAN, K. H., 1986.
Differences in phosphate metabolite levels in drug-sensitive and -resistant human breast cancer cell lines determined by P-31 magnetic resonance spectroscopy. Cancer Research, 46, 4087-4090. COZZANE, P. J. & JARDETZKY, O., 1976. Phosphorus-31 Fourier
transform nuclear magnetic resonance study of mononucleotides and dinucleotides: 1. Chemical shifts. Biochemistry, 15, 4853-4859. DEGANI, H., HOROWITZ, A. & ITZCHAK, Y.,
1986. Breast
tumours: evaluation with P-31 MR spectroscopy. Radiology, 161, 53-56. EVANOCHKO, W. T., N G , T. C , GLICKSON, J. D., DURRANT,
J. R. & CORBETT, T. H., 1982. Human tumours as examined
216
1983. Practical guide for estimating risk for familial breast cancer. Lancet, 2, 556-558. REDMOND, O., STACK, J. P., SCULLY, M., DERVAN, P., CARNEY,
D. & ENNIS, J. T., 1988. Tissue characterisation and spectral analysis of malignant tumours following chemotherapy. Tumor Diagnostik und Therapie, 9, 167-168. SlJENS, P. E., WlJRDEMAN, H. K., VERMEULEN, J. W. A H. & LUYTEN, P. R., 1987. NMR spectroscopy of human breast tumours before and after radiotherapy using a 1.5 T MRI system. In Proceedings of the Society for Magnetic Resonance in Medicine, Sixth Annual Meeting and Exhibition, New York, Vol. 2, p. 979. SlJENS, P. E., WlJRDEMAN, H. K., MOERLAND, M. A., BARKER, C. J. G., VERMEULEN, J. W. A. H. & LUYTEN, P.R., 1988. Human breast cancer in vivo: H-l and P-31 MR spectroscopy at 1.5 T. Radiology, 169, 615-620. VIHKO,
R. & APTER,
D., 1986. Hormonal
predisposition to breast cancer. Endocrine-Related Cancer, 23, 11-15.
profile and
Reviews
on
The British Journal of Radiology, March 1991
