Annals of Oncology 14: 152–158, 2003 DOI: 10.1093/annonc/mdg007

Original article

Detection of malignant bone marrow involvement with dynamic contrast-enhanced magnetic resonance imaging L. A. Moulopoulos1*, T. G. Maris3, N. Papanikolaou3, G. Panagi4, L. Vlahos1 & M. A. Dimopoulos2 Departments of 1Radiology and 2Clinical Therapeutics, Medical School, University of Athens, Athens; 3Department of Medical Physics, University Hospital of Heraklion, Heraklion, Crete; 4Department of Radiology, General Hospital of Chios, Chios, Greece Received 21 March 2002; revised 5 June 2002; accepted 17 July 2002

Background: The purpose of this study was to evaluate the role of dynamic contrast-enhanced magnetic resonance imaging (dMRI) in detecting bone marrow involvement in cancer patients.

Patients and methods: We studied 50 consecutive patients with histologically confirmed malignant dissemination to the bone marrow, using dMRI of the lumbosacral spine. Time–signal intensity curves were generated from regions of interest (ROIs) obtained from areas of obvious bone marrow disease (group B). In 16 patients from group B with focal disease, ROIs were also placed on areas with apparently normal bone marrow on static magnetic resonance images (group C). Twenty-two patients with no history of malignancy were used as a control group (group A). Wash-in (WIN) and wash-out (WOUT) rates, time to peak (TTPK), time to maximum slope (TMSP) values and WIN/TMSP ratios were calculated for each patient. Mean values for the three groups were compared statistically. Six patients from group B had follow-up dMRI after chemotherapy: four patients achieved a clinical partial response and two had resistant disease. Results: A significant difference was found between groups A and B for all values. Between groups A and C, in spite of the similar static MRI appearance, all values were significantly different. Between groups B and C, a significant difference was found for WIN, WOUT rates and WIN/TMSP ratio. Follow-up dMRI data analysis correlated well with clinical staging. Conclusions: dMRI can distinguish normal from malignant bone marrow. It may identify malignant bone marrow infiltration in patients with negative static MRI and serve as both a diagnostic and prognostic tool for patients with bone marrow malignancies. Key words: bone marrow, dynamic magnetic resonance imaging, magnetic resonance imaging

Introduction Magnetic resonance imaging (MRI) has contributed greatly to the detection and assessment of bone marrow malignancies [1, 2]. A conventional MRI study of the bone marrow includes relatively T1- and T2-weighted images, and is usually sufficient to establish a diagnosis of malignancy in the presence of abundant fatty marrow. Because T1 and T2 values of some tumors approximate those of hematopoietic bone marrow, it may be difficult to detect tumor when red marrow predominates in the skeleton. The absence of a single predictable pattern of red to yellow marrow conversion in the spine complicates the search for malignant bone marrow lesions further [3]. The institution of chemotherapy and, in particular, the concomitant use of growth factors leads to an increase in the volume of red marrow. In addition to red marrow hyperplasia, other disorders of the bone marrow that can complicate the course of disease in a cancer patient (fibrosis, infarction, edema related to recent compression fractures, infection) may also simulate tumor.

*Correspondence to: Dr Lia A. Moulopoulos, 4 Ivis Street, Ekali, Athens 14565, Greece. Tel: +30210-6228771; Fax: +30210-8131383; E-mail: [email protected] © 2003 European Society for Medical Oncology

Dynamic contrast-enhanced magnetic resonance imaging (dMRI) studies the kinetics of the distribution of paramagnetic contrast in the microvessels and in the interstitial space of the tissues being studied. dMRI has been applied to the study of musculoskeletal tumors, with encouraging results regarding the diagnosis of malignant from benign and viable from non-viable tumor, and the prediction and assessment of response to chemotherapy [4–9]. We undertook this study to investigate the potential of dMRI to separate malignant bone marrow disease from uninvolved bone marrow, and its ability to detect underlying disease in patients with known bone marrow malignancies and normal conventional MRI studies.

Materials and methods Patients Fifty consecutive patients with pathologically confirmed malignant disease of the bone marrow were enrolled in this study (group B; Table 1). Patients’ ages ranged from 24 to 80 years (mean 62 years), and the group comprised 22 men and 28 women. Twenty-two consecutive patients (14 men, 8 women), who were referred for evaluation of degenerative disc disease and who had no known history of malignancy, served as a control group (group A). Patients’ ages ranged from

153 Table 1. Primary malignancies of patients with abnormal bone marrow (group B) Primary malignancy

Patient No.

Myelodysplastic syndrome

17

Multiple myeloma

14

Breast cancer

6

Chronic lymphocytic leukemia

3

Lung cancer

3

Adenocarcinoma of unknown primary

2

Carcinoma of the cervix

2

Waldenstrom macroglobulinemia

1

Bladder cancer

1

Lymphoma

1

29 to 72 years (mean 60 years). All patients underwent dMRI studies of the bone marrow. MRI studies for group B were performed prior to initiation of treatment. Six patients (one patient each with breast cancer, lung cancer, cervical cancer and myelodysplastic syndrome, and two with multiple myeloma) had follow-up dMRI studies within a month from completion of therapy. Oral consent was obtained from all patients.

MR imaging Magnetic resonance images were obtained with a 1.5 T unit (Philips Medical Systems, Eindhoven, The Netherlands). Sequences for static MRI included sagital T1-weighted SE [repetition time (TR) = 500–600 ms, echo time (TE) = 10–20 ms] and turbo short-time inversion recovery [STIR; TR = 1620 ms, inversion time (TI) = 180 ms, TE = 70 ms, turbo spin echo factor = 12) images of the lumbosacral spine. Imaging parameters were: 4 mm section thickness, 1 mm interslice gap, 204 × 256 imaging matrix, 2–3 signals averaged. dMRI

was performed with a T1-weighted gradient-echo sequence (TR = 11 ms, TE = 4.2 ms, flip angle = 30°). The dynamic study was limited to a single spinal part to avoid inherent variations in perfusion between different spinal locations. The lumbosacral spine was selected for dMRI over other skeletal parts to match the patients’ studies with those of the control group. Five sections through the spine, carefully selected from the static study to encompass areas of abnormal marrow, were obtained every 18 s for a total of a maximum of 3 min. A bolus of gadopentetate dimeglumine 0.1 mmol/kg body weight (Magnevist; Schering, Berlin, Germany) was injected manually immediately after the end of the first dynamic acquisition. After completion of the MRI study, all contrast-enhanced dynamic images were subtracted from the first set of unenhanced images by using the subtraction function of the magnetic resonance (MR) unit. For each normal control of group A, a region of interest (ROI) was positioned on one of the lumbar vertebral bodies, avoiding the areas affected by motion artifacts. For each patient of group B, subtraction images were reviewed and an ROI was placed on a focus of maximum abnormal enhancement. Foci of abnormal enhancement that could be related to vertebral endplate perfusion changes that accompany degenerative disc disease were excluded from the study. All ROIs measured between 70 and 80 pixels, and were carefully selected to avoid the basivertebral vessels that course at the midline of each vertebra. ROIs were selected on a single image and were then plotted automatically on all images of the same dynamic series. In 16 patients from group B in whom the neoplastic process did not appear to involve the entire bone marrow on the static MR images, a second ROI was positioned at a site of apparently normal marrow (group C). For the six patients from group B who had follow-up dMRI studies, ROIs were positioned on the same vertebra as during the pretreatment study.

Data analysis The dynamic signal enhancement (DSE) of selected ROIs was plotted as a function of time (t) and signal intensity–time curves were generated (Figures 1, 2 and 3). Signal intensity–time data obtained from each ROI were fitted by

Figure 1. Mean DSE and mean first derivative d(DSE)/dt curves. Calculated fit parameters were: a = 70.70, b = 74.37, c = 8.510, d = 65.03; r = 0.9963; Fstat = 228. WIN, WOUT, TTPK and TMSP parameters were calculated from the d(DSE)/dt curve and their numerical values are: WIN = 2.19 s–1, WOUT = –0.120 s–1, TTPK = 74.37 s and TMSP = 38.46 s. WIN/TMSP = 0.057 s–2.

154

Figure 2. Mean DSE and mean first derivative d(DSE)/dt curves. Calculated fit parameters were: a = 365.83, b = 53.86, c = 4.793, d = 96.63; r = 0.9998; Fstat = 4937. WIN, WOUT, TTPK and TMSP parameters can be easily calculated from the d(DSE)/dt curve and are presented on the graph. Their numerical values are: WIN = 19.86 s–1, WOUT = –0.753 s–1, TTPK = 53.86 s and TMSP = 31.80 s. WIN/TMSP = 0.624 s–2.

Figure 3. Mean DSE and mean first derivative d(DSE)/dt curves. Calculated fit parameters were: a = 115.78, b = 55.78, c = 6.022, d = 49.31; r = 0.9961; Fstat = 211. WIN, WOUT, TTPK and TMSP parameters were calculated from the d(DSE)/dt curve and their numerical values are: WIN = 5.13 s–1, WOUT = –0.360 s–1, TTPK = 55.91 s and TMSP = 31.97 s. WIN/TMSP = 0.160 s–2.

means of a Marquardt linear regression analysis method [10] according to the formula: a DSE = ---  1 + e b

1 + d ln ( d ) – b ----------------------------------- c

t + d ln ( d ) – b

---------------------------------- c d+1  d+1  ( d + 1 ) ----------- –  ------------  e d  d  

Parameters (a, b, c, d) were calculated from the fit of DSE to t.

(1)

For all fits, r was ≥0.9. This value was used as a threshold for the estimation of goodness of each fit. Wash-in (WIN) and wash-out (WOUT) rates were calculated for each patient from the maximum and minimum slopes of signal intensity–time curves, respectively. These slopes, as well as time to peak (TTPK) and time to maximum slope (TMSP) values were obtained from the first derivative function f(t) = d(DSE)/dt of equation (1). The ratio WIN/TMSP was also calculated.

155 Table 2. Measured mean and median WIN, WOUT, TTPK, TMSP values and WIN/TMSP ratios in three subject sample groups Calculated parameters

Group A

Group B

Group C

Mean WIN ± SEM

3.09 ± 0.48

37.96 ± 6.47

9.56 ± 2.12

Median WIN

2.48

22.36

6.52

Mean WOUT ± SEM

–0.18 ± 0.03

–1.20 ± 0.22

–0.41 ± 0.07

Median WOUT

–0.15

–0.67

–0.36

Mean TTPK ± SEM

75.87 ± 5.72

58.88 ± 3.47

50.74 ± 3.05

Median TTPK

67.42

51.33

53.24

Mean TMSP ± SEM

43.17 ± 2.24

33.52 ± 1.13

32.13 ± 2.11

Median TMSP

42.33

32.66

33.09

Mean WIN/TMSP ratio

0.080 ± 0.013

1.235 ± 0.214

0.350 ± 0.081

Median WIN/TMSP ratio

0.056

0.681

0.187

WIN, wash-in rate; SEM, standard error of the mean; WOUT, wash-out rate; TTPK, time to peak; TMSP, time to maximum slope values. Group A, normal subjects; group B, abnormal bone marrow; group C, normal-appearing bone marrow.

Patients and normal controls were divided into three sample groups for statistical evaluation of the results: (i) group A (normal controls); (ii) group B (patients with abnormal bone marrow); and (iii) group C (patients from group B with areas of normal-appearing bone marrow on static MR images). Assuming the possibility of difference in standard deviations amongst compared populations, an unpaired t-test with Welch correction was used to check statistical significance amongst all groups (sample means) using WIN, WOUT, TTPK, TMSP values and the ratio of WIN/TMSP. A P value of 0.05 was considered a statistically significant threshold.

Results Mean and median values of calculated parameters WIN, WOUT, TTPK, TMSP and WIN/TMSP for the three study groups (A, B and C) are presented in Table 2. Statistical results (P values) are shown in Table 3. Graphical presentations of mean DSE and mean first derivative d(DSE)/dt curves were produced from the relevant patient samples for all three study groups (Figures 1–3). Normal compared with abnormal bone marrow (groups A and B). Mean WIN and WOUT rates and WIN/TMSP ratios were significantly lower for group A compared with those of group B. Mean TTPK and TMSP values were significantly higher for

normal subjects (group A) than for patients with abnormal bone marrow (group B). Abnormal compared with normal-appearing bone marrow (groups B and C). Mean WIN and WOUT rates and WIN/TMSP ratios were significantly higher for the group with abnormal bone marrow (group B) compared with those of the group with apparently normal bone marrow on static MR images (group C). Mean TTPK and TMSP values for groups B and C did not differ significantly. Normal compared with normal-appearing bone marrow (groups A and C). Mean WIN rates and WIN/TMSP ratios were significantly shorter for group A compared with those of group C. Mean TTPK and TMSP values were significantly higher for group A when compared with group B.

Follow-up MRI Four of six patients with malignant bone marrow disease who had follow-up MRI achieved a clinical partial response to treatment, while in two patients the disease progressed. dMRI diagnosis was consistent with the clinical evaluation in all patients (Tables 4 and 5; Figure 4).

Table 3. Statistical results (P values) using unpaired t-test with Welch correction for differences between mean WIN, WOUT, TTPK and TMSP values, and WIN/TMSP ratios amongst sample groups Mean sample group difference

WIN

WOUT

TTPK

TMSP

WIN/TMSP ratio

(Group A) – (group B)