Tumor infiltration of bone marrow in patients with hemato-logical malignancies: dynamic contrast-enhanced magnetic resonance imaging

Abstract

Background 

Conventional magnetic resonance (MR) scanning techniques can identify bone marrow (BM) containing mostly fat cells. But they are not able to differentiate BM tumor infiltration, BM fibrosis and normal red BM. This is particularly problematic in assessment of recurrent or refractory hematological malignancy. This pilot study used dynamic contrast-enhanced MR imaging (DCE-MRI) to evaluate the bone marrow status and to determine whether several calculated parameters derived from the DCE-MRI correlate with histological characteristics of marrow, especially with the tumor fraction (TF).

Methods 

DCE-MRI scans were performed in 25 patients with proven or known hematological malignancy who were about to undergo bone marrow biopsy of the posterior iliac crest. The location chosen for biopsy was examined with MRI approximately one hour prior to the biopsy. Time-signal intensity curves (TIC) were generated from the region of the iliac crest corresponding to the planned biopsy site. Enhancement parameters were calculated, including peak enhancement ratio (PER), maximum enhancement slope (Slope_max), time to peak (TTP) and mean time (MT). The biopsy specimen was reported synoptically, with relevant reported parameters including cellularity and tumor fraction (TF).

Results 

PER values were significantly higher for the bone marrow tumor infiltration group than for the normal bone marrow group (P<0.05). A significant positive correlation was found between PER and TF as well as Slope_max and TF. A negative correlation was found between TTP and TF. There was no significant difference in the mean TTP and MT values between the BM tumor infiltration group and the normal bone marrow group.

Conclusions 

The presence of diffuse bone marrow infiltration in patients with haematological malignancies could be verified using DCE-MRI.

Normal bone marrow (BM) contains predominantly fat cells (fatty marrow), hemopoietic cells (red marrow) or an intermediate combination of them. Depending on physiological (such as age) and pathological variables (such as prior irradiation, current chemotherapy, coexistent anemia or systemic illness), the proportion of non-fat cells comprising BM varies widely between 0 and 100% by volume, and is described as ‘cellularity’ in BM histopathology reports.¹ Replacement of normal hemopoietic BM by tumor cells varies from inconspicuous admixture of scattered tumor cells to solely tumor cells instead of non-fat BM cells. This is clearly a continuous variable between 0 and 100%, and is described as ‘tumor fraction’ in BM histopathology reports.

Currently the gold standard and the only reliable method of assessment of diffusely abnormal BM is BM biopsy (trephine, aspirate or a combination of them). Iliac crest biopsy is routinely performed as part of clinical work-up of patients with malignant hematology diseases.

The magnetic resonance (MR) appearance of normal BM reflects the variable amounts of its physiological components, primarily fat cells and hemopoietic cells. Conventional MR scanning techniques, including T₁weighted, T₂-weighted and fat saturated imaging, can identify BM that contains mostly fat cells. Unfortunately, conventional MR sequences are not often able to differentiate BM tumor infiltration, BM fibrosis and normal red BM. This is particularly problematic in assessment of recurrent or refractory hematological malignancy.

Dynamic contrast material-enhanced MR imaging, which depicts the physiologic features of the microcirculation, has been successfully used in the treatment management of patients with solid malignancies such as breast cancer and osteosarcoma.^2,3 Bollow et al⁴ used dynamic MR imaging findings for comparison of normal and malignant BM infiltrations. Thus dynamic MR imaging has been proved to be an effective and noninvasive method for evaluation of in vivo blood perfusion of bone, marrow, and tumors.

The purpose of this study was to evaluate dynamic contrast enhancement of the BM in patients in the recurrence or remission period with malignant hematology diseases after the chemotherapy, to determine whether several calculated parameters derived from the dynamic contrast-enhanced MR (DCE-MR) correlating with the marrow's histological characteristics, especially with the tumor fraction, in order to obviate times of BM biopsies done in the long therapy period for this indication.

METHODS

Patients

From October 2004 to July 2005, 25 patients referred to Peter MacCallum Cancer Center for BM biopsy were enrolled in this study. The inclusion criteria were: referral for iliac crest biopsy from either the left or right posterior superior iliac crest where hematological malignancy was suspected, newly diagnosed or previous histopathologically proven, irrespective of treatment status. The exclusion criteria were prior irradiation of the site to be biopsied, pregnancy and MR contraindications (cardiac pacemaker, cochlear implant, neurostimulator, intraocular metal, intracranial ferromagnetic aneurysmal clip). Failure of DCE-MR for technical reasons (injector failure) occurred in further processing patients who had not been included in the data analysis.

There were 19 men and 6 women with age ranged from 24 to 76 years (mean age, 54.9 years). Seven of the 25 subjects had non Hodgkin lymphoma (NHL), and there were 9 patients with leukemia and 9 patients with myeloma. The study protocol was conducted after the approval by the institutional review board and informed consent of all subjects had been obtained.

MR imaging

MRI examination of the area to be biopsied was done approximately one hour prior to the BM biopsy. MR imaging of the iliac crest was performed by using a 1.5-Tesla superconducting system (Signa, General Electric Medical Systems, Milwaukee, USA) and a phased-array soft coil. A routine T₁-weighted [repetition time (TR)/echo time (TE) 500/8.2 ms, echo train length (ETL) 4] and a T₂-weighted with fat saturated (T₂W+Fatsat) fast spin-echo sequences (TR/TE 4000/80 ms, ETL 16) were performed in the coronal plane with the following parameters: 256× 256 matrix, 32 cm field of view, and 8-mm-thickness sections.

Ultrafast T₁-weighted gradient-echo sequence was used in the contrast enhanced dynamic MR pulse sequence, with TR=5.6 ms, TE=1.4 ms, flip angle=45°, acquisition matrix 256×160, and 10-mm section thickness. In total, 200 dynamic images were obtained within 200 seconds (0.97 second per frame) in each of the patients. A bolus of gadopentetate dimeglumine (Magnevist, Schering, Berlin, Germany; 0.1 mmol/kg body weight) was rapidly injected at the rate of approximately 3 ml/s through a 21-gauge intravenous catheter previously inserted in the right antecubital vein. This injection was immediately followed by a 20-ml saline flush at the same injection rate. The dynamic imaging was initiated 5 seconds before the injection of the contrast medium commenced.

Qualitative analysis

Visual comparison of T₁W and T₂W signal intensity of the volume of interest was related to adjacent gluteal muscle. Each signal intensity was coded high or low and homogeneous or heterogeneous. Signal intensity (SI) measurements were made using circular regions of interest (ROIs) which were placed on the posterior-superior iliac crest corresponding to the planned biopsy site (Fig.1). The time-intensity curve (TIC) of the dynamic contrast-enhanced sequence was qualitatively visually graded against five categorical time-intensity patterns based on Chen et al.⁵

Fig. 1.: The raw image of the DCE-MRI (A), the square showing the SI change with time analysed through the MIStar software. The Slope_max image added on the raw image of the DCE-MR (B), the red indicating the faster contrast uptake. Time-intensity curve of the ROI (C).

Quantitative analysis

Semi-quantitative and quantitative analysis were performed with a bicompartmental model proposed by Tofts et al,⁶ including peak enhancement ratio [PER, PER= (SI_P—SI₀)/ SI₀), maximum enhancement slope (Slope_max, Slope_max = max {ΔS/Δt})], time to peak enhancement (TTP) and mean time [MT, MT = /.tC(t)d(t)//.C(t)d(t)]. Although the time-intensity curve could have either an equilibrium phase or a slowly rising second part, the values for PER and Slope_maxwere obtained from nonlinear fitting of the individual time-intensity curve using MIStar software (Apollo Medical Imaging Technology, Melbourne, Australia). The baseline value for signal intensity (SI₀) on a time-intensity curve was defined as the mean signal intensity of the first five images. Then the maximum signal intensity (SI_p) was defined as the maximum value of the first rapidly rising part of the time-intensity curve.

BM biopsy

The biopsy was performed by the hematologist and comprised an 11 gauge BM trephine (yielding a core of between 5 and 40 mm long) and a BM aspirate. Tumor fraction (TF) was reported as the fractional volume of the non-fat cells occupied by tumor cells (expressed as %). In our trial, TF was given a more rigorous numerical equivalent in order to standardize. The BM involvement was graded according to the TF in the BM sample: Grade 0 indicated as less than 5% of the non-fat cells occupied by tumor cells; grade 1, 5%-25%; grade 2, 25%-75%; grade 3, more than 75%. Grades 0 and 1 were regarded to represent none or mild tumor infiltration, being a complete or almost complete remission. Grades 2 and 3 were regarded to be intermediate and extensive tumor infiltration respectively, and considered to represent no remission or recurrence following therapy.

Statistical analysis

Statistical analysis was performed using SPSS 12.0 software package (SPSS Inc., Chicago, USA). Descriptive statistical data were expressed as mean ±standard deviation. The Mann-Whitney U test was performed to analyze the statistical significance observed for the small sample. Correlation between variables was assessed by the Spearman's coefficient (r). Two-sided P values of 0.05 or less was considered statistically significant.

RESULTS

Twenty-five patients were grouped into 2 groups, 16 in group A with a complete and almost complete remission BM grades (grades 0 and 1), and 9 patients in group B with non-remission tumor infiltration (grades 2 and 3) (Table 1). Among 16 remission patients, the crests SI of 5 cases were higher on both T₁W and T₂W sequence, 2 were lower on both T₁W and T₂W, and 9 were higher on T₁W and lower on T₂W with fat saturated imaging than SI of the gluteal muscles. Among 9 recurrence cases, the crest SI of 3 cases were higher on both T₁W and T₂W, 4 were lower on both T₁W and T₂W, and 2 were lower on T₁W and higher on T₂W with fat saturated imaging than SI of the gluteal muscles. The SI which was high on both T₁W and T₂W with fat saturation was defined as suspicious. The diagnostic false positive error rate was 62.5% (5/8) with conventional MR. The SI which was low on both T₁W and T₂W with fat saturation was defined as red marrow or almost normal by radiologist. The diagnostic false negative error rate was 66.7% (4/6) with conventional MR. Conventional MR strongly indicated tumor remission in BM with hematology diseases, when SI of BM on T₁W was higher than that of the adjacent gluteal muscle.

Table 1: SI features on T₁-WI and T₂WI with fat saturated in 25 patients with hematological diseases

Three patients with heterogeneous MR signal were reported by the radiologist as suspicious for tumor infiltration (Table 1). TF ≥ 25% in 2 patients confirmed tumor infiltration, but TF<25% in 1 was a false positive (Fig. 2). BM sections of these 3 patients were stained with recticulin, the latter one with severe BM fibrosis, but another 2 with no fibrosis. Heterogeneous SI on BM did not indicate existence or absence of tumor infiltration, especially after chemotherapy. Heterogeneous SI on BM may be caused by myelofibrosis.

Fig. 2.: Coronal imaging: the SI of the BM was homogeneous, high on T₁WI (A) and low on T₂WI with fat saturation (B). No evidence on conventional MR existed for residual tumor. PER after the contrast injection with DCE-MR (C) showed heterogeneity of the contrast distribution.

On comparison between groups A and B of the contrast enhancement parameters, the mean PER value in group A was 0.252±0.156, lower than that in group B, 0.592±0.433. The Slope_max was 0.204± 0.105, lower than that in group B, 0.561±0.634. There was significant difference (P=0.025, Mann-Whitney U test). The value of the PER was positively correlated with the increase of the TF (r=0.457, P<0.05). Slope_max was also correlated with the TF, but there was no significant difference (P>0.05) (Tables 2 and 3 ).

Table 2: Bone marrow TIC values for patients with tumor remission and recurrence (mean±SD)

Table 3: Mann-Whitney U test (Group A: TF<25%, n=16, group B: TF≥25%, n=9)

On comparison between groups A and B of the contrast uptake parameters, the mean TTP value in group A was 81.52 ± 38.49, higher than that in group B, 68.13±42.18. The mean MT value in group A was 83.70 ± 21.02, lower than that in group B, 86.94 ± 17.09. A negative correlation was found between TTP and TF (r=—0.161). But there was no significant difference between the mean TTP values and TF, MT values and TF for the BM tumor infiltration group and the normal BM group (P>0.05) (Tables 2 and 3).

DISCUSSION

In patients with hematologic diseases, MRI-based identification of focal lesions in axial BM is based on well-defined contours and on lower signal intensity on T₁-weighted images and higher signal intensity on T₂-weighted images, as compared with the signal intensity of the adjacent, presumed normal BM.^7,8

It was shown in our study that SI of BM on T₁W is lower than that of the adjacent gluteal muscle indicated presence of tumor infiltration in BM with hematology diseases. Furthermore, dynamic contrast-enhanced MR imaging can depict the contrast enhancement of BM in patients with BM involvement by hematology diseases despite normal signal intensity on T₁-weighted MR images obtained in these patients. Contrast enhancement parameters, such as PER and the Slope_max, were different in the patients with remission or near remission and patients with recurrence.

Previous studies have compared conventional MR sequences directly against posterior iliac crest biopsy in the diagnosis of malignant BM infiltration. In early stage during BM infiltration, tumor cells fail to displace BM fat cells, the amount of which remains normal.⁹ Subsequently, diffuse replacement of normal BM by tumor cells leads to a decrease in signal intensity on T₁-weighted MR imaging.^10,11

In our trial, the SI on T₁W in 6 of 9 patients with recurrence was low. In 16 patients with remission, the SI on T₁W of 14 patients was high. High SI on T₂W with fat saturation was also seen in the BM with tumor remission. Conventional MR strongly indicated tumor remission in BM with hematology diseases, when the SI of BM on T₁W is higher than that of the adjacent gluteal muscle. Two of 3 patients in our trial with heterogeneous MR signal had tumor infiltration with TF≥25%, but one had normal BM with TF<25%. It is possible that heterogeneous SI on BM does not indicate existence or absence of tumor infiltration, especially after chemotherapy. The BM sections were stained with recticulin, the latter one with severe BM fibrosis, but the 2 with no fibrosis. Heterogeneous SI on BM may be caused by former myelofibrosis. Further study is recommended to clarify this issue.

Comparing normal, normal-appearing with abnormal BM in cancer patients, Moulopoulos et al¹² found follow-up dynamic MRI data analysis correlated well with clinical staging. Between normal and normal-appearing groups, in spite of the similar conventional MRI appearance, wash-in and wash-out rates, time to peak were significantly different. The difference of micro-environment states could be seen among the normal, normal-appearing and abnormal bone marrow whether the subjects had hematological diseases or not.

In our trial, PER correlated well with the increase of the TF (r=0.457), after comparing remission and recurrence group using contrast enhancement parameters. PER value in remission group was significantly lower than that in the recurrence group (P<0.05). But the Slope_max, TTP and MT had no significant difference (P>0.05). Previous studies yielded some different results.^13,14 Rahmouni¹⁵ found significant differences of E_max (P <0.001), slope (P<0.001), and washout (P<0.005) between subjects with normal BM and with diffuse BM involvement among patients with lymphoproliferative diseases. In our trial, only the PER value was significantly different in remission group compared to that in the recurrence group. An explanation could be that all patients definitely had hematology diseases in our trial, such as leukemia, multiple myeloma and NHL, and all patients had recieved chemotherapy prior to MR imaging. Therefore, their BM status was different from those of the previous study groups,¹⁵ which was compared the BM of hematopoietic malignancies with the normal control BM.

Moehler et al¹³ calculated pharmacokinetic parameters and quantitative microcirculation parameters according to a 2-compartment model by DCE-MRI as markers of angiogenesis in 16 normal controls and 41 patients with active multiple myeloma. They found there was significantly higher contrast uptake and distribution in MM patients (P=0.001). In our trial, we also found the PER value in recurrence group was significantly higher than that in remission group (P<0.05). We presume the contrast enhancement features of BM in hematology patients at different post-chemotherapy stages were caused by the microvessel density change resulting from the chemotherapy.

Increased microvessel density was observed in the BM of patients with hematology disease in previous reports. Vacca¹⁴ found a significant increase of BM angiogenesis in patients with active MM compared with the nonactive MM. Pardo et al¹⁶ reported that microvessel counts were significantly higher in patients with AML compared with control patients. On day 16 of induction chemotherapy, microvessel density was reduced by 60% (P<0.001) in hypoplastic marrows without residual blasts, in contrast to only 17% reduction in hypoplastic marrows with ≥5% residual blasts (P<0.001). BM biopsies taken at the time of complete remission displayed a microvessel density in the same range as the controls. Baur et al¹⁷ reported that increases in SI after contrast administration were significantly different among the patients with low-grade, intermediate-grade and high-grade microvessel density (P< 0.0001).

In our trial, the value of TTP in group A was 81.52± 38.49, higher than that in group B, which was 68.13 ±42.18. MT value in group A was 83.70±21.02, lower than that in group B, which was 86.94±17.09. There was higher contrast uptake and distribution in the recurrence group with hematologic diseases. There was a negative correlation between TTP and tumor infiltration grade. We found that TTP was lower and MT was a little longer in the recurrence group. In Rahmouni's study,¹⁵ the washout values increased with increasing BM involvement grade.

We presume that not only the infiltrating tumor cell amounts affected the transient time changes, but the fat cell amount, the tissue interstitial pressure, microvessel membrane integrity and the microvessel permeability. Further study is needed to confirm these presumptions.

There were some limitations in our trial. First, tumor angiogenesis was not assessed in the specimens. Second, a normal control group was absent and the chemotherapeutic stage was different between the groups. This may result in the variation of the microvessel density of the BM specimens. Third, the limited number of patients within every different hematological class probably had variable tumor angiogenesis.

In conclusion, dynamic contrast enhanced MR may be helpful in the diagnosis and grading of BM involvement in patients with hematological malignancies and for evaluating the treatment response when compared with the conventional MR, especially when using the semiquantitative parameters.

Acknowledgments:

We thank Mr. Brett Frenkiel for co-ordinator of the project. We thank Dr. Con Tartaglia, Dr. Lois Comber, Dr. Lucila Zentner, Dr. Donna D'Souza, and Dr. Tak Yue Kwok for the recruitment of patients. We thank Dr. YANG Ming-shan for his help in the preparation of the manuscript.