PET/CT and bone marrow biopsy (BMB) in evaluating bone marrow in lymphoma

Bone marrow assessment is an important part in the Ann Arbor staging system in lymphoma. It is done routinely through posterior iliac crest bone marrow biopsy (BMB) which is an invasive technique with limited examination of one site. 18F-FDG PET/CT is now used for staging of lymphoma. The purpose of this study was to compare the sensitivity of PET/CT and BMB in detecting bone marrow infiltration (BMI) in lymphoma and determine agreement between both in assessing bone marrow and whether we can evaluate the bone marrow by PET/CT without the need of the routine BMB. PET/CT detected 24 (16.5%) cases with positive BMI that were missed by BMB. BMB detected only 2 (1.4%) cases that were missed by PET/CT. The PET/CT showed a higher sensitivity of 95.6% than BMB 46.7% in detecting BMI in lymphoma. We found a moderate agreement between PET/CT and BMB results in the whole cohort using Cohen’s k computation. It was found that 0.47 with p value less than 0.0001. PET/CT can detect more bone marrow involvement in lymphoma compared with BMB. It can replace the routine invasive BMB in many cases, especially those showing multifocal uptake in both Hodgkin and non-Hodgkin lymphoma. PET/CT can also help to guide the site of the biopsy in some cases. Iliac crest BMB is still needed in cases showing diffuse FDG uptake to differentiate malignant uptake from reactive hyperplasia, and in those with limited FDG avidity and in some cases with negative uptake to exclude early infiltration if management will differ.

Within the last few years, various studies have shown the superiority of PET/CT over BMB in the assessment of bone marrow infiltration in Hodgkin lymphoma (HL) in terms of diagnostic accuracy [9,10]. A number of studies have discussed this issue in the context of non-Hodgkin lymphoma (NHL), particularly in diffuse large B cell lymphoma (DLBCL) with differing results [11][12][13][14].
Until this moment, there are no definite international guidelines for the indication of PET/CT and BMB [15].
The aim of this study was to assess the sensitivity of PET/CT and BMB in detecting bone marrow infiltration (BMI) in lymphoma and determine agreement between both in assessing bone marrow in lymphoma patients and whether we can evaluate the bone marrow by PET/ CT without the need of the routine BMB.
The PET/CT imaging findings and BMB results of lymphoma patients were collected, analyzed, and correlated with each other, and with one or more of the following: follow up changes, local biopsy, targeted MRI, associated CT changes not explained by other benign findings.

Methods
This prospective study was done for 145 patients with a diagnosis of lymphoma and was approved by the Institutional review board. All subtypes of NHL and HL were included. Most of the patients with a new diagnosis of lymphoma underwent bone marrow biopsy as staging criteria. To be included in this study, patients must have had a PET/CT within 2 weeks of the bone marrow biopsy at either initial staging or relapse. Imaging may have been performed either before or after the biopsy, and must have been performed prior to the initiation of any form of therapy. Some of the relapsed cases underwent bone marrow biopsy and PET/CT imaging prior to a new line of therapy were included in this study.

Bone marrow biopsy Preparation
The patient was placed on the procedure table (prone or lateral decubitus for posterior superior iliac crest and the iliac crest exposed). Any topical ointment was removed and skin cleaned. The iliac crest was palpated and the site of aspiration or biopsy determined. The puncture site was cleaned with antiseptic solution using a circular motion from the inner to the outer area; sterile drapes were placed over the biopsy site. Anesthetizing the marrow site with 1 to 2 mL of local anesthetic (1 to 2% lidocaine without adrenaline) was done.

Technique
All BMB were acquired one sided from the posterior iliac crest. The needle was held with the proximal end in the palm and the index finger against the shaft close to the tip. With the stylet secured, the needle is presented through the skin pointing toward the anterior iliac spine. By gentle pressure, the needle was progressed with a slight turning movement until it feels anchored to the bone; the stylet was expelled, then using exchanging clockwise and counterclockwise. The needle was rotated with three turns to one side and afterward to one side without progressing; repeated once again, then the needle pulled back using a rotating movement. Utilizing a sterile gauze pad, manual pressure was applied to the site until the bleeding stops. The bone specimen was expelled from the biopsy needle by introducing a probe through the distal end. The specimen was dropped in formalin and labeled.

Interpretation of BMB
Trephine biopsy samples were then decalcified and stained with hematoxylin and eosin. They were interpreted by an expert hematopathologist for the presence of lymphomatous involvement, without knowledge of the imaging results.
FDG PET/CT acquisition FDG-PET/CT was performed on an integrated PET/CT system with 16-slice CT (GE Discovery 710 PET/CT, AW 4.6, manufactured in Chicago, Illinois, USA). This dedicated system allowed the acquisition of co-registered CT and PET images in one session.

Patient preparation
All patients instructed that they should fast for at least 4-6 h before the study (to maintain low glucose and low insulin levels), but drink water to maintain good hydration.
The fasting blood glucose level was determined. The preferred fasting blood glucose is below 150 mg/dl.

CT imaging protocol
Whole-body CT study (neck, chest, abdomen, and pelvis), typical scanning parameters: 130 KV, 40 mAs, 2.5 mm slice thickness, pitch of 1.5, gantry rotation time of 0.8 second, and field of view of 50 cm.

PET imaging
PET performed on a dedicated PET scanner with approximately six to eight bed positions that planned in the three-dimensional acquisition mode for scanning from the head to the mid-thigh with 1-2 min acquisition at each bed in a cauda-cranial direction. Each bed position is 11.5 cm following acquisition of data at each bed position, there is approximately a 4-cm (25%) overlap between table stations. The maximum length of the patient that can be scanned with the current PET/CT scanner is 180 cm, then PET and CT images were reviewed using a dedicated workstation and software (E.soft; GE medical solutions), which allowed threedimensional displays (trans-axial, coronal, and sagittal) to be constructed using CT, PET, and PET/CT images and maximum intensity projection displays of the PET data. The CT data were used for attenuation correction for the PET data and for lung lesion screening as far as possible. For image construction, the images were constructed using the standard ordered subset expectation maximization (OSEM; two iterations, eight subsets) algorithm. The reconstructed PET image voxel size was 1.7 × 1.7 × 2.4 mm. Scanning started 45-60 min after intravenously administered 3 MBq/kg body weight of 18F-FDG, image acquisition was performed (the radiation effective dose from internally administered 18F-FDG for males was 0.002 mSv/MBq and for females was 0.025 mSv/MBq, and from the CT component was 6.8 mSv in males and 7.9 mSv in females).

Imaging interpretation Qualitative (visual) assessment
PET/CT scans were interpreted as positive for BMI when the bone marrow 18F-FDG uptake was found superior to hepatic reference. The CT images were considered for corresponding CT changes.

Quantitative assessment
The maximum standardized uptake values (SUV max) were determined for the most active bony lesion after manual application of the volumetric regions of interest (ROI) on the trans-axial attenuation-corrected PET slices, around the areas showing the highest accumulation of 18F-FDG and away from any nearby overlapping activity. Another ROI was drawn over the normal liver where its SUVmax was considered a reference activity.

Data analysis
The different data used in this research methodology are demonstrated in Fig. 1.
Cases with concordant findings in both PET/CT and BMB (either positive or negative) were evaluated as true positive or true negative results. Non-concordance between these two parameters was described as false negativity or false positivity as follow: True positive PET/CT results ○ Agreed PET/CT and BMB in the detection of BMI ○ In patients with positive PET/CT and negative BMB results, we depended on one or more of the following: ➢ Regression or progression of lesions on follow-up PET/CT after chemotherapy ➢ Local biopsy or targeted MRI confirming PET/CT findings ➢ Presence of morphologic CT changes consistent with BMI not explained by other benign findings

False-positive PET/CT results
Patients with positive BMI in PET/CT study yet not meeting any of the previous criteria were considered as false positive.

True negative
If both BMB and PET/CT were negative for BMI.

False negative
PET/CT was considered as false negative if BMB was positive and PET/CT was negative for BMI.

Statistical analysis
The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPP), and accuracy with 95% confidence intervals for both PET/CT and BMB were calculated for the whole study, for Hodgkin and non-Hodgkin lymphoma then for the most common histological subtypes (nodular sclerosis, mixed cellularity, and DLBCL) in our study. Agreement between 18F-FDG PET/CT and BMB findings was assessed using Cohen's k computation for the whole study, for Hodgkin and non-Hodgkin lymphoma then for the most common histological subtypes in our study (nodular sclerosis, mixed cellularity, and DLBCL). SPSS version 23 was used in all the calculations.

Patient characteristics
This prospective study was collected from a pathologically proven Hodgkin and non-Hodgkin lymphoma patients, 57 (39.3%) of them had NHL and 88 (60.7%) had HL. The clinic-pathological characteristics are shown in Table 1. In the whole study, the frequency of B symptoms (temperature > 38°C, night sweats, weight loss > 10% of normal body weight within a period of 6 months or less) was noted in 60 patients (41.4%). The different types of HL and NHL in the study are shown in Tables 2 and 3.

Impact of combined PET/CT and BMB findings on staging
Patients were staged according to Ann Arbor staging system. It is shown in Fig. 2.
Taking both techniques (PET/CT and BMB) into account in this study, there were 19 cases diagnosed as stage IV by both PET/CT and BMB by the presence of BMI. Thirteen cases were upstaged by PET/CT from stage III to stage IV, by detecting BMI not detected by BMB. Three cases were upstaged from stage I to stage IV by detecting BMI not detected by BMB. Eight cases were diagnosed as stage IV by PET/CT, by the presence of BMI and other extranodal lesions, with negative BMB results. Two cases were diagnosed as stage IV by BMB only, as no BMI was detected by PET/CT. Five cases were diagnosed as stage IV by PET/CT by the presence of other extranodal lesions, with no bone marrow infiltration.

Analysis of bone marrow involvement
Bone marrow infiltration was detected in 45 (31%) cases. Nineteen (13.10%) cases by both PET/CT and BMB, twenty-four (16.55%) cases by PET/CT only and two (1.3%) cases by BMB only. By dividing the positive cases according to their histological type, there were 17 HL positive cases and 28 NHL positive cases (Table 4).
Bone marrow infiltration detected by PET/CT was classified as either unifocal, bifocal, multifocal (more than two lesions), or diffuse infiltration (Table 5). However, within the diffuse involvement, eight (80%) cases showed positive infiltration in BMB and two (20%) cases were negative in BMB and therefore were considered as false positive in PET/CT (Fig. 3).
Semi-quantitative assessment of BMI in PET/CT by standardized uptake value (SUV) was found ranging from 3.5 to 33.7 with a mean of 11.66 ± 7.3.
Within the 19 cases with positive BMB and PET/CT, 16 of them showed bilateral iliac crest involvement by PET/CT and three cases showed unilateral iliac crest involvement by PET/CT, and biopsy was taken from the affected side.
Within the 26 cases with positive PET/CT and negative BMB, two cases had HL and showed diffuse bone marrow uptake by PET/CT but BMB results were negative for infiltration, so they were considered as false positive. Four cases showed unilateral iliac crest involvement by PET/CT; however, BMB was taken from the contralateral side and 20 cases showed absent posterior iliac   crest involvement by PET/CT on either side, the uptake was present in other regions of the skeleton. The pattern of iliac crest involvement in positive PET/CT is illustrated in Fig. 4.

Diagnostic performance of PET/CT compared with BMB in the whole study
The PET/CT was reported as negative for BMI in 100 cases. Of these 100 patients, two patients were interpreted as positive for BMI because of positive findings on BMB. Therefore, the 18F-FDG PET/CT was considered as truly negative for 98 patients and falsely negative in 2 patients for BMI. The 18F-FDG PET/CT was reported as positive for BMI in 45 patients, two of these patients showed diffuse FDG uptake in PET/CT; however, BMB results were negative for infiltration and revealed reactive hyperplasia, so they were considered as falsely positive in PET/CT and PET/CT was considered as truly positive for 43 cases. The BMB results were positive in 21 patients. PET/CT detected BMI in 24 (16.5%) additional cases, which were not detected by BMB, so they were considered as falsely negative in BMB. In these 24 cases, one or more of the following confirmed the PET/CT findings: subsequent follow up, corresponding CT changes that cannot be explained by other benign findings, targeted magnetic resonance  imaging, and/or local biopsy. This is illustrated in Table 6. Twelve cases showed regression of the disease in the form of complete response or partial response in either the interim or end of treatment follow-up, four cases showed progression of the disease in either the interim or end of treatment follow-up, local biopsy was done for seven cases, targeted MRI was done for 5 other cases, associated CT changes were present in 14 cases. Eight cases had other extranodal lesions which indicate advanced disease. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy with 95% confidence interval of PET/CT and BMB are illustrated in Table 7.

Agreement between PET/CT and BMB in the whole study
In this study, there were 117 (80.6%) cases with concordant results in both PET/CT and BMB but discordant results were 28 (19.3%) cases (Table 8 and Fig. 5). Agreement between the PET/CT and BMB findings was assessed using Cohen's k computation and was found 0.47 with p value less than 0.0001, which indicate moderate agreement.

Diagnostic performance of PET/CT compared with BMB in HL
Regarding HL, there were 88 cases. The PET/CT was reported as negative for BMI in 70 cases, of these 70 patients, one patient was interpreted as positive for BMI because of positive findings on BMB. Therefore, the 18F-FDG PET/CT was considered as truly negative for 69 patients and falsely negative in one patient for BMI. The 18F-FDG PET/CT was reported as positive for BMI in 18 patients, two of these patients showed diffuse FDG uptake in PET/CT; however, BMB results revealed reactive hyperplasia, so they were considered as falsely positive in PET/CT and PET/CT was considered as truly positive for 16 cases. The BMB results were positive in 6 patients. PET/CT detected BMI in 11 (12.5%) additional cases, which were not detected by BMB, so they were considered as falsely negative in BMB while BMB detected only one (1.1%) additional case not detected by PET/CT.

Agreement between PET/CT and BMB in HL
In this study, there were 74 (84.1%) cases with concordant results in both PET/CT and BMB but discordant results were 14 (15.9%) cases in HL.    Agreement between PET/CT and BMB in HL cases was calculated using Cohen's kappa computation and was found 0.350 with p value less than 0.0001, which indicates fair agreement between both PET/CT and BMB (Table 9).
In our study, agreement between PET/CT and BMB was also done for nodular sclerosis and mixed cellularity subtypes (Tables 10 and 11). For nodular sclerosis type cases, it was found 0.236, with p value 0.040, which indicates fair agreement between both PET/CT and BMB results in nodular sclerosis cases.
For mixed cellularity type cases, it was found 0.523, with p value 0.002, which indicates moderate agreement between both PET/CT and BMB results in mixed cellularity cases.

Diagnostic performance of PET/CT compared with BMB in NHL
Regarding NHL, there were 57 cases. The PET/CT was reported as negative for BMI in 30 cases. Of these 30 patients, one patient was interpreted as positive for BMI because of positive findings on BMB. Therefore, the 18F-FDG PET/CT was considered as truly negative for 29 patients and falsely negative in one patient for BMI. The 18F-FDG PET/CT was reported as positive for BMI in 27 patients but BMB results were positive in 15 patients. PET/CT detected BMI in 13 (22.8%) additional cases, which were not detected by BMB, so they were considered as falsely negative in BMB while BMB detected only one (1.8%) additional case not detected by PET/CT.

Agreement between PET/CT and BMB in NHL
In this study, there were 43 (75.4%) cases with concordant results in both PET/CT and BMB but discordant results were 14 (24.5%) cases in NHL.
Agreement between PET/CT and BMB in NHL cases was calculated using Cohen's kappa computation and was found 0.496, with p value 0.040, which indicates moderate agreement between both PET/CT and BMB (Table 12).
In our study, agreement between PET/CT and BMB was also done for DLBCL (Table 13). It was found 0.236, with p value 0.040, which indicates fair agreement between both PET/CT and BMB results in DLBCL cases.

Discussion
Assessment of BMI in lymphoma plays an important role in staging as its presence upstage the disease to stage IV.
This study compares 18F-FDG PET/CT and BMB in evaluating bone marrow in 145 lymphoma patients, 88 cases were HL, and 57 cases were NHL.
BMB is an invasive technique that allows histologic examination for only a small bone marrow sample from the posterior iliac crest. This is in contrast to 18F-FDG PET/CT, which is a noninvasive technique that allows visualization of the entire bone marrow.
The blind BMB does not exclude BMI when infiltration is present in sites other than the posterior iliac crest [16].  When BMB is only used as the reference standard, the sensitivity of PET/CT declines [5]. This is why we used follow-up findings, CT changes, and local biopsy and or MRI findings in patients with negative BMB results as an additional standard for these patients.
Diffuse uptake is considered a controversial area in most of the studies. Most studies found that diffuse uptake in bone marrow in HL cases is associated with a negative BMB, while in most DLCL cases with diffuse bone marrow uptake, positive BMB were found [11,17]. Others found a diffuse pattern in HL with BMI in a small percentage of cases [18,19]. In our study, we found 2 cases with HL having diffuse bone marrow uptake and positive BMB results (Fig. 6). This is consistent with the study done by Chen-Liang et al. [20], who also detected 2 cases of HL having diffuse uptake and positive BMB findings. In our study, all the six cases of NHL with diffuse uptake had positive BMB results (Fig. 7). This is consistent with the study done by Khan et al. [12], who found five cases with diffuse uptake, all had positive BMB results. However, in the study done by Cortés-Romera et al. [16], four cases only out of nine cases with diffuse uptake pattern had positive BMB results. Also, in the study done by Cerci et al. [21], 4 cases only out of 18 cases with DLBCL had positive BMB results, so it was concluded that cases with diffuse FDG uptake should be biopsied to establish etiology.
PET/CT detected 24 (16.5%) cases with positive BMI that were missed by BMB. Local biopsy was done guided by PET/CT in seven cases (Fig. 8) and targeted MRI was done in five other patients, both confirmed bone marrow infiltration. Twenty cases had absent posterior iliac crest involvement in either side or four cases had unilateral posterior iliac crest involvement. It was interesting that we found the site of the biopsy taken was from the normal posterior iliac crest (Fig. 9).
PET/CT can detect infiltration in any region of the skeleton compared to the confined site of the posterior iliac crest by BMB. It can also guide the site of the biopsy in clinically warranted cases when management will differ.
In our study, we found moderate agreement between PET/CT and BMB results in assessment of bone marrow in the whole study using Cohen's k computation. It was found 0.47 with p value less than 0.0001. As 117 (80.7%) patients showed concordant results between PET/CT and BMB and 28 (19.3%) showed discordant result. This is consistent with the previous studies in which the discordant rate was found ranged from 13.3 to 22.9% [1,14,16,[24][25][26][27].
In our study, there were 88 patients with Hodgkin lymphoma, 17 (19.3%) of them had BMI. BMB detected only 1 (1.1%) case that was not detected by PET/CT, but PET/CT detected 11 (12.5%) additional cases with bone marrow infiltration presented outside the site of the biopsy and therefore were missed by the BMB.
In the study done by Cheng et al. [1], who included 31 cases of pediatric HL. PET/CT detected BMI in 2 (6.5 %) additional patients not detected by BMB.
Purz et al. [31] compared the difference between BMB and F-18 FDG PET/CT in the detection of BMI in 175 HL pediatric patients all with stage more than IIA. They found that F-18 FDG PET/CT detected 22% of positive cases not detected by BMB and concluded that F-18 FDG PET may replace BMB in routine staging procedure. This is also consistent with previous studies, which recommended preclusion of the routine use of BMB for staging in these patients [8,17,32,33].
In our study, we found a fair agreement between PET/ CT and BMB in HL cases by using Cohen's k, it was found 0.350 with p value less than 0.0001. This is consistent with the study done by Cistaro et al. [28], who found also fair agreement between PET/CT and BMB findings as Cohen's k was found 0.398 with p value less than 0.001.
In our study, there were 57 patients with non-Hodgkin lymphoma, 28 (49%) of them had BMI. Only one (1.8%)   This is consistent with the study made by Badr et al. [34], who included 27 patients with NHL and found the sensitivity of PET/CT was 100% while that of BMB was 42.9% and the specificity was 100% in both.
In the meta-analysis done by Adams et al. [11], who included seven studies assessing PET/CT for detection of BMI in DLBCL. They found that the PET/CT sensitivity was 78.4% (95% CI, 69.9-85.5%) and specificity was 99.7% (95% CI, 98.3-100%) when both focal and diffuse uptake were considered positive for bone marrow involvement as in our study. This is against the study done by Adams and Kwee [35], who compared PET/CT and BMB of the posterior iliac crest and showed that PET/CT may be negative in the posterior iliac crest in up to 80% of cases with a positive BMB.
The role of both PET/CT and BMB in the assessment of BMI is complementary [25]. PET/CT can be used to  guide the site of biopsy by the FDG uptake when lesions are present outside the routine iliac bone marrow sampling sites. Some studies suggest that routine BMB is still necessary in cases with a negative PET/CT scan result as PET/CT may miss early limited infiltration of the bone marrow, which can be detected by the BMB [16].

Limitations
The number of patients in our study was relatively small. We included a heterogeneous group of patients of HL and NHL. This has made our results for the separate calculation of the sensitivity, specificity, accuracy of PET/ CT, and BMB in nodular sclerosis, mixed cellularity and DLBC types not accurate and should be evaluated with a larger sample size.
Most of the previous studies were done retrospectively, so local biopsy or targeted MRI were not done. From the strengths in our study is that it was done prospectively so guided local biopsy or targeted MR imaging were done in 12 (50%) patients with positive PET/CT and negative iliac crest BMB results. This helped to confirm the positive uptake detected by PET/CT. However, this was not done for all patients, as it was not possible to take another biopsy from another site in all patients, as it is an invasive technique accompanied by pain and anxiety. Also, not all cases could underwent further targeted MRI, as it should be recommended by the multidisciplinary team for special cases, who will really benefit from it and not to take the place of other patients with different diseases who are really in need for it.

Conclusions
PET/CT showed greater sensitivity than posterior iliac crest BMB in both HL and NHL for evaluating BMI and can detect BMI that is missed by BMB, as uptake may be present in other sites of the skeleton away from the iliac crest biopsy site. PET/CT also showed moderate agreement with BMB results. PET/CT should precede the BMB as it can replace BMB in many cases, especially those showing multifocal uptake in both HL and NHL. PET/CT can help to guide the site of the biopsy in suspected infiltration, especially in unifocal FDG uptake. Iliac crest BMB is still needed in cases with diffuse FDG uptake to differentiate uptake from reactive hyperplasia, in certain types of lymphoma with known limited FDG avidity, and in some cases with negative uptake to exclude early infiltration if management will differ. The main limitation of this study is the small sample size, for which we recommend more studies with larger sample size.