PET studies are used on a regular basis in the initial staging of lymphoma patients-pretreatment-and assess response to treatment and prognosis [1-12]. BM involvement is synonymous of a stage IV and is associated with a worse short and long-term prognosis [1-12]. Concordant BM_u and BM involvement confirmed by a biopsy has been reported to correlate with worse outcomes [12].

Hong et al. have reported a shorter 2 years event free survival (EFS) and overall survival (OS) with concordant findings. Discordant findings were associated to worse outcomes only if the BM biopsy (BM_b) was positive [13]. The utility of FDG PET-CT in evaluating BM involvement has also been explored by other authors with variable results as to its benefit [6,7,12-23]. Because FDG PET-CT’s are performed routinely now in the pretherapy staging of most lymphoma patients, extracting noninvasively additional information in regards to bone marrow status would be very valuable. This is in the context of knowing that not all patients are clinically eligible to undergo a BM_b, that selecting eligible patients is not always a straightforward exercise, and without overlooking the fact that a BM_b has a small adverse event rate (ranging from pain, bleeding, infections), as well as an additional cost of the procedure and pathology review. Additionally sampling occasionally may not be adequate for pathology interpretation and a repeat procedure becomes challenging. Uncovering quantitative and qualitative parameters to elucidate the etiology of bone marrow uptake may help the radiologist and oncologist. BM_u is frequently encountered on pretherapy scans and almost always seen in post-treatment scans due to reactive changes. Pretherapy, it can be associated with disease involvement, however occasionally uptake will be related to benign etiologies (anemia, stimulation, a hypercellular marrow … etc). We decided to correlate qualitative and quantitative parameters to BM_b findings. The main objective of our study was to evaluate the diagnostic predictability of BM pathology based on the assessment of three semi-quantitative parameters, maximum standardized uptake value (SUVm), average SUV (AvSUV), and the ratio of SUVm/mediastinal blood pool (MBP) and qualitative influencing factors such as sex, bone marrow uptake type and pattern in a group of consecutive lymphoma patients who had an initial staging PET scan at Yale-New Haven Hospital and whose biopsy reports were available for review.

A review of 121 PET scans performed at Yale New Haven Hospital in patients with lymphoma was performed. A qualitative review of BM_u was performed and scans were examined for their type of uptake (none, mild, prominent), and their pattern (diffuse, patchy, focal). Semi-quantitative parameters (standardized uptake values) (SUVs) maximum (max) and average (Av) were consistently measured by drawing a VOI (volume of interest: 329 cm³) on a sagittal view from the mid-thoracic to the mid-lumbar region of the spine (Figure 1). Mediastinal blood pool (MBP) was consistently measured with an ROI (region of interest: 85cm³) at the Aortic arch level. BM_b results were analyzed from reports and classified as normal, hypercellular or positive (infiltrated). [¹⁸F]-FDG PETCT scans were reviewed retrospectively using a GE Advanced Workstation and were acquired according to standard Yale New Haven Hospital protocol on a GE Discovery (DST-E and D690) 16 and 64 slice LYSO PET-CT’s. A standard [¹⁸F]-FDG dose was administered ranging from 12 to 16 mCi. The PET portion of the scans was acquired 60 minutes post-injection using a 3 min per bed position protocol. The CT acquisition was performed using dose modulation with 120Kvp and 50 to 80 mAs and reconstructed to a 3.75mm slice thickness.

Figure 1: Normal bone marrow with mild and diffuse uptake.
MIP and Fused sagittal images of a 36 year old male showing mild and diffuse FDG uptake in the thoracic and lumbar spine. Bone marrow biopsy was negative for marrow infiltration by lymphoma.

Descriptive statistics were used to present patient characteristics for patients who had a BM_b available (n=36). A Kruskal-Wallis test was performed to evaluate if SUVm, AvSUV, or SUVm/MBP, all nonnormally distributed continuous variables, could be independently associated with BM diagnosis. Fisher’s exact test was used to evaluate the unadjusted association between BM pathology, uptake type, and uptake pattern. To evaluate the correlation between SUVm, AvSUV, and SUVm/MBP, a Pearson correlation test was performed.

A Multivariate logistic regression model was then built to evaluate the predicting effect of BM_u pattern (Diffuse and Focal versus Patchy), SUVm, AvSUV, and SUVm/MBP, on the type of BM pathology result.

Patient characteristics

We looked retrospectively at a group of consecutive lymphoma patients who had an initial staging PET scan performed at our institution. Out of 121 patients, there were 36 subjects who had a BM biopsy available for review and thus in this dataset. Patient characteristics are presented in table 1. The median age of all 36 patients was 54. 67% (24/36) were males. 58% (21/36) had a normal, and 25% (9/36) a positive BMB. 16% (6/36) had a hypercellular bone marrow (Figure 5).

Table 1: Patient Characteristics.

Figure 5: Hypercellular Bone marrow.
(A) Hypercellular Bone Marrow Mild and Focal uptake.
45 year old female with Axial and MIP PET images showing mild and focal uptake in bilateral pelvic bones, however biopsy revealed a hypercellular bone marrow.
(B): Hypercellular Bone Marrow Mild and patchy uptake.
PET images from a 32 year old female shows mild and patchy involvement of pelvic bones and lumbar spine. Bone marrow biopsy showed a hypercellular bone marrow.

BM_u type is presented in table 2. Eight subjects (22%) had no BM_u, 21 subjects (58%) had mild BM_u uptake, and 7 (20%) had prominent BM_u. Patients with different BM_u types were significantly different in terms of SUVm (p=0.0101), AvSUV (p=0.0147), SUV/ MBP (p=0.0042).

Table 2: Cross-tabulation by BM Uptake Type.

Bone Marrow pathology classification results are presented in table 3. Nine subjects (25%) had a positive diagnosis on their BM_b, 6 subjects (17%) were diagnosed as having a hypercellular BM, and 21 (58%) had a negative BM_b. In the BM positive group, surprisingly 2 had no appreciable type of uptake. No pts had patchy uptake, but 67% (6/9) had diffuse and 33% (3/9) focal patterns of uptake. 55.56% (5/9) had unexpectedly a low AvSUV < 1.5. In the BM normal group, 71.42% (15/21) had either mild or prominent type of uptake, leaving only 28.57% (6/21) with no uptake. None had a focal pattern of uptake and 90.48% (19/21) had a diffuse pattern of uptake. The majority of our cohort had an SUVm > 2.5 and SUVm/MBP > 2.5 regardless of BM pathology. 90.48% (19/21) of the normal BM group, 100.00% (6/6) of the hypercellular, and 88.89% (8/9) of the positive BMb group had SUVm > 2.5. Similarly, 52.38% (11/21), 66.67% (4/6), and 55.56% (5/9), had SUVm/MBP > 2.5, irrespective of the BM pathology groups. This suggests that SUVm is not predictive of BM pathology.

Table 3: Cross-tabulation by BM Pathology.

Relationship between BM uptake pattern and BM pathology

BM Pathology level was overall different across the BM Uptake Patterns using Fisher’s Exact Test (p=0.0192; table 3). But when drilled down, only those who had a focal uptake pattern had a borderline significantly different distribution in BM Pathology than those who had a patchy uptake pattern. Patients in this reference group using a logistic regression model (p=0.0780; table 5), were significantly different from those who had a diffuse uptake pattern (p=0.0432; table 4). However this later was not significantly predictive for BM pathology when compared to the patchy group (p=0.7093; table 5). And because we were unable to tease apart the difference between patchy and diffuse uptake groups and because of the small sample size, we could only draw a conclusion for patients in the focal groups versus the rest combined. The odds of having a hypercellular or positive BM diagnosis for those with a focal uptake pattern was 10.0 times greater (p=0.0245, 95% CI: 1.3, 74.4) than for those with either patchy or diffuse uptake. Further analysis looking at BM positive versus BM negative groups, shows BM pattern is no longer a significant variable to predict BM pathology (p=0.168). Subsequently, the odds ratio of having a positive BM pathology based on a focal vs. patchy pattern is not significant, with the estimate 6.25 and 95% CI (0.847, 46.123) (Figure 3).

Table 4: Full Logistic model on BM pathology diagnosis result based on semi-parametric variables and BM uptake pattern with Patchy as the reference group for uptake pattern.

Table 5: Full Logistic model on BM pathology diagnosis result based on semi-parametric variables and BM uptake pattern with Diffuse as the reference group for uptake pattern.

Figure 3: Positive infiltrated bone marrow with mild focal FDG uptake
MIP and axial images from a 35 year old male shows focal FDG uptake at the lower thoracic spine. Subsequent biopsy was consistent with lymphoma with bone marrow infiltration.

In our data set we only have 36 total subjects, which when divided into subgroups often yielded a small number of subjects in each category, especially in the patchy and focal groups of BM_u pattern. However this is not uncommon in the reported literature regarding this topic. Although the pattern of uptake overall seemed to correlate in our study, due to the small number of events, we should be cautious when we interpret our results. A further study with a larger sample size could prove interesting.

A correlation was found in our analysis, however that does not clearly imply that focal uptake would be predictive of bone marrow positivity, even though no patients with focal uptake had a normal bone marrow. A few patients with focality had a hypercellular marrow (Table 3). These findings are more predictive of cortical bone involvement, which also already confers a stage IV and can be associated with a worse prognosis and where a bone marrow biopsy has a less substantial value. This finding may at the most guide clinical management into pursuing a bone marrow biopsy in equivocal cases. The literature frequently implies that focal bone marrow uptake is due to focal bone marrow involvement – in other words bone marrow implants -, even though this would change significantly the current prevalence of bone marrow implants in lymphoma patients. Our findings are also supported – even though not absolutely concordant - by a recent publication by Hong et al. [13]. Purz et al. assumed in their study that all focal lesions represented bone marrow involvement by segmented BM involvement [14], even in the presence of anatomical skeletal lesions on Computed Tomography or MRI when the focality was likely to be due to cortical bone metastasis and not bone marrow infiltration/implants. A large number of patients were assumed to be BM positive by indirect evidence of focality and diffuse uptake was assumed to be reactive and while it is frequently the case, bone marrow infiltration will also appear similarly as diffuse uptake and discrimination is very difficult [14]. In our study 6 patients with diffuse uptake had a positive BM_b. In our study, also two patients with a positive bone marrow result had no appreciable uptake on PET.While most patients with a normal bone marrow had either mild or diffuse uptake (Table 3 and Figure 4 ).

Figure 4: Positive Bone marrow involvement
(A) Positive Bone Marrow Mild and Diffuse uptake.
MIP and sagittal fused images and axial PET images in a 40 year old male with a positive bone marrow biopsy, however the FDG PET images only show mild and diffuse uptake.
(B) Positive Bone Marrow Prominent and Patchy uptake.
Coronal PET and fused images and sagittal fused images in a 28 year old male shows intense and patchy uptake in the axial and appendicular skeleton. Biopsy was positive for bone marrow involvement with Lymphoma.
(C) Positive Bone Marrow Prominent and Patchy uptake.
MIP PET images of a 29 year old male with prominent FDG uptake on a pre-therapy scan. Bone marrow biopsy was consistent with lymphoma involvement of bone marrow.

Reported findings in the literature are variable in part because of variations in qualitative analysis methods; however even with this in mind it seems difficult to translate them into clinical practice with relatively good accuracy. Aware of the limitations of qualitative analysis alone, we believed that developing quantitative, semiquantitative analysis tools or a combination of qualitative analysis and quantitation (Figure 2) would be more beneficial in guiding patient care as this would allow for reproducibility and diminish variability.

Figure 2: Qualitative evaluation and Semi-Quantitation of [¹⁸F]-FDG uptake.
A qualitative review of bone marrow uptake was performed and scans were examined for their type of uptake (none, mild, prominent), and their pattern (diffuse, patchy, focal). Semi-quantitative parameters, standardized uptake values (SUVmax and AvSUV) were measured by drawing a VOI (volume of interest: 329 cm³) on a sagittal view from the mid-thoracic to the mid-lumbar region of the spine.

With that in mind we found that most patients also regardless of bone marrow pathology had high SUV values, more prominent in the normal bone marrow group and even when corrected to a MBP reference, no correlation was found. No predictive semi-quantitative values able to discriminate bone marrow involvement were found. Kand et al. recently reported a benefit in using semi-quantitative measures to predict BM involvement [15], Duarte et al. on the other hand explored quantitation in focal bony disease [16], which in our study would be concordant with our impression of focality as being a measure of bony metastasis. However in our study SUV measures were high in all groups and especially in the normal group even when corrected to a reference.

Several studies have claimed the benefit of being able to predict bone marrow pathology with FDG PET [6,7,14,17-20], however their methodologies differ, their analysis does not always use a BM_b as a gold standard, or ignores the results of the BM_b and we could not reproduce their conclusions. On the other hand other authors have proposed that qualitative or semi-quantitative BM uptake on [¹⁸F]-FDG PET was not predictive of BM pathology [12,13,16,19-24]. This is also our point of view and we believe that it should be used as an adjunct and a guide with other clinical tools and not as a confirmatory test.

Although BM uptake pattern in our study was found to be statistically significantly associated with BM pathology through Chi-square test with alpha=0.05, no other parameters or variables, including BM_u type, SUVm, AvSUV and MBP ratio, were found to be associated. Prediction of a positive BM pathology by PET was not achieved when groups were segregated into positive and negative BM status and further exploration with a larger cohort is necessary.