Association of loss of spleen visualization on whole-body diffusion-weighted imaging with prognosis and tumor burden in patients with multiple myeloma

This study investigated the clinical significance of loss of spleen visualization (LSV) on whole-body diffusion-weighted imaging (WB-DWI) in patients with multiple myeloma (MM). The WB-DWI of 96 patients with newly diagnosed MM (NDMM) and 15 patients with smoldering MM (sMM) were retrospectively reviewed. LSV was observed in 56 patients with NDMM (58.3%) and 1 patient with sMM (6.7%). Patients with NDMM with LSV had a higher median infiltration of bone marrow plasma cells (80.0% vs. 50.0%, p < 0.001) and median total diffusion volume (median; 540.2 vs. 137.0 mL, p = 0.003) than patients without LSV. Patients with LSV had a lower spleen-to-spinal cord ratio (0.36 vs. 0.96, p < 0.001) and worse 2-year overall survival (OS) (84.6% vs. 100%, p = 0.032). Patients who did not recover spleen visualization during treatment had a worse prognosis, even when they obtained very good partial response (median progression-free survival: 13.2 months). Spleen histopathological findings revealed higher cellularity and diffuse myeloma cell infiltration in a patient with LSV and splenic amyloidosis without extramedullary hematopoiesis in a patient without LSV. Therefore, LSV indicates worse prognosis for patients with MM, even when the patient responds to treatment. Further studies are warranted to clarify the immunological role of spleen in MM.

www.nature.com/scientificreports/ signal in patients with NDMM and its reappearance after treatment. Rasche et al. reported that lack of spleen visualization on WB-DWI is related to a high tumor burden and poor prognosis 11 . The previous study from Rasche et al. suggested that extramedullary hematopoiesis (EMH) in the spleen resulted in the loss of the spleen signal using Tc-99 m-labeled anti-CD66. This study aimed to investigate the association between WB-DWI, myeloma load and prognosis in patients with MM, focusing on: (1) the relationship among spleen signal loss, tDV, and tumor volume at diagnosis; (2) the change in spleen signal intensity and visualization during treatment; (3) the relationship between follow-up spleen signal, follow-up tDV, clinical data, and prognosis; and (4) the mechanism of the loss of spleen signal.

Methods
Study design and patients. The data of 96 consecutive patients with symptomatic NDMM diagnosed at Kameda Medical Center from January 2016 to December 2020, 15 consecutive patients with smoldering MM (sMM), and 2 autopsied spleens of patients with PC dysclasia (1 monoclonal gammopathy of undetermined significance and 1 primary plasma cell leukemia) were retrospectively reviewed. The diagnosis and treatment responses of patients with MM were assessed using the International Myeloma Working Group (IMWG) criteria 12,13 . The observation period was completed on October 30, 2021. All patients underwent at least one WB-DWI prior to treatment. This study was approved by the institutional review board of Kameda Medical Center (No.  and conducted in accordance with the Declaration of Helsinki. All participants provided informed consent. Cytogenetic analysis. High-risk cytogenetic abnormalities (CAs) were del(17p), t(4;14), or t(14;16) using interphase fluorescence in situ hybridization (iFISH) analysis. A gain of 1q21 was also investigated in all patients except for 2 with symptomatic MM. The iFISH analyses were performed prior to treatment using PCs purified with CD138-coated magnetic beads (Miltenyi Biotec, Bergisch Gladbach, Germany). The cutoff values of del(17p) and gain of 1q21 were 10% and 20%, respectively 14,15 . Acquisition and analysis of magnetic resonance images. As described in our previous report 10 , WB-DWI were obtained using a 1.5-Tesla unit (Magnetom Vision; Siemens Healthcare, Germany) with the following parameters: acquisition type, 2D; repetition time, 5400 ms; echo time, 74 ms; inversion time, 180 ms; slice thickness, 5 mm; and b-values, 0 and 900 s/mm 2 . The apparent diffusion coefficient (ADC) was calculated voxel by voxel for each image slice in b = 0 and 900 images. The tDV was calculated using regions of interest (ROIs) that included areas of abnormally high DWI signal intensity, which contained myeloma lesions in the BM or extramedullary tissue. The ROIs were automatically obtained using medical imaging software (BD Score; PixSpace, Japan) 16 .
Loss of spleen visualization (LSV) was defined as a visual loss of the spleen in maximum intensity projection on the WB-DWI. The spleen-to-spinal cord (SC) ratio (SSR) was calculated using the ratio of the signal intensity of the spleen and SC. The calculations are shown in Fig. 1. The spleen ROIs were defined as nonoverlapping ROIs    (Fig. 2a).
The correlations between 18 F-fluorodeoxyglucose ( 18 F-FDG) uptake in the spleen and LSV were investigated in 92 patients (95.8%) who underwent WB-DWI and PET/CT simultaneously during the initial diagnostic evaluation. LSV was observed in 53 (57.6%) patients, while 6 patients (6.5%) had higher 18 F-FDG spleen uptake than liver (equivalent to 4 or more in the Deauville 5-point scale 19 ), 5 of whom had LSV ( Table 1).
Histopathological findings. We reviewed two patients who had both received WB-DWI within 3 months before their deaths. One patient who had primary plasma cell leukemia with LSV died due to disease progression and had diffuse myeloma cell infiltration in the spleen (Supplemental Fig. S2a, c). The other patient who had a monoclonal gammopathy of undetermined significance with systemic light-chain amyloidosis without LSV died due to cardiac amyloidosis and sepsis (Supplemental Fig. S2) and had amyloid deposition in the spleen, but no myeloma cell infiltration (Supplemental Fig. S2b, d). No evidence of EMH in the spleen was detected in any patient using CD34, CD42b, or factor VIII immunohistochemistry (the CD34 immunohistochemistry results are shown in Supplemental Fig. S2e, f).

Discussion
In this study, LSV was associated with a high tumor burden and poor prognosis in patients with MM. Patients with LSV had a lower SSR and a negative correlation between the SSR and beta-2 microglobulin or BMPC was also identified, suggesting that SSR inversely reflected the myeloma burden. Visualization of the spleen during treatment was not restored in some patients (11/74, 14.9%), moreover, they had extremely worse outcomes despite achieving favorable treatment response.
Rasche et al. reported that the lack of a spleen signal was associated with a higher tumor volume and poorer prognosis, which is consistent with the results of the current study 11 . However, there was a difference in the percentage of patients with of spleen loss in our report (58.3%) and in the previous reports (24%). The current study included more older patients (median age: 75.5 years) and more advanced-stage MM (50.0% in ISS stage III) than the previous study (26% in ISS stage III). Moreover, our definition of LSV was looser than that of the previous report by Rasche. These differences in the patients characteristics and the spleen loss definition might have contributed to the difference in the percentage of patients with spleen loss between our report and in the previous report.
The median tDV decreased in patients with LSV who responded to treatment. Typically, the tDV increases in lesions with high cellularity and in patients with lesions that have an increased ADC value, such as in those with edema of the BM cells during hematopoietic recovery 20 or with changes in the microenvironment of the BM due to hematopoietic stimulating factor use. LSV may be useful in differentiating such reactive signaling from enhancement due to tumor invasion. In this study, ADC thresholds were obtained at 0 and 900 × 10 -3 mm 2 /s. However, it may be necessary to change the ADC thresholds to assess the tDV after treatment, as described in the MY-RADS 8 www.nature.com/scientificreports/ While LSV is easy to evaluate and is clinically valuable, the mechanism of LSV on WB-DWI remains unknown. Rasche et al. speculated that LSV is due to EMH in the spleen as anti-CD66 monoclonal antibodies were diffusely positive in the spleen of a patient with a large residual myeloma burden 11 . CD66 was reported to be positive in myeloid cells including promyelocytes and granulocytes 21,22 and was used as an evidence of EMH in a patient with myelofibrosis 23 . However, CD66 has been reported to be positive by flow cytometry analysis in approximately three quarters, but not all the patients with myeloma [24][25][26] . We found a patient with sMM complicated with myeloproliferative disorder who had a diffuse BM signal on WB-DWI but did not have LSV, although EMH was present (Supplemental Fig. S3). In addition, no pathological EMH was observed in the two autopsied spleens. 18 F-FDG uptake by the spleen can be observed on PET/CT scans in most patients with myeloproliferative disorders who have EMH 27 ; however, 18 F-FDG uptake at diagnosis in the spleen than in the liver was observed in only six patients with NDMM in this study despite approximately half of patients with NDMM having LSV on WB-DWI. Furthermore, spleen EMH in hematologic malignancies typically presents as a nodular, not diffuse, pattern 28 . Based on our findings and those of previous reports, the cause of LSV might not be due to EMH in the spleen.
The microenvironment of the spleen in patients with MM is largely unknown. The spleen is the organ of antibody production and some normal PCs are found in the spleen 29 . Indeed, Shapiro et al. reported that there was an increase of the percentage of myeloma cells in the splenic aspirations comparable to that in the marrow 30 . In the previous report, only 3/10 of the patients had palpable spleen. Although we considered the possibility that myeloma cells might have escaped from the BM niche into the spleen in the patients with LSV, there were no findings to support this. However, due to the poor prognosis of patients who had LSV at diagnosis and even at favorable treatment response, they should be evaluated by other methods such as measurable residual disease and be considered for change to other treatment regimens.
WB-DWI is a more feasible imaging technique than PET/CT as it is more cost-effective, less time-consuming (up to 30 min), requires no breath-holding, and does not expose the patient to radiation 31,32 . WB-MRI is currently used to detect focal lesions and determine the indication of treatment according to the IMWG criteria 12 . Moreover, recent studies demonstrated the efficacy of repetitive WB-MRI for treatment response and prognostic impact [7][8][9] . LSV, which could be additionally evaluated in the WB-DWI of diagnostic process, could become an additional and simplified prognostic indicator in patients with NDMM.
On the other hand, several hospitals are unable to perform WB-MRI. For such institutions, the IMWG consensus statement suggested the spine or pelvic MRI instead of WB-MRI 33 . In the current criteria, these MRI (WB-MRI or spinal/pelvic MRI) could only be used to evaluate treatment initiation criteria, and not used to evaluate the tumor load. Our SSR was simple to obtain (imaging in the spleen slice in b = 900 images) and could help the tumor load assessment. Moreover, in the hospitals where WB-MRI was not available, the repetitive SSR assessment could help detect patients with poor prognosis, such as those who never regain the spleen visualization (group C).
This study is not without limitations, including its retrospective nature, small sample size, and relatively short observation period. Multivariate analysis of group C in 63 patients who achieved VGPR or better could not be performed owing to the small number of events (19 progression and 6 death). The spleen pathology and WB-DWI could not be compared in most patients, as patients with MM rarely received spleen biopsy due to the risk of bleeding. While the direct infiltration of myeloma cells into the spleen has been proposed as a mechanism of LSV, the small patient population in this study could not provide enough power to verify this hypothesis. Moreover, we did not evaluate T1-or T2-weighted imaging or the potential risk of false positives or negatives of LSV. Our LSV and the "lack of a spleen signal" in the Rasche's report detected almost the same phenomenon; however, the MRI conditions or various parameters were slightly different.
In conclusion, this study showed that LSV and a low SSR on pre-treatment WB-DWI are correlated with high tumor volume and poor prognosis. As patients with LSV during treatment have very poor prognosis despite favorable treatment responses, the relationships between LSV and other variables should be investigated. The mechanism of LSV in WB-DWI remains unknown. Future studies that include pathological and microenvironment analysis of the spleen are necessary to validate our results, and to determine the unique function of the spleen in patients with MM.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.