Dispenzieri, Angela

Abstract Multiple myeloma is a highly radiosensitive malignancy but, at the present time, radionuclide-based interventions have no proven place in disease management. Bone-seeking radionuclides such as 153-Sm-EDTMP and 166-Ho-DOTMP are promising agents for systemic delivery of ionizing radiation to sites of myeloma disease activity, but they are associated with significant myelosuppression at therapeutically effective doses and have therefore been used only in the context of myeloma stem cell transplantation protocols. We previously reported that the proteasome inhibitor PS-341 potently and selectively sensitizes myeloma cell lines and primary myeloma cells to the lethal effects of ionizing X-irradiation (Goel et al, Exp Hematol. 33, 784, 2005). To determine whether PS-341 is equally effective in sensitizing myeloma cells to ionizing beta-radiation and to extend our initial observations to an in vivo model, we combined PS-341 with the bone-seeking radionuclide 153-Sm-EDTMP. In vitro clonogenic assays were performed using a panel of myeloma cell lines and demonstrated synergistic killing following co-treatment with PS-341 and 153-Sm-EDTMP. Using the orthotopic, syngeneic 5TGM1 myeloma model, the median survivals of mice treated with saline, two doses of PS-341 (0.5 mg/kg), or a single non-myeloablative dose of 153-Sm-EDTMP (22.5 MBq) were 21, 22 and 28 days, respectively. In contrast, mice treated with combination therapy comprising two doses of PS-341 (0.5 mg/kg), one day prior to and one day following 153-Sm-EDTMP (22.5 MBq) showed a greatly prolonged median survival of 49 days. Correlative studies indicated that, compared to single-agent therapy, combination treatment with PS-341 and 153-Sm-EDTMP rapidly reduced the clonogenicity of bone-marrow resident 5TGM1 cells, slowed the elevation of serum myeloma-associated paraprotein levels, and was associated with longer term preservation of bone mineral density. The myelotoxicity of single agent and combination therapies was evaluated by comparing peripheral blood cell counts in each of the treatment groups, and by clonogenicity assays of hematopoietic progenitors isolated form normal mice receiving identical treatment regimens. Treatment with 153-Sm-EDTMP led to significant, but transient, myelosuppression which did not differ between animals treated with 153-Sm-EDTMP alone versus those treated with the combination of PS-341 plus 153-Sm-EDTMP. In summary, PS-341 is a potent in vivo radiosensitizer that greatly enhances the therapeutic potency, without increasing myelotoxicity, of skeletal targeted radiotherapy in the syngeneic, orthotopic 5TGM1 myeloma model. Based on these findings, we propose to conduct a phase I clinical trial at Mayo Clinic to evaluate the combination of PS-341 plus non-myeloablative skeletal targeted radiotherapy (using 153-Sm-EDTMP) in patients with advanced or treatment-refractory multiple myeloma.

Abstract Background: Multiple Myeloma (MM) is characterized by accumulation of malignant plasma cells with relatively low proliferative and apoptotic rates. The proliferative rate of plasma cells, measured in terms of % plasma cells labeled by bromodeoxyuridine (BrdU) using slide based fluorescence microscopy (plasma cell labeling index or PCLI) has been shown to be a powerful prognostic factor in MM. We have previously shown that the % bone marrow plasma cells in apoptosis (PCAP) determined by a flow cytometry method that uses Annexin V staining or 7-amino Actinomycin D (7-AAD) correlates with disease stage in MM. The goal of this study was to examine the prognostic value of the PCAP. Methods: We studied 145 patients with MM including 74 with newly diagnosed MM, 40 with relapsed MM, and 31 with MM on treatment in whom simultaneous measurements of % plasma cell apoptosis and PCLI were available. The PCLI is expressed as the percentage of cytoplasmic immunoglobulin positive plasma cells that have taken up BrdU. The PCLI was characterized as high when >= 1%. Mononuclear cells isolated from bone marrow aspirate, following red cell lysis with ACK solution, are incubated with CD38-PE and either CD138-FITC or CD45-FITC. This is followed by incubation with 7-AAD and the plasma cells are gated using the CD38/CD45 staining pattern. The Annexin method used a similar strategy, with Annexin V in place of 7-AAD. The percentage of plasma cells in apoptosis (PCAP) was categorized as high when >5%, the median value. Results: The median follow up for the entire group was 34 months (range, 0.1 mo to 9 years) and 112 (77%) had died at the time of analysis. Forty seven patients (32%) had a high PCLI (>=1%). The median overall survival for the patients with high PCLI was 14.6 months versus 42.3 months for those with a low PCLI; P = 0.0005 (log rank test). Seventy-five (51.7%) patients had a high PCAP. The median overall survival for those patients was 28.9 months versus 40.2 months for those with a low PCAP; P = 0.078. We next examined if there was any correlation between a high PCAP and high PCLI and found none; P = NS by (Fisher’s exact test). Additionally in a multivariate Cox model using both PCLI and PCAP, both were independently prognostic for overall survival. We then developed an index by adding PCLI to log transformed PCAP (PCLI + log PCAP). Using a cutoff of 1.5 for the new index we were able identify a group of patients with poor survival. The 59 patients with a high value for the new index had a median survival of 19.3 months vs 46.1 months for the rest (P < 0.0001, log rank test). Conclusion: The measurement of plasma cell apoptosis has prognostic value in patients with MM. More importantly, we have shown that by combining measures of two important biological characteristics of the plasma cell, namely proliferation and apoptosis, patients with a poor prognosis can be identified. Such a prognostication strategy can help stratify patients in clinical trials as well as shed light on the disease biology.

Abstract Chromosome 13 (Δ13) abnormalities are found in greater than 50% of patients with Multiple Myeloma (MM). MM is most commonly defined by chromosome 13 monosomy or 13q loss (85%). Interstitial deletions comprise the remaining 15%. Many studies have revealed that Δ13 in MM are associated with poor survival and reduced response to therapy. Genes mapping to chromosome 13 may be involved in pathogenesis and/or progression of the disease due to loss of function from gene mutation or from epigenetic effects such as haploinsufficiency. In this study, array-based comparative genomic hybridization coupled with microarray technology (aCGH) is used to detect gene copy number loss on chromosome 13 from nine MM patient samples. Whole genome long-oligo microarrays constructed by Agilent Technologies were used which contain 40,000 genes that span the human genome with an average spatial resolution of ~75 kb. Using genomic DNA isolated from MM patients with interstitial deletions on chromosome 13, DNA was amplified, labeled and hybridized with a differentially labeled normal DNA reference to determine gene/genomic copy number changes. Arrays were analyzed to search for the minimum region of loss based upon single copy loss for a series of nearby mapping transcripts. A common region of loss of 2.2 Mb, at 13q14.2 was detected. Additionally, we investigated the correlation between genomic copy number change and the expression level for MM patients in the13q region. From an independent gene expression data set whose expression measurements were conducted with Affymetrix HG-U133A v2 microarrays, data was selected that corresponded to the samples used for current aCGH. Expression values from MM samples were divided by the mean expression values from 12 normal bone marrow samples for each gene and the resulting values were treated as surrogate ratios between MM and normal samples. Probes from both microarrays were then aligned according to their chromosomal positions and merged if their chromosomal positions overlapped. Composite chromosomal maps were generated that displayed the expression levels and copy number changes. The maps were used to differentiate chromosomal regions in 13q where copy number changes and expression levels show high correlation and regions where such correlation was not observed. Although the number of probes sampled in the expression microarray was much smaller than those in aCGH microarray, a chromosomal region of great interest, that encompasses 13q14.2, arose naturally from this analysis. Although the mechanism by which loss of 13q effects tumorigenesis in MM could be a haploinsufficiency model, we are not ruling out the presence of a tumor suppressor gene in this region. We are evaluating candidate tumor suppressor genes in the region for loss of function by mutational analysis and hypermethylation studies.

Abstract Chromosome 13 deletion (Δ13) is one the most common genetic abnormalities observed in multiple myeloma (MM) and confers a poor prognosis. Studies to identify critical tumor suppressor genes in MM till date have lack adequate resolution, and the molecular phenotype associated with Δ13 has not been established. In this study we seek to establish a molecular profile for Δ13 and to finely map the minimal common region of deletion (CDR) on chromosome 13 by using gene expression profiling (GEP) and array comparative genomic hybridization (aCGH). GEP was performed on RNA from purified plasma cells of 72 newly diagnosed MM and 50 human myeloma cell lines (HMCLs) using the Affymetrix U133A chip and U133plus chip (Affymetrix, Santa Clara, CA) respectively. Patients were assigned TC classes which correlates with underlying genetic subtypes. In addition, aCGH was performed on 79 MM samples (36 with GEP data) and 50 myeloma cell lines (HMCLs; 48 with GEP data) using a platform utilizing 60-mer oligonucleotides (Human Genome CGH 44B Oligo Microarrays, Agilent Technologies), which have a resolution of about 70Kb. Raw data was extracted using the Feature Extraction 8.1 and visualized using CGH Analytics 3.2 (Agilent Technologies). To define a signature that reflects the biological consequences of Δ13 and not its close association with some genetic subtypes of MM (example, t(4;14) and t(14;16)), we selected a training cohort consisting of cases belonging to D1, D2, 11q13 and none TC class such that cases with and without Δ13 are balanced for ploidy and TC classes (n=28). A 152-gene Δ13 signature was identified. Its specificity was confirmed by leave-one-out cross validation using the K-nearest neighbor (KNN) class prediction algorithm (predictive accuracy of 100% in the training cohort). It was subsequently validated in a validation cohort that includes t(4;14) and t(14;16) that were not use to derive this gene signature, maintaining a high predictive accuracy (88%). This gene-set may therefore reflect core transcriptional consequences of Δ13 and is enriched for genes involved in the cell cycle and apoptosis. Using aCGH, 20 patients have Δ13 (mostly whole chromosome) all involving the 13q14–q21 region. Of the 32 HMCLs with Δ13 (whole chromosome or interstitial), 22 have mono-allelic loss of RB1 whereas 4 had bi-allelic loss. Besides RB1, bi-allelic loss of several other genes in the 13q14–q21 region (CYSLTR2, CDADC1, ITM2B, PCDH9) was also observed. As the CDR is large and contains a number of genes, we utilized a set of criteria (bi-allelic loss in HMCLs, within 13q14–q21 CDR, gene expression significantly correlated with copy number, and loss could explain molecular phenotype) to narrow down the potential candidates. Only RB1 and ITM2B fulfilled these criteria. Furthermore, RB protein levels correlated well with mRNA levels and DNA copy suggesting that even in absence of mutations resulting in bi-allelic loss, mono-allelic loss of RB1 could be tumorigenic through a haploinsufficiency mechanism.

Abstract Background: Primary systemic amyloidosis (AL) is an incurable plasma cell disorder. Lenalidomide, especially in conjunction with dexamethasone, has been shown to be highly active in patients with multiple myeloma. Methods: We studied the toxicity and efficacy of lenalidomide in patients with symptomatic AL. Patients received single agent lenalidomide. If progression by 3 months or no evidence of hematologic response after 3 cycles, dexamethasone was added. Originally, twenty-three patients (Cohort 1) were enrolled according to study design. Because of a significant early drop out rate and notable activity of the regimen, the trial was modified to include an additional 15 patients (Cohort 2). Baseline characteristics and adverse events are available for all enrolled patients, but at the time of this writing, response data are available for Cohort 1 patients due to short follow-up of Cohort 2, but will be updated by the time of the meeting. Results: Median age was 64 years, with 69% male. Twenty-three were previously treated. Organ involvement was cardiac (67%), renal (64%), hepatic (17%), nerve (17%). Thirty-three, twenty-two, and forty-four percent of patients were cardiac biomarker stage 1, 2, and, 3 respectively. Of the 37 patients, one was a cancel, and 6 have not yet made it through 3 months of protocol treatment and event monitoring. The respective median follow-ups for Cohorts 1 and 2 are 17 and 3.4 months. Of the remaining, 30 patients, within the first 3 cycles of therapy fifteen patients discontinued treatment: 7 early deaths and 8 adverse events or other causes. Three additional patients died 0.5 to 2 months after stopping treatment. The best predictor for early withdrawal and/or death was baseline NT-proBNP and cardiac biomarker staging system (cut-offs for serum troponin T <0.035 ng/ml and NT-proBNP <332 pg/ml--Stage I neither above cut-off; Stage III, both above cut-off; and Stage II, one above cut-off). Figure1 Figure1. Figure 2 Figure 2. Of the twenty 22 patients assessed for response ten patients responded to treatment for an overall response rate of 45%, including 23% organ responders. Among the patients with organ responses, there were four renal responses, two cardiac responses and two liver responses. All but one of the responders had dexamethasone added to their treatment program. The most common grade 3−4 adverse events at least possibly attributable to lenalidomide were neutropenia (38%), thrombocytopenia (21%), fatigue (15%), and rash (12%). Conclusions: Lenalidomide and dexamethasone has significant activity in patients with AL with an overall response rate of 45%. Baseline NT−proBNP may be an effective eligibility screening strategy for subsequent trials.

Abstract Hyperdiploid MM (H-MM), characterized by recurrent trisomies constitute about 50% of MM, yet very little is known about its pathogenesis and oncogenic mechanisms. Studies in leukemia and solid tumors have shown gene dosage effect of aneuploidy on gene expression. To determine the possible gene dosage effect and deregulated cellular program in H-MM we undertook a gene expression study of CD138-enriched plasma-cell RNA from 53 hyperdiploid and 37 non-hyperdiploid MM (NH-MM) patients using the Affymetrix U133A chip (Affymetrix, Santa Clara, CA). Gene expression data was analyzed using GeneSpring 7 (Agilent Technologies, Palo Alto, CA). Genes differentially expressed between H-MM and NH-MM were obtained by t-test (p<0.01). The majority of the differentially expressed genes (57%) were under-expressed in H-MM. Genes located on the commonly trisomic chromosomes were mostly (but not always) over-expressed in H-MM and constitute 76% of over-expressed genes. Chromosome 1 contained the most differentially expressed genes (17%) followed by chromosome 12 (9%), and 19 (8%). To examine the relationship of gene copy number to gene expression, we examined the expression of genes on chromosomes 9 and 15 in subjects with 2 copies (15 normal control and 20 NH-MM) and 3 copies (12 H-MM) of each chromosome as detected by interphase FISH. We then derived a ratio of the mean expression of each gene on these chromosomes between patients with 3 copies and 2 copies of the chromosome. If a simple relationship exists between gene expression and gene copy number, one would expect the ratio of expression of most genes on these two chromosomes to be about 1.5 in H-MM compared to NH-MM. However, many genes have ratios either higher than 2 or lower than 0.5. Furthermore, when the heterogeneity of cells with underlying trisomies is taken into consideration by correcting the ratio for the number of cells with trisomies, the actual ratio is always lower than the expected ratio. When the expression of genes on a chromosome was compressed to a median value, this value was always higher in the trisomic chromosomes for H-MM compared to NH-MM. The data suggests that although gene dosage influence gene expression, the relationship is complex and some genes are more gene dosage dependent than others. Amongst the differentially expressed genes with known function, 33% are involved in mRNA translation/protein synthesis. Of note, 37 of the top 100 differentially expressed genes are involved in these processes. In particular, 60 ribosomal protein (RP) genes are significantly (p<0.05) upregulated in H-MM. This signature in H-MM is not associated with increase proliferation as measured by PCLI. This predominant signature suggests that deregulated protein synthesis may be important for the biology of H-MM. Many of these RPs are involved in the synthesis of product of oncogenic pathways (e.g. MYC, NF-KB pathways) and may mediate the growth and survival of tumor cells. It is therefore possible that these tumor cells may be sensitive to the disruption of mRNA translation/protein synthesis. Targeting the mTOR pathway with rapamycin may therefore be useful for treatment of H-MM.

Abstract Background: The serum free light chain (FLC) assay is increasingly used to monitor patients (pts) with oligo-secretory or non-secretory multiple myeloma (MM) and pts with primary amyloidosis lacking measurable monoclonal protein in the serum or urine. Criteria to use this assay to assess response to therapy have recently been proposed (Rajkumar SV, Kyle RA. Best Pr Clin Haematol2005;18:585–601) but have not been validated. The goal of this study was to validate the response criteria for the FLC assay in a prospective trial of lenalidomide plus dexamethasone in newly diagnosed MM. Methods: 34 pts were enrolled in the trial; 27 pts who had serial FLC assessments were studied. FLC estimation was carried out using the serum FLC assay (FreeliteH, The Binding Site Limited, UK) performed on a Dade-Behring Nephelometer. Pts with κ /λ FLC ratio <0.26 were defined as having monoclonal λ FLC and those with ratios >1.65 as having a monoclonal κ FLC. The monoclonal light chain isotype was considered the “involved” FLC isotype, and the opposite light chain type as the “uninvolved” FLC type. Partial response (PR) required an abnormal baseline FLC ratio and any one of the two following criteria: 1) a 50% decrease in the level of the involved FLC plus a 50% decrease (or normalization) in the ratio of involved/uninvolved FLC or 2) 50% decrease in the difference between involved and uninvolved FLC levels. Complete response (CR) required normalization of FLC ratio and negative serum and urine immunofixation. Response at 4 months or earlier by the Bladé criteria was compared to FLC response criteria from the same evaluation. Results: Three pts had normal FLC levels and ratio and were not included in the analysis. 23 of the remaining 24 achieved a PR or better by Bladé criteria. A 50% decrease in the level of the involved FLC plus a 50% decrease (or normalization) in the ratio of involved/uninvolved FLC correctly classified 20 of the 22 responding pts (sensitivity 91%); 2 pts could not be evaluated since baseline FLC ratio could not be calculated. On the other hand, a 50% decrease in the mathematical difference between involved and uninvolved FLC levels correctly classified all 24 responding pts (sensitivity 100%). The one non-responding pt by Blade criteria was correctly classified by both FLC criteria. All pts were correctly classified by both criteria when only those with a baseline “involved” FLC level of at least ≥10mg/dL (≥100mg/L) were considered (15 pts). Conclusions: This study demonstrates that the serum FLC assay can be used to assess response to therapy. A 50% decrease in the difference between “involved” and “uninvolved” FLC levels will suffice as FLC criteria for PR, eliminating the need for the alternative criteria based on the involved FLC level and the ratio. We recommend that this FLC response criteria be used only in pts not having measurable levels of serum and /or urine M protein.The FLC response criteria will now enable most patients with oligo-secretory and non-secretory MM to enter trials for which they are currently ineligible due to “lack of measurable serum or urine M protein.” This study is limited by the lack of adequate non-responders to calculate specificity and lack pts who progressed to validate progression criteria and needs further validation.

Abstract Introduction: High dose melphalan followed by autologous stem cell transplantation (ASCT) is an effective treatment for patients with light chain associated (AL) amyloidosis. Longer patient survival and higher rates of organ response have now been documented by multiple studies. However, predictors of organ response remain unknown. Previously, we have reported the benefits of renal response after ASCT in this population. This study was conducted to investigate the characteristics that would predict renal response after ASCT. Methods: The study was performed retrospectively on consecutive patients that underwent ASCT at our institution from March of 1996 to December of 2004. Exclusion criteria include baseline proteinuria < 1 g/d, dialysis prior to ASCT and lack of laboratory data at follow up to determine renal response. Renal response was defined by > 50% reduction in baseline proteinuria with < 25% decline in renal function as measured by serum creatinine. Treatment related mortality and dialysis dependence prior to meeting criteria of response were viewed as treatment failures. Hematologic response was determined by 50% reduction of monoclonal protein (free light chain) or complete eradication if the monoclonal protein was too small to be quantified. Results: A total of 135 patients met criteria for study. Median age was 56.2 years at the time of transplant, 53.7% were male. Median baseline proteinuria and GFR were 6.4 g/d and 70 ml/min/1.73m2 respectively. Renal response was achieved in 35.6% of the patients while hematologic response was 71.1% in the 128 patients evaluated. Patient’s age, sex, albumin, GFR, proteinuria, conditioning regimen, and hematologic response were evaluated and the following were found to be associated with renal response: albumin (p = 0.001), proteinuria (p = 0.008), and hematologic response (p = 0.0002). The cutoff for albumin was found to be 1.6 mg/dl and proteinuria was 3.5 g/d. Multivariate analysis using a logistic regression model showed hematologic response and proteinuria to be independent predictors of renal response. The impact of proteinuria and hypoalbuminemia was then investigated together (Table 1). When combined, they were a better predictor then either one alone (Hazard ratio = 6.34 for combined, 3.43 for proteinuria, 3.32 for hypoalbuminemia). The combination was also a better independent predictor of renal response in the multivariate analysis. In this group of patients, renal response was associated with longer survival but hematologic response was not (p = 0.02). Discussion: Our study showed that besides hematologic response, baseline serum albumin and proteinuria are independent predictors of renal response in AL patients after ASCT. Hypoalbuminemia and nephrotic range proteinuria, both markers of the severity of renal disease, have strong negative impact on response. This implies that there may be a limit to the reversibility of organ damage even when hematologic response is achieved. This study also points out the importance of organ (renal) response in this disease as hematologic response alone did not predict long term outcome. Our results suggest ASCT should be done early for AL to insure optimal organ response and patient outcome. Table 1 The Effects of Hypoalbuminemia and Proteinuria on Renal Response after ASCT Hypoalbuminemia & Proteinuria No Renal Response Renal Response None 39.3% 60.7% One 66.2% 33.8% Both 81.8% 18.2%

Abstract Introduction: Molecular cytogenetic studies have revealed that, to a great extent, the heterogeneity of myeloma is largely dictated by the underlying genetic and cytogenetic aberrations present in the clonal plasma cells. Most recently the group from the University of Arkansas identified strong prognostic associations with an increased level of gene expression of a cell cycle associated gene, CKS1B. CKS1B favors cell cycle progression by promoting degradation of p27 with release of the cyclin dependent kinases and entry into mitosis. Patients and methods: To further test this hypothesis we studied: a) via FISH for CKS1B amplification in a cohort of patients treated at the Mayo Clinic with high dose chemotherapy and stem cell support, as well as a group of patients with cytogenetically defined hypodiploidy, and b) a cohort of myeloma patients that were studied by gene expression profiling. Gene expression analysis was performed on CD138-enriched plasma-cell RNA using Affymetrix U133A chips (Affymetrix, Santa Clara, CA). Results: Of 159 patients studied 46 exhibited FISH abnormalities consistent with increase number of signals for CKS1B (30%): amplification was marginal or low in 44 cases, and in only two cases the ratio between CKS1B and control probe was greater than 2.0. Therefore, the predominant pattern is one of gene duplication rather than amplification where gene copy usually exceeds 10 and ratio with control probe exceeds 5 (e.g. HER-2/neu amplification). Patients with CKS1B duplication had a higher prevalence of chromosome 13 deletion (72%), and t(4;14) (29%). The presence of CKS1B duplication was more frequent (53%) among 19 patients with hypodiploidy. Using gene expression data for CKS1B, we found a positive association with t(14;16)(q32;q23). The level of CKS1B gene expression positively correlated with the PCLI (p<0.0001) but there was heterogeneity in this relationship (r2 = 0.34). We found no correlation between the expression level of CKS1B and p27 (r2 = 0.008, p = 0.37). While CKS1B gene duplication had a negative effect on survival this effect was weak (29.9 versus 38 months, p = 0.124. Median follow-up is 55 months) and disappeared when the variable was entered into the multivariate model including PCLI, B2-microglobulin and all the major genetic abnormalities by FISH. Likewise the level of expression of CKS1B determined by gene expression only carried prognostic significance when extreme levels of expression were utilized as a prognostic variable. Discussion: In this study we show that increase copy number of CKS1B is present in one third of patients with MM (majority of these being gene duplication) and seem to be associated with a shorter overall survival, but the net effect of this is rather weak in our series. Further study is needed to understand its potential role as a progression factor in the plasma cell neoplasms.

Abstract Background: Castleman Disease (angiofollicular hyperplasia) is a rare, lymphoproliferative disease. There has been no large study describing the natural history of this disease. Hypothesis: A retrospective analysis of clinical, pathologic, and laboratory factors predictive of outcome should be identifiable. Methods: During the period from 3/30/48 to 6/18/2002, 114 patients with Castleman Disease were seen at the Mayo Clinic (Rochester, 56; Jacksonville, 19; Arizona, 13) and at the University of Nebraska (26). Pathology, laboratory and clinical characteristics analyzed included: the presence of unicentric or multicentric disease, the pathologic variant (hyaline vascular versus plasma cell variant), co-existing POEMS syndrome, age, serum albumin, alkaline phosphatase, AST, sedimentation rate, the presence of cytopenias, organomegaly (hepatomegaly or splenomegaly), papilledema, peripheral neuropathy, sclerotic bone lesions, renal disease, B-symptoms or respiratory symptoms. The impact of these variables on overall survival from time of diagnosis was evaluated using univariate analyses, where their significance was determined based on the logrank statistic. To derive the multivariable model, the score-based model-building method was used. Based on this approach, a 3-variable model was designed, adjusting for age at diagnosis. Patients were then assigned a risk score corresponding to how many of these risk factors they possessed. A final proportional hazards model was constructed based on this risk score. Results: The median age at diagnosis was 43 (range: 5 – 78), and 48% of patients were male. Fifty-five patients had multicentric disease. Median follow-up of living patients was 5.8 years (range: 0.02 – 27). On univariate analysis, factors that predicted for overall survival included: age at diagnosis, presence of multicentric disease, presence of a monoclonal protein in the urine, co-existing POEMS syndrome, serum albumin, presence of cytopenias, organomegaly, neuropathy, thrombocytosis, and respiratory symptoms. The final multivariable model included 98 patients. Adjusting for age, the model included organomegaly, respiratory symptoms, and thrombocytosis. Patients were assigned 1 point for each risk factor they possessed based on clinical relevance and similarity of the hazard ratios for each of these 3 variables (ranging from 3.1 to 4). A final prognostic survival model was constructed using this assigned risk score. These scores were collapsed further based on observed similarities (0 factors vs. 1 or more factors). This final model yielded a hazards ratio of 5.5 (95% CI 2.3–13.4). The 10 year survival rates were 80% (95% CI: 65–98%) and 41% (95% CI: 28–59%) for the 0 factor and 1+ factor groups, respectively. The 20 year survival rates were 71% (95% CI: 52–87%) and 31% (95% CI: 19–51%), respectively. Conclusion: The most distinguishing prognostic clinical, laboratory, and pathological characteristics of Castleman’s Disease are orgnaomegaly, respiratory symptoms, and thrombocytosis. This prognostic scoring system should aid in individual cases and in assessing results of therapeutic interventions.