Citations

Abstract Reliable model systems are critical for investigating the genetic and molecular drivers of disease pathogenesis. Although the number of cell line model systems for multiple myeloma is extensive, recent studies showed patient tumors and cell lines clustering separately based on gene expression profiles. Furthermore, the hyperdiploid karyotype, present in approximately 50% of myeloma patients, is also extremely underrepresented in established human myeloma cell lines (HMCLs): only one of 59 HMCLs contains this phenotype. These data suggests that the average HMCL does not accurately represent the average myeloma patient and questions the validity of using HMCLs as model systems. However, most HMCLs were established from patients with pleural effusions or plasma cell leukemia, which are highly progressed and rare forms of myeloma. To assess how cell lines correlate with patient material, we performed a copy number abnormality (CNA) comparison between four myeloma primary tumors and their corresponding HMCLs. Here, we show that established HMCLs maintain the majority of the CNAs, contain a ten-fold greater frequency of 7q gains compared to patients, and that KP-6 is the second hyperdiploid HMCL known to date. CNAs were identified using array comparative genomic hybridization (Agilent 244k) with data analyzed at a resolution of 26kb. On average 59% (46.7-77.3) of the CNAs were shared between the tumors and their cell lines, 17% (8.5-25.9) were unique to the patient samples, and 24% (13.6-36.6) were only detected in the HMCLs. To put these differences in context we also compared the diagnostic material from which JMW-1 was established with a progression sample collected 9 months later after 3 cycles of VAD followed by an autologous peripheral blood stem cell transplant. The patient diagnostic and relapse tumors showed 58% similarity based on CNAs while the diagnostic and relapse tumors contained 8% and 34% unique events respectively. These results indicate that established HMCLs are no more divergent from their primary cells as is the same patient's relapse material. Overall the HMCLs studied appear highly representative of the patient from which they were derived, with most differences attributed to an enrichment of subclonal populations present in the primary tumor admixture. Of particular interest, ubiquitous 7q gains were an unexpected unifying event across all patient tumors and their HMCLs. We then analyzed the 729 baseline myeloma tumors with copy number data from the MMRF CoMMpass study against our panel of 59 similarly characterized HMCLs. We observed that HMCLs contained a 10-fold increase in the occurrence of 7q gains with respect to non-hyperdiploid patients with 7q gains (35/59 in HMCLs and 19/344 in patients). Because 97% of HMCLs are non-hyperdiploid and 7q gains are extremely rare in this population, over-representation of 7q gains in cell lines suggests that this event may promote extra-medullary growth and the establishment of cell lines. Profiling of patient tumors for gains on 7q may therefore increase the probability of establishing a new HMCL and help expand our model systems for understanding this disease. Further studies are required to understand the potential biological significance of this event in vitro . We also identified the KP-6 cell line as a novel hyperdiploid HMCL established from an extramedullary site of an advanced hyperdiploid patient. The presence of trisomies for at least four of the odd-numbered chromosomes is characteristic of a hyperdiploid myeloma patient. From the patient's tumor, clear trisomies were observed on chromosomes 3, 7, 11, 15, 19, and 21 and the HMCL established from this hyperdiploid patient maintains full trisomies of chromosomes 15, 19, and 21, with partial p-arm trisomies on chromosomes 7, 11, and 17. Since this is only the second bona fied hyperdiploid myeloma cell line to date, KP-6 promises to play a major role in our ability to understand the pathogenesis and progression of this prominent patient population. Overall, our research indicates that, based on CNAs, HMCLs are as representative of the primary tumor as is the same patient's tumor at a later time point. Additionally, HMCLs contain a 10-fold overabundance of 7q gains. Finally, we showed that KP-6 is the second hyperdiploid HMCL known to date. Taken together, these results suggest that HMCLs are reliable model systems for investigating the underlying biology of multiple myeloma. Disclosures No relevant conflicts of interest to declare.

Abstract Background: Historically, CLL patients with cytopenia were thought to have a poor prognosis and cytopenia is a key feature of CLL clinical staging systems. The implications of cytopenia in CLL could have changed because of improved diagnostic accuracy, increased recognition of early stage disease, and better treatments. Newer data suggest that the etiology of cytopenia determines its prognostic importance in CLL. Patients with cytopenia caused by autoimmune disease (AID) are less likely to have the poor prognosis associated with patients with bone marrow (BM) failure. To determine the prognostic significance of cytopenia in patients with CLL, we performed an observational study at Mayo Clinic Rochester (MCR). Methods: We studied all patients with CLL seen in the Division of Hematology at MCR from 1 January 1995 to 31 December 2004 (n = 1,750). Cytopenia (hemoglobin < 10 g/dL or platelet count < 100 × 09/L) was considered due to BM failure in patients with extensive BM involvement by CLL (< 20% residual myeloid tissue on BM biopsy) or reticulocytopenia without other causes persisting > 3 months after CLL therapy. Cytopenia was attributed to AID in patients with autoimmune hemolytic anemia (AIHA), immune thrombocytopenia (ITP), pure red blood cell aplasia (PRCA), or autoimmune granulocytopenia (AIG). Results: Median age at diagnosis of CLL was 63.7 years (69% male). Cytopenia occured in 423 (24.2%) patients. In 303 patients, cytopenia was due to CLL (228 BM failure and 75 AID). In 120 patients cytopenia was due to CLL related factors (splenomegaly, n = 11; possible AID, n = 18; treatment of CLL, n = 16) or non CLL causes (other malignancy, n = 12; iron deficiency, n = 11; anemia of chronic disease, n = 10; renal failure, n = 6; drugs effect, n = 4; surgical blood loss n = 4). In 28 (6.6%) patients the cause of cytopenia was not definitively determined. Of the patients with BM failure, 97 (43%) had cytopenia at CLL diagnosis (70% male, median age 63.3 years [range 32–95]), The AID patients the median age of CLL diagnosis was 66.7 years (range 30–85) with 79% male. AID diagnoses were AIHA (n = 41), ITP (n= 35), PRCA (n=8), and AIG (n=3); 9 patients had > 1 AID. Survival from diagnosis of cytopenia was significantly better for patients with AID (median 9.1 years) than BM failure (median 4.4 years, p < 0.001). AID patients had a longer survival from diagnosis of CLL (median 12.4 years) versus the rest of the CLL population (median 9.5 years, p = 0.020) but patients with BM failure at diagnosis had a worse prognosis (median 6.2 years, p < 0.001). Conclusion: We show that the etiology of cytopenia in CLL patients determines its prognostic importance. Anemia and thrombocytopenia caused by BM failure, but not AID, are associated with a significantly worse prognosis. These data show the importance of accurate determination of the etiology of cytopenia in all patients with CLL. Cytopenia in patients with AID is not a marker of advanced stage disease; these patients often need different management compared to those with BM failure. Cytopenia due to AID versus BM failure can be difficult to distinguish and a BM study should be done in all patients considered for treatment. Patients with cytopenia due to AID cannot be meaningfully classified by current clinical staging systems. We believe that efforts to review the NCI-WG 96 criteria are timely and should consider the etiology of cytopenia in CLL patients.

Abstract Introduction Approximately 2-8% chronic lymphocytic leukemia (CLL) patients develop diffuse large B-cell lymphoma (DLBCL, so-called Richter syndrome [RS]) over long-term follow-up. The outcomes of patients with RS are variable and poorly understood. Methods CLL patients seen at Mayo Clinic from January 1995 to April 2013 who developed biopsy-proven DLBCL during follow-up were included in this analysis. Clinical characteristics, therapy and overall survival (OS) of all patients with RS were analyzed. Results One hundred and twenty patients with biopsy-proven RS were identified in the Mayo Clinic database. The median age at diagnosis of RS was 67 years (range, 29-88 years), 87 (73%) were male, and 83 (69%) had received prior CLL therapy. The median duration between diagnosis of CLL and development of RS was 3.8 years (range, 0-34.5 years). Stereotyped B-cell receptors were found in 9/30 (30%) RS patients compared to 219/1294 (17%) non-RS patients (OR=2.1, p=0.07). IGHV4-39 gene usage was present in 2/26 (8%) RS patients compared to 57/1354 (4%) non-RS patients (p=0.3). Of the 30 RS biopsy samples in which Epstein Barr virus (EBV) in-situ hybridization studies were performed, EBV was detected in 7 (23%) patients. After a median follow-up of 5.2 years, 86 (72%) patients have died. Causes of death were as follows: progressive DLBCL (n=46, 53%), progressive CLL (n=2, 2%), infection (n=8, 9%), unknown (n=20, 23%) and other (n=10, 12%). The median OS of the entire cohort was 0.9 years (95% CI, 0.5-1.3 years). Tsimberidou and colleagues proposed a prognosis score (RS score) based on the following variables at the time of RS diagnosis: Eastern Co-operative Group performance status of 2-4, lactate dehydrogenase >1.5 x normal, platelets<100 x 109/L, tumour size >5 cm and >1 prior therapy for CLL. Each parameter was assigned one point, and patients were then assigned to one of four risk groups based on their calculated RS score: 0 or 1, low risk; 2, low-intermediate risk; 3, high-intermediate risk; 4 or 5, high risk. The OS according to the RS score for our group of patients is shown in Figure 1. Median survival in the low, intermediate-1, intermediate-2 and high-risk groups were 2.4 years, 0.7 years, 0.3 years and 0.2 years, respectively (p<0.001). Fifteen (12%) patients received cyclophosphamide, doxorubicin, vincristine and prednisone (CHOP) as initial therapy for RS, and 68 (57%) received rituximab in addition to CHOP (R-CHOP). There was no difference in survival among patients who received CHOP compared to R-CHOP (median, 1.8 years. vs. 1.2 years, p=0.5). Twenty-four patients (20%) received non-CHOP-based chemotherapy, and treatment was unknown in 4 (3%) patients. Nine patients (7%) died before definitive therapy could be instituted. Thirteen (11%) patients were able to proceed to stem cell transplantation (SCT): 10 (8%) autologous SCT, 2 (2%) allogeneic SCT and 1 (1%) autologous SCT followed by an allogeneic SCT. Although RS patients who underwent SCT were younger compared to those who did not undergo SCT (median, 59 years. vs. 69 years, p=0.001), no differences in prior CLL therapy or receipt of R-CHOP therapy for DLBCL was noted between the two groups. The median survival of patients who proceeded to SCT was superior compared to those who did not (5.2 years vs. 0.7 years; p=0.005). Conclusion Stereotyped B-cell receptors but not VH4-39 had a suggested evidence of an increased risk of RS in our cohort. Although 90% RS patients received intensive chemotherapy, only 11% patients were able to proceed to SCT. Median survival of RS patients remains dismal at 1 year, and newer therapies are needed to improve outcomes. Disclosures: Shanafelt: Genentech: Research Funding; Glaxo-Smith-Kline: Research Funding; Cephalon: Research Funding; Hospira: Research Funding; Celgene: Research Funding; Polyphenon E International: Research Funding.

Abstract Background: Cancer diagnosis and treatment are important risk factors for developing clinical depression. Validated tools for screening distress and depression, such as Cancer Distress Thermometer (DT) and PHQ9 (Patient Health Questionnaire), are underutilized, despite endorsement by NCCN and Institute of Medicine. We investigated patient and treatment characteristics as well as patient endorsement of depression or anhedonia to predict those at risk of having depression. Methods: The PHQ9 and/or DT were administered prospectively to patients with hematologic malignancies (HM) before they started antineoplastic therapy at Mayo Clinic in Florida. Patient endorsement of depression or anhedonia was collected from the current visit information survey. Patient demographics, disease and treatment characteristics, chronic medication burden, Charlson comorbidity index, living situation, clinic/hospital visit burden in the month prior to screening and number of psychiatric medications for every patient were recorded. Intergroup comparison of categorical and continuous variables was done by Chi-square and Wilcoxon rank-sum tests, respectively. Linear or logistic regression models were used to compare PHQ9 score with DT (continuous) or endorsing depression or anhedonia (categorical) respectively. Multivariate models were constructed using the stepwise selection technique using all potential variables in the models. All analyses were completed using SAS v9.3. Results: Final analysis included 246 patients with a median age at diagnosis 64.5 (range: 18-94) years, diagnosed between 6/30/93-10/9/14 and screened between 1/13/11-2/13/15. PHQ9 score of ≥9 and DT score ≥5 suggested a high risk of depression and distress, respectively, as per published literature. Patient characteristics at time of survey and analysis are noted in table 1. PHQ9, DT and answers to two questions about depression and anhedonia were available on 129, 129 and 246 patients, respectively. 63% of patients were chemotherapy naïve. In multivariate analysis, PHQ9 score ≥9 was associated with living alone (p=0.003) (Fig.1a) and non-White race (p=0.043) (Fig.1b), while a DT score ≥5 was associated with being currently married (p=0.048) and female gender (p=0.02). The only characteristic significant on univariate but not on multivariate analysis being chemotherapy naive, associated with a DT score ≥5 (p=0.049). Answering "no" to both the questions regarding depression or anhedonia was significantly associated with a low score on PHQ9 (p=0.007). Age at diagnosis, Charlson comorbidity score, chronic medication or visit burden, daily psychiatric medication use or type of malignancy were not associated with scores on any screen. Conclusions: Causes of depression in patients with HM have not been fully explored. We validated previously known risk factors for depression, such as living alone. We also reported for the first time that non-White race independently predicts depression in these patients. Female patients and those currently married are at a higher risk of psychological distress, possibly due to fear of abandoning family. We also found that simply asking a patient two questions about feelings of depression or anhedonia significantly correlates with the well-established PHQ9. Our analysis provides simple tools and reveals at-risk patient subgroups with HM where depression and distress screening should be aggressively instituted for better resource utilization and survivorship. Table 1. Patient Characteristic N % Gender Male Female 146 100 59.4 40.6 Race White Non-White 216 30 87.8 12.2 Marital Status Married Not married* 180 66 73.2 26.8 Living Situation Alone With others 37 209 15 85 Type of Malignancy Aggressive lymphoid Indolent lymphoid Aggressive myeloid Indolent myeloid 124 94 25 3 50.4 38.2 10.2 1.2 Prior Cancer Treatment Yes No 91 155 37 63 Daily Psych Meds Yes No 52 194 21.1 78.9 Patient Status Alive Dead 191 55 77.6 22.4 *Not currently married=single, divorced, widowed or unknown Disclosures No relevant conflicts of interest to declare.

Abstract Abstract 92 BACKGROUND: All somatic cell types, including all B lineage cells, constitutively express a number of DNA repair proteins to maintain genomic stability and thwart tumorigenesis in the face of ongoing constitutive levels of DNA damage. Compromises in expression of these essential DNA repair factors have been shown to readily induce the development of various cancers including B cell lymphomas, suggesting expression levels are tightly regulated and limited in quantity. In germinal center (GC) B cells, however, additional DNA repair capacity is likely required to counterbalance the heightened level of mutagenic activity owing to the induced expression of activation-induced cytidine deaminase (AID) during somatic hypermutation (SHM). AID is a DNA editing enzyme which introduces somatic mutations into immunoglobulin (Ig) V regions at an estimated rate which is almost a million-fold higher than the spontaneous rate in somatic cells. HYPOTHESIS: We hypothesize that to maintain the genomic wellness of GC B cells which are undergoing SHM, there is a need for induction of an accompanying robust DNA repair system. We further hypothesize that inefficient induction of these repair genes may predispose to malignant transformation. METHODS: Using tonsillar tissue sections and purified tonsillar B cell subpopulations, we compared expression levels of various DNA repair genes and related these levels to AID expression across the subsets using immunohistochemistry, real-time RT-PCR, and Western blot analysis. To characterize the nature of signals capable of inducing expression levels of AID and/or DNA repair proteins, peripheral blood B cells were activated in vitro using a panel of stimuli, including coculture with activated CD4 T cells. As a surrogate measure of mismatch repair (MMR) activity in the relative absence of T cell help, we quantitated the number of somatic hypermutations at A/T sites in the Ig heavy chain variable (IGHV) region genes in a collection of IGHV sequences obtained from normal B cells and HIV-related lymphoma cells. RESULTS: Using immunohistochemistry, we observed that, similar to the expression of AID, DNA MMR genes are significantly induced in tonsillar GC B cells. These results were further validated using a more sensitive real-time RT-PCR assay and analysis by Western blotting. By expanding our DNA repair gene panel, we observed that proteins of homologous recombination, base excision repair and DNA single strand break signalling pathway are also similarly induced in GC B cells at RNA, protein, and functional levels compared to their expression in naïve and memory B cells. By contrast, expression of non-homologous end joining and DNA double strand break signalling molecules are unchanged. We have termed this selective induction of repair mechanisms in GC B cells as somatic hyperrepair (SHR). To identify pathways that lead to the activation of AID and SHR, we used an in vitro system and a variety of stimuli and we discovered that multiple B cell stimuli including CpG, CD40L, and anti-BCR could each independently induce the expression of AID while SHR induction strictly required the engagement of CD4+ T cells. This provocative observation suggests a novel role for CD4+ T cells in mitigating tumorigenesis of post-GC B cells through their ability to induce the SHR pathway in cells that have been induced to undergo SHM. To demonstrate the possible role of SHR in lymphomagenesis, we analysed the mutation pattern of IGHV genes from a panel of B cell lymphomas obtained from HIV infected (CD4+ T cell suppressed) patients. We found that HIV-related lymphoma cells displayed a significantly lower frequency of SHM at A/T positions relative to normal memory B cells, indicative of compromised MMR of their precursor cells during GC transit. Our findings resolve the lingering paradox that B cell malignancies are overwhelmingly prevalent under T cell suppression conditions such as HIV infection, post-organ transplant, and aging. Finally, our results also suggest for the first time that mounting efficient tumor suppression for some cells may depend on signals transmitted by neighboring cells and the specific microenvironment. Disclosures: No relevant conflicts of interest to declare.

Abstract Abstract 669 Background: Chronic lymphocytic leukemia (CLL) is the most common leukemia in the Western world but its pathogenesis remains largely unknown. More recently, a widely prevalent premalignant condition termed monoclonal B cell lymphocytosis (MBL) has been defined and similarly involves expansion of a CD19+CD5+ population of B cells. MBL has been the focus of a wide number of recent studies in hopes it can provide insight into the early pathological events that lead to clonal expansion of a pre-leukemic CLL-like clone. Previously, we and others identified the transcription factor lymphocyte enhancer-binding factor-1 (LEF-1) as one of several genes significantly over expressed in CLL B cells relative to blood CD19+ B cells from healthy adults. LEF-1 is crucial for the proliferation and survival of pro-B cells during B cell development and deregulated LEF-1 activation has been directly linked with leukemogenesis in a transgenic murine model. In the present study, we addressed three critical questions concerning LEF-1 and CLL: 1) do CD5+ normal B cells express LEF-1; 2) at what stage of transformation does LEF-1 expression first appear; and 3) what mechanism(s) underlies LEF-1 expression in leukemic B cells. Methodology: The goals of our study were to: 1) determine the LEF-1 expression status of normal CD 19+/CD5+ human B cells; 2) determine the expression status of LEF-1 in clonal B cells present in patients with MBL; and 3) identify regulatory mechanisms that control aberrant expression of LEF-1 in CLL B cells. In order to achieve the first two goals, a 3-color flow cytometry assay for CD19, CD5, and intracellular LEF-1 was developed. To achieve the third goal, we performed in silico analysis of the LEF-1 locus and correlated this with publicly available genome wide methylated CpG island recovery assay (MIRA) data describing the methylome of normal human B cells (Rauch, T. A. Proc Natl Acad Sci U S A. 2009; 106(3): 671-8). Results: Analysis of human umbilical cord blood B cells, a rich source of the CD19+/CD5+ B cell subset, revealed that all normal CD 19+/CD5+ human B cells are negative for LEF-1 protein expression. These data demonstrate LEF-1 expression by CLL B cells is truly aberrant and does not simply reflect the phenotype of CD5+ lineage B cells. We next tested if the CD19+/CD5+ cells obtained from patients with MBL express LEF-1 protein. In these analyses, we restricted our study to MBL patients with absolute B cell counts of less than 2.5 ×10 9 cells/(L) (range: 0.756 - 2.44 × 10 9 cells/(L)), which is significantly below the upper limit that defines the MBL to CLL transition. Of interest, each MBL sample analyzed (n=6) revealed the presence of two populations of cells: 1) CD19+/CD5+ B cells expressing LEF-1; and 2) CD19+/CD5- B cells lacking expression of LEF-1. To confirm the clonal nature of the CD19+/CD5+ cells from MBL patients we demonstrated that the cells were light chain restricted. These data clearly indicated that LEF-1 expression becomes deregulated in the premalignant state of MBL and may therefore represent an early event in CLL leukemogenesis. In order to identify differential pathways of LEF-1 regulation that may be altered in CLL and MBL B cells, in silico promoter analysis of LEF-1 was performed. We identified a number of potential transcription factor binding sites as well as a putative CpG island in the 5' promoter region of LEF-1. Using data from a human B cell genome wide methylation array, we were able to determine that this same putative LEF-1 promoter CpG island was highly methylated in normal human B cells. Methylation of promoter region CpG islands is known to play an important role in developmental regulation of gene expression and may be the operative mechanism underlying silencing of LEF-1 expression in normal B cells. Conclusions: LEF-1 expression is deregulated in MBL and appears to be an early and possibly key event in the transition of normal B cells into a premalignant/malignant state. Our data suggest that loss of epigenetic regulation of this developmentally important locus may play a role in aberrant LEF-1 expression in MBL and CLL B cells. Ongoing studies are aimed at determining the methylation status of the LEF-1 promoter in CLL B cells and the functional role of this protein in transcriptional regulation and survival of CLL and MBL cells. Disclosures: No relevant conflicts of interest to declare.

A variety of obstacles can prevent persons or individuals with low vision from deriving the greatest possible benefit from the rehabilitation process, including inadequate understanding of their visual impairment, lack of knowledge about available services, and misconceptions about low vision devices. This study explores the use of a patient-education consultation to enable patients and their families to take better advantage of low vision services.

5574 Background: A noninvasive cell-free DNA (cfDNA)-based cancer detection assay offers the hope of a blood test that might reduce morbidity and mortality of cancers, particularly those without recommended screening tests (eg, some gynecologic cancers). CCGA (NCT02889978) is a prospective, multi-center, longitudinal, case-control study evaluating models for discriminating cancer versus non-cancer. Here, we report F/U of control participants (pts) who demonstrated a cancer-signal in CCGA. Methods: Clinically evaluable samples (N = 2508) from pts enrolled without a cancer diagnosis (dx; NC) and treatment-naive pts with newly diagnosed cancer (C) were divided into training (n = 1564; 580 NC, 984 C) and test (n = 944; 368 NC, 576 C) sets. Classification performance (cancer/non-cancer) was assessed via 3 prototype assays: whole-genome bisulfite (WGBS), whole-genome (WGS), and targeted (507 gene) sequencing. Notable outlier NC pts were identified with cancer-like scores in either ≥2 assay classification results or by the presence of known cancer drivers with ≥1 assay classification result suggesting cancer. All pts are currently in F/U in accordance with study protocol (to date: 80% with > 10 mo and 15% with > 22 mo F/U). Results: Among training and test sets, 8 ( < 1%) NC pts were identified with a cancer-like signal. To-date, 2 have been diagnosed with a gynecologic malignancy: 1 stage IIIc clear cell endometrial carcinoma and 1 stage IIIc ovarian cancer, 3 and 2 months (mo) post-enrollment [PE], respectively. Among C pts in the study, sensitivity (at 98% specificity; WGBS) in these cancer types was: uterine/endometrial: 11% (n = 27 train) and 22% (n = 9 test); ovarian: 82% (n = 17) and 71% (n = 7). In addition, a third NC pt was diagnosed with a stage IV lung cancer 15 mo PE. Conclusions: This cfDNA-based assay detected a cancer-like signal that anticipated a clinical presentation of cancer in undiagnosed pts as early as 15 months prior to the actual dx. High specificity ( > 99%) requires accounting for undiagnosed cancers in study design and analysis. Together, these data suggest that this prototype assay may have high performance detecting a variety of gynecological and other cancers. Clinical trial information: NCT02889978.

3049 Background: For multi-cancer detection using cfDNA, TOO determination is critical to enable safe and efficient diagnostic follow-up. Previous array-based studies captured < 2% of genomic CpGs. Here, we report genome-wide fragment-level methylation patterns across 811 cancer cell methylomes representing 21 tumor types (97% of SEER cancer incidence), and define effects of this methylation database on TOO prediction within a machine learning framework. Methods: Genomic DNA from 655 formalin-fixed paraffin-embedded (FFPE) tumor tissues and 156 isolated cells from tumors was subjected to a prototype 30x whole-genome bisulfite sequencing (WGBS) assay, as previously reported in the Circulating Cell-free Genome Atlas (CCGA) study (NCT02889978). Two independent TOO models, one with and one without the methylation database, were fitted on training samples; each was used to predict on the test set. A WGBS classifier was used to detect cancer at 98% specificity; reported TOO results reflect percent agreement between predicted and true TOO among those detected cancers (166 cases: 81 stage I-III, 69 stage IV, 16 non-informative). Results: Genome-wide methylation data generated from this database allowed fragment-level analysis and coverage of ~30 million CpGs across the genome (~60-fold greater than array-based approaches). Incorrect TOO assignments decreased by 35% (20% to 13%) after incorporating methylation database information into TOO classification. Improvement was observed across all cancer types and was consistent in early-stage cancers (stage I-III). Respective performances in breast cancer (n = 23) were 87% vs 96%; in lung cancer (n = 32) were 85% vs 88%; in hepatobiliary (n = 10) were 70% vs 90%; and in pancreatic cancer (n = 17) were 94% vs 100%. Results using an optimized approach informed by these results in a large cohort of CCGA participants will be reported. Conclusions: Incorporating data from a large methylation database improved TOO performance in multiple cancer types. This supports feasibility of this methylation-based approach as an early cancer detection test across cancer types. Clinical trial information: NCT02889978.