Large Granular Lymphocyte Leukemia Research Articles

Automated Peripheral Smear Recognition and Classification of LGL Leukemia Using Machine Learning

Background: Machine learning can augment clinical decision support, especially in cases where complex diagnostic criteria are necessary. In making a diagnosis of LGL leukemia, a pathologist must review the peripheral smear thoroughly, which is a tedious process. Supervised machine learning has been successfully implemented in such cases by training on very large image-based datasets, followed by a corrective phase of updating the model weights and finally demonstrating a high degree of successful predictions in an independent dataset. We sought to assess whether a supervised machine learning model can automate histopathological image analysis and provide a degree of confidence in supporting a pathologist's manual review. Methods: We searched the published literature from 2016-2022 for high-definition peripheral smear images with confirmed cases. The samples were downloaded, pre-processed with the Segment Anything Model (SAM), and labeled as training, testing, or validation data. Non-pathological controls, defined as peripheral smears confirmed negative for LGLL, were also obtained from published literature and labeled manually. The neural network was written and trained on TensorFlow using the open-source Keras library with parallel threading and on a multicore GPU machine. The resulting predictive model had a binary classification task: Does the presented image have cells that resemble LGL cells, returning true or false. Every “true” label returned from the classifier also creates a stack of probable positives that need pathology review. Feedback from each pathologist review further updates the model weights to increase the probability of successful prediction in the future. Results: We screened 57 publications to obtain 163 high-definition peripheral smear images that had confirmed LGLL and 100 non-pathological smears. Two cohorts were created: 200 smears in a training cohort comprised of 100 LGLL smears and 100 non-pathological smears, and 63 smears in the validation cohort. To remain agnostic towards image orientation, we created 7 views of each smear, and 1400 total images underwent feature extraction using SAM. The resulting data was labeled and used for model training and our process is depicted visually below in Figure 1. During validation, our model accurately predicted LGL-cells in 56/63 cases with an 8% false positive rate, 2% false negatives and 2% unclassifiable. After updating model weights and re-training with Adaptive Boosting, the accuracy increased to 60/63 smears with a false-positive rate of 3%. Our trained classification model achieved an AUC of 0.8144. Conclusion: Supervised machine learning model with Adaptive Boosting can be used to label a peripheral smear image with a high likelihood of containing LGL cells and, in turn, help a pathologist review smears with higher priority in a timely fashion. Our initial training data was limited, which hindered the performance of our model; however, with a sufficiently large training dataset, we can drastically improve our accuracy. A more extensive training set with robust hardware can transform our image-recognition model into a foundational model where other clinical parameters can be added, and the final diagnostic probability will be based on multiple data streams. Our work takes a meaningful step in this direction, and future work at Karmanos Cancer Center will focus on extending our classifier to more sophisticated deep-learning algorithms. Our open-source platform will eventually support several niches in the future that are currently only served by commercial applications.

Incidence, Clinical Features, and Outcomes of Large Granular Lymphocyte Leukemia

Introduction: - Large Granular Lymphocyte (LGL) leukemia is a rare and diverse hematologic disorder characterized by the clonal proliferation of large granular lymphocytes in the peripheral blood and bone marrow. Despite its rarity, population-based studies on LGL leukemia remain limited. This study aims to comprehensively analyze clinical characteristics and outcomes among patients with LGL leukemia. Methods: - We utilized the Surveillance, Epidemiology, and End Results (SEER) database 17 registry to identify patients diagnosed with large granular lymphocytic leukemia between 2000-2020. Data analysis was conducted using SEER*Stat version 8.4.1.2 and STATA. Univariate survival analysis was done using a Kaplan Meir plot with a log-rank test, and the Cox proportion hazard regression model was used for multivariate survival analysis. Results: - We identified a total of 2,589 LGL leukemia cases during the study period. The calculated incidence of LGL leukemia was 0.14 cases per 1,000,000 individuals (range: 0.14 - 0.15). Male patients exhibited a higher incidence rate ratio of 1.19 (95% CI: 1.11 - 1.27, p < 0.01) than females. However, females tended to be diagnosed younger than males (67 vs. 68, p < 0.01). Most cases were observed in the White population (83%) compared to other racial groups. The median age at diagnosis was 67 years (ranging from 7 to 85 years). Patients aged above 70 years showed the highest incidence of LGL leukemia (0.80) in contrast to those aged 45-54 years (0.30) and 55-64 years (0.63). Notably, there was a significant increasing trend in LGL leukemia incidence, with an incidence rate ratio of 13.50 (95% CI: 10.5-17.6, p < 0.001) in the period 2015-2019 compared to 2000-2004. Approximately 32.7% of patients received chemotherapy; the median time from diagnosis to treatment was one month. The survival rates at 12, 36, and 60 months for LGL leukemia patients were 88.9%, 76.5%, and 66.5%, respectively, with a median follow-up of 40 months. Multivariate analysis revealed that females had a more favorable survival outcome than males (HR: 0.78, 95% CI: - 0.64 - 0.97, p: 0.02). Moreover, patients who underwent chemotherapy exhibited better survival than those who did not (HR: 0.72, 95% CI: 0.56 - 0.93, p: 0.01). Elderly patients above 60 years of age had poorer survival rates (HR: 3.1, 95% CI: 2.34 - 4.01, p < 0.01), and no significant difference in outcomes was observed based on the timing of chemotherapy initiation from the time of diagnosis. Conclusion: - This study represents the most extensive and most recent population-based investigation of LGL leukemia. The findings highlight the potential benefits of chemotherapy in improving survival and indicate better outcomes in female patients. Further research is warranted to validate these promising results and gain deeper insights into the management and prognosis of LGL leukemia.

LGL in MGUS or MGUS in LGL? Lessons of Nature and Their Clinical Implications

Associations between seemingly unrelated entities often reveal important lessons of nature, instructive for the understanding of mutual pathogenic roots. While investigating the clinical features of T-cell large granular lymphocytic leukemia (LGL), we noted the frequent presence of plasma cell dyscrasias/monoclonal gammopathies (MGs), chiefly MGUS. There are several potential explanations for this co-occurrence: 1) similar demographics (both conditions mostly presenting in the elderly); 2) the possibility that the two precursors originate from an initial polyclonal immune response to a common target; and 3) as alternative, that monoclonal plasma cells or the Ig product may be the targets of cytotoxic T-LGL cells in a context of tumor surveillance reaction. For instance, clonal TCR may recognize the M protein idiotype or vice versacausing aberrant activations. In addition, while MGUS should not be associated with cytopenias, responses to anti-myeloma therapies have been observed in LGL with MGUS refractory to traditional immunosuppressive therapies. We hypothesized that the study of clinical and epidemiologic features of MGs and LGL may be helpful to elucidate some of these fundamental questions. To address this, a dual approach was applied: 1) MG-forward strategy: screening of LGL in 2348 individuals with documented diagnosis of MGs and available bone marrow exam; 2) LGL-forward strategy: search of MGs in 235 LGL cases with M protein assessment. Overall, among 2348 cases with MGs, we identified 42 cases (1.8%) with concurrent evidence of LGL. We then selected an exploratory cohort of 787 MG patients with clinical annotations for further phenotypic dissection and comparative analyses. Within 533 patients with non-active MGs (i.e., MGUS or smoldering myeloma -SM), 144 (27%) had cytopenias of any type and grade, which could not be attributable to plasma cell infiltration. Screening for comorbid conditions explaining cytopenias left 32 cases (22%) presumed to be idiopathic. LGL was overrepresented in MGUS-SM with cytopenias (6 vs 0.3% in cases without counts alterations, p< .001). This frequency was higher when considering only cases with idiopathic cytopenias (22 vs 1.8% in cases with cytopenias of other causes, p< .001) ( FigA). Reverse analysis of the LGL cohort demonstrated that 71 (30%) LGL cases had M proteins. Overall, the MG/LGL dyad was more frequently associated with concurrent autoimmune conditions (54%), as well as monoclonal B-cell lymphocytosis and other B-cell lymphoid neoplasms (19%). Next, we compared cases with MG/LGL vs those with MGs (N = 775) or LGL (N = 157) exclusively. When compared to the MGs cohort, MG/LGL cases presented a younger age (median 64 vs 70 years, p < .001), had higher frequency of M protein of the IgM type (21 vs 10%, p= .002), and oftentimes exhibited small/transient paraproteinemia (42 vs 2%, p< .001). No specific features were enriched in the MG/LGL cases when compared to the LGL cohort, except for a higher proportion of proliferations of NK-cell subtype (12 vs 4%, p= .02). Finally, to confirm the observed association, we focused on the subgroup of non-active MGs by excluding cases with initial diagnosis of MM or lymphoma. Multivariable analysis (adjusting by sex, age and MG risk) revealed a lower risk of immunoparesis (aOR: 0.4, 95%CI: 0.2 - 0.7, p= .002) in MG/LGL as opposed to MG without LGL, whereas no effect was found in progression to MM/lymphoma (aHR: 0.9, 95%CI: 0.4 - 2.1, p= .81 in a competitive-risk setting). Moreover, when opposed to its MG and LGL counterparts, MG/LGL more frequently presented with severe cytopenias (aOR vsMGs: 28.9, 95%CI: 14.0 - 59.0, p< .001; aOR vsLGL: 2.7, 95%CI: 1.7 - 5.6, p= .008), and with immune-mediated cytopenias (aOR vsMGs: 10.5, 95%CI: 4.9 - 22.2, p< .001; aOR vsLGL: 2.8, 95%CI: 1.0 - 4.8, p= .04) ( FigB). In conclusion, our study reveals that: 1) the association between MGUS and LGL is more than an age-related epidemiological coincidence; 2) distinctive nosologic features for the MG/LGL dyad; and 3) that even “non-active” MGs associated with LGL might be of hematological clinical significance through humoral or cellular immune mechanisms. These findings point towards alternative treatment strategies of cytopenias in MGUS and LGL (immunosuppression vs anti-myeloma therapies, respectively), and provide impetus for the investigation of tumor surveillance in MGUS vs common etiologic factors intertwining these clonopathies.

Ruxolitinib Promotes Clinical Responses in Large Granular Lymphocytic Leukemia Via Suppression of JAK/STAT-Dependent Inflammatory Cascades

Background: Large granular lymphocytic leukemias (LGL) are T or NK-cell lymphoproliferative disorders that are characterized by myelosuppression and frequent mutations in STAT3. Commonly used agents include low doses of methotrexate or cyclophosphamide. Among 5 patients (pts) with LGL treated on our phase II study of ruxolitinib in T-cell lymphoma (Moskowitz, et al. Blood 2021), 2 achieved partial response (PR) and 2 achieved stable disease, suggesting oncogenic dependence on JAK/STAT pathway activity in LGL. Methods: To further explore the efficacy and mechanism of JAK1/2 blockade in LGL, we expanded our phase II study to include 23 pts with relapsed/refractory LGL. This sample size was chosen to allow for narrowing of the 95% confidence interval to 54-87% (Wilson score method) if similar efficacy was observed. Pts received ruxolitinib 20mg PO BID continuously. Responses were assessed after cycles 2, 5, and every 3 cycles thereafter using the response assessment from the ECOG-5998 study for LGL. Furthermore, stable disease (SD) was defined as improvement in cytopenia(s) by 1 grade (without meeting criteria for PR) for a minimum of 6 months. Overall response rate (ORR) included complete response (CR) plus PR; clinical benefit rate (CBR) included ORR plus SD. In parallel, we performed single-cell RNA sequencing (scRNA-seq) on paired C1D1 and C1D8 peripheral blood mononuclear cells (PBMCs) from 7 patients (5 PR, 1 SD, and 1 non-responder) and plasma proteomic profiling on 5 patients (3 PR, 1 SD, and 1 POD). Results: Among 23 pts, median age was 58 (31-86), 13 (57%) were male, 19 (83%) were white, 1 (4%) was black. Median number of prior therapies was 2 (range 0-5). 12/22 (54%) had STAT3 mutations (of whom 1 also had a JAK1 mutation); 1 had a JAK2 mutation, 1 pt was missing sequencing. Baseline cytopenias included anemia (n=14, 61%), thrombocytopenia (n=12, 52%), and/or neutropenia (n=16, 70%). Among 20 pts evaluable for response (3 too early), there were 5 (25%) CRs, 6 (30%) PRs, and 4 (20%) SD. ORR and CBR were 55% and 75% respectively. Median event free survival (EFS) was not reached and was 68% (95% CI: 50-93) at 14 months ( Fig. 1A). STAT3 mutation status predicted for improved EFS (14-month EFS 100% vs 40%, p=0.007). After a median duration of treatment of 13.1 months, there were no unexpected toxicities. Severe adverse events included febrile neutropenia (n=3, 13%) and herpes zoster (n=1, 4%). To understand the mechanism by which ruxolitinib improved hematologic parameters in LGL, we performed unbiased proteomic analysis of peripheral blood plasma samples using the OLink 96 Inflammation and Immune Response Platforms. Ruxolitinib led to a decrease in several soluble mediators known to be driven by interferon-driven JAK/STAT activation, including CXCL9, CXCL10, and PD-L1, as well as TNF ( Fig. 1B). Several of these factors have been shown to suppress hematopoietic stem cell function, suggesting a JAK/STAT-dependent paracrine mechanism by which LGL leads to myelosuppression. ScRNA-seq of PBMCs identified LGL cells based on expression of cytotoxic genes, including NKG7, CST7, CCL5, GZMB and PRF1. Analysis of cell type proportions within the peripheral blood showed that patients achieving a PR to Ruxolitinib showed a decrease in the proportion of circulating myeloid cells. Gene set enrichment analysis (GSEA) of LGL cells revealed substantial downregulation of both JAK/STAT signaling and TNF signaling in LGL cells in PR but not SD patients, whereas persistently elevated TNF signaling was seen not only in LGL cells but also circulating myeloid and naïve T-cells in SD patients, suggesting that LGL cells might promote bone marrow suppression via paracrine effects on the bystander host immune system. Network analysis of LGL cells and non-malignant peripheral blood immune cells is ongoing. Conclusions: Ruxolitinib induced durable responses in pts with LGL and therefore may represent an option for pts requiring treatment. STAT3 mutation status was predictive of EFS and responses were associated with JAK/STAT pathway down-regulation, demonstrating on-target effects. Integrated immune profiling suggests that ruxolitinib may interrupt JAK/STAT-dependent paracrine inflammatory signals that promote bone marrow suppression. An ongoing phase II study of ruxolitinib in LGL will aid in clarifying optimal dosing and confirming both efficacy and mechanism of action (NCT05592015).

Preliminary Safety, Pharmacokinetics, Pharmacodynamics and Clinical Activity of KT-333, a Targeted Protein Degrader of STAT3, in Patients with Relapsed or Refractory Lymphomas, Large Granular Lymphocytic Leukemia, and Solid Tumors

Background: KT-333is a first-in-class, potent, highly selective, heterobifunctional small molecule degrader of the signal transducer and activator of transcription 3 (STAT3) protein. Aberrant activation of STAT3 resulting from activating mutations or deregulated cytokine signaling underlies various malignancies including peripheral T-cell lymphomas (PTCL), cutaneous T-cell lymphoma (CTCL), and large granular lymphocytic leukemia (LGL-L). Approximately 70% of human cancers including hematological malignancies and solid tumors exhibit increased levels of phosphorylated STAT3 (pSTAT3), a biomarker of pathway activation. In non-clinical studies, treatment with KT-333 resulted in durable tumor regressions with weekly (QW) or once every two weeks IV administration in STAT3-dependent T cell lymphomas. STAT3 degradation also sensitized immunocompetent mouse models of solid and liquid cancers to anti-PD1(ASH 2021, SITC 2021). Methods: The ongoing open-label, Phase 1a/1b study is evaluating the safety, tolerability, pharmacokinetics (PK), pharmacodynamics (PD) and preliminary clinical activity of KT-333 administered as a QW IV infusion on Day 1, 8, 15 and 22 (28-day cycle) in patients (pts) with B- and T-cell lymphomas, Hodgkin's lymphoma and advanced solid tumors (ST) relapsed/refractory (R/R) to at least two prior systemic therapies and LGL-L and T-cell prolymphocytic leukemia R/R to at least one prior therapy. Cycle 1 and 2 blood samples are collected for KT-333 plasma concentrations and to measure changes in STAT3 protein expression in peripheral blood mononuclear cells (PBMCs) using targeted mass spectrometry. Whole blood RNA sequencing measures mRNA levels of STAT3 regulated targets. Plasma levels of inflammatory biomarkers are measured with Luminex. STAT3 degradation and other related biomarker changes in tumor are assessed in patients with accessible tumors. (NCT05225584). Results: As of July 10, 2023, 21 pts were treated at five dose levels (DL) in Phase 1a with a mean number of 5.8 doses. Pts included B-cell non-Hodgkin's lymphoma (n=1: DL5), Hodgkin's lymphoma (HL) (n=1: DL4), CTCL (n=3: DL1, 2 and 4), PTCL (n=1: DL2), LGL-L (n=2: DL5) and ST (n=13: DL1-4) with median age of 61 years (range 30,77) and ECOG performance status of 0 (n=7) or 1 (n=14). No DLTs and no KT-333 related serious adverse events (SAE) were reported. The most common AEs were Grade 1 and 2 and included constipation, fatigue, nausea and anemia. Best response among pts evaluable for response at data cut-off (not including CTCL or HL pts at DL4 or any DL5 pts) included one partial response after two cycles in a CTCL pt at DL2 and SD after two cycles in three ST pts treated at DL3 and DL4. PD data in blood available for DL1-4 demonstrated robust, dose-dependent, and sustained STAT3 degradation in PBMC. The mean maximum degradation of STAT3 by targeted mass spectrometry over the first two weeks in Cycle 1 by DL was (% (range; n)): DL1: 69.9% (52.6% to 84.1%; n=4), DL2: 73.5% (65.5% to 80.7%; n=3), DL3: 79.9% (72.3% to 90.4 %; n=3) and DL4: 86.6% (78.9% to 95.9 %; n=4) with absolute quantification of STAT3 peptides falling below lower limit of quantification of the assay for one pt in DL3 and two in DL4. STAT3 pathway inhibition in blood was demonstrated via transcriptional downregulation of a canonical JAK/STAT3 target, SOCS3, which correlated with changes in STAT3 protein levels. KT-333 also resulted in dose-dependent downregulation of STAT3-regulated inflammatory biomarkers C-reactive protein and serum amyloid A protein in plasma. KT-333 demonstrated linear PK with plasma exposure increasing with dose and reaching levels close to those predicted to be efficacious. Conclusion: The emerging clinical data demonstrate that KT-333 is a potent degrader of STAT3 as demonstrated in PBMCs at doses that are well tolerated. These data provide the first evidence of STAT3 targeted protein degradation in humans with associated STAT3 pathway inhibition, along with potential early signs of antitumor activity, highlighting the potential of heterobifunctional degraders for targeting previously undruggable transcription factors implicated in diseases. Based on non-clinical data and PK/PD modeling, the high levels of degradation achieved so far are expected to be clinically efficacious in STAT3-dependent malignancies. Accrual is ongoing, and analyses from additional patients will be presented at the meeting.

Clinical Features and Outcomes in Large Granular Lymphocyte Leukemia-Associated Pure Red Cell Aplasia with STAT3 Mutation

Background Large granular lymphocyte leukemia (LGLL) is a rare disease frequently complicated with pure red cell aplasia (PRCA) [1, 2]. STAT3 mutations have been describled in 30-40% of LGLL patients [3]. The aim of this work was to evaluate whether STAT3 mutations might be associated with specific clinical features and outcomes in LGLL-associated PRCA. Methods and results From January 2016 to July 2023, 81 patients with LGLL-associated PRCA were enrolled in the China Eastern Cooperation Group for Anemia (CECGA) database (ChiCTR2100043485). As an initial step in the STAT3 mutational analysis, we screened all patients with Sanger sequencing or Next-generation sequencing (NGS). The initial dose of CsA was 3-5mg/kg/day, and the serum concentration was adjusted to be 150-200ng/mL based on adverse reactions. After 12 months of maintenance, the dosage was slowly reduced. Cyclophosphamide combined with prednisone (CP) regimen was used as a salvage therapy for patients who failed to CsA. The initial dose of cyclophosphamide and prednisone were 100mg/day and 0.5-1mg/kg/day, respectively. Among the 81 cases, 26% of them were positive for STAT3 mutations. The clinical characteristics of patients with or without STAT3 mutation were summarized in Table 1. Patients with STAT3 mutations had a higher reticulocyte percentage (0.88% vs 0.28%, P=0.039) and red cell distribution width-coefficient of variation (18.8% vs 15.8%, P=0.008) than patients without STAT3 mutations. Y640F mutation were found in 9 of 21 cases. Subgroup analysis showed that patients with Y640F mutation were associated with younger age (44 vs 65 years old, P=0.007) and higer lymphocyte percentage in peripheral blood (63.7% vs 34.4%, P=0.033). Deep sequencing of rearranged T-cell receptor Vβ complementarity-determining region 3 by NGS was conducted in 61 patients. The predominant V-gene of LGLL clonotypes belonged to the TRBV06 family gene was detected in 12 (25%) patients. The expression of TRBV06 family gene was a litter lower in patients with STAT3 mutation than non-mutant groups (8% vs 31%, P=0.189). The complete response rate (CRR) [31% (5/16) vs 33% (19/58), P=0.909] and overall response rate (ORR) [56% (9/16) vs 50% (29/58), P=0.658] of cyclosporine (CsA) treatment were similar in patients with STAT3 mutations or not. In STAT3 mutant group, the CRR [54% (7/13) vs 31% (5/16), P=0.274] and ORR [85% (11/13) vs 56% (9/16), P=0.130] of CP regimen tended to be better than CsA . 6 patients (67%) relapsed among the 9 patients who responded to CsA in STAT3 mutant group. In patients without STAT3 mutation, 19 of 29 (66%) CsA-responders relapsed. There was no siginificant difference in the relapse-free survival between the two group (Figure 1). Discussion/Conclusions In summary, STAT3 mutation was frequently recognized in LGLL-associated PRCA, and the hotspot site was Y640F. Patients with STAT3 mutations responded to CsA as well as those without the mutation. CP regimen could be used as a salvage therapy for patients who failed to CsA. Reference 1. Balasubramanian SK, Sadaps M, Thota S, et al. Rational management approach to pure red cell aplasia. Haematologica. 2018, 103(2): 221-230. 2. Wu X, Cheng L, Liu X, et al. Clinical characteristics and outcomes of 100 adult patients with pure red cell aplasia. Ann Hematol. 2022, 101(7):1493-1498. 3. Barilà G, Teramo A, Calabretto G, et al. Stat3 mutations impact on overall survival in large granular lymphocyte leukemia: a single-center experience of 205 patients. Leukemia. 2020, 34: 1116-1124.

Chronic Lymphoproliferative Disorder of Natural Killer Cells: Patient Characteristics, Laboratory Features, and Clinical Outcomes

Introduction: Chronic lymphoproliferative disorder of natural killer cells (CLPD-NK) is a rare form of large granular lymphocytic leukemia. The clinical characteristics and outcomes of patients (pts) with CLPD-NK in the literature are limited. We aim to describe the patient characteristics, laboratory features, and clinical outcomes of CLPD-NK by reporting on, to our knowledge, the largest single-center cohort of CLPD-NK in the United States. Methods: We retrospectively identified Mayo Clinic pts with CLPD-NK, defined per the WHO 5 th edition and 2022 International Consensus Classification. We conducted chart review to collect clinical and laboratory data. Overall survival (OS) was plotted using Kaplan-Meier. Time to first treatment (TTFT) was displayed using cumulative incidence methodology accounting for competing risk of death. Univariable Cox regression models investigated associations between patient characteristics and outcomes (OS and TTFT). Results: We identified 45 pts who were diagnosed with CLPD-NK between 1997 and 2022. The median age at diagnosis was 69 years (range: 25-90) and 32 pts (71%) were male. At the time of diagnosis, 10 pts (22%) had a concomitant autoimmune disorder and 9 (20%) had laboratory evidence of a serologic abnormality (monoclonal gammopathy, rheumatoid factor, etc.). In addition, 10 pts (22%) had a prior diagnosis of a hematologic malignancy and 12 (27%) had a prior diagnosis of a solid tumor. Clinically, 4 pts (9%) had B symptoms, 15 (33%) had splenomegaly, and 13 (29%) were transfusion dependent. At diagnosis, median hemoglobin (hgb) was 10.2 g/dL, with 25 pts (56%) showing a hgb <11g/dL and 8 (18%) demonstrating a hgb <8 g/dL. The median platelet count was 202x10^9/L, with 12 pts (27%) demonstrating a platelet count <150x10^9/L. The median white blood cell count was 5.8x10^9/L. The median absolute neutrophil count (ANC) was 1.2x10^9/L, with 29 pts (64%) demonstrating an ANC <1.5x10^9/L. The median absolute lymphocyte count (ALC) was 3.8x10^9/L, with 20 pts (44%) demonstrating an ALC >4x10^9/L. The median absolute abnormal NK cell count per flow cytometry was 2.2x10^9/L. Flow cytometry revealed killer cell immunoglobulin-like receptor (KIR) restriction in all pts, with 27 (61%) lacking KIR expression and 17 (39%) demonstrating KIR restriction in CD158a, CD158b, and/or CD158e. CD7 expression was present in 42 pts (93%; 31% dim/partial), CD8 expression was present in 22 (49%; 42% d/p), CD56 expression was present in 17 (39%; 23% d/p), CD57 expression was present in 26 (59%; 46% d/p), CD94 expression was present in 33 (77%; 7% d/p), and NKG2A expression was present in 21 (50%; 5% d/p). Bone marrow examination (n=30) revealed that the median number of NK infiltrate was 11% (range: 5-40%) of cellularity, with 20 demonstrating an intrasinusoidal infiltrative pattern. The median follow-up was 41 months (mo), and 23 pts required at least one line of treatment. Treatment indications included anemia (39%), neutropenia (4%), bi/pancytopenia (39%), lymphocytosis (4%), or other (13%). The median TTFT was 32 mo (95% CI 2-93 mo). Hgb <11g/dL was associated with a shorter TTFT (9.7 vs 187 mo); however, neutropenia and thrombocytopenia were not associated with TTFT. For first-line treatment, in addition to frequent steroid use, 16 pts (70%) received a methotrexate-based regimen, 6 (26%) received a cyclophosphamide-based regimen, and 1 (4%) received a cyclosporine-based regimen. The median TTFT was not reached. Of those who required systemic treatment, 9 pts required a second-line therapy. The median OS from diagnosis was 162 mo (95% CI: 84-not estimable, NE), with 33 pts alive at last follow-up (Figure 1a). Increasing age (HR 1.2, 95% CI 1.1-1.3, p=0.002) and lack of CD94 expression (HR 5.6, 95% 1.3-25.0, p=0.02) were associated with worse OS. In the 23 pts who required treatment, the median OS from treatment initiation was 84 mo (95% CI: 52-NE) (Figure 1b). Conclusions: In this large cohort of 45 pts, the majority had neutropenia and anemia, and approximately half required systemic treatment. Estimated median OS from diagnosis exceeded a decade for the entire cohort and was seven years after treatment initiation. Anemia was associated with a shorter TTFT, while increasing age and lack of CD94 expression were associated with worse OS. This rare disease warrants additional investigations to understand optimal management and improve outcomes, especially for those requiring treatment.

Inclusion body myositis accompanied with T-cell large granular lymphocyte leukemia

Inclusion body myositis (IBM) is a late-onset myopathy that manifests as distinct muscle weakness in the quadriceps, finger flexors, and ankle dorsiflexors. T-cell large granular lymphocyte (T-LGL) leukemia is a late-onset clonal disorder of CD8+ cytotoxic T-cells that is often accompanied by autoimmune diseases. To date, the association between IBM and T-LGL leukemia has been infrequently reported. Here, we report a case of a patient with T-LGL leukemia who developed IBM, along with in-depth laboratory, electrophysiological, and pathologic findings.

Analysis of Treatment Outcome in Large Granular Lymphocytic Leukemia: A Retrospective Study from India

Large granular lymphocytic (LGL) leukemia is a rare and indolent lymphoproliferative disorder that belongs to mature T and natural killer (NK) cell neoplasms, as per the World Health Organization classification. This article assesses the response to immunosuppressive therapy. In this retrospective study, various cases of chronic lymphoproliferative disorders (CLPDs) evaluated and treated in two tertiary care settings were screened and taken for subanalyses. After fulfilling the criteria for LGL leukemia, cases were further assessed for presenting features and response to treatment. Statistical analysis was performed using the Statistical Package for the Social Sciences, version 23. Out of 384 cases of CLPDs analyzed, 14 cases of LGL leukemia were identified (3.64%) and subjected to further analysis. There were six males (42.85%) and eight females (57.14%) (M: F = 1:1.33) with an age group ranging from 42 to 82 years. Thirteen cases (92.85%) were T-LGL type, and one case belonged to NK-LGL type (7.14%). Anemia was the most common presentation (92.85% of cases), followed by lymphocytosis (85.71% of cases) and neutropenia (78.57% of cases). Four patients (28.57%) presented with thrombocytopenia, and two patients presented with pancytopenia (14.28%). Splenomegaly was seen in two patients (14.28%), and lymphadenopathy in one patient (7.14%). One patient (7.14%) had rheumatoid arthritis, and one (7.14%) had recurrent chest infections. Out of the 14 cases, 5 (35.71%) attained complete remission, 5 (35.71%) attained partial remission, 2 patients were resistant to first-line treatment, and 2 patients were closely followed with observation only. The overall treatment response rate was 83.33%. This study highlights the excellent response rates with immunosuppressive therapy in a rare lymphoproliferative disorder.

A simplified prognostic score for T-cell large granular lymphocyte leukaemia

Background T-cell large granular lymphocyte leukaemia (T-LGLL) generally has a favourable prognosis, but a small proportion of patients are facing a relatively short survival time. This study aimed to identify clinical factors associated with survival in patients with T-LGLL and develop a predictive model for guiding therapeutic decision-making. Materials and Methods We conducted a retrospective study on 120 patients with T-LGLL. Lasso regression was performed for feature selection followed by univariate and multivariate Cox regression analysis. A decision tree algorithm was employed to construct a model for predicting overall survival (OS) in T-LGLL. Results The median age of diagnosis for the entire cohort was 59 years, and 76.7% of patients reported disease-related symptoms. After a median follow-up of 75 months, the median OS was not reached. The 5-year OS rate was 82.2% and the 10-year OS rate was 63.8%. Multivariate analysis revealed that an Eastern Cooperative Oncology Group performance status over two and a platelet count below 100 × 109/L were independently associated with worse OS, leading to the development of a simplified decision tree model. The model’s performance was adequate when internally validated. The median OS of the high- and intermediate-risk- risk groups was 43 and 100 months respectively, whereas the median OS of the low-risk group was not reached. Furthermore, we found that immunosuppressive agent-based conventional treatment was unsatisfactory for our high-risk patients. Conclusions Our model is an easily applicable clinical scoring system for predicting OS in patients with T-LGLL. However, external validation is essential before implementing it widely.

T-cell large granular lymphocytic leukemia in Vietnam: Using microscope-to-screen videoconferencing to improve diagnosis.

Leukemia diagnosis in Vietnam is limited by a lack of hematopathology training and expert consultation as well as the cost of high-magnification digitization of hematology slides. Screen-sharing software allows international collaboration with experienced hematopathologists for improved diagnostic accuracy. A hematopathology education and consultation program was proposed for Vietnam hospitals. By appointment, pathologists in Vietnam with access to a microscope camera, imaging software, and high-speed internet were invited to review slides and data with a volunteer board-certified hematopathologist in the United States using secure videoconferencing software. A single hospital in southern Vietnam assigned a pathologist proficient in English to access this service. All consultations from this site with clinicopathologic information were logged. After a 2-year period of online consultation, case slides for selected diagnoses were reviewed under the microscope in Vietnam to assess concordance. In total, 135 consultations were logged, 53 of which were for blood and bone marrow. T-cell large granular lymphocytic leukemia (T-LGLL) was 1 of the most frequent bone marrow consultation-related diagnoses; all diagnoses of this entity were confirmed by in-person microscopy (100% concordance). A records search and physician surveys found no prior documented diagnoses of T-LGLL made in Vietnam before this education and consultation program. Our virtual consultation model has improved patient care in Vietnam by providing correct diagnoses to inform best practices in treatment. As a result of our program, the first Vietnam diagnoses of T-LGLL were made and may help expand on the literature in this area. This model could provide cost-effective, real-time consultation and education services for pathologists in underserved communities.

A-159 Role of TCRB1 Antibody in the Evaluation Panel of T Lymphoproliferative Diseases

Abstract Background The diagnosis of T lymphoproliferative diseases is usually a challenge due to the difficulty indifferentiating pathological T cells from reaction cells and also due to the limitation of clonalitytests currently available. TCRB1 is a monoclonal antibody specific for region 1 of the TCR T cellreceptor Beta chain constant, detecting clonal populations of TCRab positive T cells in caseswhere its expression is monotypic. Its application is similar to immunoglobulin light chainrestriction in the detection of B cell lymphoproliferations and in normal T CD3 + TCRab + cells, theexpression of this marker is polytypic. The objective of the present work is to report theexperience of including this antibody in the lymphoproliferative T panel in a private laboratoryin southern Brazil. Methods Samples received with indication of lymphoproliferative disease and lymphocytosis wereevaluated using the tube Lymphocyte Screnning Tube—LST (Euroflow) and were acquired usingFACS CANTO or FACS LYRIC cytometers. Analyzes were performed using Infinicyt software. After, all, samples that showed alteration in the CD4/CD8 ratio, or antigen alteration in the expressionof pan T markers were evaluated for the complete investigation panel according to Euroflowprotocol. This panel had the incorporation of the tube for the evaluation of clonality that hasthe TCRB1 antibody in the PE channel associated with the anchor markers CD45V500, CD3PerCpcy5, CD4V450 and CD8APCH7. The channels that were not used (FITC, APC, Pecy7) areused when necessary, to evaluate antigens that can help select the desired aberrant population,such as CD5, CD2, CD7 antibodies or to assess maturation, like CD45RA and CD27. Data werecollected from the beginning of the use of this antibody from February 2021 until the presentmoment. Results During this period, 2870 samples for immunophenotyping analysis were received, of which 17cases were T chronic lymphoproliferative diseases. Of the 17 cases found, 11 (65%) were caseswith the pathological population TCRab positive and the expression of TCRB1 could beevaluated. In these cases, the expression of the patologic cells was monotypic for the TCRB1antigen. The cases were: 4 large granular lymphocytic (LGL) leukemia, 4 Adult T-cellleukemia/lymphoma /Peripheral T-cell lymphoma, unspecified , 1 Sezary syndrome, 1 T-celllymphoblastic lymphoma in cortical/common thymocyte stage and 1 T-cell prolymphocyticleukemia (T-PLL). Recent studies have demonstrated the high specificity of this test for clonality, as long as it is associated with an expanded and standardized lymphoproliferative panel. Thepresent cases demonstrated good results for peripheral blood samples, bone marrow and lymphnode biopsy. Conclusion T-lymphoproliferative disorders remain a diagnostic challenge, due to the difficulty of a simpletest to assess clonality. The inclusion of the TCRB1 antibody in the lymphoproliferative panelincreased the specificity of the test for diagnosis, being a method of rapid execution and lowcost. Another advantage of using this marker is the possibility of evaluating small cell clonesthat may not be detected in molecular biology techniques due to the sensitivity of the test.

O-110 - Coexpression of natural killer cell antigens by T-cell large granular lymphocytes in hydroa vacciniforme lymphoproliferative disorder and the involvement of Vδ1+ epithelial-type γδT cells

O-110 - Coexpression of natural killer cell antigens by T-cell large granular lymphocytes in hydroa vacciniforme lymphoproliferative disorder and the involvement of Vδ1+ epithelial-type γδT cells

Phenotype Control techniques for Boolean gene regulatory networks.

Modeling cell signal transduction pathways via Boolean networks (BNs) has become an established method for analyzing intracellular communications over the last few decades. What's more, BNs provide a course-grained approach, not only to understanding molecular communications, but also for targeting pathway components that alter the long-term outcomes of the system. This has come to be known as phenotype control theory. In this review we study the interplay of various approaches for controlling gene regulatory networks such as: algebraic methods, control kernel, feedback vertex set, and stable motifs. The study will also include comparative discussion between the methods, using an established cancer model of T-Cell Large Granular Lymphocyte Leukemia. Further, we explore possible options for making the control search more efficient using reduction and modularity. Finally, we will include challenges presented such as the complexity and the availability of software for implementing each of these control techniques.

Context-dependent T-cell Receptor Gene Repertoire Profiles in Proliferations of T Large Granular Lymphocytes.

T cell large granular lymphocyte (T-LGL) lymphoproliferations constitute a disease spectrum ranging from poly/oligo to monoclonal. Boundaries within this spectrum of proliferations are not well established. T-LGL lymphoproliferations co-occur with a wide variety of other diseases ranging from autoimmune disorders, solid tumors, hematological malignancies, post solid organ, and hematopoietic stem cell transplantation, and can therefore arise as a consequence of a wide variety of antigenic triggers. Persistence of a dominant malignant T-LGL clone is established through continuous STAT3 activation. Using next-generation sequencing, we profiled a cohort of 27 well-established patients with T-LGL lymphoproliferations, aiming to identify the subclonal architecture of the T-cell receptor beta (TRB) chain gene repertoire. Moreover, we searched for associations between TRB gene repertoire patterns and clinical manifestations, with the ultimate objective of discriminating between T-LGL lymphoproliferations developing in different clinical contexts and/or displaying distinct clinical presentation. Altogether, our data demonstrates that the TRB gene repertoire of patients with T-LGL lymphoproliferations is context-dependent, displaying distinct clonal architectures in different settings. Our results also highlight that there are monoclonal T-LGL cells with or without STAT3 mutations that cause symptoms such as neutropenia on one end of a spectrum and reactive oligoclonal T-LGL lymphoproliferations on the other. Longitudinal analysis revealed temporal clonal dynamics and showed that T-LGL cells might arise as an epiphenomenon when co-occurring with other malignancies, possibly reactive toward tumor antigens.

Efficacy and safety of TBI+rATG-based conditioning regimen for haploidentical allogeneic hematopoietic stem cell transplantation in 11 cases of chemotherapy-resistant advanced peripheral T-cell lymphoma

Objective: To evaluate the clinical outcomes and safety of haploidentical hematopoietic stem cell transplantation (haplo-HSCT) using a conditioning regimen based on total body irradiation (TBI) and rabbit anti-human thymocyte globulin (rATG) in the management of chemotherapy-resistant advanced peripheral T-cell lymphoma (PTCL) . Methods: Clinical data of 11 patients with chemotherapy-resistant advanced PTCL who underwent haplo-HSCT with a TBI+rATG-based conditioning regimen at the Department of Hematology, Shanghai Liquan Hospital and Shanghai Zhaxin Integrated Traditional Chinese and Western Medicine Hospital, from September 2019 to December 2022 were retrospectively analyzed. Results: ①Among the 11 patients (six males and five females), with a median age of 40 years (range: 22-58 years), there were six cases of PTCL, not otherwise specified (PTCL-NOS), three cases of angioimmunoblastic T-cell lymphoma (AITL), one case of large-cell transformation of mycosis fungoides (MF-LCT), and one case of T-cell large granular lymphocytic leukemia (T-LGLL). According to the Lugano staging system, all patients were in stage Ⅲ or Ⅳ, and eight patients had B symptoms. Before transplantation, the median number of prior lines of chemotherapy was 4 (range: 2-10), and all patients had progressive disease (PD). The median time from diagnosis to transplantation was 17 months (range: 6-36 months). ②The conditioning regimen consisted of a TBI dose of 10 Gy, administered at 2 Gy on day -8 and 4 Gy from day -7 to day -6, rATG was administered at a daily dose of 2.5 mg/kg from day -5 to day -2. Etoposide (VP-16) was given at a dose of 15 mg/kg/d from day -5 to day -4, while cyclophosphamide (CTX) was administered at a dose of 50 mg/kg/d from day -3 to day -2. In patients with central nervous system involvement, etoposide and cyclophosphamide were replaced with thiotepa (TT) at a dose of 5 mg/kg/d from day -5 to day -4. Additionally, cytarabine (Ara-C) was added at a dose of 2.0 g/m(2) twice a day from day -3 to day -2 into the conditioning. ③Successful engraftment was achieved in all patients, with a median time to neutrophil engraftment of 14.5 d (range: 11-16 d) and a median time to platelet engraftment of 13 days (range: 8-18 days). Acute graft-versus-host disease (aGVHD) occurred in one patient (grade Ⅰ-Ⅱ), and another patient experienced grade Ⅲ-Ⅳ aGVHD. Among the eight survivors, four developed chronic GVHD (cGVHD). ④Post-transplantation, nine patients achieved complete response (CR). ⑤Hematopoietic suppression occurred in all patients after conditioning, with three experiencing diarrhea, four developing mucositis, three exhibiting elevated transaminase/bilirubin levels, and seven developing infectious complications. These non-hematologic adverse events were effectively managed. ⑥At one year post-transplantation, the non-relapse mortality (NRM) was (22.5±14.0) %, the cumulative incidence of relapse (CIR) was (20.2±12.7) %, and overall survival (OS) rate was (72.7±13.4) %, and disease-free survival (DFS) rate was (63.6±14.5) % . Conclusion: TBI+rATG-based conditioning regimen for haplo-HSCT is an effective and safe treatment approach for patients with chemotherapy-resistant advanced PTCL.

Pro-inflammatory cells sustain leukemic clonal expansion in T-cell large granular lymphocyte leukemia.

T-cell large granular lymphocyte leukemia (T-LGLL) is a chronic lymphoproliferative disorder characterized by the clonal expansion of T-cell large granular lymphocytes (T-LGL). Immunophenotypic and genotypic features contribute to discriminate symptomatic (CD8+ STAT3-mutated T-LGLL) from clinically indolent patients, this latter group including CD8+ wildtype (wt), CD4+ STAT5B-mutated and wt cases. T-LGL lymphoproliferation is sustained both by somatic gain-offunction mutations (i.e., STAT3 and STAT5B) and by pro-inflammatory cytokines, but little information is available on the activity of T-LGLL non-leukemic cells. In this study, we characterized pro-inflammatory cells in the peripheral blood of T-LGLL patients and analyzed their role in supporting the leukemic growth. In symptomatic patients we found that cell populations not belonging to the leukemic component showed a discrete pro-inflammatory pattern. In particular, CD8+ STAT3-mutated cases showed a skewed Th17/Treg ratio and an abnormal distribution of monocyte populations characterized by increased intermediate and non-classical monocytes. We also demonstrated that monocytes released high levels of interleukin-6 after CCL5 stimulation, a chemokine specifically expressed only by leukemic LGL. Conversely, in asymptomatic cases an altered distribution of monocyte populations was not detected. Moreover, T-LGLL patients' monocytes showed abnormal activation of signaling pathways, further supporting the different pathogenic role of monocytes in patients in discrete clinical settings. Altogether, our data contribute to deepening the knowledge on the different cell subtypes in T-LGLL, focusing particularly on non-leukemic cell populations and thus offering the rationale for new therapeutic strategies.

Effective treatment with the selective cytokine inhibitor BNZ-1 reveals the cytokine dependency of T-LGL leukemia

AbstractT-cell large granular lymphocytic leukemia (T-LGLL) is a clonal proliferation of cytotoxic T lymphocytes that can result in severe neutropenia, anemia, and bone marrow failure. Strong evidence from patients and mouse models demonstrate the critical role of interleukin-15 (IL-15) in T-LGLL pathogenesis. BNZ-1 is a pegylated peptide that selectively inhibits the binding of IL-15 and other γc cytokines to their cellular receptor complex, which has demonstrated efficacy in ex vivo T-LGLL cells and transgenic mice in preclinical studies. We conducted a phase 1/2 trial of BNZ-1 in patients with T-LGLL who had hematocytopenias (anemia or neutropenia) and required therapy. Clinical responses were assessed using hematologic parameters (improvement in hematocytopenias) based on response criteria from the Eastern Cooperative Oncology Group 5998 T-LGLL trial. BNZ-1 demonstrated clinical partial responses in 20% of patients with T-LGLL with minimal toxicity and the maximum tolerated dose was not reached. Furthermore, T-LGL leukemic cells showed significantly increased apoptosis in response to BNZ-1 treatment as early as day 2, including in clinical nonresponders, with changes that remained statistically different from baseline throughout treatment (P < .005). We report first-in-human proof that T-LGL leukemic cells are dependent on IL-15 and that intervention with IL-15 inhibition with BNZ-1 in patients with T-LGLL shows therapeutic effects, which carries important implications for the understanding of the pathogenesis of this disease. This trial was registered at www.clinicaltrials.gov as #NCT03239392.

Phase 1 trial of KT‐333, a STAT3 degrader, in patients with relapsed or refractory lymphomas, large granular lymphocytic leukemia and solid tumors

Background: KT-333 is a first-in-class, potent, highly selective, heterobifunctional small molecule degrader of the signal transducer and activator of transcription 3 (STAT3) protein. Aberrant activation of STAT3 resulting from activating mutations or deregulated cytokine signaling is known to underlie a variety of malignancies including peripheral T-cell lymphomas (PTCL), cutaneous T-cell lymphoma (CTCL), and large granular lymphocytic leukemia (LGL-L). However, selective targeting of STAT3 has been a therapeutic challenge. In preclinical xenograft studies, treatment with KT-333 resulted in durable tumor regressions in STAT3-dependent T cell lymphomas. Pharmacokinetic (PK) and pharmacodynamic (PD) analysis demonstrated that STAT3 degradation of >90% in tumor for 48 h is required for anti-tumor efficacy. Methods: The ongoing multicenter Phase 1a/1b study evaluates the safety, tolerability, PK, PD and preliminary clinical activity of KT-333 administered as a weekly IV infusion in 28-day cycles. The Phase 1a will identify the recommended phase 2 dose based on dose limiting toxicity (DLT) observed in Cycle 1. Eligible patients include those with advanced solid tumors and relapsed or refractory B- and T-cell lymphomas, Hodgkin’s lymphoma, and LGL leukemia. Serial blood samples are collected in Cycle 1 and Cycle 2 to measure KT-333 plasma concentrations and STAT3 protein expression in PBMCs (using targeted mass spectrometry). STAT3 degradation and the impact on tumor biology is measured in patients with accessible tumors. Results: As of February 3, 2023, seven patients have been treated in the first two dose levels (DLs) in Phase 1a, including patients with CTCL and PTCL with median age of 65 (range 57, 74) and performance status of 0 (n = 1) and 1 (n = 6). No DLTs have been observed and no patients have discontinued KT-333 due to an adverse event. The most common adverse events are Grade 1 and 2 and include constipation, fatigue, abdominal pain, dizziness, and nausea. PD data in blood demonstrated robust and sustained STAT3 degradation with mean maximum decrease of PBMC STAT3 levels after dosing on Days 1 and 8 of 66% in DL1 (n = 4) and 70% in DL2 (n = 2). PK results were in line with the modeled predictions. Conclusion: The emerging clinical data suggest that KT-333 is a potent degrader of STAT3 as demonstrated in PBMCs at doses that are well tolerated. These data provide the first evidence of STAT3 targeted protein degradation in humans and point to the potential of heterobifunctional degraders for targeting previously undruggable targets. It is projected that higher doses of KT-333 will achieve the predicted degradation profile in tumors that may translate into clinical benefit in patients with STAT3-dependent T cell malignancies. Accrual into the study is ongoing, and analyses from additional patients will be presented at the meeting. The research was funded by: Kymera Therapeutics Keywords: Aggressive T-cell non-Hodgkin lymphoma, Cutaneous non-Hodgkin lymphoma, Therapeutics and Clinical Trials in Lymphoma Conflicts of interests pertinent to the abstract. A. Olszewski Consultant or advisory role: Schrodinger, Genmab Research funding: Genentech, Schrodinger, PrecisionBio, Adaptive biotechnologies Z. Epstein-Peterson Honoraria: OncLive Research funding: Kymera, Viracta, Amgen R. Perea Employment or leadership position: Kymera Stock ownership: Kymera; Pfizer A. Gollerkeri Employment or leadership position: Kymera Stock ownership: Kymera, Pfizer J. Dey Employment or leadership position: Kymera Therapeutics Stock ownership: Kymera Therapeutics C. Klaus Employment or leadership position: Kymera Therapeutics Stock ownership: Kymera Therapeutics S. Agarwal Employment or leadership position: Kymera Stock ownership: Kymera J. Gollob Employment or leadership position: Kymera Stock ownership: Kymera A. Shastri Consultant or advisory role: Janssen Pharmaceuticals, Gilead Sciences, Rigel Pharmaceuticals and Kymera Therapeutics Honoraria: NACE & PeerView. Research funding: Kymera Therapeutics

Analyzing free-text clinical narratives for veterans with lymphoid malignancies using natural language processing (NLP).

e13576 Background: Free-text clinical narratives contain rich patient information, which is labor-intensive to extract through chart review. We developed an NLP pipeline to enable automatic extraction of performance status (PS), staging, and diagnosis from clinical narratives from Veterans Affairs (VA) patients with lymphoid malignancies (LM). Methods: The rule-based NLP algorithm was developed and iteratively refined using a development corpus of 287 notes independently annotated by two clinicians. The F1-score for PS was 95.8 (precision 98.6, recall 93.2), 92.7 for staging (precision 94.0, recall 81.6), and 67 (precision 80.2, recall 57.9) for diagnosis. The NLP pipeline was then externally validated using an evaluation corpus of 97 notes from another group of 100 veterans with T-cell LM. Results: The results are reported. In the 97 notes, primary diagnosis was most routinely documented, with 2.76 mentions per note. In comparison, staging was most sparsely documented with only 34 mentions (note that 11 patients with large granular lymphocytic leukemia were not staged). The NLP pipeline performed relatively well in extracting PS and staging (F1-scores were 0.74 and 0.72, respectively). It also achieved high precision in extracting diagnosis information (precision 0.93). However, recall (0.44) for diagnosis was poor, likely due to the complexities and inconsistencies of how diagnoses are documented for LM. Frequency of documentation and performance in the external validation set. Conclusions: The pipeline shows promising performance on the external validation set, demonstrating the feasibility of using NLP to extract information from notes of patients with LM for clinical research. The NLP pipeline generally has lower recall than precision, indicating that the pipeline may miss clinical information that should be captured. FPs incorrectly capture entities that are easily confused with the clinical entities of interest, such as nutritional status versus performance status. Future work includes capturing more lexical variations and indicators of documentation, as well as contextual information, such as in which sections of notes elements are likely documented. In addition, we describe how diagnosis may exist as primary, secondary, and in a differential, and we are building an NLP-based classifier to distinguish between these types of diagnoses. We will use results from the rule-based NLP pipeline as labels to fine-tune transformer-based, weak-supervised models to further enhance the performance. [Table: see text]