Pavletic, Steven Z.

Objectives/Goals: Manual skin assessment in chronic graft-versus-host disease (cGVHD) can be time consuming and inconsistent (>20% affected area) even for experts. Building on previous work we explore methods to use unmarked photos to train artificial intelligence (AI) models, aiming to improve performance by expanding and diversifying the training data without additional burden on experts. Methods/Study Population: Common to many medical imaging projects, we have a small number of expert-marked patient photos (N = 36, n = 360), and many unmarked photos (N = 337, n = 25,842). Dark skin (Fitzpatrick type 4+) is underrepresented in both sets; 11% of patients in the marked set and 9% in the unmarked set. In addition, a set of 20 expert-marked photos from 20 patients were withheld from training to assess model performance, with 20% dark skin type. Our gold standard markings were manual contours around affected skin by a trained expert. Three AI training methods were tested. Our established baseline uses only the small number of marked photos (supervised method). The semi-supervised method uses a mix of marked and unmarked photos with human feedback. The self-supervised method uses only unmarked photos without any human feedback. Results/Anticipated Results: We evaluated performance by comparing predicted skin areas with expert markings. The error was given by the absolute difference between the percentage areas marked by the AI model and expert, where lower is better. Across all test patients, the median error was 19% (interquartile range 6 – 34) for the supervised method and 10% (5 – 23) for the semi-supervised method, which incorporated unmarked photos from 83 patients. On dark skin types, the median error was 36% (18 – 62) for supervised and 28% (14 – 52) for semi-supervised, compared to a median error on light skin of 18% (5 – 26) for supervised and 7% (4 – 17) for semi-supervised. Self-supervised, using all 337 unmarked patients, is expected to further improve performance and consistency due to increased data diversity. Full results will be presented at the meeting. Discussion/Significance of Impact: By automating skin assessment for cGVHD, AI could improve accuracy and consistency compared to manual methods. If translated to clinical use, this would ease clinical burden and scale to large patient cohorts. Future work will focus on ensuring equitable performance across all skin types, providing fair and accurate assessments for every patient.

Abstract Myelodysplastic syndromes/neoplasms (MDSs) are heterogeneous stem cell malignancies characterized by poor prognosis and no curative therapies outside of allogeneic hematopoietic stem cell transplantation. Despite some recent approvals by the US Food and Drug Administration, (eg, luspatercept, ivosidenib, decitabine/cedazuridine, and imetelstat), there has been little progress in the development of truly transformative therapies for the treatment of patients with MDS. Challenges to advancing drug development in MDS are multifold but may be grouped into specific categories, including criteria for risk stratification and eligibility, response definitions, time-to-event end points, transfusion end points, functional assessments, and biomarker development. Strategies to address these challenges and optimize future clinical trial design for patients with MDS are presented here.

Abstract We examined the risk of subsequent neoplasms (SNs) and late mortality in children and adolescents undergoing allogeneic hematopoietic cell transplantation (HCT) for nonmalignant diseases (NMDs). We included 6028 patients (median age, 6 years; interquartile range, 1-11; range, <1 to 20) from the Center for International Blood and Marrow Transplant Research (1995-2012) registry. Standardized mortality ratios (SMRs) in 2-year survivors and standardized incidence ratios (SIRs) were calculated to compare mortality and SN rates with expected rates in the general population. Median follow-up of survivors was 7.8 years. Diagnoses included severe aplastic anemia (SAA; 24%), Fanconi anemia (FA; 10%), other marrow failure (6%), hemoglobinopathy (15%), immunodeficiency (23%), and metabolic/leukodystrophy syndrome (22%). Ten-year survival was 93% (95% confidence interval [95% CI], 92% to 94%; SMR, 4.2; 95% CI, 3.7-4.8). Seventy-one patients developed SNs (1.2%). Incidence was highest in FA (5.5%), SAA (1.1%), and other marrow failure syndromes (1.7%); for other NMDs, incidence was <1%. Hematologic (27%), oropharyngeal (25%), and skin cancers (13%) were most common. Leukemia risk was highest in the first 5 years posttransplantation; oropharyngeal, skin, liver, and thyroid tumors primarily occurred after 5 years. Despite a low number of SNs, patients had an 11-fold increased SN risk (SIR, 11; 95% CI, 8.9-13.9) compared with the general population. We report excellent long-term survival and low SN incidence in an international cohort of children undergoing HCT for NMDs. The risk of SN development was highest in patients with FA and marrow failure syndromes, highlighting the need for long-term posttransplantation surveillance in this population.

Abstract Abstract 3094 The HLA class II DRB1 antigen DR15 is an important immunobiologic marker in immune mediated marrow failure states. DR15 has also been reported in small studies to be associated with favorable outcomes (reduction in acute GVHD and reduced relapse resulting in improved overall survival) after allogeneic hematopoietic cell transplant. To elucidate the impact of DR15 on major transplant outcomes, we conducted a retrospective study of 2, 891 recipients of first marrow or mobilized peripheral blood stem cell transplantation for the treatment of acute myeloid leukemia (n=1038), acute lymphoblastic leukemia (n=700), chronic myeloid leukemia (n=948), or myelodysplastic syndrome (n=205) between 1990–2008 and reported to the CIBMTR registry. Selection was confined to HLA-identical sibling transplantation to avoid HLA-disparity as a driving force for observed differences. All patients received conventional myeloablative conditioning, T-replete grafts and cyclosporine plus methotrexate- based GVHD prophylaxis. DNA-based HLA typing allowed categorization of 732 (25.3%) patients as positive and 2159 (74.7%) patients as negative for DRB1*15 :01 or *15 :02 (DR15). There were no significant differences in baseline characteristics between the HLA DR15-positive and -negative groups. In univariate analysis, HLA-DR15 status had no impact on neutrophil engraftment, acute graft-versus-host disease (GvHD) II-IV or III-IV, chronic GVHD, treatment related mortality, relapse, disease-free survival or overall survival. Confining the univariate analysis to myeloid malignancies did not alter these findings. Multivariate analysis models were constructed with DR15 status forced into the models in all steps of model building and the final model regardless of its statistical significance. Other variables tested included: donor/recipient age, CMV status, disease, disease stage, graft source, Karnofsky score, race and year of transplant. Variables that attained a p-value ≤0.05 were held in the final multivariate models. In multivariate analysis, DR15 status showed no significant difference in the primary outcomes of acute GVHD II-IV or III-IV, chronic GVHD, overall survival, or relapse. In conclusion, DR15 status had no impact on major HLA-matched sibling donor hematopoietic cell transplantation outcomes in this large and homogenous cohort of leukemia and MDS patients. Disclosures: No relevant conflicts of interest to declare.

Abstract Abstract 897 Recent changes in allogeneic hematopoietic cell transplantation (HCT) including increased use of reduced-intensity conditioning, peripheral blood cells as the graft source (PB) and unrelated donors (URD) warrant re-evaluation of risk factors for acute graft-versus-host disease (aGVHD) and its impact on overall survival (OS). Risk factors for these outcomes were analyzed using more recent data from CIBMTR observational database. Methods: The cohort included adult patients ≥ 20 y transplanted from an HLA-identical sibling (SD) (n=3191) or URD (N=2370) for acute myelogenous leukemia (AML), acute lymphoid leukemia (ALL), chronic myeloid leukemia (CML), or myelodsyplastic syndrome (MDS) using a T-cell replete graft from 1999–2005. Six groups were created to evaluate the impact of conditioning [(myeloablative (MA) and reduced-intensity or non-myeloablative (RIC)], total-body irradiation (TBI) and graft source [(bone marrow (BM) or PB] as follows: MA+TBI+PB (group 1), MA+TBI+BM (group 2) MA+ no TBI+PB (group 3), MA+ no-TBI+BM (group 4), RIC+PB (group 5) and RIC+BM (group 6). Separate analyses were performed for SD and URD. IBMTR grade was used to classify aGVHD. A p-value of ≤ 0.01 was considered significant. Results: Among the SD cohort, the probability of aGVHD grade B-D and grade C-D at 100 days was 39% (95% CI, 37–41%) and 16% (95% CI, 14–17%). In multivariate analyses, non-TBI based MA regimens with a BM graft (group 4), RIC conditioning with PB (group 5) and tacrolimus plus methotrexate aGVHD prophylaxis were associated with lower odds ratio of grade B-D aGVHD (Table 1). The probability of OS was 51% (95% CI: 49–53%) at 3 y and 46% (95% CI: 44–49%) at 5 y. In multivariate analyses, grade B-D aGVHD was associated with a higher risk of death. Other independent risk factors for OS are shown in Table 2. Among the URD cohort, the probability of aGVHD grade B-D and C-D at 100 days was 59% (95%, CI 57–61%) and 32% (30-34%). In multivariate analysis, BM with MA (TBI and no TBI) or RIC conditioning (groups 2, 4 and 6) were significantly associated with lower odds ratio of grade B-D aGVHD (Table 1). Other independent risk factors included a diagnosis of CML. HCT from a 7/8 HLA-mismatched URD showed a trend for higher incidence of aGVHD (p=0.02). The probability of OS was 38% (95% CI: 35–40%) at 3 y and 33% (95% CI: 31–35%) at 5 y. In multivariate analysis, grade B-D aGVHD was associated with a higher mortality. Table 2 shows other independent risk factors for OS. Conclusion: Intensity of the conditioning regimen, TBI and the graft source has a combined effect on the risk of aGVHD. In both SD and URD cohorts, BM with MA, non-TBI regimens were associated with lower risk of aGVHD. In URD cohort, BM after a RIC also was associated with a reduced risk of aGVHD. Modulation of these risk factors is needed to reduce acute GVHD incidence and death after allogeneic transplantation. Disclosures: No relevant conflicts of interest to declare.

Purpose To develop a system prognostic of outcome in those undergoing allogeneic hematopoietic cell transplantation (allo HCT) for myelodysplastic syndrome (MDS). Patients and Methods We examined 2,133 patients with MDS undergoing HLA-matched (n = 1,728) or -mismatched (n = 405) allo HCT from 2000 to 2012. We used a Cox multivariable model to identify factors prognostic of mortality in a training subset (n = 1,151) of the HLA-matched cohort. A weighted score using these factors was assigned to the remaining patients undergoing HLA-matched allo HCT (validation cohort; n = 577) as well as to patients undergoing HLA-mismatched allo HCT. Results Blood blasts greater than 3% (hazard ratio [HR], 1.41; 95% CI, 1.08 to 1.85), platelets 50 × 109/L or less at transplantation (HR, 1.37; 95% CI, 1.18 to 1.61), Karnofsky performance status less than 90% (HR, 1.25; 95% CI, 1.06 to 1.28), comprehensive cytogenetic risk score of poor or very poor (HR, 1.43; 95% CI, 1.14 to 1.80), and age 30 to 49 years (HR, 1.60; 95% CI, 1.09 to 2.35) were associated with increased hazard of death and assigned 1 point in the scoring system. Monosomal karyotype (HR, 2.01; 95% CI, 1.65 to 2.45) and age 50 years or older (HR, 1.93; 95% CI, 1.36 to 2.83) were assigned 2 points. The 3-year overall survival after transplantation in patients with low (0 to 1 points), intermediate (2 to 3), high (4 to 5) and very high (≥ 6) scores was 71% (95% CI, 58% to 85%), 49% (95% CI, 42% to 56%), 41% (95% CI, 31% to 51%), and 25% (95% CI, 4% to 46%), respectively (P < .001). Increasing score was predictive of increased relapse (P < .001) and treatment-related mortality (P < .001) in the HLA-matched set and relapse (P < .001) in the HLA-mismatched cohort. Conclusion The proposed system is prognostic of outcome in patients undergoing HLA-matched and -mismatched allo HCT for MDS.

Key Points Survival of chronic GVHD patients was predicted by clinician-assessed response and changes in patient-reported outcomes. FFS was predicted by clinician-assessed response, changes in patient-reported outcomes, and the 2014 NIH response criteria.

Abstract Risk factors for acute GVHD (AGVHD), overall survival, and transplant-related mortality were evaluated in adults receiving allogeneic hematopoietic cell transplants (1999-2005) from HLA-identical sibling donors (SDs; n = 3191) or unrelated donors (URDs; n = 2370) and reported to the Center for International Blood and Marrow Transplant Research, Minneapolis, MN. To understand the impact of transplant regimen on AGVHD risk, 6 treatment categories were evaluated: (1) myeloablative conditioning (MA) with total body irradiation (TBI) + PBSCs, (2) MA + TBI + BM, (3) MA + nonTBI + PBSCs, (4) MA + nonTBI + BM, (5) reduced intensity conditioning (RIC) + PBSCs, and (6) RIC + BM. The cumulative incidences of grades B-D AGVHD were 39% (95% confidence interval [CI], 37%-41%) in the SD cohort and 59% (95% CI, 57%-61%) in the URD cohort. Patients receiving SD transplants with MA + nonTBI + BM and RIC + PBSCs had significantly lower risks of grades B-D AGVHD than patients in other treatment categories. Those receiving URD transplants with MA + TBI + BM, MA + nonTBI + BM, RIC + BM, or RIC + PBSCs had lower risks of grades B-D AGVHD than those in other treatment categories. The 5-year probabilities of survival were 46% (95% CI, 44%-49%) with SD transplants and 33% (95% CI, 31%-35%) with URD transplants. Conditioning intensity, TBI and graft source have a combined effect on risk of AGVHD that must be considered in deciding on a treatment strategy for individual patients.