Citations

Abstract Pancreatic ductal adenocarcinoma (PDAC), the most common subtype of pancreatic cancer, is a deadly disease with a complex tumor microenvironment (TME). How chemotherapy alters the TME, and whether these changes drive chemoresistance, is poorly understood. We examined matched pre- and post-treatment tissue specimens and found near-universal enrichment of axonal guidance genes in cancer-associated fibroblasts (CAFs) after treatment. These CAFs were enriched near sites of perineural invasion, coinciding with regions of increased tumor cell proliferation, and were enriched in tumor areas distant from nerves after chemotherapy. Metastatic recurrence lesions had the highest prevalence of these CAFs versus primary tumors and untreated metastasis. These CAFs showed elevated non-canonical Wnt mediators and axonal guidance genes, which complemented matching cognate binding partners in tumor epithelial cells, suggesting a role in tumor-stroma crosstalk. Our findings implicate fibroblast-derived axonal-guidance genes in promoting PDAC invasion and point to a promising target for this disease. Statement of Significance Therapeutic resistance remains a major challenge in pancreatic cancer. Our findings shed light on the changes in the complex tumor microenvironment in response to chemotherapy and identify fibroblasts high in axonal-guidance genes that may drive tumor progression and chemoresistance, thus uncovering a potential avenue for targeting pancreatic cancer.

Abstract Background Teclistamab (Tec) and Talquetamab (Tal) are the two first FDA approved bispecific antibodies (BsAb) for the treatment of patients (pts) with relapsed/refractory multiple myeloma (RRMM). Both Tec and Tal should be initiated with step-up dosing (SUD) in an inpatient (IP) setting to mitigate the risk of cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS), per US label. However, standard of care practice continues to evolve and more pts receive SUD as an outpatient (OP) to reduce healthcare resource utilization and improve pt convenience. While most real-world evidence has focused on clinical and safety outcomes of pts treated with Tec and Tal, pt experience associated with these novel BsAb therapies is relatively unexplored. Methods HealthTree Foundation, a multiple myeloma pt advocacy group which provides enrolled pts with information on treatments, community groups, and research participation, surveyed RRMM patients who have received Tec or Tal. Eligible pts who completed the treatment experience survey had at least 4 months of follow-up after starting BsAb. The survey was fielded from May 2024 through June 2025. Data collection followed IRB-exempt protocols with informed consent. Results A total of 32 patients (24 Tec, 8 Tal) completed the survey. The median age was 70 years, 50% were male, and 84% were white. Most pts (n=24; 75%) received SUD at their primary oncology clinic (n=19, 59% at an academic center; n=5, 16% at local community center), and 8 (25%) at an academic center which was not their primary oncology clinic. All pts who received SUD at an academic center that was not their primary oncology clinic later transitioned to their local community clinic post-SUD and reported being satisfied or very satisfied with their transition of care experience. Overall, most pts (n= 21; 66%) received their entire step-up dosing during a single hospitalization, while some (n=6; 19%) received all doses of the SUD during separate hospitalizations for each dose. About 13% (n=4) received all SUD in an OP setting, and 3% (n=1) received SUD in a hybrid setting (mix of IP and OP). Although 85% (n=27) of the pts received IP SUD, 66% (n=21) would have preferred IP SUD in one hospitalization, while the remaining 34% (n=11) would have preferred OP or hybrid SUD. Among pts who received IP SUD (n=27), almost a quarter reported a preference for either OP (n=4; 15%) or hybrid SUD (n=2; 7%) and 78% (n=21) preferred one IP hospitalization for SUD. The factors influencing their preference for the setting of SUD in this IP sub-group were feeling more secure with IP monitoring (78%), convenience (26%), less worry about staying near healthcare centers (15%), being able to stay with their local physician (11%), staying closer to home (7%), and lower costs to the pt (4%). Among pts receiving OP SUD (n=4), 75% (n=3) preferred fully OP SUD and 25% (n=1) preferred hybrid SUD because of convenience (n=2; 50%), lower cost to pt (n=1; 25%), flexible dosing schedule (n=1; 25%), and feeling of security with IP monitoring (n=1; 25%). All the pts receiving hybrid SUD (n=1) preferred fully OP SUD due to convenience. Although Tec and Tal are subcutaneous (SC) injections, when patients were asked to compare the options of BsAbs with comparable efficacy and safety profile, they favored SC injection (n= 22; 69%) over intravenous infusion (n = 3; 9%), while 22% (n=7) had no preference for mode of drug administration. The challenges that impacted pt experience with Tec or Tal were treatment side effects (Tal pts: n=6, 75%; Tec pts: n=9, 38%) and distance or travel time to the treatment facility (Tal pts: n=2, 25%; Tec pts: n=4; 17%). Other reasons included caring for others, continuing to work, cost and insurance, lack of caregiver/care partner, being away from home, unavailability of the therapy in community setting near home, and eligibility criteria. Conclusions Most pts in this study received SUD at their primary oncology clinic. While most pts received IP SUD and preferred a single planned IP visit, one third preferred OP or hybrid SUD due to convenience. As the field moves toward OP-only SUD, this study underscores the need to consider pt preferences. There is an unmet need for community clinics providing SUD options closer to pt homes.

Abstract Introduction: The prognosis of patients (pts) with histological transformation (HT) of Waldenström macroglobulinemia (WM) remains dismal with standard chemoimmunotherapy. CD19-targeted chimeric antigen receptor T-cell therapies (CAR-T) can lead to durable responses in relapsed/refractory (R/R) large B-cell lymphoma (LBCL) and are approved in the second or third line settings. Pts with HT-WM were largely excluded from CAR-T trials. We previously reported high response rates in a series of 23 pts. The aim of the present study was to evaluate a larger cohort of pts with R/R HT-WM treated with CAR-T with a longer follow-up. Methods: We performed a multicenter retrospective study of pts with biopsy-proven transformed WM/lymphoplasmacytic lymphoma treated with CAR-T in the French DESCAR-T registry (NCT04328298), 4 centers from ECWM and 4 US centers. This study includes updated data from previously published cases (Durot et al, Blood, 2024). Responses were assessed according to the Lugano 2014 criteria. Cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS) were graded according to the ASTCT 2019 consensus criteria. Hematological toxicity and infections were graded according to the NCI CTCAE (version 5.0). Progression-free survival (PFS) and overall survival (OS) were calculated from the date of CAR-T infusion and estimated using the Kaplan-Meier method. Results: 50 pts were included (39 from DESCAR-T centers, 7 from US centers and 4 from ECWM centers). Median age was 60 (range: 32-72) and 68 (range: 42-82) years at WM diagnosis and CAR-T infusion, respectively. MYD88L265P and CXCR4 mutations were present in 79% (23/29) and 15% (2/13) of patients with available data, respectively. The median time from WM to HT diagnosis was 5.3 years (range: 0-32). Median prior lines of therapy was 1 (range: 0-9) for WM and 2 for HT (range: 0-4). Twenty (40%) pts previously received a BTK inhibitor for WM or HT, 12 (24%) underwent prior autologous SCT and 1 pt allogeneic SCT. 40% of pts presented primary refractory disease and 44% were refractory to last treatment before CAR-T. 40 pts (80%) received bridging therapy, of whom 17 (44%) responded (7 complete responses (CR)). 35 (70%), 10 (20%), and 5 (10%) pts received axicabtagene ciloleucel (axi-cel), tisagenlecleucel (tisa-cel), and lisocabtagene maraleucel (liso-cel), respectively. The best overall response rate (ORR) was 94%, with 44 (88%) pts reaching CR. With a median follow-up of 25.7 months (range: 1-69) from CAR-T infusion, the median PFS was 1.2 years (95% CI: 0.5-4.5) and median OS was 4.2 years (95% CI: 1.2-not reached). Median duration of response was 1.5 years (95% CI: 0.4-4.4), with a median of 4 years (95% CI: 1.5-NR) for pts still in CR at 6 months after CAR-T. 22 pts experienced relapse/progression after CAR-T, 19 related to LBCL and 3 to WM. Median survival after relapse was 7.9 months (95% CI: 3.8-25.3). At LBCL relapse after CAR-T, salvage therapy consisted of CD20xCD3 bispecific antibodies for 9 pts, lenalidomide +/- anti-CD20 or anti-CD19 antibody for 6 pts, and BTKi for 3 pts. Among 19 pts who died, 14 (74%) were due to progressive disease and 5 (26%) from infections. 37 (76%) pts experienced CRS, including 2 (4%) grade 3 events, and 22 (45%) had ICANS, with 2 (4%) ≥ grade 3 events. 17 (35%) pts presented infections, including 12 ≥ grade 3. 21 (43%) pts developed ≥ grade 3 prolonged cytopenias (ie, not resolved within 30 days post infusion). CNS involvement due to LBCL was present in 11 pts (22%) at the time of CAR-T. 6 of them had ICANS, with 1 grade 4. The best ORR was 91% (best CR 91%). Only 1 patient did not obtain a response, with stable disease at 1 month and progression at 3 months. The median PFS was 1.2 year (95% CI: 0.2-NR) and median OS was NR (95% CI: 0.5-NR). On univariate analysis, only elevated LDH was correlated with worse PFS (HR 2.5, p=0.04) and OS (HR 3.3, p=0.04). Other variables, including number of prior lines of therapy, MYD88 mutation status, primary refractory disease, refractory to last treatment, CNS involvement before CAR-T, response after bridging therapy and type of CAR-T received, were not associated with inferior PFS or OS. Conclusion: This cohort represents the largest series of pts with transformed WM treated with anti-CD19 CAR-T. We confirm the high efficacy previously reported with an acceptable safety profile, including for pts with CNS involvement due to LBCL at CAR-T infusion.

Abstract Myeloproliferative neoplasms (MPNs) are typically ascribed to hyperactive JAK-STAT signaling via mutations in JAK2, CALR, or MPL. Here, we characterize how activating mutations in the calcium sensor STIM1, previously established as the cause of Stormorken Syndrome (SS), a rare congenital disorder, are sufficient to drive development of myelofibrosis (MF). We further uncover distinct interactions between STIM1 activity and JAK2 versus CALR driven MPNs. We initially identified two SS patients with activating mutations in STIM1 (R304W and S88G) and features of MF, including marked reticulin fibrosis with megakaryocytic hyperplasia and atypia. This led us to hypothesize that altered calcium signaling may play a role in the development of MF in both SS and MPNs. Single cell RNA sequencing performed on samples from both patients demonstrated elevated NFκB inflammatory signaling in monocyte populations, suggesting a link between altered STIM1 and store operated calcium entry (SOCE) activity and aberrant inflammation, thereby contributing to MF development. To corroborate these findings from our SS patients, we generated a knock-in mouse model of the heterozygous R304W mutation. Induction of hematopoietic-specific expression of mutant Stim1 with Vav-Cre recapitulated the hematological features of SS, including elevated SOCE activity and thrombocytopenia. We also recapitulated MF features including the development of bone marrow fibrosis and severe osteosclerosis, with the mice developing splenomegaly as young as 9 weeks old. In ex vivo megakaryocyte differentiation assays with our Stim1 R304W knock-in mice, we also observed aberrant cytoskeletal organization and elevated TGFβ mRNA expression. When induction was driven universally through CMV-Cre, we also validated that mice recapitulated multiple SS features, including thrombocytopenia, tubular aggregate myopathy, growth deficiency, and the development of bone marrow fibrosis as young as 10 weeks. We additionally identified a separate cohort of 9 related individuals with SS and STIM1 mutations (L92V) as well as a separate patient with the R304W mutation. In ex vivo megakaryocytic differentiation assays, cells from SS patients demonstrated defective thrombopoiesis with significantly decreased proplatelet formation, and displayed aberrant cytoskeletal formation within the pro-platelet termini, in conjunction with increased expression of TGFβ. Further supporting a role for STIM1 in MPNs, we identified upregulation of STIM1 expression in MPN patient CD34+ hematopoietic stem/progenitors' cells (HSPCs), we well as megakaryocyte progenitors and platelets, including those with CALR or JAK2 mutations. To evaluate the functional role of STIM1 in MPNs, we generated patient-derived xenograft (PDX) models with CALR and JAK2-mutantpatient samples subjected to CRISPR ablation of STIM1. In an initial CALR-mutant PDX experiment, abrogation of STIM1 led to decreased human CD45+ cell engraftment in conjunction with prolonged survival. In a second CALR-mutant PDX experiment, STIM1 targeting did not overtly impact engraftment or survival. In contrast, CRISPR ablation of STIM1 in JAK2-mutant PDX models led to exacerbated disease phenotypes, including increased human CD45+ cell engraftment, increased splenomegaly, and early lethality. Altogether, this study represents the first demonstration of MF development in patients with Stormorken Syndrome and reveals a previously unrecognized role of aberrant SOCE underlying SS and MPNs. We recapitulated these findings in a novel conditional mouse model. These findings also illustrate unique interactions between STIM1 and JAK2 and CALR driven MPNs. These observations have important implications for the development of targeted therapeutic approaches for these disorders.

Abstract Introduction: Waldenström macroglobulinemia (WM) is characterized by the presence of a MYD88 mutation in >90% and a CXCR4 mutation in 30-40% of cases. The impact of these mutations on treatment responses and disease outcomes is well known. TP53 alterations (TP53ALT) have more recently been described in WM and are reported to occur in up to 25% of patients, with a higher frequency in patients with previously treated WM, especially those with prior alkylator or nucleoside analogue exposure (Tam et al Blood Adv 2024; Poulain et al CCR 2017; Tsakmaklis et al, ASH 2025). TP53 ALT WM is associated with a shorter time to disease progression, as well as worse overall survival (Poulain et al CCR 2017; Gustine et al, BJH 2019). Depth of response to BTK inhibitors is affected by TP53 status, with recent data reporting a very good partial response (VGPR) rate of 34% v. 25% and major response rates of 84% v. 73% in patients without or with a TP53 ALT (Tam et al Blood Adv 2024). No complete responses were reported and median PFS was shorter in those with a TP53ALT compared with wild-type (TP53WT). Loncastuximab tesirine is a humanized IgG1 anti-CD19 antibody that is conjugated to a pyrrolobenzodiazepine (PBD) dimer. Following internalization, the PBD dimer is released via proteolytic cleavage and binds to the DNA minor groove forming highly cytotoxic DNA inter-strand crosslinks. Cells deficient in DNA repair including TP53 alterations are vulnerable to the cytotoxic effects of loncastuximab tesirine (Tarantelli et al, Haematologica 2024). Methods: In an ongoing investigator-initiated, prospective phase II study in symptomatic patients with relapsed or refractory WM (NCT05190705), patients are receiving single agent loncastuximab tesirine. Loncastuximab tesirine is administered at 150 ug/kg in cycles 1 and 2 and 75 ug/kg in cycles 3 to 6. Each cycle is 28 days. The primary outcome measure of this study is attainment of VGPR or complete (VGPR/CR) response. The accrual goal of this trial is 20 patients, based on an alternative VGPR/CR rate ≥40% versus a null of ≤10%, with a double-sided alpha of 0.037 and a power of 87%. The study would meet its endpoint if ≥6 patients attained a VGPR/CR. Results: Of the 14 patients who have completed therapy on this trial to date, the median age is 71 years (range 53-81), 9 (64%) are male, and the median prior number of therapies is 4 (range 2-10). Median baseline hemoglobin is 9.5 g/dL (range 8.2-12.6), serum IgM is 2146 mg/dL (range 723-5955), and bone marrow involvement is 53% (range <5-90). Four patients had extramedullary disease. Twelve patients (86%) are MYD88 mutated; 8 patients (57%) are CXCR4 mutated, and 8 (57%) are TP53 altered. Nine patients completed all 6 cycles of therapy. Of the 5 other patients, treatment was stopped early due to GGT elevation (n=2), disease progression (n=1), patient decision (n=1), and skin toxicity (n=1). No IgM flare and no infusion reactions have occurred. For all patients, the median time to minor response, major response, and deepest response was 28, 55, and 172 days, respectively. The response rates include 1 (7%) CR; 9 (64%) VGPRs; 2 (14%) partial responses (PR); 1 (7%) minor response; and 1 (7%) progressive disease, for an overall response rate of 93% and a VGPR/CR rate of 71%. For the 8 patients with TP53ALT, the baseline characteristics were as follows: 50% male; median age of 70 years (range 53-77), median prior therapies is 4 (range 2-10); median IgM is 1877 mg/dL (range 723-5639); and median hemoglobin is 9.2 g/dL (range 8.2-11.6). All TP53ALT patients had an MYD88 mutation, and 6 (75%) carried a CXCR4 mutation. For those with TP53 ALT, the median time to minor response, major response, and deepest response was 28, 49, and 235 days, respectively. All TP53ALT patients attained a major response; 1 patient (13%) with a CR, 6 (75%) with a VGPR, and 1 (13%) with a PR, for an overall response rate of 100% and a VGPR/CR rate of 89%. One patient had splenomegaly and one had adenopathy, both of which resolved after treatment. Conclusion: Our findings from this ongoing, prospective trial demonstrate a high VGPR/CR rate of 71% in heavily pre-treated WM. Notable, is the VGPR/CR rate of 89% in high-risk patients with TP53ALT WM. No unexpected toxicities have been seen in this trial. Our findings identify a potentially novel time-limited therapy in relapsed or refractory WM, including patients with high-risk TP53ALT WM.

Abstract Background: Social determinants of health (SDOH) are critical drivers of cancer care access and outcomes. In multiple myeloma (MM), access to cellular therapies including autologous stem cell transplantation and CAR-T therapy often depends on formal or informal caregiver support. Traditional SDOH measures (e.g., race, income, or education) are often used as proxies for access to care; however, these measures may overlook individual realities pertaining to cellular therapies where a caregiver is generally required. In response to feedback from a HealthTree Patient Advisory Board questioning the practice of inferring caregiver support, this study evaluated a simple survey question by a large cross-sectional survey. Methods: After development of a dedicated single-question (“What best describes your current caregiving support?”) through iterative feedback from the Patient Advisory Board and MM experts, four key answers were settled on: (1) cohabitation with dedicated caregiver(s) to support care (“family support”); (2) presence of non-cohabiting caregiver(s) to provide dedicated support (“external support”); (3) lack of dedicated caregiver, hence support is arranged when needed (“arrangeable support ”); or (4) lack of dedicated caregiver, hence finding any support is a struggle (“inadequate support”). We deployed this new demographic question from February through July 2025 through HealthTree. Responses were analyzed across SDOH variables using Fisher exact or chi square tests, and random forest classifiers were employed for predictive modeling. Results: Of 1231 respondents, 71% had dedicated caregivers (68%, n=839, with family support, and 3%, n=41, with external support). In contrast, 23% (n=287) of patients routinely arranged support while 5% (n=64) had inadequate support (struggle to find support). The inadequate-support group had significant demographic differences from the sample as a whole: a higher proportion were women (78% vs 56%, p = 0.04), Black (9% vs 4%, p < 0.01), lived >50 miles from their treatment center (17% vs 6%, p < 0.01), and had an annual income <$50,000 (17% vs 6%, p = 0.02). This group had a numerically higher proportion of patients whose employment status was disabled (12% vs 5%, p=0.08) as well. No proportion difference in age (67 ± 9 yr), military status, or severity of diagnosis - precursor disease (17%) v. multiple myeloma (81%). Rates of inadequate support were overall low in patients who lived with only their spouse (1%) or lived with their spouse & children (3%). In contrast, rates of inadequate support were higher (than 5%) in patients who lived alone (21%), divorced (20%), lived with adult children (14%), lived with friends (13%), separated (11%), or widowed (9%) (p < 0.001). However, the arrangeable-support group comprised 12% of patients who lived only with their spouse, 19% of patients who lived with their spouse and children, and 61% of patients who lived with their adult children (p < 0.001). Finally, a random forest classifier failed to identify any individual in the inadequate-support category, suggesting that unmeasured or more complex patterns may underlie this classification. Conclusions: To our knowledge, this is the largest analysis of caregiver support and its relationship with other SDOH in patients with MM to date. Despite the importance of caregiving to transplantation and CAR-T therapy, we found that 5% of patients with MM struggle to find caregiver support; this figure likely underestimates the true prevalence, given potential selection bias among patients who are digitally engaged through the HealthTree platform. Even among patients who lived with their spouses, an additional 10-20% of patients routinely had to arrange caregiving support on a visit-by-visit basis; for patients who lived with their adult children, this proportion was over 60%. The requirement to arrange caregiving support can create time toxicity and financial toxicity for patients, although these factors were not directly queried in our study. Finally, caregiver availability cannot be routinely inferred from traditional SDOH variables. Our data highlight the importance of including caregiver status as a standard demographic question in clinical and research settings. Further interventions are needed to improve caregiver support for vulnerable patients with MM as well as to reduce the duration of caregiver requirements for procedures such as CAR-T therapy.

Abstract Introduction Detecting genetic abnormalities is mandatory during diagnostic workup and for potential individualization of therapy selection in multiple myeloma (MM) and its precursor conditions. At present, this requires invasive bone marrow (BM) biopsies, severely limiting early detection, patient monitoring, and the precise selection of therapy. The current standard for detecting genetic alterations in MM is fluorescence in situ hybridization (FISH), which cannot detect point mutations and other clinically relevant alterations. Consequently, the IMS-IMWG guidelines were recently updated to require next-generation sequencing for the classification of high-risk MM. Moreover, the importance of identifying alterations in therapeutic targets (e.g., BCMA, GPRC5D) for guiding immunotherapies is becoming increasingly recognized. To address this need, we recently launched GenoPredicta, a CLIA-approved Laboratory Developed Test that enables routine monitoring, informing diagnosis, and treatment selection by comprehensively and robustly characterizing MM genomes with whole genome sequencing (WGS) from a minimum of ~50 circulating tumor cells (CTCs) isolated from peripheral blood (PB) or tumor cells from BM. Here, we describe the characterization of clinical and research samples with GenoPredicta, highlighting resistance-conferring genetic alterations that can only be detected by sequencing and that are consistent with patients' clinical histories, while further demonstrating the complete concordance of GenoPredicta with prior FISH results. Methods Tumor cells were isolated from BM and/or PB samples using fluorescence-activated cell sorting (FACS) (after CD138+ enrichment for PB) and subjected to whole genome sequencing (WGS), from which we identified copy number alterations, structural variants, and single nucleotide variants and indels. For a subset of samples, the BM was also characterized by FISH, using a standard clinical panel for characterizing MM; for other samples prior FISH results were available from patients' clinical records. Results The analytical validation of GenoPredicta, conducted on a retrospective cohort of >60 viably frozen samples from >30 patients, showed complete concordance between BM and PB for detected genomic abnormalities, with all clinically relevant events (11 translocations and 125 CNVs) observed in BM recapitulated in PB (100% recall). All events identified by FISH in 14 samples (8 translocations and 30 CNVs) were also identified in the BM (100% recall), with 50 translocations and 30 CNVs not detected by FISH also absent in the WGS data (100% precision). The abnormalities detected in these samples also included multiple high-risk events not assayed by FISH, including biallelic TP53 loss (Del 17p coupled with a deleterious TP53 mutation, observed in 4 patients) and MYC::IGL translocations (2 patients). Moreover, for all fresh peripheral blood clinical samples with FISH results available from patients' histories, GenoPredicta recalled 100% of reported events. Among these clinical samples, we notably observed a case with both a biallelic GPRC5D deletion and a TNFRSF17 (BCMA) deletion combined with a truncating mutation on the remaining allele. This was consistent with the patient's clinical history showing progression after therapy with talquetamab (anti-GPRC5D bispecific), and previous nonresponse to teclistamab (anti-BCMA bispecific) after prior therapy with Cilta-cel (BCMA-directed CAR T). GenoPredicta also detected other mutations likely to confer resistance to immunotherapies, including loss-of-function mutations in IKZF3 and deletions of FCRL5 in frozen peripheral blood research samples. Conclusion We demonstrate that WGS-based characterization of MM from BM or CTCs is a viable replacement for FISH for clinical diagnosis and monitoring, with blood-based measurements enabling comprehensive profiling of the myeloma cancer genome. Importantly, this includes genetic alterations that confer resistance to therapy, allowing for both early detection of such alterations and more precise selection of therapy.

Abstract Background Accurate reconstruction of electronic health record (EHR) treatment sequences is essential for leveraging real-world data (RWD) in clinical research and trial-eligibility platforms in multiple myeloma (MM). However, inconsistencies in RWD treatment documentation and the complexity of line-of-treatment (LOT) assignments in MM hinder reproducibility and limit comparisons against published trials. To address these limitations, we developed HEAL-MM, an automated rule-based algorithm that systematically parses structured medication data to calculate LOTs and segment treatment into standardized, clinically meaningful phases. The algorithm is designed to be interpretable, adaptable to evolving therapeutic frameworks, and compatible with widely adopted EHR data standards. Here, we report its real-world validation against expert-adjudicated treatment sequences to assess concordance, identify edge-case failure points, and support RWD integration. Methods A stratified cohort of 100 MM patients (9,136 structured treatment events) was curated from 114 unique healthcare systems compiled from the multi-institutional HealthTree Foundation harmonized registry. Cohort selection was randomized to ensure coverage across diagnostic eras (pre-2010, 2015–2020, 2022–2025) and treatment exposures, including CAR T-cell therapies, bispecific antibodies, and anti-CD38-based regimens. HEAL-MM sequenced MM regimens chronologically and applied predefined logic rules to identify LOT transitions, the initiation of a new therapeutic course based on drug changes, treatment gaps, and procedural anchors such as stem cell transplant or CAR T. Within each LOT, the algorithm segments and labels therapy into phases, clinically distinct intervals (e.g., induction and maintenance) using rule-based clinical logic. Outputs were benchmarked against expert-labeled adjudications. Concordance was evaluated using concordance index (C-index), stratified by patient, LOT depth and diagnostic era. Results: HEAL-MM demonstrated high concordance with expert-labeled treatment sequences: 0.94 for LOT identification, 0.98 for treatment phase segmentation, and 0.96 for phase label assignment. Achieving complete agreement with all expert annotations in 90% of patients for LOT assignment, 94% for the number of treatment phases identified, 77% for the specific labels applied to each phase, and 75% across all three dimensions simultaneously. Among discordant patients, the absolute difference in final LOT count was 0.5 ± 0.9. Performance was highest in early lines (LOTs 1–3; C-index: 0.97), with seven patients showing discrepancies at this stage. Concordance declined in later lines (LOT ≥4; C-index: 0.87), where 50% of the discordant cases (n=4) were attributable to misclassifications in earlier LOTs, such as medication restarts following >3-month gaps, demonstrating the compounding effect of early errors on longitudinal accuracy. Temporal stratification revealed improved accuracy in post-2020 cases (C-index: 0.94) compared to 2015–2020 cohorts (0.91), reflecting the increasing adoption of structured treatment standards and EHR data quality improvements. To assess HEAL-MM performance at clinically significant junctures, we analyzed treatment transitions that initiated a new LOT or marked a change in phase, achieving C-indices of 0.97 for LOT transitions, 0.97 for treatment phase shifts, and 0.94 for phase label assignment. Conclusions: HEAL-MM provides a validated, scalable solution for standardizing MM treatment structures in RWD environments, as it accurately detects the exact treatment entry of a new LOT 97% of the time. This rule-based framework enables reproducible LOT and phase assignment across heterogeneous EHR systems, supporting real-time cohort identification across institutions and EHR platforms. The algorithm's modular design allows for continuous integration of novel therapeutic categories and evolving treatment logic. While the algorithm maintained high fidelity overall, early-stage misclassifications consistent with increased heterogeneity and documentation ambiguity occasionally propagated downstream, underscoring the compounding effect of upstream errors in longitudinal reconstruction. With its generalizable structure and compatibility with industry-standard data models, HEAL-MM lays the foundation for structured treatment reconstruction in MM across RWD infrastructure and serves as a transferable blueprint for other hematologic malignancies.

BACKGROUND Implementing Mohs case complexity grading systems is critical due to inherent complexities of Mohs workflow coordination. Precise scheduling can improve wait times and resource allocation. OBJECTIVE To predict Mohs number of stages and reconstruction type as surrogates of case complexity and to use information derived from model training and practice to develop scheduling recommendations. METHODS Machine learning models were trained to predict number of Mohs stages and complex reconstruction across multiple practices and surgeons in an institutional enterprise, using information available on pathology reports. A workflow analysis by health systems engineering was conducted to identify improvement areas for efficiency. Information from modeling and workflow analysis was combined to develop fixed rules for scheduling recommendations. RESULTS In total, 18,473 Mohs cases were used for training/validation and 2,053 for testing. Models achieved an AUC-ROC of 0.83 (95% CI: 0.81–0.85) for prediction of complex reconstruction and 0.62 (0.60–0.66) for number of stages (1 vs 2 vs 3 or more). Anatomical location was the most valuable feature, with midface locations predicting higher case complexity and longer surgical time. CONCLUSION We demonstrated feasibility for prediction of Mohs surgery case complexity. Combining machine and human-derived knowledge provides complementary insights that can assist with improving scheduling and workflow.