Citations

Over the last decade, there have been widespread efforts to reduce non-evidence-based therapies for viral bronchiolitis. We question whether this change in practice has inadvertently impacted the diagnosis and management of other pediatric lower-respiratory-tract illnesses. We used the Pediatric Health Information System (PHIS) and logistic regression to describe trends in relative diagnosis frequency of bronchiolitis, viral pneumonia, and reactive airway disease (RAD)/asthma as well as systemic corticosteroid use among children of the age range 1 to 4 years over a 10-year period. Among 169,207 children, the relative frequency of asthma/RAD diagnoses declined over a 10-year period, while bronchiolitis and viral pneumonia diagnoses increased among children of the age range 1 to 3 years and 2 to 4 years, respectively. Frequency of systemic corticosteroid use declined. We question whether the observed shift in diagnosis from asthma/RAD to bronchiolitis or viral pneumonia is reflective of true disease pathophysiology or if it represents an unintended consequence of campaigns surrounding bronchiolitis.

Context.— Machine learning applications in the pathology clinical domain are emerging rapidly. As decision support systems continue to mature, laboratories will increasingly need guidance to evaluate their performance in clinical practice. Currently there are no formal guidelines to assist pathology laboratories in verification and/or validation of such systems. These recommendations are being proposed for the evaluation of machine learning systems in the clinical practice of pathology. Objective.— To propose recommendations for performance evaluation of in vitro diagnostic tests on patient samples that incorporate machine learning as part of the preanalytical, analytical, or postanalytical phases of the laboratory workflow. Topics described include considerations for machine learning model evaluation including risk assessment, predeployment requirements, data sourcing and curation, verification and validation, change control management, human-computer interaction, practitioner training, and competency evaluation. Data Sources.— An expert panel performed a review of the literature, Clinical and Laboratory Standards Institute guidance, and laboratory and government regulatory frameworks. Conclusions.— Review of the literature and existing documents enabled the development of proposed recommendations. This white paper pertains to performance evaluation of machine learning systems intended to be implemented for clinical patient testing. Further studies with real-world clinical data are encouraged to support these proposed recommendations. Performance evaluation of machine learning models is critical to verification and/or validation of in vitro diagnostic tests using machine learning intended for clinical practice.

Background: A pivotal clinical trial comparing the efficacy and safety of ibrutinib with zanubrutinib in patients with Waldenström macroglobulinemia (WM) did not achieve its primary endpoint of complete response or very good partial response. In addition, median progression-free survival was not reached in either treatment arm. In the absence of treatment effectiveness differentiation based on clinical trial evidence, we conducted a real-world study to assess patient characteristics, treatment patterns, and all-cause healthcare resource utilization (HRU) in patients with WM who received ibrutinib or zanubrutinib as first-line (1L) therapy. Methods: In this retrospective cohort study, we used the ConcertAI Database to collect claims data of patients with WM aged ≥18 years who initiated single-agent ibrutinib (420 mg/day) or zanubrutinib (320 mg/day) as 1L therapy on or after 9/1/2021 (index date). A 6-month washout period of guideline-recommended WM antineoplastic agents was required to confirm 1L status. Patients were excluded if they were receiving ibrutinib and zanubrutinib concurrently, died within 30 days of the index date, had a record of hematopoietic stem cell transplant, had a record of other ibrutinib indications in the 6 months before index, or were participating in a clinical trial. Baseline demographics and clinical characteristics, therapy duration (number of days between the first and last day of the medication supply), and HRU (all-cause and WM-specific) per-patient-per-month (PPPM) were assessed. Results: The study included 51 patients (ibrutinib, n=30; zanubrutinib, n=21). In both cohorts, most participants were male (ibrutinib, n=20 [67%]; zanubrutinib, n=16 [76%]), and the mean age at baseline was 73 years and 75 years, respectively. At baseline, the median Charlson Comorbidity Index was 0, and the WM-related and ibrutinib/zanubrutinib-related clinical events were similar between the two cohorts. While the zanubrutinib cohort had a higher prevalence of infections (33% vs 3%) and musculoskeletal pain (24% vs 7%) at baseline, the ibrutinib cohort had a higher prevalence of hypertension (20% vs 14%). The baseline prevalence of anemia was similar in both cohorts (ibrutinib, 27%; zanubrutinib, 29%). Mean follow-up was 6.9 months (ibrutinib) and 6.3 months (zanubrutinib). The mean ± SD duration of therapy was 152 ± 112 days (ibrutinib) and 113 ± 99 days (zanubrutinib; p=0.195). The ibrutinib cohort had a significantly lower all-cause HRU PPPM (4.53 ± 3.96 vs 8.53 ± 6.66, p=0.020) and lower WM-specific HRU PPPM (0.11 ± 0.26 vs 0.54 ± 1.14, p=0.104). Conclusions: In this real-world analysis, patients with WM treated with 1L ibrutinib had a numerically longer duration of therapy and lower all-cause and WM-specific HRU PPPM, compared to those treated with 1L zanubrutinib. Due to a small sample size and short follow-up, these findings should be interpreted cautiously.

Introduction: Molecular classifications have been developed to characterize the heterogeneity of diffuse large B cell lymphoma (DLBCL), with activation of BCL6 and Notch signaling pathways acting as driving pathogenic features in C1/BN2 subtype tumors. Among C1/BN2 defining variants are truncating/inactivating SPEN mutations. SPEN regulates the Notch signaling pathway by initiating formation of a repressive complex that facilitates transcriptional repression of Notch target genes. Our group has identified SPEN mutations enriched in DLBCL cases failing to achieve event-free status at 24 months (EFS24). This contrasted with generally favorable outcomes in C1/BN2 tumors, suggesting a potential high-risk molecular feature with unexplored significance. Presenting an opportunity to expand underlying mechanisms and risk stratification in C1/BN2 tumors, we aimed to use a functional genomic approach, combined with clinical data, to characterize the alternative biology stemming from SPEN truncating mutations. Methods: FFPE tumor biopsies from newly diagnosed DLBCL cases (N=444) enrolled in the Mayo Clinic/Iowa Lymphoma Molecular Epidemiology Resource (MER) program were analyzed by WES (n=404) and RNA-seq (n=321). Biologic and clinical variables were annotated for all cases. CRISPR/Cas9 introduced truncating mutations to the endogenous SPEN locus in OCI-LY3 & SU-DHL-2 ABC-type human cell lines to establish models of SPEN depletion. Functional in vitro analysis utilized standard methods for RNA-seq, cell proliferation, western blot, and flow cytometry. Results: WES identified 29/404 ndDLBCL (7.2%) harboring SPEN mutations; n=17 missense and n=12 nonsense or frameshift insertions/deletions (truncating variants - [trunc]), the later considered as the focal point of our study. Against all cases, SPEN trunc cases displayed enrichment for high-risk factors including non-GCB COO (p=0.06, OR=3.7, 95% CI [0.9, 22.2]) and primary refractory disease (PTR; p=0.02, OR=5.0, 95% CI [1.05, 19.9]), and demonstrated inferior OS (p=0.06, HR=2.2), where a majority of clinical progression events occurred within 12 months of diagnosis. Leveraging RNA-seq, we observed 11 of 11 SPEN trunc cases exhibiting a double-hit negative gene expression signature, 9/11 cases with inflammatory or depleted microenvironment (LME), and 8/11 with a signature associating with intermediate/high risk of EFS24 failure. Applying the LymphGen algorithm, 5/12 SPEN trunc cases received the BN2 classification, observing co-occurring mutations with CD70 (p<0.01, OR=9.68), BRCA2 (p=0.03, OR=5.84), TET2 (p=0.02, OR=4.91), IRF4 (p=0.02, OR=7.27), and PTPRD (p<0.01, OR=8.31) genes, and were mutually exclusive with ARID1A, BCL2, EZH2, FAS, HLA-B, IRF8, and TMEM30A mutations ( Figure 1A). Isolating SPEN trunc against BN2 cases without SPEN variants (n=14), despite similar clustering of non-GCB COO (p=0.7), PTR was exclusively observed in SPEN trunc cases, accompanied again by inferior OS (p=0.07, HR=4.2; Figure 1B). In the OCI-LY3 SPEN depletion model, GSEA analysis from RNA-seq revealed activation of E2F (NES=2.7; p=9.4E-20) and MYC target genes (NES=2.3; p=4.1E-6), while both OCI-LY3 and SU-DHL-2 models featured depletion of antigen presentation genes (NES= -1.8, p=5.7E-4; NES= -2.2, p=3.8E-7, respectively). A decrease of MHC class II proteins (HLA-DR, HLA-DP) was confirmed by flow cytometry in both models (p<0.01). Hypothesizing loss of SPEN enhances reliance on the BCL6-HDAC3 axis featured in BN2 tumors, we treated both SPEN depletion models with a class-I HDAC inhibitor highly selective for HDAC3, BRD-3308, and indeed observed a 1.4 and 2.4-fold increase in sensitivity compared to WT cells (p=0.07, p<0.01, respectively). Significance: We find SPEN trunc mutations in poor acting DLBCL cases with clinical and biologic characteristics that diverge from C1/BN2 biology. The enhanced rate of clinical progression highlights urgency to better characterize these tumors. Our data suggests loss of SPEN regulatory mechanisms may act cooperatively with C1/BN2 biology to facilitate tumor immune evasion and drive progression through Notch-mediated pathways. Importantly, in vitro evidence shows SPEN deficient lymphoma cells are susceptible to targeted therapeutics against dependent pathways, providing rationale for advancing individualized care strategies for affected patients.

Background: QOL can be affected by disease burden, side effects of treatment, and psychosocial effects of living with cancer. Although MZL is considered incurable, we previously found that patients diagnosed from 2002-2015 had a good QOL both at baseline (BL) and the first decade after diagnosis, regardless of treatment choice (Bommier et al., submitted). Here, we evaluate whether these findings can be extended to the 2015-2020 treatment era and a more diverse US cohort. Methods: Adults with newly diagnosed MZL from 8 academic centers were prospectively enrolled within 6 months of diagnosis from 2015-2020 in the LEO cohort (NCT02736357). Participants were actively followed for disease progression/relapse, retreatment and death. QOL was measured at BL, and at 1, 2, 3 and 5 years after diagnosis using the Functional Assessment of Cancer Therapy-Lymphoma (FACT-Lym) scale (range 0-164), which consists of the FACT-General (FACT-G, range 0-104) plus a lymphoma-specific scale (range 0-60). Total FACT-G measures 4 QOL domains (range): physical (0-28), social/family (0-28), emotional (0-24), and functional (0-28) well-being (WB). We also used the Trial Outcome Index (TOI; range 0-112), which consists of FACT-G physical and functional WB plus the lymphoma-specific scale and is commonly used as an endpoint in treatment trials. All MZL patients with both BL and at least one follow-up QOL questionnaire were eligible for analysis. Patients were grouped by initial management: active surveillance (observation), local/antibiotic therapy (radiation, surgery, or anti- H. pylori therapy), or systemic therapy (chemotherapy, targeted agents, or antibody therapy). Change in QOL scores over time was analyzed using mixed models for repeated measurements, and initial treatment groups were compared. Results: Of 384 patients, 60% were female, 91% were White, 6.6% were Black/African American, and 13% were Hispanic (all races). The median age was 64 years (range 19-94). For initial management, 28% (N=106) were observed, 31% (N=118) received local/antibiotic therapy, and 42% (N=160) received systemic therapy. Across treatment groups, there were no significant differences by age, gender, race, or performance status, while there were differences by Hispanic ethnicity (p=0.006), Ann Arbor stage (p<0.001) and MALT-IPI score (p<0.001) (Table). At BL, patients receiving systemic therapy had lower physical WB (p=0.008), emotional WB (p=0.04), functional WB (p=0.03), FACT-G total score (p=0.008), lymphoma subscale (p<0.001), FACT-Lym (p=0.001) and TOI (p<0.001) compared to observation and local/antibiotic therapy (Table). While FACT-G total scores were largely stable through 5 years after diagnosis, social/family WB declined from BL for patients initially receiving either local/antibiotic therapy (mean at 1-year -1.4, p<0.05; 2-year -2.1, p<0.05; 3-year -1.6, p<0.05; 5-year-2.9, p<0.05) or systemic therapy (mean at 2-year -1.2 , p<0.05; 3-year -1.4, p<0.05; and 5-year -1.2, p<0.05), while this was stable for observed patients. In patients receiving initial systemic therapy, there was a transient improvement in the lymphoma-specific scale (mean +1.7, p<0.05) and TOI (mean +3.7, p<0.05). Otherwise, QOL remained stable over time and global QoL (FACT-G, FACT-Lym), TOI, Lymphoma-specific scale (Figure), and the other FACT-G subscales (physical, emotional, or functional WB) did not differ by initial treatment strategy. Conclusions: In the diverse LEO cohort from the recent treatment era, we confirm our prior findings for newly diagnosed MZL patients that QOL is globally stable over the first 5 years after diagnosis; emotional WB is not associated with initial treatment type, providing evidence that being initially observed does not negatively impact emotional health; and there is a decline in social/family WB (support from friends and family, family acceptance of illness, family communication, close to partner), which supports the need to consider interventions for patients to address these unmet needs. In addition, this study extends these findings to the lymphoma-specific scale, FACT-Lym, and TOI, which all showing largely stable trends in the 5 years after diagnosis and suggest that recent MZL treatment strategies, including systemic therapies, neither impair nor improve QOL.

Introduction: High dose chemotherapy and autologous stem cell transplantation (ASCT) is one of the most effective therapies in light chain (AL) amyloidosis. It is associated with a high hematologic response rate and leads to durable remissions. High dose melphalan is the most frequently used conditioning chemotherapy, as has been used in multiple myeloma (MM). Melphalan dose is based on renal function, age, degree of cardiac involvement and patients' frailty and is predictive of outcomes. While studies that assessed the conditioning schedule have been conducted in MM, with variable results, no such studies were done in AL amyloidosis. Methods: In this retrospective study, we explored the impact of the conditioning schedule on post-transplant outcomes in a large multicenter study. Patients with AL amyloidosis who underwent ASCT for the primary diagnosis of AL amyloidosis between January 1, 2003, and December 31, 2020, were included in this study. Results: One thousand seven hundred and eighteen (n=1718) patients from 9 centers are included. The median age was 58 years, and 61% were male. Most patients had one (43%) or two (33%) involved organs, with the most involved organs were the kidneys (69%) followed by the heart (52%). Induction therapy was given to 52% of patients. The conditioning regimen was almost exclusively melphalan (99.2%), with 65% receiving full dose melphalan (≥170 mg/m 2). Melphalan was given over two days in most patients (80%). A rest day between conditioning and stem cell infusion was provided in 53% of patients, mostly in those who received melphalan on a single day rather than over 2 days (87% vs 44%, P<0.001). Post-transplant hematologic response was available for 1569 patients (91% of patients) and were assessed from baseline. Complete response (CR) and very good partial responses (VGPR) were achieved post-ASCT in 33.7% of patients each. Low dFLC-partial response (PR) was achieved in 2.7% and PR in 18% of patients. Altogether, satisfactory post-ASCT response (CR, VGPR, or low-dFLC PR combined) was achieved in 70% of patients. Compared to no induction, less than a PR to induction was associated with a lower probability of achieving satisfactory post-ASCT response, while PR or better to induction had higher chances of satisfactory post-ASCT response (65% vs 46% vs 84%, P<0.001). A single day of melphalan conditioning was associated with a higher probability of satisfactory post-ASCT response compared to two days of conditioning (82% vs 67%, P<0.001). Omitting a day of rest between conditioning and stem cell infusion was also associated with higher satisfactory post-ASCT response compared to having a rest day (76% vs 65%, P<0.001). The Table lists other predictors for a post-ASCT satisfactory hematologic response. In a nominal multivariate logistic regression analysis, independent predictors for higher post-ASCT satisfactory hematological response rate included: PR or better to induction therapy, female gender, kappa light chain isotype, lower dFLC at baseline, and kidney and liver involvement. In addition, full dose melphalan, conditioning given over a single day and having no rest day between conditioning and stem cell infusion were also associated with a higher likelihood of satisfactory post-ASCT response. Post-ASCT hematologic CR rate was higher in those receiving full dose conditioning (36% vs 28%, P<0.001), in those having no rest day (40% vs 28%, P<0.001) but not in those who received one day of conditioning versus two days (35% vs 33%, P=0.5). The rate of ICU admission was higher in those receiving no day rest (17.7% vs 12.6%), P=0.003) and among those who received reduced dose melphalan (18.1% vs 13.3%, P=0.008) but not based on number of conditioning days (P=0.56). Day 100 mortality was significantly higher in the no rest day group (5.6% vs 3.2%, P=0.01) and among those receiving reduced dose melphalan (6.8% vs 3.1%, P=0.004) compared to their counterparts, but not based on the number of days of conditioning (P=0.89). The median follow-up in the entire cohort was 104 months (95% CI 98-110). In a multivariate analysis for overall survival (OS), the conditioning dose but not the conditioning schedule was independently associated with OS. Conclusions: This study is the first to demonstrate the effect of melphalan conditioning schedule on post-transplant response. This study supports a single day melphalan conditioning in AL amyloidosis to improve post-ASCT response.