Citations

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.

Introduction: DH/THL also classified as high-grade B cell lymphomas (HGBCL) with MYC and BCL2 and/or BCL6 rearrangements have higher risk of progression after initial treatment and poor response to salvage therapies compared to diffuse large B cell lymphoma (DLBCL), NOS. Recently, CART cell therapy has shown encouraging responses in treatment of relapsed refractory (R/R) DLBCL. In this retrospective study, we evaluated whether survival of (R/R) DH/THL is improved in the CART era. Methods: In this multicenter retrospective study, adult pts (pts) with R/R DH/THL after anthracycline-based therapy were included. Demographic and clinical variables were obtained from electronic health records. Study objectives was overall survival (OS) at 2 year landmark. We utilized full propensity score matching to compare progression free survival (PFS) and OS of CART with 2 nd line chemoimmunotherapy. Results: 278 DH/THL pts treated between 2010-2020 were included. Median age was 63 years (range 52-74), 163 (58%) were male, 119 (43%) were classified as DLBCL and 116 (42%) as HGBCL; 40 (14%) were transformed from follicular lymphoma. 225 (82%) had advanced stage, 161 (67%) had elevated LDH, 85 (33%) had bulky disease, 74 (59%) had B symptoms. 105 (38%) pts treated for relapsed disease before 2018 were included in the pre-CART era cohort. 168 (60%) pts treated from 2018 onwards were included in the CART era cohort. 102 (60%) pts in CART cell era received CART cells, 40 (14%) received polatuzumab vedotin (pola), 10 (3.5%) received loncastuximab tesirine, tafasitamab and bispecific antibody each. Characteristics at relapse were similar in both groups. Median follow up was 8.9 (0.2-146) months. 2-year OS estimate was 36% (CI 95: 30-42%) in entire population. Patients treated in CART cell era had significantly higher OS (2 yr OS: 43% [CI 95: 35-53]) compared to pts treated in pre-CART cells era (2-year OS: 24% [CI 95: 16-34], p=0.0006) (Figure 1). In multivariable analyses, increasing age (HR = 1.02 (CI 95: 1.01, 1.04), p=0.003), extranodal disease (HR = 1.41 (CI 95: 1.00, 1.99), p<0.001), elevated LDH (HR = 1.57 (CI 95: 1.11, 2.24), p=0.01) were associated with inferior OS. Increasing time to relapse (HR = 0.98 (CI 95: 0.97, 0.99), p<0.001) was associated with superior OS. BCL6 rearrangement did not affect OS. When adjusted for these factors, treatment with CART cells was associated with higher 2 yr OS (HR = 0.33 (CI 95: 0.22, 0.49), p<0.001). Treatment with polatuzumab (HR = 1.07 (CI 95: 0.71, 1.6), p=0.07) was not significantly associated with OS in univariable analysis. 102 pts underwent CART cell therapy at any point for relapse; 94 (93%) were refractory to prior chemotherapy. 62 (60%) responded and 52 (50%) had CR after CART cells. 2-year PFS and OS after CART cell therapy were 28% (CI 95: 19-40) and 40% (CI 95: 30-53), respectively. Increasing time to CART was associated with inferior OS (HR: 1.02 (CI 95: 1.0-1.03, p=0.03) but not PFS (HR: 1.02 (CI 95: 0.99-1.04), p=0.2). 219 (79%) pts received curative intent salvage chemotherapy for 2 nd line. 2-year PFS and OS were 15% (CI 95: 11-20) and 38% (CI 95: 31-45), respectively. 21 pts received CART cell therapy for 2 nd line. 2 year PFS and OS in this small number of pts were 52% (CI 95: 35-79) and 70% (CI 95: 52-94), respectively. Due in part to the small number of pts who received CART cells in 2 nd line, we utilized full propensity score matched analysis to compare each patient in CART cell group to all possible matched controls in salvage chemotherapy group. Variables matched were age, LDH, extra nodal disease and time to relapse. In propensity score matched analysis, pts who received CART cells had significantly higher 2 year PFS (HR: 10.5 (CI 95: 3.4-17.5), p=002) and 2 year OS (HR: 7.5 (CI 95: 1.4-13), p=0.01). Conclusions: Survival of r/r DH/THL appears to have improved in the era of CART cell therapy. About 2/3 of pts refractory to chemotherapy respond to CART cells, and in multivariable analysis, receipt of CART cells was associated with improved OS. Increasing time to CART cells was associated with inferior survival. In a propensity score matched analysis utilizing full matching, CART cell therapy in 2 nd line was associated with significantly higher 2 year PFS and OS compared to curative intent salvage therapy. Other CD 19 and CD 20 targeting agents approved after CART cell therapy might contribute to favorable survival, effect of these agents could not be assessed because of small numbers.

While the cure rate of pediatric acute lymphoblastic leukemia (ALL) now exceeds ~90% with contemporary combination chemotherapy, the prognosis for adults with ALL remains significantly inferior with long-term overall survival ranging from 50% to 70%. Recent studies have uncovered marked differences in ALL genomics between children and adults, with some high-risk subtypes becoming more prevalent with age. However, the underlying biology of age-related disparities in ALL is not fully understood, especially with regard to differences in leukemia sensitivity to chemotherapy. To address this knowledge gap, we performed ex vivo drug sensitivity profiling (i.e., pharmacotyping) of 21 anti-leukemia agents on primary B-ALL diagnostic samples from 767 pediatric (age, 0-18 years) and 309 adult (19-84 years) patients. Drug sensitivity was measured as LC 50: the concentration of drug required to kill 50% of the leukemia cells (PMID: 36604538). A total of 7,975 unique LC 50 values were experimentally determined. RNA-seq was used for subtype classification and gene expression analysis. Among 21 drugs, seven showed significant differences in overall LC 50 between children and adults ( P<0.05 after Bonferroni correction): children displayed higher sensitivity to asparaginase, prednisolone, mercaptopurine, daunorubicin, and inotuzumab, while adults showed higher sensitivity to dasatinib and nelarabine. In multivariate models adjusting for 23 ALL molecular subtypes, only mercaptopurine remained significantly associated with age ( P=1.5×10 -5), suggesting that age-related differences in drug sensitivity can be primarily attributed to the variation in ALL subtypes between children and adults. For mercaptopurine, within KMT2A, CRLF2, and DUX4 subtypes, pediatric samples consistently showed a lower LC 50 than adults carrying the same genomic abnormality ( P=0.032, 0.0045, and 0.02, respectively). To explore intra-subtype heterogeneity, we performed unsupervised clustering using gene expression data for each of these three subtypes. Remarkably, within each of these subtypes, we identified two clusters with distinct transcriptomic profiles that were also largely segregated by age group, i.e., an adult-dominated cluster (C-a) and a pediatric-dominated cluster (C-p). In the KMT2A subtype, cases in C-a exhibited an over-representation of the KMT2A:: AFF1 fusion, and resistance to mercaptopurine ( P=0.029), prednisolone ( P=0.0039), vincristine ( P=0.046) and cytarabine ( P=0.0037). Within CRLF2 ALL, cases in C-a were associated with the presence of BCR:: ABL1-like signature and IGH:: CRLF2 rearrangements, and were more resistant to mercaptopurine ( P=0.0073) and prednisolone ( P=0.00036) compared to those in C-p. For DUX4 ALL, C-a was characterized by an under-representation of ERG deletions and resistance to mercaptopurine ( P=0.0056) and prednisolone ( P=0.0031), compared to C-p within DUX4. To explore the clinical relevance of this heterogeneity, we analyzed the in vivo treatment response of KMT2A (N=35), CRLF2 (N=59) and DUX4 (N=118) B-ALL enrolled in six frontline ALL trials. Compared to cases in C-p (usually drug-sensitive), those in C-a (usually drug-resistant) consistently had significantly poorer initial treatment responses as measured by persistent end-of-induction minimal residual disease (≥0.01%) in KMT2A (58% vs 9%; P=0.0063), CRLF2 (74% vs 41%; P=0.030), and DUX4 (66% vs 40%; P=0.0058) ALL. In conclusion, these studies have revealed important new insights into the pharmacogenomic basis of age-related differences in B-ALL treatment response. These results indicate that both inter- and intra-subtype heterogeneity contribute to inferior prognosis in adults with ALL, but also point to therapeutic opportunities to improve their outcomes.

Introduction Frontline therapy for pts with follicular lymphoma (FL) is not based on the molecular profile of the tumor. The PI3K pathway is central to FL biology, but the molecular profile of tumors most sensitive to PI3K inhibition (PI3Ki) is unknown. Copanlisib inhibits both PI3Kα and δ isoforms, and we hypothesized that most pts with treatment naïve FL will have PI3Ki-sensitive tumors and durable complete responses (CRs) may be achievable with short durations of targeted therapy. A gene expression profiling (GEP) predictor utilizes a combination of the expression of 45 selected informative genes and an additional 19 housekeeping genes to identify FL pts at high-risk for early progression within 24 months (Wright et al. ASH Annual Meeting2022). We also hypothesized that copanlisib would induce durable CRs in high-risk pts by the GEP assay. Here, we report a preliminary analysis of an ongoing “window of opportunity” study of copanlisib followed by response-adapted copanlisib and rituximab in treatment-naïve pts with FL (NCT03789240). Methods Pts with untreated grade 1-2, 3A FL, ≥stage 2 are eligible if systemic therapy is indicated: symptoms, progressive lymphadenopathy, or organ compromise. Prior radiation is permitted. Eligibility includes age ≥18 and adequate organ function unless due to lymphoma. HIV, CMV, Hep B or C, and autoimmune conditions are excluded. Pts first receive copanlisib 60mg on days 1, 8, and 15 of a 28-day “window of opportunity” to explore the activity of copanlisib. Following the window, pts receive 6 cycles of copanlisib 60mg on days 1, 8, and 15 of a 28-day cycle along with rituximab 375mg weekly x 4 then on day 1 of each cycle and response is assessed with FDG-PET and CT scans. Pts without response after C6 are taken off study and those with a partial response (PR) can continue for an additional 6 cycles and then stop therapy. Pts who achieve a CR after 6 cycles stop therapy. No maintenance is given. Supportive care includesPCP prophylaxis. The primary endpoint is the CR rate with secondary endpoints of safety, duration of CR, and PFS. Exploratory objectives include identification of a predictive signature of PI3Ki response and the response rate in high-risk pts identified by the GEP assay. Results Twenty-eight pts have enrolled. Median age was 56y (range, 24-80). Seventeen (61%) and 12 (43%) pts had high-risk FLIPI and FLIPI-2 scores ≥3, respectively; 6 (21%) pts had Grade 3A FL, and 17 (61%) pts met two or more GELF criteria. After copanlisib monotherapy in the window, 27 (96%) pts had tumor reductions by CT with a median change of -38% (-62% to +16%) (Figure 1). Among 22 pts who completed 6 cycles, 21 (96%) responded, including 9 (41%) pts with PR and 12 (55%) pts with CR. Of 9 pts with PR who received an additional 6 cycles of therapy, only 1 (11%) converted to a CR bringing the overall CR rate to 59% (13/22 pts). After a median follow-up of 25.8 months, the 2-year PFS was 43.2% (95% CI, 20.7-64.0%). The median duration of CR was 29.7 months (95% CI, 6.3 months - not estimable) with 8 (62%) CRs ongoing (Figure 2). Toxicity was evaluated in 28 pts across 220 cycles. Most common were rash (64%), mucositis (50%), and nausea/vomiting (50%), all G1-2 except 1 (4%) pt with G3 mucositis. The most common ≥G3 toxicities were neutropenia and infections in 5 (18%) pts each. ≥G3 neutropenia occurred in 10 (5%) cycles, responded to therapy interruption and/or growth factors, and did not require dose reduction. Two (8%) pts had copanlisib dose reduced, 1 (4%) each due to rash and mucositis. Therapy was prematurely discontinued in 5 pts: 3 due to COVID-19, 1 for hepatotoxicity that resolved, and 1 due to infusion reaction to rituximab. Nine of 24 (38%) pts were poor risk by the GEP assay. Complete responses were observed in 71%, 67%, and 57% of pts categorized as high-risk by FLIPI, FLIPI-2, and GEP assay, respectively. The CR rate was not affected by bulky disease (any LN >7cm; p=1.00), high-risk FLIPI (p=0.19), high-risk FLIPI-2 (p=0.67), poor-risk GEP (p=1.00), or metabolic tumor volume (p=0.33). Conclusion Nearly all pts with treatment naïve FL have PI3Ki-responsive tumors, and the combination of copanlisib with rituximab achieved a CR rate of 59% at the end of induction therapy. The median duration of CR is nearly 30 months, including ongoing remissions without maintenance. CRs were observed in pts deemed high-risk by the GEP assay and FLIPI score. The safety profile is excellent, with mostly G1-2 toxicities manageable with supportive care.

Healthcare inequities are often attributed to large-scale structural barriers for access to medical care or patient-medical staff interactions. However, the role of the electronic medical record (EMR) in perpetuating hidden biases has been poorly characterized. This passive form of sharing implicit biases with unknown readers may have profound ripple effects asynchronous in place and time from the note writer. This study aims to illustrate racial imbalances in ad hominem patient characterizations as documented by medical staff, regardless of role, in a large metro healthcare system. The use of certain potentially stigmatizing language in emergency department (ED) and inpatient notes for hematology/oncology patients was investigated. Retrospective chart review of four hospitals within a university network in a major metro area between 1/1/2019 - 6/30/2021 was undertaken. Patients ≥22 years old were included with two primary hypotheses in mind: 1) Negative terms are used more frequently for underrepresented minorities, and 2) words with shock value (i.e. curse words) are used more often than more general descriptors of these terms (e.g. “cursing” or “profanity”). Nine positive and twelve negative terms were evaluated, based on the authors' observations of their use in the EMR in past encounters (Table 1). Some were personal descriptors in which the ED or inpatient note writer would be less likely than outpatient providers to have a full sense of outpatient care or social determinants of health (e.g. compliant). Others were behavioral descriptions (e.g. hostile) or words with minimal medical relevance (e.g. curse words). If multiple notes in an encounter included the same word(s), the encounter was only analyzed once for each word to avoid word count inflation, though re-occurrences for the same patient in subsequent encounters were counted. Authors' manual data review using natural language processing software confirmed appropriate use of terms. Generalized estimating equations were used for race-based comparisons to account for same-patient clustered encounters. Odds ratios (OR) and 95% confidence intervals (CI) were calculated for self-reported demographic comparisons with White as the reference category. P≤0.05 was considered significant. A total of 1,059 encounters for 889 hematology/oncology patients were analyzed. Mean age was 61 years (standard deviation 16.7) and 49.4% were female. White race was reported for 80.2% of patients (72% system-wide prevalence during same timeframe) and Black for 11.4% (15.2% system-wide); all others were ≤4% so only White/Black comparisons were included. Positive terms were more prevalent than negative ones (n=535 vs 459). However, negative terms were found more often for Black individuals (OR 2.24, CI 1.41 - 3.56, P<0.001), while the positive terms were less frequently documented in Black patients relative to White patients (OR 0.51, CI 0.33 - 0.77, P=0.001), and neither category was concordant with the overall patient racial demographics. Other significant racial differences existed for the terms “Argumentative”, “Disrespectful”, “Noncompliant”, and “Polite”, all discrepancies which favored White patients (Table 2). Specific curse words were documented nearly as often as “cursing”, “swearing”, or “profanity”, but no racial differences were observed. Some words in the initial search (“Brave”, “Courageous”, “Empathetic”, and “Ambitious”) were not found in hematology/oncology notes in this timeframe. EMR documentation in hospital-based hematology/oncology care demonstrated evidence of racial discrepancies in language that may carry stigma, positively or negatively. Importantly, many of these terms have questionable, if any, medical relevance. The use of stigmatizing terminology in the EMR may reflect implicit biases that extend past the direct patient encounter, passively and subconsciously prejudicing future care from other staff members. Time-saving heuristics and documentation shortcuts like copy-forwarding may further ossify healthcare inequities. This phenomenon is undoubtedly not limited to this hospital system or the hematology/oncology field, but quantitative real-world evidence has been scant. Improved awareness of stigmatizing language in documentation may be one step towards mitigating this contributor to unequal care for hematology/oncology providers and their patients.

ABSTRACT Standardized approaches to phage susceptibility testing (PST) are essential to inform selection of phages for study in patients with bacterial infections. There is no reference standard for assessing bacterial susceptibility to phage. We compared agreement between PST performed at three centers: two centers using a liquid assay standardized between the sites with the third, a plaque assay. Four Pseudomonas aeruginosa phages: PaWRA01ø11 (EPa11), PaWRA01ø39 (EPa39), PaWRA02ø83 (EPa83), PaWRA02ø87 (EPa87), and a cocktail of all four phages were tested against 145 P . aeruginosa isolates. Comparisons were made within measurements at the two sites performing the liquid assay and between these two sites. Agreement was assessed based on coverage probability (CP 8 ), total deviation index, concordance correlation coefficient (CCC), measurement accuracy, and precision. For the liquid assay, there was satisfactory agreement among triplicate measurements made on different days at site 1, and high agreement based on accuracy and precision between duplicate measurements made on the same run at site 2. There was fair accuracy between measurements of the two sites performing the liquid assay, with CCCs below 0.6 for all phages tested. When compared to the plaque assay (performed once at site 3), there was less agreement between results of the liquid and plaque assays than between the two sites performing the liquid assay. Similar findings to the larger group were noted in the subset of 46 P . aeruginosa isolates from cystic fibrosis. Results of this study suggest that reproducibility of PST methods needs further development.

Abstract During aging, muscle gradually undergoes sarcopenia, the loss of function associated with loss of mass, strength, endurance, and oxidative capacity. However, the 3D structural alterations of mitochondria associated with aging in skeletal muscle and cardiac tissues are not well described. Although mitochondrial aging is associated with decreased mitochondrial capacity, the genes responsible for the morphological changes in mitochondria during aging are poorly characterized. We measured changes in mitochondrial morphology in aged murine gastrocnemius, soleus, and cardiac tissues using serial block‐face scanning electron microscopy and 3D reconstructions. We also used reverse transcriptase‐quantitative PCR, transmission electron microscopy quantification, Seahorse analysis, and metabolomics and lipidomics to measure changes in mitochondrial morphology and function after loss of mitochondria contact site and cristae organizing system (MICOS) complex genes, Chchd3 , Chchd6 , and Mitofilin . We identified significant changes in mitochondrial size in aged murine gastrocnemius, soleus, and cardiac tissues. We found that both age‐related loss of the MICOS complex and knockouts of MICOS genes in mice altered mitochondrial morphology. Given the critical role of mitochondria in maintaining cellular metabolism, we characterized the metabolomes and lipidomes of young and aged mouse tissues, which showed profound alterations consistent with changes in membrane integrity, supporting our observations of age‐related changes in muscle tissues. We found a relationship between changes in the MICOS complex and aging. Thus, it is important to understand the mechanisms that underlie the tissue‐dependent 3D mitochondrial phenotypic changes that occur in aging and the evolutionary conservation of these mechanisms between Drosophila and mammals.