Comparative Analysis of Machine-Learning Model Performance in Image Analysis: The Impact of Dataset Diversity and Size.

BACKGROUND: This study presents an analysis of machine-learning model performance in image analysis, with a specific focus on videolaryngoscopy procedures. The research aimed to explore how dataset diversity and size affect the performance of machine-learning models, an issue vital to the advancement of clinical artificial intelligence tools. METHODS: A total of 377 videolaryngoscopy videos from YouTube were used to create 6 varied datasets, each differing in patient diversity and image count. The study also incorporates data augmentation techniques to enhance these datasets further. Two machine-learning models, YOLOv5-Small and YOLOv8-Small, were trained and evaluated on metrics such as F1 score (a statistical measure that combines the precision and recall of the model into a single metric, reflecting its overall accuracy), precision, recall, mAP@50, and mAP@50-95. RESULTS: The findings indicate a significant impact of dataset configuration on model performance, especially the balance between diversity and quantity. The Multi-25 × 10 dataset, featuring 25 images from 10 different patients, demonstrates superior performance, highlighting the value of a well-balanced dataset. The study also finds that the effects of data augmentation vary across different types of datasets. CONCLUSIONS: Overall, this study emphasizes the critical role of dataset structure in the performance of machine-learning models in medical image analysis. It underscores the necessity of striking an optimal balance between dataset size and diversity, thereby illuminating the complexities inherent in data-driven machine-learning development.

Information-theoretical limit on the estimates of dissipation by molecular machines using single-molecule fluorescence resonance energy transfer experiments.

Single-molecule fluorescence resonance energy transfer (FRET) experiments are commonly used to study the dynamics of molecular machines. While in vivo molecular processes often break time-reversal symmetry, the temporal directionality of cyclically operating molecular machines is often not evident from single-molecule FRET trajectories, especially in the most common two-color FRET studies. Solving a more quantitative problem of estimating the energy dissipation/entropy production by a molecular machine from single-molecule data is even more challenging. Here, we present a critical assessment of several practical methods of doing so, including Markov-model-based methods and a model-free approach based on an information-theoretical measure of entropy production that quantifies how (statistically) dissimilar observed photon sequences are from their time reverses. The Markov model approach is computationally feasible and may outperform model free approaches, but its performance strongly depends on how well the assumed model approximates the true microscopic dynamics. Markov models are also not guaranteed to give a lower bound on dissipation. Meanwhile, model-free, information-theoretical methods systematically underestimate entropy production at low photoemission rates, and long memory effects in the photon sequences make these methods demanding computationally. There is no clear winner among the approaches studied here, and all methods deserve to belong to a comprehensive data analysis toolkit.

Impact of debulking therapy on the clinical outcomes of axicabtagene ciloleucel in the treatment of relapsed or refractory large B-cell lymphoma.

Axicabtagene ciloleucel (axi-cel), an autologous anti-CD19 chimeric antigen receptor T-cell therapy, was approved for relapsed/refractory (R/R) large B-cell lymphoma (LBCL) based on the results from pivotal Cohorts 1+2 of ZUMA-1 (NCT02348216). ZUMA-1 was expanded to investigate safety management strategies aimed at reducing the incidence and severity of cytokine release syndrome (CRS) and neurologic events (NEs). Prospective safety expansion Cohort 5 evaluated the impact of debulking therapy, including rituximab-containing immunochemotherapy regimens and radiotherapy, in axi-cel-treated patients; the CRS and NE management strategy paralleled those in Cohorts 1+2. Among the 50 patients in Cohort 5 who received axi-cel, 40% received ≥3 prior lines of chemotherapy, and 40% had disease that progressed while on the most recent chemotherapy. Forty-eight patients (96%) received debulking therapy, 14 (28%) radiotherapy only, and 34 (71%) systemic immunochemotherapy. Median decrease in tumor burden (per sum of product of diameters of target lesions) relative to screening was 17.4% with R-ICE/R-GDP, 4.3% with other debulking chemotherapies, and 6.3% with radiotherapy only. All patients were followed for ≥8 months. CRS was reported in 43 patients (86%), with 1 patient (2%) experiencing grade ≥3. NEs were reported in 28 patients (56%), with 6 (12%) experiencing grade ≥3. Cytopenias were the most frequent grade ≥3 adverse event (AE); 19 (38%) and 18 (36%) treated patients had any and grade ≥3 prolonged thrombocytopenia, respectively, and 25 (50%) and 24 (48%) patients had any and grade ≥3 prolonged neutropenia, respectively. Overall, patients who received debulking chemotherapy had higher incidences of serious treatment-emergent AEs than those who received radiotherapy only. At the 24-month analysis, objective response rate was 72%, and complete response rate was 56%. Median duration of response, progression-free survival, and overall survival were 25.8, 3.1, and 20.6 months, respectively. These results from exploratory Cohort 5 demonstrate the feasibility of debulking prior to axi-cel, and together with current real-world evidence, suggest that debulking regimens may help minimize the frequency and severity of CRS and NEs in patients with R/R LBCL. The incidence of other AEs observed in Cohort 5 suggest the risk/benefit profile was not improved via the debulking regimens studied here.

Tandem Autologous Stem Cell Transplantation Does Not Benefit High-Risk Myeloma Patients in the Maintenance Era: Real-World Results from The Canadian Myeloma Research Group Database.

In patients with multiple myeloma (MM), the presence of high-risk cytogenetic abnormalities is associated with worse disease control and survival. Autologous stem cell transplant (ASCT) does benefit these patients. Tandem transplantation has been explored as a means to deepen responses and further improve survival however, its role remains controversial. This is particularly true in the era of novel agent induction and post-transplant maintenance therapy. The aim of this study was to use the Canadian Myeloma Research Group database and examine a large cohort of real-world patients comparing the outcomes of tandem versus single ASCT specifically in high-risk patients receiving novel agent-based induction and post-transplant maintenance. The data for this study was derived retrospectively from a comprehensive national-level database of Canadian patients with MM. High-risk cytogenetics was defined as presence of del17p, t(4;14), or t(14;16). Those receiving allogeneic transplant were excluded. Tandem transplantation was defined as a second ASCT performed consecutively without interim relapse or progression after first ASCT. Those with relapse or progressive disease within 3 months of completing a first transplant were excluded. We compared response depth, progression-free, and overall survival (OS) based on single or tandem transplantation procedures. The impact of covariates of interest was also assessed. A total of 381 patients with high-risk cytogenetics were identified. A total of 242 received single and 139 patients received tandem transplants. All received post-transplant maintenance. The most common induction regimen for these patients was cyclophosphamide, bortezomib, and steroids (CyBorD, 87%). Forty-one patients (10.8%) required reinduction prior to first ASCT. The best overall responses at any time were 98.3% (90.5% ≥ very good partial response [VGPR]) and 98.6% (89.9% ≥ VGPR) in the single and tandem ASCT groups, respectively. Survival outcomes were similar with the median progression-free survival for single or tandem ASCT of 35.2 and 35.3 months (P = .88) and the median OS were 92.6 and 88.9 months, respectively (P = .72). No statistically significant differences were seen based on type of cytogenetic abnormality or type of maintenance. This was confirmed on multivariate analysis. In the real-world setting, tandem ASCT does not improve outcomes for MM patients with high-risk cytogenetics. This may be driven by the use of effective pre- and post-ASCT therapies. The development of more potent induction and consolidation along with current nearly ubiquitous continuous maintenance therapies until disease progression does not support the use of a second high-dose procedure.

Nucleolytic processing of abasic sites underlies PARP inhibitor hypersensitivity in ALC1-deficient BRCA mutant cancer cells.

Clinical success with poly (ADP-ribose) polymerase inhibitors (PARPi) is impeded by inevitable resistance and associated cytotoxicity. Depletion of Amplified in Liver Cancer 1 (ALC1), a chromatin-remodeling enzyme, can overcome these limitations by hypersensitizing BReast CAncer genes 1/2 (BRCA1/2) mutant cells to PARPi. Here, we demonstrate that PARPi hypersensitivity upon ALC1 loss is reliant on its role in promoting the repair of chromatin buried abasic sites. We show that ALC1 enhances the ability of the abasic site processing enzyme, Apurinic/Apyrimidinic endonuclease 1 (APE1) to cleave nucleosome-occluded abasic sites. However, unrepaired abasic sites in ALC1-deficient cells are readily accessed by APE1 at the nucleosome-free replication forks. APE1 cleavage leads to fork breakage and trapping of PARP1/2 upon PARPi treatment, resulting in hypersensitivity. Collectively, our studies reveal how cells overcome the chromatin barrier to repair abasic lesions and uncover cleavage of abasic sites as a mechanism to overcome limitations of PARPi.

Long-term follow-up of outcomes including progression-free survival 2 in patients with transplant-ineligible multiple myeloma in the real-world practice: A multi-institutional report from the Canadian Myeloma Research Group (CMRG) database.

Multiple myeloma remains an incurable cancer mostly affecting older adults and is characterized by a series of remission inductions and relapses. This study aims to evaluate the outcomes in newly diagnosed transplant-ineligible patients using bortezomib/lenalidomide-based regimens in the Canadian real world as well as their outcomes in the second line. The Canadian Myeloma Research Group Database (CMRG-DB) is a national database with input from multiple Canadian Centres with now up to 8000 patients entered. A total of 1980 transplant ineligible patients were identified in the CMRG-DB between the years of 2007-2021. The four most commonly used induction regimens are bortezomib/melphalan/prednisone (VMP) (23%), cyclophosphamide/bortezomib/dexamethasone (CyBorD) (47%), lenalidomide/dexamethasone (Rd) (24%), and bortezomib/lenalidomide/dexamethasone (VRd) (6%). After a median follow-up of 30.46 months (0.89-168.42), the median progression-free survival (mPFS) and median overall survival (mOS) of each cohort are 23.5, 22.9, 34.0 months, and not reached (NR) and 64.1, 51.1, 61.5 months, and NR respectively. At the time of data cut-off, 1128 patients had gone on to second-line therapy. The mPFS2 based on first-line therapy, VMP, CyBorD, Rd, and VRd is 53.3, 48.4, 62.7 months, and NR respectively. The most common second-line regimens are Rd (47.4%), DRd (12.9%), CyBorD (10.3%), and RVd (8.9%) with a mPFS and a mOS of 17.0, 31.1, 15.4, and 14.0 months and 34.7, NR, 47.6, 33.4 months, respectively. This study represents the real-world outcomes in newly diagnosed transplant-ineligible myeloma patients in Canada. The spectra of therapy presented here reflect the regimens still widely used around the world. While this is sure to change with anti-CD38 monoclonal antibodies now reflecting a new standard of care in frontline therapy, this cohort is reflective of the type of multiple myeloma patient currently experiencing relapse in the real-world setting.

Outcomes with allogeneic stem cell transplant using cryopreserved versus fresh hematopoietic progenitor cell products.

BACKGROUND: Allogeneic hematopoietic stem cell transplant (alloHSCT) is a mainstay of treatment for hematologic malignancies such as acute leukemias and aggressive lymphomas. Historically, fresh hematopoietic progenitor cell (HPC) products have been preferred to cryopreserved products (cryo-HPC) due to concerns of loss of stem cell viability and number with the cryopreservation procedure. OBJECTIVE: We aimed to analyze the outcomes of patients who received cryo-HPCs during the COVID-19 pandemic and compare this against historical cohorts that received fresh HPC. STUDY DESIGN: A retrospective chart review was conducted on all adult patients who received a peripheral blood alloHSCT in British Columbia, Canada between June 2017 and November 2021. Baseline characteristics, Kaplan-Meier (KM) overall survival (OS), engraftment, and incidences of acute and chronic graft versus host disease were compared between patients who received cryo-HPCs and fresh HPCs. Univariable analysis followed by multivariable analysis was performed using a backward stepwise selection procedure to generate predictors of OS, cumulative incidence of relapse (CIR), nonrelapse mortality (NRM), and primary and secondary graft failure. RESULTS: Three hundred eighty-three patients were included in the analysis, with cryo-HPC representing 40%. Median viability was higher in the fresh-HPC group at 99.2% (IQR 98.3-99.5) versus cryo-HPCs at 97.0% (96.0, 98.6) (P < 0.01). The 12-month actuarial survivals were 77% in the fresh HPC and 75% in the cryo-HPC groups (P = 0.21). There were no differences between cryo-HPCs and fresh HPCs on univariable analysis of OS, CIR, or NRM. There was a shorter median time to platelet engraftment in patients receiving fresh HPC at 17 days (IQR 16, 20) versus cryo-HPC at 21 days (IQR 18, 29), P < 0.001. There was a shorter median time to neutrophil engraftment in the fresh HPC group at 17 days (IQR 14, 20) versus 20 days (17, 23), P < 0.001. Cryo-HPC accounted for 5 out of 6 cases of primary graft failure (P = 0.04), and 3 out of five cases of secondary graft failure (P = 0.39). There were no significant differences in acute GVHD between the fresh HPC and cryo-HPC groups (P = 0.34). The incidence of moderate or severe chronic GVHD was 32% in the fresh-HPC group and 17% in the cryo-HPC group (P < 0.001). In multivariable analysis, cryopreservation did not emerge as an independent predictor of OS, CIR, NRM, primary GF or secondary GF. However, viability <90% on arrival at our center was a significant predictor of OS (HR 5.3, 2.3-12.3, P < 0.01), primary graft failure (OR 36.3, 5.4-210.2, P < 0.01), and secondary graft failure (OR 18.4, 1.7-121.1, P < 0.01). CONCLUSIONS: Patients who received cryo-HPCs had similar OS and relapse rates to those who received fresh-HPCs but typically took 2-3 days longer to achieve engraftment of platelets or neutrophils and were associated increased primary graft failure. However, after accounting for multiple variables, cryopreservation was no longer a significant predictor of survival or engraftment while viability <90% emerged as an important predictor of OS, primary graft failure, and secondary graft failure. If confirmed, this suggests that viability on arrival at the infusion center may be a good quality control indicator used to identify HPC products that may warrant recollection if the risk of graft failure is sufficiently increased.

Milestoning estimators of dissipation in systems observed at a coarse resolution.

Many nonequilibrium, active processes are observed at a coarse-grained level, where different microscopic configurations are projected onto the same observable state. Such "lumped" observables display memory, and in many cases, the irreversible character of the underlying microscopic dynamics becomes blurred, e.g., when the projection hides dissipative cycles. As a result, the observations appear less irreversible, and it is very challenging to infer the degree of broken time-reversal symmetry. Here we show, contrary to intuition, that by ignoring parts of the already coarse-grained state space we may-via a process called milestoning-improve entropy-production estimates. We present diverse examples where milestoning systematically renders observations "closer to underlying microscopic dynamics" and thereby improves thermodynamic inference from lumped data assuming a given range of memory, and we hypothesize that this effect is quite general. Moreover, whereas the correct general physical definition of time reversal in the presence of memory remains unknown, we here show by means of physically relevant examples that at least for semi-Markov processes of first and second order, waiting-time contributions arising from adopting a naive Markovian definition of time reversal generally must be discarded.

Selective refueling of CAR T cells using ADA1 and CD26 boosts antitumor immunity.

Chimeric antigen receptor (CAR) T cell therapy is hindered in solid tumor treatment due to the immunosuppressive tumor microenvironment and suboptimal T cell persistence. Current strategies do not address nutrient competition in the microenvironment. Hence, we present a metabolic refueling approach using inosine as an alternative fuel. CAR T cells were engineered to express membrane-bound CD26 and cytoplasmic adenosine deaminase 1 (ADA1), converting adenosine to inosine. Autocrine secretion of ADA1 upon CD3/CD26 stimulation activates CAR T cells, improving migration and resistance to transforming growth factor ß1 suppression. Fusion of ADA1 with anti-CD3 scFv further boosts inosine production and minimizes tumor cell feeding. In mouse models of hepatocellular carcinoma and non-small cell lung cancer, metabolically refueled CAR T cells exhibit superior tumor reduction compared to unmodified CAR T cells. Overall, our study highlights the potential of selective inosine refueling to enhance CAR T therapy efficacy against solid tumors.

Arsenate Anion-π Interactions on Amine-Modified Polydivinylbenzene in Aqueous Systems: Experimental and Theoretical Investigation.

Anion-π interactions aiding in the adsorption of anions in the solution phase, though challenging to quantify, have attracted a lot of attention in supramolecular chemistry. We present the design of a polymer adsorbent that quantifies the adsorption of arsenate ions experimentally by optimizing anion-π interactions in a purely aqueous system and use density functional theory to compare these results with theoretical data. Arsenate anions are removed from water by amine-functionalized polydivinylbenzene using the comonomer 1-vinyl-1,2,4-triazole, which was cross-linked with divinylbenzene via radical polymerization in a hydrothermal procedure. The amine-functionalized polydivinylbenzene successfully removed arsenate anions from water with a capacity of 46 mg g-1, a 70% increase compared to the nonfunctionalized polydivinylbenzene (27 mg g-1) capacity under the same conditions. Adsorption is best described by the Sips isotherm model with a correlation coefficient R2 factor of 0.99, indicating that adsorption sites are homogeneous, and adsorption occurred by forming a monolayer. Kinetic studies indicated that adsorption is second order in the amine-functionalized polydivinylbenzene. Computational studies using density functional theory showed that the 1-vinyl-1,2,4-triazole comonomer improved the thermodynamic stability of the anionic-π interactions of polydivinylbenzene with arsenate anions. Electrostatic interactions dominate the mechanism of adsorption in polydivinylbenzene compared to the anion-induced interactions that dominate adsorption in amine-functionalized polydivinylbenzene.

The influence of immunodeficiency, disease features, and patient characteristics on survival in plasmablastic lymphoma.

ABSTRACT: Plasmablastic lymphoma (PBL) is a rare and aggressive non-Hodgkin lymphoma associated with immunodeficiency, characterized by uncertain treatment approaches and an unfavorable prognosis. We conducted a multicenter, international, retrospective cohort study, aiming to characterize the clinical features, risk factors, and outcomes of patients with PBL. Data were collected from 22 institutions across 4 countries regarding patients diagnosed with PBL between 1 January 1999 and 31 December 2020. Survival risk factors were analyzed using both univariate and multivariate regression models. Overall survival (OS) was calculated using Kaplan-Meier statistics. First-line treatment regimens were stratified into standard- and higher-intensity regimens, and based on whether they incorporated a proteasome inhibitor (PI). A total of 281 patients (median age, 55 years) were included. Immunodeficiency of any kind was identified in 144 patients (51%), and 99 patients (35%) had HIV-positive results. The 5-year OS for the entire cohort was 36% (95% confidence interval, 30%-42%). In multivariate analysis, inferior OS was associated with Epstein-Barr virus-negative lymphoma, poor performance status, advanced stage, and bone marrow involvement. In an independent univariate analysis, the international prognostic index was associated with OS outcomes. Neither immunosuppression nor HIV infection, specifically, influenced OS. Among patients treated with curative intent (n = 234), the overall response rate was 72%. Neither the intensity of the treatment regimen nor the inclusion of PIs in first-line therapy was associated with OS. In this large retrospective study of patients with PBL, we identified novel risk factors for survival. PBL remains a challenging disease with poor long-term outcomes.

Utilizing Deep Learning to Identify an Ultrasound-guided Nerve Block Target Zone.

Ultrasound guided nerve blocks are increasingly being used in perioperative care as a means of safely delivering analgesia. Unfortunately, identifying nerves in ultrasound images presents a challenging task for novice anesthesiologists. Drawing from online resources, here we attempted to address this issue by developing a deep learning algorithm capable of automatically identifying the transversus abdominis plane region in ultrasound images. Training of our dataset was done using the U-Net architecture and artificial augmentation was done to optimize our training dataset. The Dice score coefficient was used to evaluate our model, with further evaluation against a test set composed of manually drawn labels from a pool of (n=10) expert anesthesiologists.Across all labelers the model achieved a global Dice score of 73.31% over the entire test set. These preliminary results highlight the potential effectiveness of this model as a future ultrasound decision support system in the field of anesthesia.

Real-world results of autologous stem cell transplantation in newly diagnosed multiple myeloma: a report from the Canadian Myeloma Research Group database.

Autologous stem cell transplant (ASCT) remains an important option for eligible multiple myeloma (MM) patients as part of initial therapy. Using the Canadian Myeloma Research Group (CMRG) national database, we examined the details and outcomes of ASCT performed as first-line therapy in eligible Canadian MM patients between 2007 to 2021. We included 3821 patients with 72% receiving CyBorD induction and 2061 patients receiving maintenance, consisting of lenalidomide +/- steroids in 78.3%. The median PFS and OS for patients given a single ASCT were 35.4 and 126 months. Those receiving a second induction regimen had significantly inferior outcomes, although when maintenance was used, results were comparable regardless of the number of induction regimens administered (median PFS 55.3 vs 51.1 months [p = 0.11]; median OS 158.6 vs not yet reached [p = 0.13]). Consolidation patients had a longer median PFS (55.3 vs 34.4 months [p = 0.001]), but no significant gain in median OS (p = 0.065). Patients who received lenalidomide-based maintenance experienced a median PFS of 53.7 months and OS of 159 months. In the multivariable analysis, use of any type of maintenance therapy vs no maintenance was associated with a lower risk of progression (HR 0.52 (95% CI 0.47-0.57)) and death (HR 0.58 (95% CI 0.51-0.67)). This real-world study demonstrates that, overall, first-line treatment sequence in transplant-eligible patients produces a median OS of ≥10 years. It also highlights the contribution of post-ASCT maintenance, particularly lenalidomide given until progression.

Outcome of carfilzomib/pomalidomide-based regimens after daratumumab-based treatment in relapsed multiple myeloma: A Canadian Myeloma Research Group Database analysis.

INTRODUCTION: Although daratumumab-containing regimens improve multiple myeloma (MM) outcomes, recurrence is inevitable. METHODS AND OBJECTIVE: We performed a retrospective study using the Canadian Myeloma Research Group Database to benchmark the efficacy of carfilzomib- or pomalidomide-based therapies immediately following progression on daratumumab treatment. RESULTS: We identified 178 such patients; median number of prior lines of therapy was 3, 97% triple-class exposed, and 60% triple-class refractory. In our cohort, 75 received a subsequent carfilzomib-based therapy, 79 received a pomalidomide-based therapy, and 24 received a treatment with both immunomodulatory drug (IMiD) and proteasome inhibitor (PI) using carfilzomib and/or pomalidomide. The median progression-free survival (PFS) and overall survival (OS) for the entire cohort were 4.5 and 14.2 months, respectively. Carfilzomib-based therapy yielded a median PFS and OS of 4.5 and 10.2 months, respectively, compared to 5.2 and 21.7 months for pomalidomide-based therapy. Patients who received both IMiD and PI with carfilzomib and/or pomalidomide had a median PFS and OS of 4.1 and 14.5 months, respectively. CONCLUSION: Our observations demonstrate the poor outcome of MM patients when standard regimens based on carfilzomib and/or pomalidomide are utilized directly after daratumumab-based therapy given in the relapsed setting. Novel therapies, including immune therapies, are urgently needed to improve the outcomes of these daratumumab-exposed patients.

Non-Markov models of single-molecule dynamics from information-theoretical analysis of trajectories.

Whether single-molecule trajectories, observed experimentally or in molecular simulations, can be described using simple models such as biased diffusion is a subject of considerable debate. Memory effects and anomalous diffusion have been reported in a number of studies, but directly inferring such effects from trajectories, especially given limited temporal and/or spatial resolution, has been a challenge. Recently, we proposed that this can be achieved with information-theoretical analysis of trajectories, which is based on the general observation that non-Markov effects make trajectories more predictable and, thus, more "compressible" by lossless compression algorithms. Toy models where discrete molecular states evolve in time were shown to be amenable to such analysis, but its application to continuous trajectories presents a challenge: the trajectories need to be digitized first, and digitization itself introduces non-Markov effects that depend on the specifics of how trajectories are sampled. Here we develop a milestoning-based method for information-theoretical analysis of continuous trajectories and show its utility in application to Markov and non-Markov models and to trajectories obtained from molecular simulations.

Redefining attrition in multiple myeloma (MM): a Canadian Myeloma Research Group (CMRG) analysis.

While most patients diagnosed with multiple myeloma (MM) receive initial therapy, reported attrition rates are high. Understanding attrition rates and characteristics of patients not receiving subsequent therapy is useful for MM stakeholders. We performed an analysis of attrition rates in a large disease-specific database of patients with newly diagnosed MM who received at least one line of therapy between Jan 1/10-Dec 31/20. Attrition was defined as failure to receive a subsequent line of therapy despite progression of MM or due to death. A total of 5548 patients were identified, 3111 autologous stem cell transplant (ASCT) patients and 2437 non-ASCT. In the ASCT cohort, the attrition rate was 7% after line 1, 12% after line 2, and 23% after line 3. In non-ASCT patients, the attrition rate was 19% after line 1, 26% after line 2, and 40% after line 3. Death was the dominant contributor to attrition across all cohorts, with a minority of patients alive with progressive disease in the absence of further therapy at each line. Multivariable analysis identified older age, shorter time to progression, and inferior response as independent risk factors for attrition. Our data show that attrition rates increase with each line of therapy and are higher in non-ASCT patients but are appreciably lower than previously reported. This study supports a revision of the previous definition of attrition, demonstrating that most patients who do not receive subsequent therapy are either continuing their current therapy and/or are in remission off-treatment rather than being irreversibly lost to attrition.

How should multiple myeloma research change in a patient-oriented world? Findings and lessons from the pan-Canadian myeloma priority setting partnership.

BACKGROUND: Over the last decade there has been considerable research into the treatment, management, and quality of life of people living with multiple myeloma. However, there has been limited investigation into topics deemed important to patients and caregivers within this community. We conducted a James Lind Alliance Priority Setting Partnership to establish the 'Top 10 Priorities for Myeloma Research', informed by patient and public partners. METHODS: A research team and steering group were established in 2019 to conduct the myeloma priority setting partnership. Steering group members included patients, caregivers, and healthcare providers who advised the research team and oversaw the scope of the project, grounded on their lived experience. Following the James Lind Alliance guidelines for identification and ranking of research questions, we used surveys and a virtual workshop to collect and prioritize questions posed by myeloma patients, caregivers, and healthcare providers across Canada. RESULTS: The Top 10 list of priorities for myeloma research was finalized at the consensus-building workshop and encompassed questions related to diagnosis, treatment, management, and living well with myeloma. A final participant evaluation survey elicited a positive response. INTERPRETATION: The myeloma priority setting partnership identified the research priorities of people living with myeloma, caregivers, and healthcare providers to inform clinical research on this disease going forward. This project underscores the importance of patient and public engagement in the identification of research questions, highlighting the concerns of people affected by myeloma to ultimately improve the lives of people living with this disease.

Research on multiple myeloma, a rare blood cancer, rarely focuses on topics that are personally important to people living with myeloma, their caregivers, and their healthcare providers. The purpose of this study was to complete a priority setting partnership, following guidelines from the James Lind Alliance, to identify unanswered research questions from people directly affected by myeloma. The project was guided by a steering group of people living with myeloma, their caregivers, and healthcare providers who informed all stages of this process including two questionnaires to collect and rank questions and a priority setting workshop. The workshop brought together representatives from each group to form the final 'Top 10 Priorities for Myeloma Research', which included questions related to diagnosis, treatment, management, and living well with myeloma. This process allowed us to identify the research priorities of the myeloma community and highlighted the importance of including patients and the public as team members in the research process. We encourage other myeloma researchers to do the same to ensure research is meaningful and relevant to the communities who rely on it.