Individual Patient-level Data Research Articles

Predicting chronic responses to calcium channel blockade with a virtual population of African Americans with hypertensive chronic kidney disease.

Chronic kidney disease (CKD) is associated with the progressive loss of functional nephrons and hypertension (HTN). Clinical studies demonstrate calcium channel blocker (CCB) therapy mitigates the decline in renal function in humans with essential HTN. However, there are few long-term clinical studies that determine the impact of CCBs in patients with hypertensive CKD. African Americans (AA) have a higher prevalence of CKD and a faster progression to total kidney failure as compared to the white population but the mechanisms are poorly understood. Both clinical evidence (the African American Study of Kidney Disease and Hypertension, or AASK trial) and experimental studies have demonstrated that CCB may expose glomerular capillaries to high systemic pressures and exacerbate CKD progression. Therefore, using a large physiological model, we set out to replicate the AASK trial findings, predict renal hemodynamic responses and the role of the renin-angiotensin system during CCB antihypertensive therapy in a virtual population, and hypothesize mechanisms underlying those findings. Our current mathematical model, HumMod, is comprised of integrated systems that play an integral role in long-term blood pressure (BP) control such as neural, endocrine, circulatory, and renal systems. Parameters (n=341) that control these systems were randomly varied and resulted in 1400 unique models that we define as a virtual population. We calibrated these models to individual patient level data from the AASK trial: BP and glomerular filtration rate (GFR) before and after 3 years of amlodipine (10 mg/day). After calibration, the new virtual population (n=165) was associated with statistically similar BP and GFR before and after CCB. Baseline factors such as elevated single nephron GFR and low tubuloglomerular feedback were correlated with greater declines in renal function and increased glomerulosclerosis after 3 years of CCB. Blocking the renin-angiotensin system (RAS) in the virtual population decreased glomerular pressure, limited glomerular damage, and further decreased BP (-14 ± 8 mmHg) as compared to CCB alone (-11 ± 9 mmHg). Our simulations echo the potential risk of CCB monotherapy in AA CKD patients and support blockade of the renin angiotensin system as a valuable tool in renal disease treatment when combined with CCB therapy.

Data sharing and reuse in clinical research: Are we there yet? A cross-sectional study on progress, challenges and opportunities in LMICs.

Data sharing holds promise to accelerate innovative discoveries through artificial intelligence (AI) and traditional analytics. However, it remains unclear whether these prospects translate into tangible benefits in improving health care and scientific progress. In this cross-sectional study, we investigate current data reuse practices and explore ways to enhance the use of existing data in clinical research, focusing on low- and middle-income countries. 643 clinical researchers and data professionals participated in the study. 55.5% analysed clinical trial data. 75.3% of data users analysed data from observational studies obtained mainly through personal requests or downloads from publicly available sources. Data was mainly used to influence the design of new studies or in pooled and individual patient-level data meta-analyses. Key benefits realised were career progression and academic qualification, with more gains reported by users affiliated with high-income and upper-middle-income countries (p = 0.046, chi = 8.0). Scientific progress through publications and collaborations was associated with gender (p = 0.012, chi = 10.9), with males more likely to contribute. Benefits to the public although minimal, were associated with career seniority (p = 0.001, chi = 18.8), with works by senior researchers being more likely to influence health policy or treatment guidelines. Although 54% of the respondents accessed at least 3 datasets in the past 5 years, 79.4% of data users encountered difficulty finding relevant data for planned analyses. Researchers affiliated with low and middle income institutions reported more difficulty interpreting data (p = 0.012, chi = 25.7), while challenges with language were regionally influenced (p = 0.000, chi = 51.3) and more commonly reported by researchers in Latin America and South and East Asia institutions. While the utilisation of shared data is lower than expected, focused efforts to enrich existing data with extensive metadata using standard terminologies can enhance data findability. Investment in training programmes, building professional networks, and mentorship in data science may improve the quality of data generated and increase researchers' ability to use existing datasets.

Evaluating dynamic patterns in mortality before and after reconfiguration of the Danish emergency healthcare system.

This study explored changes in short-term mortality during a national reconfiguration of emergency care starting in 2007. Unplanned hospital contacts at emergency departments across Denmark from 2007 to 2016. The reconfiguration was a natural experiment, resulting in individual timelines for each hospital. The outcome was in-hospital and 30-day mortality. Individual patient-level data included 9,745,603 unplanned hospital contacts from 2007 to 2016 at 20 hospitals with emergency departments. We observed a sharp downwards shift in in-hospital mortality and 30-day mortality in three hospitals in relation to the reconfiguration. This nationwide study identified three hospitals where the reconfiguration was closely associated with reduced in-hospital and 30-day mortality. In contrast, no major effects were identified for the remaining hospitals.

The Comparative Effects of Monotherapy with Topical Minoxidil, Oral Finasteride, and Topical Finasteride in Postmenopausal Women with Pattern Hair Loss: A Retrospective Cohort Study

Introduction: Oral finasteride and topical minoxidil are long-standing androgenetic alopecia (AGA) treatments; topical finasteride is a more recent medicine. Few studies have compared their therapeutic effects in postmenopausal women. We compared the therapeutic impact of topical finasteride (1–4 sprays of 0.25% topical finasteride solution daily for 12 months), oral finasteride (2.5 mg oral finasteride once daily for 12 months), and topical minoxidil (1 mL of topical minoxidil 5% twice daily for 12 months) in postmenopausal women with AGA. Methods: We conducted Bayesian network meta-analyses of individual patient-level data insofar as four clinically relevant endpoints, namely, 12-month change in (1) total hair density, (2) hair diameter, (3) clinical photographs, and (4) patients’ opinion of efficacy. Data were obtained through medical charts. Regimens’ surface under the cumulative ranking distribution (SUCRA) values and relative effects – as per odds ratios – were computed. Results: As per SUCRA, the most and least effective regimens – across the four outcomes – were oral finasteride, and topical finasteride, respectively; however, no significant statistical differences were found (i.e., p > 0.05). Conclusion: Oral finasteride is ranked more effective than the topical forms of minoxidil and finasteride; however, more studies are needed to confirm this result.

Maintenance of zilucoplan efficacy in patients with generalised myasthenia gravis up to 24 weeks: a model-informed analysis.

Clinical efficacy of zilucoplan has been demonstrated in a 12-week, placebo-controlled, phase III study in patients with acetylcholine receptor autoantibody-positive generalised myasthenia gravis (gMG). However, placebo-controlled zilucoplan data past 12 weeks are not available. Predict the treatment effect of zilucoplan versus control (placebo or standard of care) in patients with gMG up to 24 weeks. A model-informed analysis (MIA) within a Bayesian framework. Part 1 of the MIA comprised a control meta-regression using aggregate data on control response over time from randomised studies and a national myasthenia gravis (MG) registry. In Part 2, a combined Bayesian analysis of individual patient-level data from the phase II (NCT03315130), RAISE (NCT04115293) and RAISE-XT (NCT04225871) studies of zilucoplan was conducted using posterior distributions from Part 1 as informative priors. Population mean treatment effect in the change from baseline (CFB) at week 24 in MG-Activities of Daily Living (MG-ADL) and quantitative MG (QMG) scores for zilucoplan versus control were assessed. At week 24, the predicted mean CFB in MG-ADL score was -4.55 (95% credible interval: -6.04, -3.13) with zilucoplan versus -2.00 (-3.35, -0.64) with control (difference: -2.55 [-3.76, -1.40]). The probability of a favourable treatment effect as measured by MG-ADL score at week 24 with zilucoplan versus control was >99.9%. There was an 82.8% probability that the difference in the predicted mean CFB in MG-ADL score at week 24 was greater than the clinically meaningful threshold (⩾2.0-point improvement). Comparable results were observed with QMG. This MIA demonstrates the maintenance of efficacy with zilucoplan versus control up to 24 weeks. Through combining real-world evidence with data from randomised studies, this novel method to estimate long-term treatment efficacy facilitated reduced exposure to placebo in the phase III RAISE study. This methodology could be used to reduce the length of future placebo-controlled studies.

Learning from vertically distributed data across multiple sites: An efficient privacy-preserving algorithm for Cox proportional hazards model with variable selection

ObjectiveTo develop a lossless distributed algorithm for regularized Cox proportional hazards model with variable selection to support federated learning for vertically distributed data. MethodsWe propose a novel distributed algorithm for fitting regularized Cox proportional hazards model when data sharing among different data providers is restricted. Based on cyclical coordinate descent, the proposed algorithm computes intermediary statistics by each site and then exchanges them to update the model parameters in other sites without accessing individual patient-level data. We evaluate the performance of the proposed algorithm with (1) a simulation study and (2) a real-world data analysis predicting the risk of Alzheimer’s dementia from the Religious Orders Study and Rush Memory and Aging Project (ROSMAP). Moreover, we compared the performance of our method with existing privacy-preserving models. ResultsOur algorithm achieves privacy-preserving variable selection for time-to-event data in the vertically distributed setting, without degradation of accuracy compared with a centralized approach. Simulation demonstrates that our algorithm is highly efficient in analyzing high-dimensional datasets. Real-world data analysis reveals that our distributed Cox model yields higher accuracy in predicting the risk of Alzheimer’s dementia than the conventional Cox model built by each data provider without data sharing. Moreover, our algorithm is computationally more efficient compared with existing privacy-preserving Cox models with or without regularization term. ConclusionThe proposed algorithm is lossless, privacy-preserving and highly efficient to fit regularized Cox model for vertically distributed data. It provides a suitable and convenient approach for modeling time-to-event data in a distributed manner.

Generalizability of clinical trial efficacy results to a real-world population: An example in migraine prevention.

Health care decision makers are often concerned about the external validity of randomized controlled trials (RCTs), as their results may not apply to certain patients in the real world who intend to receive treatment. To demonstrate a methodology for assessing the generalizability of clinical trial results to a real-world population, before sufficient and appropriate real-world effectiveness data are available, using individual patient-level data from an RCT and aggregated baseline data from a real-world French registry in migraine. The analyses were conducted in 2 steps. First, individual patient-level baseline data from the multinational CONQUER RCT were weighted to match aggregated real-world InovPain registry patient characteristic data. Matched patient characteristics were sex, age, migraine type and duration, number of monthly migraine headache days, and number of monthly headache days at baseline. Second, the weighted CONQUER patient data were used to reanalyze the primary endpoint of CONQUER (least squares mean change from baseline in the number of monthly migraine headache days during the 3-month double-blind treatment phase) using predefined methodology. Sensitivity analyses were conducted to assess the robustness of findings. A total of 462 patients with migraine were randomized and treated with galcanezumab or placebo in CONQUER; aggregated InovPain data were available from 130 patients with migraine. We identified no important differences in baseline patient characteristics between the 2 prespecified populations, suggesting good external validity for CONQUER. Although this limited the extent of observed differences between the original and matched CONQUER populations, weighting of CONQUER data did help harmonize the 2 datasets and allow the results obtained in CONQUER to be generalized to patients more representative of the real-world French population with migraine. Results of weighted analyses suggested that galcanezumab would be superior to placebo for reducing monthly migraine headache days in a clinical trial in patients with episodic or chronic migraine who reflected the characteristics of patients eligible to receive the drug in France. Findings suggest that our methods may be helpful for assessing the generalizability of clinical trial results to a real-world population before the availability of substantial real-world clinical data.

Prognostic significance of lymph nodes assessment during pulmonary metastasectomy: a systematic review and meta-analysis.

Lung metastasectomy is an accepted treatment modality worldwide. Whether the addition of lymph node dissection to the procedure is useful remains, however, unknown. We performed a systematic review of the literature analyzing MEDLINE, Embase, until 31st October 2021. We included all studies which met the inclusion criteria aiming to determine if the addition of lymph node tissue dissection/sampling to lung metastasectomy offers survival benefits when compared to patients who do receive lymph node tissue dissection. Secondary outcomes were 3- and 5-year overall survival (OS) and disease-free survival (DFS). Each study was assessed for risk of bias. The data collected from the included studies were pooled using reconstruction of individual-level patient data and pooling of reported 5-year odds ratios (ORs). Interstudy heterogeneity was estimated with visual inspection of forest plots and calculation of the I2 inconsistency statistic. We found 11 eligible studies that included a total of 3,310 patients. The most common primary tumor type was colorectal cancer (1,740 patients) and the most commonly performed operative procedure was wedge resection (57%) followed by lobectomy (39%). When resection status was reported, R0 resection was achieved in 90% of the cases. Nine studies did not show a statistically significant effect of lymph nodes dissection or sampling on the 5-year OS with a pooled hazard ratio (HR) of 0.94 [95% confidence interval (CI): 0.82, 1.08; I2=26%; 95% prediction interval (PI): 0.70, 1.24]. Regarding DFS, the pooled HR 0.60 (95% CI: 0.44, 0.80; I2=31%; 95% PI: 0.12, 2.09). The addition of lymph node tissue dissection during lung metastasectomy was not associated with a significant benefit in OS and showed a slight tendency towards a better DFS.

Comparative Efficacy of Ciltacabtagene Autoleucel Versus Elotuzumab, Isatuximab, and Selinexor-Based Regimens in the Treatment of Patients with Relapsed or Refractory Multiple Myeloma with 1-3 Prior Lines of Therapy Using a Matching-Adjusted Indirect Comparison

Introduction: Ciltacabtagene autoleucel (cilta-cel; Carvykti) was assessed in the phase 3 randomized controlled trial CARTITUDE-4 for adult patients with relapsed and refractory multiple myeloma (RRMM) who have received 1-3 prior line(s) of therapy (LOT) that included an immunomodulatory agent (IMiD), and a proteasome inhibitor (PI) and are refractory to lenalidomide. Other therapies used to treat patients with multiple myeloma who have received at least one other therapy and are lenalidomide-refractory vary by country, but broadly include doublet and triplet regimens based on daratumumab, pomalidomide, bortezomib, and carfilzomib. This analysis aims to compare efficacy outcomes in patients randomized to the cilta-cel arm of CARTITUDE-4 versus relevant comparators based on their target trial populations. Methods: Comparators for the matching adjusted indirect comparisons (MAICs) were identified a priori based on regimens used in the target population including; SVd, EloPd, IsaKd, DKd, DVd, Kd, Vd, Pd, and Ide-cel. PVd and DPd were not of interest given their inclusion in the CARTITUDE-4 trial. Melflufen+d, PanVd, VenVd, PCd, IxaKd, D-monotherapy, V-monotherapy, CyVd, and DVCd were not of interest for the US, EU-5, the Netherlands, and Canada. A systematic literature review identified publications of relevant clinical trials for these treatments, which were assessed for feasibility of an indirect treatment comparison. Feasibility was determined based on study design and degree of overlap with the CARTITUDE-4 population (prior LOT, lenalidomide-refractoriness, and exclusion criteria around previous therapies, i.e., prior daratumumab). We report here on comparisons vs. EloPd (ELOQUENT-3), IsaPd (ICARIA-MM), and SVd (BOSTON). Due to differences in methods for the comparisons versus ide-cel (KarMMa-3) and versus DKd, Kd, DVd, Vd, and Pd, these are reported separately. Comparison to IsaKd was not feasible due to a lack of published data in the lenalidomide-refractory subgroup (baseline characteristics and efficacy data), and key differences in exclusion criteria. Given the lack of a common comparator between CARTITUDE-4 and the comparator trials, unanchored MAICs were performed utilizing individual patient-level data (IPD) from patients randomized to the cilta-cel arm of CARTITUDE-4 (N=208). The eligibility criteria from each of the comparator trials were applied to the cilta-cel IPD, and outcomes for the matched population were compared against published summary data for EloPd (N=60), IsaPd (N=154), and SVd (N=195) using reconstructed IPD for response endpoints and simulated IPD from published Kaplan-Meier (KM) curves for progression-free survival (PFS). Further imbalances in patient characteristics were adjusted by weighting the cilta-cel data to match the reported baseline characteristics of the comparator trials in terms of refractory status, cytogenetic risk, ISS stage, and presence of extramedullary disease. Comparative efficacy was estimated for overall response rate (ORR), very good partial response or better (≥VGPR) rate, complete response or better (≥CR) rate, and PFS. For binary endpoints, relative effects were quantified using relative response ratios (RRs) with 95% confidence intervals (CIs) derived from a weighted logistic regression analysis. For PFS, HRs including 95% CIs were estimated using a weighted Cox proportional hazards model. Results: After adjustment, patients in the cilta-cel arm were significantly more likely to achieve an overall response than patients in EloPd or IsaPd arms, and to achieve ≥VGPR and ≥CR versus patients in all other comparators arms ( Table 1). In addition, cilta-cel was associated with a significant reduction in the risk of disease progression or death (PFS) versus all comparators, ranging from 39% (versus SVd) to 68% (versus IsaPd) ( Table 2). Conclusions: Cilta-cel demonstrated clinical benefit over EloPd, IsaPd, and SVd for response outcomes and PFS, highlighting its potential as an effective treatment option for patients with RRMM who have received at least one other therapy and are lenalidomide-refractory. These comparisons provide valuable information to contextualize the efficacy of cilta-cel in countries where treatment for these patients may be different from DPd or PVd.

Matching-Adjusted Indirect Comparison of Personalized Prophylaxis with Simoctocog Alfa Versus Standard Prophylaxis with Efanesoctocog Alfa in Adults with Severe Hemophilia a

Background: Personalized factor VIII (FVIII) prophylaxis is currently the gold standard approach for the treatment of patients with severe hemophilia A. Although several FVIII concentrates are available, head-to-head comparative studies of their relative efficacy are not feasible in rare diseases such as hemophilia A. In such situation, the use of validated indirect comparison methods such as matching-adjusted indirect comparison (MAIC) methodology can be used to indirectly compare outcomes from separate clinical studies. MAIC can be used to adjust for differences in patient characteristics between studies, with the goal to reduce bias in the treatment effect estimates which can occur with comparison of efficacy data across clinical trials. Aims: To indirectly compare the efficacy of a pharmacokinetic (PK)-guided personalized prophylaxis regimen with simoctocog alfa (Nuwiq ®; a 4 th-generation recombinant FVIII [rFVIII] concentrate with no chemical modifications of fusion proteins) against a standard prophylaxis regimen with efanesoctocog alfa (Altuviiio ®; an Fc-fusion rFVIII concentrate with chemical modifications including Von Willebrand factor independent-XTEN fusion protein), in previously-treated patients (PTPs) with severe hemophilia A. Methods: After matching study populations by accounting for clinically important baseline characteristics,individual patient-level data (IPD) from 65 patients treated with simoctocog alfa from the NuPreviq study were compared against aggregate data from the once-weekly efanesoctocog alfa arm of the XTEND-1 study (n=133). All patients enrolled in both studies were male, with the exception of one female in XTEND-1. Baseline age and body weight were used to re-weight the simoctocog alfa IPD to match the aggregate data reported for efanesoctocog alfa from the XTEND-1 study. Unanchored indirect treatment comparisons were performed for various clinical outcomes using the re-weighted data for simoctocog alfa and the published comparator study data. The endpoints analysed were annualized bleeding rates (ABRs), percentage of patients with zero bleeds, and weekly consumption of FVIII. All efficacy endpoints reported here consider treated and untreated bleeds. Results: After matching populations, the effective sample size for simoctocog alfa was 35.3. Although, the percentage of patients with zero bleeds was higher with simoctocog alfa versus efanesoctocog alfa (64.1% vs 55.5%), this difference was not statistically significant (Table 1). The mean total ABRs were similar between the two products: 1.5 for simoctocog alfa and 1.1 for efanesoctocog alfa (Table 1). The mean weekly dose was significantly higher in patients treated with simoctocog alfa versus efanesoctocog alfa (98.3 IU/kg vs 52.2 IU/kg, respectively; p<0.0001). Conclusion: An indirect comparison analysis demonstrated that a PK-guided, personalized prophylaxis with simoctocog alfa in adult PTPs resulted in zero bleed rates and ABRs that are not statistically different from those with efanesoctocog alfa. The higher total weekly dose observed with simoctocog alfa is to be expected based on the dosing recommendations for each respective FVIII replacement product. This MAIC analysis provides important comparative efficacy and utilization data, which can help guide patients and physicians in making decisions regarding product choice for prophylaxis regimens.

Updated Results from a Matching-Adjusted Indirect Comparison of Efficacy Outcomes for Isatuximab Plus Pomalidomide and Dexamethasone Versus Daratumumab Plus Pomalidomide and Dexamethasone in Patients with Relapsed and/or Refractory Multiple Myeloma

Introduction: Findings from a previously conducted matching-adjusted indirect comparison (MAIC) between isatuximab plus pomalidomide and dexamethasone (IsaPd) and daratumumab plus pomalidomide and dexamethasone (DaraPd) among relapsed and/or refractory multiple myeloma (RRMM) patients receiving at least two prior lines of therapy (LoT) including lenalidomide and a proteasome inhibitor (PI) demonstrated a significantly better overall survival (OS) and a positive trend for progression-free survival (PFS), overall response rate (ORR), and very good partial response or better (≥VGPR) rates with IsaPd than that with DaraPd, with a similar safety profile for both treatments [Leleu X et al, Value in Health 2022;25(1):S39]. The present study used longer follow-up data from the phase 3 ICARIA trial (NCT02990338) to update the previous MAIC, evaluating the comparative efficacy of IsaPd versus DaraPd. Methods: A MAIC was performed with the latest available individual patient-level data (IPD) from the ICARIA trial (IsaPd arm, n = 154; Pd arm, n = 153), data cutoff of March 14, 2022, and aggregate data from study publications of the phase 3 APOLLO trial (DaraPd arm, n = 151; Pd arm, n = 153, NCT03180736) [Dimopoulos MA, et al. Blood 2022;140 (Suppl 1):7272-7274; Dimopoulos MA, et al. Lancet Oncol. 2021;22(6):801-812; Dimopoulos MA, et al. Blood 2020;136(Suppl 1):5-6]. To match patients with APOLLO, patients in ICARIA who did not meet the inclusion criteria of APOLLO were excluded. In APOLLO, 11% patients received only one prior LoT. As patients with one prior LoT were not enrolled in ICARIA, it was not feasible to match on this factor; however, the treatment effect of DaraPd versus Pd on PFS among these patients was similar to patients in APOLLO with ≥2 prior LoTs. Therefore, this difference in study populations is unlikely to bias the MAIC. Remaining patients in ICARIA were then weighted using standardized MAIC weights (mean value = 1.0) generated using the method of moments, with active treatment and control arms matched separately to ensure any residual imbalances between active and control arms in APOLLO were replicated in the matched ICARIA sample. Anchored MAICs for the hazard ratio (HR) for OS and PFS, odds ratio (OR) for ORR, and ≥VGPR for IsaPd versus DaraPd were calculated using the Bucher method [Bucher HC, et al. J Clin Epidemiol. 1997;50(6):683-691]. Results: MAIC weighting was able to match all characteristics included in the matching algorithm, with standardized mean difference (SMDs) <0.0001 after weighting. The effective sample size after weighting was 28.1 for IsaPd and 43.1 for Pd. Based on an anchored MAIC, OS was significantly better with IsaPd versus DaraPd (HR [95% CI]: 0.49 [0.24-0.98]; p = 0.04) (Figure 1). A numerical trend in favor of IsaPd versus DaraPd was observed for PFS, ORR, and ≥VGPR, although the differences were statistically non-significant (Table 1). Conclusions: This MAIC analysis demonstrated a significant benefit in OS and a positive trend in the improvement of PFS, ORR, and ≥VGPR in favor of IsaPd compared with DaraPd. These findings were consistent with the previous MAIC with a longer follow-up data (present analysis used a maximum follow-up of 4.5 years of OS data from both trials, whereas the previous analysis used a follow-up of 3 years from ICARIA and unknown from APOLLO), which highlights the therapeutic potential of IsaPd in RRMM, particularly for patients who become refractory to lenalidomide and a PI.

Efficacy and Safety Outcomes in Patients with Cancer-Associated Thromboembolism Treated with Tinzaparin According to the Tumor Site: Meta-Analysis of Prospective Studies Involving 1413 Individual Patients' Data

Introduction DOAC (direct oral anticoagulant) and LMWH (Low Molecular Weight Heparin) are recommended by international guidelines as first line therapy in patients with CAT (Cancer Associated Thrombosis), but the risk of venous thromboembolic recurrence (rVTE) and the risk of major bleeding (MB) remains high. Several scores have attempted to predict these risks with disappointing results, probably because of heterogeneous treatments and small sample sizes representing various tumor sites. We conduct a meta-analysis on individual patient level data to report the rate of rVTE and MB by 6 months according to site of cancer from prospective studies involving CAT patients on tinzaparin. Methods The meta-analysis is based on individual patient data from prospective studies that included patients with CAT treated with tinzaparin who were followed for 6 months (PROSPERO registration CRD42019119907). Eligible studies must include controlled data and a central decision for study outcomes. Main outcomes were cumulative incidence of rVTE and MB at 6 months. Time-to-event outcomes were estimated using competing risk regression (Fine and Gray's method with death as competing risk). Cumulative incidences were presented with corresponding 95% confidence interval (95% CI). Time-to-event outcomes were estimated using competing risk regression (Fine and Gray's method with death as competing risk). Cumulative incidences were presented with corresponding 95% confidence interval (95% CI). Results Three prospective cohort studies (AXA - NCT02898051, 308 patients; PREDICARE - NCT03099031, 409 patients (1); TICAT - NCT01164046, 247 patients (2) and the Tinzaparin arm of the CATCH study (NCT01130025, 449 patients (3)) were included. A total of 1413 patients (median age 65.0 [IQR 56.0-73.5], 51.9% female) were included in the meta-analysis. Cancers were mostly solid tumors (91.7%); primary sites were gastrointestinal (27.0%), lung (18.0%), breast (11.8%), genitourinary (14.4%); ECOG performance status was ≥ 2 for 30.2%. At 6 months, 74 patients experienced a rVTE, for a cumulative incidence of 5.4% [95% CI: 4.3%; 6.7%], while 47 patients had a MB, for a cumulative incidence of 3.4% [95% CI: 2.6%; 4.6%]. Clinical outcomes according to the type of cancer are summarized in the Table. rVTE cumulative incidence varied from 0.6 to 8.5% and rVTE occurred more frequently in patients with genitourinary, lung and gynecological cancers, while MB cumulative incidence ranged from 2.1% to 4.2% across different cancer sites (Table). Overall mortality ranged from less than 18% (breast and hematological cancers) to 45.2% (lung cancer) within 6 months of the index VTE. Conclusion In CAT patients receiving tinzaparin for up to 6 months, rVTE and MB vary according to tumor site. The CAT entity is not sufficient to describe a situation at risk of recurrence or bleeding for patient management.

Pomalidomide, Dexamethasone, and Daratumumab Versus Other Triplet Regimens in Patients with Relapsed or Refractory Multiple Myeloma: A Population-Adjusted Indirect Treatment Comparison

Introduction: Multiple myeloma (MM) is an incurable malignant plasma cell disorder in which most patients eventually relapse or become refractory to treatment. Lenalidomide (LEN) is typically administered as part of first-line regimens; however, patients who receive LEN usually relapse, necessitating improved options in later lines of therapy. The phase 2, non-randomized MM-014 trial (NCT01946477) demonstrated promising efficacy of pomalidomide (POM) plus dexamethasone and daratumumab (DPd) among patients with relapsed or refractory MM (RRMM), with an overall response rate of 77.7% and a median overall survival (OS) of 56.7 months at a median follow-up of 41.9 months. With the evolving treatment landscape of RRMM, however, there is limited head-to-head data comparing DPd to other standard of care therapies to inform clinical decision-making. Objective: To investigate the comparative efficacy of DPd versus non-POM-containing triplets in adult patients with RRMM using population-adjusted indirect comparison (PAIC) approaches. Methods: A PAIC was conducted using individual patient-level data of 112 adult patients with RRMM who were exposed to LEN (75.9% refractory to LEN) from cohort B of the MM-014 trial. Investigator-assessed OS was the primary endpoint of this analysis. A systematic literature review was performed to identify comparable interventional studies assessing the efficacy of non-POM-containing triplets for RRMM as well as potential prognostic variables (PVs). PVs verified by clinicians were selected for inclusion if significantly associated with OS (p < 0.1) in the MM-014 trial based on univariate Cox regression model. The suitability of these potential trials for PAIC was assessed by considering the comparability of study design, inclusion/exclusion criteria, patients' baseline characteristics, and outcomes. A retrospective analysis utilizing Flatiron electronic health record data from Jan 2011 to July 2022 was previously performed to evaluate the effect of LEN-refractory status on OS using multivariable Cox regression. Given the lack of a common comparator, unanchored PAIC was performed for OS using matching-adjusted indirect comparison (MAIC) and confirmatory simulated trial comparison (STC). A sensitivity analysis was conducted to explore the impact of LEN-refractory status as a PV. Results: Six trials were systematically identified as eligible comparators in initial screening including TOURMALINE-MM1, CANDOR, CASTOR, POLLUX, ASPIRE, and LEPUS. The analyses using Flatiron data showed that patients who were refractory to LEN had significantly shorter OS than those exposed to LEN but not refractory (Hazard ratio: 1.51; 95% CI: 1.17, 1.94), indicating the unsuitability to compare treatments based on studies with large variations in proportion of LEN-refractory patients. Four trials were excluded due to differences violating the transitivity assumption (0% LEN-refractory patients in TOURMALINE-MM1 and POLLUX; LEN-refractory status not reported in ASPIRE; only Chinese patients in LEPUS). The remaining two trials are phase 3 randomized controlled trials investigating carfilzomib plus dexamethasone and daratumumab (DKd) (CANDOR; NCT03158688; Usmani et al. Blood Adv. 2023; Epub ahead of print) and daratumumab plus bortezomib and dexamethasone (DVd) (CASTOR; NCT02136134; Sonneveld et al. J Clin Oncol. 2023; 41:1600-1609). In the base case MAIC adjusting for creatinine clearance, time from diagnosis, International Staging System score, and refractory to LEN, DPd was associated with numerical reduction in the risk of death compared to DKd and DVd (Table). In addition, the base case STC showed significant improvement of OS for DPd over DKd and DVd, confirming the findings from MAIC (Table). In the sensitivity analyses excluding LEN-refractory status, the effect estimates were substantially different from those in the base case analyses (>40% difference), demonstrating the importance of adjusting for LEN refractory status. Conclusions: In patients with RRMM, consistent OS benefit was found for DPd versus DKd and DVd regardless of the PAIC technique used. These findings support the use of DPd over DKd and DVd in treating patients with RRMM, especially given the oral route of administration of pomalidomide. This study also highlighted the importance of considering LEN refractory status when comparing efficacy of treatments for RRMM across trials.

Virtual Patient Simulation Using Copula Modeling.

Virtual patient simulation is increasingly performed to support model-based optimization of clinical trial designs or individualized dosing strategies. Quantitative pharmacological models typically incorporate individual-level patient characteristics, or covariates, which enable the generation of virtual patient cohorts. The individual-level patient characteristics, or covariates, used as input for such simulations should accurately reflect the values seen in real patient populations. Current methods often make unrealistic assumptions about the correlation between patient's covariates or require direct access to actual data sets with individual-level patient data, which may often be limited by data sharing limitations. We propose and evaluate the use of copulas to address current shortcomings in simulation of patient-associated covariates for virtual patient simulations for model-based dose and trial optimization in clinical pharmacology. Copulas are multivariate distribution functions that can capture joint distributions, including the correlation, of covariate sets. We compare the performance of copulas to alternative simulation strategies, and we demonstrate their utility in several case studies. Our work demonstrates that copulas can reproduce realistic patient characteristics, both in terms of individual covariates and the dependence structure between different covariates, outperforming alternative methods, in particular when aiming to reproduce high-dimensional covariate sets. In conclusion, copulas represent a versatile and generalizable approach for virtual patient simulation which preserve relationships between covariates, and offer an open science strategy to facilitate re-use of patient data sets.

An economic evaluation of first-line cryoballoon ablation versus antiarrhythmic drug therapy for the treatment of paroxysmal atrial fibrillation from a German healthcare service payer perspective

Abstract Background Three recent randomized controlled trials demonstrated that, in patients with symptomatic paroxysmal atrial fibrillation (PAF), first-line pulmonary vein isolation (PVI) with cryoballoon ablation reduces atrial arrhythmia recurrence when compared to initial antiarrhythmic drug (AAD) therapy. Purpose To evaluate the cost-effectiveness of initial rhythm control therapy from a German healthcare service payer perspective. Methods The cost-effectiveness model (CEM) structure consisted of a hybrid decision tree and Markov model, where the decision tree informed the initial health state allocation in the first cycle of the Markov model and had a one-year time horizon. The Markov model had a 40-year time horizon using a three-month cycle length. Health benefits were expressed in quality-adjusted life years (QALYs). Costs and benefits were discounted at 3.5% p.a. Individual patient-level data from 703 participants with untreated PAF enrolled into Cryo-FIRST (NCT01803438), STOP AF First (NCT03118518) and EARLY-AF (NCT02825979) were used to estimate efficacy, resource use and health-related quality of life parameters. Cost inputs were sourced from diagnosis-related groups and the Institute for the Hospital Remuneration System (InEK). Where parameters could not be derived, inputs were sourced from available published literature or determined through clinical expert opinion. Probabilistic sensitivity analyses were conducted to explore the impact of any assumptions on model outputs. Results In those treated with cryoablation, the three-month rate of AF recurrence was reduced by 46.7% (p<0.001) on average. Similarly, the average monthly rate of receiving an ablation following initial treatment was reduced by 72.8% (p<0.001). Cryoablation was also associated with a 4.3% (p=0.025) increase in health-related quality of life at 12 months, assessed through the standard EQ-5D-3L instrument. There was no difference in the rate of AF resolution in those who failed initial treatment. CEM results are shown in Table 1. Analysis shows that cryoablation is cost-effective, incurring a cost of ∼€1,000 per patient over a lifetime compared to AADs, while offering an increase in QALYs. Cryoablation attains an average ICER of ∼€5,500, with a 94.1% probability of being cost-effective at a willingness-to-pay threshold of €35,000 per QALY gained. Through 5,000 iterations, the probabilistic sensitivity analysis indicates that cryoablation has ∼20% probability of being cost-saving. Individuals are expected to receive a total of ∼1.2 ablations over a lifetime, regardless of initial treatment. Although, those initially treated with cryoablation as opposed to AADs experience a 45% reduction in time spent in AF health states. Conclusion Initial rhythm control with cryoballoon ablation in PAF is a cost-effective treatment option in a German healthcare setting.Table 1:Key results (per patient)

Cardiovascular risk prediction in a New Zealand cohort of patients with known cardiovascular disease - a regional individual patient-level data linkage study

Abstract Background National health databases have been underutilised in the development of cardiovascular disease (CVD) risk prediction equations for people with known CVD, yet can inform population level practice and policy—including identification of health equity gaps, high-risk groups, and high-risk localities. With the recent introduction of novel and expensive preventive cardiovascular drugs, stratifying risk in patients with prior CVD could be required to inform cost-effective use of these medications in the future. Purpose We sought to develop 5-year CVD risk equations in adults with known history of CVD in Aotearoa New Zealand, using data from administrative health databases and characterize distribution of risk according to sex. Methods All people above 18 years of age with a history of CVD hospitalization, living in the Auckland and Northland regions of Aotearoa New Zealand (approximately one-third of the national population) on 1 January 2014 were included. Participants were followed for 5 years (31 December 2018) or until death or the outcome of interest, being any fatal or non-fatal CVD hospitalization (ischemic heart disease, heart failure, ischemic stroke, or peripheral artery disease) during follow-up. Sex-specific 5-year CVD risk prediction equations were developed using multivariable Fine and Gray models taking competing risk of non-cardiovascular death into account. Predictor variables included were age; ethnicity; deprivation level; time since most recent CVD hospitalization, history of smoking, heart failure, atrial fibrillation, chronic renal disease, and diabetes mellitus; use of anti-thrombotic, blood pressure-lowering, and lipid-lowering drugs; and blood levels of glycosylated hemoglobin (HbA1c), non-HDL cholesterol, and creatinine. Results The cohort comprised 42,130 men (median age: 68 years [interquartile range (IQR): 59–76]) and 31,717 women (median age: 70 years [IQR: 60–78]) with 12,875 and 8,734 observed events during follow-up, respectively. Sex-specific 5-year CVD risk prediction equations with excellent calibration (accuracy) were developed (Figure 1). Increasing age; high deprivation level; more recent CVD hospitalization; history of smoking, heart failure, atrial fibrillation, chronic renal disease, diabetes mellitus; use of blood-pressure lowering drugs; and high blood levels of HbA1c, non-HDL cholesterol, or creatinine were predictors of elevated risk. There was a wide range of predicted 5-year risk, with 15% of men and 24% of women having a predicted risk below 15%, and 24% and 20%, respectively, having a predicted risk of 40% or above (Figure 2). Conclusion This study demonstrates that it is possible to develop accurate CVD risk prediction equations from administrative health databases for patients with known CVD. Marked heterogeneity in 5-year risk of a subsequent CVD event emphasise that the current ‘one size fits all’ approach to secondary prevention of CVD requires review.Figure 1Figure 2

Comparative Efficacy of Ciltacabtagene Autoleucel Versus Idecabtagene Vicleucel in the Treatment of Patients with Relapsed or Refractory Multiple Myeloma Previously Treated with 2-4 Prior Lines of Therapy Using a Matching-Adjusted Indirect Comparison

Introduction: Two novel B-cell maturation antigen (BCMA)-targeting chimeric antigen receptor T-cell (CAR-T) therapies, ciltacabtagene autoleucel (cilta-cel; Carvykti) and idecabtagene vicleucel (ide-cel; Abecma), were recently developed and approved in 2022 and 2021, respectively, for heavily pre-treated relapsed/refractory multiple myeloma (RRMM) based on the CARTITUDE-1 and KarMMa studies. A previous comparison of these therapies in triple-class exposed patients demonstrated improved efficacy with cilta-cel versus ide-cel (Martin T, et al. Curr Med Res Opin. 2023;39(1):81-89). More recently, the efficacy and safety of these CAR-Ts were demonstrated earlier in the treatment pathway in lenalidomide-refractory, proteasome inhibitor (PI)-exposed patients with 1-3 prior line(s) of treatment (LOT; cilta-cel; CARTITUDE-4), and in triple-class exposed patients with 2-4 prior LOT (ide-cel; KarMMa-3). In the absence of a head-to-head comparison of these two therapies, an unanchored matching-adjusted indirect comparison (MAIC) was used to estimate their relative efficacy using data from CARTITUDE-4, CARTITUDE-1, and KarMMa-3. Methods: While KarMMa-3 and CARTITUDE-4 both included DPd in the physician's choice control arm, there is no published data on this subgroup from KarMMa-3. As such, an unanchored MAIC was performed utilizing individual patient-level data (IPD) from 245 patients who received cilta-cel; 208 patients from CARTITUDE-4 who had received 1-3 prior LOT and 37 patients from CARTITUDE-1 who had received 3-4 prior LOT. From these data, cilta-cel patients who fulfilled the inclusion criteria from KarMMa-3 (2-4 prior LOT and triple-class exposed) were selected, and their outcomes were compared against published summary data for ide-cel (N=254) from KarMMa-3 using reconstructed IPD for response endpoints and simulated IPD from published Kaplan-Meier (KM) curves for progression-free survival (PFS). Further imbalances in patient characteristics were adjusted for by weighting the cilta-cel IPD to match the reported baseline characteristics of KarMMa-3 in terms of refractory status, ISS stage, time to progression on prior LOT, cytogenetic risk, and presence of extramedullary disease. Comparative efficacy was estimated for overall response rate (ORR), very good partial response or better (≥VGPR) rate, complete response or better (≥CR) rate, and PFS. For binary endpoints, relative effects were quantified using relative response ratios (RRs) with 95% confidence intervals (CIs) derived from a weighted logistic regression analysis. For PFS, the hazard ratio (HR) with 95% CIs for cilta-cel versus ide-cel was estimated using a weighted Cox proportional hazards model. Results: After applying the KarMMa-3 inclusion and exclusion criteria to the CARTITUDE-1 and CARTITUDE-4 IPD (excluding patients with only 1 prior LOT or no prior daratumumab), 36 patients and 49 patients were included from the cilta-cel trials, respectively. After adjustment, patients in the cilta-cel group (effective sample size [ESS]=46) were significantly more likely to achieve an overall response (RR: 1.18 [95% CI: 1.04, 1.33]; p=0.0404), and to reach deeper levels of response as demonstrated by ≥VGPR (RR: 1.34 [95% CI: 1.16, 1.56]; p=0.0027), and ≥CR (RR: 1.91 [95% CI: 1.56, 2.34]; p<0.0001) compared to patients in the ide-cel group ( Table). In addition, a statistically significant reduction of 49% in the risk of disease progression or death was observed with cilta-cel versus ide-cel (HR: 0.51 [95% CI: 0.31, 0.84]; p=0.0078; Figure). Conclusions: This unanchored MAIC showed that cilta-cel provides clinical benefit over ide-cel across response outcomes and PFS for patients with triple-class exposed RRMM treated with 2-4 prior LOT. These results are consistent with the previous comparison in later line, triple-class exposed patients, continuing to highlight the value of cilta-cel for these patients. Furthermore, these comparisons provide valuable information to contextualize the efficacy of cilta-cel compared to other available CAR-Ts for multiple myeloma.

Efficacy of Subcutaneous Epcoritamab Vs Tisa-Cel in R/R LBCL CAR T-Naive and CAR T-Eligible Patients: An Indirect Comparison

Background: Epcoritamab is a subcutaneous, off-the-shelf, T-cell-engaging CD3xCD20 bispecific antibody (bsAb) with powerful single-agent activity and a manageable safety profile in relapsed/refractory (R/R) large B-cell lymphoma (LBCL), recently approved in the US for the treatment of adults with R/R diffuse large B-cell lymphoma (DLBCL), not otherwise specified, including DLBCL arising from indolent lymphoma, and high-grade B-cell lymphoma after ≥2 lines of systemic therapy. While chimeric antigen receptor T-cell therapy (CAR T) has demonstrated remarkable efficacy in R/R LBCL, there are still barriers to wide adoption, including logistical restrictions and manufacturing limitations. In the absence of head-to-head clinical trial data comparing CAR T with T-cell-engaging bsAbs, we conducted an indirect treatment comparison between CAR T-naive patients treated with epcoritamab in EPCORE NHL-1 (NCT03625037), including a subgroup of CAR T-eligible patients, identified according to ZUMA-1 (NCT02348216) eligibility criteria and utilized in a prior unanchored, matching-adjusted indirect comparison (MAIC) of epcoritamab and axicabtagene ciloleucel (axi-cel) in patients with R/R DLBCL (Thieblemont et al EHA 2023 #P1154), and patients treated with tisagenlecleucel (tisa-cel) in the JULIET trial (NCT02445248). The CAR T-eligible population was of interest for the comparison of epcoritamab vs tisa-cel as it was most similar in baseline clinical characteristics and comparable to patients examined in CAR T trials. Methods: Published data on the overall response rate (ORR), complete response (CR) rate, digitized progression-free survival (PFS), and overall survival (OS) for tisa-cel from JULIET publications were used in MAIC vs individual patient-level data (IPD) of CAR T-naive and CAR T-eligible patients from EPCORE NHL-1 (Nov 2022 data cut). Analyses were adjusted for imbalances in the following baseline characteristics between IPD from EPCORE NHL-1 and aggregate data from JULIET: age 65 years or older, gender, Eastern Cooperative Oncology Group performance status, disease stage, DLBCL, refractory to last therapy, and prior autologous stem cell transplant. Kaplan-Meier methodology was used to estimate survival. Results: A total of 96 CAR T-naive patients from EPCORE NHL-1 were included in the analysis, with an effective sample size of 33 CAR T-naive patients after adjustment. The CAR T-eligible subgroup from EPCORE NHL-1 included 57 patients with an effective sample size of 21 patients after adjustment. After adjustment there was a significant difference in ORR for epcoritamab vs tisa-cel (77.9% vs 53.0%, respectively; difference [95% confidence interval (CI)]: 24.8% [9.5, 40.2]; P=0.002) in the CAR T-naive cohort, and in CR rate (52.3% vs 39.1%, respectively; difference [95% CI]: 13.2% [ 5.9, 32.3]; P=0.174). Also, in the CAR T-naive cohort there was a trend toward PFS benefit for epcoritamab vs tisa-cel after adjustment (hazard ratio [HR]: 0.725; 95% CI: 0.447, 1.177; P=0.194) and OS (HR: 0.611; 95% CI: 0.356, 1.049; P=0.074). In the CAR T-eligible subgroup there was a significant difference in ORR (80.8% vs 53.0%, respectively; difference [95% CI] 27.7% [11.0, 44.4]; P=0.001) and CR rate (61.9% vs 39.1%, respectively; difference [95% CI]: 22.8% [1.5, 44.1]; P=0.036) for epcoritamab vs tisa-cel after adjustment. Also, there was a statistically significant survival benefit for OS (HR: 0.450; 95% CI: 0.227, 0.891; P=0.022) and a numerical trend toward benefit for PFS (HR: 0.548; 95% CI: 0.300, 1.003; P=0.051) when comparing epcoritamab vs tisa-cel after adjustment. Conclusion: In the absence of head-to-head data, this MAIC of the R/R LBCL CAR T-naive cohort and the eligible subgroup treated with epcoritamab vs tisa-cel demonstrated improved response rates and survival outcomes. This study underscores the therapeutic potential of epcoritamab as a novel, subcutaneous, off-the-shelf, core therapy for the treatments of patients with R/R LBCL. Indirect comparisons can be biased by cross-trial differences and further comparisons with longer follow-up are warranted.

Updated Results from a Matching-Adjusted Indirect Comparison of Efficacy Outcomes of Isatuximab Plus Carfilzomib and Dexamethasone Versus Daratumumab Plus Lenalidomide and Dexamethasone in Patients with Relapsed and/or Refractory Multiple Myeloma

Introduction: Findings from a previously conducted matching-adjusted indirect comparison (MAIC) between isatuximab plus carfilzomib with dexamethasone (IsaKd) and daratumumab plus lenalidomide and dexamethasone (DRd) among relapsed and/or refractory multiple myeloma (RRMM) patients receiving one to three prior lines of therapy (LoT) demonstrated a significantly better progression-free survival (PFS) and a positive trend for overall survival (OS) in favor of IsaKd compared with DRd (median follow up of 20.7 months for IKEMA trial). Additionally, IsaKd demonstrated an acceptable safety profile with significantly lower rates of several treatment-emergent adverse events than DRd [Richter J et al, Cancer Med 2023;12(7):8005-8017; Richter J et al.(Publication Number 1962), 63 rd ASH Annual Meeting and Exposition, 2021; Richter J, et al. P-213, Clin Lymphoma Myeloma Leuk. 2021;21(Suppl 2):S156-S157]. The present study used longer follow-up data from the phase 3 IKEMA trial (NCT03275285) to update the previous MAIC evaluating the comparative efficacy of IsaKd versus DRd. Methods: A MAIC was performed with the latest available individual patient-level data (IPD) from the IKEMA trial (IsaKd arm, n = 179) (data cutoff: February 2023 for OS and January 2022 for PFS), and aggregate data from study publications of phase 3 POLLUX trial (DRd arm, n = 286, NCT02076009) [Bahlis NJ et al. Leukemia 2020;34(7):1875-1884; Dimopoulos MA et al. Hemasphere 2022;6:13]. Patients were selected from the IsaKd arm of the IKEMA trial to match the inclusion and exclusion criteria of the POLLUX trial. Individual patients from the IKEMA trial were then assigned weights such that (1) weighted mean baseline characteristics for patients in the IsaKd arm matched those reported for patients in the DRd arm, and (2) each patient's weight was equal to their estimated odds (propensity) of being in the POLLUX trial versus the IKEMA trial. After matching, outcomes were compared across the two trial populations. For PFS and OS, hazard ratio (HR) and 95% confidence interval (CI) were reported from a Cox proportional hazards (PH) model using the weighted sample for IsaKd and the pseudo-IPD for DRd extracted from published Kaplan-Meier curves via digitization software [Mark M et al, Engauge Digitizer Software; http://markummitchell.github.io/engauge-digitizer]. For any violations of the PH assumption, the restricted mean survival time (RMST) at median follow-up of IKEMA for the outcome of interest was calculated instead. Results: After the MAIC adjustment, all available baseline characteristics were well balanced across the trial populations. For PFS, as the PH assumption was violated ( χ2 = 5.52, p = 0.019), RMST method was used to compare the two treatments. PFS was significantly better with IsaKd versus DRd (mean difference [95% CI]: 3.63 months [0.14-7.12]; p = 0.04) at 44 months follow-up using RMST method. For OS, the PH assumption ( χ2 = 0.89, p =0.35) was not violated. A numeric, statistically non-significant trend in favor of IsaKd versus DRd was observed for OS (Figure 1). Conclusions: This MAIC analysis demonstrated a significant benefit in PFS and a positive trend in the improvement of OS with IsaKd compared with that with DRd. These findings with a longer follow-up were consistent with the previous analyses conducted with a follow-up of 2 years. These data highlight the therapeutic potential of IsaKd, a lenalidomide-free regimen, in RRMM patients receiving one to three prior LoT compared with a lenalidomide-based regimen.

Matching-Adjusted Indirect Comparison (MAIC) of Efficacy and Safety of Lisocabtagene Maraleucel (liso-cel) and Mosunetuzumab for the Treatment (Tx) of Third Line or Later (3L+) Relapsed or Refractory (R/R) Follicular Lymphoma (FL)

Background:The tx landscape of R/R FL is evolving with the availability of anti-CD19 chimeric antigen receptor (CAR) T cell therapies and CD20xCD3 bispecific T-cell-engaging monoclonal antibodies (mAb). Liso-cel is an autologous, CD19-directed, 4-1BB CAR T cell therapy which recently reported a positive benefit-risk tx profile for the 3L+ tx of patients with R/R FL (Morschhauser F, et al. ICML 2023; LBA4). Mosunetuzumab, a CD20xCD3 bispecific T-cell-engaging mAb, is approved for the tx of R/R FL after 2 or more lines of therapy (LOT). In the absence of head-to-head data, we conducted an unanchored MAIC to compare the efficacy and safety of liso-cel and mosunetuzumab for the 3L+ tx of patients with R/R FL. Methods: Unanchored MAICs were used to estimate population-adjusted relative tx effects between individual patient-level data from TRANSCEND FL (NCT04245839; data as of January 27, 2023; median follow-up of 19.3 months) and aggregate data from GO29781 (NCT02500407; data as of August 27, 2021; median follow-up of 18.3 months; N = 90). For TRANSCEND FL, the leukapheresed (intention to treat [ITT]) set (N = 114) was used for primary comparisons of efficacy endpoints: Progression-free survival (PFS), objective response rate, and complete response rate (all per independent review committee assessment). Overall survival was not analyzed due to the low rate of events. The treated efficacy set (n = 101) was used for a sensitivity analysis of efficacy comparisons. The treated set (n = 107) was used for comparisons of safety endpoints: cytokine release syndrome (CRS), neurological events (NEs), serious infections, and use of corticosteroids or tocilizumab for CRS. Baseline characteristics and outcome measures in TRANSCEND FL were redefined to align with those reported in GO29781. Data from TRANSCEND FL were weighted using a method-of-moments propensity score model to match the marginal distribution (ie, mean, variance) of clinical factors among patients from GO29781. A panel of expert clinicians selected and ranked tx effect modifiers separately for efficacy and safety endpoints. For efficacy, factors adjusted for included number of prior systemic LOTs, progression of disease ≤ 24 months from initiation of immunochemotherapy, bulky disease at screening, FL International Prognostic Index risk factor, refractory to last therapy, double refractory to an anti-CD20 antibody and alkylating agent, ECOG PS at screening, sex, and prior autologous HSCT. For safety, factors adjusted for included age, bulky disease at screening, refractory to last therapy, number of prior systemic LOTs, ECOG PS at screening, Ann Arbor stage, and sex. Hazard ratios (HR) were used to compare PFS, and odds ratios (OR) were used for response rates and safety outcomes. Results: After adjusting for all available factors, when compared with mosunetuzumab, liso-cel was associated with higher objective response rate (OR, 3.78 [95% CI, 1.48‒9.67]), and complete response rate (OR, 6.46 [2.85‒14.65]), and improved PFS (HR, 0.28 [0.16‒0.49]) based on an effective sample size (ESS) of 71.4 (Figure). Results were consistent across all scenario analyses, including unadjusted comparisons and sensitivity analyses using the treated efficacy set for TRANSCEND FL. Although liso-cel was associated with higher incidence of any-grade CRS (OR, 1.86 [1.01‒3.43]) and any-grade NEs (OR, 2.16 [0.72‒6.44]), it demonstrated a lower incidence of grade ≥ 3 CRS (OR, 0.45 [0.04‒5.13]) and grade 3-4 serious infections (OR, 0.35 [0.12‒1.03]) based on an ESS of 81.9 compared with mosunetuzumab. Furthermore, liso-cel was associated with reduced use of corticosteroids for CRS management (OR, 0.14 [0.03‒0.65]), while the use of tocilizumab was higher in patients treated with liso-cel (OR, 2.27 [0.86‒5.99]). Conclusions: In this MAIC, liso-cel was associated with improved efficacy compared with mosunetuzumab. Liso-cel had a lower incidence of grade ≥ 3 CRS, grade 3-4 serious infections, and steroid use for management of CRS; however, liso-cel exhibited higher incidence of any-grade CRS, any-grade NEs, and tocilizumab use for CRS management. Safety comparisons were confounded by preventative use of corticosteroids in all patients in GO29781 and definition differences, most notably for NEs. These findings highlight a potential positive benefit-risk profile of liso-cel over mosunetuzumab as a 3L+ tx option for R/R FL.