Network Meta-Analysis (NMA) of Chimeric Antigen Receptor (CAR) T-Cell Therapy for the Treatment of Relapsed/Refractory Diffuse Large B-Cell Lymphoma (DLBCL) after 2 Prior Treatments Using Published Comparative Studies

Abstract

INTRODUCTION Historically, the clinical prognosis for patients with relapsed/refractory (r/r) diffuse large B-cell lymphoma (DLBCL) has been poor, with limited curative treatment options. However, the introduction of chimeric antigen receptor CAR T-cell (CAR-T) therapy is changing how r/r DLBCL patients are treated. Regulatory approvals of axicabtagene ciloleucel (axi-cel), lisocabtagene maraleucel (liso-cel) and tisagenlecleucel (tisa-cel) in patients with DLBCL with two or more prior lines of treatment were based on their respective non-comparative trials: ZUMA-1, JULIET, and TRANSCEND NHL 001 (TRANSCEND). To understand the comparative efficacy of these CAR-T therapies, treatment comparisons across trials are necessary. In the absence of direct evidence, several matching-adjusted indirect comparisons (MAICs) have been conducted, but have led to conflicting results. This may be, in part, due to the lack of a common comparator. In parallel, to characterize the efficacy of CAR-T compared to other available treatment options, comparisons of these single-arm studies to external historical standard-of-care (SoC) cohorts have been published. Use of this evidence base allows the construction of a network to with a common comparator, which may reduce bias in between treatment estimates. We leveraged this available evidence to conduct an adjusted indirect comparison of axi-cel, liso-cel, and tisa-cel using published comparative studies of CAR-T products to historical SoC cohorts. METHODS On 17th September 2021, we systematically searched EMBASE and MEDLINE databases. Three additional conferences were searched. Eligible studies enrolled patients with r/r DLBCL and compared approved CAR-T therapies to SoC. Outcomes of interest were response and time-to-event outcomes. Safety outcomes were not reported in SCHOLAR-1 and published comparisons to SCHOLAR-1 so safety could not be explored in this analysis. The systematic search followed the Preferred Reporting Item for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. All study selection and data extraction steps were conducted in dual and independently. For indirect treatment comparisons, network meta-analyses (NMA) were conducted using a Bayesian framework. For dichotomous outcomes, we used logistic regression with binomial link function. A linear regression on log-transformed hazard ratios (HR) were used for available time-to-event outcomes. The NMA used an anchored network with the historical SoC serving as the common comparator. RESULTS The search identified 467 publications, of which 3 were included in the evidence base (Figure 1). In the first study, axi-cel individual patient data (IPD) was compared to SCHOLAR-1 (a historical SoC cohort) IPD using propensity score methods. For the second, a matching adjusted indirect comparison (MAIC) was used to compare liso-cel IPD to SCHOLAR-1 summary data. For the third, tisa-cel IPD was compared to CORAL IPD, a historical SoC cohort (and a sub-cohort of SCHOLAR-1). The results from the ITT population of JULIET were used in this analysis. Available outcomes across all three studies included overall survival (OS) and overall response rate (ORR). Complete response (CR) was analyzed where possible. All three treatments were superior to SoC across all outcomes (Table 1). For OS, axi-cel had a significantly lower hazard ratio for death relative to liso-cel and tisa-cel. Both axi-cel and liso-cel had improved ORR relative to tisa-cel, but were not statistically different from one another. CONCLUSIONS Results of the analyses suggest that axi-cel leads to improved OS in r/r DLBCL relative to liso-cel and tisa-cel. Axi-cel and liso-cel were comparable with respect to response outcomes, showing favorable ORR relative to tisa-cel. These results are in line with MAIC results, where efficacy between CAR-T treatments have been directly compared, but offer the advantage of being able to include a common comparator in the absence of placebo controlled RCTs. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal