Efficacy of zanubrutinib versus acalabrutinib for relapsed or refractory chronic lymphocytic leukemia (R/R CLL): a matching-adjusted indirect comparison (MAIC)

Background

Zanubrutinib is a next-generation covalent Bruton tyrosine kinase inhibitor (BTKi). Zanubrutinib represents an appropriate BTKi for patients with relapsed or refractory chronic lymphocytic leukemia (R/R CLL) due to its high affinity and inhibition potency for BTK, and because it is the only BTKi that maintains plasma levels above IC₅₀ for 24 h.^1,2 Zanubrutinib is the only BTKi to demonstrate progression-free survival (PFS) superiority versus ibrutinib in R/R CLL in ALPINE, ³ which was sustained with an extended follow-up of 39 months. ⁴ Acalabrutinib, a second-generation BTKi, showed improved PFS versus rituximab-idelalisib/bendamustine in R/R CLL in ASCEND, ⁵ but PFS noninferiority to ibrutinib in patients with R/R CLL and del(17p) or del(11q) mutations in ELEVATE-RR.^6,7 No head-to-head study of zanubrutinib and acalabrutinib in R/R CLL exists.

Objective

An indirect treatment comparison (ITC) was performed to evaluate the relative efficacy of zanubrutinib and acalabrutinib in R/R CLL. This ITC showed a significant investigator-assessed PFS (PFS-INV) and complete response (CR) advantage and a trend toward improvement in overall survival (OS) for zanubrutinib compared with acalabrutinib.

Design

An unanchored matching-adjusted indirect comparison (MAIC) was performed, using data sourced from individual patient-level data (IPD) from ALPINE for zanubrutinib, and aggregate data from ASCEND for acalabrutinib. This comparison was preferred over ALPINE versus ELEVATE-RR due to a greater similarity of the patient populations. The inclusion criteria for ELEVATE-RR were limited to high-risk patients, whereas ALPINE and ASCEND were not restricted to a risk profile. ALPINE (NCT03734016) is a phase III, open-label, randomized, multicenter trial comparing the efficacy and safety of zanubrutinib (160 mg, twice daily) with ibrutinib (420 mg, once daily) in patients with R/R CLL or small lymphocytic lymphoma (SLL). ³ Adults with a confirmed diagnosis of CLL/SLL, ⩾1 previous line of therapy, and no previous treatment with BTKi were eligible for the study. The primary endpoint was investigator-assessed overall response, and key secondary endpoints included PFS-INV, independent review committee-assessed PFS (PFS-IRC), and OS. ASCEND (NCT02970318) is a phase III, open-label, randomized, multicenter trial comparing the efficacy and safety of acalabrutinib monotherapy (100 mg twice daily) with investigator’s choice of idelalisib plus rituximab or bendamustine plus rituximab in patients with R/R CLL. ⁵ Adults with CLL who had received ⩾1 previous systemic therapy were eligible for the study. The primary endpoint was PFS-IRC, and key secondary endpoints included investigator-assessed overall response, PFS-INV, and OS.

Methods

This analysis used an unanchored MAIC due to the lack of a common comparator arm between ALPINE and ASCEND, and was conducted in accordance with the guidance from the National Institute for Health and Care Excellence (DSU TSD 18). ⁸ The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for cohort studies ⁹ (see Supplementary Material Table S1 for a completed checklist).

The MAIC used similar median follow-up times (ALPINE data cutoff (DCO) September 2023, 39 months; ASCEND DCO October 2020, 36 months). An adjustment was made to correct for the impact of COVID-19 on ALPINE outcomes, as the entire trial was conducted during the COVID-19 pandemic. Adjustments within ALPINE were made by neutralizing the impact of COVID-19 on ALPINE in terms of death, treatment discontinuation, and COVID-related dose holds. COVID-19 adjustment was not feasible for ASCEND given the unavailability of IPD. Regardless, COVID-19 adjustment was deemed unnecessary for ASCEND as (1) only a few months of the ASCEND trial follow-up used in this MAIC overlapped with the COVID-19 pandemic, (2) the DCO did not overlap with any of the peaks of the COVID-19 pandemic in terms of number of global cases and excess mortality,^10,11 and (3) only one COVID-19 death was reported in ASCEND in the latest reported DCO of September 2021 (46.5 months of follow-up). ⁴

IPD from the zanubrutinib arm of ALPINE (N = 327) were reweighted to match the profile of acalabrutinib-treated patients in ASCEND (N = 155) and adjusted for variables identified as predictors of treatment effect or prognostic factors. The variables matched were considered complete and representative of the true differences between the two trials. Variables adjusted for in the base case analysis included age, sex, Eastern Cooperative Oncology Group performance status, geographic region, immunoglobulin heavy chain variable (IGHV) mutation status, del(17p), del(11q), TP53 mutation status, bulky disease, Rai/Binet stage, cancer type (CLL/SLL), absolute lymphocyte and neutrophil counts, platelet count, and number and type of prior therapies. Sensitivity analyses included several other scenarios to consider the impact of matching for different sets of variables (see Supplementary Material Table S2).

Pseudo IPD for PFS and OS in the acalabrutinib arm of ASCEND were reconstructed from the digitized Kaplan–Meier (KM) curves reported in the ASCEND publication using the algorithm by Guyot et al. ¹² A weighted Cox proportional hazard model was used to compare PFS-INV and OS, and a weighted logistic regression model was used to compare CR.

Results

The zanubrutinib population in ALPINE was filtered to patients with existing data on the selected baseline characteristics, reducing the number of patients from 327 to 308. After population adjustment, the effective sample size (ESS) for zanubrutinib was 185 (60% of the starting filtered population). Before matching and adjustment, the baseline characteristics of the zanubrutinib and acalabrutinib populations were different with regard to IGHV and TP53 mutation status, and geographic region. After matching, there were no differences between the variables included in the matching. The differences between unmatched variables in the base case (i.e., complex karyotype and beta 2-microglobulin) were not significant pre- and post-matching; however, these two variables were matched in the sensitivity analysis (See Supplementary Material Table S3).

In the unadjusted population, the hazard ratio (HR) of zanubrutinib (N = 327) versus acalabrutinib (N = 155) for PFS-INV was 0.77 (95% confidence interval (CI): 0.55–1.07; p = 0.1213). The HR for OS was 0.60 (95% CI: 0.37–0.97; p = 0.0354). Post-matching, PFS-INV was improved significantly for zanubrutinib versus acalabrutinib (HR = 0.68 (95% CI: 0.46–0.99); p = 0.0448; Figure 1(a)). OS showed a trend toward improvement for zanubrutinib post-matching (HR = 0.60; (95% CI: 0.35–1.02); p = 0.0575; Figure 1(b)). The odds ratio (OR) for CR significantly favored zanubrutinib over acalabrutinib in both unadjusted (OR = 2.88 (95% CI: 1.18–7.02); p = 0.0198) and base-case adjusted analyses (OR = 2.90 (95% CI: 1.13–7.43); p = 0.0270; Table 1). The robustness of the base-case findings was verified by the sensitivity analyses; results were consistent with the base case, showing favorable PFS-INV and CR for zanubrutinib (Table 1).

OPEN IN VIEWER

Figure 1.

Unadjusted and base-case adjusted clinical outcomes comparing zanubrutinib and acalabrutinib. (a) PFS-INV. (b) OS.

OPEN IN VIEWER

Table 1.

Relative treatment effects for base case and sensitivity analyses.

PFS, OS, and CR outcomes were calculated using the ESS before rounding (i.e., ESSs used in these calculations were: base-case adjusted population = 184.8; S1 = 188.9; S2 = 210.3; S3 = 208.1; S4 = 188.2; S5 = 187.4; S6 = 78.2). Cells shaded gray indicate p < 0.05.

a

Covariates matched in the base-case and sensitivity analyses are shown in Supplementary Material Table S2.

b

ESSs are rounded up to the nearest integer to ensure clinical relevance.

CI, confidence interval; CR, complete response; ESS, effective sample size; HR, hazard ratio; ITT, intent-to-treat; OR, odds ratio; OS, overall survival; PFS-INV, investigator-assessed progression-free survival.

Discussion

The results of this MAIC showed a significant PFS-INV and CR advantage and a trend toward improvement in OS for zanubrutinib compared with acalabrutinib. The trend in OS without reaching statistical significance is consistent with the expected survival outcomes for indolent lymphomas. ¹³ This MAIC has limitations typical of cross-trial comparisons. First, in the absence of head-to-head trials, MAIC provides a valuable method for indirect comparisons by adjusting for cross-trial differences; however, it remains subject to potential biases and confounding factors. Despite adjustments for key prognostic factors, residual bias remains, making MAIC less robust than a well-powered randomized controlled trial. Second, the unanchored MAIC methodology assumes that differences between the two treatment arms across two different trials can be entirely explained by the variables selected for matching. This assumption could lead to potentially unmeasurable bias. To reduce this bias, the MAIC included numerous variables for matching. Any variable not matched in the base case was included in the sensitivity analysis. However, not all variables (e.g., R/R status; i.e., relapsed vs refractory) were reported in the ASCEND study and therefore could not be matched. Adjustment for the impact of COVID-19 and considering DCOs with similar follow-up across the two trials further reduces possible biases in the comparison. Third, MAICs are designed for situations when IPD are available for one trial while only aggregate data are available for another ⁸ ; however, aggregate data include less information than IPD. Specifically, potential minor measurement errors may arise due to the reconstruction of digitized KM curves. This analysis used the Guyot algorithm to reconstruct the survival curves in the ASCEND publication, a method that is commonly used in MAICs. Maximum effort was made to recover the KM curves and ensure minimal measurement error. Fourth, a comprehensive efficacy comparison should include PFS-IRC outcomes. However, PFS-IRC was not available in the analyzed ASCEND or ALPINE DCOs and therefore was not analyzed in this MAIC. Last, this study did not compare safety for zanubrutinib versus acalabrutinib, given different treatment exposure times across the two trials. Safety of a drug is best evaluated via meta-analyses that use all available evidence across all indications. ¹⁴ A recent meta-analysis of 61 trials involving 6959 patients who received ibrutinib ± anti-CD20 antibody, acalabrutinib, and zanubrutinib extensively analyzed the adverse event (AE) profiles of zanubrutinib and acalabrutinib across several indications and reported differences between the two treatments. ¹⁵

A previous MAIC by Kittai et al. ¹⁶ compared the efficacy and safety of zanubrutinib (aggregate data from ALPINE) versus acalabrutinib (IPD from ASCEND) in R/R CLL. Findings showed that zanubrutinib and acalabrutinib had similar efficacy (PFS-INV) and different AE profiles; however, the MAIC by Kittai et al. had several important limitations. Those relevant to the efficacy comparison include the following: First, length of follow-up is a known treatment effect modifier ¹⁷ ; however, Kittai et al. did not include data from the final analysis of ALPINE, as these data were not published at the time of study. Consequently, median follow-up for ASCEND was notably longer than that for ALPINE (46.5 versus 29.6 months, respectively). Previous examples of acalabrutinib MAICs applied this criterion in selecting trials to be compared in the analysis. ¹⁸ Second, the authors acknowledged that outcomes did not adjust for the timing of ASCEND and ALPINE or the impact of COVID-19. Third, differences in patient characteristics between studies were not fully accounted for, such as insufficient granularity in distinguishing between geographic regions and the number and types of previous treatments. Fourth, outcomes were estimated based on a relatively small ESS of 99 for acalabrutinib post-matching. Lastly, important efficacy outcomes, such as response and OS, were not reported. Regarding the safety comparison, a hypothetical DCO was used to compare AEs across treatments. Additionally, the same adjustments were used to compare safety and efficacy across the trials, even though matched variables impacting treatment efficacy may not impact safety. The rationale for analyzing only a selective set of AEs was also lacking.

Conclusion

Findings from the current MAIC showed a significant PFS-INV and CR advantage and a trend toward improvement in OS for zanubrutinib compared with acalabrutinib. This study provides an understanding of the relative efficacy of zanubrutinib and acalabrutinib and may be considered a more robust analysis than the previous MAIC presented by Kittai et al., as it addresses the limitations of the latter study. MAICs are used for generating hypotheses; ¹⁹ however, all MAICs have limitations. Although randomized controlled trials remain the gold standard for generating relative efficacy evidence, future investigations may benefit from utilizing real-world analyses to gain deeper insights into the comparative effectiveness of zanubrutinib and acalabrutinib. Despite their limitations, real-world studies complement MAIC by providing valuable perspectives on treatment effectiveness in diverse real-world settings.

Supplemental Material