A matching‐adjusted indirect comparison of acalabrutinib versus zanubrutinib in relapsed or refractory chronic lymphocytic leukemia

second-generation


C O R R E S P O N D E N C E A matching-adjusted indirect comparison of acalabrutinib versus zanubrutinib in relapsed or refractory chronic lymphocytic leukemia
To the Editor: The second-generation Bruton tyrosine kinase inhibitors (BTKis), acalabrutinib and zanubrutinib, have not been compared with each other in head-to-head randomized controlled trials (RCTs) in relapsed or refractory (R/R) chronic lymphocytic leukemia (CLL), so how they compare in terms of efficacy or safety is unclear.In the absence of headto-head RCTs, anchored or unanchored indirect treatment comparison (ITC) methods can be used to compare treatments. 1 ELEVATE-RR, 2 which evaluated acalabrutinib versus ibrutinib, and ALPINE, 3 which evaluated zanubrutinib versus ibrutinib, have ibrutinib as a common comparator, potentially allowing the treatments to be compared using an anchored ITC. 1 However, these two RCTs are too different to conduct an anchored ITC without a high risk of bias (Table S1): • ELEVATE-RR exclusively enrolled patients with del(17p) and/or del(11q), while ALPINE did not restrict enrollment by these genetic abnormalities.
• The median number of previous lines of treatment was higher in ELEVATE-RR versus ALPINE (2 vs. 1).
In the ASCEND RCT, acalabrutinib was compared with idelalisib plus rituximab or bendamustine plus rituximab in patients with R/R CLL enrolled without enrichment for del(17p) and/or del(11q), meaning the population overlaps considerably with the ALPINE population.ASCEND and ALPINE do not share a common comparator arm, so an unanchored ITC was used to compare outcomes from the acalabrutinib and zanubrutinib arms from these RCTs.To minimize the potential selection bias caused by differences in patient characteristics between studies, a matching-adjusted indirect comparison (MAIC) assigns weights to the population with available individual patient-level data (IPD) so that it matches the aggregated baseline data of the population with which it will be compared. 1,4This study aimed to compare the efficacy and safety of acalabrutinib with zanubrutinib in R/R CLL using IPD from ASCEND and published aggregate data from ALPINE using an unanchored MAIC, as summarized in the Video S1.
An overview of the methodology is presented in Figure 1 and full details of the methods and results are included in the Tables S2-S6 and Figures S1-S7.In the unanchored MAIC, acalabrutinib IPD from ASCEND for the subset of randomized patients with complete baseline data (n = 149) was weighted to match aggregate baseline data from all patients randomized to zanubrutinib in ALPINE (n = 327).
Variables identified as prognostic or predictive of investigatorassessed progression-free survival (INV-PFS) in an exploratory multivariate Cox regression analysis of ASCEND were included in the matching.
INV-PFS was assessed before and after matching and pseudo-IPD for INV-PFS for zanubrutinib were obtained from Kaplan-Meier curves using the algorithm by Guyot et al. 5 Sensitivity analyses were performed to assess the robustness of the results from the primary analysis.The first sensitivity analysis included variables that were imbalanced between ALPINE and ASCEND and that were identified as not being prognostic and/or predictive in the exploratory Cox regression analysis.The second sensitivity analysis combined ASCEND and ELEVATE-RR IPD to create a pooled acalabrutinib arm (n = 406) that was matched to zanubrutinib data from ALPINE using the same matching variables as the primary analysis.In a secondary analysis, acalabrutinib IPD from ASCEND was compared with aggregate ibrutinib data from ALPINE (n = 325).
A safety analysis reported Odds ratios (ORs) of adverse events (AEs) in the subset of treated patients with complete baseline data (acalabrutinib, n = 148; zanubrutinib, n = 324).To control for imbalances in the duration of treatment exposure, an artificial data cut-off (February 21, 2020) was imposed for acalabrutinib to match the zanubrutinib median treatment exposure (both 28.4 months).The safety analysis used the same matching variables as the primary efficacy analysis to ensure consistent populations across the analyses.
The following baseline variables were used for matching: sex, ECOG PS, bulky disease, Rai stage, del(11q), del(17p) and/or TP53 mutations, TP53 mutations without del(17p), immunoglobulin heavy chain variable (IGHV) mutation status, geographical region, age (continuous and categorical), number of prior lines of therapy and prior chemoimmunotherapy (Table S2).Weights were estimated for the acalabrutinib population and used to create a re-weighted population whose characteristics included in the matching were identical to those in the zanubrutinib population.The median weight was 0.83 (range: 0.29-3.87)and there were no excessive weights (>10; Figure S1), meaning that none of the patients had an excessive influence on outcomes.After matching, the effective sample size (ESS) of the acalabrutinib arm was 99 (66% of the original efficacy sample).Baseline characteristics pre-and post-matching are reported in Table S3.
Pre-matching, most baseline characteristics were similar across the studies, except for geographical region and beta-2 microglobulin, for which there were notable differences.After matching, there were no differences between variables included in the matching, and the differences between the unmatched variables were reduced.

We can use MAIC to minimize differences between the ASCEND and ALPINE RCT populations
Step with acalabrutinib post-matching versus zanubrutinib (Figure 1).
In summary, this MAIC showed that, when matching on patient baseline characteristics known to be prognostic and/or predictive of INV-PFS, acalabrutinib and zanubrutinib have a similar efficacy in the treatment of R/R CLL when evaluated using INV-PFS.The risk of having AEs was broadly comparable between acalabrutinib and zanubrutinib, except for having an SAE, any grade and grade ≥3 hypertension, any grade hemorrhage, and dose reduction due to AEs, which were lower with acalabrutinib.A strength of our study was that it followed the published National Institute for Health and Care Excellence guidance on MAIC methodology (DSU TSD 18). 1,6ALPINE and ASCEND also had very similar inclusion criteria, study design, and patient characteristics.This was highlighted by the fact that the matching led to a small reduction in ESS and there was little difference between the acalabrutinib results before and after matching.
This study has limitations that are inherent to the methodology and specific to this analysis.The unanchored MAIC methodology makes strong and untestable assumptions, and it is not possible to determine the extent of bias.Despite matching observed patient variables that are prognostic and/or predictive at baseline, unobserved variables or variables reported by only one study cannot be controlled via MAIC.Variables that could not be matched were complex karyotype (due to >30% missing data in ALPINE) and R/R status (whether patients had relapsed or were refractory to treatment, which was not reported in ASCEND).

Complex karyotype was not prognostic or predictive in the exploratory
Cox regression analysis using ASCEND data, so not including it in the matching should have had minimal impact on the results, while R/R status could not be evaluated, making the impact of this omission unclear.In addition, the MAIC was unable to adjust for differences in the period when ASCEND and ALPINE were conducted (enrollment: 2017-2018 vs. 2018-2020), which may have affected outcomes.
We were also unable to adjust for the impact of COVID-19 infections on INV-PFS.We do not have IPD for ALPINE, therefore, it is unclear if the MAIC results would be different if patients were matched in the zanubrutinib arm of ALPINE to the acalabrutinib arm in ASCEND.However, given the similarity between the populations, this would be unlikely to yield different conclusions.Finally, AEs not reported in ALPINE, potentially because they were not common with zanubrutinib (e.g., headache 3 ), were not evaluated.
The limitations of MAIC analyses mean these results should be seen as hypothesis-generating, and ideally require testing in prospective clinical trials.Further research is needed to consolidate our understanding of the comparative efficacy and safety of next-generation BTKis in the treatment of R/R CLL.

leading to dose reduction AE leading to treatment interruption
F I G U R E 1 Overview of the matching-adjusted indirect comparison (MAIC).Odds ratios in bold with an asterisk are statistically significant.AE, adverse event; CI, confidence interval; ECOG PS, Eastern Cooperative Oncology Group Performance Status; ESS, effective sample size; IGHV, immunoglobulin heavy chain variable; INV-PFS, investigator-assessed progression-free survival; IPD, individual patient-level data;