Matching‐adjusted indirect comparisons of efficacy and safety for zanubrutinib versus the combination of fixed‐duration venetoclax and ibrutinib in patients with treatment‐naïve chronic lymphocytic leukaemia

Chronic lymphocytic leukaemia (CLL) is the most common leukaemia in adults, with an incidence of 5 per 100 000 people per year in Europe.1 Novel agents, such as Bruton tyrosine kinase inhibitors (BTKis) and B-cell lymphoma 2 inhibitors (BCL2is), have changed the therapeutic landscape for CLL/small lymphocytic lymphoma (SLL).2 A recent network meta-analysis from 10 randomized controlled trials (RCTs) in treatment-naïve (TN) CLL found zanubrutinib (next-generation continuous BTKi) to have the most favourable safety profile. The combination of ibrutinib (BTKi) plus venetoclax (BCL2i; V + I) as a fixed-duration regimen was associated with the highest risk of adverse events (AEs) leading to discontinuation and a higher risk of any AE than with zanubrutinib (comparisons were not adjusted for drug exposure).3 This study conducted a matching-adjusted indirect comparison (MAIC) to assess the efficacy and safety of zanubrutinib compared with V + I in the absence of head-to-head studies. We utilized patient-level data (PLD) from SEQUOIA (NCT03336333), a phase 3 RCT evaluating the efficacy and safety of zanubrutinib in TN patients with CLL/SLL. SEQUOIA comprised Cohort 1 (479 patients without deletion of the short arm of chromosome 17 [del17p] randomized 1:1 to receive zanubrutinib [Arm A; n = 241] or bendamustine and rituximab [BR; Arm B; n = 238]) and Cohort 2 (111 patients with del17p treated with zanubrutinib; Arm C). Two V + I studies were identified as relevant: GLOW and CAPTIVATE. GLOW (NCT03462719)4, 5 was a phase 3 RCT of V + I (n = 106) versus obinutuzumab + chlorambucil (n = 105); it included older TN patients with CLL/SLL and comorbidities and excluded patients with del17p or TP53 mutation. CAPTIVATE (NCT02910583)6, 7 was a phase 2 single-arm study investigating V + I (n = 159); it included younger and fitter TN patients with CLL without restrictions on del17p and/or TP53 mutation. The MAICs assessed the comparative efficacy and safety of zanubrutinib and V + I. The MAIC versus GLOW utilized PLD from SEQUOIA Cohort 1 (n = 241; no del17p), with a median follow-up of 43.7 months and a median treatment duration of 43.3 months; the median follow-up for GLOW was 27.7 months. The MAIC versus CAPTIVATE utilized PLD from SEQUOIA Cohorts 1 and 2 (n = 352), with a median follow-up of 45.0 months and a median treatment duration of 44.1 months; the median follow-up for CAPTIVATE was 49.8 months. The median treatment duration for V + I was 13.8 months in both GLOW and CAPTIVATE, reflecting fixed-duration treatment and a substantially shorter exposure than that of zanubrutinib in SEQUOIA. MAIC analyses were conducted in accordance with the UK's National Institute for Health and Care Excellence Decision Support Unit (NICE DSU),8 using methods described by Signorovitch et al.9 First, weights were derived for patients receiving zanubrutinib to match the preselected set of population characteristics of the comparator population. Second, weighted Cox proportional hazards regression was used to estimate the relative treatment effect in a comparator-like population. Weights were also used to calculate the effective sample size (ESS) achieved after re-weighting patients. Unanchored MAICs were performed as there was no common control arm between studies. A feasibility assessment found differences in eligibility criteria (e.g. exclusion of patients with del17p or TP53 mutation) that led to significant imbalances in certain population characteristics. Population characteristics for matching were selected based on availability, clinical relevance and degree of imbalance. The unanchored MAIC aimed to adjust for imbalances that are treatment effect modifiers or prognostic factors while preserving the ESS. Factors not selected for matching in the base-case analysis were explored in alternative matching sensitivity analyses. Efficacy outcomes included investigator-assessed progression-free survival (PFS-INV); overall survival (OS) data were deemed immature for the MAIC and were therefore not investigated. The primary safety comparison was conducted without applying weights, as weights were based on efficacy factors. Simple logistic regression was used to estimate the relative effect on treatment-emergent AEs. As part of the sensitivity analysis, weighted logistic regression was conducted using weights from the efficacy comparison. Odds ratios (ORs) derived from logistic regression, along with 95% confidence intervals (CIs), were reported for AEs where the naïve comparison yielded p > 0.2. Data analyses were conducted using R 4.0.2; the R code of NICE DSU Technical Support Document 18 appendix D8 was used to derive the MAIC weights. In the base-case analysis between SEQUOIA and GLOW, populations were adjusted for factors listed in Table S1. The inclusion of the Cumulative Illness Rating Scale (CIRS) as a matching factor in the base-case analysis led to a drop in ESS from 152 to 85 due to high imbalance in the proportion of patients with CIRS >6; hence, it was explored as part of the sensitivity analyses. In the base-case analysis between SEQUOIA and CAPTIVATE, populations were adjusted for the factors shown in Table S2. While no adjustment was made for age and proportion with bulky disease, these characteristics were reasonably well balanced after matching for other characteristics, so they were not included in matching to avoid a further drop in ESS. Complex karyotype had a high proportion of missing values (28%) and thus could not be added to the base-case analysis without a substantial drop in ESS. Additional characteristics (e.g. beta2-microglobulin, geographic region, time from diagnosis, creatinine clearance level, CIRS) were not available in CAPTIVATE. The impact of population matching on PFS-INV in the base-case analysis is illustrated in Figure 1. For the comparison of zanubrutinib versus V + I in GLOW, the hazard ratio (HR) was 0.71 (95% CI: 0.39–1.28) before matching and 0.84 (95% CI: 0.45–1.59) after matching. The comparison of zanubrutinib versus V + I in CAPTIVATE showed a potential trend in favour of zanubrutinib for PFS-INV: the HR was 1.00 (95% CI: 0.65–1.52) before matching and 0.69 (95% CI: 0.34–1.40) after matching. Changes in HRs were not statistically significant for either comparison. Sensitivity analyses were consistent with those of the base-case analysis. When including the proportion of patients with CIRS >6 in the comparison versus GLOW, the HR for PFS-INV was 0.72 (95% CI: 0.34–1.52). In a simplified matching scenario versus CAPTIVATE, sex, Eastern Cooperative Oncology Group performance status, trisomy 12, del13q and TP53 mutation were excluded, which increased the ESS from 59 to 90; the resulting HR for PFS-INV was 0.67 (95% CI: 0.35–1.28). A second sensitivity analysis also included complex karyotype, with the ESS decreasing to 50, resulting in a PFS-INV HR of 0.49 (95% CI: 0.21–1.13). The unadjusted comparison of rates of grade 3+ AEs in the zanubrutinib arm of SEQUOIA and the V + I arms of GLOW and CAPTIVATE revealed a lower incidence of AEs for zanubrutinib. The unadjusted OR estimates of preferred term AEs are presented in Figure S1. The upper bound of the 95% CI was <1 for diarrhoea, neutropenia, nausea, anaemia, atrial fibrillation, decreased appetite and arthralgia, indicating robust evidence of a substantial advantage for zanubrutinib. A numerically higher incidence of cough was found for zanubrutinib; however, the 95% CI was wide and crossed 1, indicating no robust evidence in favour of V + I. As shown in Figure 2, sensitivity analyses using matching weights from the base-case analysis for efficacy led to consistent results, with the exception of arthralgia, for which the upper bound of the 95% CI was >1. To our knowledge, no head-to-head studies have compared the efficacy and safety of zanubrutinib with V + I in TN CLL. The unanchored MAIC presented here indicates comparable treatment effects with a trend in favour of zanubrutinib. Furthermore, a notably better safety profile was observed for zanubrutinib than for V + I. Although no quality-of-life outcomes were available for comparison, we anticipate a positive impact from zanubrutinib based on safety results, an aspect that may influence treatment decisions of clinicians and patients for TN CLL. After adjusting for population characteristics for PFS-INV, the HR estimates favoured zanubrutinib numerically regardless of the choice of comparator study or matching set. The relatively small sample size in the V + I arms and the notable imbalance in certain key characteristics across the populations led to substantial uncertainty in the estimation of HRs and contributed to the lack of significant differences in treatment effects. When comparing against CAPTIVATE, numerical benefit in favour of zanubrutinib was observed from 3 years onwards; further follow-up is needed to confirm the observed benefit. Furthermore, zanubrutinib showed a favourable safety profile versus V + I despite the longer treatment exposure in SEQUOIA versus V + I as a fixed-duration regimen. It should also be noted that AEs in CAPTIVATE were reported only if the incidence was >30% for any grade or >5% for grade 3+, further supporting the finding of a better safety profile for zanubrutinib. A prospective comparison between these two options is warranted to confirm our findings. This MAIC supports the use of both zanubrutinib and V + I as effective therapeutic options for TN CLL; longer follow-up is needed to observe any differences in PFS-INV. The observed safety differences between zanubrutinib and V + I suggest that zanubrutinib may have an impact on a patient's quality of life due to AEs, which may be considered at the time of treatment decision-making in TN CLL. Tahla Munir, Leyla Mohseninejad, Balazs Dobi, Sheng Xu, Remus Vezan and Nicolás Martinez Calle contributed to the conception and design of the study. Leyla Mohseninejad, Pal Rakonczai, Balazs Dobi and Sheng Xu analysed the data. Tahla Munir, Leyla Mohseninejad, Balazs Dobi, Keri Yang, Remus Vezan and Nicolás Martinez Calle interpreted the data. All authors contributed to the writing and final approval of this article. Editorial support was provided by Amiculum, funded by BeOne Medicines Ltd. This study was sponsored by BeOne Medicines Ltd. Tahla Munir has received research grants from AbbVie and Janssen; has received consulting fees from AbbVie, BeOne Medicines, Sobi, Alexion, Novartis, Janssen, AstraZeneca, Eli Lilly and Company and Roche; has received honoraria from BeOne Medicines, AstraZeneca, Sobi, Roche, Janssen, AbbVie and Eli Lilly and Company; has received travel support from Alexion, BeOne Medicines, AbbVie, Janssen and AstraZeneca; and has participated in advisory boards for AbbVie, BeOne Medicines, AstraZeneca and Janssen. Leyla Mohseninejad, Sheng Xu and Keri Yang are employees of and stockholders in BeOne Medicines Ltd. Balazs Dobi and Pal Rakonczai are employees of Evidera, which was hired by the sponsor, BeOne Medicines Ltd, to conduct this research. Remus Vezan is an employee of and stockholder in BeOne Medicines Ltd; is a stockholder in AbbVie and Gilead; and had a leadership role for which they were compensated at Carina Bio. Nicolás Martinez Calle has participated in advisory boards for AbbVie, AstraZeneca, BeOne Medicines, CSL Behring and Takeda; and has received speaker fees/honoraria from AbbVie, AstraZeneca, BeOne Medicines and Janssen. The studies included here were conducted in accordance with the good clinical practice guidelines and the Declaration of Helsinki. Study protocols were approved by the institutional review boards of participating study centres, and all participants provided written informed consent. SEQUOIA, ClinicalTrials.gov NCT03336333; GLOW, ClinicalTrials.gov NCT03462719; CAPTIVATE, ClinicalTrials.gov NCT02910583. On request, and subject to certain criteria, conditions and exceptions, BeOne Medicines Ltd will provide access to individual de-identified participant data from BeOne-sponsored global interventional clinical studies conducted (1) for indications that have been approved based on the BeOne data sharing policy or (2) in programmes that have been terminated. BeOne shares data only when permitted by applicable data privacy and security laws and regulations, shares when it is feasible to do so without compromising the privacy of the study participants and other considerations. Data requests may be submitted to [email protected]. Data S1. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.