Letter to the Editor: Matching-adjusted indirect comparison between garadacimab and lanadelumab for the long-term prophylactic treatment of patients with hereditary angioedema

We read with interest the recent article titled “Matching-adjusted indirect comparison between garadacimab and lanadelumab for the long-term prophylactic treatment of patients with hereditary angioedema” [1]. While indirect treatment comparisons such as matching-adjusted indirect comparisons (MAICs) provide valuable evidence in the absence of head-to-head clinical trials, the robustness of findings depends on adherence to core methodological principles of the comparison. We have identified several areas where we believe these methodological requirements may not be fully met, potentially affecting the validity of the conclusions, and we identify other aspects that limit the relevance of the findings to clinical practice.

Walsh et al. compared five outcomes relating to efficacy and quality of life (QoL) between garadacimab (200 mg once monthly) and two lanadelumab dosing intervals (300 mg every 2 weeks [Q2W] or every 4 weeks [Q4W]) as long-term prophylaxis in patients aged ≥12 years with hereditary angioedema (HAE) [1]. The MAIC used data from participants enrolled in the 6-month Phase III VANGUARD Study [2] and a 12-week Phase II study [3] of garadacimab (with data pooled across both), and from participants enrolled in the 6-month Phase III HELP Study [4] of lanadelumab. First, we wish to explain that because of fundamental differences between the HELP Study and the real-world use of the Q4W regimen, the indirect comparison between garadacimab and lanadelumab Q4W has limited clinical relevance. We highlight this because, for the primary end point (time-normalized number of HAE attacks), the authors found no difference between garadacimab and lanadelumab Q2W, but reported a statistically significant improvement with garadacimab compared with lanadelumab Q4W.

At initiation of the HELP Study [4], participants were randomly assigned to receive lanadelumab 150 mg Q4W, lanadelumab 300 mg Q2W, lanadelumab 300 mg Q4W or placebo. There were no patients who initiated lanadelumab 300 mg on a Q2W regimen and moved to Q4W. All three lanadelumab regimens resulted in significant reductions in HAE attack rates versus placebo, with the 300 mg Q2W regimen resulting in the numerically largest reduction. Thus, lanadelumab was approved with a starting dose of 300 mg Q2W with a dose-interval extension to 300 mg Q4W allowed in patients who are ‘stably attack free’ or ‘well controlled (e.g., attack free)’ on Q2W dosing [5,6]. Consequently, efficacy and QoL outcomes in study participants randomized directly to lanadelumab Q4W in HELP likely do not reflect outcomes in patients treated with lanadelumab Q4W in clinical practice, who have started on Q2W dosing. Indeed, the independent Expert Advisory Group of the UK National Institute for Health and Care Excellence (NICE) stated it is appropriate to consider equal efficacy between lanadelumab 300 mg Q4W and Q2W in clinical practice [7], rather than use Q4W data from HELP, which shows a numerically smaller attack rate reduction versus the Q2W regimen. In summary, lanadelumab 300 mg Q4W outcome data from HELP do not provide an appropriate substitute for real-world Q4W outcomes, undermining the validity of the indirect comparison between garadacimab and lanadelumab 300 mg Q4W described by Walsh et al. Therefore, for the remainder of our commentary, we will focus on the results and interpretation of the indirect comparison between garadacimab and lanadelumab Q2W.

MAICs use individual patient-level data from the trial(s) of one treatment to match the aggregated baseline characteristics of patients in the trial(s) of another treatment by weighting individual patient data based on treatment effect modifiers [8]. Bodies such as the NICE Decision Support Unit have developed best practices for MAIC reporting, including presentation of the distribution of weights (and reporting the number of individuals assigned zero weight) [9]. However, this is not reported by Walsh et al., limiting the understanding of the calculation of the effective sample size.

Although other types of heterogeneity remain, the goal of a MAIC is to balance the heterogeneity between different study populations. The garadacimab Phase II study did not include any patients in the 12 to <18 age group. Methodologically, while the MAIC matched the VANGUARD and HELP study populations according to the proportions of patients aged <40 years, this approach was not consistent with the more sophisticated matching using categories of 12 to <18, 18 to <40, 40 to <65 and ≥65 years conducted in the recent MAIC of garadacimab and donidalorsen conducted by the same author group [10]. Inconsistency in approaches raises questions. The primary HELP publication provided the age distributions of 12 to <18, 18 to <65, and ≥65 years [4], and the subgroup analysis further provided the 18 to <40 and 40 to <65-years distributions [11]. Using more detailed categorization, particularly both 12 to <18 and 18 to <40-years age groups, would have provided more comparable populations, particularly given different trial age eligibility criteria in the garadacimab Phase II study.

The rationale for the authors' inclusion of the 12-week Phase II garadacimab study and pooling data with those from the Phase III VANGUARD study is unclear, beyond increasing the initial sample size to mitigate reductions in the effective sample size after adjusting for effect modifiers. Pooling garadacimab data from 12-week and 26-week trials inevitably introduces unnecessary heterogeneity when making a comparison with the exclusively 26-week data from the lanadelumab study. Calculations by Walsh and colleagues would have benefitted substantially from a sensitivity analysis based on a MAIC including only Phase III data.

After weighting the garadacimab pooled Phase II and III data to match the HELP Study population characteristics for age <40 years, sex, body weight <75 kg and HAE attack rate during run-in, Walsh et al. reported no statistical differences between garadacimab and lanadelumab Q2W for the primary outcome of time-normalized number of HAE attacks [1]. The MAIC also showed no significant differences between garadacimab and lanadelumab Q2W for the outcomes of time-normalized number of HAE attacks that required on-demand treatment or proportion of attack-free patients over the trial period. Of note, assumptions, and reliance upon them, differ between anchored and unanchored comparisons when trial follow-up is not the same. In the unanchored case, used for the proportion of attack-free patients and where there is no ‘placebo-anchoring’, the differences between the studies are assumed to affect both the treatment and the anchor arm similarly, preserving the relative effect. In an unanchored analysis, differences between the arms that could impact outcomes are indistinguishable from the treatment effect. Traditional survival models resulting in calculation of hazard ratios can produce biased or uninterpretable results if the trial duration is too short to be consistent with longer follow-up. Importantly, the MAIC method requires assumption of proportional hazards [12], which is not necessarily valid for all time-to-event trials. This is particularly relevant in studying long-term prophylaxis because HAE is characterized by unpredictable attacks rather than a continuous disease state. For the binary end point of the proportion of attack-free patients, the authors generated simulated individual patient data for lanadelumab. A clearer justification for this, and details about simulation methodology, assumptions, distributions and covariates are not provided, introducing additional uncertainty and preventing independent validation.

Although the authors did report statistically significant differences in favor of garadacimab over lanadelumab Q2W in time-normalized number of moderate and/or severe HAE attacks and change in Angioedema QoL Questionnaire (AE-QoL), we believe there are fundamental issues with the analyses for both. In VANGUARD, with an attack that was categorized as mild, on-demand medication to treat the attack may not have been necessary, but was not prohibited according to the definition. Whereas, with a mild attack in the HELP Study (transient or with mild discomfort), no medical intervention/therapy was required. Therefore, an attack of the same severity could be classified as mild in VANGUARD but moderate in HELP. Furthermore, while the MAIC matched populations from VANGUARD to HELP with respect to age, sex, bodyweight and baseline attack rate, other critical treatment effect modifiers were not included. Overall, 16.5% of participants in the HELP Study had a medical history of depression and 13.8% had a history of anxiety [13], higher than rates for both in the garadacimab trials, highlighting the importance of further population matching. Additionally, the participants in HELP had a mean baseline AE-QoL score of 43.8–48.8 [14] versus 38.8 [2] in garadacimab-treated patients in VANGUARD, indicating greater QoL impairment in participants enrolled in HELP. However, baseline AE-QoL was not included as a treatment effect modifier, potentially biasing the comparative AE-QoL conclusions. This is noteworthy as baseline AE-QoL scores have been found to be predictive of change in AE-QoL scores at Week 24 [15].

The authors state that garadacimab may provide improved therapeutic benefit compared with lanadelumab as long-term prophylaxis in patients with HAE. However, the 2025/2026 guidelines recommend both lanadelumab and garadacimab as first-line HAE LTP options and the general consensus is of equivalent effectiveness [16–19]. This reinforces the challenges in the clinical interpretation of indirect treatment comparisons in general. Given the methodological limitations raised, particularly the limited clinical relevance of lanadelumab Q4W comparisons, non-comparable treatment durations, outcome definition heterogeneity, missing adjustments for baseline QoL, and use of simulated data without transparency, we feel that it may be misleading to summarise the MAIC as showing ‘improved efficacy and QoL with garadacimab compared with lanadelumab across multiple end points’. The validity of comparative effectiveness conclusions, especially for secondary outcomes, has substantial uncertainty and should be interpreted with appropriate caution.