A scientific dispute over how two modern preventive drugs for hereditary angioedema compare against each other has spilled into print, with the authors of a network meta-analysis issuing a detailed rebuttal to critics who argued the analysis overstated the case for garadacimab over lanadelumab. The responding team, led by Imtiaz A. Samjoo, addresses four distinct criticisms raised in a letter to the editor by Andresen and colleagues, who had titled their critique with the claim that treatment outcomes with lanadelumab every two weeks and garadacimab are very similar in patients with hereditary angioedema.
The first point of contention concerns dosing. The critics questioned the inclusion of a every-four-weeks, or Q4W, dosing regimen for lanadelumab, arguing that such dosing does not reflect clinical practice or regulatory guidance. In their response, the analysts defend the inclusion on the grounds of what a network meta-analysis is designed to do: synthesize randomized evidence while preserving internal validity. They argue that randomized controlled trials remain the gold standard for indirect comparisons because randomization minimizes bias and confounding, and that health technology assessment agencies worldwide prefer randomized evidence precisely for these reasons. Excluding evidence simply because it does not mirror everyday practice, they contend, would introduce selective interpretation and undermine the scientific purpose of evidence synthesis. They add an important factual clarification: the only randomized evidence for Q4W dosing comes from the HELP study, where it was tested as a starting dose, while the only maintenance dose studied in clinical trials for lanadelumab is every two weeks, examined in HELP-OLE.
The second criticism targeted the combination of studies with different follow-up durations. Andresen and colleagues suggested that pooling data from a 12-week phase II study and a six-month phase III trial favored garadacimab on endpoints that were not normalized for time. The responding authors call this critique unclear, explaining that their statistical model uses time offsets to normalize exposure in rate models based on Poisson and negative binomial distributions, and in cumulative risk models using a cloglog link with a log-exposure offset. This, they note, is a standard approach that adjusts for differential follow-up times between trials. Still, they concede the underlying point has merit: homogeneous follow-up times matter, and they state that they made all efforts possible to minimize this potential bias in their analysis.
The third dispute is more granular and turns on the precise wording of a clinical trial definition. The critics asserted that the HELP trial defined mild attacks as those requiring no medical intervention. The responding authors say this was a misquotation. The actual definition, they state, was transient or mild discomfort lasting fewer than 48 hours, with no medical intervention or therapy required. The distinction matters because, in clinical research, therapy typically refers to disease-specific medication used to treat hereditary angioedema attacks, such as intravenous C1-inhibitors or bradykinin B2 receptor antagonists. Analgesics, by contrast, are not considered disease-specific therapy; they relieve pain but do not treat the underlying condition. In hereditary angioedema trials, analgesics are usually categorized as concomitant medication, a category the authors note is often under-reported in clinical studies.
On the broader question of attack severity definitions, the authors acknowledge that minor differences exist between the HELP and VANGUARD trials in how mild and moderate attacks were defined, and that such differences may contribute to cross-trial heterogeneity. But they argue this is not unusual when synthesizing clinical data from several trials and does not necessarily invalidate the evidence synthesis, particularly given what they describe as the limited clinical impact of these definitional differences. They further contend that, given the size of the treatment effects favoring garadacimab over lanadelumab at both Q2W and Q4W dosing for the time-normalized number of moderate or severe attacks, their conclusions are robust to this type of heterogeneity.
The fourth and most technically involved criticism concerns the choice between fixed-effect and random-effects models, a perennial fault line in meta-analysis. The responding authors defend their use of a Bayesian framework as appropriate for representing uncertainty in sparse networks with limited information, where frequentist approaches carry their own limitations, including unstable estimators of heterogeneity and questionable reliance on asymptotic approximations, along with a degree of subjectivity embedded in estimator assumptions. The critics had noted that random-effects models should have been used where fit statistics favored them, and had highlighted how statistically significant findings differed between fixed-effects and random-effects specifications.
In response, the authors attribute the apparent disagreement between models to well-known challenges in Bayesian random-effects meta-analysis when vague priors are used for between-trial heterogeneity. Vague priors are chosen to let the data drive the results, but in sparse networks they can instead dominate the posterior distributions, producing unstable estimates. The authors report that, in their supplementary tables, the between-trial standard deviation estimates were nearly identical to the vague prior they used, a uniform distribution on the interval from zero to five, which they say implies implausible uncertainty. The consequence, they argue, was posterior distributions with limited clinical interpretability, including rate ratios ranging from three to more than 75,000. On this reading, the wide intervals reported in the supplementary random-effects model might reflect prior-driven uncertainty inflation rather than genuine evidence of large heterogeneity.
The authors bolster this argument with empirical context. Research on heterogeneity parameters in randomized controlled trials suggests these tend to be small, they note, and informative priors could provide a more plausible representation of between-study variability than very diffuse ones. When such empirically grounded priors are used, random-effects estimates typically align more closely with fixed-effect results, which suggests, in their view, that the conclusions of their manuscript may be less sensitive to model choice than the critics implied through analyses based on vague priors. This is a technical defense with practical stakes: whether one model or the other is deemed appropriate can determine whether a treatment difference appears statistically significant or dissolves into statistical noise.
On treatment ranking, the authors strike a conciliatory note. They agree that SUCRA values and p-best statistics, common tools for ranking treatments in network meta-analysis, should not be interpreted in isolation. Rankings, they explain, simplify reporting by ordering treatments and are less affected by variations in baseline risk, but they should complement, not replace, effect size estimates in clinical decision making. Notably, they acknowledge that the observation underlying the critics’ letter, that mean monthly attack rates for garadacimab and lanadelumab every two weeks are similar, aligns with their own interpretation. They add a caution, however: the analyses reported in their paper confirm that cross-trial comparisons should not be made naively.
The response closes with a broad defense of the work. The authors describe their manuscript as a rigorous synthesis of randomized evidence for long-term prophylaxis in hereditary angioedema, state that their network meta-analysis methodology has been externally validated by experts and health technology assessment agencies, and assert that the findings remain robust under reasonable assumptions and align with independent analyses. Readers should keep the nature of this document firmly in view: it is a commentary and author response, not a new clinical study, and it reports no new patient data. Its claims about robustness are the analysts’ own assessments of their previously published modeling work, offered in rebuttal to a critical letter.
Subject of Research: A methodological defense of a network meta-analysis comparing lanadelumab and garadacimab for hereditary angioedema prophylaxis
Article Title: Response to “Treatment Outcomes with Lanadelumab Every Two Weeks and Garadacimab Are Very Similar in Patients with Hereditary Angioedema”
Article References: Samjoo, I. A. (2026). Response to “Treatment Outcomes with Lanadelumab Every Two Weeks and Garadacimab Are Very Similar in Patients with Hereditary Angioedema”. Drugs in R&D, 26(1), 35-37. https://doi.org/10.1007/s40268-026-00536-x
Image Credits: AI Generated
DOI: 10.1007/s40268-026-00536-x
Keywords: hereditary angioedema, lanadelumab, garadacimab, network meta-analysis, Bayesian statistics, randomized controlled trials, heterogeneity, attack severity, prophylaxis, methodology, commentary, Drugs in R&D
News Source: Ophelia Keating. (October 29, 2026). Researchers Defend Network Meta-Analysis Comparing Hereditary Angioedema Drugs. Scienmag.
