Assembling the evidence…
Factor XII (FXII) is a serine protease involved in multiple cascades, including the kallikrein-kinin system. It may play a role in diseases in which the downstream cascades are dysregulated, such as hereditary angioedema. Garadacimab (CSL312) is a first-in-class, fully human, monoclonal antibody targeting activated FXII (FXIIa). We describe how translational pharmacokinetic (PK) and pharmacodynamic (PD) modeling enabled dose selection for the phase I, first-in-human trial of garadacimab. The PK/PD data used for modeling were derived from preclinical PK/PD and safety studies. Garadacimab plasma concentrations rose with increasing dose, and clear dose-related PD effects were observed (e.g., a mechanism-based prolongation of activated partial thromboplastin time). The PK/PD profile from cynomolgus monkeys was used to generate minimal physiologically-based pharmacokinetic (mPBPK) models with target-mediated drug disposition (TMDD) for data prediction in cynomolgus monkeys. These models were later adapted for prediction of human data to establish dose selection. Based on the final mPBPK model with TMDD and assuming a weight of 70 kg for an adult human, a minimal inhibition (<10%) of FXIIa with a starting dose of 0.1 mg/kg garadacimab and a near maximal inhibition (>95%) at 10 mg/kg garadacimab were predicted. The phase I study is complete, and data on exposure profiles and inhibition of FXIIa-mediated kallikrein activity observed in the trial support and validate these simulations. This emphasizes the utility and relevance of translational modeling and simulation in drug development.
The phase I study is complete, and data on exposure profiles and inhibition of FXIIa-mediated kallikrein activity observed in the trial support and validate these simulations.
broad validation language despite terminal-PK and lower-dose-PD discrepancies in an unpublished clinical dataset
The predicted PK parameters were within two-fold of the observed values, which is the industry standard.
standard PK metrics agree within two-fold, but exact comparisons and uncertainty rely on unpublished human data
A dose of 0.1 mg/kg i.v. was predicted to cause slight (<10%) inhibition of FXIIa-mediated kallikrein activity; 10 mg/kg i.v. was predicted to cause near maximal (>95%) inhibition.
the claim accurately reports the model-generated dose-response predictions displayed in Figure 4
Inhibition of FXIIa-mediated kallikrein activity in the first-in-human study was predicted reasonably well, with the exception of the lower doses where the model underpredicted the inhibition.
the wording directly acknowledges the lower-dose discrepancy, though comparison rests on unpublished clinical observations
Garadacimab plasma concentrations rose with increasing dose, and clear dose-related PD effects were observed (e.g., a mechanism-based prolongation of activated partial thromboplastin time).
dose-related preclinical PK and PD patterns are reported across the cynomolgus studies
Derived from the full evaluation — not a separate score.
Strengths
Predictions derived from monkey data were used to select the phase I dose range and were subsequently compared with observed human PK/PD, reducing the risk of post-hoc fitting to clinical outcomes.
↳ Methods, First-in-Human Dose Selection; Results, Figures 4–5
The paper presents the mPBPK and TMDD equations, variable definitions, fixed physiological parameters, allometric scaling, and software versions in a stepwise framework.
↳ Methods, Steps I–III, Model Prediction in Humans, and Software
Results identify a faster predicted terminal PK decline and underprediction of inhibition at lower doses rather than presenting only the favorable Cmax and AUC comparisons.
↳ Results, Figure 4a–b and Figure 5
Limitations
The central validation claim relies on phase I results designated as unpublished and deferred to a separate report, limiting independent inspection of observed values and predictive uncertainty.
↳ Methods, Comparison of Predicted versus Observed Garadacimab PK in Humans; Results, Figures 4–5
The Discussion states that a two-compartment model fit the observed data equally well but does not show its development or results, weakening transparency around model selection.
↳ Discussion, paragraph beginning “Initially, both the mPBPK model”
The broad claim that the human observations validate the simulations is not fully aligned with underprediction of lower-dose and overall PD inhibition and the terminal-phase PK mismatch.
↳ Abstract; Results, Figure 4; Discussion, final paragraph
The principal contribution is the prospective use of a monkey-derived mPBPK/TMDD model to select doses later administered in a completed first-in-human study. Methods provide a detailed stepwise model and the Results report standard Cmax and AUC agreement, but the equally fitting two-compartment model is not shown and clinical predictive uncertainty is limited. Figure 4 also documents faster predicted terminal decline and underprediction of lower-dose PD inhibition, making the broad validation language stronger than the complete result. The single-compound, single-dose evidence supports practical relevance for early development more strongly than broad methodological transferability.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Sound3.4
Confidence mediumThe score reflects a meaningful applied contribution: monkey-derived predictions informed a completed first-in-human dose range and were subsequently compared with clinical observations. The advance remains incremental because mPBPK, TMDD, allometric scaling, and MABEL are established methods applied to one antibody.
“The predicted PK parameters were within two-fold of the observed values”
The model equations, stepwise fitting sequence, assumptions, software, and cross-species scaling are reported, and the evaluation uses predictions generated before human outcomes. Methodological Rigour is constrained by the unshown two-compartment comparison, limited uncertainty reporting for clinical validation, and reliance on unpublished human data.
“model building and results of the two-compartment model not shown”
The preclinical-data-to-model-to-human-prediction sequence is easy to follow, with equations and parameter roles organized into three explicit steps. The broad validation wording is stronger than the mixed PK/PD performance reported in Results.
“support and validate these simulations”
The paper connects its approach to relevant mPBPK, TMDD, MABEL, and translational-modeling literature and names several study limitations. It does not fully trace the consequences of single-dose data, absent placebo baselines, normalization assumptions, and ex vivo readouts into the strength of its conclusion.
“does not take baseline variability into account”
Caveats4 of 4 checks
The reported modeling sequence, dose ranges, exposure margins, and prediction-versus-observation narrative are internally consistent. Discrepancies in terminal PK decline and lower-dose PD inhibition are disclosed rather than contradicted elsewhere.
Animal ethics and conflicts of interest are disclosed, but the human validation dataset is described as unpublished and no human ethics or consent declaration appears for the phase I study. This limits independent verification and constitutes a reporting concern without showing that approval or consent was absent.
Flags: 1 declared / 5 total
29 of 29 checkable references verified
32 references in manuscript 3 are books, websites or datasets — counted, but not index-checkable
No retraction notice found in Retraction Watch.
Sources: Retraction Watch ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
The model moved beyond preclinical simulation by informing doses administered in a completed healthy-volunteer trial and being compared with observed human PK/PD. It remains an early clinical-development application rather than deployable treatment guidance.
“the first-in-human, phase I study has been completed”
AI-generated, human-governed. Something look off? Contact us to request a review.