Exposure–Response Analysis of Donidalorsen for the Treatment of Hereditary Angioedema

ABSTRACT

Hereditary angioedema (HAE) is characterized by recurrent attacks of severe tissue swelling. In the OASIS‐HAE phase 3 study (NCT05139810), donidalorsen, an RNA‐targeted antisense oligonucleotide that reduces prekallikrein production in the liver, significantly reduced HAE attack rates. To characterize the relationship between prekallikrein concentrations and HAE attack rates following donidalorsen and predict the efficacy of potential dosing regimens, an exposure–response model was developed using data from OASIS‐HAE. Simulations were conducted to evaluate the following regimens: 80 mg once every 4 weeks (Q4W), 8 weeks (Q8W), 1 month (Q1M), and 2 months (Q2M), and a switch to Q2M dosing for patients who were attack‐free for 3 months on the Q1M regimen. The relationship between prekallikrein concentrations and HAE attack rates was well characterized by a sigmoidal E_max (maximum effect) model with baseline attack rate and baseline prekallikrein concentration included as covariates. The average prekallikrein concentration estimated to result in a 90% reduction in attack rate (EC₁₀) was 47.1 mg/L. Predicted percent reductions in attack rates at steady state were similar for Q4W (84.1%) vs. Q1M (82.9%) and Q8W (72.6%) vs. Q2M (70.2%) dosing. Predicted reductions in attack rate remained similar and clinically meaningful in patients who switched from Q1M to Q2M dosing (94.0% in month 1 and 91.3% in month 2 of the 2‐month dosing interval at steady state). Overall, exposure–response analyses supported the efficacy of Q1M and Q2M dosing and indicated that switching to Q2M dosing could be a viable approach for patients who are attack‐free on the Q1M regimen.

Study Highlights.

• What is the current knowledge on this topic?

  • ○
    Hereditary angioedema (HAE) is a rare disease characterized by recurrent episodes of severe tissue swelling.
  • ○
    Donidalorsen, an RNA‐targeted antisense oligonucleotide that specifically reduces plasma prekallikrein production in the liver, significantly reduced HAE attack rates with an acceptable safety profile when administered once every 4 weeks (Q4W) or once every 8 weeks (Q8W) in clinical trials.
• What question did this study address?

  • ○
    This study used exposure–response analyses to quantify the relationship between HAE attack rates and prekallikrein concentrations after treatment with donidalorsen and to predict the efficacy of several potential donidalorsen dosing regimens.
• What does this study add to our knowledge?

  • ○
    The relationship between prekallikrein concentrations and HAE attack rates was well characterized by a sigmoidal E_max model with baseline attack rate and baseline prekallikrein concentration as covariates and with no significant effects of age, body weight, sex, or antidrug antibodies, suggesting a generally similar response to donidalorsen across the patient population.
  • ○
    Simulations conducted using this model indicated that the efficacy of once‐every‐month (Q1M) and once‐every‐2‐months (Q2M) dosing is similar to the efficacy of Q4W and Q8W dosing, respectively. Further, simulated attack rates remained low in patients who switched from Q1M to Q2M dosing if they were attack‐free for ≥ 3 months, indicating patients may benefit from an initial period of more frequent dosing to achieve a rapid clinical response followed by a switch to less frequent dosing for convenience.
• How might this change clinical pharmacology or translational science?

  • ○
    These results describe a quantitative relationship between prekallikrein concentrations and HAE attack rates after donidalorsen treatment and provide support for potential dosing regimens for donidalorsen.

1. Introduction

Hereditary angioedema (HAE) is a rare, chronic disease characterized by severe, unpredictable attacks of tissue swelling [1]. HAE attacks are painful, debilitating, and potentially life‐threatening [1, 2]. As a result, HAE can have a substantial negative impact on patients' quality of life. Patients with HAE experience increased levels of depression and anxiety, and HAE can interfere with their ability to participate in work, school, and other activities of daily living [2, 3, 4]. Several long‐term prophylactic (LTP) treatment options that reduce the number and severity of HAE attacks exist, including once‐daily berotralstat, plasma‐derived C1 inhibitor (C1INH) once every 3 to 4 days, and lanadelumab once every 2 weeks [5, 6, 7]. However, breakthrough attacks still occur, and there remains a need for LTPs that provide a more patient‐friendly experience, including reduced dosing frequency and greater efficacy, safety, and tolerability [8, 9].

HAE is most commonly caused by C1INH protein deficiency (HAE‐C1INH‐Type1) or dysfunction (HAE‐C1INH‐Type2) resulting from mutations of the SERPING1 gene [1, 10, 11, 12]. In the absence of normal C1INH function, the kallikrein–kinin system is dysregulated, leading to increases in plasma prekallikrein and kallikrein that result in increased bradykinin production. Overproduction of bradykinin causes vasodilation and increases vascular permeability, ultimately leading to subcutaneous and submucosal edema [11, 13].

Donidalorsen is a N‐acetyl galactosamine–conjugated antisense oligonucleotide designed to reduce prekallikrein production by selectively binding to prekallikrein messenger RNA in the liver, leading to its subsequent degradation by ribonuclease H1 [14, 15]. Donidalorsen is approved by the US Food and Drug Administration (FDA) for prophylaxis to prevent attacks of HAE in adult and pediatric patients 12 years of age or older [16]. In the phase 3, randomized, placebo‐controlled clinical trial OASIS‐HAE (NCT05139810), donidalorsen 80 mg administered subcutaneously once every 4 weeks (Q4W) or once every 8 weeks (Q8W) for 24 weeks significantly reduced HAE attack rates by 81% and 55%, respectively, and had an acceptable safety and tolerability profile [17]. These reductions in HAE attack rates were associated with corresponding reductions in mean plasma prekallikrein trough concentrations of 73% in the donidalorsen Q4W group and 47% in the donidalorsen Q8W group. Similar results were observed in the phase 2 trial (NCT04030598), in which donidalorsen 80 mg Q4W for 16 weeks reduced the mean HAE attack rate by 90% and mean plasma prekallikrein concentration by 61% [14]. Reductions in HAE attack rates and plasma prekallikrein were correlated in this study. In the open‐label extension of the phase 2 study (NCT04307381), patients could switch from Q4W dosing to Q8W dosing if they experienced no HAE attacks in the first 12 weeks of the study. Per a recent report, attack rate and prekallikrein reductions have been maintained in patients in this study up to the 2‐year data cutoff, and 5/8 patients who switched from Q4W dosing to Q8W dosing remained attack‐free [18]. These data suggest that patients could benefit from a dosing regimen in which they start on Q4W dosing to drive down prekallikrein concentrations and reduce HAE attack rates and then, if attack‐free for at least 3 months while receiving donidalorsen, switch to Q8W dosing to maintain low attack rates at a more convenient dosing frequency.

Here we report the results of an exposure–response analysis using data from the OASIS‐HAE study to characterize the quantitative relationship between HAE attack rate and predicted plasma prekallikrein concentrations following treatment with donidalorsen. The exposure–response model developed based on the OASIS‐HAE study was further validated using data from the phase 2 donidalorsen study. The efficacy of the 80 mg Q4W and Q8W dosing schedules used in OASIS‐HAE was evaluated and compared with dosing regimens of 80 mg once every month (Q1M) and once every 2 months (Q2M), and the effect of switching dosing regimens from Q1M to Q2M to facilitate an initial clinical effect followed by a less frequent maintenance dosing schedule was simulated.

2. Methods

2.1. Study Populations

Patient data from the phase 3 OASIS‐HAE trial were used to perform the exposure–response analysis, and the model was externally validated against data from the phase 2 trial. The analysis population included patients who were included in the pharmacokinetic (PK) and pharmacodynamic (PD) models for whom individual exposure metrics could be calculated. The details of these studies have been described previously [14, 17]. Briefly, OASIS‐HAE was a phase 3, double‐blind, randomized, placebo‐controlled trial in which placebo or 80 mg donidalorsen was administered subcutaneously Q4W or Q8W for 24 weeks. Patients were ≥ 12 years of age with HAE‐C1INH‐Type1 or HAE‐C1INH‐Type2 and had experienced ≥ 2 investigator‐confirmed HAE attacks during the 56‐day screening period. In the phase 2 study, patients ≥ 18 years of age with a documented diagnosis of HAE were randomized to receive donidalorsen 80 mg Q4W or placebo for 16 weeks. These studies were conducted in accordance with ethical principles of the Declaration of Helsinki, Good Clinical Practice, and the International Conference for Harmonization guidelines, and were approved by the relevant institutional review boards. All patients provided written informed consent, or assent if < 18 years of age.

2.2. Exposure–Response Model Development and Evaluation

Population PK and PD models based on multiple clinical studies of donidalorsen have been developed independent of the current analysis [19]. The PK model structure was a standard 2‐compartment disposition model with first‐order absorption and elimination. The PD model structure was an indirect response model with inhibition of prekallikrein production rate by donidalorsen. Goodness‐of‐fit plots and visual predictive checks indicated that the PKPD model adequately fit the observed data with minimal bias. Full details of the model development and results will be published separately (data on file). Four‐week average systemic prekallikrein concentrations (PKK_avg,4W) for each 4‐week period from Weeks 0–4 to Weeks 20–24 were predicted for each individual donidalorsen‐treated patient using the PKPD model. Prekallikrein concentrations for placebo patients were based on the observed data.

A longitudinal exposure–response model with per‐4‐week normalized HAE attack rate as a dependent variable and PKK_avg,4W as an independent variable was developed and evaluated using generalized Poisson regression. Poisson regression was used because attacks are counts, which are by nature discrete, left censored at 0, and highly skewed. Generalized Poisson regression rather than standard Poisson regression was used to account for overdispersion in the data. The relationship between prekallikrein concentration and per‐4‐week HAE attack rate was analyzed using a nonlinear mixed‐effects modeling approach, which accounted for structural (fixed) effects and intra‐/interindividual variability.

Throughout model development, diagnostic plots were used to assess the model's ability to describe the observed data. Structural, statistical, and covariate model components were selected such that the models provided an adequate and unbiased description of the observed data. Alternative models were compared based on overall model fit (objective function value), precision of parameter estimates, condition number, correlations between parameters, random effect, and residual error model variances. A model was discarded if the relative uncertainty of any fixed model parameters was not acceptable, if any parameter estimate appeared implausible, or if any of the diagnostic plots indicated that the model showed substantial bias, was nonidentifiable or numerically unstable, or was unable to describe the trends and variability of the observed data.

The initial model evaluated a linear relationship between PKK_avg,4W and per‐4‐week HAE attack rate. Using an E_max and subsequently a sigmoid E_max relation improved the model fit; random effects for E_max and Hill coefficient were included. This model was utilized to perform an exploratory univariate assessment of the sources of variability in per‐4‐week normalized HAE attack rate response. However, to assess potential covariate effects, a model that included interindividual variability (IIV) on the effective PKK_avg,4W associated with 50% of the maximum effect (EC₅₀) was also evaluated. This model showed a further reduction of objective function value, but it was found to be unstable during the stepwise covariate evaluation. Therefore, the model without IIV on EC₅₀ was used for a formal stepwise covariate evaluation.

Body weight, age, sex, baseline prekallikrein concentration, baseline per‐4‐week HAE attack rate, time, and antidrug antibody (ADA) status, as defined by the presence of treatment‐emergent ADAs, were assessed in the model as potential covariates. Treatment‐emergent ADAs were defined as ADAs that developed any time after the initial drug administration in a patient without preexisting ADAs or a preexisting ADA titer that increased to ≥ 8‐fold greater than baseline. ADA status was included as a time‐independent variable controlling for immunogenicity risk, as this simpler approach is more efficient than including time‐varying ADA status considering the high patient‐to‐patient variability in ADA onset, magnitude of titer response, and that there was no loss of efficacy over time irrespective of the presence of ADAs. Continuous covariates were evaluated in the model as power functions, while binary covariates were evaluated as factors. Covariates were assessed for inclusion in the model by a stepwise forward inclusion/backward elimination strategy based on a log‐likelihood ratio test. Statistical significance of covariates was evaluated using the chi‐square test, with a threshold of p < 0.01 during the forward inclusion phase and p < 0.001 during the backward elimination phase. Model development also considered a time‐dependent placebo effect, which ultimately did not improve fit and was thus not included.

The final model equation that best characterized the relationship between prekallikrein and the per‐4‐week normalized HAE attack rate was a direct‐effect model in which the effect of prekallikrein concentration on the per‐4‐week normalized HAE attack rate was modeled with a sigmoidal E_max (maximum effect) relationship. The final model equation was as follows:

$$\begin{array}{c}\mathit{per}4\mathit{\text{week}}\mathit{HAE}\mathit{\text{attack rate}}=\hfill \\ {\left(\frac{\mathit{\text{BLRATE}}}{3}\right)}^{b_{\mathit{HAE}}}·E_{\mathit{max}}·\left(\frac{{\mathit{PKK}}^{\mathit{\text{Hill}}}}{{\left(\frac{\mathit{\text{BLPKK}}}{122}\right)}^{b_{\mathit{PKK}}}{·\mathit{EC}}_{50}^{\mathit{\text{Hill}}}+{\mathit{PKK}}^{\mathit{\text{Hill}}}}\right)\hfill \end{array}$$

where BLRATE is the baseline per‐4‐week normalized HAE attack rate, b_HAE is the coefficient for the effect of baseline per‐4‐week normalized HAE attack rate on E_max, E_max is the per‐4‐week normalized HAE attack rate at infinite prekallikrein concentration, PKK is the per‐4‐week average prekallikrein concentration, Hill is the Hill coefficient, BLPKK is the baseline prekallikrein concentration, b_PKK is the coefficient for the effect of baseline prekallikrein concentration on EC₅₀, and EC₅₀ is the prekallikrein concentration associated with a 50% reduction from the maximum attack rate. Per the phase 3 OASIS‐HAE study protocol, BLRATE was calculated as the number of investigator‐confirmed HAE attacks during the 56‐day run‐in period divided by the number of days contributed to the run‐in period multiplied by 28 days.

The predictive ability of the model was evaluated using visual predictive checks comparing plots of data distributions from 1000 simulations with observed data distributions. The robustness of the model was evaluated by a nonparametric bootstrap procedure in which parameter estimates obtained with the original dataset were compared with the median and 95% confidence interval of estimates derived from a bootstrap based on 1000 data replicates. External model validation was performed using patient data from the phase 2 study [14]. The model, together with individual dosing history and individual covariates from the phase 2 study, was used to generate predicted per‐4‐week normalized HAE attack rates, which were plotted against the observed phase 2 data.

2.3. Exposure–Response Simulations

Exposure–response simulations of per‐4‐week normalized HAE attack rate over time were conducted using 10,000 virtual patients generated by sampling the eta distributions (a random variable describing interindividual variability) from the PKPD and exposure–response models. Body weight, baseline per‐4‐week normalized HAE attack rate, and baseline PKK concentration were set to 80 kg, 3.0 attacks per 4 weeks, and 122 mg/L, respectively, approximately reflecting the typical patient in OASIS‐HAE. Three simulation scenarios were evaluated to provide insights into dose levels and dosing regimens: a reference scenario for a regimen of 80 mg donidalorsen Q4W (defined as 28 days) or Q8W (defined as 56 days), a regimen of 80 mg donidalorsen Q1M (defined as 31 days) or Q2M (defined as 62 days), and a regimen‐switching scenario in which virtual patients started on donidalorsen 80 mg Q1M and were switched to donidalorsen 80 mg Q2M if they had an attack rate of < 0.333, corresponding to < 1 attack within the first 3 months, because the number of HAE attacks observed in an individual patient is a count and not a continuous variable. In the switching simulation, patients who were not attack‐free for 3 months remained on the Q1M dosing regimen.

2.4. Software

Nonlinear mixed‐effects modeling software (NONMEM version 7.5.1; ICON, Hanover, MD, USA) was used for the development of the exposure–response model. Assembly of the analysis datasets was performed using SAS (SAS/STAT software, Version 9.4; SAS System for Microsoft Windows 2002–2012; SAS Institute Inc., Cary, NC, USA). R (version 4.3.1) was used for statistical summaries and graphical exploration in addition to model diagnostics, predictions, and exposure–response simulations.

3. Results

3.1. Baseline Characteristics and Demographics

A summary of patient demographics and baseline characteristics is presented in Table S1. Of the 84 patients included in the analysis, 41 received donidalorsen 80 mg Q4W, 21 received donidalorsen 80 mg Q8W, and 22 received placebo (Table S2). Mean (standard deviation [SD]) baseline per‐4‐week normalized HAE attack rates were 3.62 (2.18) in the Q4W group, 3.17 (2.14) in the Q8W group, and 2.90 (1.66) in the placebo group. Mean (SD) predicted baseline prekallikrein concentrations (mg/L) were 126 (31.8), 144 (41.1), and 118 (27.0) in the Q4W, Q8W, and placebo groups, respectively. External model validation used data from 11 patients who received donidalorsen 80 mg Q4W and 6 who received placebo in the phase 2 study.

3.2. Exposure–Response Model

The final exposure–response model described HAE attack rate as a function of average prekallikrein concentrations modeled as a sigmoidal E_max relationship with baseline attack rate and baseline prekallikrein concentration as covariates. Age, body weight, sex, and treatment‐emergent ADA status were evaluated but not identified as significant covariates. Parameter estimates for the final model are presented in Table 1. Reductions in predicted prekallikrein concentrations were correlated with reductions in predicted HAE attack rates (Figure 1). The average effective prekallikrein concentrations leading to 50% (EC₅₀) and 90% (EC₁₀) reductions in HAE attack rate from the maximal rate were estimated to be 110 mg/L and 47.1 mg/L, respectively (Table 1). The exponent for the effect of baseline HAE attack rate on E_max was estimated to be 1.03, indicating a linear correlation between individual E_max and baseline HAE attack rate. The exponent for the effect of baseline prekallikrein concentration on EC₅₀ was estimated to be 0.13, indicating a weak dependence of EC₅₀ on baseline prekallikrein concentration such that relatively insignificant increases in EC₅₀ are expected with increasing baseline prekallikrein concentrations.

TABLE 1.

Parameter estimates of the exposure–response model.

Parameter	Estimate (RSE%)	Bootstrap statistics a		Shrinkage
		Median	95% CI
Emax	4.50 (9.6%)	4.54	3.31, 5.78	—
EC50 (mg/L)	110 (0.8%)	111	85.7, 124	—
Hill	2.60 (16%)	2.61	1.84, 4.20	—
bHAE	1.03 (7.9%)	1.03	0.85, 1.19	—
bPKK	0.13 (14.3%)	0.13	0.04, 0.18	—
Interindividual variability
IIV on Emax (CV%)	27.9 (28.8%)	26.5	7.0, 42.2	42.7
IIV on Hill (CV%)	131 (14.6%)	129	82.2, 255	27.1
Secondary parameters
EC10 (mg/L)	47.1	—	—	—
EC1 (mg/L)	18.7	—	—	—

Abbreviations: b_HAE, coefficient for the effect of baseline per‐4‐week normalized HAE attack rate; b_PKK, coefficient for the effect of baseline prekallikrein concentration; CI, confidence interval; CV%, percent coefficient of variation; EC₁, effective PKK_avg,4W associated with 99% reduction from the maximum attack rate; EC₁₀, effective PKK_avg,4W associated with 90% reduction from the maximum attack rate; EC₅₀, effective PKK_avg,4W associated with 50% reduction from the maximum attack rate; E_max, per‐4‐week HAE attack rate at infinite prekallikrein concentration; HAE, hereditary angioedema; IIV, interindividual variability; PKK_avg,4W, per‐4‐week average prekallikrein concentration; RSE%, percent relative standard error.

^aBased on 909 of 1000 replicates that minimized successfully.

FIGURE 1.

Relationship between HAE attack rates and predicted prekallikrein concentrations. (a) The solid line represents the model‐predicted relationship between 4‐week normalized HAE attack rate and predicted prekallikrein concentrations for a typical patient with a baseline concentration of 132 mg/L and baseline per‐4‐week normalized HAE attack rate of 3.47. The gray area represents the 90% prediction interval of the predicted relationship. The red and blue dots represent the observed mean per‐4‐week normalized HAE attack rates and time‐matched predicted per‐4‐week average prekallikrein concentrations for the patients in the Q4W and Q8W dosing regimens at baseline and in each 4‐week period of OASIS‐HAE. (b) Corresponding plot for change from baseline in per‐4‐week normalized HAE attack rate vs. time‐matched change from baseline in predicted per‐4‐week average prekallikrein concentration for the typical patient (solid line) and for each 4‐week period of OASIS‐HAE (dots). The dots at the origin (0,0) represent baseline. CFB, change from baseline; HAE, hereditary angioedema; PKK, prekallikrein; Q4W, once every 4 weeks; Q8W, once every 8 weeks.

A visual predictive check of HAE attack rate vs. prekallikrein concentration showed that the model predictions were in adequate agreement with the observed data and the model captured the relationship accurately (Figure S1). No bias was observed in the goodness‐of‐fit plots, which also showed the model accurately captured the observed HAE attack rates (Figure 2A,B). The model evaluation also showed no trends in Pearson residuals vs. time or vs. predictions (Figure S2). For external validation, the model was applied to predict HAE attack rate data from the phase 2 randomized study that were not included in model development. Observed and predicted HAE attack rates were in good agreement overall (Figure 2C). Furthermore, the model predicted the individual observed per‐4‐week HAE attack rates over time and vs. prekallikrein concentration reasonably well (Figure 3; Figure S3).

FIGURE 2.

Goodness of fit and external validation of the exposure–response model. (a) Comparison of the observed per‐4‐week normalized HAE attack rate in OASIS‐HAE and the model population‐predicted per‐4‐week normalized HAE attack rate. (b) Comparison of the observed per‐4‐week normalized HAE attack rate in OASIS‐HAE and the model individual‐predicted per‐4‐week normalized HAE attack rate. (c) Comparison of the observed per‐4‐week normalized HAE attack rate and the model‐predicted per‐4‐week normalized HAE attack rate in the phase 2 donidalorsen clinical study. Black dots are individual data points. The black dashed line is the identity line. The red line is a LOESS regression line in (a) and (b) and a linear regression line in (c). HAE, hereditary angioedema; LOESS, locally estimated scatterplot smoothing.

FIGURE 3.

Representative individual plots of predicted and observed per‐4‐week HAE attack rate vs. study week in the OASIS‐HAE study. Representative individual plots of predicted and observed per‐4‐week HAE attack rates vs. study week for patients in the (a) Q4W and (b) Q8W dosing groups. Observed per‐4‐week normalized HAE attack rates are represented by the black dots. The green and blue solid lines represent the population‐predicted and individual‐predicted per‐4‐week normalized HAE attack rates, respectively. BL, baseline; HAE, hereditary angioedema; Q4W, once every 4 weeks; Q8W, once every 8 weeks.

3.3. Exposure–Response Simulations

Simulation of 10,000 virtual patients on 80 mg Q4W or 80 mg Q1M dosing indicated that these dosing regimens were nearly identical with regard to efficacy (Figure 4A). Simulated mean per‐4‐week HAE attack rates at steady state were 0.39 for Q4W dosing and 0.42 for Q1M dosing, corresponding to mean percent reductions in HAE attack rates of 84.1% and 82.9% from baseline. Median HAE attack rates at steady state were 0.11 and 0.13, corresponding to median percent reductions in HAE attack rates of 95.6% and 94.6% from baseline with Q4W and Q1M dosing, respectively. Simulated efficacy results were also similar between Q8W and Q2M dosing (Figure 4B). Mean per‐4‐week HAE attack rates at steady state were 0.69 for Q8W dosing and 0.75 for Q2M dosing. Mean percent reductions from baseline in HAE attack rates were 72.6% for Q8W dosing and 70.2% for Q2M dosing. Median attack rates at steady state were 0.47 (Q8W) and 0.57 (Q2M), and median percent reductions from baseline were 81.1% (Q8W) and 77.3% (Q2M).

FIGURE 4.

Simulated HAE attack rate profiles by dosing regimen. Simulated per‐4‐week normalized HAE attack rates over time for (a) Q4W and Q1M dosing and (b) Q8W and Q2M dosing. The dots represent the median, and the error bars represent the 10th and 90th percentiles. HAE, hereditary angioedema; Q1M, once every month; Q2M, once every 2 months; Q4W, once every 4 weeks; Q8W, once every 8 weeks.

In a simulation of virtual patients switching from Q1M to Q2M dosing if attack‐free for 3 months, predicted response remained robust after the switch to less frequent dosing. In total, 56.4% of virtual patients had predicted per‐4‐week HAE attack rates < 0.333 in the first 3 months and were therefore eligible to switch to the Q2M dosing regimen. Predicted prekallikrein concentrations rose slightly but remained below the level required for clinically meaningful effect after the dosing regimen switch (Figure 5). Before the switch at 3 months, the simulated mean HAE attack rate for switch‐eligible patients was 0.07, and the median HAE attack rate was 0.02. After the switch, at steady state for the Q2M dosing regimen, the mean HAE attack rate was estimated to be 0.15 in the first month of the dosing interval after receiving a dose and 0.22 in the second month after receiving a dose. Median HAE attack rates after the switch at steady state were 0.07 in the first month of the dosing interval and 0.12 in the second month (Figure 6). These rates corresponded to mean percent reductions from baseline of 94.0% and 91.3% and median percent reductions from baseline of 97.4% and 95.5%, respectively. For comparison, if these patients did not switch to Q2M dosing, the mean steady‐state HAE attack rate was 0.04 and the median steady‐state HAE attack rate was 0.01, corresponding to a mean percent reduction from baseline of 98.2% and a median percent reduction of 99.7%. These data indicate that HAE attack rates remained low in switch‐eligible patients who switched from Q1M to Q2M dosing after 3 months without an attack.

FIGURE 5.

Predicted prekallikrein concentrations over time in simulated patients who switched dosing regimens from Q1M to Q2M. The dots represent the median, and the error bars represent the 10th and 90th percentiles. PKK, prekallikrein; Q1M, once every month; Q2M, once every 2 months.

FIGURE 6.

Simulated per‐4‐week HAE attack rate profiles for switching dosing regimens from Q1M to Q2M. The dots represent the median, and the error bars represent the 10th and 90th percentiles. HAE, hereditary angioedema; Q1M, once every month; Q2M, once every 2 months.

4. Discussion

In this study, a quantitative relationship between plasma prekallikrein concentrations and HAE attack rate following treatment with donidalorsen was established based on data from the phase 3 OASIS‐HAE clinical trial and validated using data from the phase 2 donidalorsen trial. Simulations conducted based on the developed exposure–response model showed similar efficacy between Q4W and Q1M dosing and between Q8W and Q2M dosing. Further simulations showed that the predicted reduction in HAE attack rate remained robust and clinically meaningful when switching virtual patients without an attack after 3 months on Q1M dosing to Q2M dosing.

Donidalorsen is a N‐acetyl galactosamine–conjugated antisense oligonucleotide designed to reduce HAE attacks by reducing the production of plasma prekallikrein in the liver, its main site of production [14]. Reduction of prekallikrein decreases the capacity for the generation of bradykinin, a potent vasodilator that increases vascular permeability and fluid extravasation, leading to the attacks of swelling associated with HAE [11, 20]. Several studies have demonstrated that donidalorsen effectively reduces plasma prekallikrein concentrations over time [14, 15, 17]. The phase 1 clinical study of unconjugated donidalorsen in healthy volunteers demonstrated that reductions in plasma prekallikrein concentration are accompanied by corresponding decreases in bradykinin generation capacity [15]. The phase 2 and phase 3 donidalorsen clinical trials have demonstrated that donidalorsen is effective in reducing HAE attack rates and that these reductions in attack rates are correlated with the observed reductions in plasma prekallikrein concentration [14, 17]. However, prior to the current study, a formal quantitative relationship between plasma prekallikrein concentrations and HAE attack rates had not been established.

The exposure–response model developed here successfully captured the relationship between the systemic concentration of plasma prekallikrein and the clinical efficacy endpoint of normalized per‐4‐week HAE attack rate. The final model described the time course of HAE attack rate as a function of prekallikrein concentration with a sigmoidal E_max relationship and had good agreement with the observed data. Statistically significant effects of baseline HAE attack rate on E_max and of baseline prekallikrein concentration on EC₅₀ were identified. The effect of baseline prekallikrein concentration on EC₅₀ was small (exponent of 0.13) and suggested that patients with higher baseline prekallikrein concentrations reach meaningful reductions in maximal HAE attack rate at prekallikrein concentrations similar to those of patients with lower baseline prekallikrein. Moreover, age, body weight, sex, and treatment‐emergent ADA status were not identified as significant covariates in the model. One important limitation of this study is that the clinical trials for donidalorsen had relatively small sample sizes due to the rarity of HAE, and therefore correspondingly low numbers of observations were available for modeling. A low sample size can limit the ability to estimate the impact of covariates on the relationship between prekallikrein and HAE attack rates. On the individual level, heterogeneity in HAE attack rates over time was observed for some patients, resulting in a weaker fit for those patients. Though the underlying genetic cause of HAE (deficiency or dysfunction of the C1 inhibitor protein) is the primary determinant of disease severity, known triggers for HAE attacks, such as stress, physical trauma, infections, and hormonal fluctuations, vary in their impact and could influence HAE attack rates even while patients are on a prophylactic therapy such as donidalorsen [21]. This can result in individual variability in attack rates over time that cannot be fully accounted for by a model. These limitations notwithstanding, the results suggest a generally comparable and clinically meaningful response to donidalorsen across the patient population irrespective of the magnitude of the baseline rate of HAE attacks or prekallikrein levels at baseline.

While several approved prophylactic treatments for HAE are available, significant disease burden remains, including a substantial burden of treatment [22, 23, 24]. Patients generally report reduction of HAE attacks as their highest priority for treatment, but also cite the inconvenience of dosing route and frequency as a disadvantage of currently available treatments and as a factor that could influence their choice to switch medications [8, 23, 24]. The simulations reported here support the Q4W and Q8W dosing regimens tested in the donidalorsen clinical trials [14, 17, 18]. The simulations also indicate that the efficacy of Q1M or Q2M dosing is similar to that of Q4W and Q8W dosing, respectively. Donidalorsen has now been approved by the FDA as a prophylaxis to prevent HAE attacks, with recommended dosing frequencies of Q4W or Q8W [16]. While these dosing intervals were ultimately recommended in the label based on direct support from the trials, there is a possibility patients may miss their exact day of Q4W or Q8W dosing, and these patients may land closer to once a month or every other month. These modeling results underscore that the small difference in days between these dosing intervals (Q4W and Q1M, Q8W and Q2M) would not be expected to impact efficacy, and the simulations further suggest that a switch in dosing can be safely managed without an increase in HAE attack rate.

The option of donidalorsen dosing as infrequently as Q8W could be a substantial improvement in convenience for patients whose current treatments require dosing every day, every 3 to 4 days, or every 2 weeks [25, 26, 27]. The dose‐switching simulation showed that HAE attack rates remained low in patients who switched from donidalorsen Q1M to donidalorsen Q2M after experiencing no attacks for 3 months. A 3‐month timeframe was chosen based on the results of the donidalorsen phase 2 open‐label extension study, in which 5/8 patients who changed from Q4W to Q8W dosing after being attack‐free for at least 12 weeks remained attack‐free up to 2 years [18]. Moreover, the results of both the phase 2 and phase 3 randomized, placebo‐controlled clinical studies suggest that near‐maximal clinical effect is reached after approximately 12 weeks [14, 17]. Therefore, this period is considered sufficient to confirm whether patients have achieved a stable response with the initial donidalorsen dosing regimen and can be considered for a regimen switch. However, the transition from donidalorsen Q4W to Q8W dosing is an individualized clinical decision that should be based on specific treatment goals and needs. These simulation results support that this dose‐switching regimen could benefit patients who experience no attacks for the first 3 months by combining initially more frequent Q4W dosing to achieve a rapid reduction in prekallikrein and a robust clinical response, with long‐term Q8W dosing to maintain the clinical response on a more convenient and less disruptive dosing schedule. This dosing flexibility would allow individualization of treatment based on shared decision‐making between patients and their healthcare providers.

Taken together, these results provide further support for donidalorsen as a potential LTP for the treatment of HAE, describe a quantitative relationship between reductions in plasma prekallikrein concentrations and reductions in HAE attack rate, and offer additional evidence to support a patient‐friendly Q1M‐to‐Q2M dose‐switching regimen for donidalorsen.

Author Contributions

All authors wrote the manuscript. Laura Bordone, Kenneth B. Newman, and Pratap Singh designed and performed the research. Laura Bordone, Huub Jan Kleijn, Kenneth B. Newman, Pratap Singh, and Han Witjes analyzed the data.

Conflicts of Interest

P.S., J.K.D., L.B., K.B.N., and X.G. are employees of Ionis and hold shares and/or options in Ionis. H.W. and H.J.K. are employees of Certara. D.M.C. has received speaker or consultancy fees from Astria, BioCryst Pharmaceuticals, CSL Behring, Intellia Therapeutics, Ionis, KalVista Pharmaceuticals, Pharvaris, and Takeda.

Supporting information

Data S1: Supporting Information.

Singh P., Witjes H., Kleijn H. J., et al., “Exposure–Response Analysis of Donidalorsen for the Treatment of Hereditary Angioedema,” Clinical and Translational Science 18, no. 11 (2025): e70388, 10.1111/cts.70388.