1 Introduction
Finerenone is a nonsteroidal selective mineralocorticoid receptor antagonist (MRA) that recently demonstrated efficacy in delaying the progression of kidney disease and reducing the risk of cardiovascular events in patients with chronic kidney disease (CKD) and type 2 diabetes (T2D) in the pivotal phase III trial FIDELIO-DKD (ClinicalTrials.gov number, NCT02540993) [
1]. FIDELIO-DKD was an event-driven trial randomizing 5734 patients with advanced CKD to study kidney (primary) and cardiovascular (key secondary) outcomes. Design details of relevance for this manuscript are summarized in Sect. 2; further details have been published [
2].
The efficacy and safety of finerenone is also investigated in three additional phase III studies: the recently completed FIGARO-DKD study (NCT02545049) [
3] in patients with less advanced CKD and T2DM compared with patients in FIDELIO-DKD, FIND-CKD (NCT05047263) in patients with nondiabetic CKD, and FINEARTS-HF (NCT04435626) in patients with heart failure with a left ventricular ejection fraction of ≥ 40%.
Finerenone binds to the mineralocorticoid receptor (MR), which is expressed in many tissues and cells, including in the heart, kidneys, blood vessels, and immune cells, and blocks the effects of its natural hormone ligands, aldosterone and cortisol. Besides beneficial effects on inflammation and fibrosis, blockade of the MR at the collecting duct and distal nephron in the kidneys leads to a decreased reabsorption of sodium and consequently a decreased excretion of potassium into the urine. Thus, the use of finerenone cannot be devoid of any effect of increasing serum potassium [
4,
5].
In the advanced CKD population studied in FIDELIO-DKD (mean estimated glomerular filtration rate [eGFR] 44.3 mL/min/1.73m
2, median urine albumin-to-creatinine ratio [UACR] 851 mg/g) with a median follow-up of 2.6 years, hyperkalemia-related treatment-emergent adverse events (TEAEs; including Medical Dictionary for Regulatory Activities preferred terms “hyperkalemia” and “blood potassium increased”) and serum potassium increases of > 5.5 and > 6.0 mmol/L were collected. The number of subjects with hyperkalemia-related TEAEs was higher in the finerenone arm (18.3%) than in the placebo arm (9.0%) [
1,
2]. Most of these events were nonserious and mild or moderate in intensity. Hyperkalemia TEAEs leading to permanent discontinuation of study drug (2.3% of subjects) or hospitalization (1.4% of subjects) constituted a small proportion of these events in the finerenone group. No fatal TEAEs of hyperkalemia were observed in either treatment group, and no evidence was found for an increased incidence of any severe clinical cardiac manifestations of hyperkalemia (e.g., ventricular arrhythmia or sudden cardiac death). The incidences of serum potassium > 5.5 and > 6.0 mmol/L (based on central laboratory measurements) were 21.7 and 4.5% in the finerenone group and 9.8 and 1.4% in the placebo group, respectively [
1].
Since assessment of hyperkalemia-related TEAEs was subject to the investigator’s clinical judgment and provides limited binary information only, serum potassium was chosen to study the relation to finerenone exposure in a population pharmacokinetic and pharmacodynamic (popPKPD) model informed by more than 148,000 local and central serum potassium measurements. In this manuscript, serum potassium > 5.5 and > 6.0 mmol/L are generally referred to as hyperkalemia of different severity.
We previously studied serum potassium popPKPD based on phase IIb data (ARTS-DN and ARTS-DN Japan studies) [
6]. The analyses were extended to FIDELIO-DKD data to better understand the relationship between finerenone dose and exposure to serum potassium and hyperkalemia, qualitatively and quantitatively. Clinical trial simulations quantified the influence of serum potassium thresholds for inclusion and dose titration. The analyses also explained and highlighted the role of dose titration, which was implemented in phase III, in the management of serum potassium. In van den Berg et al. [
7], we reported the population pharmacokinetic analysis of FIDELIO-DKD along with exposure-response (ER) analysis for the primary renal time-to-event endpoint. That analysis also provided the pharmacokinetic parameter post hoc estimates as a basis for the serum potassium popPKPD we describe here.
2 Methods
2.1 Clinical Study
Study design, patient characteristics, and the main study results have been published [
1,
2,
8]. Information on informed consent, ethics, and analytical methods are also summarized in the electronic supplementary material (ESM).
In brief, FIDELIO-DKD was a randomized, double-blind, placebo-controlled phase III study investigating the efficacy and safety of finerenone, in addition to standard of care (SoC), on the progression of kidney disease in patients with T2DM and CKD. For inclusion, patients were required to have central laboratory serum potassium levels ≤ 4.8 mmol/L at both the run-in and the screening visits, although a single re-assessment was allowed for each visit. However, patients with serum potassium levels > 4.8 mmol/L at baseline (≥ 2 weeks after screening) were allowed to start treatment at the investigator’s discretion. The starting dose of finerenone was dependent on the eGFR values at screening visits. At randomization (baseline), subjects with an eGFR 25 to < 60 mL/min/1.73 m2 (central laboratory measurements) at the screening visits were assigned to the lower dose of finerenone (10 mg) or placebo in addition to SoC, whereas subjects with an eGFR ≥ 60 mL/min/1.73m2 were assigned to the higher dose of finerenone (20 mg) or placebo in addition to SoC, which was the individually maximum tolerated labeled dose of either an angiotensin-converting enzyme inhibitor or an angiotensin II receptor blocker.
During the study, potassium was measured in the local laboratories of each study site—for titration and safety monitoring—as well as in the central laboratory. From the month 1 visit onwards, patients were eligible for uptitration of finerenone dose (from 10 to 20 mg) if local laboratory serum potassium levels were ≤ 4.8 mmol/L and if eGFR (local laboratory value) had not decreased by more than 30% compared with the value measured at the last regular visit. If local laboratory serum potassium levels were > 5.5 mmol/L, study drug was interrupted and an additional re-test was scheduled within 3 days. After study drug interruption, subjects were allowed to re-start study drug at the lower dose of 10 mg when serum potassium levels were ≤ 5.0 mmol/L. Interruption of study drug was permitted at any time during the study for any safety reasons according to the investigator’s clinical judgment. A safety visit for a local potassium assessment was scheduled ± 4 weeks after each uptitration or re-initiation of treatment after an interruption of more than 7 days.
2.2 Model Development
A popPKPD model was developed to characterize the effect of finerenone exposure on serum potassium in NONMEM 7.4. The individual post hoc estimates of pharmacokinetic parameters from the FIDELIO-DKD population pharmacokinetic model were used to calculate the area under the finerenone concentration–time curve at steady state with the current dose level [
7], which was used as the exposure metric to drive the effect of finerenone in the PKPD potassium model. Full simulated concentration–time profiles were also evaluated but did not improve the data description and were clearly computationally inferior. The pharmacokinetic model considered dose titration; with a half-life of 2.7 h, a new pharmacokinetic steady state is rapidly attained [
6,
7]. As the model was supposed to simulate data for altered inclusion thresholds for serum potassium, data from screening failures (subjects who participated in run-in and screening visits but were not included in FIDELIO-DKD) were also considered for this analysis, with the exception of subjects who did not meet the eGFR inclusion criteria (≥ 25 and < 75 mL/min/1.73 m
2 at run-in and screening visits).
Model development occurred according to a stepwise approach, starting with the development of the structural model using the central laboratory data only. In the second and third steps, the local laboratory data were included and a covariate analysis performed, respectively. As a final step, potassium data from the run-in and screening visits (including data of screening failures) were added to the dataset. For nested models, during model development, additional parameters were only added to the model if this significantly (p < 0.001) improved the model fit, as determined by the objective function value (− 2 times the log-likelihood). If the models being compared were structurally different (and thus not nested), the models were compared using the Akaike information criterion.
The structural model included a turnover model for the effect of finerenone on serum potassium, similar to the phase IIb PKPD model [
6]. The effect of finerenone was included as an effect on the first order dissipation rate constant (
kout) parameter in the turnover model, and the following relationships were tested to characterize the ER relationship: linear, power, maximum drug effect (
Emax), and sigmoid
Emax. The structural model development also tested the inclusion of a linear progression slope (TSLOPE) of serum potassium over time.
A limited number of known risk factors for hyperkalemia were tested as potential covariates on parameters for which interindividual variability was included in the model: age, baseline eGFR, baseline UACR, and sex. Based on diagnostic plots, the possible effect of baseline UACR and treatment randomization (active vs. placebo) were also tested on TSLOPE, even though interindividual variability for TSLOPE was not included in the model. Additionally, for comparability with a previously developed potassium PKPD model that included phase IIb data from a Japanese sister study (ARTS-DN Japan) [
6], Japanese ethnicity was evaluated as a covariate for baseline serum potassium and residual variability.
2.3 Model Evaluation
The performance of the model was evaluated by performing simulations and comparing these with observations, hereafter referred to as a visual predictive check (VPC). The VPC simulations included residual error and interindividual variability.
Simulated subjects were “screened” at the run-in and screening visits according to the FIDELIO-DKD protocol criteria and removed from the simulated study if they did not meet the serum potassium inclusion criterion of ≤ 4.8 mmol/L. To adequately reflect the FIDELIO-DKD dosing algorithm, the simulations for the VPC included simulation of dose titration decisions based on simulated local laboratory serum potassium observations. The occurrence of potassium re-tests and uptitration safety visits was also dependent on the simulated serum potassium observations and uptitration decisions, in line with the FIDELIO-DKD study design. Simulated central laboratory values were used to assess safety in terms of reaching serum potassium thresholds. A more detailed description of the VPC simulation workflow and computational methods and a NONMEM control stream are provided in the ESM.
2.4 Potassium Threshold Simulations
Additional simulations were performed to virtually explore the safety impact of increasing the serum potassium threshold from ≤ 4.8 to ≤ 5.0 mmol/L to determine eligibility for inclusion and uptitration of finerenone. The same simulation workflow was used as to generate the simulated data for the VPCs, except that parameter uncertainty was included to reflect the whole estimated uncertainty, resulting in a scenario similar to that in FIDELIO-DKD (i.e., number of subjects, frequency of visits, duration of the study, covariate distribution, screening, and uptitration procedure). In line with the FIDELIO-DKD protocol, simulated local laboratory values were used to guide dose titration, and central laboratory values were used to describe safety results.
4 Discussion
Finerenone, with its inherent mode of action as an MRA, increases serum potassium, particularly in patients with advanced CKD who are on maximum tolerated labeled doses of a renin–angiotensin system inhibitor, as studied in FIDELIO-DKD. Accordingly, the analysis showed increased serum potassium levels and hyperkalemia rates with finerenone treatment compared with placebo. However, the analysis of serum potassium values and finerenone exposure or dose revealed an apparent inverse relationship not in line with a naïve “traditional” assumption, where the highest hyperkalemia risk would be expected at the highest dose level (Fig.
1).
The model-based analysis can explain this phenomenon. In this model, finerenone increased serum potassium levels in an exposure-dependent fashion, with the strongest effect being predicted at the higher dose level (20 mg), as shown in Fig.
5a illustrating the intrinsic dose–response relationship, i.e., the relationship with a constant dose assuming no dose titration. However, patients whose serum potassium levels were too high discontinued or interrupted treatment and re-started at the lower dose level, whereas patients with lower serum potassium levels were uptitrated to the higher dose. By simulating the dose-titration decisions based on simulated serum potassium levels, the model re-created the pattern in the observations where the patients on the higher dose level had the lower serum potassium levels as shown in Fig.
5b. Although dose titration was a design element of FIDELIO-DKD intended to manage hyperkalemia, it inverted the relation between dose and serum potassium, which may not have been expected a priori. Although the risk for an individual patient followed the intrinsic dose–response relationship and thus increased with dose, in the apparent behavior of the FIDELIO-DKD population, patients receiving higher doses had a lower hyperkalemia risk compared with patients receiving a lower dose. Interestingly, independent of the findings described here, Schnider et al. [
9] also recently described and conceptually proved in a related setting how dose titration can lead to the described phenomenon, referring to the “drug titration paradox,” and related work providing hints has been published [
10‐
12]. From a systems theory perspective, dose titration introduces negative feedback, which is the basis of control and supports stabilization of systems, i.e., reducing fluctuations and supporting equilibration. The concept is well-established for various biological systems and, in physiology, is often relevant to the support of homeostasis, including in the face of perturbations [
13‐
15].
Besides providing qualitative insight into serum potassium dynamics under finerenone treatment, the model-based covariate analysis quantified the effect of important risk factors. Most prominently, in the model, patients with low eGFR and/or high UACR had an increased risk of hyperkalemia because of effects on baseline serum potassium, finerenone effect, or disease progression. The identified effects were in line with general expectations, as low eGFR and high UACR characterize disease progression and are known to be associated with an increased risk for hyperkalemia [
16‐
18]. The number of covariates tested in the current study was limited, and evaluating additional factors such as comedications or comorbidities in the future could be of interest to further support the understanding of serum potassium biology and hyperkalemia risk.
In FIDELIO-DKD, patients who received finerenone had higher mean serum potassium levels than those who received placebo. A maximum mean difference between both groups, of 0.23 mmol/L, was observed at month 4, and the difference remained largely stable thereafter [
1]. With a slower disease progression rate in patients receiving finerenone than in those receiving placebo, the current analysis also identified an opposite, i.e., lowering, effect on potassium on the long term. This potassium-lowering effect likely reflected the kidney-protective effect of finerenone treatment [
1,
7]. Although the net effect on the mean potassium values was an increase with finerenone treatment within the timescale investigated in FIDELIO-DKD (median follow-up 2.6 years), the effect on potassium lowering may attenuate the potassium-increasing effect of finerenone compared with placebo during long-term treatment.
In FIDELIO-DKD, serum potassium ≤ 4.8 mmol/L was required for inclusion into the study, based on values at run-in and screening. Similarly, dose uptitration was limited to subjects with values below that threshold. However, because of the variability in serum potassium, 13.6% of subjects at baseline (randomization) had a serum potassium level > 4.8 mmol/L. Variability in the model was considered random and reflected the unknown effects of, for example, amounts of potassium in food or patient hydration status. These subjects also informed the model on treatment effects with high baseline values. Generally, the model adequately described the potassium dynamics and treatment effect, including pretreatment, start, uptitration, interruption, or discontinuation (example shown in Fig.
2). Therefore, the model was considered suitable to assess the influences of increased inclusion or uptitration thresholds. Model-based analyses showed that the absolute risk for hyperkalemia increased with increasing baseline serum potassium values, whereas the risk ratio of hyperkalemia for patients receiving finerenone relative to those receiving placebo did not (but rather decreased according to the model), because the hyperkalemia risk increased disproportionately for patients receiving placebo compared with patients receiving finerenone (Fig.
3). This was also seen in simulations with a FIDELIO-DKD-like population, where inclusion and uptitration limits (serum potassium ≤ 4.8 mmol/L) were compared with simulations of an extended scenario with these limits increased to serum potassium ≤ 5.0 mmol/L (Table
2). The absolute risk for hyperkalemia also (substantially) increased in the subset of this population with serum potassium between > 4.8 and ≤ 5.0 mmol/L, whereas the risk ratio decreased compared with the FIDELIO-DKD setting. Based on these findings, model-based analyses indicated that patients with high serum potassium baseline values, especially > 4.8 mmol, had a high risk for hyperkalemia independent of treatment assignment, and the risk increased disproportionately in the placebo group. FIDELIO-DKD studied a population with advanced CKD, but very recently published FIGARO-DKD data indicated a lower finerenone-related absolute hyperkalemia risk in patients with better preserved renal function [
3].
FIDELIO-DKD was a population at increased risk of hyperkalemia, as also shown by the hyperkalemia rates in patients receiving placebo, where the use of steroidal MRAs is typically avoided [
19], or in case of eplerenone contraindicated (US and UK prescribing information). A meta-analysis indicated a fourfold increase in hyperkalemia with steroidal MRAs in CKD with T2DM [
20], whereas the data for finerenone presented in Sects. 1 and 3 point more towards a twofold increase.
Acknowledgements
The authors thank the FIDELIO-DKD committees, investigators, and patients; Sissy Stauffenberg (ClinStat), Martin Gebel, Cosima Klein, Aziz Tuermer, Jacobus Buytendach, and Patrick Schloemer (Bayer) for derived variables, information about data, and alignment with statistical analysis; Martijn van Noort (Leiden Experts on Advanced Pharmacokinetics and Pharmacodynamics) for critical discussion of popPKPD methods and results; Udo Kuhnert and Nicholas Downie (Bayer) for supporting pharmacometric analysis dataset creation; Nurahamid Siddiki, Md. Aminul Islam, and Rumana Khatun (Shafi Consultancy) for data set quality control; Hauke Ruehs, Yang Zhang, and Alexander Solms (Bayer) for critical discussion of methods and results; Roland Heinig and Susanne Metzger for critical discussion of results; Simone Steinbach and Ui Yen Morgenthaler (Bayer) for medical writing support.