Hendrik Maxime Lagraauw (1), F. Lee Hodge (1), Stefan Hantel (2), Jan-Georg Wojtyniak (2), Silke Retlich (2)
(1) qPharmetra, LCC. (2) Boehringer Ingelheim Pharma GmbH & Co. KG
Objectives: Chronic kidney disease (CKD) is a prevalent, progressive condition with no cure, causing significant morbidity and mortality, particularly in individuals with diabetes and hypertension. Preservation of kidney function can improve outcomes and can be achieved through non-pharmacological strategies (e.g., dietary and lifestyle adjustments) and chronic kidney disease-targeted and kidney disease specific pharmacological interventions, including angiotensin-converting enzyme (ACE) inhibitors, angiotensin receptor blockers (ARBs), sodium-glucose transport protein 2 (SGLT2) inhibitors and mineralocorticoid receptor antagonists (MRAs).
This analysis utilized a model-based meta-analysis (MBMA) approach to develop longitudinal models for these drug classes, focusing on the percentage change from baseline in glomerular filtration rate (GFR) and urine albumin-to-creatinine ratio (UACR) using summary-level data extracted from CKD studies conducted worldwide between 1992 and 2022.
Methods: Two distinct analysis datasets were created from a licensed CKD database (CODEX), encompassing 52 studies (112 treatment arms) with longitudinal kidney function endpoints and 48 studies (103 treatment arms) with longitudinal proteinuria endpoints. Efforts were dedicated to standardizing units and summary metrics to leverage as much of the available data as possible. For UACR, conversion of UPCR to UACR values was implemented following the equation published by Sumida et al. [1]. The multiple endpoint metrics and various reported summary statistics were assumed comparable and compatible when modeling percentage change from baseline. To acknowledge that variability between treatment
arms within a study is likely less than across different studies, a nested study and arm variability structure was introduced via a super ID, and the $LEVEL functionality in NONMEM. Residual unexplained variability (RUV) and inter-arm variability were weighted by the (time-varying) treatment arm sample size.
Results: The GFR model demonstrated alignment with published structural models, incorporating linear disease progression [-0.122 %/week], slowed down by a linear improvement in GFR under active treatment and accounting for an acute decline in GFR after start of active treatment due to intrarenal hemodynamic changes. Unique drug class treatment effects were estimated for SGLT2 inhibitors [0.091 %/week; -5.13 %] and MRAs, ACE & ARBs (combined) [0.028 %/week; -2.56 %], treatment effect slope and hemodynamic effect offset parameters, respectively. A dose-response relationship for the non-steroidal MRA finerenone was introduced to allow predictions across a range of dose levels. Although covariates were limited, baseline GFR was identified as a significant covariate, resulting in faster relative disease progression at lower baseline values [power model exponent estimated to -0.391, standardized to a median GFR of 51.45].
The final UACR model consisted of a linear disease progression [-0.006 %/week] counteracted by an (Emax-like) active treatment response relationship that saturates over time. Separate drug class treatment effects were estimated for SGLT2 inhibitors [-36.4 %; 1.6 weeks], MRAs [-49.6 %; 3.4 weeks], and ACE & ARBs (combined) [-36.7 %; 10.8 weeks], for the maximal effect and time at which half the maximal effect is reached parameters, respectively. A finerenone dose-response relationship was estimated to allow future benchmarking against a range of finerenone dose levels. Covariate modeling for UACR did not identify significant covariates that could be retained in a stable model.
Clinical trial simulations, employing the developed models, offered valuable insights into the impact of different drug classes on kidney function and proteinuria outcomes in CKD patients, paving the way for benchmarking of emerging treatments.
Conclusions: Summary-level literature data were used to successfully develop longitudinal models for GFR and UACR percentage change from baseline. This model-based meta-analysis will facilitate the future benchmarking of novel CKD treatments against existing drugs.
References:
[1] Sumida K, Nadkarni GN, Grams ME, et al. Conversion of Urine Protein-Creatinine Ratio or Urine Dipstick Protein to Urine Albumin-Creatinine Ratio for Use in Chronic Kidney Disease Screening and Prognosis : An Individual Participant-Based Meta-analysis. Ann Intern Med. 2020;173(6):426-435. doi:10.7326/M20-0529
Reference: PAGE 32 (2024) Abstr 10835 [www.page-meeting.org/?abstract=10835]
Poster: Drug/Disease Modelling - Other Topics