Clinical trial designs to assess treatment effects on glomerular filtration rate decline

Abstract

Glomerular filtration rate (GFR) decline is used as surrogate endpoint for kidney failure. Interventions that reduce chronic kidney disease (CKD) progression often exert acute GFR reductions which differ from their long-term benefits and complicate the estimation of long-term benefit. Here, we assessed the utility of two alternative trial designs (wash-out design and active run-in randomized withdrawal design) that attempt to exclude the impact of acute effects. Post-hoc analyses of two clinical trials that characterized the effect of an intervention with acute reductions in GFR were conducted. The two trials included a wash-out period (EMPA-REG Outcome testing empagliflozin vs placebo) or an active run-in period with a randomized withdrawal (SONAR testing atrasentan vs placebo). We compared the drug effect on GFR decline calculated from the first on-treatment visit to the end of treatment (chronic slope in a standard randomized trial design) with GFR change calculated from randomization to end of wash out, or GFR change from treatment-specific baseline GFR values (GFR at start-of-run-in for placebo and end-of-run-in for atrasentan) until end-of-treatment. The effect of empagliflozin versus placebo on chronic GFR slope was 1.72 (95% confidence interval 1.49-1.94) mL/min/1.73 m2/year, similar to total GFR decline from baseline to the end of wash-out period using a linear mixed model 1.64 (1.44-1.85) mL/min/1.73 m2/year). The effect of atrasentan versus placebo on chronic GFR slope was 0.72 (0.32-1.11) mL/min/1.73 m2/year, similar to total slope from a single slope model when estimated from treatment specific baseline GFR values 0.77 (0.39-1.14) mL/min/1.73 m2/year). Statistical power of the two designs outperformed the standard randomized design. Thus, wash-out and active-run-in randomized-withdrawal trial designs are appropriate models to compute treatment effects on GFR decline. Glomerular filtration rate (GFR) decline is used as surrogate endpoint for kidney failure. Interventions that reduce chronic kidney disease (CKD) progression often exert acute GFR reductions which differ from their long-term benefits and complicate the estimation of long-term benefit. Here, we assessed the utility of two alternative trial designs (wash-out design and active run-in randomized withdrawal design) that attempt to exclude the impact of acute effects. Post-hoc analyses of two clinical trials that characterized the effect of an intervention with acute reductions in GFR were conducted. The two trials included a wash-out period (EMPA-REG Outcome testing empagliflozin vs placebo) or an active run-in period with a randomized withdrawal (SONAR testing atrasentan vs placebo). We compared the drug effect on GFR decline calculated from the first on-treatment visit to the end of treatment (chronic slope in a standard randomized trial design) with GFR change calculated from randomization to end of wash out, or GFR change from treatment-specific baseline GFR values (GFR at start-of-run-in for placebo and end-of-run-in for atrasentan) until end-of-treatment. The effect of empagliflozin versus placebo on chronic GFR slope was 1.72 (95% confidence interval 1.49-1.94) mL/min/1.73 m2/year, similar to total GFR decline from baseline to the end of wash-out period using a linear mixed model 1.64 (1.44-1.85) mL/min/1.73 m2/year). The effect of atrasentan versus placebo on chronic GFR slope was 0.72 (0.32-1.11) mL/min/1.73 m2/year, similar to total slope from a single slope model when estimated from treatment specific baseline GFR values 0.77 (0.39-1.14) mL/min/1.73 m2/year). Statistical power of the two designs outperformed the standard randomized design. Thus, wash-out and active-run-in randomized-withdrawal trial designs are appropriate models to compute treatment effects on GFR decline. Lay SummaryGlomerular filtration rate (GFR) decline is a surrogate endpoint in clinical trials but can produce ambiguous results due to acute treatment effects of interventions that differ from longer-term treatment effects. We demonstrate that 2 alternative trial designs (washout design and active run-in randomized withdrawal design) that attempt to exclude the impact of acute effects on longer-term treatment effects are appropriate study designs to compute treatment effects on GFR decline. We also showed that the 2 alternative trial designs provide greater statistical power than the standard randomized design. These results may serve as an impetus to design more efficient clinical trials to assess the efficacy of existing and new therapies to slow kidney function decline. Glomerular filtration rate (GFR) decline is a surrogate endpoint in clinical trials but can produce ambiguous results due to acute treatment effects of interventions that differ from longer-term treatment effects. We demonstrate that 2 alternative trial designs (washout design and active run-in randomized withdrawal design) that attempt to exclude the impact of acute effects on longer-term treatment effects are appropriate study designs to compute treatment effects on GFR decline. We also showed that the 2 alternative trial designs provide greater statistical power than the standard randomized design. These results may serve as an impetus to design more efficient clinical trials to assess the efficacy of existing and new therapies to slow kidney function decline. Clinical endpoints in trials of chronic kidney disease (CKD) progression include 57% glomerular filtration rate (GFR) decline (doubling of serum creatinine) and kidney failure.1Levey A.S. Gansevoort R.T. Coresh J. et al.Change in albuminuria and GFR as end points for clinical trials in early stages of CKD: a scientific workshop sponsored by the National Kidney Foundation in collaboration with the US Food and Drug Administration and European Medicines Agency.Am J Kidney Dis. 2020; 75: 84-104Abstract Full Text Full Text PDF PubMed Scopus (330) Google Scholar These endpoints are late manifestations of CKD progression and lead to large clinical trials with substantial follow-up to obtain sufficient clinical events to achieve adequate statistical power. The change in GFR over time (usually analyzed via total GFR slope) is now employed as a surrogate endpoint to facilitate design of more efficient clinical trials, including trials of rare kidney diseases.1Levey A.S. Gansevoort R.T. Coresh J. et al.Change in albuminuria and GFR as end points for clinical trials in early stages of CKD: a scientific workshop sponsored by the National Kidney Foundation in collaboration with the US Food and Drug Administration and European Medicines Agency.Am J Kidney Dis. 2020; 75: 84-104Abstract Full Text Full Text PDF PubMed Scopus (330) Google Scholar, 2Heerspink H.J. Agarwal R. Bakris G.L. et al.Design and baseline characteristics of the FIND-CKD Trial: efficacy of Finerenone on Kidney Disease Progression in People with Non-Diabetic CKD.J Am Soc Nephro. 2023; 34: S678Crossref Google Scholar, 3Inker L.A. Collier W. Greene T. et al.A meta-analysis of GFR slope as a surrogate endpoint for kidney failure.Nat Med. 2023; 29: 1867-1876Crossref PubMed Scopus (26) Google Scholar However, interventions that affect CKD progression typically produce short-term effects on GFR (referred to as acute effects) that can differ from their long-term treatment effects (referred to as chronic GFR slope). These acute effects complicate the interpretation and assessment of long-term drug efficacy. In the absence of an acute effect due to treatment, GFR declines linearly over time during the follow-up period of a clinical trial, such that a linear single-slope model accurately characterizes the GFR trajectory. However, in the setting of acute effects, a linear model may not accurately capture the GFR trajectory. In such cases, a piecewise linear 2-slope model has been used to separate the acute and chronic effects.4Heerspink H.J.L. Stefansson B.V. Correa-Rotter R. et al.Dapagliflozin in patients with chronic kidney disease.N Engl J Med. 2020; 383: 1436-1446Crossref PubMed Scopus (2715) Google Scholar,5Perkovic V. Jardine M.J. Neal B. et al.Canagliflozin and renal outcomes in type 2 diabetes and nephropathy.N Engl J Med. 2019; 380: 2295-2306Crossref PubMed Scopus (3932) Google Scholar However, the optimal use of the 2-slope model requires specification of the length of the acute phase in the analysis, which may not be fully understood at the time of trial design. In addition, interventions that are meant to slow CKD progression are typically intended for long-term chronic use, a setting in which the impact of an acute effect on the overall progression of disease is likely to be minimized. However, such long-term treatment is often infeasible in a clinical trial setting, meaning that acute effects may exert undue influence on clinical trial results, compared with typical longer-term treatment in clinical practice. To overcome the need for specification of the length of the acute phase and to optimize the estimation of the long-term benefit of chronic treatments for interventions with acute effects, 2 different clinical trial designs have been used (Figure In the first to as the the change in GFR is estimated from randomization to an follow-up which a the on-treatment visit The clinical trial to as the active run-in withdrawal an active run-in period with randomized withdrawal at the end of the In the GFR at the of the run-in is used as baseline for the placebo The GFR at the end of the active run-in is used as the baseline for the active treatment The in GFR slope GFR values randomization and not during is calculated with baseline values as a designs overcome the of an acute effect on the rate of GFR the use of a piecewise model designs and for GFR over randomized design is at a time slope from a single-slope is as the the of treatment (usually at the randomization slope from a 2-slope piecewise linear is as the the of the model to acute and chronic the end of slope from a is as the the of from baseline to is as the the of from the period but from baseline to visit including is as the the of from the period and run-in with randomized slope from a single-slope for is as the the of the for active treatment is as the GFR at the end of the the randomized The GFR at randomization for the placebo is from the slope is are using active study from baseline to the end of from the treatment period are not included in the from baseline to the end of treatment including of glomerular filtration mixed models for in a new of glomerular filtration mixed models for and clinical trial designs have been used in and V. et al.A phase randomized trial of in to kidney disease Am Soc Nephro. Google in patients with Scholar are the and of and In addition, are of the or active run-in with randomized withdrawal designs compared with the standard randomized trial which is in designs in clinical trial We analyzed 2 clinical trials to the impact of different trial design on GFR decline. We analyses to the in of statistical power and of the different clinical trial In we used from 2 randomized clinical trials that assessed the efficacy and of the empagliflozin in the Outcome in 2 (EMPA-REG and the atrasentan in the with et and progression of kidney disease in type 2 Engl J Med. PubMed Scopus Google H.J.L. et and renal events in patients with type 2 diabetes and chronic kidney disease a 2019; Full Text Full Text PDF PubMed Scopus Google Scholar We phase clinical trials included a period (EMPA-REG and an active run-in and withdrawal design The were with a linear mixed effects model with values estimated from the of the Outcome and The Outcome trial randomized patients were at of with a of or an estimated GFR of and were to empagliflozin or or placebo in a the empagliflozin and were was that the efficacy or of empagliflozin the 2 The was and the trial patients to with a of type 2 estimated GFR and and and were using an or and atrasentan during an active run-in period to patients likely to to as a in of or and to exclude patients to as an of at in or an in to at patients atrasentan and a of patients to the randomization visit and were in a to atrasentan or to to we the and the effect of atrasentan on the kidney not differ in and H.J.L. et and renal events in patients with type 2 diabetes and chronic kidney disease a 2019; Full Text Full Text PDF PubMed Scopus Google Scholar The of follow-up was We calculated for the standard randomized design the total GFR decline as the change from randomization until the on-treatment and chronic GFR decline as the change from the first to on-treatment using mixed effects models over a or We the total GFR decline for the design as the change from randomization to the end of We the total GFR decline for the active run-in randomized withdrawal design as the change from treatment-specific baseline GFR values (GFR at the of the run-in for placebo and end of run-in for atrasentan) until the on-treatment the total GFR decline for the and active run-in randomized withdrawal designs attempt to exclude the impact of the acute we compared the total GFR decline for designs with the chronic GFR decline for the standard randomized design. GFR was estimated using the Kidney Disease GFR over time was analyzed using mixed effects models with a for the effects and a for the J. et models for endpoints in clinical trials of chronic kidney Med. 2019; PubMed Scopus Google Scholar the Outcome trial, 2 endpoints were the total rate of change in GFR from baseline to the end of treatment at and and the total rate of change in GFR from baseline to the end of over treatment of and different of the mixed effects model were a linear mixed effects model with a single slope treatment a piecewise 2-slope linear model with a at the acute chronic GFR and total GFR slope from baseline to the end of and a piecewise linear model with at and at the on-treatment visit GFR slope from baseline to the end of The total GFR decline over the period or was using a of the acute and chronic the total as acute effect chronic slope the overall GFR slope over the first 2 the the total slope was calculated as the of the and We also used mixed models for analyses to assess active and placebo the trial, the GFR from randomization to the end of treatment were analyzed with the mixed effects models and models as for the Outcome In addition, a linear mixed effects model with a single slope treatment was to the GFR from baseline to the end of with treatment-specific GFR values at the randomization visit for the atrasentan and the of the active run-in period for the placebo The GFR values at randomization for the placebo were for baseline as values were patients were using analyses were in more of the different statistical models and is in the Statistical of the were as to the and M.J. to statistical Med. 2019; PubMed Scopus Google Scholar To assess the of different statistical in the change of GFR over for different effects and study GFR were using the linear mixed effects the GFR for the on the the time of that a treatment on a of for to the treatment and a of for to the treatment, and an to a acute in GFR for in the treatment for the effects for the treatment, the for the effects, and the were by the Outcome The effect for the was to a of to a chronic slope in of the treatment of The of the effect for the effect was as of the randomized in total slope as the change from randomization until the on-treatment the treatment and the in total slope the treatment and the treatment from baseline to end of in and randomized withdrawal in total slope the treatment and the treatment from baseline to end of the that the acute effect is fully the total slope with the design and the active run-in randomized withdrawal design are to the chronic slope the 2 in the chronic slope as the were for of the the study design randomized or active run-in randomized the from to in of and the of the acute effect effect of or The overall study was to be with the first to the or the placebo GFR were at and until treatment to end of study or of study the an GFR was included at the end of the the active run-in randomized withdrawal an GFR was included at the of the of study drug was by a time with a The statistical models that were to the from the Outcome trial and the trial were also to the of models with the rate of GFR change baseline and the end of as the were the for a more model the and statistical power of the different study designs and statistical were assessed on was as the the and estimated treatment effect in was as the of that in a of the treatment effect at the The baseline characteristics of the in the Outcome trial are in of GFR for compared with placebo GFR at by (95% confidence interval the GFR of randomized to placebo by 1.72 (95% and in randomized to empagliflozin by (95% to a in the chronic GFR slope of 1.72 (95% (Figure treatment in the GFR by (95% in the placebo and (95% in the empagliflozin The change in GFR slope from baseline to the end of the period estimated from a mixed effects model linear slope) was (95% in the placebo and (95% in the empagliflozin to a similar of 1.64 (95% compared with the treatment effect on the chronic GFR slope The model showed similar results with an estimated change in estimated GFR from baseline to end of over a period of (95% in the empagliflozin and (95% in the placebo to a of (95% estimated glomerular filtration Outcome in 2 with in a new effects of empagliflozin compared with placebo on GFR slope in the Outcome clinical to end of to end of study (95% (95% (95% (95% (95% (95% mixed effects model model 2-slope piecewise linear model was used to the and total slope in the baseline to piecewise linear model was used to the slope in the baseline to 2-slope piecewise linear model was used to the and total slope in the baseline to piecewise linear model was used to the slope in the baseline to from baseline to end of treatment GFR change GFR change confidence Outcome in 2 glomerular filtration mixed models for not in the treatment effect of empagliflozin compared with placebo on GFR decline in the design with the treatment effect on the chronic GFR slope in the standard randomized 2-slope piecewise linear model was used to the and total slope in the baseline to piecewise linear model was used to the slope in the baseline to in a new estimated glomerular filtration Outcome in 2 with confidence Outcome in 2 glomerular filtration mixed models for not in the treatment effect of empagliflozin compared with placebo on GFR decline in the design with the treatment effect on the chronic GFR slope in the standard randomized design. of the showed in the treatment effect when the change in GFR from randomization to end of was estimated using a single linear or model (Figure The statistical power was for the design than a standard randomized trial design that the treatment effect on GFR decline from randomization to end of treatment using a piecewise 2-slope linear model (Figure and the the treatment effect with a single linear slope model or outperformed the statistical power of the for the and of acute The also showed that for the model using the of 2 baseline and 2 GFR statistical power the of for the the power to a treatment effect of with for an intervention with a acute effect was when a single baseline and GFR was used and to when using the of 2 GFR (Figure power for different clinical trial Statistical power to assess treatment effects on glomerular filtration rate (GFR) decline in the design using a linear mixed models for or of outperformed the randomized clinical trial design using a piecewise 2-slope linear model or for the for with acute effects of Statistical power in the active run-in randomized withdrawal design using treatment-specific baseline GFR values versus GFR values at end of end of In the trial, were of GFR at characteristics of were to atrasentan or to placebo are in treatment with atrasentan during the active run-in GFR by (95% GFR by (95% in to GFR was in GFR of randomized to placebo by (95% and in randomized to atrasentan by (95% to a of 0.72 (95% in of atrasentan (Figure the total GFR slope at calculated for the standard randomized design with a piecewise 2-slope linear by (95% compared with the treatment-specific baseline GFR values and a linear mixed effects model linear the GFR decline was (95% in the placebo and was to (95% in the atrasentan to a similar of 0.77 (95% as compared with the treatment effect on the chronic GFR slope effects of atrasentan compared with placebo on GFR slope in the clinical to end of specific baseline to end of (95% (95% (95% (95% (95% (95% mixed effects model piecewise model slope slope from baseline to end of treatment GFR change GFR change confidence glomerular filtration mixed models for not with in the treatment effect of atrasentan compared with placebo on GFR decline in the active run-in randomized withdrawal design with the treatment effect on the chronic GFR slope in the standard randomized design. the chronic treatment effect of the in a new confidence glomerular filtration mixed models for not with in the treatment effect of atrasentan compared with placebo on GFR decline in the active run-in randomized withdrawal design with the treatment effect on the chronic GFR slope in the standard randomized design. the chronic treatment effect of the of the showed in the estimated treatment effect in the run-in active withdrawal design when estimated with a linear slope or (Figure was for assessed of acute the estimated treatment effect with a linear slope or model statistical power in the baseline design compared with the models the GFR from a randomization visit in the standard clinical trial design (Figure of change in GFR is used as a surrogate endpoint in clinical trials to obtain for new interventions for the treatment of The interpretation of treatment effects on GFR decline can be in GFR are often the of an intervention with and of compared with the rate of GFR change during chronic These acute GFR can be to effects or be the of in of or L.A. Greene T. A.S. end points for clinical trials of kidney disease J Am Soc PubMed Scopus Google Scholar The effect of an intervention on the rate of GFR decline during the chronic treatment phase is to the impact of the intervention on the of and the of the intervention to slow the progression of To the in clinical trial results in the setting of acute effects, the treatment effect on chronic GFR slope by using the first on-treatment GFR the acute phase as baseline be but is not as the of randomized study design the treatment can differ at time is to a 2-slope model for the acute and chronic phase of the treatment but requires of the length of the acute which is not clinical trials with that exert acute effects, alternative trial such as the design and active run-in randomized withdrawal be a to of of GFR decline. The results that the treatment effect on GFR change over time when calculated from randomization to the end of a single linear slope model or was similar to the treatment effect on chronic slope in the standard randomized trial design or calculated in a design with an active run-in phase with treatment-specific baseline GFR time points using a single linear slope The and active run-in randomized withdrawal designs also in a greater statistical power than the standard randomized trial design that the total GFR change from randomization to end of The statistical power was of the single linear slope or to GFR change over in the the statistical power of the model be the of 2 GFR were used at baseline and end of compared with using single GFR at time Kidney interventions are intended to slow the of over which is by the effect of treatments on the rate of GFR decline during chronic analyses on the chronic slope in the standard randomized design. treatment effects on total GFR slope showed a with effects on kidney outcomes compared with treatment effects on chronic slope in clinical trials the effects on chronic slope are likely more during follow-up and likely are patients in clinical practice. from 2 large clinical trials that the treatment effect from baseline until the end of are similar to the treatment effects on chronic slope in the standard randomized trial design. The of the design is that the randomization is in to treatment effects on chronic slope in the standard randomized design. the design for the acute effects, the statistical power is compared with the standard randomized design in which the treatment effect is calculated from randomization and the acute The design also be used for with acute effects as the GFR slope calculated in the standard randomized trial design may the treatment The of the design is that the acute treatment effects are fully during the Clinical trials with and have that the acute reductions in GFR on of interventions are during to of short-term in glomerular filtration rate long-term of renal Full Text PDF PubMed Google Scholar, V. et al.Canagliflozin and renal outcomes in type 2 results from the clinical Full Text Full Text PDF PubMed Scopus Google Scholar, H.J.L. B. et and kidney function decline in patients with type 2 a slope from the Outcome Am Soc 29: PubMed Scopus Google Scholar In addition, the time of the GFR be to the of the that be when a clinical trial using a design is that a follow-up period for is the of patients in a phase trial of CKD progression often more than a the follow-up of the first the follow-up be included in the efficacy analysis, analyses can be to include follow-up for the kidney or during GFR during the phase likely be of GFR values using the of the The active run-in and withdrawal design with treatment-specific baseline GFR is design type to for acute treatment effects. to the that design results in treatment effects that were similar to the treatment effects on chronic GFR slope and in statistical power compared with the standard design. design was at a National Kidney Foundation workshop by the US Food and Drug Administration and European and has been in a clinical trial, the trial was early due to drug A.S. Gansevoort R.T. Coresh J. et al.Change in albuminuria and GFR as end points for clinical trials in early stages of CKD: a scientific workshop sponsored by the National Kidney Foundation in collaboration with the US Food and Drug Administration and European Medicines Agency.Am J Kidney Dis. 2020; 75: 84-104Abstract Full Text Full Text PDF PubMed Scopus (330) Google V. et al.A phase randomized trial of in to kidney disease Am Soc Nephro. Google Scholar In the active run-in randomized withdrawal the active run-in period be to baseline design is not for for that slow In addition, of outcomes during the run-in phase or of patients due to effects is that be similar to active run-in the of effects, effects, be patients are to the intervention during the run-in period and to placebo at similar to the study be to in the GFR slope study has we trials with acute effects. We not have to trials with acute effects the alternative design the as for trials with acute effects. at are kidney trials with interventions that produce acute effects or using the baseline or design. We used from the Outcome trial, we that Outcome was a trial and not a kidney in the trial, atrasentan a acute We were to assess the of the active run-in randomized withdrawal design for interventions with acute effects. In study designs and are to compute treatment effects on GFR decline over time for interventions that exert acute effects that differ from their longer-term treatment effects. of designs and has and that be during the design phase and on the and drug the and of the acute effects is a for the design of clinical has as a for and and has from and from and are for and from and The of the Outcome trial is to of clinical study clinical and clinical study are to via the The clinical trial can be by in scientific and be and of a and Statistical and of a by the and the of the can be at time and the be for with We study and patients in the Outcome and The Outcome trial was sponsored by The trial was sponsored by was for specific

Topics & Concepts