Lead Investigator: Lesley Inker, Division of Nephrology, Tufts Medical Center
Title of Research Proposal: Chronic Kidney Disease Epidemiology – Clinical Trials Consortium (CKD-EPI CT).
Vivli Data Request: 4113
Funding Source: Additional Contracts or Consultancies – We have funding for this work from the National Kidney Foundation. We have submitted proposals to NIH for this work in the past but none are pending at present.
Potential Conflicts of Interest: LAI reports funding to Tufts Medical Center for research and contracts with the National Institutes of Health (NIH), National Kidney Foundation (NKF), Retrophin, Omeros and Reata Pharmaceuticals. She has consulting agreements with Tricida Inc. and Omeros Corp. Tufts Medical Center, John Hopkins University and Metabolon Inc. have a collaboration agreement to develop a product to estimate GFR from a panel of markers.
Summary of the Proposed Research:
Chronic kidney disease (CKD) is any condition that causes reduced kidney function over a period of time. CKD is common and harmful, but with few therapies. Conduct of randomized clinical trials (RCTs) in CKD is hampered, in part, because CKD often progresses slowly, with insufficient clinical endpoints (i.e., kidney failure or doubling of serum creatinine) in feasible time frames. In addition, there is insufficient data to guide sponsors or investigators in design of new RCTs.
The Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) was established in 2003 to address fundamental questions in CKD epidemiology using individual patient data and rigorous methods in clinical chemistry and statistics to create useful tools for research, patient care, and public health.
Proteinuria, or the occurrence of abnormal amounts of protein in urine, and decline in glomerular filtration rate (GFR), or the rate at which kidneys filter blood, are signs of kidney disease. A critical focus of CKD-EPI has been evaluation of surrogate endpoints for kidney disease progression, focusing on early changes in proteinuria (albuminuria) and GFR decline (time to 30% and 40% decline in eGFR and eGFR slope) as the key endpoints of interest, using individual patient meta-analysis of RCTs. The RCTs included in CKD-EPI Clinical Trials Consortium (CKD-EPI CT) have been identified through systematic literature searches with the goal to include all available studies, thereby decreasing potential bias of the analyses. The systematic search was initially conducted in 2007, then updated for the more recent analyses and was last performed in 2017. To our knowledge, this is the largest and most diverse collection of RCTs for CKD progression. This work was presented at a series of workshops (May 2008, December 2012, March 2018) sponsored by the National Kidney Foundation (NKF) in collaboration with the Food and Drug Administration (FDA) and European Medicines Agency (EMA).
Results to date were able to comment on the general validity of these endpoints, but were not able to make specific recommendations for how to select specific endpoints in specific trials and populations. The overall goal of our next phase is to develop evidence-based tools to rigorously evaluate trade-offs between alternative design strategies, including requirements for sample size and study duration, thereby improving feasibility and decreasing cost without inflating the risk of false positive conclusions. We anticipate that such tools could facilitate drug development even at the earlier stages as it enables sponsors to make calculations as to the optimal path from early phase investigations to regulatory approval. More efficient trials should spur an increased number of RCTs, speeding development of effective therapies for CKD.
Statistical Analysis Plan:
Study populations: RCT previously included in our meta-analyses and studies newly identified in this new phase of work.
Clinical Endpoints: The primary clinical endpoint will be defined as a composite of any of the following events over the full study duration: End-stage kidney disease (ESKD) (initiation of chronic treatment with dialysis or kidney transplantation); GFR < 15 mL/min per 1.73 m2; or sustained doubling of serum creatinine. In a sensitivity analyses, we will use an alternative clinical endpoint, defined as ESKD, eGFR < 15 mL/min per 1.73 m2 and time to 40% eGFR decline.
Change in Albuminuria: The measures of albuminuria varies across studies, with most prior studies measuring protein excretion rate (PER). Because recent guidelines recommend use of albumin to creatinine ratio (ACR), we will express all measures to ACR using a commonly used conversion: ACR mg/g (mg/mmol) = 0.6 * PER mg/day.4 Early change will be quantified as the change in log-transformed ACR from baseline to the measurement closest to 6 months (within 2.5 and 14 months) or 12 months (within 2.5 to 19 months).
GFR: GFR will be estimated using the CKD-EPI 2009 creatinine equation. Scr will be standardized to isotope dilution mass spectroscopy traceable reference methods using direct comparison or will be reduced by 5%.
Estimation of GFR slope: As we have done previously, we will use a linear mixed effects model based on a single slope starting at three months post randomization adjusted for baseline GFR.8,9 The model will consider a) informative censoring by applying shared parameter mixed effects models assuming Weibell distributions for the time from baseline to clinical events leading to early termination of GFR follow-up for trials with ≥ 15 ESRD or death events; b) between-subject variability in slopes by including random effects using mixed models; c) heteroscedasticity of within-subject variation in individual GFR measurements about each patient’s underlying slope with the use of power-of-mean (POM) variance structure and used a population level POM structure if the individual variation did not converge; and finally d) non uniform treatment effects by including different slope variances in the treatment group compared to the control group to accommodate. Simplified models without one or more of these features were used in cases where convergence could not be obtained with the full model. Because the underlying mixed-effects model is not strictly linear in the random effects nor in the model parameters, the model was fit using the SAS (version 9.4) nonlinear mixed-effects regression procedure, NLMIXED.
Trial level analysis: The trial level analysis requires two steps: intent-to-treat estimation of the treatment effects on the surrogate and clinical endpoints within each RCT and a meta-regression to relate the treatment effects on the surrogate and clinical endpoints across RCTs. These methods have been previously described.
1. Estimation of treatment effects on the endpoints:
a. Albuminuria: Treatment effects on ACR change will be estimated by performing analyses of covariance within each study, with log ACR change as the endpoint adjusting for treatment and log baseline ACR. Treatment effects on ACR will be expressed as geometric mean ratios (GMR). Treatment effects on the clinical endpoint will be estimated by Cox proportional hazard regressions to estimate hazard ratios for the treatment in each study.
b. Slope: Treatment effects on GFR slopes will be estimated using the mixed effects models described above and will be expressed as mean differences in the GFR slopes between the treatment vs control groups, in units of mL/min per 1.73m2/year.
c. Clinical Endpoint: Treatment effects on the clinical endpoint will be estimated by performing separate Cox proportional hazard regressions to estimate log hazard ratios for the treatment in each trial. Summary estimates of treatment effects will be obtained by use of random effects models.
2. Meta-regression to relate the treatment effects on the surrogate and clinical endpoints across RCTs:
a. Meta-regression: A Bayesian mixed effects meta-regression will relate the estimated treatment effects on the clinical endpoint to the estimated treatment effects on surrogate with study as the unit of analysis. For aim 1, the surrogate is either GFR slope or albuminuria. For Aim 2, the surrogate is both changes in albuminuria and GFR slope. For the description below, we therefore use the term surrogate which would differ by analysis.
The model relates the treatment effects on the two endpoints after accounting for random errors in the estimated effects in each RCT. The meta-regression supports validity of the surrogate endpoint if 1) the slope of the meta-regression line is statistically significant as defined by Bayesian credible intervals that do not cross 0, with a large magnitude, 2) the intercept is close to 0, implying absence of an average effect on the clinical endpoint when the treatment does not affect the surrogate, 3) the R2 is high, so that treatment effects on surrogate account for most of the variation in treatment effects on the clinical endpoint, and 4) the root mean square error (RMSE) is low, assuring low variation in the clinical endpoint given a fixed treatment effect on surrogate
Positive predictive value: We will use positive predictive values (PPV) to describe the uncertainty in predicting the treatment effect on the clinical endpoint from the treatment effect on the GFR slope. From the trial level meta-regression, we will compute 95% Bayesian prediction intervals and estimate the probabilities of clinical benefit (defined as HR < 1 or slope > 0) for different sized trials. We will compute the threshold associated with the smallest observed treatment effect on the surrogate that may assure a high probability of benefit of the treatment on the reference endpoint (ie clinical endpoint or slope which will vary by analysis), which we define as the treatment effect providing a PPV of 97.5%.
b. Subgroup and Sensitivity analyses: Where there is sufficient data, we will perform the trial level analysis for the primary analytical dataset overall and by subgroups defined by average study level of baseline ACR (< or ≥ 30 mg/g) or (< or ≥ 3.4 mg/mmol), GFR (< or ≥ 60 mL/min per 1.73m2), cause (diabetes and diabetic kidney disease, glomerular diseases, or other causes of CKD) and intervention. Because of differences in the ranges of treatment effects, accuracy in predicting the treatment effects on the clinical endpoint is best compared between subgroups using the RMSE.
Preliminary thoughts about analyses for adaptive design (to be updated pending results from above analyses): We will apply the joint trial-level analysis relating treatment effects on the clinical endpoint to treatment effects on both ACR change and GFR slope to develop an approach for Bayesian adaptive trial design. Instead of calculating positive predictive values (PPV) once, at the end of a new RCT, the trial-level meta-regression can be used to repeatedly update the estimated probability of clinical benefit at a sequence of interim time points as the RCT proceeds. At each interim time point, the estimated probability of benefit will account for updated estimates of the treatment effects on ACR change and GFR slope using albuminuria and GFR measurements obtained through that time. At each interim analysis, the adaptive trial either a) terminates with a conclusion of clinical benefit if the updated probability of benefit is sufficiently high, b) terminates with a conclusion of no benefit if the probability of benefit is sufficiently low, or c) continues to accrue new data if the probability of benefit is intermediate. We will develop a methodology for updating the meta-regression and estimated probability of benefit as data accumulates and for determining thresholds for early stopping at each analysis. Because precise estimates of treatment effects on ACR change can be obtained with shorter follow-up and smaller sample size than slope, in early interim analyses our procedure will be dominated by the estimated treatment effect on ACR change. Because the trial-level R2 is higher for slope than ACR change, the procedure will be progressively more strongly influenced by the estimated effect on GFR slope as sample size and follow-up become sufficient for well-powered slope analysis. We will also develop an approach for applying the meta-regression to calculate the adaptive design’s overall risk of falsely concluding clinical benefit (type-1 error).
A Randomized, Double-Blind Study to Compare the Durability of Glucose Lowering and Preservation of Pancreatic Beta-Cell Function of Rosiglitazone Monotherapy Compared to Metformin or Glyburide/Glibenclamide in Patients With Drug-Naive, Recently Diagnosed Type 2 Diabetes Mellitus
Study ID: NCT00279045
Sponsor ID: 49653/048
A Multicenter, Randomized, Double-Blind, Placebo-Controlled Study to Evaluate Cardiovascular Outcomes Following Treatment With Alogliptin in Addition to Standard of Care in Subjects With Type 2 Diabetes and Acute Coronary Syndrome
Study ID: NCT00968708
Sponsor ID: SYR-322_402
A Long Term, Randomised, Double Blind, Placebo-controlled Study to Determine the Effect of Albiglutide, When Added to Standard Blood Glucose Lowering Therapies, on Major Cardiovascular Events in Patients With Type 2 Diabetes Mellitus
Study ID: NCT02465515
Sponsor ID: 116174
- Collier W, Inker LA, Haaland B, Appel GB, Badve SV, Caravaca-Fontán F, Chalmers J, Floege J, Goicoechea M, Imai E, Jafar TH, Lewis JB, Li PKT, Locatelli F, Maes BD, Neuen BL, Perrone RD, Remuzzi G, Schena FP, Wanner C, Heerspink HJL, Greene T. Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI). Evaluation of Variation in the Performance of GFR Slope as a Surrogate End Point for Kidney Failure in Clinical Trials that Differ by Severity of CKD. Clin J Am Soc Nephrol. 2023 Feb 1;18(2):183-192. doi: 10.2215/CJN.0000000000000050
- Inker, L.A., Collier, W., Greene, T., Miao, S., Chaudhari, J., Appel, G.B., Badve, S.V., Caravaca-Fontán, F., Del Vecchio, L., Floege, J. and Goicoechea, M., 2023. A meta-analysis of GFR slope as a surrogate endpoint for kidney failure. Nature Medicine, pp.1-10. Doi: 10.1038/s41591-023-02418-0