A Combination of Change in Albuminuria and GFR as a Surrogate End Point for Progression of CKD

CKD remains an important worldwide public health problem, particularly in people with diabetes, with few therapies to slow the progression to kidney failure. Despite this unmet clinical need, there are fewer clinical trials in nephrology than in other fields of medicine, in part because of a paucity of end points for CKD progression that are accepted by regulatory agencies for drug development (1). Kidney failure remains the only accepted clinical end point, because earlier stages of kidney disease cause few symptoms. Surrogate end points could improve the feasibility of clinical trials by reducing the number of participants and shortening the duration of follow-up in the trial, but they must be validated to assure that conclusions about the efficacy of interventions on the surrogates in trials will apply to the clinical end points in practice. Criteria for validation include biologic plausibility, strong associations between the risk of occurrence of the surrogate and clinical end point in observational studies, and consistent effects of interventions on the risks of occurrence of surrogate and clinical end points in clinical trials.

A large body of work has investigated the role of changes in GFR and albuminuria as candidate surrogate end points for CKD progression. Biologic plausibility is well established for both. GFR decline is on the path to kidney failure, and regulatory agencies accept a large decline in GFR, such as a doubling of serum creatinine concentration (equivalent to a 57% decline in eGFR using the Chronic Kidney Disease Epidemiology Collaboration [CKD-EPI] equation) as a validated surrogate end point. A 2012 scientific workshop sponsored by the US National Kidney Foundation (NKF) and the Food and Drug Administration (FDA) proposed that a 40% eGFR decline and in some circumstances, a 30% eGFR decline could also be valid surrogate end points (2). Confirmation of the eGFR decline is important to minimize the effects of biologic variability and analytic variability (measurement error). However, these end points are less applicable for early stages of kidney disease and interventions that have a short-term (acute) effect on GFR decline that differs from the long-term (chronic) effect. In these circumstances, other surrogates may be helpful. (1) Early changes in albuminuria, because a rise in albuminuria occurs before GFR decline and is on the path to kidney failure in many causes of CKD. However, biologic and analytic variabilities are greater for albuminuria than for GFR, and the risk associations with the clinical end point are not as strong for albuminuria as for GFR decline. (2) The group average rate of GFR decline (slope), because it may have greater statistical power than time to specified decline in an individual. However, there are various methods for computing slopes, and interpretation of acute and chronic effects may be difficult. (3) A combination of change in albuminuria and GFR, but there is not as much empirical data on the combinations as on changes in albuminuria and GFR separately.

A 2018 scientific workshop sponsored by the NKF in collaboration with the FDA and the European Medicines Agency evaluated the role of early change in albuminuria and GFR slope as surrogate end points. The analyses of GFR slope are still under review, but the results of the analyses of albuminuria change have recently been reported in Lancet Diabetes and Endocrinology (3,4). The results show that a 30% decline in albuminuria over 2 years (adjusted for measurement error) was associated with an adjusted hazard ratio of 0.78 (95% confidence interval, 0.66 to 0.95) for the clinical end point (ESKD) in observational studies, and an intervention that reduced average albuminuria by 30% was associated with an adjusted hazard ratio of 0.73 (95% Bayesian credible interval, 0.55 to 0.95) in the clinical end point (doubling of serum creatinine, GFR<15 ml/min per 1.73 m², or ESKD) in clinical trials. The absolute risk implications were stronger for participants with higher baseline albuminuria. The authors concluded that a 30% reduction in albuminuria could be a reasonably likely valid surrogate for CKD progression under some circumstances and that it could be applicable to clinical trials of diabetic kidney disease.

In this issue of CJASN, Ohkuma et al. (5) evaluate the combination of changes in GFR and albuminuria and the risk of major clinical outcomes in a large study of people with type 2 diabetes in the Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation (ADVANCE) randomized clinical trial and the Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation Post Trial Observational study. The authors had previously reported that changes in albuminuria over 2 years were independently associated with risk of major cardiovascular events (defined as nonfatal or fatal myocardial infarction, nonfatal or fatal stroke, or cardiovascular death), major kidney events (defined as requirement for kidney replacement therapy or kidney death), and all-cause mortality (6). For this study, the authors related 2-year changes in albuminuria, GFR, or both to these outcomes. A total of 8776 (79%) of the 11,140 participants in the ADVANCE trial were included. Baseline mean (SD) age was 66 (6) years old, and duration of diabetes was 7.8 (6.3) years. Median (interquartile range) urine albumin-to-creatinine ratio (ACR) was 14.1 (7.1–37.1) mg/g, and mean (SD) eGFR (using the CKD-EPI equation) was 75 (17) ml/min per 1.73 m². Changes in ACR classified as ≥40% decrease (improvement) and ≥40% increase (worsening), compared to <40% change (minor change) they were observed in 2515 (29%) and 3249 (37%), respectively. Changes in eGFR were classified similarly, but they were much less common: an improvement in 304 (4%) patients and a worsening in 276 (3%) patients. Only 108 (1%) patients experienced a worsening in both ACR and eGFR, and 91 (1%) patients experienced an improvement in both ACR and eGFR. During a median follow-up of 7.7 years, 2191 (25%) patients developed the primary outcome (a composite of the following components): a major cardiovascular event (1392; 16%), a major kidney event (108; 1%), and all-cause mortality (1416; 16%). The major findings were that risk associations of ACR changes and eGFR changes with the composite outcome and with each component were significant (P value for trend <0.02 for each) and that risk prediction was significantly stronger for the combination of ACR and eGFR changes than for either change alone (as judged by differences in C statistic, integrated discrimination improvement, and net reclassification improvement). Of note, risk associations (multivariable adjusted hazard ratios and 95% confidence intervals) seemed to be stronger for worsening ACR or eGFR (multivariable adjusted hazard ratio, 1.32; 95% confidence interval, 1.19 to 1.46 and multivariable adjusted hazard ratio, 1.58; 95% confidence interval, 1.27 to 1.95, respectively) than for improving ACR or eGFR (multivariable adjusted hazard ratio, 0.96; 95% confidence interval, 0.85 to 1.07 and multivariable adjusted hazard ratio, 0.82; 95% confidence interval, 0.64 to 1.04, respectively), possibly because the baseline ACR and eGFR were only mildly abnormal. As expected, risk associations seemed to be stronger for kidney than cardiovascular outcomes, because albuminuria and GFR are primarily kidney measures. Risk associations were stronger for eGFR than for ACR, possibly because of larger measurement error in ACR than in eGFR. The authors concluded that a clinically meaningful increase in ACR and a decrease in eGFR over 2 years, independent and in combination, were significantly associated with higher risk of major clinical outcomes, and they suggested that a combined assessment of changes in ACR and eGFR could add prognostic utility compared with separate assessments of the changes in people with type 2 diabetes.

These findings provide some support for a combination of changes in albuminuria and GFR as a candidate surrogate end point, but many important questions remain to be answered. The primary outcome of the study by Ohkuma et al. (5) is broad, whereas changes in albuminuria and GFR are under consideration as surrogate end points only for the more narrow clinical outcomes of CKD progression. Evaluation in other studies with larger numbers of CKD events will be required. Minimizing biologic and analytic variability is important for evaluation of changes in albuminuria and GFR. Considering group average changes rather than individual changes can allow improved precision as can performing multiple measurements at selected time points or computing slopes. The study by Ohkuma et al. (5) does not directly address the consistency of the effect of the randomized intervention on changes in ACR and eGFR with the effect on the clinical outcomes in the ADVANCE trial. The primary and secondary results of the ADVANCE trial showed that both interventions (strict glycemic control and combined angiotensin-converting enzyme inhibitor and diuretic therapy) significantly improved the composite outcome, with generally stronger effects on changes in ACR than in eGFR (7–10). However, there are limitations to interpretation of the consistency of effects within a single clinical trial; thus, the evaluation for surrogacy requires consistency of these effects across multiple trials. Finally, although combining two end points can provide more specificity (protection against type 1 error [false positive conclusions]) compared with a single end point, by definition, there are fewer combined end points than single end points, resulting in lower sensitivity (increased type 2 error [false negative conclusions and loss of power]). Careful consideration of this tradeoff will be necessary in deciding how to combine ACR and eGFR changes. Strategies, such as sequential evaluation of end points (for example, testing for an early change in albuminuria with continued follow-up to test later changes in GFR), may be useful in some CKD settings.

In conclusion, we believe that there is value in considering combinations of changes in albuminuria and GFR as surrogate end points for CKD progression. The study by Ohkuma et al. (5) provides useful results showing improved risk prediction of clinical outcomes by using both changes compared with either alone in people with type 2 diabetes. Subsequent studies should define when such risk prediction is clinically useful and how and when the combination of changes in ACR and eGFR can be used as an outcome for clinical trials.

Disclosures

Dr. Levey reports financial support from Siemens (research and contracts to Tufts Medical Center). Dr. Coresh and Dr. Levey also report a provisional patent (filed August 15, 2014; precise estimation of GFR from multiple biomarkers; patent no. PCT/US2015/044567).