Creatinine and cystatin C measurements are widely used to assess GFR in clinical or research settings, in every relevant patient population (including kidney transplant recipients). Creatinine is a breakdown product of creatine from muscle cells and is largely removed from the blood by glomerular filtration. Therefore, the serum creatinine level, which is easily measured, is a useful reflection of GFR and is traditionally analyzed as an indicator of kidney function. Several decades ago, cystatin C — released by all nucleated cells — was also shown to be a reasonable marker of kidney function. Importantly, however, neither creatinine nor cystatin C meet the requirements of an ideal filtration marker (4). Creatinine levels depend not only on GFR, but also on muscle mass and dietary meat intake; cystatin C levels can also increase with corticosteroid treatment, which is frequently administered after kidney transplantation (5).

Creatinine clearance over a 24-h period can be used as a surrogate marker of GFR. However, creatinine is also secreted, which leads to GFR overestimation. Moreover, 24-h urine collection is burdensome, and inaccuracies in collection cause discrepancies between creatinine clearance and GFR. To overcome such limitations, equations have been developed to calculate eGFR (3, 6, 7). Common denominators in these formulas are serum creatinine and/or cystatin C levels; additional factors used in different combinations are sex, weight, age, and ethnicity. The most frequently used equations derive from the Modification of Diet in Renal Disease (MDRD) study (6) or the CKD-EPI formula (7), which have been validated in transplanted populations (8, 9). However, difficulties associated with measurements based on creatinine and cystatin C levels translate into limitations when applying these formulas, resulting in suboptimal agreement between eGFR and mGFR in the individual patient. Importantly, when it is essential to determine GFR precisely (e.g., when an anticipated decline in function will occur slowly, in longitudinal studies, or when there is considerable variation in non-GFR determinants of biomarkers employed for estimation), the EMA recommends mGFR rather than eGFR (3).

In kidney transplantation populations, the use of formulas such as MDRD and CKD-EPI to calculate eGFR (8, 9) is hampered by post-transplant differences in body composition (caused by protein catabolic effects of corticosteroids or edema) and inhibition of tubular creatinine secretion by trimethoprim (which is frequently administered). Nevertheless, they are widely used to evaluate GFR in clinical trials of kidney transplantation.

For accurate assessment of kidney function at a given time point in an individual, mGFR is undoubtedly the best available method (10), but this is difficult to undertake in routine practice. However, for comparing cohorts in clinical trial settings, the precise value in the individual may not be required: average eGFR values perform as well in study group comparisons as average mGFR values (11–15). The EMA/CHMP proposes that mGFR is performed in a prespecified subset of patients to confirm eGFR, with creatinine-based eGFR used in preference to cystatin C-based estimations (as creatinine-based eGFR is better characterized). Regardless of the methodology used for eGFR, the influence of confounders on data interpretation should be considered (3).

When selecting an outcome measure of kidney function for clinical research, it is important to know the strength of the relationship between each measure and the occurrence of hard endpoints such as ESRD. No studies show that one-time determination of mGFR is more strongly associated with future ESRD than eGFR, and mGFR has potential limitations, such as the complexity of evaluating large trials. Another major drawback of mGFR evaluation is the impossibility of calculating slopes over time (see below), which requires many repeated measurements.

There is limited agreement between decline in mGFR and eGFR. The Chronic Renal Insufficiency Cohort (CRIC) study compared associations between longitudinal changes (two measurements in 24 months) in eGFR and mGFR (urinary iothalamate clearance) with ESRD risk (16). The strongest association was found for changes in eGFR, which may be explained by higher precision (i.e., less variability) in GFR measurement using eGFR compared with mGFR. In a study of octreotide long-acting release in patients with autosomal-dominant polycystic kidney disease, similar slopes were observed for mGFR and eGFR in intervention and control groups during the 3-year follow-up period (17); comparable studies are not available for the transplant population. In accordance with the prevailing view in the nephrological community, we consider the use of eGFR as a suitable alternative for the practically cumbersome and more costly mGFR, in longitudinal studies and for comparison of investigational groups.

Doubling of the serum creatinine level is commonly an endpoint in clinical studies of kidney disease, including transplantation: it is considered analogous to a prespecified reduction in eGFR, and is often part of a composite outcome together with initiation of kidney replacement therapy and death from a renal cause (31, 32). However, doubling of serum creatinine is a late event in the progression of kidney insufficiency, and using this measure requires lengthy follow-up and/or very large numbers of patients. Alternative endpoints include the use of less steep (e.g., 30%) declines in eGFR, which have been strongly associated with ESRD (33, 34).

Consistent with findings observed in CKD (33), others have demonstrated that graft failure can be predicted not only by eGFR level at a given time point, but also by decline in eGFR within a relatively short period. For example, data from the Australian and New Zealand Dialysis and Transplant Registry (7,949 transplants) indicated that a ≥30% decline in eGFR between years 1 and 3 post-transplantation was strongly associated with subsequent death-censored graft failure and mortality (35).

The Clinical Trials in Organ Transplantation consortium showed that a 20–40% decline in eGFR at 3–24 or 6–24 months post transplantation was significantly associated with graft loss at 2–5 years, and with absolute eGFR at 5 years (36). The relationship between changes in eGFR and graft loss at 5 years was confirmed in the Genomics of Chronic Renal Allograft Rejection study (36). This suggests that the recommendation of a ≥30% eGFR decline over 2 years as being an acceptable outcome measure in CKD trials could be extended to studies involving kidney transplant recipients.

In some circumstances eGFR may not be a valid surrogate endpoint, although potential solutions for some situations have been formulated (34). For example, in kidney transplantation studies, eGFR decline may not be an ideal surrogate endpoint when drugs that affect muscle mass are administered, such as when muscle atrophy results from corticosteroid treatment (37). In addition, commencing or discontinuing angiotensin-converting enzyme inhibitors, angiotensin receptor blockers (ARBs), or calcineurin inhibitors can acutely affect GFR (38), which could have implications for trial design when eGFR is a surrogate endpoint: inclusion of a run-in period may be warranted.

Several RCTs involving kidney transplant recipients have used eGFR, or a change in eGFR, as a single primary endpoint (23–25, 27, 29, 30, 39). In most cases, eGFR was calculated using the MDRD or Nankivell formulas. Studies compared different immunosuppressive regimens, either de novo or starting at a specific time after transplantation. The fact that a significant difference in the primary outcome measure was absent in all but one study is probably not a weakness of the chosen endpoint, but rather illustrates that current immunosuppressive regimens are generally equivalent with respect to short-term graft function.

Therefore, in most circumstances, post-transplantation eGFR seems to be an appropriate endpoint for evaluating graft dysfunction. A systematic review concluded that post-transplantation eGFR (at 12 months) is associated with risk for overall or death-censored graft loss, and all-cause mortality, in univariate and multivariate analyses (40), although such a highly significant association does not necessarily translate into good predictive capability (41). The magnitude of the association between reduced GFR and outcomes was greater for death-censored graft loss versus overall loss, and for graft loss compared with overall patient mortality (40).

Clinical studies in diabetic nephropathy, hypertension and CKD, and polycystic kidney disease use the eGFR slope to evaluate the efficacy of interventions that aim to slow progression of kidney insufficiency (32, 42–45).

In 2017, the KDIGO conference Challenges in the Conduct of Clinical Trials in Nephrology concluded that change in eGFR over time was a practical and acceptable method for assessing kidney disease progression (46). KDIGO considered CKD stage and progression rate as determinants of the most useful outcome measure (46). When there is a markedly reduced kidney function (eGFR <45 ml/min/1.73 m²) and/or a rapid decline in GFR (>5 ml/min/1.73 m² per year), a composite endpoint consisting of a 30–40% decline in eGFR or the occurrence of ESRD failure is a robust and feasible outcome. When there is a slow decline in kidney function, using the GFR slope as the outcome measure may circumvent the need for lengthy follow-up and/or recruitment of very large numbers of patients. The same considerations can be applied in kidney transplantation studies.

In March 2018, several meta-analyses based on individual patient data were conducted in preparation for the workshop Change in Albuminuria and GFR as End Points for Clinical Trials in Early Stages of CKD, which evaluated surrogate endpoints for trials of CKD progression (47, 48). One meta-analysis (14 CKD cohorts) showed that, compared with a rapid decline in eGFR, a slower decline is associated with a lower risk of subsequent ESRD, even in participants with eGFR ≥60 ml/min/1.73 m² (49). A second meta-analysis of 47 RCTs evaluated the GFR slope as a surrogate endpoint for trials examining effects on CKD progression (50). This showed, with sufficiently large sample sizes, that treatment effects on the GFR slope from baseline and 3-month follow-up of 0.5–1.0 ml/min/1.73 m² per year strongly predicted benefits on clinical endpoints such as doubling of serum creatinine, GFR <15 ml/min/1.73 m², or ESRD. Using statistical simulations of GFR trajectories based on data from the 47 RCTs, the GFR slope performed better than clinical endpoints when patients’ initial GFRs were high and not acutely affected by treatment (51). Although cohorts that formed the basis of these meta-analyses did not include kidney transplant recipients, there was a large variation in underlying causes of CKD. This makes it reasonable to assume that the study conclusions would apply to patients with graft dysfunction as a particular form of CKD.

A theoretical advantage of using the eGFR slope as the endpoint, rather than a time-to-event endpoint (e.g., ESRD), is that the decision to initiate dialysis or (re)transplantation can be affected by factors other than GFR. The effect of an intervention on the eGFR slope may therefore better reflect the true effect on kidney graft function.

Importantly, suitability of the eGFR slope as surrogate endpoint depends on patterns of acute and chronic phases of the slope, in the context of the specific disease and potential pharmacodynamic impact of the investigational compound on the slope. The total eGFR slope reflects the slope from time of randomization, i.e., across the entire study period; the chronic slope calculation starts later and is less affected by acute changes in eGFR during the initial phase post randomization. In this respect, transplant recipients likely differ from patients with native CKD. The chance of a non-linear decline in eGFR is probably higher in kidney transplant recipients as a result of acute events such as infection, rejection and initiation or withdrawal of drugs that have acute effects on kidney function, including immunosuppressive agents. This can also impact the number of eGFR measurements needed to calculate the eGFR slope. Such information should be available to evaluate the usefulness of eGFR slope after kidney transplantation and to judge the validity of the chronic slope, versus the total slope. The EMA has indicated that the total slope is generally favored over the chronic slope, because the total slope minimizes possible biases introduced when post-randomization events (e.g., death) or acute changes in eGFR on investigational drug initiation are not considered (48).

The only study in kidney transplant recipients to use the eGFR slope as the primary outcome measure was a clinical trial of bortezomib in late antibody-mediated rejection (AMR) (29). In preparation for a placebo-controlled trial investigating clazakizumab as a treatment for chronic AMR, a data-modeling exercise evaluated the relationship between rate of eGFR decline and risk of graft failure. This was a historical prospective cohort study investigating the relationship between change in eGFR (estimated using the MDRD 4 equation) and risk of graft failure in kidney transplant recipients diagnosed with acute/active (a)AMR (52). The primary analysis used data from 91 patients with biopsy-proven aAMR and baseline eGFR ≥25 ml/min/1.73 m², with a minimum of 3 years’ follow-up data. Both a linear mixed-effects model to describe eGFR decline and a joint model, involving longitudinal eGFR and its rate of decline, were constructed. The joint model predicted that the baseline eGFR and its rate of decline (slope change per month) following an aAMR diagnosis significantly predicts risk of both death-censored and all-cause graft failure. Using the modeling results for all-cause graft loss, the mean eGFR decline from baseline to month 12 after AMR diagnosis was 0.75 ml/min/1.73 m² per month. Using these data, and assuming a 50% reduction in the rate of eGFR decline with clazakizumab, a sample size was calculated for an interim analysis of the 52-week eGFR endpoint. This is an example of how modern statistical techniques can optimize study design in a specific patient population. For these techniques to be used, data are required on the natural course (i.e., without intervention) of kidney function in the population of interest, which should be sufficiently large to accurately define the natural disease course.

Few RCTs in kidney transplantation have used a change in mGFR as the primary endpoint; mGFR assessment was based on iohexol or ⁵¹Cr-EDTA clearance (14, 15). Although eGFR values were markedly higher than mGFR values in both studies, the conclusions were not affected when eGFR was used instead of mGFR; thus, no advantage for using mGFR was demonstrated. The BENEFIT and BENEFIT-EXT studies (11, 12) and the Spare The Nephron trial (13) also used mGFR as primary endpoint; the latter illustrated the difficulty in obtaining mGFR data, as no values were available in nine of 112 (8.0%) and 25/116 (21.6%) patients in the mycophenolate mofetil (MMF)/sirolimus and MMF/calcineurin inhibitor (CNI) groups (13). In the BENEFIT study, missing mGFR values were imputed using eGFR values, although the exact magnitude of this imputation is not available (11). Measured GFR is no longer used as a kidney function endpoint in kidney transplantation studies because of limitations mentioned earlier and availability of different methods. In summary, ESOT considers eGFR as the most useful marker for post-transplantation kidney function.

While proteinuria can be considered as a surrogate marker for severity of glomerular damage, proteinuria can also directly contribute to kidney injury and decline in kidney function (53). Although not formally proven, this finding probably also holds true for kidney transplantation populations.

In a large cohort (31,372 individuals from a general population; two or more ACR measurements in 2 years), a fourfold increase in ACR was associated with a threefold heightened risk of ESRD during a median 3 years of follow-up (54). A reduction in proteinuria is also known to protect patients with various forms of renal disease from kidney function decline (55).

In IgA nephropathy, proteinuria is the most widely recognized risk factor for progression to ESRD. Analysis of 13 controlled intervention trials in IgA nephropathy showed an association between treatment effects on percentage reduction of proteinuria, and on a composite of time to doubling of serum creatinine, ESRD, or death (56). Similarly, a meta-analysis of 41 randomized trials in CKD supported use of change in albuminuria level as a surrogate endpoint for CKD progression, particularly in patients with high baseline albuminuria (57). A European Regulator’s perspective on the potential of change in albuminuria as endpoint for clinical trials in CKD has been published (48).

Unlike specific diseases in native kidneys, causes of proteinuria after kidney transplantation are diverse. During the early months after transplantation there may be some contribution from proteinuric native kidneys, but major causes of proteinuria are chronic rejection, recurrence of proteinuric disease, or de novo glomerulopathy. Nevertheless, an association between proteinuria and progression to ESRD (demonstrated particularly in diabetes mellitus and IgA nephropathy) has been observed in several cohorts of kidney transplant recipients (58–60). Such findings were confirmed in a post hoc analysis of the FAVORIT trial (3,511 participants followed over 4 years), which found that an elevated baseline ACR is independently associated with graft failure, cardiovascular disease, and death (61).

In contrast to studies investigating chronic disease in native kidneys, no studies in kidney transplantation have demonstrated a beneficial effect of proteinuria reduction on progression to ESRD. A clinical trial of ramipril versus placebo in 213 kidney transplant recipients with (mean proteinuria ≥0.2 g/day) showed no difference in the primary outcome (a composite of doubling of serum creatinine, ESRD or death), despite some reduction in mean proteinuria (26). Evidential support for using proteinuria as a post–kidney transplantation endpoint is weaker than evidence for its use in studies of CKD in native kidneys. More evidence is needed from larger cohorts before proteinuria could be proposed for use in kidney transplantation clinical trials.

The KDIGO 2012 guidelines updated the classification system for CKD to include albuminuria, stating that, for the general population, risk of adverse outcomes (mortality, progression to ESRD) at a given eGFR increases with higher levels of albuminuria.

Although studies indicate its promise (26, 61), the combination of eGFR and proteinuria (either as absolute values or as changes from baseline) has not been used as an endpoint in kidney disease clinical trials. Data from the ADVANCE study showed that, in patients with type 2 diabetes mellitus, the 2-year change in eGFR and ACR more strongly predicted the risk of ESRD during a median follow-up of 7.7 years than either of these changes alone (62). A limitation of this study is that the combination of worsening of eGFR and increase in urinary ACR, as well as major kidney events, occurred in only 1% of the study population. Additional studies are required to determine when the combination of changes in proteinuria and eGFR can be used as a surrogate outcome in a broad spectrum of kidney diseases.

The interaction between eGFR and proteinuria — as demonstrated in participants with diabetes in the ADVANCE study — was also observed in kidney transplant recipients. An analysis of linked databases in Canada (N = 900) found that rates of death-censored graft loss also increased with lower levels of kidney function at 1 year after transplantation (63). Moreover, within each eGFR category, adjusted rates increased with higher levels of proteinuria. Risk of death-censored graft loss was 49-fold higher for kidney transplant recipients with an eGFR of 15–29 ml/min/1.73 m² and severely increased albuminuria, compared with recipients with an eGFR ≥60 ml/min/1.73 m² and normal protein excretion (62) Although the integration of proteinuria and eGFR assessment has been shown to be a very good predictor of graft outcome (61, 63), more data must be collected before this combination can be advocated as a study endpoint in kidney transplantation clinical trials.

Finally, the causes of long-term graft failure are complex, and there are good arguments to capture this heterogeneity in more integrated composite scoring systems. For this, we refer to the paper in this supplement on surrogate endpoints (64).

This paper provides a contemporary discussion of graft functional parameters as primary endpoints in clinical trials as endpoints for clinical trials in kidney transplantation. ESOT has come to the following conclusions:

• The CHMP agreed that endpoints to assess efficacy of medicinal products to slow progression of chronic renal insufficiency (3) can be adopted to trials of kidney transplantation.