Modernization of Chronic Allograft Injury Research: Better Biomarkers, Better Studies, Better Outcomes

Abstract

Despite dramatic improvements in acute rejection rates and short-term allograft survival, long-term allograft survival remains unchanged in the modern era, largely due to chronic allograft injury, a progressive disease that is common across all solid organ transplantation but has no proven treatment. Studies of novel diagnostic and therapeutic strategies for chronic allograft injury have been relatively sparse, in part due to the time and expense required to conduct traditional long-term clinical studies of a variably progressive disease. In this article, we review the pathophysiology of chronic allograft injury, including recent insights into key mechanisms of the disease. We discuss the barriers to progress in chronic allograft injury research and present alternative approaches to study design that could accelerate improvements in diagnosis, prevention, or treatment of the disease. We integrate these approaches with emerging biomarkers and surrogate endpoints into a model clinical study of chronic renal allograft injury, providing a framework for modern study design in solid organ transplantation.

INTRODUCTION

Since the 1980s, increased potency of modern immunosuppression has dramatically improved short-term acute rejection rates with concurrent improvements in 1-year patient and graft survival (1). A common and recurring concern among transplant professionals is that expected similar improvements in 5- and 10-year graft survivals have not been realized (1–3), thus preventing necessary improvements in long-term patient survival in end-organ failure. The probability of death after kidney transplant is 37% and 57% at 5 and 10 years, respectively (4), which is unfortunately comparable to mortality rates from brain, kidney, or colorectal cancer (based on the Surveillance, Epidemiology, and End Results database from http://www.cancer.gov, accessed 1/20/2016). The majority of allograft failures leading to this increased mortality can now be attributed to an entity known as chronic allograft injury, which is therefore a principal barrier to progress in our field (2,5).

There are currently no effective therapies for chronic allograft injury and existing diagnostic strategies are largely inadequate. Tissue biopsies remain the gold standard for diagnosis, but are expensive, invasive, and often suffer from under-sampling errors and variance in interpretation (6). Other less-invasive diagnostic methods (e.g., creatinine-based estimates of kidney allograft function) are often inaccurate and provide no therapeutic target or insight into mechanisms of disease. Recent studies of therapeutic interventions for chronic allograft injury have shown promise, but most are beset by weaknesses in study design, reproducibility, and generalizability (7–11).

Much of this lack of progress in chronic allograft injury research can be attributed to inherent difficulties in using a traditional clinical trial design that employs clinical outcomes for studying the efficacy of a given diagnostic test or therapeutic agent. Chronic allograft injury is a heterogeneous, irreversible disease with a long latency period (developing as late as 10 years post-transplant in some subjects), thereby limiting the feasibility of clinical trials with chronic allograft injury as a primary endpoint. There is an urgent need for novel study designs that implement alternative outcomes, surrogate endpoints, and study populations and that will accelerate discoveries in the field that translate more efficiently to the clinical realm. These of course, will require validation in traditional clinical outcome studies before being accepted into practice.

The purpose of this manuscript is to briefly examine the pathophysiology of chronic allograft injury, including the historical evolution of our understanding of the disease and its contribution to late allograft failure, focusing on kidney transplantation as the most prevalent transplant. We will then explore the barriers to progress in chronic allograft injury research, including a critical discussion of approaches to study design (including a review of emerging biomarkers and alternate surrogate endpoints) that could speed advancements in diagnosis, prevention, or treatment of the disease. Finally, we will illustrate these principles in a proposed “mock study” in chronic renal allograft injury, providing a framework for modern study design in solid organ transplantation.

AN EVOLVING UNDERSTANDING OF CHRONIC ALLOGRAFT INJURY AND LATE ALLOGRAFT FAILURE

Chronic allograft injury is common to all solid organ platforms, which share many common clinicopathologic features of the disease (Table 1). The disease may be best characterized as an “obliterative lumenopathy,” manifesting in a progressive concentric narrowing and fibrosis of endothelial- and/or epithelial-lined structures (e.g., blood vessels and tubular lumens) that are critical to maintaining allograft function (12–14). Endothelial cells line the macro- and microvasculature of the allograft and serve as the final barrier between the external environment and the internal graft milieu. Therefore, the allograft endothelium is the final common target for a myriad of acute and chronic insults that are known contributors to chronic allograft injury, including: innate, humoral, and cell-mediated immunity; drug toxicity; ischemiareperfusion injury; infections; and, recurrent primary organ disease. This cumulative injury results in chronic endothelial dysfunction and inflammation that leads to local tissue hypoxia, enhanced alloimmune responses, and progressive tissue fibrosis with loss of allograft function (15,16).

Table 1.

Histologic features of chronic allograft injury across solid organ transplantation.

Solid Organ Platform	Endothelial and Vascular Lesions	Epithelial and Tubular Lesions
Kidney	Transplant glomerulopathy Chronic active antibody-mediated rejection Allograft vasculopathy Chronic active T-cell mediated rejection	Interstitial fibrosis and tubular atrophy
Heart	Cardiac allograft vasculopathy Transplant arterio-/atherosclerosis Stenotic microvasculopathy
Lung	Pulmonary vascular atherosclerosis Miscrovascular destruction	Bronchiolitis obliterans syndrome Insterstitial fibrosis
Liver	Allograft vasculopathy Obliterative/foam cell arteriopathy	Vanishing bile duct syndrome

In kidney transplants, chronic allograft injury is characterized by the presence of one or more major histologic lesions: interstitial fibrosis/tubular atrophy, transplant glomerulopathy, or allograft vasculopathy (17–19). It has been detected in allograft biopsies as early as a few months or as late as 10 years post-transplant, providing an extremely wide diagnostic window in which to study the disease. The nomenclature for chronic renal allograft injury has evolved to reflect important changes in our understanding of the disease. Terms such as “chronic rejection” and “chronic allograft nephropathy” have given way to chronic allograft injury to better represent the totality of immune and non-immune mechanisms of disease (17). There is a renewed focus on assigning a specific cause or causes for chronic allograft injury in each patient, as recent studies have demonstrated that few cases are truly idiopathic (5,20,21). This is in contrast to the historical belief that much of late allograft failure was due to either severe, irreversible acute rejection or non-specific interstitial fibrosis and tubular atrophy.

Both surveillance and indication biopsy studies have elucidated the most common causes of chronic renal allograft injury and late allograft failure. Early allograft failure (< 1 year post-transplant) is often attributable to recurrent glomerular and nonglomerular diseases, infections, inflammation, and irreversible acute rejection. Chronic allograft injury lesions are seen much less often in this early time frame. In contrast, late allograft failure is often due to causes directly related to chronic allograft injury lesions: glomerular disease (recurrent/de novo glomerulonephritis or transplant glomerulopathy), chronic active antibody-mediated injury, and specific yet discrete causes of interstitial fibrosis and tubular atrophy [BK virus (BKV)-associated nephropathy (BKVAN), recurrent acute rejection, recurrent acute pyelonephritis] but not isolated calcineurin-inhibitor toxicity (21–24). A recent study has suggested that the presence of calcineurin inhibitor toxicity is less of a “death sentence” than previously considered, and that long term allograft failure due to alloantibody has been overlooked (22). These concepts in kidney transplantation are now being explored and confirmed in heart, lung, and liver transplantation (25–27).

Based on these recent insights, chronic allograft injury can be viewed as the epicenter for multiple distinct pathophysiologic processes that all contribute to allograft failure with variable importance and timing (24). While the histologic lesions of chronic allograft injury represent the final endgame of these processes, the path to late allograft failure includes much interpatient variability. This raises significant concerns for the approach of using a histological diagnosis of “chronic allograft injury” or an equivalent historical term as a primary endpoint in prospective clinical studies. For instance, an interventional trial of a novel biologic agent to reduce biopsy-proven chronic renal allograft injury will achieve widely variable results depending on whether the study population is enriched for subjects with antibody-mediated injury, a key contributor to the disease.

BARRIERS TO SUCCESS IN CLINICAL STUDIES OF CHRONIC ALLOGRAFT INJURY AND LATE GRAFT FAILURE

Clinical Endpoints

In the current clinical research climate, human investigation entails substantial scrutiny and oversight from reviewers, regulators, funding agencies, pharmaceutical companies, and other stakeholders (28). The desire and need to generate data that provide meaningful conclusions regarding the value of new therapies in a timely fashion limits the options for clinical endpoints. This is most apparent in the study of effective therapies to address long-term transplant outcomes. In the past, investigations of new immunosuppressant efficacy focused on primary clinical endpoints such as patient and graft survival and reduction in acute rejection rates. As these early outcomes have improved substantially with infrequent episodes of patient death, graft loss, or acute rejection, clinical trials utilizing non-inferiority designs have evolved as the typical pathway for drug approval. These non-inferiority trials typically compare the effectiveness of an established treatment to a novel treatment rather than using a placebo-controlled, superiority trial design. Non-inferiority trials are designed to demonstrate that any difference between the two treatments is negligible enough to conclude that the novel drug has at least some effect that is similar to the established treatment. Importantly, these trials only aim to demonstrate similarity between the established and novel treatments, regardless of whether they are effective or ineffective versus placebo. Typically, the clinical efficacy is estimated from prior placebo-controlled studies of the established treatment such that one can conclude that both established and novel treatments are similarly more effective than placebo. Given these challenges, the non-inferiority trial is useful and ethical in transplant studies where there is a lack of equipoise between an established treatment and placebo (29).

Alternatively, one could employ a primary composite outcome using the clinical endpoints noted above. Such a strategy may allow for performance of studies with smaller sample sizes and reduced costs compared to those based on a single, low-frequency primary endpoint. For example, to detect a 20% reduction in 1-year acute rejection with 80% power (assuming an alpha of 0.05 and 15% annual incidence of acute rejection), a total sample size of over 4,000 subjects is required. If a triple composite outcome of death, graft loss, and acute rejection is used, a 20% reduction in the outcome with 80% power (assuming a 15% incidence of acute rejection, 10% incidence of graft loss, and 5% incidence of death in the first post-transplant year), a total sample size of only 1,800 subjects may be required, significantly shortening the time to study completion and reducing costs.

Despite these advantages, there are difficulties using this strategy. For example, each component of the composite is weighted equally in the analysis. However, an intervention that reduces the incidence of a triple composite outcome by reducing acute rejection rates, for example, is less impactful than an intervention that achieves the same endpoint by reducing patient mortality (30). Therefore, the typical practice is to report findings on each component of the composite as secondary clinical endpoints in addition to the combined primary outcome.

Secondly, when using a composite outcome in a longitudinal study, one must consider the influence of one component of the composite on the other endpoints. For example, death with a functioning graft prohibits that same patient from being at risk for subsequent acute rejection or graft failure in the study period, a phenomenon referred to as competing risks or informative censoring (31). Studies often account for this by censoring graft failure for death, but this construct requires the censoring event (death) to be independent of the outcome event (graft failure), an invalid assumption. In other words, death is a competing risk for allograft failure and should be handled in the analysis phase of a study using a competing risks model or cause-specific hazards model (31). However, these analyses can be cumbersome, time-consuming, are often not available in standard statistical software packages, and require input from a trained statistician.

Alternate Clinical and Patient-Centered Outcomes

As reviewed above, clinical studies in solid organ transplantation are often centered on graft outcomes such as acute rejection and short- and long-term graft survival since prolonging graft survival is akin to prolonging patient survival. This is certainly true in kidney transplantation, where a return to dialysis due to transplant failure is associated with substantially increased mortality risk (2, 3). However, nearly half of kidney allograft failures are due to death of recipients with a functioning allograft, most often attributed to cardiovascular causes (32,33). Another concern is the 1.7-fold increased rate of suicide in kidney transplant recipients compared to the general population, most often occurring less than 5 years post-transplant (32). These data identify a need for additional studies beyond traditional endpoints of acute rejection and graft loss; namely trials aimed at understanding and reducing cardiovascular risk and increasing quality of life in transplant recipients.

Cardiovascular risk is an important consideration for kidney transplant recipients since it remains the major contributor to mortality just as in native chronic kidney disease patients. Therefore, across the spectrum of native and transplant kidney disease, patients are more likely to die with kidney disease than of kidney disease. Cardiovascular risk studies are an excellent format for deploying surrogate endpoints that have been rigorously validated as predictors of clinical endpoints such as cardiovascular morbidity and mortality. In an effort to improve the totality of late graft loss from both allograft failure and death with a functioning graft, studies designed to improve surrogates for cardiovascular risk must play a significant role. Such studies may include interventions to reduce blood pressure, left ventricular mass, cholesterol, or vascular stiffness that have been explored in native kidney disease (34–38).

In other chronic disease fields such as oncology, investigators are shifting their focus toward improving quality of life metrics in addition to traditional notions of prolonging patient survival. Several studies have demonstrated that kidney transplant recipients have lower health-related quality of life compared to the general population, which is independently associated with reduced graft and patient survival (32,39–42). Accordingly, newer trials are incorporating longitudinal changes in health-related quality of life as important clinical endpoints, a practice strongly supported by the Food and Drug Administration (FDA) (43–45). Routinely gathering data on quality of life when testing new therapies in clinical trials will enable greater recognition of the complete benefit of such treatments, not only focused on quantity but also quality of patients’ survival.

Important Subgroup Differences

Further complicating the proper design of future clinical transplant studies are the racial disparities in access and outcomes that continue to plague African American and Hispanic populations. In these subgroups, waitlist times are longer, access to transplantation is lower, and outcomes remain inferior compared to Caucasian recipients. Importantly, the causes of accelerated graft loss in these subgroups are biologically complex and cannot be explained by poor socioeconomic status (46–48). Researchers are increasingly encouraged by funding agencies to increase proportions of these minority subgroups in overall study populations. Specific studies in these racial subgroups are essential to better understand how to tailor therapies that are selectively beneficial for them versus the general transplant population.

Vulnerable populations such as pediatric recipients are an important subgroup biologically distinct from adult recipients and must be properly studied as a separate entity. Our understanding of the pediatric response to an allograft has evolved over the last several years. Even though the immunosuppressive protocols, diagnostic tests, mechanisms of allograft disease, and outcome measures are similar to adults, it is clear that children have several important differences from adults that necessitate pediatric-specific studies rather than application of findings in large adult cohorts to pediatric recipients (49). Compared to adults, children have a more naïve immune system with a reduced fraction of antigen-specific T-cells, B-cells, and macrophages. This modifies their response to immunosuppressive drugs and alters their risk of acute rejection and chronic allograft injury. The causes of end-stage organ failure in children are often quite different, favoring metabolic, structural, or glomerular diseases over the typical causes in adults of diabetes and hypertension. Children are also more likely to be naïve to important post-transplant infections, such as Epstein-Barr virus (EBV) and cytomegalovirus (CMV), with important implications for long-term outcomes (49). Conducting trials in pediatric transplantation is a challenging endeavor, as the highest-volume programs in the nation perform just a few dozen solid organ transplants each year. Thus, generating meaningful conclusions in clinical pediatric transplant studies will require coordination of large multicenter consortia.

Surrogate Endpoints

Another solution to shorten the time to reach meaningful conclusions in clinical studies is the use of surrogate endpoints for the primary clinical outcome. A surrogate endpoint is defined by multiple regulatory and funding agencies as a biomarker intended to substitute for a clinical outcome and to reasonably predict clinical benefit of an intervention under study (30). In other words, the effect of the intervention on the surrogate endpoint should predict the effect of the intervention on the primary clinical outcome. An example from the chronic kidney disease literature is the improvement in blood pressure, cholesterol, or vascular stiffness as a surrogate endpoint for cardiovascular risk and long-term patient survival (34,35,37,38). This strategy allows for shortened time intervals to determine the efficacy of a given intervention, but the findings must eventually be validated for the related clinical outcome to confirm the predicted clinical benefit (50–52). Since chronic allograft injury and graft failure are often late events occurring more than 5 years post-transplant, surrogate endpoints are an attractive approach to efficient trial design.

Numerous surrogate endpoints for chronic allograft injury and late graft failure have been considered in kidney transplant research. Problems arise when there is poor correlation between the surrogate endpoint and the primary clinical outcome, as is often the case when the surrogate is non-specific for the outcome of interest. For example, absolute values or longitudinal changes in kidney function are commonly deployed as noninvasive, inexpensive surrogate endpoints for late graft failure (20,53). However, there are many contributors to changes in kidney function over time, limiting their specificity for graft failure. Proteinuria has been used as a surrogate endpoint for chronic allograft injury but can be influenced by multiple pathophysiologic processes that may contribute variably to progression of the disease (54). Both surveillance and indication biopsy findings such as subclinical inflammation, C4d deposition, and microvascular injury have been shown to correlate with future development of chronic allograft injury and graft loss (55–59) and could serve as effective surrogate endpoints if not for the expense, sampling, and interpretation issues with kidney biopsies. Given the emerging importance of antibody-mediated damage in chronic allograft injury, there is growing interest in measuring changes in donor-specific antibody (DSA) presence and/or intensity as a surrogate endpoint for the disease. However, like the aforementioned surrogates, DSA measurements are limited by their lack of correlation with clinical disease and lack of reproducible results between laboratories (23,60,61). In addition to these commonly used surrogates, recent translational studies have identified numerous candidate biomarkers for chronic allograft injury that require further validation before being widely accepted, as summarized in Table 2 (23,24,53,62–89). In general, given what we know about the complexity of the immune response in individual transplant recipients, it is unlikely that a single biomarker will be a sufficient criterion for enrollment or surrogate endpoint for determining efficacy for all patients. More likely, panels of biomarkers will be required for a comprehensive understanding of which subgroups of patients are the best candidates for enrollment in a particular clinical study (6,90). Critically, these suggested biomarkers require more thorough validation.

Table 2.

Selected biomarkers and surrogate endpoints for chronic allograft injury and late graft failure in kidney transplantation.

Biomarker/surrogate	Source	Characteristics/peformance	Refs.
Clinical
Allograft Function (eGFR)	Blood	Decline in function over time is more predictive of late graft loss than values at a single time point	53
Proteinuria	Urine	Higher levels after 1 year post-transplant are associated with late graft loss, independent of function and histology	62
DSA	Blood	Detection of de novo DSA associated with chronic allograft injury scores and late graft loss; clinical benefit of therapies to reduce DSA intensity requires further validation	23, 24, 63, 64
Viral reactivation	Blood, urine	Subclinical EBV, CMV, and BKV viremia have been associated with chronic allograft injury and late graft loss	65, 66, 67
Histology	Tissue
Total interstitial inflammation		Banff ti score in late for-cause biopsies is associated with late graft loss	68
Subclinical inflammation		Banff i, t scores in early protocol biopsies are associated with chronic allograft injury and graft loss	69, 70, 71
Microcirculation injury		Banff g, ptc scores are associated with development of chronic allograft injury and late graft loss, especially in concert with de novo DSA	72, 73
Banff chronic injury scores		Banff cg, ct, ci, cv scores in early and late biopsies are predictive of graft loss	70, 74
Peritubular capillary C4d deposition		C4d+ for-cause biopsies are associated with late graft loss, especially in concert with de novo DSA	75, 76
Translational (individual biomarkers)
Antibodies to self-antigens	Blood	Association of early and late responses to self-antigens in chronic allograft injury and late graft loss	77, 78
CXCL9	Urine	Lower urine value at 6 months post-transplant is associated with lower risk for future rejection and decline in allograft function	79
CCL2	Urine	Higher urine value at 6 months post-transplant is associated with chronic allograft injury scores (Banff ci, cg) at 24 months and late graft loss	80, 81
FGF23	Blood	Higher circulating values at a single time point are associated with chronic allograft injury and late graft loss	82, 83
Translational (biomarker panels)
Gene expression microarray	Blood, urine, tissue	Endothelial-associated transcripts are associated with chronic allograft injury scores and late graft loss	84, 85, 86, 87
Proteomic array	Urine	Urine peptide patterns associated with chronic allograft injury scores and late graft loss	88, 89

Importantly, given the long latency period between diagnosis of chronic allograft injury and graft failure, chronic allograft injury itself can also serve as a surrogate endpoint, in addition to its role as a specific clinical outcome (91). As reviewed above, due to complexity and heterogeneity of the disease, it may be more advantageous to target improvement in a specific component of chronic allograft injury as a surrogate endpoint, such as the interstitial fibrosis (ci) or tubular atrophy (ct) Banff severity scores, rather than a generic diagnosis of chronic allograft injury (92). Therefore, study designs could employ a short-term surrogate endpoint such as subclinical inflammation on an early surveillance biopsy that is linked to an intermediate surrogate endpoint such as the ci/ct score on a late surveillance or indication biopsy, both as predictors of clinical benefit for reducing late graft failure. This could be further strengthened by the use of a central pathologist reading all biopsies in a study, as interobserver variability exists using such scoring schema (93).

Enrichment Strategies to Optimize Enrollment in Clinical Studies

Another approach to optimize the execution of clinical studies of chronic allograft injury is enrichment. This involves focused enrollment of patients with a clearly defined phenotype that places them at high risk for a subsequent surrogate endpoint or primary clinical outcome of interest. Because the majority of kidney transplant recipients have excellent overall 5- and 10-year death-censored graft survival, enrolling an unselected cohort of these patients in a clinical study of chronic allograft injury or graft failure will result in a large sample size and a long follow-up period to power a meaningful change in the primary outcome. Such studies are often time- and cost-prohibitive, and result in (at best) critical delays in bringing new therapeutics to the clinic or (at worst) failure to detect a clinical benefit for a specific subpopulation of recipients. This is unfortunately typical of a randomized controlled trial that provides valuable evidence for how a particular intervention would perform in a random sample of the population at large, not taking into account important positive and negative responses in subgroups of the population. Given the vast heterogeneity in kidney transplant recipients, it is much more feasible to define therapies that benefit a specific subgroup of patients than finding an intervention that benefits the entire population equally.

In 2010, the FDA outlined several enrichment strategies to strengthen the signal-to-noise ratio of clinical studies. Increasing homogeneity involves excluding highly variable patients from the study population. Prognostic enrichment is an approach that focuses on high-risk patients with a higher probability of experiencing the endpoint of interest. Predictive enrichment refers to selecting patients with a high probability of responding to a given treatment (94). The FDA also encouraged the use of emerging biomarkers to select a uniform study population that follows this guidance, particularly those derived from genomic and proteomic methods. Importantly, since any of these approaches affect patient selection before the start of a clinical study, there is little concern for statistical validity or application of the findings in the specific subgroup that was enriched for in the study design (94). An example in chronic allograft injury research would be an intervention to reduce the Banff chronic glomerulopathy (cg) severity score in a study of human leukocyte antigen incompatible recipients at high risk for antibody-mediated injury and cg lesions on biopsy. Novel therapies (or application of existing therapies in novel ways) can be improved much more rapidly with these methods under FDA Code of Federal Regulations Title 21, Subpart H, “Accelerated approval for serious or life-threatening illnesses.” (30) The caveat to such accelerated approvals based on surrogate endpoints is that they require verification to describe the benefit in primary clinical outcome studies.

Adaptive Clinical Trial Design

The FDA’s 2010 guidance on adaptive clinical trials reviewed planning, models, protocols, analysis, and troubleshooting methods for such studies (94). Adaptive clinical trials can reduce time, costs, and sample sizes required to demonstrate or predict clinical benefit (or alternatively, to demonstrate a lack of efficacy) for an intervention, and have garnered high interest from both the pharmaceutical industry and the FDA. The FDA guidance defines an exploratory study (Phases 2a and 2b) as one that does not rigorously control the probability of Type I errors (“false positive” conclusions). These studies provide room for creativity and learning about efficacy, safety, effects on multiple different endpoints, and variability between patient subgroups, and are often the subject of an adaptive trial design.

According to its guidance, adaptive trials include a prospectively planned opportunity for modification of one or more aspects of the study design and hypotheses based on interim data analyses. These are conducted at pre-specified time points and can be performed in a fully blinded or unblinded manner, and do not require statistical analysis. Potential adaptations may include changes in inclusion criteria, randomization, regimens, sample size, visit intervals, primary and secondary endpoints, and early termination of the study (for excessive benefit or harm). Any number of adaptations can be made, provided they are considered and deployed at pre-specified time points agreed upon prior to study initiation; otherwise the integrity of the study would be compromised. Compared to non-adaptive (traditional) studies, adaptive trials can reach meaningful efficacy conclusions more efficiently and improve knowledge regarding the effect of an intervention at various doses or in different patient subgroups.

One such adaptive trial design that could accelerate drug discovery in chronic allograft injury research is known as a group sequential design (GSD). Instead of waiting until the end of a trial period, the GSD allows for an unblinded interim analysis of study data to determine whether to terminate early, either based on a robust efficacy signal or an obvious lack of benefit or even enhanced harm. Early determination that an intervention is futile would save valuable time and resources to apply toward other research efforts. Likewise, early findings of efficacy can hasten the process of bringing new therapeutics to the clinical realm where they can benefit patients. An important caveat is that early termination may result in analyzing a smaller sample size than was intended at the onset of the study. Also, there will be less data regarding safety and clinical outcomes may be underpowered. These points highlight the importance of rigorous statistical pre-planning in the design phase of an adaptive clinical trial to ensure that such steps are taken without compromising the validity of the trial results.

One recent example in kidney transplantation is a trial in which deceased organ donors were randomized to either a hypothermia protocol or standard body temperature to decrease the rate of delayed graft function (95). This study employed an adaptive trial design using the GSD method, which used a planned interim analysis to examine pre-specified stopping criteria for either success or failure of the study. In this study, the interim analysis was performed when half of the planned number of subjects was enrolled. There was an overwhelming efficacy signal (hypothermia reduced delayed graft function by nearly 50%) that met stopping criteria for early success of the trial. This enabled early completion of the study with a significantly reduced sample size, a tremendous savings in terms of both time and cost, resulting in rapid delivery of an important advance to our field. The final analysis of the study showed a mildly attenuated benefit compared to the interim analysis, with hypothermia reducing delayed graft function by 38% compared to kidneys from normothermic donors (95).

There are several concerns regarding these adaptive designs that must be considered, including correcting for multiple hypothesis testing, selection bias with adjusted doses of an intervention under study, and operational bias resulting from unblinded interim analyses. These adaptive trial designs require much care and rigor in the design phase to ensure their validity is maintained, but have the potential to greatly accelerate our acquisition of knowledge regarding new therapeutic interventions in solid organ transplantation, as demonstrated in the aforementioned study by Niemann et al. (95).

INTEGRATION OF EMERGING CONCEPTS INTO IMPROVED STUDY DESIGNS FOR CHRONIC ALLOGRAFT INJURY RESEARCH

Despite tremendous advances in our understanding of the immune response to an allograft, improved organ preservation and surgical techniques, novel immunosuppressive targets and agents, and an expanding cadre of diagnostic biomarkers, there has been little improvement in chronic allograft injury or late graft failure over the last few decades. As a transplant research community, this issue requires a shift in our thinking and in our approach to how we generate new knowledge going forward. A traditional study design that measures the efficacy of a single immunosuppressive agent in a broad transplant population using chronic allograft injury or graft loss ignores what we have learned about the disease, as well as how best to conduct clinical studies in general.

An example of a modern study design for chronic allograft injury would use enrichment strategies to enroll relevant subgroups of patients at high risk for a component of the disease and define a surrogate endpoint(s) within an adaptive trial design to allow for rapid dissemination of any findings of robust efficacy. One such approach is detailed in Table 3. In this example, the goal of therapy is to reduce BKVAN. This study would enroll recipients with BKV viruria and viremia, a subgroup that is enriched for development of BKVAN and therefore at increased risk for chronic allograft injury and subsequent graft loss. The primary surrogate endpoint would be a significant reduction in BK viral load, with secondary surrogate endpoints of reduced incidence of BKVAN, change in allograft function, and change in ci/ct biopsy scores. Prespecified interim analyses would be conducted to determine the efficacy of the intervention once enrollment reached 50% of the planned cohort, with evaluation of stopping criteria for either robust efficacy or a clear lack of benefit. Successful reduction in BK viremia would be sufficient for initial approval of the intervention, with long-term follow-up to determine its effect on graft survival in patients with BKV viremia. Similar studies could be designed to study focused problems in specific subgroups, such as recurrent glomerular disease, transplant glomerulopathy, delayed graft function, and subclinical rejection. Alternatively, enrollment could be based on a novel biomarker of disease, such as a gene expression profile or donor-specific immunity assay hypothesized to increase risk for a surrogate endpoint of chronic allograft injury (79, 84, 96). Conducting more focused studies such as these will provide important knowledge in specific subgroups of patients at risk for chronic allograft injury and late graft failure and better elucidate the role of various contributors to the disease. This would allow for more personalized management of solid organ transplant recipients to improve long- term outcomes, akin to the explosion of innovation in precision diagnostics and targeted therapeutics that has taken place in cancer research.

Table 3.

Proposed clinical trial design to evaluate a new therapy to prevent chronic renal allograft injury and late graft failure.

Study Hypothesis: Intravenous immune globulin (IVIG) prevents progression of chronic renal allograft injury in patients with BKV viremia.

Enrichment. Target subjects at increased risk for BKV viremia and BKVAN.

Proanostic enrichment: enroll kidney recipients < 6 months post-transplant with BKV viruria (estimated 30–40% of recipients who are at increased risk for BKV viremia).

Increased homoaeneitv: perform allograft biopsy at time of BKV viruria to document Banff ci, ct, and ti scores are all ≤ 1 and absence of BKVAN and acute rejection at enrollment.

Adaptive trial design. Pre-specify surrogate endpoints and an interim analysis.

Study intervention: randomize subjects with BKV viremia > 2,000 copies/mL to IVIG versus placebo in addition to standard-of-care (reduction in immunosuppression).

Primary surroaate endpoint: remission of BKV viremia, defined as decrease in blood BK viral load to < 2,000 copies/mL.

Secondary surrogate endpoints:

Progression of BKV viremia, defined as blood BK viral load > 10,000 copies/mL Biopsy-proven BKVAN Stable Banff ci, ct, ti scores at 1 year post-enrollment biopsy Stable allograft function at 1 year post-enrollment.

Interim analysis: blinded analysis when 50% of enrollment target is reached to determine if adequate efficacy (or futility) signal to support early study cessation.

Expedited approval. Initial approval for prevention of progression of chronic renal allograft injury in BKV viremic recipients is based on primary surrogate endpoint.

Final approval. Follow-up studies are required to demonstrate that IVIG prevents long-term progression of chronic allograft injury and improves 5-year allograft survival in a subgroup of kidney transplant recipients with early BKV viremia.

CONCLUSIONS AND FUTURE DIRECTIONS

There are several principal needs for improving clinical studies in chronic allograft injury and late graft failure. There must be appropriate use of typical primary clinical endpoints and composite outcomes as well as prioritization of relevant patient-centered outcomes such as cardiovascular risk and quality of life. Consideration must be given to inclusion of high-risk populations in clinical transplant studies, perhaps as separate cohorts rather than as greater proportions of an overall study cohort, which would allow better understanding of best practices for these unique groups. Modern study designs are needed that not only incorporate relevant and valid surrogate endpoints, but also make use of novel methods such as emerging biomarkers, adaptive trials and enrichment strategies. We must continue to be vigilant and willing to consider developing and implementing new approaches to solve the most important problems facing our field now and in the years to come.

Abbreviations:

BKV

BK virus

BKVAN

BKV-associated nephropathy

cg

Banff chronic glomerulopathy score

ci

Banff interstitial fibrosis score

CMV

cytomegalovirus

ct

Banff tubular atrophy score

cv

Banff chronic vasculopathy score

DSA

donor-specific antibody

EBV

Epstein-Barr virus

eGFR

estimated glomerular filtration rate

FDA

Food and Drug Administration

FGF23

fibroblast growth factor-23

g

Banff glomerulitis score

GSD

group sequential design

i

Banff interstitial inflammation score

IVIG

intravenous immune globulin

ptc

Banff peritubular capillaritis score

t

Banff tubulitis score

ti

Banff total interstitial inflammation score