Precision Medicine in Acute Kidney Injury: A Promising Future?

Despite increased focus over the past decade, the management of patients with acute kidney injury (AKI) remains largely supportive, including dialysis for severe cases. Clinical trials in AKI examining timing of dialysis, intensity of dialysis, pharmacotherapy, and novel biologics have been consistently negative (1–6). One postulated reason for this dearth of positive trials is the inherent delay in intervention for patients with AKI due to a reliance on serum creatinine, and researchers have embarked on a decades-long journey to identify a biomarker of AKI that would identify patients while kidney damage was actively ongoing and before serum creatinine increases. Numerous biomarkers of tubular injury have now been identified (7, 8), and in the ICU these biomarkers have modest sensitivity, specificity, and association with outcomes (9, 10). However, these biomarkers have so far failed to recategorize the heterogeneous syndrome of AKI into more clinically useful subtypes or be incorporated into clinical practice. There has been some progress with biomarkers of cell cycle arrest, most notably TIMP2*IGFBP7 (tissue inhibitor of metalloproteinase-2*insulin growth factor binding protein-7), to identify patients at high risk of AKI (11, 12). In patients after cardiac bypass surgery at high risk for AKI (as denoted by elevated TIMP2*IGFBP7), there was a lower incidence and decreased severity of AKI in patients who were randomized to a “KDIGO (Kidney Disease: Improving Global Outcomes) bundle,” which included monitoring of hemodynamic parameters, avoidance of nephrotoxins, and holding angiotensin-converting-enzyme inhibitors (13). Although prevention may be possible, the role of biomarkers in guiding treatments or response to therapy remains unclear.

For this reason, the article by Bhatraju and colleagues (pp. 863–872) in this issue of the Journal represents meaningful progress (14). The authors applied latent class analysis to a discovery group of 794 patients admitted with systemic inflammatory response syndrome to the ICU and a replication cohort of 425 patients with acute respiratory distress syndrome (ARDS) and identified two subphenotypes of AKI (AKI-SP1 and AKI-SP2). The patients in AKI-SP2 were sicker and had worse renal function; higher rates of sepsis, ARDS, and mortality; and lower rates of renal recovery. The authors determined via least absolute shrinkage and selection operator method that the ratio of angiopoietin-1 and angiopoietin-2 (Ang1/Ang2) and sTNFR-1 (soluble tumor necrosis factor receptor-1) were sufficient to accurately distinguish between the two subphenotypes of AKI (c-statistic > 0.93).

Ang1 and Ang2 are endothelial growth factors, which both bind to the extracellular portion of the Tie-2 receptor. They have opposing actions: Ang-1 stabilizes the vascular endothelium, and Ang-2 destabilizes the vascular endothelium. Consequently, the ratio of these endothelial growth factors provides an assessment of endothelial dysfunction and is associated with prognosis in several cohorts of critically ill patients with and without AKI (15, 16).

These sophisticated statistical techniques and biomarkers determined what clinicians intuitively understand: patients with more severe inflammation do worse. The authors then reidentified the subphenotypes in a random subset of 328 patients from the VASST (Vasopressin in Septic Shock Trial) who had measurements available for Ang1/Ang2 and IL-8 (17). (Soluble tumor necrosis factor receptor-1 was not available in the VASST cohort, but IL-8 was notably different between AKI-SP1 and AKI-SP2 in the discovery and replication cohorts.) This clinical trial was a randomized, double-blind study comparing vasopressin and norepinephrine infusions to norepinephrine alone in 776 patients with septic shock. The study had shown no differences in mortality or rates of renal failure between patients in either treatment group. Once patients were recategorized into the AKI subphenotypes, patients in AKI-SP1 (the less ill group) had improved 90-day mortality with early addition of vasopressin compared with norepinephrine alone. This association persisted after adjustment for Acute Physiology and Chronic Health Evaluation II score, suggesting discriminating ability of the AKI subphenotypes beyond simply severity of illness. If replicated, these findings hold the potential to guide clinicians in more accurately assessing prognosis of patients as well as expected response to therapy.

Despite these findings, much work remains. There were no urine specimens available from these cohorts, and it is unclear if the particular model the authors identified is truly the best model to distinguish subphenotypes of AKI in patients with systemic inflammatory response syndrome. Moreover, given the heterogeneity of AKI and the innumerable settings in which it occurs, this particular model may not be helpful for all patients with AKI, such as those post cardiac surgery, post contrast exposure, or with heart failure. Other investigators will have to identify subphenotypes of patients with AKI in these settings, and these subphenotypes may be identified by a combination of biomarkers of tubular injury, cell cycle arrest, cardiac dysfunction, clinical variables, or a novel biomarker. Incorporation of phenotyping may therefore be an important component of future trials for AKI and could possibly be the key to breaking the trend of negative clinical trials in AKI.

It is important to note that our molecular understanding of acute tubular injury, which, if biopsies were available in these cohorts, probably would have been the most common histologic entity, remains suboptimal. Inflammation and endothelial dysfunction are typically considered a systemic response instead of intrinsic to the kidney. The KPMP (Kidney Precision Medicine Project), which will integrate molecular, structural, and clinical information from kidney biopsy specimens of patients with AKI, will be critically important to untangle the molecular underpinnings of acute tubular injury. Ideally, these data will improve our ability to identify disease subgroups that would respond to therapy.

Instead, this particular article can serve as a framework for other investigators to attempt to identify other subphenotypes of AKI. To move forward, we recommend the following next steps. First, investigators should attempt to identify novel subphenotypes of AKI in other common AKI settings using currently existing biomarkers. As our molecular understanding of AKI improves from studies like KPMP, these subphenotypes should ultimately be identified based on underlying molecular mechanisms. Second, investigators should demonstrate that these subphenotypes respond differently to therapies in previously completed clinical trials. This is an important step and, given the plethora of negative clinical trials in AKI, there is ample opportunity. Ultimately, we hope these findings would change the way patients are selected and enrolled into future clinical trials. Although biorepositories require time, money, and effort, we urge investigators to continue maintaining and creating them, as they can yield valuable information years later.

It remains to be proven if the subphenotypes of AKI in patients with critical illness identified by this study will be useful in clinical practice. Regardless, the findings are important because they suggest that it is possible to untangle the complex and heterogeneous clinical syndrome of AKI into groups of patients who respond to a particular therapy. In other words, we may be one step closer to personalized and precision medicine for AKI.