PROJECT SUMMARY/ABSTRACT
Acute kidney injury (AKI) occurs in over 50% of patients with sepsis and is associated with increased risk of
mortality and other complications, including longer hospital length of stay and the development or progression
of chronic kidney disease. To date, therapies for sepsis associated AKI (SA-AKI) remain largely supportive.
One possible reason for the failure of targeted therapies may be the currently limited understanding of the
pathophysiology of the condition. Traditionally, SA-AKI has been thought to be largely due to renal ischemia
secondary to hypoperfusion. However, animal models show that SA-AKI is actually a state characterized by
hyperdynamic circulation, and review of histological changes in septic patients shows a lack of acute tubular
necrosis, a pathological finding often associated with renal ischemia. Rather, it appears that a combination of
immunologic, toxic and inflammatory factors likely leads to SA-AKI, and intervening along these pathways may
improve patient outcomes for SA-AKI.
We hypothesize that within a population with SA-AKI, there are subgroups who have particular inflammatory
and immunological profiles that are associated with differential outcomes and response to therapies. In the
fields of asthma and the acute respiratory distress syndrome, the use of clinical data and biomarkers to identify
such subpopulations, or sub-phenotypes, of disease, has led to new treatment paradigms, where individuals
with a specific biological profile receive and benefit from a targeted therapy.
We propose to use data from a large, established cohort of patients admitted to the Intensive Care Unit (the
Early Assessment of Renal and Lung Injury [EARLI] cohort) to define sub-phenotypes of SA-AKI using latent
class analysis. Latent class analysis is an established method that uses mixture models to identify sub-
phenotypes within a heterogenous population in an unbiased manner. Prior work using this methodology has
allowed for discovery of new disease entities that inform biological pathways of cellular injury leading to organ
failure.
Our specific aims are:
Aim 1: To identify novel sub-phenotypes of patients with SA-AKI and discover molecular targets for
therapy through the incorporation of clinical and biomarker data into our latent class analysis.
Aim 2: To determine whether certain sub-phenotypes are associated with differential outcomes with
regards to mortality, need for renal replacement therapy, or duration of AKI.
Through the studies outlined in this F32 proposal, Dr. Kwong will learn how to rigorously conduct latent class
analysis as well as other unbiased analytic techniques (e.g., machine learning) and to apply these techniques
to AKI. This work will effectively position Dr. Kwong for her future research studies, where she wants to identify
appropriate patients for clinical trials of AKI therapies and to better define the pathophysiology of human AKI.