Clinical and Translational Science Pilot Award Round 2 Awardees

The Michigan Institute of Clinical & Health Research (MICHR)’s Clinical and Translational Science Pilot Award supports innovative research projects in clinical & translational science (CTS). CTS is a field of investigation focused on understanding and removing barriers that slow translational research so that new treatments and other health solutions reach people faster. Projects typically propose addressing a common cause of inefficiency or failure in research projects at any stage of translation, with the goal of producing foundational knowledge that will ultimately improve health.

We are pleased to announce that the following proposals have been funded in the second round:

Advancing Precision Medicine: A Microdose Approach for Universal Kidney Function Measurement

Investigators: Levi Hooper, PharmD; Michael Heung, MD; Lu Wang, PhD; Aleksas Matvekas, BS; Radin Alikhani, PharmD; Michael T. Kenes, PharmD; Kathleen A. Stringer, PharmD; Bruce A. Mueller, PharmD; Manjunath P. Pai, PharmD, FCP – University of Michigan College of Pharmacy, Michigan Medicine, and School of Medicine.

Misclassification of kidney function has serious consequences such as unnecessary exclusion from kidney transplantation and clinical trial participation, underdosing or overdosing medications, and delays in treatment initiation. Current approaches to kidney function assessment rely on estimation rather than direct measurement due to convenience. However, contemporary equations used for these estimations are fraught with limitations and misclassify kidney function in many specific populations, such as those with obesity, kidney disease, older adults, and children. These inaccuracies increase the failure risk of critical medications like antibiotics, anticancer drugs, and direct oral anticoagulants that need precise dosing. Our proposed solution aims to address these RMD challenges by developing a cost-effective and precise method for measuring kidney function. Our method relies on a microdose approach using the commonly used contrast agents iopamidol and iohexol, coupled with a sensitive assay method to enhance efficiency. Our methodology will compare the precision of these kidney function biomarkers in a cohort of adult participants (n=24) within the expected clinical range of kidney function. Creating a sensitive analytical method and establishing the interchangeability of iohexol and iopamidol has major translational implications. If successful, it could allow for the repurposing of routine blood samples for kidney function measurement after patients receive a contrast-enhanced CT scan for another diagnostic purpose. The development of techniques like the one proposed will advance translatability and enhance the convenience of measured kidney function, overcoming fundamental biological variables that contribute to large variabilities between patients in the estimation of kidney function, ultimately allowing for broader generalizability.

A Human Intestinal Organoid Screening Platform for Drug Discovery and Pre-Clinical Development Research Plan

Investigators: Jason Spence, PhD, Professor, University of Michigan; CSO, Intero Biosytstems, Inc.; Charlie Childs, PhD, CEO, Intero Biosystems, Inc.; Madeline Eiken, PhD, COO, Intero Biosystems, Inc.

The drug development pipeline currently tests drug candidates in animals before seeking FDA approval for clinical trials, yet 90% of new drugs fail clinical trials, wasting over $1 billion per drug on average. In late 2022, the FDA Modernization Act was signed, allowing alternative methods to replace animal testing in preclinical trials. This legislation will reduce the number of animals sacrificed in research and promote the use of more accurate predictive models to mimic complex drug behavior in humans, with cell-based assays and organoid models as expected alternatives. Organoids are complex “organ-like” 3D structures that mimic aspects of the native organ in vitro, but often lack the ability to mimic many physiologic functions of an organ because they are missing cell types such as blood vessels, muscle, neurons, and immune cells. The foundation of this MICHR application is a novel stem cell derived Human Intestinal Organoid (HIO), the first of its kind to contain many tissue types found in the native intestine, mimicking the human intestine “in a dish”. This makes them the perfect platform to predict how drugs will behave in humans. These HIOs are cultured from a highly tractable and scalable source, human pluripotent stem cells, and grown using our novel method to create complex structures including the inner lining of the intestine and supporting tissue including smooth muscle, blood vessels, and neurons. We have shown that these HIOs recapitulate important functions of the human intestine; they spontaneously and rhythmically contract without stimulus, suggesting functional muscle and neuronal connections and they have also been successfully attached to a simulated circulatory system (via microfluidic devices), confirming blood vessel function. Furthermore, these HIOs can be stimulated to mimic fibrosis, a common complication associated with Inflammatory Bowel Diseases (IBD) such as Crohn’s and Ulcerative. We have also shown proof of concept that the HIOs can be used to test anti-fibrotic drugs and believe we can push this technology further to aid in many therapeutic areas where patients are in desperate need of effective drugs. We hypothesize that the HIO can model complex intestinal diseases that cannot be accurately mimicked in animals, such as fibrosis, and HIO will be suitable for rapid screening to determine efficacy and safety concerns of antifibrotic drug candidates in a preclinical setting medical research, here at the University of Michigan and beyond.

Med-Cox-Nnet: A Missing Value-Aware Competing Risk Survival Prediction Model For Longitudinal EHR Data

Investigators: Dr. Lana Garmire, Associate Professor, Department of Computational Medicine & Bioinformatics, University of Michigan; Xiaotong Yang, PhD Candidate, Department of Computational Medicine & Bioinformatics, University of Michigan

Electronic health records (EHR) is a digital version of a patient’s medical history, including information on diagnoses, medications, lab results, and other healthcare-related data. EHR data are an invaluable resource for precision health because they offer larger patient cohorts and more comprehensive information at lower costs, compared to genomics. One of the most essential applications of EHR is clinical survival models. Survival models can predict the onset time, prognosis and survivability of diseases and assist clinical decision-making, treatment planning, and resource allocation. However, conventional survival methods cannot effectively capture the missing patterns in EHR data nor incorporate temporal clinical measurements. 1) Missing data in clinical research often follows non-random patterns due to patient non-response, follow-up loss, errors, or selective testing. Simply removing or filling these gaps risks information loss or bias. 2) Many EHR data are irregular time series data. The time difference between each measurement or patient varies greatly, not fitting for conventional time-series analysis. Researchers often take the summary statistics (i.e. Mean, max, min) or divide time series into bins of a fixed period, resulting in considerable information loss. A comprehensive algorithm that learns from both the irregular longitudinal clinical data and its missing patterns is pressingly needed to improve prediction performance in EHR data. We will create a new deep-learning survival prediction model that can incorporate both the temporal clinical features (e.g. vital signs, lab results) as well as time-invariant variables (e.g. demographics, medical history). Additionally, this model will learn from both the sampling and missing patterns in the temporal input. We will then compare the performance of our model with other state-of- the-art survival methods. The successful development of this model will provide clinicians and researchers with a powerful tool for time-to-event prediction using EHR data, in clinical ap- plications such as disease onset, prognosis, adverse outcome occurrence, transfer of patient care and hospital readmission. It will improve patient outcomes by enabling more informed decision-making, reducing healthcare costs, and ultimately advancing precision health. Moreover, it promotes the development of innovative methodologies for leveraging EHR data to enhance patient care and medical re- search, here at the University of Michigan and beyond.

Utilizing a Smart Intervention Design to Optimize Participant Adherence in Home-Based Behavioral Interventions

Investigators: Rebecca Hasson, PhD; Penelope Friday, PhD; Alissa Li, BS; Mariam Habring, BS; Kelley Kidwell, PhD; Alison Miller, PhD; Leah Robinson, PhD; Andria Eisman, PhD, MPH, MS; Lauren Oshman, MD; Wala Tout, MD

Adherence refers to how well a person follows a recommended and agreed-upon course of action. Adherence is necessary for desired clinical outcomes and can include attending appointments, making lifestyle changes, and following home-based regimens for themselves or someone for whom they care for. Adherence to home exercise programs can be as low as 50%, directly impacting program effectiveness. There are many reasons why an individual may struggle to adhere, which may include perceived barriers such as not having enough time, reduced self-efficacy, the belief that the program is ineffective, or seeing early positive results so they then feel they do not need to continue with the program. For children, low skill level and lack of parental support are additional barriers to program adherence. A Sequential Multiple Assignment Randomized Trial (SMART) intervention is an intervention design that guides the adaptation of treatments over time. Adaptation refers to the use of dynamic information about a person (or family) to decide whether and how to intervene. SMART interventions are intended to address the unique and changing needs of individuals. The proposed study will examine the feasibility of using a 12-week (60-day) adaptive intervention design to optimize child adherence to a home exercise program. To advance equitable adherence to home exercise programs, there is a pressing need to develop strategies that support all children and families. The proposed study will provide new knowledge regarding the utility of adaptive interventions to optimize participation engagement in behavioral interventions in community contexts.