Project Summary/Abstract
Conventional thresholds of physiological parameters for clinical diagnosis have become a root cause of
racial/ethnic disparities in healthcare. These thresholds were established decades ago using limited samples of
Caucasian subjects and simplistic statistics without considering genetic effects. Consequently, these thresholds
have no capacity to account for the normal variability intrinsic in race/ethnicity, genes, and environments, which
can lead to errors in disease diagnosis and risk assessment when applied to racial/ethnic minorities. As the
minority population grows and becomes even more diverse, continuing to use Caucasian-based, “one-size-fits-
all” thresholds creates the risk of exacerbating racial/ethnic disparities in healthcare. Therefore, an effective
alternative must be developed to address this issue. Hence, we propose creating Person-specific Precision
Thresholds (PPTs), which are designed to account for normal variability in genetic and biological makeup,
variation in race/ethnicity, sex, age, and other characteristics. PPTs will eliminate the root cause of healthcare
disparities caused by conventional thresholds for diagnosis and risk assessment. The innovative precision of
PPTs will fundamentally shift current research paradigms regarding clinical thresholds from a single cutoff point
for everyone to precise, person-specific thresholds matched to individual genetics and phenotypic
characteristics. PPTs will effectively improve disease diagnosis and risk assessment while reducing healthcare
disparities. The first PPT iteration will focus on women's bone mineral density (BMD), as osteoporosis primarily
affects women the most. The existing T-score threshold of BMD has become controversial because many women
who sustain fragility fractures have a “normal” BMD value, as defined by the commonly employed T-score
method. Because genetic factors contribute to 50-85% of BMD variation, genetic variants must be considered
when creating precision thresholds. With feasibility tested by pilot studies, we hypothesize that PPTs will enhance
the ability of a single BMD measure to predict fractures in minority women compared to conventional T-score
and prior model-based methods, which lack consideration of genetics. We will pursue the following three specific
aims: (1) Develop a modeling framework for PPTs of BMD using novel machine-learning techniques; (2) Calibrate
PPTs of BMD using clinical data; (3) Validate the PPTs of BMD using genomic cohort data. In addition, we will
create a publicly available, cloud-enabled PPT application for community investigators. This proposed approach
is innovative because it integrates individual genetics into clinical precision thresholds and uses novel machine-
learning techniques to create PPTs. By replacing the traditional one-size-fits-all threshold, the innovative PPTs
will shift existing research paradigms regarding how clinical thresholds have been developed and implemented.
The knowledge gained from this research can be utilized to create many other PPTs to improve the diagnosis
and prevention of various diseases and illnesses. These next-generation PPTs will help significantly improve
patient outcomes, leading to better health, a higher quality of life, and improved health equity.
