Association Between HSV‐1 and Risk of Alzheimer’s Disease Detected by a Propensity‐Score Matched Case Design
Marlene Tejeda, John Farrell, Congcong Zhu, Lee Wetzler, Kathryn L. Lunetta, William S. Bush, Eden R. Martin, Li‐San Wang, Gerald D. Schellenberg, Margaret A. Pericak‐Vance, Jonathan L. Haines, Lindsay A. Farrer, Richard Sherva- Psychiatry and Mental health
- Cellular and Molecular Neuroscience
- Geriatrics and Gerontology
- Neurology (clinical)
- Developmental Neuroscience
- Health Policy
- Epidemiology
Abstract
Background
Herpes simplex virus 1 (HSV‐1) is a neurotropic and neuroinvasive species that can enter the brain triggering inflammation and an immune response, increasing the risk of Alzheimer disease (AD). Previously, we found a significant association between HSV‐1 and AD in whole‐exome sequencing (WES) data (OR = 3.69, P = 6.71×10−4) but not whole‐genome sequenced (WGS) data (OR = 0.92, P = 0.27) from the AD Sequencing Project (ADSP).
Method
To understand this discrepancy, and to better control for the numerous covariates and technical differences that influenced viral quantity, we created a novel algorithm using propensity score analysis (PSA) which permits analysis of observational study data comparable to a randomized controlled trial. WGS reads from 6,541 AD cases and 6,855 controls from the ADSP (including 7,395 non‐Hispanic Whites, 2,961 African Americans, and 3,040 Caribbean Hispanics) that did not align to the human genome were mapped to viral reference sequences and quantified. Propensity scores were derived for 7,619 individuals with at least one HSV‐1 read and 5,777 individuals lacking any HSV‐1 reads using a binary treatment indicator for HSV‐1 status. We applied forward stepwise regression including known confounders (APOE ε4 status, sex, age, tissue source (blood or brain), population ancestry, sequencing center, and PCR amplification) to identify other potential confounders and/or interactions and quadratic terms for creating propensity scores. We tried several matching algorithms to generate matched sets of individuals with and without HSV‐1. Utilizing the best quality matched set, we evaluated the association of AD with HSV‐1 status using a weighted generalized linear model.
Result
The final regression model included terms for the covariates mentioned above that were available for most of the original sample (5828 AD cases and 6487 controls). We found a significant association between HSV‐1 and AD (OR = 1.10, P = 0.02).
Conclusion
Results from WGS data confirm our previous findings obtained from WES data and support the hypothesis that HSV‐1 is associated with an increased AD risk, highlighting the utility of PSA for analyzing association of risk factors with complex disorders like AD. Further studies are needed to replicate this finding in other ancestral populations and explain the lack of association in WGS data.