Suicide risk modification by statin prescriptions in US Veterans with common inflammation-mediated clinical conditions- a controlled, quasi-randomized epidemiological approach

Baltimore Va Medical Center, Baltimore MD

Investigators

Abstract

In addition to their metabolic and cardiovascular protective effects, statins reproducibly engage multiple pathophysiological factors implicated in suicidal behavior - neuroinflammation, increased oxidative stress, excitotoxicity, and endothelial dysfunction. Add-on statins have been also reported to improve therapeutic control in physical and mental health. The Veterans’ persistent higher rates of suicide have remained unabated challenges and, and thus, demanding new ways of understanding and engaging in preventative efforts. The long-term objective of our group is to uncovering new modifiable targets, novel and repurposed treatments in suicide prevention, and identifying individuals at risk who are likely to most benefit from specific interventions. Macro-epidemiological approaches using electronic medical records in suicide research are irreplaceable for their capability to account for multiple interactive risk factors, moderators and confounders, and potential for immediate impact. The primary aims of the proposed research project are to: 1) Estimate potentiating interactions between traumatic brain injury (TBI), a very common condition in US Veterans, and inflammation-mediated medical conditions (IMCs: allergies, infection, and autoimmune conditions), in predicting suicide in US Veterans. Our preliminary data support hypothesizing synergistic interactions. 2) Estimate the suicide protective effect of sustained vs. unsustained statin treatment 3) Identify demographic and clinical Veteran characteristics and pharmacological statin features (dose, lipophilia, potency, duration) conducive to stronger attenuating effects of statins on suicidal behavior. We will test these hypotheses on a Veterans Health Administration (VHA) retrospective cohort (individuals with clinical encounters in VA Medical Centers nationwide beginning in 2004 and followed for 13 years) including 5,446,318 Veterans with 28,749 suicides. The Cox proportional hazard model will be applied to evaluate the interactions between TBI immune mediated conditions , with Relative Excess Risk due to Interaction (RERI), the Attributable Proportion (AP) due to interaction, and the Synergy Index (SI) to test synergism on an additive scale (Aim 1). A Cox proportional hazard model will also be applied to testing risk attenuation with statins, with propensity scoring for time-independent confounding and marginal structural Cox proportional hazards (Aim 2). Finally, we will identify the demographic, clinical (diagnostic codes, medications, laboratory markers of inflammation (e.g., white blood count) and pharmacological characteristic of Veterans expected to benefit the most from sustained statin treatment using an aggregate machine learning approach (the SuperLearner integrative methodology). Considering the high prevalence of TBI history and its ongoing sequelae,( “a silent epidemic”) , especially in the VA, and confirming their synergistic interaction with IMCs may contribute to developing suicide risk-attenuating interventions specifically for those subpopulations. The PI’s preliminary data nested in Danish registers, our team’s piloting confirming preliminarily a reduction in rates of psychiatric hospitalization (considered a proxy measure of suicide risk) with statins in US Veterans diagnosed with schizophrenia or bipolar disorder and treated with psychotropic medication (Appendix 4C), and our successful evaluation of potential heterogenous effects of an alternative modifiable suicide risk using the specific machine learning algorithms proposed in this project (Appendix 4B) support our hypotheses, integration, and purpose, and overall, project completion capability. Using tailored repurposed medications, such as statins, targeting specifically molecular, cellular and histological mechanisms directly implicated in suicidal behavior, to individuals at risk who are identified by machine learning to potentially derive the greatest benefit from treatment , may provide a much-needed breakthrough in suicide risk management and prevention.