4011 — High Dimensional Empirical Bayes Screening of Medications for Risk of Suicidal Behavior

Lead/Presenter: Jill Lavigne,  Center of Excellence for Suicide Prevention
All Authors: Gibbons RG (University of Chicago Center for Health Statistics), Hur K (Center for Medication Safety, Pharmacy Benefit Management), Lavigne JE (Center of Excellence for Suicide Prevention) Mann JJ (Columbia University Department of Psychiatry

Objectives:
To develop a statistical surveillance methodology based on population medical claims and electronic medical records that identifies drugs associated with increased and decreased risk of suicidal events.

Methods:
Design: We use a within-person incident-user cohort design to simultaneously examine the relationship between 923 drugs and suicide attempts. Setting: A pharmacoepidemiologic study of U.S. medical claims for private health insurance (MarketScan - MS) and medical records from the Veterans Administration (VA) from 2003-2014. Participants: 198,652 subjects with a suicidal behavior event and days' supply of any medication documented in either North American commercial claims data (MarketScan (MS)) (n = 923) or VA electronic medical records (n = 513). Exposure: The first filled prescription of each of 923 drugs in MS data, 513 of which were also documented in VA electronic medical records. Main Outcome: Suicidal behavior events (suicide attempts and intentional self-harm) were identified based on the following ICD-9 codes (E950-E959). Statistical analysis: Mixed-effects logistic regression was used to estimate empirical Bayes (EB) odds ratios (ORs) and 95% confidence intervals (CIs) adjusted for multiple comparisons for the individual drug effects. The interpretation of the EB OR is the increased or decreased likelihood that the adverse event will occur following drug exposure.

Results:
This new statistical surveillance system identified 24 drugs with increased risk and 36 with reduced risk of suicide attempt following exposure in either the MS or VA datasets. Among the common set of 513 drugs in both the North American commercially insured patients and VA patients, 7 were associated with increased risk and 20 with decreased risk. The strongest increased risk signals occurred in alprazolam (OR 1.83 (1.57-2.12) (MS), OR 1.70 (1.34-2.16) (VA)), diazepam (OR 1.36 (1.09-1.70)(MS), OR 1.74 (1.42-2.13) (VA)), and hydromorphone (1.72 (1.07-2.79)(MS) and 1.79 (1.21-2.64) (VA)). Among the potentially most protective drugs were folic acid (OR 0.37 (0.25-0.56) (MS), OR 0.56 (0.51-0.62)), mirtazapine (OR 0.37 (0.31-0.45) (MS), OR 0.75 (0.67-0.83) (VA)) and naltrexone (OR 0.43 (0.30-0.60) (MS), OR 0.70 (0.57-0.87)). Of those with decreased risk signals, 22 of 36 (61.1%) are approved psychotropic medications, providing both a degree of validation of the method and reassurance to clinicians about the effectiveness and safety of these drugs in suicidal patients.

Implications:
High-dimensional drug safety surveillance using extensive observational data is feasible and generates statistically and clinically significant signals of possible risks and benefits of drugs on risk of suicide behavior.

Impacts:
Suicide is the tenth leading cause of death in the US and the rate has been rising for 16 years. Most suicides occur in the context of a psychiatric disorder, and yet the effect of many medications on suicide risk is intensely debated. This high dimensional drug safety surveillance method screened nearly 1,000 drugs and found similar results in VHA patients as well as North American commercially insured adults.