Degree of serotonin reuptake inhibition of antidepressants and ischemic risk

Data source

We conducted a retrospective cohort study with a nested case-control analysis using the UK Clinical Practice Research Datalink (CPRD). The CPRD contains the electronic medical records of >15 million people enrolled across 700 general practices.¹⁶ Age and sex distributions of patients broadly reflect those of the UK population. Medical diagnoses and procedures are recorded by general practitioners using the Read code classification. Drugs prescribed by general practitioners are coded with the UK Prescription Pricing Authority dictionary and are automatically recorded in the CPRD. The CPRD contains also information on anthropometric variables such as body mass index, lifestyle variables such as smoking or alcohol consumption, laboratory test results, and referrals to specialists and hospitals. Moreover, the CPRD undergoes regular quality controls, and its data have been previously validated and shown to be of high quality.¹⁷

Study population

We assembled a study cohort including all patients in the CPRD who were ≥18 years of age and newly prescribed an SSRI or a third-generation antidepressant between January 1, 1995, and June 30, 2014. Cohort entry was defined as the date of the first antidepressant prescription within the study period. Patients with <1 year of information in the CPRD before cohort entry date and patients who had been prescribed an antidepressant before cohort entry were excluded to form a cohort of new users of an SSRI or a third-generation antidepressant. We also excluded patients initiating treatment with the following antidepressants: trazodone, because it is mostly used for sleep disorders¹⁸; monoamine oxidase inhibitors, because initial pharmacotherapy with these drugs could be a marker of severe disease¹⁹; and tricyclic antidepressants, because they have previously been associated with an increased cardiovascular risk.²⁰ Finally, we excluded patients with a history of the respective outcome (i.e., patients with a history of myocardial infarction for the analysis of the risk of myocardial infarction and patients with a history of ischemic stroke for the analysis of the risk of ischemic stroke). Patients were followed up until the date of their first outcome event (depending on the outcome under study), end of registration with the general practice, death resulting from any cause, or end of the study period (October 31, 2014), whichever occurred first.

Case definition

Within this cohort, we identified all patients with a first-time Read code related to ischemic stroke (including Read codes for TIAs and for strokes not otherwise specified, given that most strokes are ischemic) and myocardial infarction recorded at any time after cohort entry, depending on the outcome being studied. The index date was defined as the calendar date of the first recorded code for each outcome being studied. The recording of cerebrovascular and cardiovascular diseases in the CPRD has been shown to be accurate, with positive predictive values up to 93%.^17,21,22

Control selection

For each case, up to 30 controls were randomly selected among the cohort members in the risk sets defined by the case after matching on sex, age (±1 year), cohort entry date (±1 year), and duration of follow-up. Matching on age and cohort entry date was relaxed to 2 years when no controls were found. Because controls were selected from the risk set defined by each case, which included all individuals still at risk of the event at index date, all controls were necessarily alive, active in the practice, and event-free when matched to their corresponding case. The date resulting in the same duration of follow-up for the case and controls defined the index date for the controls.

Exposure definition

For all cases and their matched controls, we identified all prescriptions for an antidepressant of interest (SSRI or a third-generation antidepressant) between cohort entry and the index date. Patients were considered current users if the duration of their last prescription included the index date or ended within 30 days before the index date. Patients currently exposed to >1 antidepressant were classified into a separate category (multiple users). Antidepressants, regardless of their class, were classified according to the degree of serotonin reuptake inhibition: strong inhibitors (duloxetine, fluoxetine, paroxetine, and sertraline; dissociation constant K_D < 1 nmol), intermediate inhibitors (citalopram, escitalopram, fluvoxamine, and venlafaxine; K_D = 1–10 nmol), and weak inhibitors (mianserin, mirtazapine, nefazodone, reboxetine, agomelatine, and viloxazine; K_D > 10 nmol).²³

In a subsequent analysis, current exposure to antidepressants (based on the exposure definition above) was categorized according to duration of use. The duration of continuous use of an antidepressant was defined by the number of days exposed before the index date and was calculated from the number and duration of successive prescriptions for the antidepressant of interest. Two prescriptions were considered continuous if the duration of 1 prescription overlapped with the date of the next prescription. For nonoverlapping prescriptions, we allowed a 30-day grace period between any 2 prescriptions and considered such patients still under treatment. Duration of use was stratified into 3 categories according to the tertiles of the distribution of use in the controls.

Statistical analysis

We used a nested case-control approach for the analysis of the cohort because of the time-varying nature of exposure, the large size of the cohort, and the long duration of follow-up.^24,25 Thus, conditional logistic regression was used to compute odds ratios (ORs) with 95% confidence intervals (CIs) of ischemic stroke/TIA and myocardial infarction separately. These ORs are unbiased estimators of incidence rate ratios (RRs), with little or no loss in precision.²⁶ Current use of antidepressants strongly inhibiting serotonin reuptake was compared with current use of antidepressants weakly inhibiting serotonin reuptake. By the matching process, all RRs were inherently matched on sex, age, duration of follow-up, and calendar time.

In addition, the following potential confounders were included in the multivariable analysis, measured 366 to 730 days before the index date: obesity (body mass index ≥30 kg/m²), smoking status, alcohol abuse, arterial hypertension, atrial fibrillation, congestive heart failure, coronary artery disease excluding prior myocardial infarction, peripheral vascular disease, hyperlipidemia, diabetes mellitus, chronic kidney disease, liver disease, depression, chronic obstructive pulmonary disease, and cancer. When ischemic stroke/TIA was studied, myocardial infarction was additionally included in the model; when myocardial infarction was studied, ischemic stroke and TIA were additionally included. We also measured the use of the following drugs in the year before index date: anticoagulants, antiplatelet agents, nonsteroidal anti-inflammatory drugs, opioids, lipid-lowering drugs, β-blockers, thiazides, angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, calcium channel blockers, antipsychotics, other antidepressants (tricyclic antidepressants, monoamine oxidase inhibitors, trazodone), lithium, and hormone replacement therapy, in addition to measures of health care use (number of physician visits).

All analyses were repeated to assess the rate of ischemic events according to the strength of serotonin reuptake inhibition of SSRIs only (i.e., excluding third-generation antidepressants). Specifically, we estimated separately the rate of stroke/TIA and myocardial infarction associated with current use of SSRIs strongly inhibiting serotonin reuptake (fluoxetine, paroxetine, and sertraline) compared with current use of SSRIs intermediately inhibiting serotonin reuptake (citalopram, escitalopram, and fluvoxamine) because no SSRIs weakly inhibit serotonin reuptake.

Secondary analyses

We conducted 2 secondary analyses. First, we assessed the risk for each of the 2 outcomes according to duration of current use of antidepressants strongly inhibiting serotonin reuptake. Second, to assess effect modification, we conducted stratified analyses with respect to sex, chronic kidney disease, and history of cardiovascular disease.

Sensitivity analyses

We performed 5 prespecified sensitivity analyses to assess the robustness of our results. First, to assess potential exposure misclassification, we extended the definition of current use to 60 days before the index date for all exposed cases and controls. Second, we restricted our outcome definition to ischemic stroke only by excluding Read codes for TIA. Third, to assess the effect of the potential adjustment for covariates in the causal pathway, we repeated the primary analysis while adjusting for covariates measured at cohort entry. Fourth, we used a multiple imputation technique for missing data (i.e., smoking status and body mass index).^27,28 Finally, to assess the potential for residual confounding, we repeated the primary analysis after trimming the cohorts using the high-dimensional propensity score (hdPS) approach.²⁹ Briefly, the hdPS algorithm identified the top 500 covariates that were most likely to bias the exposure-outcome association and was estimated with information up to and including the 365 days before cohort entry. The cohorts were trimmed by excluding the 2.5% of patients at either end of the distribution of the hdPS to ensure overlap in the distribution of covariates for the 2 groups being compared (strong inhibitors of serotonin reuptake vs weak inhibitors of serotonin reuptake or SSRIs strongly inhibiting serotonin reuptake vs SSRIs intermediately inhibiting serotonin reuptake).

We also performed a post hoc sensitivity analysis using propensity score matching as an alternative means to control for confounding. First, we calculated the probability (propensity) of initiating a strong inhibitor of serotonin reuptake vs a weak inhibitor using logistic regression and based on the variables listed in table 1. Patients initiating a strong inhibitor of serotonin reuptake and patients initiating a weak inhibitor of serotonin reuptake were matched 1:1 on closest propensity score using a caliper width equal to 0.2 of the pooled SD of the logit of the propensity score. Potential imbalance in baseline characteristics after matching was assessed with standardized differences. Patients were followed up from cohort entry, that is, drug initiation, until an incident stroke/TIA (or myocardial infarction, depending on the outcome being studied), drug discontinuation, switch to or add on of an antidepressant with a different degree of serotonin reuptake inhibition (e.g., from a strong to an intermediate or a weak inhibitor), or end of the study period, whichever occurred first. Using a Cox proportional hazards model, we estimated separately hazard ratios (HRs) and corresponding 95% CIs of ischemic stroke/TIA and myocardial infarction associated with the use of antidepressants strongly inhibiting serotonin reuptake compared with the use of antidepressants weakly inhibiting serotonin reuptake. All analyses were conducted with SAS version 9.4 (SAS institute, Cary, NC) and R (R Foundation for Statistical Computing, Vienna, Austria).

Standard protocol approvals, registrations, and patient consents

The study protocol was approved by the Independent Scientific Advisory Committee of the CPRD (protocol 17_219R2) and by the Research Ethics Committee of the Jewish General Hospital, Montreal, Québec, Canada. Written consent from participants was not required because of the use of anonymized data and no direct patient involvement.

Data availability

No additional data are available.