Rationale and background
Venlafaxine is a dual-acting serotonin and norepinephrine reuptake inhibitor (SNRI) indicated for the treatment of depression, prevention of relapse and prevention of recurrence of depression, anxiety or generalized anxiety disorder, social anxiety disorder, and panic disorder.
Venlafaxine has a dose-dependent pharmacodynamic profile. At lower doses it primarily inhibits serotonin reuptake, while at higher doses (=150 mg/day), it additionally inhibits norepinephrine reuptake and exerts weak inhibitory effects on dopamine reuptake.
Several cardiovascular effects are already labelled in section 4.8 of the Summaries of product characteristics (SmPC ) in Europe, e.g., tachycardia and hypertension, Torsade de pointes, ventricular tachycardia, ventricular fibrillation, prolonged QT, and stress cardiomyopathy (Takotsubo cardiomyopathy).
A published case series and spontaneously reported cases (e.g., EudraVigilance, Vigibase, company global pharmacovigilance safety database) suggest an association between venlafaxine and cardiotoxicity (heart failure, cardiomyopathy other than Takotsubo). An observational study using high-quality data from the DARWIN EU network could add important further evidence to evaluate the potential association.
The European Medicines Agency (EMA) has therefore requested a study to estimate the association between venlafaxine and heart failure/cardiomyopathy, using mirtazapine, which is indicated for the treatment of depression, as active comparator.
Research question and objectives
Research question
Is the new use of venlafaxine associated with an increased risk of incident heart failure compared to new use of mirtazapine?
Objectives
The study aims to assess the potential association between venlafaxine and incident heart failure among adults.
The specific objectives of this study are:
1) To characterise both new venlafaxine and mirtazapine users separately at the time of treatment start (in terms of demographics, pre-defined comorbidities/conditions of interest and drug utilisation), overall and stratified for condition of interest, age group and sex.
2) To assess the risk of incident heart failure and cardiomyopathy between new users of venlafaxine vs. new users of mirtazapine, stratified for condition of interest, age group, sex, history of cardiovascular (CV) disease and previous Selective Serotonin Reuptake Inhibitor (SSRI) use.
Methods
Study design
•Patient-level characterisation (Objective 1)
•New user active comparator cohort study (Objective 2)
Index date will be the date of first prescription of venlafaxine or mirtazapine, and individuals are followed up until date of last data availability, drug discontinuation, outcome occurrence, death, or end of study period.
Population
IMP-226-CT V04
The study population will include all individuals with a recorded first prescription for venlafaxine or mirtazapine within the study period from 01/01/2010 up to the end of data availability who meet the eligibility criteria at study entry.
Other eligibility criteria are:
•At least 365 days of data source history prior to index date.
•Aged = 18 years at index date.
•No prescription / dispensation of other antidepressants recorded in the data source in the year before index date – except SSRIs, which are permitted.
•No diagnosis or record of heart failure or cardiomyopathy ever recorded in the data source before index date (Objective 2).
•No prescription/dispensation record of both venlafaxine and mirtazapine at index date.
For Objective 2, a sensitivity analysis will be conducted requiring 5 years of prior history for those countries where such data is typically available (Denmark, Sweden, Finland).
Variables
Exposure:
•Venlafaxine (exposure) [ATC-code: N06AX16]
•Mirtazapine (active comparator) [ATC-code: N06AX11]
Outcome:
•Objective 1: Summarised patient characteristics at index date
– Objective 2: o Primary outcome: Incident heart failure (any)
– Secondary outcomes: broad/narrow definitions for incident heart failure; incident cardiomyopathy (broad, primary), HF hospitalisation
Data sources
1. Denmark: Danish Data Health Registries (DK-DHR)
2. Finland: Finnish Care Register for Health Care (FinOMOP-THL)
3. Spain: Base de Datos para la Investigación Farmacoepidemiológica en el Ámbito Público (BIFAP)
4. Spain: The Information System for Research on Primary Care (SIDIAP)
5. Sweden: Health Impact – Swedish Population Evidence Enabling Data-linkage (HI-SPEED)
6. The United Kingdom: Clinical Practice Research Datalink GOLD (CPRD GOLD)
Study size
Based on a preliminary feasibility assessment, the expected number of person counts for venlafaxine in the data sources included in this study range from 184,600 (HI-SPEED) to 303,200 (CPRD GOLD). The expected number of person counts for mirtazapine in the data sources included in this study range from 266,900 (SIDIAP) to 1,088,000 (BIFAP). The expected number of person counts for heart failure in the data sources included in this study range from 252,200 (CPRD GOLD) to 936,600 (BIFAP).
In the DARWIN P3-C3-001 (Adverse events of special interest) study, background rates of incident heart failure were estimated to be 138 per 100,000 person-years in CPRD GOLD and 386 per 100,000 person-years in SIDIAP. Using these incidence rates, with an alpha of 0.05, power of 80%, and follow up of 10 years, 30,806 and 10,708 patients in each treatment group are required to detect a hazard ratio of 1.2 in CPRD GOLD and SIDIAP, respectively. Using the same parameters, to detect a hazard ratio of 1.5, 5,589 patients in CPRD GOLD and 1,936 patients in SIDIAP for each treatment group are required, respectively.
Statistical analysis
IMP-226-CT V04
A minimum cell count of 5 will be used when reporting results, with any smaller count reported as “<5” and zero counts as “0”. All results will be presented separately by data source. For objective 2, random-effects meta analyses will be conducted to pool results.
Objective 1:
A summary of patient characteristics will be provided. Covariates of interest will also be reported as counts and proportions, including demographics, pre-defined comorbidities (incl. cardiovascular diseases) and comedication, and drug utilisation, stratified separately by indication (depression/anxiety) and age group. CohortCharacteristics and DrugUtilisation R packages will be used.
Objective 2:
We will use large-scale propensity score matching (PS) to address confounding from observed covariates, and match venlafaxine vs. mirtazapine patients based on a combination of pre-defined key variables for exact matching [age bands, sex, calendar year, history of CV disease, previous SSRI use, condition of interest (depression/anxiety)], and data-driven analyses of confounders based on the exposure with treatment use.
Among the PS matched population, Poisson regression will be used to estimate incidence rates for the outcomes of incident heart failure (HF) and cardiomyopathy among the two treatment groups of new users of venlafaxine and mirtazapine.
Cox proportional hazards modelling will then be used to estimate hazard ratios (HRs) for incident HF and cardiomyopathy, respectively. Results will be stratified for age, sex, history of CV disease, previous SSRI use and condition of interest (depression/anxiety).
The proportional hazards assumption will be tested using visual inspection of log-log plots. In the scenario that the assumption is violated, Incidence Rate Ratios (IRR) will be calculated using Poisson regression and presented alongside the HR.
To detect residual confounding, Negative Control Outcome (NCO) analyses will be conducted for each outcome among the overall population and subgroups of previous SSRI use and condition of interest.
A sensitivity analysis with a population restricted to 5 years of prior history will be conducted for the overall population for each outcome in HI-SPEED, DK-DHR and FinOMOP-THL
All results will be presented separately by data source. For objective 2, random-effects meta analyses will be conducted to pool results.