Study design, setting, and population
We conducted an international, retrospective, pharmacovigilance disproportionality analysis using the WHO’s VigiBase® pharmacovigilance database. VigiBase® contains more than 30 million individual case safety reports received from 160 members countries since 1967.
We used included VigiBase® data from the database’s inception to March 1st, 2022. We restricted our analysis to the individual case safety reports on people aged 65 or over. The study protocol was registered at ClinicalTrials.gov (NCT05356078). In line with the French legislation on retrospective, anonymized studies of routine medical practice (MR-004), and in accordance with the European regulation of April 27, 2016 on the protection of individuals with regard to the processing of data to personal character, the study protocol was approved by a hospital committee (C.L.E.R.S Comité Local d’Ethique de la Recherche en Santé) with competency for research not requiring authorization by an Institutional Review Board (University of Caen Normandy (Caen, France); reference: 2646, dated July 15th, 2021).
Variables
Each individual case safety report include administrative data (country, type of report, and type of reporter), sociodemographic data (age and sex), the time to the onset of the ADE, the outcome (coded according to the Medical Dictionary for Regulatory Activities (MedDRA) version 24.0), the WHO causality assessment, and the drug(s) involved (drug name, start and stop dates, time to onset, indication, dose, dechallenge, and rechallenge).
In VigiBase®, drugs are coded using the WHODrug Global dictionary. Antidepressant classes were based on the Anatomical Therapeutic Chemical (ATC) hierarchical classification and classified into NSMRIs, SSRIs, SNRIs, monoamine oxidase inhibitors (MAOIs), alpha-2-adrenergic receptor antagonists, and other antidepressants (for details, see Supplementary Table 1). Most studies included SNRIs and alpha-2-adrenergic receptor antagonists in an “other antidepressants” class; however, in view of their specific pharmacodynamic properties and their frequency of use, we decided to consider these two classes in their own right [5, 8]. We also noted the most frequently prescribed antidepressant drugs within each class, defined as those mentioned in more than 1000 reports (regardless of the type of ADE) in VigiBase®.
ADEs were coded according to the MedDRA terminology. In the present study, the event “delirium” encompassed the MedDRA terms “Delirium”, “Confusional state” and “Disorientation”. A concomitant delirium-hyponatremia event was defined as a report in which delirium (as defined above) and hyponatremia at the same time were reported.
We selected all types of reports, regardless of whether the antidepressant was suspected to be responsible for delirium, concomitantly prescribed with another suspected drug, or thought to be interacting with another drug.
Outcomes
The primary outcome was the association between antidepressant use and reports of delirium among people aged 65 or over. The secondary outcomes included (i) the association between the most frequently prescribed antidepressants and reports of delirium, (ii) the association between antidepressant classes and concomitant delirium-hyponatremia events, and (iii) the same associations by age class (65 to 74 vs. 75 and over).
Statistical methods
A case/non-case disproportionality analysis was used to probe the effect of antidepressant prescription on reports of delirium; this type of analysis has been described in detail previously [9]. The case/non-case disproportionality method is recommended for detecting a signal for an association between a drug and an adverse event in a pharmacovigilance database. A signal is present when the number of reports of an adverse event is greater than expected. This is referred to as a disproportionate reporting rate of an adverse event, relative to others. In the present study, a signal corresponded to a statistically significant difference in the distribution of cases of delirium related to antidepressants or classes of antidepressant vs. cases of delirium related to drugs other than antidepressants. The strength of the disproportionality was quantified as the reporting odds ratio (r-OR) and their 95% confidence interval ([95%CI]) estimated with univariate and multivariate logistic regression models. An r-OR was considered statistically significant if it was higher than 1 and if the lower boundary of its 95%CI did not include 1. An r-OR lower than 1 was considered statistically unsignificant, as the method was unable to detect the absence of a signal.
Disproportionality analyses typically include positive controls (i.e. drugs or drugs classes established to trigger the ADE of interest) and negative controls (i.e. drugs that are not known to trigger the ADE of interest); if the obtained results and the expected results are consistent, major sources of bias are likely to be absent. In the present multivariate analysis, we chose natural opium alkaloids (ATC N02AA) as positive controls and bisphosphonates (ATC M05BA) as negative controls [3].
Our models were adjusted for potential confounders: age class (65-74, 75 and over), sex, geographic region, and the major potentially associated prescriptions of drugs and illnesses reported in the literature as inducing delirium (opioids (ATC N02), antipsychotics (ATC N05A), anxiolytics (ATC N05B), and hypnotics (ATC N05C), constipation, acute urinary retention, alcohol use, unspecified infections, drug misuse, dementia, dehydration, hyponatremia (“hyponatraemia”, according to the spelling used in MedDRA), anticholinergic syndrome, hypoglycemia (“hypoglycaemia”, according to the spelling used in MedDRA), seizure disorder, drug abuse, drug dependence, drug withdrawal, central nervous system vascular disorders, hearing impairment, and visual impairment [2, 3, 10,11,12]; for details, see Supplementary Table 2. We studied the collinearity of our final models by computing the variance-inflation factors.
Sensitivity analyses were conducted on a multivariate model for the primary outcome by using three different definitions of delirium: (i) “Delirium” or “Confusional state”; (ii) “Delirium” or “Confusional state” or “Disorientation” or “Circadian rhythm sleep disorder”; and (iii) “Delirium” or “Confusional state” or “Disorientation” or “Circadian rhythm sleep disorder” or “Hallucinations”.
In analyses of pharmacovigilance databases, it is sometimes not possible to assess certain variables because the latter data are missing or because the patient did not meet certain clinical or pharmacologic criteria. In the present study, the person’s age was specified for all reports; hence, sex was the only variable for which missing data could have been imputed. As less than 1% of the data for this variable were missing, we decided not to impute them and so performed a complete case analysis.
All statistical analyses were performed using R software (v 4.0.2, R Studio v1.4.1717) and its packages rlang, dplyr, stats, base, fst, data.table, magrittr, openxlsx, carData, car, grid, and checkmate [13].
Add Comment