Citalopram and the Risk of Serious Adverse Events

Many older adults experience mental health disorders, including depression and anxiety, in conjunction with reduced kidney function. Citalopram, a selective serotonin reuptake inhibitor, is frequently prescribed to this population for the treatment of depression and anxiety. However, both aging and impaired kidney function can result in elevated plasma concentrations of citalopram, increasing the risk of adverse events. Prescribing references, product monographs, and other standard sources provide inconsistent recommendations regarding initial dose adjustments for patients with low kidney function. This inconsistency may contribute to variable dosing practices in clinical settings, where many older adults with reduced kidney function are started on citalopram at a dose of 20 mg per day. Compared to a lower dose of 10 mg per day, this higher dose may be associated with increased rates of falls, fractures, gastrointestinal bleeding, arrhythmias, hospitalization, and mortality. Despite these potential risks, real-world evaluations remain limited. Recent analyses using a high-throughput approach with Ontario healthcare databases identified an increased risk of adverse outcomes in patients with advanced chronic kidney disease (CKD), defined as an estimated glomerular filtration rate (eGFR) less than 45 mL/min/1.73 m² without a history of dialysis or kidney transplant, who were newly prescribed citalopram compared to a cohort of non-users. To further investigate and validate these preliminary findings, a population-based retrospective cohort study is planned among older adults in Ontario, Canada. This study will assess the 30-day risk of serious adverse events following initiation of high versus low doses of citalopram in older adults with reduced kidney function, with the goal of informing safer prescribing practices. Methods: This population-based retrospective cohort study will include older adults (aged 66 years or older) with an estimated glomerular filtration rate (eGFR) less than 45 mL/min/1.73 m2, who are not receiving dialysis and have no history of kidney transplantation. Participants will be identified from Ontario, and potentially Alberta, who have been dispensed a new outpatient prescription for oral citalopram (10 or 20 mg per day) with a minimum day supply of seven days. In Ontario, participant accrual will occur from January 1, 2008, to January 1, 2025. Based on the prescribed daily dose, individuals will be categorized into low-dose (10 mg/day) and high-dose (20 mg/day) groups. Propensity score weighting will be applied to balance baseline characteristics between groups. The primary outcome will be a 30-day composite of all-cause hospitalization, emergency department visit, or mortality. Objectives: The study will assess whether initiation of high-dose (20 mg/day) versus low-dose (10 mg/day) citalopram is associated with a higher 30-day risk of the composite outcome among older adults with reduced kidney function. Additionally, the study will evaluate whether baseline eGFR category modifies the risk of the 30-day composite outcome among patients initiating higher versus lower doses of citalopram in the outpatient setting. The cohort will be expanded to include all levels of baseline eGFR, and eGFR will be categorized as ≥60, 45-59, and \<45 mL/min per 1.73 m². Study Design and Setting: Data Sources: This study will use a population-based retrospective cohort design with linked administrative health data from Ontario, Canada. If the Ontario cohort does not provide sufficient sample size, Alberta health administrative data will be included using the same methodology. Ontario data will be sourced from ICES (ices.on.ca), which securely manages individual-level data for all Ontario residents covered by the province's single-payer healthcare system. Data use is authorized under section 45 of Ontario's Personal Health Information Protection Act, which does not require research ethics board review. The study will draw on several Ontario health databases: the Ontario Drug Benefit (ODB) for prescriptions, the Registered Persons Database (RPDB) for demographics and vital status, the Canadian Institute for Health Information Discharge Abstract Database (CIHI-DAD) and National Ambulatory Care Reporting System (NACRS) for hospitalizations and emergency visits, the Ontario Health Insurance Plan (OHIP) Database for physician billing, and the Ontario Laboratories Information System (OLIS) for laboratory data, including serum creatinine to estimate kidney function. If Alberta data are used, they will be accessed through the Alberta Kidney Disease Network (AKDN), with data available up to approximately 2021. The study protocol, including the description, design, and statistical analysis plan, will be publicly registered before outcome analyses. Results will be reported according to RECORD guidelines. Study Population: All older adults (≥66 years) with an eGFR less than 45 mL/min per 1.73 m2, who are not receiving dialysis and have no history of kidney transplantation, and who received a new outpatient prescription for oral citalopram at an initial dose of 10 or 20 mg/day with a day supply of at least 7 days between 2008 and 2024, will be included. The dispensing date of the prescription will be defined as the index date. Only the first eligible prescription per individual will be included to ensure unique cohort entry. Each individual will enter the cohort only once. Patients will be classified as either "high-dose" or "low-dose" citalopram users. Baseline Characteristics: Baseline variables will be obtained from health records, census files, hospital records, laboratory data, and physician claims. These variables will include demographic characteristics such as age, sex, rurality, and neighborhood income quintile, as well as comorbidities and medication use. Baseline comorbidities and healthcare utilization will be assessed using 5-year and 1-year look-back periods from the index date, respectively. Baseline medication use will be evaluated within 120 days prior to the index date. Eligible patients must have an outpatient serum creatinine measurement within 365 days prior to cohort entry to enable estimation of eGFR using the 2021 CKD-EPI equation. Patients with end-stage kidney disease, defined as those on dialysis or who have received a kidney transplant, will be excluded. Outcomes: The primary outcome is the 30-day composite of all-cause hospitalization, emergency department visit, or mortality. Only the first hospitalization or emergency department visit within 30 days is counted. The components of the composite outcome are accurately coded in ICES data. The observation window begins on the prescription fill date and continues for 30 days. Loss to follow-up is expected to be less than 0.3%, as the observation period is under 90 days and annual provincial emigration in this age group is below 0.5%. Participant observation will be censored at the first occurrence of the outcome of interest, death (if not the outcome of interest), or 30 days after the index date, whichever occurs first. Secondary outcomes will include the following: 1) Each component of the primary composite outcome, specifically 30-day all-cause hospitalization, all-cause emergency department visit, and all-cause mortality, analyzed separately. 2) A 30-day composite outcome of hospital admission or emergency department visit with delirium or encephalopathy (including disorientation, transient alteration of awareness, or other cognitive symptoms), or a hospital encounter involving an urgent computed tomography scan of the head. 3) A 30-day composite outcome of hospital admission or emergency department visit with fracture (hip, femur, humerus, wrist/forearm, pelvis, or spine), falls, hypotension, or syncope. 4) A 30-day hospital admission or emergency department visit with a composite outcome of atrial fibrillation/flutter, ventricular arrhythmia, or other arrhythmias (including pacemaker insertion, palpitations, unspecified tachycardia, atrioventricular block, supraventricular tachycardia, other conduction disorders, or implantable cardiac defibrillator). Statistical analysis plan: Categorical variables will be reported as frequencies and proportions. Continuous variables will be presented as means with standard deviations (SD) or medians with interquartile ranges (IQR), as appropriate. Baseline characteristics will be compared between the high-dose (20 mg/day) and low-dose (10 mg/day) groups using standardized mean differences (SMDs). An absolute SMD greater than 10% will indicate a meaningful imbalance. Balancing comparator group: Inverse probability of treatment weighting (IPTW) based on the propensity score will be used to balance baseline characteristics between the high-dose (20 mg/day) and low-dose (10 mg/day) groups, including all known predictors of citalopram use. Propensity scores will be estimated using a multivariable logistic regression model incorporating all baseline characteristics. The average treatment effect in the treated (ATT) weights will be applied, assigning patients in the low-dose group weights calculated as (propensity score / \[1 - propensity score\]), while patients in the high-dose group will receive a weight of 1. This approach will generate a weighted pseudo-sample of low-dose citalopram (10 mg/day) patients with a distribution of measured baseline characteristics similar to that of the high-dose citalopram (20 mg/day) group. Baseline characteristics will be compared between groups using standardized differences in both unweighted and weighted samples. Regression analysis: To assess the primary composite outcome of all-cause hospitalization, all-cause emergency department visit, or all-cause mortality, a modified Poisson regression will be used to estimate the risk ratio (95% CI), and binomial regression will be used to estimate the risk difference (95% CI) in the weighted cohort, with the low-dose citalopram (10 mg/day) group as the referent. Secondary analyses: All secondary outcomes will be tested independently without adjustment for multiple comparisons. Each test will be conducted and reported separately. Consistent with best practices, the primary outcome will be presented with its P-value, and all secondary outcomes will be reported as point estimates with 95% confidence intervals. Additional analyses: Five supplementary analyses are planned. 1. Effect measure modification (EMM): The cohort will be expanded to include all estimated glomerular filtration rate (eGFR) levels, categorized into three groups: eGFR ≥60, 45-\<60, and \<45 mL/min/1.73 m². Baseline characteristics between high-dose (20 mg/day) and low-dose (10 mg/day) citalopram groups will be assessed using standardized differences for all renal function categories combined and within each eGFR category. To balance baseline characteristics between high-dose and low-dose citalopram groups, the inverse probability of treatment weighting (IPTW) method, based on propensity scores, will be applied for all eGFR categories combined and within each category. EMM will be evaluated on both additive and multiplicative scales. For additive interaction, binomial regression with an identity link will be used to estimate risk differences, including an interaction term between dose groups and eGFR strata. For multiplicative interaction, modified Poisson regression will estimate risk ratios, including an interaction term between dose groups and eGFR strata. The estimated risk difference and risk ratio will be reported as point estimates with their 95% CI in the subgroups. 2. A high-throughput computing analysis involving over 700 population-based, new-user cohort studies, each comparing 74 acute (30-day) outcomes across kidney function strata (PMID: 38186562), suggested that citalopram use versus no use is associated with harm in patients with low kidney function. This analysis used Ontario data from January 1, 2008 to March 1, 2020. A subgroup analysis will restrict the cohort to periods before March 1, 2020 (overlapping with the high-throughput computing period) and after March 1, 2020 (no overlap). 3. Hazard ratios for all outcomes within 30 days will be computed to illustrate the effect of the intervention over time. Hazard ratios are expected to be similar to risk ratios. 4. E-value analyses will be performed to determine the minimum association strength an unmeasured confounder would require with both the prescription drug and the outcome to eliminate the observed association. 5. The risk of the primary outcome will be compared between new users of 20 mg/day of citalopram and older adults with no citalopram use (non-users). The same comparison will be conducted for new users of 10 mg/day of citalopram versus non-users. The same baseline characteristics and statistical analysis methods as in the primary analysis will be employed. \*This approach will be implemented only if analysis of Alberta data proceeds.\* Due to privacy regulations in Canada, individual-level patient data from different provinces cannot be linked. Instead, results from Ontario and Alberta will be combined using a privacy-preserving method to ensure data privacy and regulatory compliance. The proposed approach involves a single transfer of summary-level outputs from each province to the research team and will be adapted from the techniques described by Shu et al. 2025.