ABSTRACT
Importance On the grounds of their anti-inflammatory and potential antiviral effects, we hypothesized that SSRIs and SNRIs might be effective treatments for Covid-19.
Methods We examined the association of antidepressant use with intubation or death in hospitalized patients with COVID-19. The primary endpoint was a composite of intubation or death in time-to-event analyses adjusted for numerous potential confounders. We compared this endpoint between patients who received antidepressants and those who did not. The primary analyses were multivariable Cox models with inverse probability weighting.
Results Of the 9,509 inpatients with positive Covid-19 RT-PCR test, 2,164 patients (22.8%) were excluded because of missing data or exclusion criterion (i.e. less than 18 years old of age). Of the remaining 7,345 adult inpatients, 460 patients (6.3%) received an antidepressant during the hospitalization. The primary endpoint occurred respectively in 143 patients exposed to antidepressants (31.1%) and 1,188 patients who were not (17.3%). After adjusting for the older age [74.8 (SD=15.5) versus 56.8 (SD=19.3); Welch’s t-test=23.7; p<0.001] and the greater medical severity of patients receiving antidepressants, the primary analyses showed a significant negative association of the composite endpoint with exposure to any antidepressant (HR, 0.64; p<0.001), SSRIs (HR, 0.56; p<0.001), and SNRIs (HR, 0.57; p=0.034), but not with other antidepressant classes. Exposure to escitalopram, fluoxetine, and venlafaxine was significantly associated with lower risk of intubation or death (all p<0.05). Results were similar in multiple sensitivity analyses.
Conclusions SSRIs and SNRIs may be associated with lower risk of death or intubation in patients with COVID-19.
1. Introduction
Global spread of the novel coronavirus SARS-CoV-2, the causative agent of coronavirus disease 2019 (COVID-19), has created an unprecedented infectious disease crisis worldwide. In the current absence of a vaccine or curative treatment with published evidence-based clinical efficacy, the search for an effective treatment for patients with COVID-19 among all available medications is urgently needed.1,2
Both COVID-19 and severe acute respiratory syndrome (SARS) are characterized by an overexuberant inflammatory response3 and, for COVID-19, viral load is associated with the worsening of symptoms.4
Prior work suggests a significant reduction of overactive inflammatory processes observed in individuals with major depressive disorder following antidepressant treatment.5,6 A recent meta-analysis5 of studies conducted in this population, mainly including selective serotonin reuptake inhibitors (SSRIs) and serotonin–norepinephrine reuptake inhibitors (SNRIs), supports that antidepressants in general may lower plasma levels of several pro-inflammatory cytokines, including IL-6, TNF-α, and CCL-2, which are thought to be involved in the pathogenesis of severe COVID-19.7 However, a prior study suggests that these anti-inflammatory effects could be stronger for SSRIs, which could more potently inhibit microglial TNF-α and NO production through cAMP signaling regulation, than SNRIs.8
A recent in-vitro study9 also suggests antiviral effects of fluoxetine on SARS-CoV-2, although this effect was not observed for other serotonin reuptake inhibitors, including paroxetine and escitalopram.
On the grounds of their anti-inflammatory and potential antiviral effects, we hypothesized that antidepressants, and more specifically SSRIs and SNRIs, could be effective treatments in reducing the risk of respiratory failure in patients with COVID-19. Short-term use of low to moderate doses of antidepressants, and particularly of SSRIs and SNRIs, is generally well tolerated,10 and notably in older adults,11,12 who are the most prone to developing severe COVID-19.13
To our knowledge, no study has examined to date the efficacy of these medications for COVID-19. Observational studies of patients with COVID-19 taking medications for other indications can help determine their efficacy for COVID-19, decide which should be prioritized for randomized clinical trials, and minimize the risk for patients of being exposed to potentially harmful and ineffective treatments.
We took advantage of the Assistance Publique-Hôpitaux de Paris (AP-HP) Health Data Warehouse, which includes data on all patients with COVID-19 who had been consecutively admitted to any of the 39 Greater Paris University hospitals.
In this report, we examine the association between the use of any antidepressant, each class of antidepressants (i.e. SSRIs, SNRIs, tricyclic, tetracyclic and α2-antagonist antidepressants), and individual medications with the risk of intubation or death among adult patients who have been admitted to these medical centers with COVID-19.
We hypothesized that antidepressant use, and specifically SSRIs and SNRIs, would be associated with reduced risk of intubation or death in time-to-event analyses adjusted for potential confounders, including sex, age, obesity, smoking status, any medical condition associated with increased risk of severe COVID-19, any medication prescribed according to compassionate use or as part of a clinical trial, clinical and biological severity of COVID-19 at admission, mood or anxiety or other current psychiatric disorder, and any prescribed benzodiazepine or Z-drug, mood stabilizer, or antipsychotic medication.
2. Methods
2.1. Setting
We conducted this study at AP-HP, which comprises 39 hospitals, 23 of which are acute, 20 adult and 3 pediatric hospitals. We included all adults aged 18 years or over who have been admitted with COVID-19 to these medical centers from the beginning of the epidemic in France, i.e. January 24st, until April 1st. COVID-19 was ascertained by a positive reverse-transcriptase–polymerase-chain-reaction (RT-PCR) test from analysis of nasopharyngeal or oropharyngeal swab specimens. The Institutional Review Board of the AP-HP Clinical Data Warehouse approved this study on April 8th (CSE-20-20_COVID19). All procedures related to this work adhered to the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
2.2. Data sources
We used data from the AP-HP Health Data Warehouse (‘Entrepôt de Données de Santé (EDS)’). This warehouse contains all the clinical data available on all inpatient visits for COVID-19 to any of the 39 Greater Paris University hospitals. The data obtained included patients’ demographic characteristics, vital signs, laboratory test and RT-PCR test results, medication administration data, past and current medication lists, past and current diagnoses, discharge disposition, ventilator use data, and death certificates.
2.3. Variables assessed
We obtained the following data for each patient at the time of the hospitalization: sex; age, which was categorized based on the OpenSAFELY study results13 (i.e. 18-50, 51-70, 71-80, 81+); obesity, defined as having a body-mass index higher than 30 kg/m2 or an International Statistical Classification of Diseases and Related Health Problems (ICD-10) diagnosis code for obesity (E66.0, E66.1, E66.2, E66.8, E66.9); self-reported smoking status; any medical condition associated with increased risk of severe COVID-1913-16 based on ICD-10 diagnosis codes, including diabetes mellitus (E11), diseases of the circulatory system (I00-I99), diseases of the respiratory system (J00-J99), neoplasms (C00-D49), and diseases of the blood and blood-forming organs and certain disorders involving the immune mechanism (D5-D8); any medication prescribed according to compassionate use or as part of a clinical trial (e.g. hydroxychloroquine, azithromycin, remdesivir, tocilizumab, or sarilumab); clinical severity of COVID-19 at admission, defined as having at least one of the following criteria:17 respiratory rate > 24 breaths/min or < 12 breaths/min, resting peripheral capillary oxygen saturation in ambient air < 90%, temperature > 40°C, or systolic blood pressure < 100 mm Hg; and biological severity of COVID-19 at admission, defined as having at least one of the following criteria:17,18 high neutrophil-to-lymphocyte ratio or low lymphocyte-to-C-reactive protein ratio (both variables were dichotomized at the median of the values observed in the full sample), or plasma lactate levels higher than 2 mmol/L. To take into account possible confounding by indication bias for antidepressants, we recorded whether patients had any current mood or anxiety disorder (F30-F48) or any other current psychiatric disorder (F00-F29 and F50-F99) based on ICD-10 diagnosis codes, and whether they were prescribed any benzodiazepine or Z-drug, any mood stabilizer (i.e. lithium or antiepileptic medications with mood stabilizing effects), or any antipsychotic medication.
All medical notes and prescriptions are computerized in Greater Paris University hospitals. Medications and their mode of administration (i.e., dosage, frequency, date, condition of intake) were identified from medication administration data or scanned hand-written medical prescriptions, through two deep learning models based on BERT contextual embeddings,19 one for the medications and another for their mode of administration. The model was trained on the APmed corpus,20 a previously annotated dataset for this task. Extracted medications names were then normalized to the Anatomical Therapeutic Chemical (ATC) terminology using approximate string matching.
2.4. Exposure to antidepressants
Study baseline was defined as the date of hospital admission. Antidepressant exposure was defined as receiving any antidepressant at any time during the follow-up period, from study baseline to the end of the hospitalization or intubation or death.
2.5. Endpoint
The primary endpoint was the time from study baseline to intubation or death. For patients who died after intubation, the timing of the primary endpoint was defined as the time of intubation. Patients without an end-point event had their data censored on May 20th, 2020.
2.6. Statistical analysis
We calculated frequencies and means (± standard deviations (SD)) of each baseline characteristic described above in patients receiving or not receiving antidepressants and compared them using chi-square tests or Welch’s t-tests.
To examine the associations between the use of any antidepressant, each class of antidepressants, and individual medications with the composite endpoint of intubation or death, we performed Cox proportional-hazards regression models. To help account for the nonrandomized prescription of antidepressants and reduce the effects of confounding, the primary analysis used propensity score analysis with inverse probability weighting.21,22 The individual propensities for exposure were estimated by multivariable logistic regression models that included sex, age, obesity, smoking status, any medical condition associated with increased risk of severe COVID-19, any medication prescribed according to compassionate use or as part of a clinical trial, clinical and biological severity of COVID-19, mood or anxiety or other current psychiatric disorder, and any prescribed benzodiazepine or Z-drug, mood stabilizer, and antipsychotic medication. In the inverse-probability-weighted analyses, the predicted probabilities from the propensity-score models were used to calculate the stabilized inverse-probability-weighting weights.21 Associations between any antidepressant, each class of antidepressants, and individual treatments with the primary endpoint were then estimated using multivariable Cox regression models including the inverse-probability-weighting weights. Natural logarithm transformation of weights was used when proportional hazards assumption was not met. Kaplan-Meier curves were performed using the inverse-probability-weighting weights,23 and their pointwise 95% confidence intervals were estimated using the nonparametric bootstrap method.24
We conducted sensitivity analyses, including multivariable Cox regression models comprising as covariates the same variables as the inverse-probability-weighted analyses, and univariate Cox regression models in matched analytic samples. For this latter analysis, we selected one control for each exposed case for exposures to any antidepressant and each class of antidepressants, and two controls for each exposed case for individual antidepressant medications, based on the same variables used for both the inverse-probability-weighted and the multivariable Cox regression analyses. Weighted Cox regression models were used when proportional hazards assumption was not met. To reduce the effects of confounding, optimal matching was used in order to obtain the smallest average absolute distance across all clinical characteristics between exposed patient and non-exposed matched controls. We also performed multivariable Cox regression models including interaction terms to examine whether the association between antidepressant exposure and the primary endpoint differed across subgroups defined by baseline characteristics.
Within patients exposed to antidepressants, we tested the association of daily dose (converted into fluoxetine-equivalent dose25 and dichotomized at the median value) with the primary endpoint. Furthermore, we examined whether exposure to a combination of antidepressants was associated with a different risk of intubation or death than exposure to only one antidepressant.
We reproduced these analyses (i) among patients with critical COVID-19 hospitalized in intensive care units (ICUs), and (ii) using death and intubation as separate endpoints. Finally, among patients with COVID-19 not exposed to antidepressants during the hospitalization, we examined whether the risk of intubation or death differed between patients taking or not antidepressants in the 3 months before hospital admission.
For all associations, we performed residual analyses to assess the fit of the data, check assumptions, including proportional hazards assumptions, and examined the potential influence of outliers. To improve the quality of result reporting, we followed the recommendations of The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Initiative.26 Statistical significance was fixed a priori at two-sided p-value<0.05. All analyses were conducted between May 28th and June 20th in R software version 2.4.3 (R Project for Statistical Computing).
3. Results
3.1. Characteristics of the cohort
Of the 9,509 patients with a positive COVID-19 RT-PCR test consecutively admitted to the hospital, a total of 2,164 patients (22.8%) were excluded because of missing data or their young age (i.e. less than 18 years old of age). Of the remaining 7,345 adult inpatients, 460 patients (6.3%) received an antidepressant during the hospitalization, at a mean fluoxetine-equivalent dose of 21.4 mg (SD=13.6) per day (Figure 1). Doses of each antidepressant are shown in eTable 1. Among patients exposed to antidepressants, 391 (85.0%) were exposed to only one antidepressant, 69 (14.1%) received two antidepressants, and 4 (0.9%) patients were exposed to more than two different antidepressants. Mean age of patients exposed to antidepressants was 74.8 (SD=15.5) years, whereas it was 56.8 (SD=19.3) years in those who were not (Welch’s t-test=−23.70, p<0.001).
RT-PCR test results were obtained after a median delay of 1 day (SD=12.1) from the date of hospital admission. This delay was not significantly different between patients receiving or not receiving antidepressants [median in the exposed group = 1 day (SD=14.1); median in the non-exposed group = 1 day (SD=11.9); Mood’s median test Chi-square=0.21, p=0.650)].
Over a mean follow-up of 18.5 days (SD=27.1; median=4 days; range: 1 day to 117 days), 1,331 patients (18.1%) had a primary end-point event at the time of data cutoff on May 20th. In patients exposed to antidepressants, the mean follow-up was 20.1 days (SD=23.2; median=11 days; range: 1 day to 112 days), while it was of 18.4 days (SD=27.4; median=4 days; range: 1 day to 117 days) in those who were not. The times to follow-up by treatment exposure are shown in eTable 2.
All baseline characteristics were independently and significantly associated with the primary endpoint, except for smoking, any medication prescribed according to compassionate use or as part of a clinical trial, any current mood or anxiety disorder, and any other current psychiatric disorder (Table 1).
The distributions of the patients’ characteristics according to antidepressant exposure, and exposure by class of antidepressants and by individual medication are shown in eTables 3 to 13. In the full sample, antidepressant exposure significantly differed according to all baseline characteristics, as did exposure to SSRIs (except for any medication prescribed according to compassionate use or as part of a clinical trial), SNRIs (except for sex and any medication prescribed according to compassionate use or as part of a clinical trial), tricyclic antidepressants (except for smoking and any current psychiatric disorder outside mood and anxiety disorders), tetracyclic antidepressant (except for sex, obesity, smoking, and any medication prescribed according to compassionate use or as part of a clinical trial), and α2-antagonist antidepressants (except for obesity, smoking, any medication prescribed according to compassionate use or as part of a clinical trial, any current psychiatric disorder outside mood and anxiety disorders, and clinical severity of COVID-19 at admission), and the direction of associations indicated older age and greater medical severity of patients exposed to antidepressants than those who were not. After applying the propensity score weights, these differences were substantially reduced (eTables 3 to 13). In the matched analytic samples, there were no significant differences in any characteristic across the different exposures, except for antidepressant exposure, for which sex ratio significantly differed (eTables 3 to 13).
3.2. Study endpoint
Among patients receiving antidepressants, SSRIs, SNRIs, tricyclic antidepressants, tetracyclic antidepressants, and α2-antagonist antidepressants, the primary endpoint of intubation or death occurred respectively in 143 patients (31.1%), 75 patients (29.2%), 18 patients (25.4%), 19 patients (32.2%), 43 patients (45.7%) and 12 patients (27.3%), while 1,188 non-exposed patients (17.3%) had this outcome (Table 2). When adjusting for the older age and the greater medical severity of patients who received an antidepressant than those who did not, the primary multivariable analyses with inverse probability weighting showed a significant negative association of the composite endpoint with exposure to any antidepressant (HR, 0.64; 95% CI, 0.51 to 0.80, p<0.001) (Figure 2), SSRIs (HR, 0.56; 95% CI, 0.42 to 0.75, p<0.001), and SNRIs (HR, 0.57; 95% CI, 0.34 to 0.96, p=0.034) (Figure 3), and no significant association with exposure to tricyclic antidepressants (HR, 0.96; 95% CI, 0.57 to 1.61, p=0.880), tetracyclic antidepressants (HR, 0.87; 95% CI, 0.61 to 1.25, p=0.454), and α2-antagonist antidepressants (HR, 0.29; 95% CI, 0.28 to 1.02, p=0.058) (Table 2; Figure 3). Among hospitalized patients with COVID-19, exposure to escitalopram (HR, 0.61; 95% CI, 0.40 to 0.92, p=0.018), fluoxetine (HR, 0.32; 95% CI, 0.14 to 0.73, p=0.007), or venlafaxine (HR, 0.47; 95% CI, 0.24 to 0.91, p=0.025) was significantly associated with a reduced risk of intubation or death. There were no significant differences in this risk according to exposure to any other molecule (Table 2; eFigures 1 to 5). The association between antidepressant exposure and the outcome did not significantly differ across subgroups defined by baseline characteristics, except for biological severity of COVID-19 at admission (eTable 14).
In sensitivity analyses, multivariable Cox regression models in the full sample yielded similar results, as did univariate Cox regression models in the matched analytic samples, except for exposure to SNRIs which was not significant (HR, 0.85; 95% CI, 0.32 to 1.05, p=0.074), and exposures to paroxetine (HR, 0.62; 95% CI, 0.40 to 0.96, p=0.034) and mirtazapine (HR, 0.52; 95% CI, 0.27 to 0.99, p=0.046), which were significantly associated with reduced risk of intubation or death in these analyses (Table 2; eFigures 1 to 5).
Exposure to higher rather than lower doses of antidepressants was not significantly associated with the composite outcome, except for tetracyclic antidepressants, for which higher doses were associated with increased risk of intubation or death (eTable 15). Exposure to a combination of antidepressants did not significantly modify this risk compared to exposure to only one antidepressant (eTable 16).
When using intubation and death as separate endpoints, we found that exposures to any antidepressant and SSRIs, but not to other classes of antidepressants, were significantly associated with lower risks (eTable 17).
Among the 593 patients admitted to ICUs, antidepressant exposure was significantly associated with reduced risk of intubation or death in the primary multivariable Cox model with inverse probability weighting, but not in other multivariable analyses, possibly because of a lack of statistical power, as only 28 of these patients had an antidepressant (eTable 18).
Finally, among patients with COVID-19 not exposed to any antidepressant during the hospitalization, there was no significant association between use of antidepressants or SSRIs in the 3 months before hospital admission and the primary endpoint (eTable 19).
Discussion
In this observational multicenter study involving a large sample of patients admitted to the hospital with COVID-19, we found that exposure to SSRIs or SNRIs, and particularly to fluoxetine, escitalopram and venlafaxine, may be associated with reduced risk of intubation or death. This association was observed at a mean dosage of 21.4 (SD=13.6) fluoxetine-equivalent milligrams in patients receiving antidepressants. Exposure to any other class of antidepressants was not associated with this risk. These findings should be interpreted with caution due to the observational design and the fact that this is, to our knowledge, the first study examining the efficacy of antidepressants in a clinical population of patients with COVID-19. However, our results suggest that SSRIs and SNRIs may be associated with reduced risk of intubation or death for COVID-19.
In the analyses, we tried to minimize the effects of confounding in several different ways. First, we used multivariable regression models with inverse probability weighting to minimize the effects of confounding by indication.21,22 We also performed sensitivity analyses, including multivariable Cox regression models and univariate Cox regression models in matched analytic samples, that showed similar results, giving strength to our conclusions. Second, although some amount of unmeasured confounding may remain, our analyses adjusted for numerous potential confounders, including sex, age, obesity, smoking status, any medical condition associated with increased risk of severe COVID-19, any medication prescribed according to compassionate use or as part of a clinical trial, clinical and biological severity of COVID-19, mood or anxiety or other current psychiatric disorder, and any prescribed benzodiazepine or Z-drug, mood stabilizer, and antipsychotic medication. Furthermore, our findings indicate that these associations were not significantly different across subgroups defined by baseline characteristics, except for biological severity of COVID-19 at hospital admission, and only observed for patients receiving these medications during the visit, but not if they were only exposed before the date of hospital admission. Third, these associations remained significant for any antidepressant and SSRIs when using death and intubation as separate endpoints. Finally, the low number of patients with antidepressants admitted to ICUs (N=28) in our study, and the significant negative association of antidepressant exposure with reduced risk of intubation or death among patients admitted to ICUs in the propensity score analysis with inverse probability weighting, further give strength to our conclusion.
Additional limitations of our study include missing data for some baseline characteristic variables, including baseline clinical and biological severity of COVID-19, which may be explained by the overwhelming of all hospital units during the COVID-19 peak incidence, and potential for inaccuracies in the electronic health records in this context, such as the possible lack of documentation of illnesses or medications, or the misidentification of treatments’ mode of administration (e.g., dosage, frequency), especially for hand-written medical prescriptions. Furthermore, type I error inflation due to multiple testing may have occurred in our study. However, our analyses were exploratory, and results were similar across different statistical approaches. Finally, despite the multicenter design, our results may not be generalizable to other settings or regions.
In this multicenter observational study involving patients admitted to the hospital with COVID-19, exposure to SSRIs or SNRIs at usual antidepressant doses during the visit may be associated with lower risk of intubation or death. Our results suggest that double-blind controlled randomized clinical trials of these specific medications for COVID-19 are urgently needed.
Data Availability
Data from the AP-HP Health Data Warehouse can be obtained at https://eds.aphp.fr//.
Acknowledgments
The authors thank the EDS APHP Covid consortium integrating the APHP Health Data Warehouse team as well as all the APHP staff and volunteers who contributed to the implementation of the EDS-Covid database and operating solutions for this database. Collaborators of the EDS APHP Covid consortium are: Pierre-Yves ANCEL, Alain BAUCHET, Nathanaël BEEKER, Vincent BENOIT, Mélodie BERNAUX, Ali BELLAMINE, Romain BEY, Aurélie BOURMAUD, Stéphane BREANT, Anita BURGUN, Fabrice CARRAT, Charlotte CAUCHETEUX, Julien CHAMP, Sylvie CORMONT, Christel DANIEL, Julien DUBIEL, Catherine DUCLOAS, Loic ESTEVE, Marie FRANK, Nicolas GARCELON, Alexandre GRAMFORT, Nicolas GRIFFON, Olivier GRISEL, Martin GUILBAUD, Claire HASSEN-KHODJA, François HEMERY, Martin HILKA, Anne Sophie JANNOT, Jerome LAMBERT, Richard LAYESE, Judith LEBLANC, Léo LEBOUTER, Guillaume LEMAITRE, Damien LEPROVOST, Ivan LERNER, Kankoe LEVI SALLAH, Aurélien MAIRE, Marie-France MAMZER, Patricia MARTEL, Arthur MENSCH, Thomas MOREAU, Antoine NEURAZ, Nina ORLOVA, Nicolas PARIS, Bastien RANCE, Hélène RAVERA, Antoine ROZES, Elisa SALAMANCA, Arnaud SANDRIN, Patricia SERRE, Xavier TANNIER, Jean-Marc TRELUYER, Damien VAN GYSEL, Gaël VAROQUAUX, Jill Jen VIE, Maxime WACK, Perceval WAJSBURT, Demian WASSERMANN, Eric ZAPLETAL.
Footnotes
Conflicts of interest NH has received personal fees and non-financial support from Lundbeck, outside the submitted work. FL has received speaker and consulting fees from Janssen-Cilag, Euthérapie-Servier, and Lundbeck, outside the submitted work. CL reports personal fees and non-financial support from Janssen-Cilag, Lundbeck, Otsuka Pharmaceutical, and Boehringer Ingelheim, outside the submitted work. GA reports personal fees from Pfizer, Pierre Fabre and Lundbeck, outside the submitted work. Other authors declare no competing interests.
Data Availability Statement Data from the AP-HP Health Data Warehouse can be obtained at https://eds.aphp.fr//.
Authorship NH designed the study, performed statistical analyses, and wrote the first draft of the manuscript. MSR contributed to study design, performed statistical analyses and critically revised the manuscript. FL contributed to study design and critically revised the manuscript for scientific content. RV contributed to statistical analyses and critically revised the manuscript for scientific content. NB and ASJ contributed to study design and critically revised the manuscript for scientific content. NB, ASJ, AN, NP, CD, AG, GL, MB, and AB contributed to database build process. AN, CB, MO, CL, GA, NP, CD, AG, GL, MB, and AB critically revised the manuscript for scientific content.
Funding source This work did not receive any external funding.