Fluoxetine pharmacokinetics and tissue distribution quantitatively supports a therapeutic role in COVID-19 at a minimum dose of 20 mg per day

Pharmacokinetic model

The pharmacometrics model that incorporates differential equation variables and the respective variances for a structural one-compartment pharmacokinetic model with first-order absorption was referenced from the United State Food and Drug Administration’s (FDA) Center for Drug Evaluation and Research Clinical Pharmacology and Biopharmaceutics Review(s) for a New Drug Application (NDA) 19-936 SE5-064 for Prozac (Fluoxetine Hydrochloride) that was submitted by Eli Lilly (https://www.accessdata.fda.gov/drugsatfda_docs/nda/2003/18936S064_Prozac%20Pulvules_biopharmr.pdf) (Center for Drug Evaluation and Research, 2002). For this study, the pharmacokinetic parameters referenced from the NDA were calculated by the FDA after combining 3 pharmacokinetic study datasets resulting in the following pharmacometric parameters: volume of distribution (Vd) value of 1,480 liters (variance [ω], 0.22), clearance rate (CL) value of 29.1 liters/hour (ω=0.376), and fixed absorption rate (Ka) of 0.67 (1/hour) (ω=not applicable as value was fixed) (Center for Drug Evaluation and Research, 2002). As noted in the fluoxetine NDA, the influence of age was neither relevant for clearance nor volume of distribution (Center for Drug Evaluation and Research, 2002).

Target fluoxetine plasma concentration to achieve the EC50 and EC90 lung concentrations

The molecular weight of fluoxetine hydrochloride is 345.8 g/mol and the reported EC50 (0.82 μM) and EC90 (4.02 μM) values from the Schloer et al. study are equivalent to EC50=283.6 ng/ml and EC90=1390.1 ng/ml, respectively. For all calculations, the trough target plasma concentrations will be referenced from the Schloer et al. study who reported after a 48-hour incubation period in Calu-3 lung cells (Schloer et al., 2020) which are significantly higher than the EC90 in Vero E6 cell results from Zimniak et al. study (Schloer et al., 2020; Zimniak et al., 2020, 2021).

Dosing simulations

To estimate the percentage of patients from a population of one thousand simulated patients who would achieve the trough target EC90 concentration, pharmacokinetic dosing of fluoxetine consisted of three dosing trials of fluoxetine: 20 mg/day, 30 mg/day, 40 mg/day, 50 mg/day, and lastly 60 mg/day.

Software and statistics

All pharmacokinetic dosing simulations are conducted with a population of 1,000 patients using mrgsolve and pharmacokinetic parameter estimates using PKNCA in R version 3.6.3 (R Core Team, 2015). The overall R script has been adapted from a study published in Clinical Pharmacology and Therapeutics using hydroxychloroquine (Al-Kofahi et al., 2020). Statistical results providing percentage estimates are calculated from trough concentrations of patients achieving the effective concentrations and is referenced from the Schloer et al. study reporting the EC50 and EC90 values in human-lung Calu-3 cells (Schloer et al., 2020).

Brain EC50 and EC90 concentrations

Extrapolating from in vitro to in vivo concentrations are dependent on intracellular versus extracellular concentrations, as well as the methodology of quantifying either whole-blood versus plasma concentrations in human pharmacokinetic studies. The EC50 and EC90 target concentrations represent the extracellular fluoxetine concentrations in the SARS-CoV-2 cell culture media. As COVID-19 is known to affect the brain during active infection and in post-COVID-19, adequate brain concentrations would be clinically important in patients who may experience depression. Bolo et al. reported fluoxetine brain concentrations, at steady-state, using fluorine magnetic spectroscopy and showed fluoxetine concentrations were 10-times higher in the brain than in human plasma (Bolo et al., 2000). Specifically, Bolo et al. found in study participants taking oral doses (10 mg, n=1; 20 mg, n=1; 40 mg, n=2) with a treatment period ranging from three months to 12-months that fluoxetine human brain concentrations were 13 μM (SD=7) versus 1.73 μM (SD=1.00) in human plasma fluoxetine (Bolo et al., 2000). In comparison, Johnson et al. found the coefficients for tissue distribution of fluoxetine relative to whole-blood was: 60× higher for the lung, 15× for brain, 10× for heart, 38× for liver, 20× for spleen, and 9× higher for the kidneys (Johnson, Lewis & Angier, 2007). Of note, all fluoxetine doses of at least 20 mg/day exceeds the EC50 in the brain; however, only a fluoxetine dose of 60 mg/day exceeds the EC90 in the brain resulting in 90% inhibition of the SARS-CoV-2 titers and likely most beneficial for long COVID-19 and likely treatment-resistant depression associated with inflammatory biological markers.

Fluoxetine adverse drug reactions and drug-drug-interactions

According to the United States Food and Drug Administration Adverse Events Reporting System (FAERS) during the window period of 1982 to June 30, 2022, fluoxetine was reported to have a total of 85,407 cases, 67,924 serious cases, and 11,046 end of life cases (https://www.fda.gov/drugs/questions-and-answers-fdas-adverse-event-reporting-system-faers/fda-adverse-event-reporting-system-faers-public-dashboard). Females (n=49,467) represented 58% of the adverse drug reaction (ADR) cases, males (n=23,240) represented 27% of the ADRs cases, and 15% of the ADR cases did not specify (n=12,700) a gender. The most common adverse drug event reported for fluoxetine is Drug Interaction and amounts to 4,347 cases (5.1% of total). Given this information, drug interactions associated with fluoxetine are due to inhibition of the cytochrome P450 (CYP) system. Specifically, CYP2C19 and CYP2D6 may have interactions such as in patients taking tamoxifen for breast cancer by inhibiting conversion to the active endoxifen metabolite via CYP2D6 or in cases of clopidogrel in cardiology by inhibiting the conversion of clopidogrel to the active 2-oxo-clopidogrel metabolite (Spina, Trifirò & Caraci, 2012; Eugene, 2019).

Overall, from a drug-safety perspective, prior to administering fluoxetine, a careful review of all current medications and clinical status by a physician clinical pharmacologist to avoid drug interactions due to fluoxetine’s ability to strongly inhibit CYP2C19 and CYP2D6 (Hefner, 2018). Compounds that are sensitive and moderate CYP2C19 substrates (e.g. omeprazole, diazepam, lansoprazole, rabeprazole, voriconazole) and CYP2D6 substrates (e.g. dextromethorphan, eliglustat, nebivolol, tolterodine, encainide, metoprolol, propranolol, tramadol) will have an increased total area under the concentration-time curve of ≥ 5-fold drug exposure when treated with fluoxetine (https://www.fda.gov/drugs/drug-interactions-labeling/drug-development-and-drug-interactions-table-substrates-inhibitors-and-inducers). Lastly, patients who have a pharmacogenomic profile of being a CYP2D6 Ultra-rapid Metabolizers may have sub-therapeutic fluoxetine concentrations; while, CYP2D6 Poor Metabolizer or Intermediate Metabolizers may have supratherapeutic concentrations and should be monitored for potential fluoxetine side-effects; but may also have a higher rate of achieving the target trough EC90 concentration at a 20 mg/day fluoxetine dose relative to CYP2D6 Normal (Extensive) Metabolizers.

Long COVID-19 and fluoxetine

As patients recover from the acute COVID-19 symptoms, long-term sequelae are being documented and in one of the long COVID-19 study in young patients reported that 92% were found to have ongoing cardiorespiratory symptoms with organ dysfunction and impairment in the lungs (33%), heart (32%), kidneys (12%) (Dennis et al., 2020). In another long COVID-19 syndrome study, 96% of the patients were female and experienced statistically significant exercise intolerance, dyspnea, and chest pain when compared to those not diagnosed with COVID-19 (Walsh-Messinger et al., 2020). Moreover, Walsh-Messinger et al. found patients with long COVID-19 syndrome had higher ratings of depression subscale markers of altered sleep and thinking, but depression severity was not significantly different with patients not diagnosed with COVID-19 (Walsh-Messinger et al., 2020). As shown in Table 1, a fluoxetine dose of 60 mg/day achieves the EC90 concentration in the brain and would likely benefit patients with neuropsychiatric symptoms of COVID-19 at pharmacokinetic steady-state.

Opportunities for repurposing compounds in COVID-19

In addition to fluoxetine, other psychotropics (fluvoxamine, hydroxyzine, and trazodone) and one antihypertensive (amlodipine) are reported to be associated with reducing risk of death in patients with COVID-19 (Zhang et al., 2020; Darquennes et al., 2021; Sánchez-Rico et al., 2021; Clelland et al., 2022; Reis et al., 2022). Further, Reznikov et al. showed three antihistamines (azelastine, diphenhydramine, and hydroxyzine) have direct inhibitory activity against SARS-CoV-2 in vitro; while clinically, Reznikov et al. showed azelastine, cetirizine, diphenhydramine, hydroxyzine, and loratadine was significantly associated with a lower incidence of testing positive for COVID-19 in patients 61-years-old and older, but only cetirizine had the same association in patients 31-years-old and older (Reznikov et al., 2020).

Further, of note, the over-the-counter compound famotidine (PEPCID^®), a histamine type 2 H₂ receptor antagonists, which has been shown to inhibit the human immunodeficiency virus (HIV) (Bourinbaiar & Fruhstorfer, 1996), has also been shown to provide clinical benefit in patients with COVID-19 – particularly COVID-19 pneumonia – and future retrospective studies should evaluate potential synergistic benefit as famotidine is available in both oral and intravenous formulations (Freedberg et al., 2020; Janowitz et al., 2020; Mather, Seip & McKay, 2020). Lastly, doxycycline has also been shown to provide benefit in patients with COVID-19 pneumonia even in high-risk elderly nursing home patients and achieves therapeutic levels in lungs as found in vitro (Alam et al., 2020; Gendrot et al., 2020; Stricker & Fesler, 2020; Yates et al., 2020; Alexpandi et al., 2022).

Previously known antimicrobial properties of fluoxetine

Antimicrobial properties of fluoxetine are well reported in the biomedical literature. Carpinteiro et al. reported that fluoxetine inhibits acid sphingomyelinase preventing infection of both cultured cells and human nasal epithelial cells in SARS-CoV-2, as well as in vesicular stomatitis virus pseudoviral particles presenting the SARS-CoV-2 spike protein (Carpinteiro et al., 2020b). A study by Zuo et al. showed fluoxetine resulted in potent inhibition of the coxsackievirus by reducing both synthesis of viral RNA and protein (EC50 of 2.3 μM or 795.34 ng/ml) exhibiting peak antiviral properties at 6.25 μM (2161.25 ng/ml) (Zuo et al., 2012). Bauer et al. showed, in a broad-spectrum manner, fluoxetine inhibited enterovirus (picornaviridae family) replication with the S-fluoxetine enantiomer exhibiting a 5-fold lower EC50 than the racemic mixture of R- and S-fluoxetine (Bauer et al., 2019). Further, Bauer et al. found the following fluoxetine EC50 values for the following pathogens: coxsackievirus B3 (racemate-EC50=2.02 μM or 698.5 ng/ml, S-fluoxetine-EC50=0.42 μM or 145.2 ng/ml), enterovirus EV-D68 (racemate-EC50=1.85 μM or 639.7 ng/ml, S-fluoxetine-EC50=0.67 μM or 231.7 ng/ml), and S-fluoxetine values alone for rhinovirus HRV-A2 (EC50=7.95 μM or 2749.1 mg/ml) and HRV-B14 (EC50=6.34 μM or 2192.4 ng/ml) (Bauer et al., 2019). Notably, Zimniak et al. found that individual stereoisomers, R-fluoxetine and S-fluoxetine, inhibited the SARS-CoV-2 viral load; however, in contrast, fluoxetine could not inhibit gene expression of the herpes simplex-1 virus, human herpes virus-8, rabies virus, nor the respiratory syncytial virus (Zimniak et al., 2020, 2021). As shown in Table 1, standard fluoxetine doses are capable of achieving the aforementioned EC50s for all of the aforementioned microbes and fluoxetine may be combined with antipsychotics (e.g. olanzapine) to treat bipolar depression, treatment-resistant depression, schizophrenia in general and these antipsychotics may have notable associations with sedation and somnolence (Eugene et al., 2021).

Direct clinical translation of this current pharmacokinetic study supports the findings from a retrospective multicenter observational study by Hoertel et al., who found a median fluoxetine dose of 20 mg/day resulted in a significantly lower risk of intubation and death in a population composed of 63% women and 37% men (Hoertel et al., 2020, 2021b). Comparing the Hoertel et al. and Zimniak et al. publications, Hoertel et al. found that in addition to fluoxetine, venlafaxine (median dose of 75 mg/day) and escitalopram (median dose of 10 mg) were also associated with a lower risk of intubation and death; however, Zimniak et al. showed that neither escitalopram nor paroxetine inhibited SARS-CoV-2 in vitro (Hoertel et al., 2020, 2021b; Zimniak et al., 2020, 2021). There have been authors who have suggested that the effect of fluoxetine in treating depression is due to a placebo effect; however, with all due respect, it suggests of the lack of knowledge of the etiologies of depression and clinical pharmacology of fluoxetine beyond serotonin transporters to inhibiting inflammatory cytokines and more that will be idientied (Kirsch & Sapirstein, 2004).

Fluoxetine dose for COVID-19 clinical trials

Optimal dose selection for clinical trials are of importance to ensure the maximum number of patients achieve the effective concentration resulting in 90% inhibition of the SARS-CoV-2 pathogen similarly to that reported in Calu-3 human lung cells. Based on the study findings showing the percentage of the population achieving the EC50 and EC90 inhibitory concentrations in the lungs, a dose of 40 mg/day would be recommended to achieve therapeutic potential inhibiting SARS-CoV-2 titers. Overall, given the abundance of clinical trial data and published clinical studies investigating various dosing approaches (e.g. loading dose) for fluoxetine, this author defers all final dosing recommendations for fluoxetine to the FDA approved package-insert for adults and for the pediatrics population (Eli Lilly / Dista Products Company, 2021). For an application in pediatrics, fluoxetine doses in children are recommended to start at 10 mg/day and then titrate to 20 mg/day, whereas in adults with depression, a 20 mg/day is recommended as a starting dose (Eli Lilly/Dista Products Company, 2021). Lastly, given pharmacogenomic (drug-gene), DDIs, and drug-drug-gene interactions that are likely in patients taking multiple medications, therapeutic drug monitoring of fluoxetine and other co-administered medications are recommended.

Study limitations

A limitation of this study is in respect to how fast the trough concentrations are achieved which assumes that fluoxetine immediately distributes to lung tissue at the estimated 60-fold concentration compared to plasma levels. There is some evidence that suggests fluoxetine may accumulate into tissues over time as is reported by Erb et al. who found that certain antidepressants accumulate in lipid rafts in vitro over several days of exposure (Erb, Schappi & Rasenick, 2016). In reference to human brain fluoxetine concentrations, Karson et al. used spectroscopy and determined that fluoxetine and the norfluoxetine metabolite are not detectable in human brain at about 1 week, but actually detectable at about 2- to 3-weeks of treatment the concentrations were able to be assayed indicating that fluoxetine accumulates in brain over time (Karson et al., 1993). Therefore, with this evidence, it is likely that the fluoxetine lung concentrations would not be evident as quickly as reported in the simulations and the data should be interpreted assuming the target EC50 and EC90 concentrations inhibiting SARS-CoV-2 at fluoxetine’s pharmacokinetic stead-state concentrations – that is between 10-days and 21-days of treatment – and patients should in-turn be treated for 21-days with fluoxetine in order to have a valid clinical trial with a dose of at least 20 mg/day and ideally at 40 mg/day.

Another limitation of this study is that the results are purely quantitative using pharmacometrics simulations based on differential equations and not a randomized controlled clinical trial. Despite this limitation, it is important for the reader to note that the methodology used in this study is standard practice for all drugs undergoing Phase-1 Clinical Trials in drug development by clinical pharmacologists for later regulatory approval for human use. Moreover, the pharmacometric parameters used were estimated from an NDA submitted to the United States Food and Drug Administration with parameter estimates actually calculated by the FDA clinical pharmacologists as reported in the Clinical Pharmacology and Biopharmaceutics Review after combining three human study datasets provided by the study sponsor (https://www.accessdata.fda.gov/drugsatfda_docs/nda/2003/18936S064_Prozac%20Pulvules_biopharmr.pdf).

Underlying data

Open Science Framework: All pharmacokinetic data presented in this study are available for download at the following link: https://doi.org/10.17605/OSF.IO/RVYPZ.

This project contains the following underlying data:

Software availability

Open Science Framework: The R programming language pharmacokinetic script that produces the general results of this study are available for download at: https://doi.org/10.17605/OSF.IO/RVYPZ (Eugene, 2021).

This project contains the following software:

Data are available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication).