How Can Pharmacology Help Us Overcome the Challenges of Drug Repositioning as Antivirals to Treat Emerging Pathogens? The Example of Covid‐19

ABSTRACT

The Covid‐19 pandemic highlighted the urgent need for effective therapies against emerging pathogens. Drug repurposing, defined as the use of existing medications for new therapeutic purposes, was extensively pursued for SARS‐CoV‐2 but has not yielded successful treatments. This narrative review critically examines the pharmacological and methodological factors that contributed to these unsuccessful outcomes, paying particular attention to tests of azithromycin and hydroxychloroquine. There are many reasons the promise of repurposed drugs was not realized. Many repurposed compounds displayed promising in vitro antiviral activity that did not translate into clinical efficacy. Major pharmacokinetic (PK) limitations, for example, poor oral bioavailability, low concentrations in pulmonary tissue, and extensive plasma protein binding, prevented these drugs from reaching therapeutic levels in humans. Preclinical research often relied on non‐human cell lines and animal models that inadequately reflected human physiology, leading to misleading experimental outcomes. Clinical trials were often undermined by methodological limitations, including endpoints with uncertain clinical significance, suboptimal comparators, and insufficient attention paid to key PK and pharmacodynamic (PD) parameters such as half maximal effective concentration (EC50) values. This narrative review emphasizes the importance of integrating comprehensive PK/PD assessments, relevant experimental models, and rigorous trial design to strengthen drug development during future health crises. The relative success of antivirals including molnupiravir, nirmatrelvir, and remdesivir, which were either novel or previously unapproved compounds, suggests the value of designing and developing targeted antivirals. We must coordinate global research, develop pharmacologically sound strategies, and use evidence‐based decision‐making to effectively prepare for future pandemics and quickly produce effective treatments.

Pharmacological pitfalls and challenges in drug repurposing for emerging pathogens: lessons from COVID‐19.

1. Introduction

The emergence of severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) in 2019 precipitated a global pandemic by 2020. SARS‐CoV‐2 caused widespread upper and lower respiratory tract infections which were called coronavirus disease 2019 (Covid‐19). In the most severe cases, Covid‐19 may cause multi‐organ dysfunction and death [1]. From the start of the pandemic, it was evident that effective curative and preventive treatments were urgently needed to address this escalating crisis. Many existing drugs were quickly proposed as potentially effective treatments against Covid‐19, often based on in vitro data or deductive hypotheses. Urgency spurred an unprecedented wave of clinical studies and trials to meet expectations and address the concerns of the public, healthcare professionals, and policy makers [2, 3]. Different therapeutic classes were clinically evaluated, including antibiotics [4], antivirals [5, 6, 7, 8, 9, 10, 11, 12], antiparasitics [13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29] and calcium channel blockers [13, 30]; these studies mainly produced inconclusive results [31].

Some randomized controlled trials (RCTs) faced design limitations, including small sample sizes [32, 33] or questionable comparators [8, 34, 35]. Non‐randomized and non‐comparative trials were subject to confounding biases [16, 34]. Open‐label trials had to be cautiously interpreted [5, 8, 11, 12, 24, 34, 35, 36, 37, 38, 39]. Some studies lacked clinically relevant endpoints, for example, mortality or hospitalization rates; instead, they focused on SARS‐CoV‐2 seropositivity or on the time it took to achieve a negative nasopharyngeal swab [8, 35, 40]. Many concurrent studies overlapped in scope or tested the same drugs; duplication created recruitment challenges and caused funding shortages. New unfavorable data (e.g., early results that showed a drug was ineffective or unsafe) sometimes rendered ongoing trials unviable [41]. Some trials were withdrawn without explanation, creating gaps in our understanding of drug efficacy [2]. Many published studies were later retracted due to methodological or statistical errors, data fabrication, or plagiarism [42, 43].

Even ahead of clinical evaluation, certain repurposing failures could have been anticipated. We hypothesized that more comprehensive evaluation of preclinical and pharmacological data early in the pandemic could have helped researchers quickly eliminate less promising candidate drugs, leaving them free to focus their efforts on drug interventions with greater potential. This narrative review uses the SARS‐CoV‐2 pandemic to explore the pharmacological pitfalls and challenges of repurposing drugs against emerging pathogens. Note that we do not classify as “repurposed” any compounds already under development as antivirals for other pathogens pre‐pandemic: “repurposed” is defined as established drugs investigated for new therapeutic uses.

2. Materials and Methods

This narrative review aims to provide a critical overview of pharmacological challenges associated with drug repurposing against emerging pathogens, using the example of the SARS‐CoV‐2 pandemic. The literature was identified through a broad search on PubMed databases and clinicaltrials.gov, spanning July 2003 to November 2025. Search terms included “SARS,” “SARS‐CoV‐2,” “Covid‐19,” “antiviral,” “drug,” “treatment,” and “therapy.” We considered in silico studies, in vitro studies, animal studies, retrospective clinical studies, prospective clinical studies, including RCTs, reviews, and guidelines in English or French. Additional sources were identified through recursive searches, manual searches, and cross‐referencing of relevant articles. This narrative review focuses on repurposed drugs with alleged antiviral activity, while drugs primarily investigated for anti‐inflammatory or immunomodulatory actions fall outside the scope of this review.

3. Results

3.1. Overreliance on Limited Preclinical Models

Several clinical studies were initiated based on preclinical data that were too limited; some inherent limitations of the preclinical models were insufficiently considered.

3.1.1. Variable Suitability of Cell Lines

The relevance of isolated cell lines for studying SARS‐CoV‐2 may vary. Not all constitute equally suitable models. For example, azithromycin, a macrolide antibiotic used to treat a wide range of bacterial infections and extensively studied to treat Covid‐19, has demonstrated antiviral activity, primarily in vitro, against various pathogens, especially certain rhinoviruses and flaviviruses [44]. In 2020, Touret et al. [45] found azithromycin had promising effects in African green monkey kidney cells (Vero E6 cells), with a half maximal effective concentration (EC50) of 2.1 μM. Andreani et al. [46] tested the antiviral activity of azithromycin (5 and 10 μM) in combination with hydroxychloroquine (1, 2, or 5 μM) against SARS‐CoV‐2 in Vero E6 cells, and found that their combined effect was strongly antiviral, suggesting a synergistic relationship that could offer benefits, especially in the early stages of infection. The azithromycin EC50 that inhibited SARS‐CoV‐2 in Vero E6 cells is likely achievable in lung tissues, where azithromycin diffuses well. In adult patients who received a 500 mg oral dose of azithromycin daily for 3 days, peak lung tissue concentration of azithromycin on day 5 was 12 μM, which was higher than the azithromycin EC50 against SARS‐CoV‐2 in Vero E6 cells, and the area under the curve (AUC) value was high (approximately 1300 hx μg g⁻¹) [45, 47].

All these pharmacokinetic (PK)/pharmacodynamic (PD) parameters were studied in Vero E6 cells. Many antiviral assays have used Vero E6 cells as the testing model [48] because these cells express high levels of angiotensin‐converting enzyme 2 (ACE2), the receptor that allows SARS‐CoV‐2 to enter via clathrin‐mediated endocytosis [49]. But Vero E6 cells do not express the transmembrane protease serine 2 (TMPRSS2), which primes the Spike protein of the SARS‐CoV‐2 to enter the cell through direct fusion with the plasma membrane, independently of the endosomal pathway. In most of the SARS‐CoV‐2 variants, except the first Omicron [50], TMPRSS2 was the primary pathway by which the virus entered human cells [51, 52, 53]. Vero E6 cells are an unreliable model for studying SARS‐CoV‐2 in human cells because they do not accurately replicate expression of key receptors and other human cell characteristics [51]. Mautner et al. [54] recommend using Vero E6 cells only for isolating and propagating the virus, and not for drug testing. When tested in human‐derived cell lines such as Caco‐2 cells, azithromycin showed only moderate activity at concentrations of 5 and 10 μM [45]. EC50 evaluated on Vero E6 cells should have been reassessed in more relevant human cell models, for example, Calu‐3 or Caco‐2 cells, which naturally express ACE2 and TMPRSS‐2.

3.1.2. Use of Isolated Cell Lines Without Validation in Complex Models, for Example, Human Airway Epithelium (HAE) or Animal Models

Isolated cell lines have more inherent limitations than complex, physiologically realistic models, for example, HAE cultures. HAE cultured at the air‐liquid interface are a more accurate pathophysiological model than isolated cells because HAE cultures mimic the diverse cell types in target tissues [55]. Cochin et al. [56] tested different concentrations of azithromycin (from 0 to 80 μM) in an HAE model reconstituted with primary human bronchial cells (an ex vivo model that represented target cells of infection and viral replication in humans) and did not report antiviral activity of azithromycin against SARS‐CoV‐2. Cochin et al. [56], who evaluated azithromycin in a hamster model (90.6 ± 6.9 mg/kg/day), reported no antiviral activity. Andreani et al.'s study (which appears to be an outlier) evaluated azithromycin in combination with hydroxychloroquine [46], with a notable limitation: the EC50 they observed in vitro in VeroE6 cells was not validated in primary cell models, such as HAE models.

Other agents that received significant attention during the Covid‐19 pandemic include chloroquine and hydroxychloroquine. These drugs are quinine analogs used to treat malaria and some autoimmune diseases [57]. Both were heavily promoted as potential treatments [15, 16]. Preclinical data on chloroquine's and hydroxychloroquine's efficacy against SARS‐CoV‐2 were limited. Wang et al. [58] found that chloroquine had high antiviral activity against SARS‐CoV‐2 in Vero E6 cells (an EC50 of 1.13 μM and an EC90 of 6.90 μM), but chloroquine and hydroxychloroquine failed to demonstrate antiviral efficacy against SARS‐CoV‐2 in HAE or in animal models. Early in the pandemic, Maisonnasse et al. [59] showed that hydroxychloroquine did not cause antiviral activity in HAE or primate models (macaque monkeys). Similarly, Cochin et al.'s 2021 study [56] showed no antiviral activity of hydroxychloroquine alone or combined with azithromycin against SARS‐CoV‐2 in HAE or Syrian hamster models.

The mechanism of action and the site of action of chloroquine in Covid‐19 remains unclear [60]. Chloroquine acts on the endosomal entry pathway, so it may not work in vivo because the virus enters lung cells by directly fusing with the membrane when TMPRSS2 is present, avoiding the endosomal pathway entirely [61, 62]. Chloroquine's anti‐inflammatory and immunomodulatory properties also raise concerns; in animal models, under some conditions chloroquine can increase the replication of certain viruses [63]. Despite lack of evidence of efficacy in HAE and animal models, many clinical studies and trials continued to evaluate hydroxychloroquine [15].

Together, these examples show that first, rapid antiviral screening results should be confirmed in primary human cells such as HAE cultures or in animal models. Results from these more advanced models should then be used carefully in subsequent studies.

3.1.3. Phospholipidosis as a Source of Misleading Results

Phospholipidosis can create false antiviral signals in vitro, so results may not translate in vivo. Phospholipidosis is defined by intracellular accumulation of phospholipids and lamellar bodies (lysosomal inclusion bodies). Many drugs can trigger this effect, particularly cationic amphiphilic drugs such as azithromycin, which interfere with lipid metabolism [64, 65]. Because viral replication and RNA synthesis rely on membrane compartments composed of phospholipids, phospholipidosis correlates strongly with antiviral activity in cell‐culture experiments [66]. But this phenomenon, which does not occur in vivo, is a confounding factor in in vitro antiviral drug research. Phospholipidosis could have induced azithromycin antiviral activity in vitro, which may explain the lack of antiviral activity observed in animal and clinical studies [4, 14, 15, 16, 36].

Phospholipidosis may also explain the apparent antiviral activity reported for chloroquine and hydroxychloroquine, both of which are cationic amphiphilic drugs [67]. Liu et al. [68] reported antiviral activity for both chloroquine and hydroxychloroquine in VeroE6 cells, and observed “dilated cytoplasmic vesicles” in treated cells—a hallmark of phospholipidosis. Their finding could prompt us to ask if the apparent antiviral effects of these treatments were attributable to pharmacological activity or if they were artifacts of phospholipidosis.

The phospholipidosis mechanism should be explicitly tested in early in vitro tests for antiviral assays. If a drug's apparent antiviral effect in vitro is associated with phospholipidosis and does not rely on specific on‐target effects, the effect is unlikely to be reproduced in the clinic [69]. Studies should thus either exclude such drugs or confirm that their results are not caused by phospholipidosis.

3.1.4. Insufficient Data on Appropriate Animal Models

Drugs that show antiviral activity in vitro, in isolated cell lines or in primary human cells, should then be tested in vivo, in animal models. Murine models have been used for antiviral research against SARS‐CoV‐2, but rodent cells express an ACE2 that cannot bind the SARS‐CoV‐2 spike protein. Genetically engineered mice can express human ACE2, using tissue‐specific promoters or genetic substitutions. In mice, human ACE2 can be mildly to severely overexpressed or only ectopically expressed, resulting in variable lethality after SARS‐CoV‐2 infection and sometimes a change in viral tropism [70]. Syrian hamsters may be a more accurate pathophysiological model of human SARS‐CoV‐2 infection, as they show similar clinical, radiological and histopathological signs, and a similar demographic pattern (older age as a risk factor). Yet, the most appropriate animal models remain non‐human primate models (rhesus macaques, African green monkeys) that provide an accurate pathophysiological model of human SARS‐CoV‐2 infection. These primate models show similar clinical signs, cellular and humoral immune responses, and demographic patterns [70] but their cost and ethical considerations limit their availability [71].

For example, the antiviral activity of favipiravir against SARS‐CoV‐2 was evaluated in both Syrian hamster and macaque models, yielding conflicting results. In the Syrian hamster model, favipiravir showed dose‐dependent antiviral activity; a dose of 700 to 1400 mg/kg/day reduced infectious titers [72]. When favipiravir was initiated on the day of infection or a day before, clinical improvement was evident, but doses far exceeded the approved posology for humans (1600 mg orally twice daily for 1 day followed by 600 mg twice daily for 4 days) [73]. In contrast, a study that used a macaque model found no antiviral or clinical benefits of prophylactic favipiravir therapy in SARS‐CoV‐2 infected animals [74]; four treated animals exhibited rapid clinical deterioration, suggesting favipiravir may worsen disease in SARS‐CoV‐2 infected animals [74]. Differences in drug pharmacokinetics between the two animal models (hamsters quickly metabolize favipiravir so they can tolerate higher doses) may explain these conflicting results [74].

Drugs should first be screened in rodents, as they are the most readily available animal models; then they should be tested in other non‐rodent models, particularly primate models.

3.2. Lack of Consideration of PK/PD Data

3.2.1. Gap Between EC50 and Clinically Achievable Blood Concentrations

For many drugs investigated in RCTs, reported EC50 against SARS‐CoV‐2 in preclinical models was not sufficiently considered in relation to clinically achievable plasma concentrations. Several drugs provide illustrative examples.

First, lopinavir is ineffective for treating Covid‐19 because patients cannot achieve plasma concentrations high enough to inhibit SARS‐CoV‐2. Lopinavir inhibits HIV protease when combined with ritonavir, which boosts its effect by inhibiting cytochrome P450 (CYP) 3A4 and improving lopinavir's PKs [75], but this mechanism does not translate to effective antiviral activity against SARS‐CoV‐2. In vitro studies of the effect of lopinavir on SARS‐CoV‐2 in Vero E6 cell models showed antiviral activity, which the in silico molecular dynamic simulations anticipated [76], but the reported EC50 of 26 μM was relatively high [77]. In patients with HIV who received 400/100 mg lopinavir/ritonavir twice daily, mean lopinavir peak plasma concentrations (Cmax) ranged from 15 μM to 19 μM; steady‐state trough plasma concentrations (Cmin) prior to morning dose were between 11 μM and 13 μM [78, 79]. But the EC50 of lopinavir for inhibiting HIV, which is about 100 nM [80], is much lower than the concentration needed to inhibit SARS‐CoV‐2 [81]. In three small clinical studies of Covid‐19 patients (33 patients) treated with lopinavir/ritonavir 400/100 mg bid, plasma trough concentration ranged from 21 to 48 μM, with high variability [82, 83, 84]. In one of these studies, plasma levels were less than half of the EC50 of lopinavir for SARS‐CoV‐2 in about half of patients [84]. This study revealed that concentrations high enough to effectively inhibit SARS‐CoV‐2 cannot realistically be achieved in vivo, which may explain why lopinavir did not reduce mortality or clinical outcomes in two notable RCTs [5, 6].

Second, favipiravir, a purine base analog, has shown no efficacy in treating early or mild Covid‐19 despite its activity against RNA viruses including influenza [73]. Although favipiravir demonstrated antiviral activity against SARS‐CoV‐2 in Vero E6 cells by lethal mutagenesis [85] and Driouich et al. [72] reported favipiravir was active against SARS‐CoV‐2 in Vero E6 cells (EC50 was 204 μM), effective levels of the drug may be difficult to achieve in humans. In healthy volunteers administered a single favipiravir oral dose of 1600 mg, Cmax was about 380 mM; the area under the curve (AUC) was 2.53 mmol.h/L [86]. However, in 66 patients with Ebola virus disease treated with much higher doses (6000 mg at day 0, followed by 2400 mg from days 1 to 9), Cmin dropped to 293 μM by day 2 and to 165 μM by day 4, failing to meet predefined target exposure levels [87]. Given the EC50 for SARS‐CoV‐2, these low Cmin indicated an unfavorable inhibitory quotient (Cmin/EC50) [88]. The gap between the EC50 required to effectively inhibit SARS‐CoV‐2 and the Cmin and Cmax achieved in vivo could have predicted the lack of efficacy of favipiravir on viral clearance and clinical outcomes in patients with Covid‐19 [7, 9].

Third, tenofovir/emtricitabine yielded mixed results against SARS‐CoV‐2 [11, 12, 39] because efficient blood concentrations could not be reached. Tenofovir/emtricitabine is a combination of antiretroviral drugs commonly used against HIV at a dose of 300 mg of tenofovir disoproxil fumarate and 200 mg of emtricitabine once daily. Tenofovir, a nucleotide analog, and emtricitabine, a nucleoside analog, both inhibit HIV‐1 reverse transcriptase. Tenofovir (nucleoside monophosphate) is metabolized intracellularly into its active anabolite, tenofovir diphosphate, which exerts its antiviral effects [89]. Preclinical studies explored tenofovir's antiviral activity against SARS‐CoV‐2. Molecular docking studies showed that tenofovir diphosphate (tenofovir's active triphosphate form) can bind tightly to the RNA dependent RNA polymerase (RdRp) of SARS‐CoV‐2 [90]. However, this active tenofovir triphosphate form is only obtained in vivo after enzymatic transformation: plasma‐based for tenofovir disoproxil fumarate and intracellular enzymatic transformation for tenofovir alafenamide [91]. In vitro, tenofovir disoproxil fumarate demonstrated antiviral activity against SARS‐CoV‐2 for concentrations of between 3 and 90 μM [92]. But in healthy volunteers who received 300 mg oral tenofovir disoproxil fumarate daily, Cmax was only about 1 μM, far below the effective range in vitro [93]. The gap between the tenofovir EC50 and clinically achievable blood concentrations should have raised questions about its use in clinical practice.

Fourth, in preclinical studies ribavirin demonstrated antiviral activity in vitro against SARS‐CoV‐2 but results suggested it may be difficult to reach effective concentrations in vivo. Molecular docking studies conducted when SARS‐CoV‐2 emerged found that ribavirin triphosphate, the active intracellular form of the drug, could bind the SARS‐CoV‐2 RdRp [90]. Wang et al. [58] reported that ribavirin acted against the SARS‐CoV‐2 virus in Vero E6 cells; EC50 was 110 μM and the selectivity index was > 3.7. Similarly, Unal et al. [94] found antiviral effects in Vero E6 cells and Caco‐2 cells at high concentrations (200 μM and 750 μM). But these effective concentrations far exceeded the ribavirin plasma Cmax of about 11 μM observed in patients treated for hepatitis C [95, 96]. Again, in vitro studies did not translate into clinically relevant evidence for SARS‐CoV‐2 [8, 35, 96].

Fifth, ivermectin was unlikely to be a viable treatment for patients with Covid‐19 because the drug could not clinically achieve the EC50 needed to inhibit SARS‐CoV‐2 at doses safe for humans. Ivermectin is a macrocyclical lactone antiparasitic drug used against a variety of parasites. In vitro studies that administered a 5 μM ivermectin concentration to Vero/hSLAM cells found significant reduction (~5000‐fold) in viral RNA [97]. EC50 was between 2.2 and 2.5 μM at 48 h. But these concentrations far exceed those that can be reached in humans, as even a dose of 600 μg/kg (3× the usual dose) brings the Cmax only up to about 130 nM [98], well below the EC50; Cmin would be even lower. In an evaluation of in vitro viral efficacy, EC50 values should always be related to clinically achievable drug concentrations. Both Cmax and Cmin should be assessed because the inhibitory quotient (Cmin/EC50) appears to predict antiviral efficacy [88, 99].

The gap between achievable plasma concentrations and the EC50 rendered questionable any consideration of these drugs as Covid‐19 treatments.

3.2.2. Failure to Consider Protein Binding and Active Metabolites

One cannot directly compare the EC50 obtained in cell culture, where the drug comes into direct contact with infected cells, to plasma concentration in vivo. After the drug moves from the plasma to the target organ, tissue concentrations drop well below plasma levels. Most in vitro drug screening studies failed to account for protein binding and drug metabolism, which may have greatly reduced the amount of diffusible active drug. For example, less than 2% of total lopinavir is diffusible [100], so EC50 should be adjusted for protein binding. PK studies indicated that unbound plasma concentrations of lopinavir were 22‐fold to 84‐fold lower than in vitro EC50 for SARS‐CoV‐2 [84, 100]. As another example, more than 50% of favipiravir binds to plasma proteins in humans [101] and favipiravir is a prodrug that requires intracellular phosphorylation to form its active metabolite [102]. So, if the Cmin of favipiravir is low, as already mentioned, the concentration of the free and active form of favipiravir is likely even lower. Thus, EC50 values should be interpreted in light of protein binding and drug metabolism, since these determine the availability of active drug at the target site.

3.2.3. Low Oral Bioavailability

Some drugs were tested against SARS‐CoV‐2 by the oral route, although their oral bioavailability was known to be low. Among these was niclosamide, a salicylanilide antiparasitic drug widely used to treat tapeworm infection. Niclosamide attracted attention for its potential antiviral and anticancer properties [103], but failed to show efficacy in SARS‐CoV‐2 clearance, symptom duration, or mortality in patients with Covid‐19 [24, 25]. Niclosamide's poor water solubility, low oral bioavailability [104], and strong plasma protein binding (> 99.8%) [105] severely limit its systemic exposure. Pharmacokinetic studies showed that oral doses of 2 g produced highly variable Cmax in healthy volunteers, ranging from 0.8 to 18.4 μM [106]. A phase 1 study in patients with metastatic prostate cancer found even lower Cmax values (110–550 nM) when niclosamide was given orally at doses of 0.5–1.5 g twice daily alongside enzalutamide [107]. A phase 1b trial in castration‐resistant prostate cancer patients receiving reformulated niclosamide (1.2 g thrice daily) reported trough concentrations of 0.3–0.65 μM [108], well below the 1.25–2.5 μM required for effectiveness in HAE models [109]. Though niclosamide has very low oral bioavailability [110], RCTs that tested oral niclosamide formulations in patients with Covid‐19 did not measure niclosamide blood concentrations [24, 25], and the drug did not prove to reduce mortality in patients with Covid‐19 [24].

To address oral bioavailability issues, researchers developed inhalable formulations of niclosamide, which also failed to show efficacy in reducing risk of infection in a RCT [111]. The phase 1 trial reported a mean Cmax of about 1.3 μM after a single dose [112]. This Cmax was also likely insufficient since niclosamide concentrations must be higher than 1.25–2.5 μM to significantly reduce SARS‐CoV‐2 infectious titers [109]. Even via alternative delivery methods, it is unlikely effective concentrations can be achieved because of the drug's high EC50 values, low systemic absorption, and strong plasma protein binding. Given these limitations, even intranasal and inhalable formulations may not overcome niclosamide's pharmacokinetic constraints.

3.2.4. Low Pulmonary Diffusion

Because SARS‐CoV‐2 mainly infects the lungs, antivirals need to penetrate the lungs sufficiently to be effective [113, 114]. It is likely that insufficient concentration in the lungs accounts for nitazoxanide's failure to reduce viral load, disease progression, or death in RCTs [115]. Nitazoxanide, a broad‐spectrum thiazolide antiparasitic, demonstrated antiviral activity against SARS‐CoV‐2 in vitro; EC50 was 1.29 μM and the selectivity index was high (> 515) [116]. Driouich et al. [117], confirmed the antiviral activity of nitazoxanide in Vero E6 (EC50 = 3.2 μM) and Caco‐2 cells (EC50 = 0.6 μM), with a selectivity index of > 13. Antiviral activity was corroborated in an HAE model [117]. But animal models yielded disappointing results: oral and intranasal nitazoxanide failed to demonstrate antiviral efficacy, as measured by infectious titers, lung and plasma viral RNA yields, and clinical outcomes [117]. Pharmacokinetic data suggested that tizoxanide, nitazoxanide's active metabolite, achieved insufficient pulmonary concentrations to exceed the in vitro EC50 after 3 days of multiple‐dose treatment in hamsters [117]. The drug is moderately lipophilic and has high plasma protein binding (99%), which may limit pulmonary diffusion [117, 118].

Pharmacokinetics make it difficult to determine the antiviral potential of chloroquine and hydroxychloroquine. After 1 month of treatment, a daily oral dose of chloroquine will keep plasma concentrations at equilibrium between 0.3 μM and 3.1 μM, which is likely to fall below the reported EC50 of 1.13 μM and EC90 of 6.90 μM [58, 67]. But chloroquine accumulates in tissues, including in the lungs; in rats, after 1–168 h, a single 10 mg/kg intraperitoneal dose raised chloroquine lung tissue concentrations 11–547 times higher than plasma concentrations [119]. Yet few clinical studies have examined lung concentrations of chloroquine or hydroxychloroquine in Covid‐19 patients in clinical settings.

At the start of the pandemic, Ruiz et al. [120] measured hydroxychloroquine levels in plasma and bronchoalveolar lavage fluids of 22 critically ill, intubated Covid‐19 patients who had received different dosages of hydroxychloroquine at the start of the pandemic; median hydroxychloroquine concentration in bronchoalveolar lavage fluids was about 9 μM—above the EC50 reported in Vero E6 cells. Assessing the clinical relevance of these results is difficult because EC50 values were derived from non‐human cells and the relationship between bronchoalveolar lavage fluid concentration and lung tissue levels is unclear [60]. Effective antiviral therapy for Covid 19 requires adequate drug penetration into the lungs. Pulmonary drug distribution should be carefully evaluated, and lung concentrations should be consistently measured in animal models and in patients.

3.2.5. Low Selectivity Index

Some of the drugs under study had a poor selectivity index (50% cytotoxic concentration [CC50]/EC50): their antiviral activity was associated with high cellular toxicity [121, 122]. Diltiazem, a non‐dihydropyridine calcium channel blocker, is such a drug. Diltiazem's antiviral effects are hypothesized to result from stimulation of interferon production, which could limit viral replication and promote clearance [123]. In 2020, Pizzorno et al. [124], showed that diltiazem induces type III interferon‐related gene expression in an HAE model infected with SARS‐CoV‐2. The group then tested diltiazem alone and combined with remdesivir against SARS‐CoV‐2 in Vero E6 cells and in an HAE model; diltiazem monotherapy did not instigate antiviral activity against SARS‐CoV‐2 [124]. Diltiazem EC50 was > 45 μM and the CC50 was 424 μM, so diltiazem had a very low selectivity index (CC50/EC50) of < 9.42 [125]. Combining diltiazem and remdesivir appeared to potentiate remdesivir's antiviral activity, reducing its EC50 by 2‐ to 3‐fold [125]. But diltiazem's poor selectivity index may limit its potential as a monotherapy for Covid‐19 treatment. We know of no clinical trials that evaluated diltiazem as a Covid‐19 treatment, but evidence suggests that such trials would have inconclusive or unfavorable outcomes.

Since selectivity index is critical to evaluating antiviral drug candidates: drugs with a selectivity index < 100 should be discarded [121].

3.2.6. Narrow Therapeutic Index

Because antivirals need to be both safe and efficacious against SARS‐CoV‐2, their therapeutic index matters. But even though preclinical and PK data on chloroquine, hydroxychloroquine, and SARS‐CoV‐2 were inconclusive, these drugs were evaluated in more than 29 clinical trials and observational studies [15]. The narrow therapeutic index of chloroquine and hydroxychloroquine is well‐known. Delivering a single dose three times higher than the usual dose can be fatal, and this should have raised safety concerns at the outset [60, 126]. In 2021, Fiolet et al. [15] conducted a systematic review and meta‐analysis of four clinical trials and 25 observational studies, which included 11,932 patients treated with hydroxychloroquine and 8081 patients treated with combined hydroxychloroquine and azithromycin. Findings indicated that hydroxychloroquine monotherapy did not reduce mortality in hospitalized patients with Covid‐19 (pooled RR = 0.83; 95% CI 0.65–1.06). Moreover, combined hydroxychloroquine and azithromycin was associated with increased mortality (RR = 1.27; 95% CI 1.04–1.54) [15]. Pradelle et al. [127] later estimated that hydroxychloroquine use during the first wave of the Covid‐19 pandemic may have contributed to 16,990 in‐hospital deaths (ranging from 6267 to 19,256) across the six countries under study (Belgium, France, Italy, Spain, Turkey, USA). Some deaths were linked to cardiac adverse events (conduction and rhythm disorders), particularly when hydroxychloroquine was combined with azithromycin, which prolongs QT intervals [14].

When considering drugs for evaluation in clinical trials, researchers should pay close attention to PK/PD parameters, including comparing in vitro EC50 with clinically achievable drug concentrations, and assessing bioavailability, pulmonary diffusion, selectivity index, and therapeutic index.

3.3. Limitations of Clinical Retrospective Observational Studies

Some RCTs were based on clinical retrospective observational studies that had inherent biases and limitations. Several clinical studies evaluated tenofovir/emtricitabine for potential effectiveness against SARS‐CoV‐2. Early clinical retrospective observational studies in HIV patients treated with tenofovir/emtricitabine had returned conflicting results. Berenguer et al. [40] reported that treatment with tenofovir disoproxil fumarate/emtricitabine was associated with SARS‐CoV‐2 seropositivity rates two times lower than rates obtained with tenofovir alafenamide/emtricitabine, possibly because tenofovir plasma concentrations were higher [128]. Even though tenofovir disoproxil fumarate was associated with higher tenofovir plasma concentrations in persons living with HIV, it was also associated with lower intracellular concentrations in peripheral blood mononuclear cells (PBMCs). This lowered its antiviral potency against HIV‐1 [128], but the focus on SARS‐CoV‐2 seropositivity instead of SARS‐CoV‐2 symptoms may have reduced the direct clinical relevance of this study's findings. Del Amo et al. [129] reported that a tenofovir/emtricitabine treatment was associated with lower infection and hospitalization risk among patients living with HIV. Both observational studies may be subject to bias because antiretroviral therapy regimens may be confounded by a patient's clinical characteristics [130] and results of studies of HIV patients with compromised immune systems cannot be generalized to the whole population. Open‐label RCTs that evaluated tenofovir/emtricitabine in patients with confirmed Covid‐19 have yielded inconsistent results on mortality [11, 12, 39].

4. Discussion and Propositions

All clinical trials that aimed to repurpose existing drugs for Covid‐19 were unsuccessful. In this narrative review, we outlined key challenges in repurposing drugs during the pandemic (Table 1). We argue that future therapeutic investigations should be grounded in robust data, even during a global health crisis. A more methodological approach that prioritised drug candidates with sound PK/PD profiles could have increased the efficiency and effectiveness of research.

TABLE 1.

List of main repositioned drugs for the treatment of Covid‐19, main clinical trials and preclinical data.

Repositioned drug	Mechanism of action	Main clinical trials	Major shortcomings according to preclinical and long‐standing clinical data
Antibiotic
Azithromycin	Macrolide	Cavalcanti et al. [14] Gautret et al. [16] Furtado et al. (COALITION II) [36]	No antiviral activity in HAE or animal models [56] Phospholipidosis [65]
Antivirals
Lopinavir	Protease inhibitor	RECOVERY [5] Ader et al. [6]	Gap between EC50 and clinically achievable blood concentration [77, 84] No consideration of protein binding [100]
Favipiravir	Purine base analog	McMahon et al. [9] Bosaeed et al. [7]	Gap between EC50 and clinically achievable blood concentrations [72, 87] No consideration of protein binding [101] No consideration of the active metabolite [102] No antiviral activity in non‐human primate models [72]
Tenofovir/emtricitabine	Nucleoside analogs	Parienti et al. [12] Gaitán‐Duarte et al. [39] Montejano et al. (PANCOVID) [11]	Gap between EC50 and clinically achievable blood concentrations [92, 93] Use of ferret models, no data on other animal models, for example, non‐human primate models [131] Biased clinical retrospective observational studies in persons living with HIV [40, 129]
Ribavirin	Nucleoside analog	Hung et al. [8] Huang et al. [35]	Gap between EC50 and clinically achievable blood concentration [94, 95] No HAE or animal models
Antiparasitics
Chloroquine and hydroxychloroquine	Quinine analogs	Cavalcanti et al. [14] Horby et al. (RECOVERY) [18] Skipper et al. [19]	Phospholipidosis [68] No antiviral activity in human respiratory epithelial models or animal models [56, 59] Narrow therapeutic index [60]
Ivermectin	Macrocyclical lactone	Ravikirti et al. [22] Reis et al. (TOGETHER) [23] Lim et al. (I‐TECH) [21]	Gap between EC50 and clinically achievable blood concentration [97, 98]
Niclosamide	Salicylanilide	Abdulamir et al. [24] Cairns et al. [25]	Low oral bioavailability [110] Gap between EC50 and clinically achievable blood concentration [108, 109, 132] No consideration of protein binding [105] Use of murine models, no non‐human primate models [133]
Nitazoxanide	Thiazolide	Rocco et al. [28] Rocco et al. [27] Rossignol et al. [29]	Low pulmonary diffusion [118] Gap between EC50 and clinically achievable concentration [117] Antiviral activity only in vitro [116, 117] No antiviral activity in animal models [117]
Other drug
Diltiazem	Non‐dihydropyridine calcium channel blocker	Withdrawn [13, 30]	Low selectivity index [125]

Abbreviations: EC50, half maximal effective concentration; HAE, human airway epithelium.

Study designers frequently overlooked PK limitations and instead relied on in vitro findings suggesting antiviral activity at high concentrations, though these were unlikely to be achieved in vivo. Though plasma concentrations play a key role in therapeutic efficacy, they were rarely measured. Drug EC50 values against SARS‐CoV‐2 were often much higher than for the drug's original target, complicating extrapolation. Studies often neglected to consider protein binding, bioavailability, and lung diffusion. Cytokine storms in severe Covid‐19 cases can also alter drug pharmacokinetics by suppressing hepatic enzymes like CYP450, increasing patients' exposure to certain drugs, and potentially leading to toxicity [134].

Many preclinical models were of limited relevance. Studies often relied on non‐human or irrelevant cell lines that provided limited clinical insight and overlooked more suitable models, such as primary HAE or Calu‐3 cells. Critical evaluations of the limitations of in vitro results were sometimes lacking, for example, not taking into account antiviral effects caused by drug‐induced phospholipidosis. Because animal models typically mimicked mild‐to‐moderate infections, their predictive value was limited. Murine models were often used, though mouse ACE2 receptors differ from human ACE2 receptors, limiting their relevance to humans with Covid 19 [70]. Syrian hamsters were more pathophysiologically similar, but non‐human primates (rhesus macaques, African green monkeys) offered the closest match [70], though, ethical and cost concerns limited their use [71].

Once sufficient preclinical and PK/PD data become available, repurposed drugs should be evaluated in phase II and III RCTs with validated virological and clinical criteria. Even fast‐tracked studies should rely on strong preclinical or clinical evidence. Trials based on insufficient evidence should be avoided, as they risk diverting valuable resources that could be directed toward more promising therapeutic avenues. A massive number of clinical trials were conceived and initiated, mobilized vast human, logistic, and financial resources, but remained incomplete because recruitment targets were not met [41], resources ran out, or new unfavorable data was published. These trials were usually withdrawn without explanation [2].

Even trials that ran until the pandemic ended produced some published articles that were later retracted due to, for example, fraudulent data, methodological or statistical errors, or plagiarism [42, 43]. Enrolling patients in such trials raises serious ethical concerns. Most patients who agree to participate in clinical trials expect that they and society will benefit from their participation and actively intend to contribute to advancing scientific knowledge. Enrolling patients and exposing them to potential risks is unethical if the trial is unlikely to generate useful knowledge [135, 136].

Trials based on insufficient data may also produce inconclusive results or misleading evidence [15, 137] that can confuse the public. Many people are uninformed about the scientific process, which depends upon testing hypotheses to generate evidence that may later be supported or refuted [138]. During the Covid‐19 pandemic, when fragmented and overstated findings were rapidly disseminated and then quickly contradicted by the next set of findings, growing public mistrust in science and scientists was exacerbated. To prevent further erosion of trust and misunderstandings, scientists should communicate their findings appropriately and transparently. We must be cautious in our conclusions, contextualize our findings, and avoid overstating the benefits or implications of our work. We must be clear about what can and cannot be concluded from our work when interacting with the popular press, engaging on social media, and crafting press releases [139].

Health authorities should critically evaluate proposed studies, prioritizing those with strong preclinical or clinical support. Large coordination initiatives like COVID‐CIRCLE (UKCDR and GloPID‐R) were designed to align research agendas, optimize efforts and minimize duplication [140]. “Accelerating COVID‐19 Therapeutic Interventions and Vaccines” (ACTIV), set up by the United States National Institutes of Health (NIH), initiated a public‐private partnership in April 2020 to rigorously choose the most promising drug candidates. ACTIV chose 34 of 800 candidates to include in master protocol clinical trials, but most ultimately showed no clinical benefit [141]. Regulatory and funding agencies should promote transparency in trial reporting and publication to ensure that positive and negative study findings are widely disseminated and accessible.

Intellectual property restrictions also hindered antiviral research. Patents limited access to antiviral drugs, restricting broader research efforts [142]. It may be necessary to introduce public health emergency clauses that mandate research availability for critical antiviral and antimicrobial agents, so the field can swiftly respond during crises. Public‐private partnerships could facilitate collaboration and resource‐sharing between academia and industry, so researchers have more access to promising drugs [142, 143].

Repurposing drugs with, for example, known safety profiles and production infrastructure, offers theoretical advantages, but most repurposing efforts for SARS‐CoV‐2 yielded inconclusive results. The three antiviral drugs—molnupiravir, nirmatrelvir, and remdesivir—that did prove effective against SARS‐CoV‐2 were not repurposed. They were newly developed (nirmatrelvir) or already in the industry pipeline (molnupiravir, remdesivir). Because repurposing efforts failed, the focus should be on continuing antiviral drug research. However tempting, repurposing is not a guaranteed shortcut to effective treatments, especially when drugs are repurposed without rigorous preclinical assessments and controlled clinical trials.

Molnupiravir, a nucleoside analog prodrug, initially showed antiviral efficacy against influenza and respiratory syncytial virus [144, 145]. Research by Sheahan et al. [146] confirmed its activity against SARS‐CoV‐2 in HAE models and in mice. Molnupiravir targets the SARS‐CoV‐2 RNA‐dependent polymerase RdRp [147]. A retrospective study of 192 patients found early molnupiravir use reduced disease progression risk [148]. The phase III MOVe‐OUT trial (n = 1433, including only unvaccinated patients) showed a 50% reduction in hospitalization or death versus placebo [149]. However, the AGILE CST‐2 phase II trial (n = 180) found no significant difference in viral clearance between molnupiravir and placebo [150]. Sanderson et al. [151] suggested molnupiravir‐induced mutations could increase viral transmissibility and immune escape, raising concerns about its potential long‐term risks. Molnupiravir was invented at Emory University in Atlanta, USA, then licensed to Ridgeback Biotherapeutics in 2020 after the drug demonstrated activity against SARS‐CoV‐2. Ridgeback Biotherapeutics then partnered with Merck to rapidly initiate clinical trials and accelerate molnupiravir's development [152, 153].

Nirmatrelvir progressed to efficacy trials based on its pharmacological properties. Designed by Pfizer to inhibit SARS‐CoV‐2 main protease (3CLpro), it demonstrated strong antiviral activity in cell models, with a low EC50 of 77.9 nM [154]. In mouse models, it reduced weight loss and lung viral titers [154]. A phase I RCT confirmed its efficacy with co‐administered ritonavir [155], which enhances plasma concentrations by inhibiting CYP3A4 metabolism [156]. Several RCTs in high‐risk, non‐hospitalized Covid‐19 patients indicated reduced hospitalization and mortality [157], but recent studies suggest diminishing benefits against new variants like Omicron [158].

Remdesivir, an adenosine nucleotide prodrug, was initially developed for Ebola [159]. An RCT of patients with Ebola failed to demonstrate it was as effective in reducing mortality as monoclonal antibodies [160] but Covid‐19 studies suggested remdesivir could bind SARS‐CoV‐2 RdRp [90]. Wang et al. [58] reported antiviral activity in VeroE6 cells with an EC50 of 0.77 μM and a high selectivity index (> 129). Pizzorno et al. [124] confirmed activity in an HAE model. In healthy subjects, a single intravenous dose (75 or 150 mg) yielded Cmax levels over the EC50 [161]. However, in four RCTs with over 8500 hospitalized Covid‐19 patients, remdesivir did not show an effect on clinical progression or mortality [162, 163, 164, 165]. Another RCT that included 562 non‐hospitalized high‐risk patients found remdesivir reduced hospitalization or death by 87% [166]. While remdesivir was initially developed by Gilead Sciences, the NIH helped to quickly initiate clinical trials, launching the Adaptive COVID‐19 Treatment Trial in February 2020; results suggested remdesivir might help accelerate recovery among patients hospitalized with Covid‐19 [141].

While molnupiravir originated in an academic setting and nirmatrelvir and remdesivir were industry discoveries, successful development of all three antivirals ultimately relied on strong industrial engagement throughout drug development and approval. Clinical trials were also supported by public agencies, so trials with adaptative or sequential designs could be rapidly initiated [141]. Close interactions with regulatory bodies were crucial to the process, enabling fast‐track procedures and rolling reviews that sped up drug approval [167, 168, 169]. Of all Covid‐19 clinical trials, a small minority (approximately 12%) were sponsored by biopharmaceutical companies [3]. Drug repurposing efforts might have been more successful with more industry involvement because public‐private partnerships can increase access to drugs and expedite clinical trials. Policymakers should establish frameworks that balance innovation incentives with public health needs, so nations can respond more effectively to future pandemics.

This narrative review makes clear the challenges of drug repurposing during public health emergencies by using Covid‐19 as a case study. While repurposed drugs may offer advantages, pandemic response efforts paid insufficient attention to existing pharmacological data and repurposing efforts were largely inconclusive; only newly developed antivirals showed efficacy. Major obstacles to progress included inadequate preclinical models, unaddressed pharmacokinetics, and clinical trials with design limitations, all of which must be addressed to prepare for future pandemics. Even in crises, researchers should prioritize appropriate preclinical models, address pharmacokinetic challenges, and conduct well‐designed RCTs to ensure rigorous, evidence‐based treatment strategies that improve patient outcomes.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Ben Ghezala I., Peiffer‐Smadja N., Solas C., Nougairède A., Touret F., and Bardou M., “How Can Pharmacology Help Us Overcome the Challenges of Drug Repositioning as Antivirals to Treat Emerging Pathogens? The Example of Covid‐19,” Clinical and Translational Science 19, no. 2 (2026): e70505, 10.1111/cts.70505.