Oxidative stress, i.e., overproduction of free radicals, is believed to be a significant predictor and/or a source of secondary pathologies in human disease. However, the role of oxidative stress has only been proven for some types of cancer, neurodegenerative disorders (e.g., Alzheimer’s, Parkinson’s, and Huntington’s disease), and conditions involving chronic inflammation (Halliwell, 2012; Islam et al., 2024). There is an ongoing debate for cardiovascular diseases (e.g., atherosclerosis) and eye disorders (Cheah and Halliwell, 2021). For other human diseases, including diabetes, the role of oxidative stress on the onset and secondary pathology is not that important (or at least the results are contradictory) (Seet et al., 2010; Monnier et al., 2011; Halliwell, 2024).

Free radicals are molecular entities containing at least one unpaired electron. Free radicals tend to be highly unstable and chemically reactive (however, this is not true for all free radicals). The biologically important free radicals include the oxygen-containing radical species (reactive species: RS, or reactive oxygen species: ROS). These include hydroxyl, superoxide and nitric oxide radicals, hydrogen peroxide, singlet oxygen, hypochlorous acid (HOCl), and peroxynitrite (Figure 1). They may damage important biomolecules, including lipids, proteins, and DNA (Lobo et al., 2010). However, free radicals also have indispensable physiological functions (e.g., acting as signaling molecules and regulating immune responses), and their deficiency may lead to health complications (Valko et al., 2007).

Although both animals and humans possess endogenous antioxidant defense systems, the prevailing view remains that additional antioxidant support must come from exogenous sources, particularly diet. However, many antioxidants, including so-called dietary antioxidants, have failed to demonstrate therapeutic benefits in human clinical trials (Halliwell, 2012; Halliwell et al., 2018). In some cases, they have even been associated with adverse effects, such as the increased risk of prostate cancer linked to vitamin E supplementation (Klein et al., 2011) (for more details, see Table 1). In recent years, the food and pharmaceutical industries have increasingly favored the use of natural products over synthetic compounds, based on the assumption that synthetic agents are inherently more toxic. However, this assumption is not always valid; some natural compounds such as benzoic and propionic acids can also exhibit toxicity (Shaw, 2018). From a therapeutic perspective, natural antioxidants may often be ineffective or even detrimental. Therefore, considerable attention is also being paid to synthetic antioxidants, which can be designed or modified to exert more favorable pharmacological and safety effects than their natural counterparts (see Table 2).

Synthetic AT may be divided into two main categories: (i) those that have been primarily developed and synthetized as AT and (ii) compounds that primarily act via different mechanisms of action, while their AT properties were discovered later, which may contribute to their biological activity (as observed for some clinically used drugs, e.g., 4-aminosalicylic acid with inflammatory bowel disease (IBD)). For any synthetic AT developed, it is important to know which biomolecule the agent is designed to protect, by which mechanism (scavenging RS, preventing RS formation, increasing endogenous defense mechanisms, and/or supporting oxidative damage repair), whether it can generate RS, and if there are any adverse effects associated with RS suppression. Many biologically active compounds whose benefits have been observed in clinical trials/animal studies might or might not exert antioxidant effects. It may be important to include measurements of biomarkers of oxidative damage (e.g., products of DNA and lipid oxidation, such as 8-hydroxy-2′-deoxyguanosine, 8-OHdG, or F₂-isoprostanes) in observational studies, suggesting that the therapeutic benefit may actually be associated with an antioxidant effect (Moosmann and Behl, 2002; Halliwell and Gutteridge, 2015).

Many synthetic compounds with primary or secondary antioxidant activity have been developed. The aim of this review is to summarize the knowledge on synthetic AT. Since there are many ATs, including those that were not originally developed as ATs, some substances may have been inadvertently omitted. However, the most important examples of promising AT substances will be included, focusing on those for which clinical trials have been conducted. However, most synthetic ATs have not yet progressed to clinical trials. In these cases, animal and/or in vitro studies are considered. Unfortunately, they have many limitations, and their evidence is not nearly as high compared to that of clinical studies (this is further elaborated in the Future Prospects section). In some cases, interest in the substance has ended, and there is a lack of more recent studies—for this reason, an obsolete reference is used (because a newer one does not exist).

The information for this review was obtained by performing a through literature review and search of relevant books and articles using the Web of Knowledge, SciVerse Scopus, and PubMed databases. Keyword searches were done using the names of different ATs (search period: 1960–2024; Figure 2). Chemical structures have been accurately depicted using ChemDraw software (ver 12.0.2; CambridgeSoft; Cambridge, USA). Some figures have been created with the help of BioRender image and illustration software (BioRender.com).

Superoxide dismutases (SOD; copper-, zinc-, or manganese-based) are part of the first line of AT defense in various biological systems, including humans (Figure 3; Table 3). A recombinant variant of SOD has been tested for therapeutic purposes, although it had limitations. The recombinant SOD (e.g., obtained from mice milk) has a short plasma half-life when injected into animals (Halliwell and Gutteridge, 2015). Therefore, conjugates with longer half-lives with varying therapeutic efficiency have been developed (Younus, 2018). Clinical trials with these recombinant SODs have been performed. For example, administration of recombinant CuZnSOD to premature infants attenuated inflammation, with subsequent improvement in clinical outcomes in later stages of life. Children treated with SOD had reduced incidences of pulmonary conditions (Li and Davis, 2003). Therefore, compounds mimicking the structure and activity of SOD as AT in the treatment of various diseases might give better results than recombinant variants. These are discussed below.

SOD/catalase mimetics are low-molecular compounds that usually contain transition metals, i.e., manganese, iron, copper, or zinc. Their mechanism of action is similar to the naturally occurring counterparts they are designed to mimic, i.e., to neutralize superoxide radical (O₂ ^•−) and/or hydrogen peroxide (H₂O₂) and convert them to water (Figure 3A). SOD/catalase mimetics that are based on manganese and iron porphyrins unlike earlier copper derivatives might avoid high degrading activity and releasing copper ions in vivo, thus potentially providing a pro-oxidant effect. The metal centers of SOD/catalase mimetics are more open and can participate in redox reactions more readily with more different compounds than the original SOD/catalase enzymes. Additionally, as SOD/catalase mimetics are low-molecular weight compounds, they can enter cells more easily than their natural counterparts (Day, 2004; Halliwell and Gutteridge, 2015).

Some of the SOD/catalase mimetics, e.g., SC-52608, SC-55858 (Valdivia et al., 2009), SC-54417 (Doggrell, 2002), M40401, or M40403 (Shimizu et al., 2003) (Figure 3B), will react selectively with superoxide, but not with other RS (e.g., hydroxyl radicals, hydrogen peroxide, nitric oxide, or peroxynitrite). In these structures, manganese is held by five coordination bonds and is thus able to transfer just one electron. Although considered a specific superoxide scavenger, they are able to undergo one-electron transfers with other cellular redox-active agents and enzymes (including cytochrome P450 enzymes) (Day, 2004).

Non-selective manganese-based SOD mimetics capable of reacting with other free radicals have also been produced. Eukarion has developed a series of tetra-coordinated manganese structures (designated as EUK: e.g., EUK8, EUK134, EUK139, EUK189, and EUK207; Figure 3C) that were shown to react with superoxide, H₂O_2, and peroxynitrite (Doctrow et al., 2002). The AEOL series (named after Aeolus Pharmaceuticals; Mission Viejo, California, USA; Figure 3D) contains a porphyrin system and scavenges these radicals as well (Day, 2004; Zhang et al., 2018; Forman and Zhang, 2021). Other related porphyrin ring-containing structures, such as FeTMPyP and MnTMPyP (Figure 3D), were shown to interfere with peroxynitirite. The AT mechanism of all of the above-mentioned compounds usually involves conversion of manganese to an Mn^IV-oxo form, which is then reducible by endogenous or exogenous AT (e.g., glutathione or vitamin C) activity (Halliwell and Gutteridge, 2015). The SOD mimetics in animal models have benefits with a range of oxidative stress-related diseases, including inflammation, ischemia/reperfusion, shock, thrombosis, and diabetes. However, these compounds have thus far not had been subjected to clinical trials, possibly because of unwanted side effects, including lower Ca²⁺ transport or increased heme-oxygenase (HO-1) levels (Konorev et al., 2002).

In summary, since EUK series appear to be unique, in that they can scavenge both O₂ ^•− and H₂O₂, they may be potentially more effective in conditions where the levels of both types of ROS are elevated (e.g., neurodegenerative diseases and ischemic injuries). The AEOL series and pentaazamacrocyclic ligand-based mimetics (SC series and M404 series), on the other hand, are more specialized toward O₂ ^•− (though many can also address other ROS as well) and generally lack catalase-like activity (targeted to eliminate H₂O₂), and this may limit their therapeutic use (e.g., to more acute conditions such as lung injury or effects of ionizing radiation) (Day, 2008). However, AEOL series has shown to have better stability (are excreted unchanged in the urine) and oral bioavailability compared to the EUK series, which can suffer from toxicity and instability under certain conditions (Liang et al., 2007; 2021). The SC and M404 series also face solubility and bioavailability issues. All three series have shown varying degrees of efficacy in specific tissues, such as the lungs, brain, and cardiovascular system (Day, 2004). Enhancing their ability to reach specific tissues and sites of oxidative damage is an ongoing area of research. The proposed therapeutic benefit of SOD mimetics, including with animal models, has been summarized elsewhere (Halliwell and Gutteridge, 2015).

Spin traps are used to detect free radicals both in vitro and in vivo. The idea of using spin traps as therapeutic agents arose when α-phenyl-tert-butylnitrone (PBN) was shown to have protective effects in various animal models of ischemia–reperfusion (including intestinal, cardiac, and cerebral). Many spin traps react relatively ineffectively with superoxide (e.g., PBN has weak in vitro antioxidant activity), so high doses are required to achieve a therapeutic benefit. They accumulate rapidly in various tissues, although some spin traps/nitroxides have problems crossing the blood–brain barrier (BBB). On the other hand, many spin traps seem to be safe even in large doses. Their mode of action may not be directly related to an AT mechanism but may involve other effects, as some of them (including PBN) were shown to release NO (nitric oxide), which can inhibit ROS-producing enzymatic activity. PBN was also shown to interfere with genes encoding inducible nitric oxide synthase (iNOS), cyclooxygenase-2 (COX-2), and proinflammatory cytokines (Maples et al., 2004; Halliwell and Gutteridge, 2015).

In addition, PBN administration to old gerbils decreases the levels of brain protein carbonyls and improves cognitive functions. It can cross the BBB (concentrations in gerbil brain are estimated to approach 0.5 mM after an injection of 150 mg PBN/kg body weight (Yue et al., 1992). Several derivatives of PBN have been developed as potential therapeutic agents. CPI-1429 was shown to delay mortality as well as memory impairment in an aging mouse model (Floyd et al., 2002). After that, the interest appeared to cease. It is not known whether CPI-1429 is able to cross the BBB. However, the improved state of learning and improvement in memory deficits in the mice indicate the presence of some ability. NXY-059 (also known as Cerovive^®) showed promising results in a primate model for stroke. It advanced to human clinical trials where it demonstrated effectiveness in the treatment of related ischemic injuries (Maples et al., 2004). However, it was not possible to obtain the initial results again, and this compound was also excluded from clinical trials (Antonic et al., 2018). NXY-059 is currently in human clinical trials for certain types of cancers (glioma) and auditory disorders (e.g., tinnitus and hearing loss). As for PBN, NXY-059 is a poor AT in vitro, so its activity may be due to other mechanisms of action (Maples et al., 2001). It is also hydrophilic, suggesting problems in transporting it across the BBB (unlike PBN) (Halliwell and Gutteridge, 2015). LPBNAH is another spin trap derivative. It ameliorated injury in isolated perfused rat heart (Tanguy et al., 2006) and increased the lifespan and neuroprotective action in a Philodina acuticornis (a species of freshwater bdelloid rotifers) model (Poeggeler et al., 2005). However, LPBNAH was more hydrophilic than NXY-059, so its ability to cross the BBB may be even worse. Stilbazulenyl nitrone (STAZN) showed neuroprotective effects in various animal models of ischemia/reperfusion (Belayev et al., 2002; Yang et al., 2005; Ley et al., 2007; 2008). Compared to NXY-059 and LPBNAH, STAZN is highly lipophilic and has been shown to cross the BBB (it reached a plasma concentration of ≈2.5% in the forebrain within 2–3 h after intravenous administration) (Ley et al., 2005).

The PBN molecule is converted to nitroxides with reactions with radicals. Nitroxides are radicals proposed as ATs. Examples are OXANO (Hasaniya et al., 2011), TEMPO (Yonekuta et al., 2007), and TEMPOL (Ciriminna and Pagliaro, 2010). As for PBN, the reaction with superoxide is quite slow (thus requiring large doses to be efficient in vivo) (Day, 2004). Nitroxides have been found to be capable of undergoing other redox reactions, including those involving peroxynitrite, carbonate radicals, and nitrogen dioxide. They can also interact with other entities, including vitamin C, Fe²⁺, NAD(P)H, and thiols, and they can inhibit myeloperoxidase activity. Since they show a range of redox reactions, their mechanism of action in vivo may not be related to an antioxidant effect. Some derivatives, e.g., 3-nitratomethylproxyl, in addition to typical nitroxide activity, can also donate NO. Available data suggest that these substances are capable of crossing the BBB (Zhang et al., 1998; Kwon et al., 2003). Some nitroxides (e.g., TEMPO) are used as stabilizers in plastics and polymers, and also as polymerization inhibitors (Moad et al., 2008).

The reaction products of nitroxides, hydroxylamines, may have an antioxidant effect as well. PBN undergoes degradation to benzaldehyde and N-tert-butyl-hydroxylamine, which may be a stronger AT than PBN (Atamna et al., 2000). N-tert-butyl-hydroxylamine was found to cross the BBB and showed a protective effect in a mouse model of infantile neuronal ceroid lipofuscinosis (Sarkar et al., 2013). IAC, the reaction product of TEMPO and presumably also TEMPOL, showed protective effects in animal models of various diseases, including colitis, Alzheimer’s disease, and diabetes (Novelli et al., 2007; Vasina et al., 2009; Puoliväli et al., 2011). It can cross the BBB as well (Canistro et al., 2010). Some hydroxamates, including N-methyl acetohydroxamic, N-methyl butyrohydroxamic, and N-methyl hexanoylhydroxamic acids, were also considered ATs and tested with several reperfusion models (Halliwell and Gutteridge, 2015). All these compounds are shown in Figure 4.

While spin traps and nitroxides may possess significant therapeutic potential, their clinical applications may be limited by poor bioavailability and targeting and inadequate BBB penetration (mainly due to low lipophilicity). To overcome these problems, various halogenated derivatives (e.g., chlorinated or boronated PBN), alkylated derivatives having a lipophilic tail (e.g., those with dodecyl chains), bifunctional derivatives (e.g., GS-PBN, Mito-DEPMPO, 5-ChEPMPO, DECPO, and 4-HMDEPMPO), conjugates with drugs or targeting molecules (such as PEG or heparin, that both can cross the BBB), dimers, and cyclodextrin polymers have been developed in the last few years (Han et al., 2009; Kleschyov et al., 2012; Wang et al., 2021; Marco-Contelles, 2024). Several nanoparticles with enhanced delivery abilities have been developed as well, which includes those attached to the silica core protected by poly(ethylene glycol) chains (PluS–NO), CdSe quantum dots, rotaxane-branched radical dendrimers (Gn-TEMPO), G4-polyamidoamine dendrimers, 3Gc0T zero-generation dendrimer, polyurethane dendrimers, gold nanoparticles, liquid crystal nanoparticles, nanosized sterically stabilized liposomes, poly[oligo(ethylene glycol)methyl ether acrylate] and poly(2-hydroxyethyl acrylate) conjugates, and redox nanoparticles. These derivatives showed better ROS scavenging ability (toward various radicals, such as H₂O₂) in vitro. Activities of these are thoroughly summarized elsewhere (Sadowska-Bartosz and Bartosz, 2024).

The main function of the enzyme glutathione peroxidase (GPx) is the elimination of hydrogen peroxide (see Figure 5; Table 3). Selenocysteine has an important role at the active sites of GPx (Orian et al., 2015). Consequently, low-molecular-weight selenium-containing compounds that mimic GPx’s hydrogen peroxide-scavenging activity may hold therapeutic potential. Ebselen (2-phenylbenzo[d][1,2]selenazol-3(2H)-one; Figure 5) was one of the first such compounds developed (Parnham and Sies, 2013). It can decompose peroxides. Prior to this action, ebselen needs to be reduced, i.e., the Se-containing ring is opened, and the Se is converted to selenol (-SeH). The reduced selenol then reacts with peroxide, and ebselen is regenerated. Selenol can also react with another ebselen molecule to form a diselenide, which may contribute to the catalytic cycle. Some studies, however, suggested that ebselen is an inefficient catalyst due to formation of various unreactive intermediates that prevent the regeneration of the original compound (Sarma and Mugesh, 2005; Kumar et al., 2014). The main reducing agent of ebselen in vivo is GSH. However, other compounds may have a similar role including N-acetylcysteine (NAC), reduced thioredoxin, and dihydrolipoate (Parnham and Sies, 2013). Ebselen has shown positive effects with numerous animal models of diseases, including metabolic syndrome, noise-induced hearing loss, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-induced Parkinson’s disease, alcohol-induced liver injury, atherosclerosis, and diabetes (Day, 2008; Halliwell and Gutteridge, 2015). It also showed some therapeutic benefit in clinical studies of various diseases, mainly stroke, hearing loss, Meniere’s disease, and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (Forman and Zhang, 2021; Ramli et al., 2022; Sahoo et al., 2023); however, it was ineffective in diabetes (Beckman et al., 2016). These results indicate that ebselen (and GPx mimetics in general) could be particularly useful in neurodegenerative and respiratory disorders. Ebselen has AT properties in vitro against various radical species (including HOCl, singlet oxygen, and peroxynitrite). However, it also showed anti-inflammatory actions, such as inhibition of 5- and 15-lipoxygenases (LOX), NOS, and phagocyte-related ROS production (Day, 2008) (see Section 11.2). Its therapeutic activity may not involve an AT mechanism.

Ebselen has also been tested in combination with other ATs that may improve its efficacy. Together with vitamin E, it modulated the activity of acetylcholinesterase and reduced demyelinating events in different areas of rat brains (Mazzanti et al., 2009). The combined oral formulation of ebselen/allopurinol reduced multiple cisplatin toxicities in rat breast and ovarian cancers (Lynch et al., 2005). Inhalable microparticles combining remdesivir and ebselen demonstrated antiviral properties against SARS-CoV-2 infection (Saha et al., 2023). In addition, ebselen was also tested together with various antibacterials such as daptomycin, retapamulin, fusidic acid, and mupirocin and exhibited synergistic effects with these drugs (Thakare et al., 2020). Similarly, ebselen also showed synergistic effects with silver, making it more selective for pathogenic bacteria than for mammalian cells (Zou et al., 2017). There are also reports of ebselen conjugates with commonly used drugs, such as clioquinol, a compound known for its ability to chelate metal ions and inhibit Aβ deposition. The conjugate was able to penetrate the CNS without inducing toxicity in vivo and demonstrated inhibition of Aβ aggregation and H₂O₂ scavenging activity in in vitro conditions (Wang et al., 2016). Ebselen has also been formulated to nanoemuslsions that showed antifungal activity against vulvovaginal candidiasis in a mouse model (Menon et al., 2021). Again, the above studies suggest that ebselen may be valuable in the treatment of neurological disorders and as an antimicrobial agent in various conditions (especially respiratory disorders).

Another compound that has GPx-like activity is BXT-51072 (Figure 5). It has a better protective effect than ebselen, being several-fold more reactive in catalyzing the peroxide reaction. BXT-51072 may help with several diseases including IBD, asthma, chronic obstructive pulmonary disease, and stroke. It showed promising results in clinical trials of ulcerative colitis. Currently, its further development is not happening (May, 2016; Forman and Zhang, 2021).

Tellurium has similar chemical properties as Se. Various Te derivatives have been developed and may eliminate various radical species, such as peroxides and peroxynitrite, in vitro (Pariagh et al., 2005; Lu et al., 2017). Research on these compounds is still limited.

Vitamin E (tocopherols) deficiency is associated with neurological and reproductive abnormalities. These conditions usually disappear after adequate supplementation with vitamin E (Dewick, 2009). Similarly, administration of vitamin E to infants and adults suffering from inborn glutathione deficiency have provided therapeutic benefits (although quite limited) (Halliwell and Gutteridge, 2015). However, with other diseases putatively associated with oxidative stress, including diabetes, cardiovascular disorders, cancer, and various neurodegenerative disorders, vitamin E does not appear to have a beneficial effect (Robinson et al., 2006; Steinhubl, 2008; Halliwell, 2024). Additionally, vitamin E supplementation is not entirely safe as higher doses have been associated with specific side effects, including hemorrhagic stroke (Sesso et al., 2008) and increased risk of prostatic cancer (Klein et al., 2011) (Table 1). There are several possible reasons why vitamin E is ineffective in the treatment of human diseases. Its bioavailability is very less, and it takes a relatively long time for significant concentrations of vitamin E to accumulate in the target tissues (Halliwell and Gutteridge, 2015; Halliwell, 2024). Vitamin E and its derivatives (e.g., α-tocopheryl acetate) are commonly used in the food industry as preservatives to prevent rancidity by reducing the rate of lipid peroxidation (Dewick, 2009). However, vitamin E appears to be ineffective at inhibiting lipid peroxidation in humans, or at least its efficacy may vary depending on the specific disease. Additionally, vitamin E may not be biologically important as an AT, but rather as a signaling molecule and regulator of specific gene expressions (Azzi and Zingg, 2005).

Various structurally related derivatives of vitamin E have been developed, some of which have improved antioxidant effects with in vitro AT assays or animal models. Some have a benzofuran ring instead of the chroman ring of vitamin E, e.g., BO-653. It showed an antiatherosclerosis effect with animal models and lowered F₂-isoprostane levels (that are often used as biomarkers for oxidative stress) in vitamin E-deficient mice (Halliwell and Gutteridge, 2015; Valgimigli and Amorati, 2019). BO-653 was able to suppress hepatitis C replication (Yasui et al., 2013). However, no further clinical development has taken place since then. Raxofelast is a water-soluble form of vitamin E that improves vascular endothelial dysfunction in diabetes (Hadi and Al Suwaidi, 2007) and aids wound healing (Bitto et al., 2007). It undergoes deacetylation hydrolysis to form IRFI-005 in vivo. IRFI-005 may have greater antioxidant action than raxofelast (Barzegar et al., 2011). Trolox is another hydrophilic analog of vitamin E that is a good scavenger of peroxyl and alkoxyl radicals. Upon reaction with them, a resonance-stabilized Trolox radical will be formed, which is presumably relatively easily regenerated by vitamin C (or natural vitamin E; Figure 6A). However, due to its high hydrophilicity, Trolox enters cells with difficulty. It is extensively used as a reference compound in a number of in vitro AT assays (MacDonald-Wicks et al., 2006).

MDL 74,405 contains a quaternary ammonium group. It is highly cardioselective as it is predominantly deposited in the heart relative to other tissues and blood (Chan et al., 1994; Kuo et al., 1995). It showed positive effects in dog models of heart ischemia/reperfusion (Tang et al., 1995a; 1995b). Further work has not been carried out. Troglitazone is an antidiabetic drug with AT properties, which exerts its blood sugar-lowering activity by enhancing insulin sensitivity. It shows significant hepatotoxicity as it readily oxidizes to form various radicals, including quinones (Kassahun et al., 2001). It exerted a protective effect with brain ischemia in rats. However, further work has not been carried out. CX-659S showed anti-inflammatory activity in animals (contact hypersensitivity and allergic reactions) (Inoue et al., 2003), but this has not been followed-up. ETS-GS is a combination of vitamin E, taurine, and GSH that showed protective effects with various models of ischemia/reperfusion and inflammation (Sugita et al., 2013). All agents discussed in this section are shown in Figure 6B.

Many animals can synthesize vitamin C. Some, however, including certain birds, guinea pigs, bats, primates, and humans, lack the necessary biosynthetic apparatus and must acquire vitamin C through external sources, i.e., diet. Vitamin C deficiency leads to serious health complications, including scurvy, skin lesions, fragile blood vessels, bleeding gums, and tooth loss. The beneficial effects of vitamin C are a result of its antioxidant activity, especially as it provides a regenerating system for vitamin E. However, vitamin C has other important biological roles. It is an important cofactor in the hydroxylation of proline to 4-hydroxyproline and lysine to 5-hydroxylysine, which accounts for ∼25% of the collagen structure composition. Vitamin C is also associated with hydroxylation of tyrosine in the synthesis of catecholamines (dopamine, noradrenaline, and adrenaline) and homogentisic acid, the precursor of tocopherols and plastoquinones (Dewick, 2009). Its therapeutic benefit may, therefore, not always be necessarily associated with its antioxidant action (Table 1).

Various lipophilic esters of vitamin C have been synthesized. Examples are ascorbyl palmitate and 2-octadecylascorbate. Both have been used in the food industry as food preservatives. They have also been tested as ATs with various animal disease models (Halliwell and Gutteridge, 2015). However, it seems that they are not of interest as medicinal agents. EPC-K1 is a phosphate ester derivative of vitamins C and E. It was extensively investigated with animal models of ischemia/reperfusion and stroke (Zhang et al., 2001; Kato et al., 2003; Yamamoto et al., 2011), although there have been no human clinical trials. Synthetic vitamin C analogs are shown in Figure 6C.

As indicated above, vitamins E and C and their respective derivatives have largely failed to provide therapeutic benefits in many studies because of some possible reasons. Inefficiency may be caused by limited tissue retention. For example, it is known that α-tocopherol can enter the brain, but it is unclear how far its supplementation elevates human brain levels (Halliwell, 2024). In contrast, vitamin C supplementation does not appear to increase levels in the brain at all (Terpstra et al., 2011). Even if the compound enters the target tissue/organ, it may not reach the active sites of oxidative damage (e.g., mitochondria). It may reach the correct site, but may not be targeted toward the specific agent responsible for oxidative damage, as observed in the case of vitamin E that inhibits lipid peroxidation, but may not prevent the increased oxidative DNA, RNA, and protein damage seen in Alzheimer’s disease (Praticò, 2008). Intervention with vitamin E/C derivatives could have been initiated at advanced stages of the disease and not during the onset and early stages (in other words, it was too late to be effective). High doses of vitamin E/C derivatives were studied. Some in vitro studies suggest that high doses of vitamin E and C can be pro-oxidant in nature (Witting et al., 1999; Pearson et al., 2021), and this may also apply for the semisynthetic variants. Studies may also have used high doses of one agent, which may have resulted in a reduced uptake and/or distribution of other substances with important metabolic activity. For example, high doses of a semi-synthetic variant of vitamin E/C could have lead to reduced levels of endogenous natural tocopherols. There may be also issues with instability and degradation (vitamins E and C are quite prone to degradation in solutions) or low interaction of the semi-synthetic derivatives with the delivery systems (such as sodium-dependent vitamin C transporters).

To overcome some of these limitations, several advanced formulation strategies have been developed to increase the bioavailability, stability, and targeted delivery of these vitamins and their semisynthetic forms. These include the incorporation of the vitamins to the liposomes (Marsanasco et al., 2011), poly(lactic-co-glycolic) acid (PLGA) nanoparticles (Astete et al., 2011), solid lipid nanoparticles (Gambaro et al., 2022), nanostructured lipid carriers (Saez et al., 1984), chitosan carriers (Sherif et al., 2024), nanocapsules (Khayata et al., 2012), cyclodextrin complexes (Celebioglu and Uyar, 2017; Khan et al., 2023), and peptides and antibody conjugates (Jung et al., 2018). These systems may represent new horizons for compounds that have not been clinically tested thus far or that have previously failed in clinical settings as potential ATs. Whether these advanced formulations solve the above discussed issues of vitamin E and C analogs remains to be known.

Various drug classes have been suggested as candidates to develop analogs of commonly used pharmaceuticals with enhanced AT properties, including hypolipidemic agents, anesthetics, non-steroidal anti-inflammatory drugs, or antiarrhythmic agents. Methylprednisolone exhibits anti-inflammatory activity by inhibiting the phospholipase enzymatic activity. It also inhibited lipid peroxidation in the brain during traumatic events. These results led to the development of lazaroids (21-aminosteroids), in which the steroid skeleton has been improved with addition of an AT component (Kavanagh and Kam, 2001). Many lazaroid derivatives have been developed that inhibit iron-dependent lipid peroxidation in brain cells in vitro, as well as show neuroprotective activity in various animal models of traumatic brain injury (TBI) and of the spinal cord. Perhaps the most studied lazaroid is the tirilazad mesylate (Freedox; U-74006F, U series named after the Upjohn Company). It was subjected to several clinical trials for stroke (Putman et al., 1994), spinal cord injury (Bracken et al., 1998), and TBI (Marshall and Marshall, 1995). However, it failed to provide any notable therapeutic benefit in most patients and in some even produced adverse effects. However, tirilazad mesylate may be helpful in treating subarachnoid hemorrhage in men, but not in women, as it appears that women metabolize it much faster (Cahill and Hall, 2017). There is little evidence that lazaroids function as ATs in vivo; therefore, their therapeutic benefits are likely due to other mechanisms.

Tirilazad mesylate accumulates at the BBB and has limited penetration into the brain . Some related derivatives, such as U-101033E or U-104067F, cross the BBB more readily (Halliwell and Gutteridge, 2015). So far, they have not been tested in clinical studies. However, preclinical data on animals (gerbils) showed that both U-101033E and U-104067F can significantly attenuate the post-ischemic loss of dopaminergic nigrostriatal neurons (Andrus et al., 1997). Another analog that, like tirilazad mesylate, has been intensively tested in neurological conditions is edaravone. It has shown some therapeutic benefits in clinical trials of stroke and amyotrophic lateral sclerosis (Miyaji et al., 2015; Kobayashi et al., 2019; Soares et al., 2023). In some countries, it is used to treat amyotrophic lateral sclerosis and is administered to patients after stroke to facilitate their recovery. Edaravone is the only member of the lazaroid compound which is used clinically as an AT (Soares et al., 2023; Petrov et al., 2017). It may also act as an anti-inflammatory drug owing to its beneficial effects in animals with inflammatory conditions (Yuan et al., 2014). Additionally, the available clinical trials did not measure the biomarkers of oxidative damage, so its mechanism of action may not be indeed related to the antioxidant effects. The clinical efficacy of edaravone has been thoroughly reviewed elsewhere (Halliwell and Gutteridge, 2015; Halliwell, 2024). Even though edavarone seems to be of value in neurological disorders, it has limited ability to cross the BBB. Quite recently, some edaravone analogs with improved BBB penetration and overall efficacy were developed. For example, intranasal administration of edaravone PLGA nanoparticles showed improved brain stability and bioavailability and reduced H₂O₂-induced oxidative stress toxicity in mouse microglial cell line BV-2 (Lu et al., 2023). Edaravone-MIL-53(Cr) nanoparticles alleviated brain injury and cognitive dysfunction in mice receiving whole-brain irradiation (Li et al., 2025). It was also conjugated with several drugs, including glutathione, in the form of poly(methacrylic) nanogel, which had improved targeted delivery to the brain and elevated cognitive function in Wistar rats (Mozafari et al., 2023). Administration of an ionic liquid formulation of edaravone, combined with tetrabutylphosphonium cation, resulted in prolonged blood circulation time and reduced kidney distribution. It also demonstrated cerebroprotective effects comparable to those of edaravone in a rat model of cerebral ischemia/reperfusion injury (Fukuta et al., 2023). A series of edaravone derivatives bearing N-benzyl pyridinium moieties demonstrated significant in vitro acetylcholine inhibitory activity (with low activity toward butyrylcholinesterase) and antioxidant effect (Zondagh et al., 2020). Edaravone was also used in combination with the anti-inflammatory drug dexborneol and showed promising results in clinical trials of stroke. This combination was even more effective than edaravone alone (Xu et al., 2021). Structures of lazaroids are shown in Figure 8.

Glutathione (GSH) is an endogenous AT that is clinically useful. It provides the substrate for glutathione peroxidase (GPx; Figure 4 in Section 5) but can also directly scavenge several ROS, including peroxynitrite, HOCl, and OH^• (Tambyraja et al., 2007). It also protects tissues against the effects of various cytotoxic agents, such as cyclophosphamide (Lopez and Luderer, 2004). GSH also protects the lungs against RS-induced damage (e.g., in cystic fibrosis, where patients suffering from this condition tend to have lower GSH levels) (Bozic et al., 2020). However, despite having endogenous AT activity, GSH may induce serious side effects. For example, it caused bronchoconstriction in asthma patients (Halliwell and Gutteridge, 2015).

GSH does not readily enter cells. Some derivatives with improved membrane-crossing properties were developed, including methyl, isopropyl, and ethyl monoesters. Upon entering the cells, they are hydrolyzed to glutathione. Diethylester derivatives were seen to be more easily delivered into cells (Cacciatore et al., 2010).

2-Oxothiazolidine-4-carboxylic acid (OTC) is hydrolyzed to cysteine in vivo. Administration of OTC may increase endogenous GSH synthesis as cysteine is one of the three amino acids of GSH. The therapeutic efficiency of OTC is ambiguous. It decreased allergen-induced airway injury in an animal model of asthma. However, it showed no benefit in reducing oxidative stress in HIV patients (Halliwell and Gutteridge, 2015). Structures of compounds discussed in this section are shown in Figure 9A.

N-acetylcysteine (NAC) is used as a standard reference compound in many AT assays and is claimed to be effective in the treatment of paracetamol (acetaminophen) overdosage and toxicity. The mode of therapeutic action of NAC is believed to involve its hydrolysis in the affected cells to cysteine, leading to increased synthesis of GSH. NAC is also capable of scavenging various radicals (Forman and Zhang, 2021). However, these results were obtained using methods with serious limitations (e.g., based on boronate or dichloroflorescin diacetate probes) and may not be relevant in vivo, as previously reviewed elsewhere (Murphy et al., 2011; Halliwell, 2012). NAC was also shown to have non-AT activities, including interaction with receptor for N-methyl-D-aspartate (NMDA) and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA) and inhibition of NF-κB activity. Its therapeutic efficiency is also dubious. It has been tested for the treatment of various respiratory disorders (e.g., bronchopulmonary dysplasia in infants) (Ahola et al., 2003). However, it showed contradictory results. NAC has also been tested for the prevention of cardiovascular disorders in patients with kidney failure, where the results were more convincing (Ye et al., 2021). However, in one study, the observed positive effect was not associated with a reduction in urinary F₂-isoprostane levels (Efrati et al., 2003), suggesting that NAC’s protective action may involve mechanisms other than antioxidant activity. This is consistent with the fact that NAC, like most thiols, is a poor scavenger of hydrogen peroxide and superoxide (Winterbourn and Metodiewa, 1999). NAC also reduced the rate of lung deterioration in patients with idiopathic pulmonary fibrosis (Demedts et al., 2005). Few NAC derivatives have been developed, the most notable being N-acetylcysteine amide (AD4). It can cross the blood–brain barrier (unlike NAC). AD4 may have therapeutic benefits in various neurodegenerative disorders and inflammatory conditions (Sunitha et al., 2013).

The mitochondria are significant producers of reactive species, and mitochondrial damage is one of the major contributors to the onset of various diseases and aging. Therefore, developing antioxidants specifically designed to selectively target the mitochondria could be beneficial for treating the aforementioned conditions. Indeed, research attention in the last decade has focused significantly on antioxidants targeting the mitochondria, and it appears that the focus is shifting from more traditional antioxidants (see Figure 2). One approach in creating a mitochondria-targeted AT is to bind an antioxidant molecule (e.g., SOD, thiol, vitamin E, BHT, ebselen, CoQ, or various spin traps, such as TEMPO) with a hydrocarbon chain of varying length to the phosphorus of the lipophilic compound triphenylphosphonium (TPP) such as mitoquinol (MitoQ₁₀; Figure 10). These compounds accumulate in the mitochondria and were shown to protect cell cultures from damage caused by the addition of peroxide or hypoxia. Animal studies reveal that they can enter all tissues, including the brain. MitoQ₁₀ is the most studied of them all. MitoQ₁₀ has been investigated in phase II clinical trials in humans for the treatment of Parkinson’s disease and chronic hepatitis C (Kezic et al., 2016). In addition, it also showed therapeutic benefits in various animal disease models, including cardiovascular disorders (Silva et al., 2016); conditions involving ischemia/reperfusion (Kezic et al., 2016); neurodegenerative disorders, e.g., Alzheimer’s (Oliver and Reddy, 2019) and Parkinson’s disease (Liu and Wang, 2014); and traumatic brain injury (Ismail et al., 2020). However, the results are not conclusive, and some studies have observed no protective effects at all (Adlam et al., 2005). Mitochondria-targeted antioxidants can be valuable tools for studying the effects of mitochondrial ROS removal on both normal physiology and pathological conditions. Compounds like TPP and similar molecules can be used to make ROS-detecting agents, such as dihydroethidium, mitochondria-specific (Smith et al., 2012; Halliwell and Gutteridge, 2015).

Various chelating agents have been used to inhibit metal-dependent oxidative damage. Metals like iron and copper are considered significant contributors to oxidative stress. The antioxidant action of chelators may be achieved using different mechanisms. These include the binding of a metal to a given AT, leading to decreased production of RS or the ability to scavenge RS. The former mechanism is preferred because RS scavengers can be consumed during the reaction and may generate toxic radicals derived from the chelators (Halliwell and Gutteridge, 2015).

One of the first chelating agents documented to reduce RS production (specifically the hydroxyl radical) in vitro was diethylenetriaminepentaacetic acid (DTPA). The antioxidant action of DTPA is derived from its ability to chelate Fe³⁺ and thus prevent its reaction with superoxide and hydrogen peroxide, leading to formation of hydroxyl radicals (i.e., preventing Haber–Weiss and Fenton’s reactions) (Cohen and Sinet, 1982). The resulting molecule, Fe³⁺–DTPA is slowly reduced by superoxide, which results in the production of fewer hydroxyl radicals (Rahhal and Richter, 1988). However, hydroxyl radical may still be formed, i.e., during the reaction of Fe²⁺–DTPA with hydrogen peroxide (Cohen and Sinet, 1982; Egan et al., 1992). Superoxide is not the only reducing agent of the Fe³⁺–DTPA complex. It may be reduced by other, more powerful agents (Sutton and Winterbourn, 1984). DTPA, therefore, is not a general inhibitor of iron-dependent hydroxyl radical production. DTPA is currently used only marginally as it causes depletion of certain important metals, such as zinc (Arts et al., 2018). It has been used to ward off lead, plutonium (De Bruin, 1967; Deblonde et al., 2018), and also iron poisoning (McDonald, 1966).

Ethylenediaminetetraacetic acid (EDTA) is a common iron chelator, which is a structural analog of DTPA. It is reduced by superoxide more rapidly than DTPA (Halliwell, 1978). EDTA chelates several metal ions, and its calcium disodium derivative was used in chelation therapy, such as for treating lead poisoning (Bjørklund et al., 2017). It was also used in the treatment of an iron overdose and cardiovascular disorders. However, some studies have suggested its poor efficiency (Matteucci et al., 2006; Lamas et al., 2013).

Other agents such as phytic acid (inositol hexaphosphate), 1,10-phenanthroline, and deferoxamine (desferrioxamine; DFO) are better inhibitors of iron-dependent hydroxyl radical generation than DTPA. Phytic acid has a history of use in the food industry as an AT. Its use has been discontinued in many countries because of its antinutrient properties, decreasing the absorption of various dietary metals (e.g., phosphorus, zinc, calcium, and copper) in the gastrointestinal tract (Bloot et al., 2023). On the other hand, phytic acid can provide a protective effect in the gut by sequestrating iron and preventing its pro-oxidant effect (Halliwell and Gutteridge, 2015).

1,10-Phenantroline readily chelates zinc, iron, and copper ions (Bencini and Lippolis, 2010). It was shown to prevent DNA degradation mediated by hydrogen peroxide (Mello-Filho and Meneghini, 1991). On the other hand, Cu²⁺–phenantroline complexes can lead to DNA damage (Bales et al., 2005).

DFO is a strong (but not entirely specific) chelator of Fe³⁺. It inhibits iron-dependent lipid peroxidation and conversion of hydrogen peroxide to hydroxyl radicals (Fenton’s reaction) in the presence of physiological buffer systems. DFO is a natural compound produced by Streptomyces pilosus that is of value in the prevention and treatment of acute iron poisoning, and in cases of repeated blood transfusion, e.g., in thalassemia, an inherited blood disorder characterized by decreased hemoglobin production. DFO seems to be safe, the daily doses can go up to 50 mg/kg body weight. However, there is a chance that patients will develop iron deficiency. DFO is not orally active and needs to be administered intravenously or subcutaneously (Merlot et al., 2013). Large doses can lead to hearing and visual disorders. These effects usually disappear on removal of the drug (Chen et al., 2005). A DFO overdose is also associated with certain bacterial and fungal infections, including those caused by Yersinia enterocolitica, Vibrio vulnificus, and Rhizopus spp. (Neupane and Kim, 2009). The Fe³⁺–DFO complex (known as feroxamine or ferrioxamine) is difficult to reduce, not only by superoxide but also other reducing agents (compared with Fe³⁺–DTPA). Despite having a high affinity for iron, DFO is unable to remove iron from important structures, such as transferrin. DFO is hydrophilic and therefore does not readily enter cell membranes (Halliwell and Gutteridge, 2015). It is sometimes part of organ preservation fluids, e.g., for heart transplantation (Chen et al., 2019).

DFO was shown to scavenge various radicals, including superoxide, hydroxyl radical, and HOCl. It does not respond to peroxynitrite, although it can react with the RS derived from it, such as nitrogen dioxide and the carbonate radical. Again, the antioxidant action of DFO in vivo may be of little importance because the blood levels reached during therapy (∼20 μM) are too low to achieve the desired effect (Halliwell and Gutteridge, 2015). However, it may have other activities. DFO reduced chronic inflammation and symptoms of autoimmune diseases in animal studies (Di Paola et al., 2022). It also inhibited cell proliferation. Thus, the mechanism underlying the anti-inflammatory effect of DFO may be due to a reduction in the number of inflammatory cells (e.g., lymphocytes) (Choi et al., 2004). DFO also increases the levels of hypoxia-inducible factor (HIF-1) in various cell types (Figure 11A), resulting in amplified transcription of genes promoting erythropoiesis, glycolysis, and angiogenesis (Woo et al., 2006). The therapeutic benefits of DFO may be due to other non-AT activities. Increased HIF-1 activity is also observed for other iron chelators, such as salicylaldehyde isonicotinoyl hydrazone (SIH) and deferiprone (Creighton-Gutteridge and Tyrrell, 2002).

Among the DFO analogs are those where the DFO molecule has been attached to various high-molecular weight polymers, including cellulose, dextran, or hydroxyethyl starch. In these derivatives, the iron-binding activity has been significantly reduced. However, toxicity has also been decreased while the circulating plasma half-life has increased, and therefore, higher doses of these agents can be administered. They supposedly enter cells more readily than the original DFO molecule (Kalinowski and Richardson, 2005). These derivatives have been tested with various animal models of disease, e.g., septic shock and diabetes, where they showed a positive effect. DFO and its derivatives still have limited clinical use outside of its original indication, i.e., iron poisoning. They have been studied as a possible treatment for coronary artery disease (Duffy et al., 2001), spinal cord injury (Yao et al., 2019), and intracerebral hemorrhage (Zeng et al., 2018; Ren et al., 2020). Metal chelators discussed in this section are shown in Figure 11A.

ATs are of interest both as therapeutic agents and as food preservatives. Natural products, as well as semi-synthetic and synthetic analogs, are being actively investigated—based on the premise that structural modifications may yield compounds with superior properties compared to their natural counterparts. The proposed mechanisms of action and therapeutic benefits of synthetic ATs are summarized in Table 4. Despite their prevalence in the human diet, natural antioxidants, particularly dietary flavonoids, have largely failed to demonstrate therapeutic efficacy and may not serve as suitable templates for drug development. Although a wide range of semi-synthetic and synthetic derivatives have been developed and studied, the discovery of a medicinally effective antioxidant remains elusive. There is currently little conclusive evidence that oxidative stress-related diseases can be effectively treated by administering antioxidants. Many compounds have failed during preclinical or clinical evaluation. In cases where therapeutic benefits were observed in clinical trials, it often became apparent that these effects were unrelated to the compound’s antioxidant activity. For some agents, antioxidant activity may contribute to their efficacy, but therapeutic doses are frequently too low to elicit a meaningful antioxidant response. Additional limitations include poor solubility, low oral bioavailability, limited tissue distribution, and serious adverse effects. As a result, only a small subset of (semi-)synthetic antioxidant analogs appears to warrant further investigation. Among them, ebselen, along with other selenium- and tellurium-containing compounds, has shown some promise in both animal models and clinical trials targeting oxidative stress-related conditions. It is clinically used to treat stroke in some countries (Japan). More studies are, however, needed to prove its efficacy in other clinical scenarios and if its therapeutic benefits are indeed associated with antioxidant activity. Edaravone is used in the treatment of stroke and amyotrophic lateral sclerosis in some countries. However, it remains unclear whether it truly functions as an antioxidant–it is still not known whether it can scavenge free radicals at concentrations achievable in the body after a normal dose. No clinical study has measured oxidative stress levels following its administration, and its exact mechanism of action remains unknown. NAC is of value in the treatment of paracetamol overdose and is also used as a mucolytic agent. In the literature, NAC is often, albeit somewhat exaggeratedly, referred to almost synonymously with the term antioxidant. However, this compound also exhibits a range of other biological activities. As with edaravone and ebselen, it is therefore questionable whether its therapeutic benefits are primarily due to its antioxidant properties. Its ability to mitigate paracetamol toxicity is likely related to its role as a cysteine substitute in the synthesis of GSH. Paradoxically, this same mechanism may underlie its potential pro-cancer effects: by boosting GSH production, cancer cells can better protect themselves from oxidative damage caused by ROS (Ramachandran and Jaeschke, 2021). Mitochondria-targeted ATs (e.g., mitoquinol) have also shown potential. Additionally, various metal chelators are already used clinically to treat specific conditions. However, their therapeutic index is often narrow—as exemplified by deferoxamine, which, despite being well-studied, is currently limited to treating acute iron poisoning and conditions involving chronic blood transfusions (e.g., thalassemia). Enzymes of the NOX family have emerged as promising targets for antioxidant therapy. Compounds that selectively inhibit specific NOX isoforms, such as S17834 or GKT137831, may be particularly effective. Ebselen and structurally related compounds may exert their therapeutic effects, at least in part, through NOX inhibition. While NOX inhibitors hold significant promise, they are relatively new to antioxidant research, and additional studies are necessary to validate their clinical efficacy. In contrast, antioxidant-based food preservatives appear to be more successful in practical application. Nevertheless, safety concerns remain as some compounds—such as BHT and BHA—have been associated with potential toxicity. However, the concentration used in food is typically so low that toxic effects are considered unlikely. Still, several preservatives, such as nordihydroguaiaretic acid, have been withdrawn due to safety concerns.