Oxidative stress levels in long COVID patients during the Omicron wave in Japan

Coronavirus disease 2019 (COVID-19) causes a wide variety of symptoms, not only during the acute phase following severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, but also during the chronic phase, affecting up to 20% to 40% of patients [1,2,3,4]. This persistent illness, referred to as post-COVID-19 condition (PCC), post-acute sequelae of COVID-19 (PASC), or long COVID [5,6], manifests as a broad spectrum of symptoms, including fatigue, dyspnea, myalgia, arthralgia, and cognitive dysfunction, including concentration deficit. Long COVID typically begins within two to three months of initial infection and cannot be explained by an alternative diagnosis [7]. The pathophysiology of long COVID remains unclear, although various mechanisms, including microthrombus, persistent inflammation, autoimmunity, viral persistence, neurological damage, vascular dysfunction, and hypochondria, have been proposed [8,9,10].

Epidemiological data suggest that, partly due to the impact of vaccination, the overall risk of developing long COVID has decreased from early pandemic strains to the Omicron phase, both in Europe and in Japan [11,12,13,14]. Nevertheless, long COVID still occurs in approximately 10% of individuals infected with COVID-19 [11,13], indicating that it remains a significant public health concern. Furthermore, in Japan, the risk of cognitive symptoms, such as brain fog, appears to be relatively higher during the Omicron period compared to previous variants [15].

This complex mechanism suggests the need for multidisciplinary rather than single-modality therapeutic approaches. Patients with long COVID frequently report symptoms that resemble those of myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) [16,17]. Degenerative conditions characterized by persistent multisystemic symptoms lasting longer than six months include pathological fatigue, mental and physical exhaustion, sleep disturbances, pain, autonomic dysfunction, and cognitive impairment [18,19]. Although some screening tools for symptom differentiation have been established, and elevated serum ferritin has been proposed as a possible biomarker for ME/CFS in patients with long COVID, no definitive biomarkers have been identified to date [11,20,21,22]. Various endocrine factors, including those related to the hypothalamo–pituitary–adrenal (HPA) axis, sex steroids, and thyroid hormones, may also be involved in the development and persistence of some of the long COVID symptoms [23,24,25,26,27].

In the present study, we aimed to investigate whether markers of oxidative stress and/or antioxidant activity, which are frequently altered in infectious diseases, could serve as potential monitoring biomarkers for distinguishing subgroups of patients with long COVID. Furthermore, we assessed whether these markers could help exclude cases that mimic long COVID but are in fact false positives. There have been only a few reports evaluating oxidative stress markers in patients with COVID-19, primarily during the acute phase. These studies aimed to assess clinical management strategies of inflammatory conditions and disease severity, facilitate the stratification of at-risk individuals, and optimize the allocation of public health resources [28,29]. However, fundamental data on patients with long COVID remain lacking. We therefore assessed oxidative stress markers, including serum levels of diacron-reactive oxygen metabolites (d-ROMs) and biological antioxidant potential (BAP), and calculated oxidative stress index (OSI) based on these values [30,31,32]. We then examined how these markers were associated with clinical background and individual patient characteristics.

As shown in Table 1, 77 patients with long COVID (41 [53%] females and 36 [47%] males; median age: 44 years) who visited our outpatient clinic were included in the present study. Smoking habits were reported by 30% (23 cases) of the patients, and 94% (72 cases) of them experienced only mild severity during the acute phase of COVID-19 (Table 1). Among the participants, 64% (49 cases) visited the hospital >90 days after the onset of infection of SARS-CoV-2, 16% (12 cases) had received no COVID-19 vaccination, none had received only one dose, and 84% (65 cases) had received at least two doses of COVID-19 vaccinations (Table 1).

As shown in Table 2, in 77 patients with long COVID, the median [interquartile range (IQR)] serum levels of d-ROMs (CARR Unit), BAP (μmol/L), and OSI were 533.8 [454.9–627.6], 2385.8 [2169.2–2558.1], and 2.0 [1.7–2.5], respectively (Table 2). Reference data from a healthy population (control group), obtained using the same methods, have been reported in previous studies, as shown in Table 2 [31,32]. The control group had a mean age of 36.7 years (standard deviation 8.8), consisting of 164 women and 148 men. Referential data are shown as median [IQR] values from healthy controls. Median [IQR] d-ROM, BAP, and OSI values in the 312 healthy cases were 287.4 [252.8–314.5] CARR Units, 2545.7 [2503.1–2583.5] mol/L, 1.0 [0.9–1.1], respectively [31,32] (Table 2). On the basis of the referential data, d-ROM and OSI values in the current 77 patients with long COVID, including 53% female cases, were significantly higher (p < 0.01) than those previously reported in the healthy population, while the levels of BAP in the patients were also significantly lower (p < 0.01) than the healthy controls (Table 2). Using the same measurement methods in a separate cohort of 12 healthy female participants (mean age 20.4 ± 0.5 years), pre-experimental acute fatigue measurements were as follows (mean ± standard deviation): d-ROMs, 303.4 ± 24.4 CARR Units; BAP, 2371.8 ± 77.6 μmol/L; OSI, 1.13 ± 0.07 [30]. Thus, the present data on the oxidative stress markers differed apparently compared to the healthy controls.

In addition, there were no significant differences in oxidative stress markers between individuals with (65 cases) and without (12 cases) a history of two and more vaccinations (Table 3).

As shown in Figure 1, d-ROM (median: 579.8 vs. 462.2, ** p < 0.01) and OSI (median: 2.3 vs. 1.8, * p < 0.05) levels were significantly higher in female than in male patients with long COVID, while BAP levels (median: 2403.4 vs. 2352.6) did not significantly differ by sex (Figure 1A). OSI levels positively correlated with patient age (R = 0.33; ** p < 0.01) and BMI (R = 0.27; * p < 0.05), whereas BAP levels negatively correlated with age (R= −0.37; ** p < 0.01) and BMI (R = −0.45; ** p < 0.01) (Figure 1B,C). The levels of d-ROM were not significantly correlated with the patients' age or BMI. In multivariable linear regression models restricted to patients with long COVID (Table 4), OSI was not significantly associated with age (β = 0.007; p = 0.063). Male sex was associated with lower OSI values compared to female sex (β = −0.322; p < 0.05), and BMI was positively associated with OSI (β = 0.032; p < 0.05). These findings indicate that sex and BMI may modestly influence OSI, whereas age is not a significant predictor.

As shown in Figure 2, d-ROM and OSI were significantly associated with inflammatory markers: CRP (d-ROM: R = 0.48; ** p < 0.01; OSI: R = 0.57; ** p < 0.01) and fibrinogen (d-ROM: R = 0.62; ** p < 0.01; OSI: R = 0.60; ** p < 0.01), while BAP showed negative correlations with CRP (R = −0.27; * p < 0.05) and ferritin (R = −0.25; * p < 0.05). In addition, a negative correlation between d-ROM and ferritin (R = −0.30; * p < 0.05) was observed (Figure 2). There were no significant correlations between the levels of BAP and fibrinogen and those between OSI and ferritin. From the endocrinological aspect, serum free thyroxine levels were negatively correlated with d-ROMs (R = −0.31; ** p < 0.01) and OSI (R = −0.36; ** p < 0.01), as shown in Figure 3, whereas serum cortisol levels were positively correlated with d-ROMs (R = 0.29; * p < 0.05). On the other hand, BAP levels were not significantly correlated with serum levels of free thyroxine and cortisol (Figure 3).

As shown in Figure 4, the clinical associations between the levels of oxidation markers and clinical characteristics were examined. Among the symptoms related to long COVID, the six most frequently reported were fatigue (75%), brain fog (57%), headache (22%), insomnia (21%), dizziness (18%), and depression (17%) (Figure 4A). As other minor symptoms, fever, dysgeusia, memory disturbance, dyspnea, and dysosmia were also included. To assess the utility of oxidative stress markers in evaluating these symptoms, OSI levels were compared between patients with and without each symptom listed in Figure 4A. As a result, only the patients' group with brain fog exhibited significantly higher OSI levels than those without brain fog (median: 2.2 vs. 1.8, * p < 0.05) (Figure 4B). On the contrary, there were no significant differences in OSI data between patients with and without each representative symptoms: fatigue (median: 2.1 vs. 2.0), headache (2.3 vs. 2.0), insomnia (2.3 vs. 2.0), dizziness (2.1 vs. 2.0), and depression (2.0 vs. 2.0) (Figure 4B).

Finally, to examine sex-specific associations, we compared oxidative markers with or without brain fog in female and male patients. Of note, in female patients, d-ROM and OSI levels were significantly higher in those with brain fog than in those without brain fog (median d-ROM: 625.6 vs. 513.0, ** p < 0.01; BAP: 2329.6 vs. 2498.8; OSI: 2.4 vs. 2.0, ** p < 0.01), whereas no significant difference was observed between the groups with and without brain fog in male patients (median: d-ROM: 480.9 vs. 456.6; BAP: 2363.6 vs. 2350.7; OSI: 1.8 vs. 1.8) (Figure 5). These findings suggest that oxidative stress markers may be useful for identifying or predicting brain fog in long COVID, particularly in a sex-dependent manner.

As shown in Figure 6, ROC curve analysis of OSI levels was performed to compare our patients with long COVID to healthy controls. Figure 6A demonstrates an area under the curve (AUC) of 0.99 (95% Confidence Interval, 0.98–1.00), indicating a high discriminative ability for the diagnosis of long COVID. Using Youden's J statistic, the optimal cut-off value for OSI was 1.32, which yielded a sensitivity of 0.96 and a specificity of 0.97. Moreover, as shown in Figure 6B, ROC analysis was performed to discriminate patients with brain fog among those with long COVID, in which the AUC was 0.64 (95% CI, 0.51–0.76), and an optimal cut-off value of OSI was 1.92, yielding a sensitivity of 0.70 and a specificity of 0.61.

In this study, we evaluated the clinical utility of oxidative stress markers in patients with long COVID, particularly those infected during the Omicron phase. Serum levels of d-ROMs and OSI, but not BAP, were elevated in patients with long COVID compared with reference values derived from healthy individuals, with d-ROMs and OSI being notably higher in female patients with long COVID. Oxidative parameters were associated with age, BMI, inflammatory factors, and endocrine functions. Oxidative stress positively correlated with adrenocortical function and negatively correlated with thyroid function. Additionally, brain fog was significantly associated with elevated oxidative stress levels, especially in female patients with long COVID. Furthermore, ROC curve analysis demonstrated discriminative abilities for diagnosing long COVID and for identifying patients with brain fog among the patients with long COVID.

Fatigue, a major feature of long COVID, is a significant symptom with 46% of patients complaining of fatigue lasting for several months, and its pathophysiology has yet to be elucidated [22]. A more detailed characterization of post-COVID-19 fatigue requires multiple clinical and laboratory approaches to identify physical damage and mental health conditions. In addition to widely used screening tools for symptom differentiation [11,22], it has been revealed that oxidative stress markers might help distinguish acute, subacute, and resting fatigue in healthy individuals by comparing d-ROMs and BAP levels [30]. Furthermore, combining d-ROMs and BAP with OSI has proven useful in differentiating CFS from other fatigue-related conditions [30]. In that study, patients with CFS tended to exhibit higher d-ROM and OSI values and lower BAP values at rest than age- and sex-matched controls [30].

Although differentiating the causes of post-COVID-19 fatigue remains a clinical challenge, the measurement of d-ROMs and BAP may aid in distinguishing among various fatigue-related conditions in patients with long COVID. Our findings extend this application by demonstrating that these markers can also reflect brain fatigue, as evidenced by elevated d-ROMs and OSI in patients with long COVID experiencing brain fog. According to the Italian cohort study, the risk and prevalence of long COVID during the Omicron phase are highly affected by viral shedding from infections and vaccination status [11]. However, there were no significant differences in the levels of oxidative markers in our present long COVID patients with or without vaccinations.

In the present study, serum levels of d-ROMs and OSI were comparably higher than reference data from the healthy population, while the levels of BAP were moderately low compared to the healthy controls. In an earlier study of 312 healthy subjects, oxidative stress levels detected by OSI increased significantly with age [32], and the OSI levels were significantly higher in women compared to men [31]. The same trend was observed in the present study on patients with long COVID. Of note, serum BAP levels were found to be significantly decreased with age in patients with long COVID, suggesting the age-dependent impairment of antioxidative capacity in patients with long COVID.

One of the hallmarks of unexplained fatigue is ME/CFS, which includes symptoms such as pain, depression, and neurocognitive dysfunction. Although the etiology remains unclear, inflammation and oxidative/nitrosative stress are suspected contributors [39]. Nearly half of patients with long COVID meet the diagnostic criteria for ME/CFS [40,41], although the prevalence rates vary depending on the criteria applied. Our previous study showed that 8.4% of patients met the ME/CFS criteria [15], with a lower prevalence observed during the Omicron phase compared with earlier variants [42].

However, patients with long COVID during the Omicron phase reported brain fog more frequently [15], suggesting a variant-specific neurological effect [43]. This may involve direct viral entry into the central nervous system via angiotensin-converting enzyme 2 (ACE2) receptors in the alveolar epithelium, leading to disruption of the blood–brain barrier (BBB) and inflammation in the periventricular and choroid plexus regions [44,45,46]. These pathophysiological processes may explain the elevated oxidative stress and reduced antioxidant defenses observed in our study.

Recent investigations further provide indirect evidence that pathological fatigue is a neurological disorder, involving diminished muscular force and sensory attenuation [47]. Several studies support a relationship between brain fog and oxidative stress [48]. After SARS-CoV-2 infection, endothelial cells release pro-inflammatory cytokines and chemokines, which initiate cytokine storms and elevate reactive oxygen species (ROS) [49,50]. This leads to calcium imbalance and increased BBB permeability [51,52]. Viral binding to ACE2 receptors further disrupts renin-angiotensin homeostasis and enhances ROS production via NADPH oxidase [53]. Microglial activation and interleukin-1 up-regulation exacerbate oxidative damage, phagocytosis, and neural apoptosis [54]. These mechanisms may underlie vascular leakage and neuroinflammation [55], contributing to brain fog.

Endocrine functions may also affect the endogenous oxidative stress [56]. We previously reported that depression and fatigue conditions were correlated with adrenocortical and thyroid functions in patients with long COVID [27], in which the severity of infection and immunoreactions were associated with changes in HPA activity and thyroidal function [27]. These findings suggest that endocrine dysfunction is likely to be involved in the persistence of long COVID symptoms. Of interest, hypothyroidism has been linked to increased ROS, oxidative stress, and lipid peroxidation [57,58], while hyperthyroidism also shifts the pro-oxidant/antioxidant balance towards oxidative damage [56]. Our present results demonstrated that serum thyroxine levels inversely correlated with d-ROMs and OSI, suggesting a link between hypothyroidism and oxidative stress in long COVID. On the other hand, glucocorticoids (GCs) are known to promote neuronal oxidative stress by enhancing mitochondrial respiration and ROS generation [59]. Chronic GC excess is associated with various ROS-related pathologies, including myopathy, osteoporosis, diabetes, and hypertension [59]. The brain is particularly vulnerable to GC-induced oxidative damage owing to its high metabolic activity, low antioxidant capacity, and abundance of GC receptors [60], which may contribute to the linkage between oxidative stress and neurological symptoms in a state of hypercortisolemia. Although cortisol is known as a representative stress hormone, our study found that cortisol levels correlated with oxidative markers, indicating that hypercortisolemia may exacerbate endogenous oxidative stress in long COVID. Thus, our present results suggest that thyroid and adrenal dysfunctions exert opposing effects on redox balance in patients with long COVID.

This study has several limitations. First, it was a single-center, retrospective study conducted in Japan. As most participants were referred cases, patients with more severe long COVID symptoms may have been over-represented. Moreover, because this study was based in an outpatient clinic, individuals experiencing prolonged symptoms after severe acute COVID-19 may also have been over-represented. Information regarding the type and batch of mRNA vaccines administered was unavailable. Therefore, multicenter prospective studies are needed to validate these findings. Second, patient characteristics and comorbidities were not analyzed in detail. In addition, covariate data such as age, sex, and BMI were not available for the healthy control group, which made it impossible to perform fully adjusted multivariable analyses across patients and controls. We therefore conducted additional multivariable regression analyses within the patient group to account for patient background factors, but residual confounding cannot be fully excluded. Third, the timing and handling of blood samples varied and were not standardized across participants. Fourth, the cutoff values were determined in the laboratory based on previous reports by the co-authors, and no universally accepted standards have been established. Fifth, patients with non-specific symptoms showed OSI levels nearly overlapping with those without such symptoms, raising the possibility of false positives if OSI is used alone. These limitations should be addressed in future studies employing an appropriate study design.

In conclusion, oxidative stress markers, such as d-ROMs, BAP, and OSI, may serve as useful indicators for detecting or predicting neurologic symptoms of long COVID, particularly brain fog, in a gender-dependent manner. These markers may also be useful for assessing the pathophysiology and the severity of long COVID, especially when combined with inflammatory markers and endocrine data. Further studies are warranted to elucidate the oxidative stress–related pathophysiology underlying long COVID.