We recruited 119 symptomatic adult inpatients (76 male, 43 female, mean age 59.3 ± 16.7 years) with mental disorders treated for COVID-19 at Alekseev Psychiatric Clinical Hospital no. 1 (Moscow, Russia) in June–July 2020; patients were at hospital within the treatment course. The inclusion criteria were as follows: (Mehta et al., 2020): at least 18 years of age; (Dubina, 2022); confirmed COVID-19 diagnosis (a positive test for SARS-CoV-2 RNA detected by RT-PCR collected from the upper respiratory tract; or pulmonary radiological data specific for COVID-19; or antibodies ratio specific for acute viral infection). Exclusion criteria: (Mehta et al., 2020): known or suspected active viral, bacterial, mycobacterial, or fungal infection other than COVID-19, including Epstein-Barr virus, cytomegalovirus, herpesvirus family, HIV, hepatitis C virus, etc; (Dubina, 2022); pregnancy and/or breastfeeding; (Giamarellos-Bourboulis et al., 2020); oncology diseases. The participants were assigned to experimental (50 patients) and control (69 patients) groups. All of these patients received antiviral treatment according to the national clinical guideline (Ministry of Health of Rusussian Federation, 2020), namely chloroquine (500 mg twice a day for 7 days), hydroxychloroquine (400 mg twice at the first day, then 200 mg twice at the next 6 days), lopinavir-ritonavir combination (400 mg + 100 mg p. o. every 12 h within 14 days), azithromycin (500 mg p. o, 5 days, in combination with hydroxyloroquine), and interferon preparations (IFN-β1b, 0.25 mg/ml, 8,000,000 ME, 14 days; IFN-α2b, 3,000 ME, 5 times a day, 5 days). Additionally, dexamethasone (12 mg per day p. o. or 4 mg three times a day i. v.) was used in case of CRP value growth. In addition, patients in the experimental group received the full course of riboflavin supplementation on medical advice (10 mg flavin mononucleotide intramuscularly twice a day).

Demographic, clinical, treatment, and outcome data were obtained from the electronic medical records of Alekseev Psychiatric Clinical Hospital no. 1 (Moscow, Russia). Data were anonymized by removing personally identifiable information prior to processing. Laboratory data included RBC, platelets, WBC, lymphocytes, neutrophils, and monocytes counts determined using a SYSMEX hematological analyzer (Japan); serum levels of C-reactive protein and hemoglobin; serum levels of IL-1β, IL-2, IL-6, MCP-1, TNF-α, and IFN-γ cytokines assessed using ELISA reagent kits (Vector-Best, Novosibirsk, Russia); chest computed tomography (CT) scans and their description by radiologists (unilateral and bilateral ground glass opacity, and lung involvement).

Statistical processing of the results was carried out using the GraphPad Prism software, version 6.01. The p values were estimated using the Wilcoxon Matched-Pairs Signed Ranks Test (paired, non-parametric, two-tailed) or Mann–Whitney U test (non-paired, non-parametric, two-tailed). The data are presented as median [IQR] values or box-and-whiskers plots using Tukey’s modification.

We analyzed demographic, clinical, treatment, and outcome data for 119 patients with mental disorders treated for COVID-19 at Alekseev Psychiatric Clinical Hospital no. 1 (Moscow, Russia) in June–July 2020. All of these patients received antiviral treatment according to the national clinical guideline (Ministry of Health of Rusussian Federation, 2020). We retrospectively divided this cohort to two groups: the first one additionally received 2 × 10 mg/day FMN intramuscularly (10 mg twice a day) on medical advice within 1 week; the second one did not receive FMN supplementation (Table 1).

The groups were comparable in terms of age [median (IQR) 59.00 (48.25; 68.60) for the experimental group vs 62.50 (47.00; 72.75) for the control group, p = 0.4192 in Mann–Whitney U test), CRP value on admission to the hospital [18.20 (2.570; 64.24) vs. 14.21 (3.285; 37.11), p = 0.5331], CT grade 1/CT grade 2–4 ratio (two-tailed p value of 0.0761 in Fisher’s exact test), and outcome (two-tailed p value of 0.6486 in Fisher’s exact test). The only statistically significant difference between the experimental and control groups is the male/female ratio with a predominance of men in the experimental group (two-tailed p value of 0.0214 in Fisher’s exact test).

The distribution of the study sample patients according to ICD-10 diagnoses of mental disorders is listed in Table 2. In total, most patients (n = 65; 55.1%) were diagnosed with schizophrenia spectrum disorders, F2 in ICD-10. Different types of organic mental disorders, F0 in ICD-10, accounted for 26.7% of cases. Other disorders, including affective disorders (8.8%) and intellectual disability (2.2%), were less frequent. The distribution of ICD-10 diagnoses within the FMN and control groups can also be considered comparable [F (1,3) = 9.417; p = 0.0546 in two-way ANOVA]. Based on these demographical and clinical parameters, we assume that both studied groups are similar in their demographic and clinical characteristics.

We analyzed a number of the main blood and biochemical parameters, namely RBC, platelets, WBC, lymphocytes, neutrophils, monocytes, and hemoglobin values in both groups. The correlation of these parameters with the two most clinically relevant characteristics (C-reactive protein value and lung involvement in chest computed tomography) was evaluated. It was found that neutrophils values positively correlated with the severity of the disease (with both CRP value and lung involvement), while lymphocytes values correlated with these parameters in a negative manner (Figure 1, 2). Accordingly, the neutrophils to lymphocytes ratio also correlated positively.

The only blood test parameter that correlated with CRP values, but not with the lung involvement, was the total leucocytes count. This could be explained by a better sensitivity of the CRP value to the inflammation status. Indeed, for all studied parameters the correlation with CRP values was clearer than with the lung involvement (e.g., for neutrophils r = 0.3771, p < 0.0001 vs. r = 0.2407, p = 0.004). Finally, lung involvement positively correlated with CRP values (r = 0.2375, p = 0.0076). We can conclude that in our cohort we had four parameters that correlated with the severity of the disease (in addition to well-established clinically relevant lung involvement and CRP values), namely total leucocytes, neutrophils, lymphocytes, and NLR, while RBC, platelets, monocytes and hemoglobin did not correlate with COVID-19 progression. This is in agreement with the literature data: neutrophils and lymphocytes counts, as well as their ratio, were the parameters that predicted the severity and outcome of the COVID-19 infection (Li et al., 2020; Yang et al., 2020; Imran et al., 2021).

The median values of the main blood and biochemical parameters, namely RBC, platelets, WBC, lymphocytes, neutrophils, monocytes, hemoglobin, and CRP, were estimated and compared in each group (day 1 vs. day 7) and in each time point (experimental group vs. control group), and summarized in Table 3. Neutrophils and lymphocytes counts, as well as their ratio, provide the most promising data for the correlation between the COVID-19 course and FMN supplementation. Neutrophils decreased in both cohorts, but only in the experimental group was this decrease statistically significant in the Mann–Whitney U test [67.00 (63.00; 73.00) to 63.00 (57.00; 69.00), p = 0.0406 vs. 69.00 (58.75; 72.25) to 66.00 (58.50; 71.50), p = 0.7904]. Lymphocytes also increased in both groups, but statistical significance has been found for the FMN group only [21.00 (19.00; 27.00) to 27.00 (20.00; 33.00), p = 0.0262 vs. 21.50 (19.00; 30.25) to 24.00 (19.00; 31.50), p = 0.5651]. Finally, the NLR demonstrated a similar pattern for the experimental and control groups [3.190 (2.267; 3.789) to 2.296 (1.781; 3.190), p = 0.0152 vs. 3.117 (1.935; 3.798) to 2.680 (1.888; 3.789), p = 0.6224].

Surprisingly, we have found a statistically significant increase in the platelets count [224.0 (182.0; 284) to 275.0 (207.0; 341.0), p = 0.0363] and a statistically significant decrease in the RBC count [4.500 (4.210; 4.840) to 4.380 (3.950; 4.630), p = 0.0406] and hemoglobin value [134.0 (120.0; 145.0) to 124.0 (118.0; 134.0), p = 0.0420]. None of these parameters changed significantly in the control group and none of them correlated with COVID-19 severity in our cohort. Although thrombocytopenia (platelets count <150×10⁹ cells per liter) is one of the COVID-19 symptoms (Mei et al., 2020; Zong et al., 2021), we had only seven patients that demonstrated thrombocytopenia within the COVID-19 course. Therefore, we split the normal platelets range (150–450×10⁹ cells per liter) into two parts and evaluated groups with lower (<300×10⁹ cells per liter) and higher (>300×10⁹ cells per liter) platelet counts separately (Supplementary Data S1). In fact, platelets in the lower (<300×10⁹ platelets per liter) FMN group increased from 205.5 [178.5; 240.3] to 272.0 [197.3; 327.3], p = 0.0014, while in the higher (>300×10⁹ platelets per liter) FMN group they decreased from 336.0 [325.0; 353.0] to 291.0 [217.0; 352.0], p = 0.1657), so median values shifted to the center of normal range.

Moreover, we found a decrease in both hemoglobin and RBC count in the experimental group, but not in the control group. Since hemoglobin and RBC values did not correlate with the severity of COVID-19 (CRP values or lung involvement), we assume that this decrease could be explained with fractional eryptosis occurring via visible light irradiation of the patients’ skin. It is known that the light illumination of red blood cells incubated with riboflavin results in partial cell death (Qadri et al., 2017). Fortunately, both median RBC and hemoglobin changed slightly in absolute values and generally remained in the normal range (Supplementary Data S2). Moreover, riboflavin plays an important role in erythropoiesis, as it improves iron absorption and ferritin mobilization (Suwannasom et al., 2020), so we can assume an increase in the RBC count in the FMN group soon.

We additionally picked out patients with advanced severity of COVID-19 (CRP value >10 mg/l and/or CT grade 2–4 at any time of treatment) and evaluated the median values of all studied biochemical parameters in this sub-cohort (Table 4). The obtained data were similar to Table 3 with statistically significant changes for RBC, platelets, neutrophils, lymphocytes, NLR, and hemoglobin in the experimental group, but not in the control group. However, we additionally found statistically significant differences between the experimental and control groups on day 7 for neutrophils and NLR. On day 1 median values of neutrophils were 68.00 [63.00; 79.00] in the experimental group and 70.00 [63.75; 76.00] in the control group (p = 0.7504), while on day 7 median values were 64.00 [55.00; 71.00] and 68.00 [63.00; 77.00], respectively (p = 0.0317). Similarly, median NLR on day 1 was 3.342 [2.222; 4.139] in the experimental group and 3.600 [2.680; 4.676] in the control group (p = 0.2834), while on day 7 median NLR was 2.240 [1.610; 3.300] and 3.000 [2.207; 4.588], respectively (p = 0.0284). We assume this result is direct evidence of the anti-inflammatory efficacy of high doses of riboflavin in COVID-19 patients.

Mental disorders vary greatly in their characteristics, including behavior, pathogenesis, and even biochemical markers (García-Gutiérrez et al., 2020). Based on this knowledge, we can assume alterations in the course of the disease in patients with different psychiatric diagnoses. Our cohort included two main types of diagnosis, namely organic mental spectrum disorders (F0, n = 41) and schizophrenia spectrum disorders (F2, n = 65), while affective disorders (F3, n = 11) and oligophrenia (F7, n = 2) were minor fractions. Here we compared main blood and biochemical parameters in patients with organic mental disorders and schizophrenia. The median values are summarized in Table 5. Indeed, F0 patients demonstrated higher inflammation than F2 patients, which was expressed in higher neutrophils [72.00 (66.00; 79.25) vs. 67.00 (60.75; 72.00), p = 0.0026], lower lymphocytes [19.00 (13.50; 25.25) vs. 23.35 (19.00; 31.00), p = 0.0029], and a higher NLR ratio [3.642 (2.703; 5.923) vs. 2.867 (2.006; 3.640), 0.0020]. In addition, the median CRP value was also higher in the F0 subgroup [24.30 (5.750; 63.14) vs. 16.20 (1.900; 40.74)], although this difference was not significant (p = 0.2111). Higher inflammation in the F0 subgroup can be explained by background neuroinflammation, which is one of the central mechanisms in dementia or Alzheimer’s disease (Kinney et al., 2018; Hur et al., 2020). However, schizophrenia is also often discussed in terms of inflammatory biomarkers (Müller et al., 2015; Miller and Goldsmith, 2019), so this issue needs further evaluation. Additionally, median RBC value was lower in F0 subgroup in comparison to F2 [4.320 (3.620; 4.640) vs. 4.540 (4.218; 4.860), 0.0072]. Anemia conditions are often found in chronic psychiatry patients (Korkmaz et al., 2015), so we can suppose mild anemia in F0 subgroup. The other possible explanation is statistical bias since absolute values are rather close.

A panel of cytokines, including IL-1β, IL-2, IL-6, TNF-α, МРС-1, and IFN-γ, was also monitored over the hospital stay for a limited number of patients that were recruited to the trial at a late stage (July 2020). All of these cytokines have been earlier studied for clinical relevance in COVID-19. E.g., IL-6 has been previously shown as a predictor for respiratory failure (Jøntvedt Jørgensen et al., 2020) and disease outcome (Del Valle et al., 2020). TNF-α serum levels have been described as an independent and significant predictor of disease severity and death (Del Valle et al., 2020). Studies involving a larger cohort of severe COVID-19 patients showed the importance of IL-2 levels (Shi et al., 2020), while IL-1β has been demonstrated as a low predictive value marker due to its minor expression in patients (Del Valle et al., 2020). In the current research, IL-6 measurements showed the most promising results, decreasing in the experimental group and growing in control group (Table 6). The limited number of patients with cytokine data resulted in a lack of statistical significance, but in line with absolute values changes, we assume IL-6 as the most interesting marker for further research; TNF-α, МСP-1, and МСP-1 as of moderate interest; and IL-2 and IL1β as of low interest.