No significant benefit of moderate-dose vitamin C on severe COVID-19 cases

2.1 Study design

Data of all patients with COVID-19 consecutively admitted to the Cancer Centre, a designated clinical center for COVID-19 of the Union Hospital of Huazhong University of Science and Technology (Wuhan, China) between February 13 and February 29, 2020, were collected. Patients aged ≥18 years and diagnosed as severe COVID-19 were screened for this study. Exclusion criteria were (a) the duration of hospitalization was less than 3 days; (b) vitamin C treatment started before admission; and (c) the length of vitamin C use was less than 3 days. Definite outcomes (discharge or death) were followed up until March 15, 2020.

All patients in this study were diagnosed according to the Guidelines of the Diagnosis and Treatment of Novel Coronavirus Pneumonia released by the China NHC [10]. The definition of severe COVID-19 was patients who met any of the following criteria: (a) demonstrating shortness of breath with a respiratory rate ≥30 beats/min; (b) pulse oxygen saturation (SpO₂) ≤93% in resting state; (c) arterial blood oxygen partial pressure (PaO₂)/oxygen concentration (FiO₂) ≤300 mm Hg; (d) pulmonary imaging demonstrating significant >50% increase of lesions within 24–48 h.

Data on demographic characteristics, clinical features, comorbidities, laboratory findings, medications, and outcomes were obtained from the patients’ electronic medical records and evaluated by two trained physicians (CXG and JZL) independently. This study was approved by the Ethics Committee of Tongji Medical College of Huazhong University of Science and Technology in Wuhan (2020-0058). Written informed consent was waived due to the rapid emergence of this infectious disease.

2.2 Vitamin C exposure

Patients treated with intravenous vitamin C (2–4 g/day) at study baseline or received it during follow-up before death or discharge were defined as vitamin C exposure. The study baseline for each patient was 24 h after admission. As there was no powerful evidence for vitamin C use in patients with COVID-19, the duration of vitamin C treatment was not standardized and was determined by the clinicians according to the condition of each patient.

2.3 Outcomes

The primary outcome was all-cause death during the follow-up period. The secondary outcome was clinical improvement assessed at the end of follow-up. Clinical improvement was defined as live discharge from the hospital, or a reduction of at least 2 points from baseline on a 6-point ordinal scale [11], whichever occurred first.

The 6-point ordinal scale was as follow: (1) discharge; (2) hospitalized, breathing ambient air without supplemental oxygen; (3) hospitalized, receiving low-flow supplemental oxygen; (4) hospitalized, requiring noninvasive mechanical ventilation or high-flow oxygen; (5) hospitalized, requiring invasive ventilation or extracorporeal membrane oxygenation (ECMO); and (6) death [11].

2.4 Statistical analysis

Discrete variables were presented as frequency and proportion, and the continuous were described using median and interquartile range (IQR). First, we applied univariate Cox proportional-hazards regression models to evaluate the association between vitamin C use and outcomes. For controlling the potential confounders, two propensity-score based strategies were adopted, one of which is IPTW and another is 1:1 propensity-score matching. The propensity score concerning the assignment of vitamin C was modelled by a non-parsimonious multivariable logistic regression incorporating patients’ demographics, comorbidities, laboratory testing, and medications at study baseline. We assessed the distribution of the weights as well as the propensity scores and then conducted the covariate balance diagnostics using standardized differences with the threshold of 0.25 [12]. In IPTW analysis, we adjusted the weights of samples by basic stabilized inverse probability treatment weights without any truncation. The weighted hazard ratios (HR) were estimated using the Cox model, and Kaplan–Meier curves were generated to compare survival rates of the patients in different groups. In 1:1 matching analysis, we matched the patients using the nearest neighbor matching with the caliper of 0.2, after which we clustered every single pair in the Cox model to improve the statistical power.

We tested the proportional assumption of the Cox model based on Schoenfeld residuals against the transformed time and found no evidence of violation during the analysis. For paired survival analysis, the robust variance was estimated. Multiple imputations were used for the missing data at the study baseline, and the coefficients and standard errors were all estimated based on Rubin’s rules. All statistical analyses were performed with the software R version 4.0.2 (https://www.r-project.org/).

3.1 Characteristics of the patients

Among 955 patients with COVID-19 who were admitted consecutively to the hospital between February 13 and February 29, 2020, 427 were diagnosed as severe type on admission. Of the 427 patients, among which 30 were excluded from the study (5 started vitamin C use before admission, 21 were treated with vitamin C for less than 3 days, and 4 had a length of stay in hospital of less than 3 days) (Figure 1).

Of the 397 patients, in addition to standard therapies, 70 (17.6%) received intravenous vitamin C (2–4 g/day) and 327 (82.4%) did not. The median duration of vitamin C treatment was 8.8 days (interquartile range [IQR], 5.1–16.0 days). The distribution of duration of vitamin C use is shown in Figure A1. Among the 70 patients who were exposed to vitamin C, 33 (47.1%) received it within 48 h of admission. The timing of the initiation of vitamin C use after admission is provided in Figure A2. The baseline demographic and clinical characteristics of the patients with severe COVID-19 according to vitamin C exposure are shown in Table 1.

In the unadjusted population, the median age of the patients was 67.0 (IQR, 61.0–74.0) years, and 207 (52.1%) were men. Vitamin C-treated patients had a higher level of C-reactive protein at study baseline than the patients who did not receive it. Interferon α-2b was administered to 35.7% of the patients in the vitamin C treatment group versus 10.4% in the control group, and low molecular weight heparin was given to 32.9% versus 16.2%, respectively (Table 1).

Figure A3 shows the distribution of estimated propensity scores for the assignment of intravenous vitamin C among the severe COVID-19 population before and after matching. After matching, there is a balanced propensity score distribution between the exposed and unexposed groups. Particularly, the absolute standardized mean difference (aSMD) of the variables incorporated in the propensity score model indicated that it is of comparability between two groups (aSMD <0.25) (Figure A4). Figure A5 displays the distribution of weights for IPTW analysis on a log scale. Particularly, the mean of stabilized weighting is approximately 1, indicating that we did not observe the obvious weighting discrepancy from the theoretical framework after standardization. Also, the maximum of the weighting is less than 10 and the minimum of the weighting is greater than 0.1, which shows us less risk of positivity assumption violation. Hence, we did not observe the distinct evidence indicating the absence of the IPTW assumption violation. Figure A6 also reflects that the two groups are comparable based on the threshold of 0.25 for aSMD.

Figure 1

Flow chart of study participants in the study cohort. The study baseline was defined as 24 h after admission. COVID-19 denotes coronavirus disease 2019.

3.2 Primary outcome

Over a median follow-up of 29.3 days (IQR, 28.5–30.1 days), the primary outcome developed in 19 patients (4.8%) of the 397 enrolled in our study. At the end of follow-up on 15 March 2020, 360 (90.7%) patients were discharged alive, and 18 (4.5%) were still in the hospital.

Among 397 patients in two groups, in contrast to 0.206 deaths/person-months in the population receiving intravenous vitamin C, those not receiving vitamin C is 0.02 deaths/person-months only. In the crude analysis, according to the Kaplan–Meier curves in the left panel of Figure 2, it seems to be a significant difference between the two groups (HR, 8.64; 95% confidence interval [CI], 3.40–21.94). However, after adjusting potential confounders, IPTW analysis (Figure 2, right panel) shows the weighted HR and its 95% CI as 2.69 (0.91–7.89). A similar non-significant result could also be found in the 1:1 matching analysis. After matching, the paired Cox model indicates that the HR is 2.57 (95% CI, 0.39–16.79) (Table 2), without significance.

Figure 2

Kaplan–Meier curves for overall survival during follow-up before and after inverse probability of treatment weighting (IPTW) analysis. On unadjusted analysis in the left panel, patients treated with vitamin C had higher risk of death than those who did not receive it (HR, 8.64; 95% CI, 3.40–21.94), while there was no significant association between vitamin C use and death on IPTW analysis in the right panel (weighted HR, 2.69; 95% CI, 0.91–7.89). The shaded areas represent 95% CI.

3.3 Secondary outcome

By 15 March 2020, clinical outcomes of the 397 patients according to the 6-point ordinal scale are shown in Table A1. In the unadjusted analysis, the cumulative incidence of clinical improvement was 74.3% (52/70) among patients who received vitamin C, and 95.1% (311/327) among patients who did not receive it. Clinical improvement was less frequent among patients exposed to vitamin C than among those who were not exposed to it (HR, 0.72; 95% CI, 0.53–0.96) (Figure 3, left panel). However, there was no significant difference in the incidence of clinical improvement between two groups on IPTW analysis (weighted HR, 0.76; 95% CI, 0.55–1.07) (Figure 3, right panel) and on propensity-score matching analysis (HR, 0.74; 95% CI, 0.48–1.14) (Table 2).

Figure 3

Kaplan–Meier curves for clinical improvement at the end of follow-up before and after inverse probability of treatment weighting (IPTW) analysis. On unadjusted analysis in the left panel, clinical improvement was less frequent among patients exposed to vitamin C than among those who were not exposed to it (HR, 0.72; 95% CI, 0.53–0.96). But no significant difference was observed between the two groups on IPTW analysis in the right panel (weighted HR, 0.76; 95% CI, 0.55–1.07). The shaded areas represent 95% CI.

3.4 Safety

During the follow-up period, adverse events which may be caused by vitamin C such as abdominal pain, diarrhea, or nausea [13] did not occur in patients with the use of vitamin C.