3.1. Description of the Trials
We identified 18 publications, which reported 19 trials on the effect of vitamin C on the length of ICU stay or on the duration of mechanical ventilation (Figure 1
; Table 1
). The trials are summarized in Table S1 of Supplementary file S1
. One trial was excluded because of severe violation of the ITT principle [123
]; thus, 18 trials remained for our analyses.
Length of ICU stay was studied in 17 trials, and duration of mechanical ventilation in six trials. Patients undergoing cardiac surgery were studied in 13 trials, patients with sepsis in two trials, lung contusion patients in two trials, and burns patients in one trial.
Eight trials were carried out in Iran, four in the USA, two in Egypt, and one trial in each of China, Greece, Japan, and Slovenia. The total number of patients in the 18 trials was 2004, with 1835 patients in studies related to cardiac surgery and 169 patients in other settings. The mean age in the trials ranged from 40 to 70 years (median 60 years). Of the trials that reported sex distribution, the proportion of males varied from 57% to 79% (median 70%) (Table S1
Vitamin C was administered orally in seven trials and intravenously in 11 trials. All oral administration trials investigated cardiac patients and the dose ranged from 1 to 3 g/day. The 11 intravenous administration trials include all five non-cardiac studies and the dose ranged from 0.5 to 110 g/day. Three trials administered vitamin C on one single day only, and 11 administered vitamin C for four days or longer (Table 1
and Table S1
There was substantial variation in patient selection between the trials (Table 1
). The cardiac trials enrolled mostly patients undergoing elective surgery and so the patients were in reasonably good health and the length of ICU stay was short; one trial enrolled both elective and urgent cardiac surgery patients [112
]. In nine cardiac surgery trials the duration of ICU stay was two days or less in the control group, whereas in four trials the stay was three–four days. In the two lung contusion trials, the mean ICU stay was five days in the control groups. The longest reported ICU stay was in the two trials that enrolled patients with sepsis, with ICU stay of 11 days [115
] and 20 days [117
]. The trial with burns patients did not report the duration of ICU stay, but the duration of mechanical ventilation was 21 days in the control group [107
], which is about 10 times longer than the ICU stay in the majority of cardiac surgery trials. Thus, ICU stay was evidently the longest in the trial with the burns patients [107
The majority of the trials were randomized (Table S1
). Two trials used quasi-randomization. Tanaka et al. allocated patients to study groups by the month of admission [107
] and Mirmohammadsadeghi allocated patients by their person numbers [121
]. The method of allocation was not reported by Ebade et al. [114
] and by Alsfahey [118
]. Most trials reported no or few drop-outs (Table S1
The risk of bias assessment of the trials is shown in Figure 2
. The justifications for the quality assessments are shown in Table S1
. An explicit placebo was used in nine of the trials [107
]. Nevertheless, ICU patients receive a large number of treatments, and thus, it is highly unlikely that receiving one additional tablet or infusion would produce a placebo effect that could influence the length of ICU stay. Therefore, we did not consider that the lack of a placebo undermines the validity of the comparisons in those trials that did not have placebo. In 14 trials we concluded that there was no reason to assume that blinding at the stage of allocation, in the treatment phase, or in the outcome assessment might have led to bias in the findings.
Four trials did not report the methods in sufficient detail to enable conclusions to be drawn about the possibility of bias caused by poor blinding [106
]. In our sensitivity analyses for vitamin C effect on ICU stay, we excluded the two particularly poorly reported trials [114
], and a quasi-randomized trial [121
Although plasma vitamin C levels at baseline and during treatment are highly informative characteristics of studies on vitamin C, only one trial reported plasma levels at the baseline. Fowler reported that at enrollment the mean plasma vitamin C level of their patients was 18 μmol/L [115
]. Plasma vitamin C level in the placebo group fell from 20.2 μmol/L at entry to 15.6 μmol/L on study day four. In contrast, on day four, the plasma vitamin C in the 3.5 g/day group had increased to 330 μmol/L, and in the 14 g/day group to 3080 μmol/L [115
]. Tanaka published plasma vitamin C levels during the study in a figure but did not provide numbers. With the intravenous administration of 66 mg/hour/kg over a 24 h treatment period, the plasma vitamin C level increased to about 500 μmol/L in the vitamin C group, but remained close to zero in the placebo group [107
3.2. Effect of Vitamin C Administration on the Length of ICU Stay
In our meta-analyses we pooled the results of the trials on the relative scale, calculating the ratio of means (RoM). For small effects, it is useful to transform the RoM values to a percentage scale. For example, RoM = 0.922 corresponds to a 7.8% mean decrease in the length of ICU stay.
Over the 17 trials reporting on length of ICU stay, there is highly significant heterogeneity in the effect of vitamin C with I2
= 90% (p
) (Table 2
and Figure S1 in Supplementary file S1
). The singular reason for this very high heterogeneity is the oldest trial by Dingchao (1994) that was carried out in China [106
]. When it was excluded, the level of heterogeneity decreased dramatically to I2
= 58% (p
= 0.002). In addition, when the Abdoulhossein (2018) trial with patients administered vitamin E [122
] was also excluded, the level of heterogeneity decreased further to I2
= 35% (p
= 0.088). Finally, three trials were very small [109
] (Table 1
), with a total of just 76 patients; with each of them having weight <1% in the meta-analysis that included all trials in Figure S1
. These trials did not contribute to the estimate of effect, but they inflated the degrees of freedom for the test of heterogeneity (Table 2
). Thus, their exclusion had no influence on the point estimate or the p
-value, but moderately increased heterogeneity. We also excluded these three trials, and our final meta-analysis is shown in Figure 3
. The excluded five trials are discussed separately below and the forest plot of all 17 trials is shown in Figure S1
In the final meta-analysis of 12 trials in Figure 3
, vitamin C shortened the duration of ICU stay on average by 7.8% (95% CI: 4.2% to 11.2%; p
= 0.00003). There are substantial variations in the trials of the final meta-analysis, and moderate statistical heterogeneity over the 12 studies with I2
= 43%. Nevertheless, the confidence intervals of all the 12 trials are consistent with the pooled effect.
We calculated the estimate of effect for selected subgroups (Table 2
). In a sensitivity analysis, we excluded three trials; however, the estimate of vitamin C effect on ICU stay was essentially the same in the nine better-quality trials compared with the final meta-analysis of the 12 trials: 7.3% versus 7.8%, respectively. Eleven of the 12 trials investigated ICU stay associated with cardiac surgery. Six trials administered vitamin C orally and six administered intravenously (Figure 3
). Six of the trials were carried out in Iran, and six outside of Iran. For each of these subgroup selections, the estimate of effect ranged from a statistically significant 7% to 8% reduction in the length of ICU stay.
All of the six oral administration trials used vitamin C doses of 1 to 3 g/day. Overall, in these low dose oral administration studies, vitamin C reduced the length of ICU stay by 8.6% (95% CI: 3.0% to 14.0%; p = 0.003). For these six trials, the weighted mean dose of vitamin C was 2.0 g/day. Thus, the effective dose of 2.0 g/day was associated with the 8.6% reduction in ICU stay.
In seven trials with control group ICU stay from 1 to 2 days, corresponding to less sick patients, vitamin C reduced ICU stay by 5.7% (p
= 0.027). In five trials with control group ICU stay from 3 to 5 days, corresponding to sicker patients, vitamin C reduced ICU stay by 10.1% (p
= 0.0001). However, the confidence intervals for these groups were substantially overlapping (Table 2
), and the subgroup comparison gives p
= 0.21 (see Supplementary file S1
). We did not carry out meta-regression by the dose of vitamin C, since few trials used doses over 3 g/day (Table 1
). In addition, we did not carry out meta-regression by the proportion of males, since the range was very narrow in the studies (57% to 79% males) (Table S1
Five of the 17 trials that reported length of ICU stay were excluded from the meta-analysis in Figure 3
. The Dingchao trial is the oldest, carried out in China in the early 1990s [106
]. They reported that intravenous administration of 17 g/day of vitamin C for one single day reduced the length of ICU stay by 44% (95% CI 40% to 49%). On the basis of the comparison of the confidence intervals, and the heterogeneity test, the Dingchao trial is fundamentally inconsistent with all the trials shown in Figure 3
. The methods of the Dingchao trial are poorly reported; however, we do not consider that there is reasonable justification to exclude it. However, even if the study findings were valid in the context of China in the 1990s, the results cannot be generalized to the contexts examined in the more recent trials. All the more recent trials found the effect of vitamin C to be much smaller (Figure 3
). Nevertheless, the Dingchao study supports the concept that vitamin C can influence ICU stay.
] reported a 2 × 2 factorial trial with vitamin C and vitamin E. There was a statistically significant interaction between the two vitamins (Table 3
). Intravenous 0.5 g/day vitamin C shortened the duration of ICU stay only in patients who were simultaneously administered 1 g/day vitamin E. In Figure 3
, we excluded the vitamin E patients of that study, because this considerably decreased the heterogeneity of the final meta-analysis (Table 2
). Nevertheless, the benefit of vitamin C in the patients administered vitamin E supports the concept that vitamin C can influence ICU stay in some contexts.
Three particularly small trials [109
] were excluded from Figure 3
since they had weight <1% in the meta-analysis of the 17 comparisons. They had such wide 95% CIs that they were uninformative about any possible effects of vitamin C on the length of ICU stay, see Figure S1
3.4. Other Effects in the Included Trials
Our primary and secondary outcomes were the length of ICU stay and the duration of mechanical ventilation. Additionally, we collected data on other outcomes relevant in the ICU context; however, there are few and they are so heterogeneous that further meta-analyses were not feasible.
Two small trials with sepsis patients were not informative about ICU stay and were excluded from Figure 3
], though another of the sepsis trials contributed to the analysis of mechanical ventilation (Figure 4
). Nevertheless, certain of their other outcomes are relevant for the ICU context.
] reported that 3.5 and 14 g/day of vitamin C for 4 days significantly reduced multi-organ failure scores, and the rate of decline with the higher dose was twice the rate with the lower dose. They also reported decreased inflammation markers C-reactive protein and procalcitonin, and the effects of the higher dose appeared greater also on these outcomes.
In another trial with sepsis patients, Zabet [117
] reported that 7 g/day of vitamin C for 3 days significantly reduced the mean daily dose and mean duration of norepinephrine administration (p
< 0.01 for both). Mortality over 28 days in the vitamin C group was reduced by RR = 0.22 (95% CI 0.06–0.85; p
= 0.018 Fisher exact test) based on 2/14 deaths in the vitamin C group and 9/14 in the placebo group.
In the earliest study, Dingchao [106
] reported that 17 g vitamin C on a single day decreased CK-MB levels and cardiac index during follow-up.
] reported that in their trial with patients with severe burns, the requirement for fluid infusions in the vitamin C group was decreased by 45% (p
< 0.01). In addition, fluid retention was decreased by 68% (p
< 0.01) and weight gain was 48% lower (p
< 0.01) in the vitamin C group.
] reported that the combination of vitamin C and E increased oxygen saturation and decreased carbon dioxide levels.