Table 2.

Meta-analyses of all included trials and selected subgroups on the length of ICU stay.

Selection of Trials	N Trials	N Patients	Estimate of Effect			Heterogeneity
			RoM	95% CI	p	I 2	p
All	17	1967	0.835	0.81–0.86	10−26	90%	10−24
Exclusion1 (a)	16	1882	0.908	0.87–0.94	10−6	58%	0.002
Exclusion2 (a)	15	1842	0.923	0.89–0.96	0.00003	35%	0.088
Final meta-analysis (b)	12	1766	0.922	0.89–0.96	0.00003	43%	0.055
Sensitivity analysis (c)	9	1312	0.927	0.88–0.98	0.005	51%	0.036
Cardiac trials	11	1726	0.918	0.88–0.95	0.00001	43%	0.066
Oral vitamin C	6	927	0.914	0.86–0.97	0.003	49%	0.08
Intravenous vitamin C	6	839	0.928	0.88–0.97	0.003	47%	0.10
Trials in Iran	6	862	0.927	0.88–0.98	0.005	43%	0.12
Trials out of Iran	6	904	0.917	0.87–0.97	0.002	53%	0.06
1-2 days ICU (d)	7	1231	0.943	0.89–0.99	0.027	52%	0.05
3-5 days ICU (d)	5	535	0.899	0.85–0.95	0.0001	23%	0.3

^(a) Exclusion1 is the meta-analysis in which the Dingchao trial [106] is excluded. Exclusion2 is the meta-analysis in which the Abdoulhossein study vitamin E patients [122] is further excluded. ^(b) The final meta-analysis excludes the three small trials by Colby [109], Fowler [115], and Zabet [117] in addition to the two trials excluded in Exclusion2. Note that the exclusion of these three small trials has no meaningful effect on the point estimate of effect (RoM: 0.9227 versus 0.9220) or on the p-value, but their exclusion increases the level of heterogeneity from I² = 35% to 43%. The Final meta-analysis, indicated by bold, is shown as Figure 3. The subgroup analyses below the final meta-analysis are based on the final meta-analysis. ^(c) In the sensitivity analysis, we excluded two trials with no description of allocation [114,118], and a quasi-randomized trial [121]. ^(d) Length of ICU stay in the control group of the trial; see the comparison of the subgroups in Supplementary file S1. Abbreviations: RoM, ratio of means; for example, RoM = 0.922 of the final model indicates that vitamin C shortens the mean length of ICU stay by 7.8%.