Vitamin C may reduce the duration of mechanical ventilation in critically ill patients: a meta-regression analysis

Background Our recent meta-analysis indicated that vitamin C may shorten the length of ICU stay and the duration of mechanical ventilation. Here we analyze modification of the vitamin C effect on ventilation time, by the control group ventilation time (which we used as a proxy for severity of disease in the patients of each trial). Methods We searched MEDLINE, Scopus, and the Cochrane Central Register of Controlled Trials and reference lists of relevant publications. We included controlled trials in which the administration of vitamin C was the only difference between the study groups. We did not limit our search to randomized trials and did not require placebo control. We included all doses and all durations of vitamin C administration. One author extracted study characteristics and outcomes from the trial reports and entered the data in a spreadsheet. Both authors checked the data entered against the original reports. We used meta-regression to examine whether the vitamin C effect on ventilation time depends on the duration of ventilation in the control group. Results We identified nine potentially eligible trials, eight of which were included in the meta-analysis. We pooled the results of the eight trials, including 685 patients in total, and found that vitamin C shortened the length of mechanical ventilation on average by 14% (P = 0.00001). However, there was significant heterogeneity in the effect of vitamin C between the trials. Heterogeneity was fully explained by the ventilation time in the untreated control group. Vitamin C was most beneficial for patients with the longest ventilation, corresponding to the most severely ill patients. In five trials including 471 patients requiring ventilation for over 10 h, a dosage of 1–6 g/day of vitamin C shortened ventilation time on average by 25% (P < 0.0001). Conclusions We found strong evidence that vitamin C shortens the duration of mechanical ventilation, but the magnitude of the effect seems to depend on the duration of ventilation in the untreated control group. The level of baseline illness severity should be considered in further research. Different doses should be compared directly in future trials.


Standard meta-analysis of 8 trials:
> venMA <-metagen (TE, seTE, studlab,data=Ven,sm="RR", comb  Exclusion: Change from off-pump to on-pump surgery, known drug allergy, history of COPD, anemia, congestive heart failure (CHF), active sepsis, preoperative ejection fraction lower than 40%, preoperative creatinine above 1.3 mg/dL, use of nephrotoxic drugs, coronary angiography, intraoperative transfusion of more than 2 units of red blood cells, perioperative use of intra-aortic balloon counterpulsation, perioperative requirement for high-dose vasopressors, and any intraoperative life-threatening events such as fatal arrhythmias, excessive bleeding, or desaturation.

Interventions
Vit C before the operation: Dose: 3 g/d "vitamin C... 1500 mg ... tablets, twice a day, from 24 hours before the operation until two postoperative days." Method: po Timing: "24 h before operation" Vit C after operation: Dose: 3 g/d "1500 mg .. tablets ... twice a day" Method: po Duration: 2 d "until two postoperative days" Control group: no tablets. The patients received many drugs and it is unlikely that they identified vitamin C among all the other administered drugs. We classify that all the other drugs serve as a functional placebo to vitamin C.

Length of mechanical ventilation
Notes "Ventilation times were 7.33±6.02, 10.68±27.15, 5.90±3.13, and 25.36±157.5 hours in the vitamin C, NAC, selenium and control groups, respectively (P=0.429)." (p 131) Thus, the SD for ventilation in the control group was much greater than the SD for each of the other 3 groups.
We were able to contact Dr. Amini and he kindly sent us their data set (email 2018-11-6). We found that there was one patient (#198 male) in the control group who had ventilation time 1333 hours. We removed this outlier and calculated that mean ventilation time 6.68 (SD 4.26) hours in the control group. This revised SD values is closely similar to the SD values of the three other arms, compare above.
We were also able to get more information of the Methods from Dr. Amini (email 2018-12-13), see below.

Bias
Authors' judgement Support for judgement

Random sequence generation (selection bias)
Low risk "Using a computer-based randomization method" (p 130).
Groups were balanced (Table 2) for eg. age, sex, weight, height, baseline diseases, ejection fraction Allocation concealment (selection bias) Low risk "The patients, the researcher who collected the data and the one who analyzed the data were blinded to the study. However, the physicians and nurses in charge were aware of the allocated groups" (email 2018-12-13).

Blinding of participants and personnel (performance bias)
Low risk "The patients, the researcher who collected the data and the one who analyzed the data were blinded to the study. However, the physicians and nurses in charge were aware of the allocated groups" (email 2018-12-13).
Given that no benefit was found from vitamin C, we do not consider that this lack of benefit could be biased by "the physicians and nurses in charge were aware of the allocated groups". Therefore we classify this as low risk for personnel. Blinding of outcome assessment (detection bias) Low risk "The patients, the researcher who collected the data and the one who analyzed the data were blinded to the study. However, the physicians and nurses in charge were aware of the allocated groups" (email 2018-12-13).
Given that no benefit was found from vitamin C, we do not consider that this lack of benefit could be biased by "the physicians and nurses in charge were aware of the allocated groups". Therefore we classify this as low risk for outcome assessment. Incomplete outcome data (attrition bias) Low risk 6 vitamin C participants and 2 control participants discontinued (p 131; Fig 1) and we (HH and EC) removed 1 control participant, see Notes above. Inclusion: >18 yr who were scheduled to undergo CABG.
Exclusion: current AF, temporary or permanent pacemaker, life expectancy <1 month, emergency surgery precluding the initiation of study protocol the evening before surgery, current pregnancy.

Interventions
Vit C before the operation: Dose: 2 g Method: po Timing: "the evening before surgery" Vit C after operation: Dose: 2 g/d "1 gram twice daily" Method: po Duration: 5 d Placebo: "identical placebo capsules at the same intervals"; "the inert substance for both treatment and placebo capsules was talc."

Length of mechanical ventilation
Notes "Both ascorbic acid and inert placebo capsules were prepared by a custom pharmacy" (p 863) indicates that the products were not commercial. Bjorndahl reported the duration of ventilation in vitamin C group as 1.2 (SD 0.8) d and in control group as 1.4 (SD 1.0) d.
Calculation of the P-value with the Taylor series formula from these reported mean and SD-values gives P = 0.13. Bjorndahl reported P = 0.032 (Mann Whitney) for the effect of vitamin C on ICU stay. We used this P = 0.032 to calculate the consistent SE(log(RoM), see Supplementary file. "The pharmacy department maintained the randomization list and assigned participants to the placebo and treatment arms of the study in a blinded fashion. Participants, clinicians, and evaluators were blinded to the treatment assignments and the blind was not broken until after data analyses were complete" (p 863). Blinding of participants and personnel (performance bias)

Interventions
Vit C before the operation: Dose: 2 g "2 g of vitamin C tablets before the surgery" Method: po Timing: "All patients took the tablets within 12 hours before surgery" (email 2015-9-9) Vit C after operation: Dose: 1 g/d "500 mg twice daily" Method: po Duration: 5 d Placebo: No tablets. The patients received many drugs and it is unlikely that they identified vitamin C among all the other administered drugs. We classify that all the other drugs serve as a functional placebo to vitamin C.

Length of mechanical ventilation
Notes "Our study was funded by Urmia University of Medical Sciences, Iran" (email 2015-10-1).

Risk of bias table Bias
Authors' judgement Support for judgement

Random sequence generation (selection bias)
Low risk "All patients were randomized into two groups in a 1:1 ratio using randomnumber table" (p 493).
Groups were balanced (Table 1) for eg. age, sex, smoking, baseline diseases, ejection fraction Allocation concealment (selection bias) Low risk "Neither ward physician nor Holter interpreter were aware of the patients' group. Only one who analyzed data was aware of the patients' group" (email 2015-4-11) and "we did not let ward physician and surgeons to know which of patients taking vitamin c or not, except for being informed about the conduction of our trial and prescribing some of patients to take vitamin c. Furthermore, patients were informed that they would be included in our trial to be prescribed vitamin c" (email 2015-4-22).

Blinding of participants and personnel (performance bias)
Low risk "Neither ward physician ... were aware of the patients' group. " (email 2015-4-11) and "we did not let ward physician and surgeons to know which of patients taking vitamin c or not, except for being informed about the conduction of our trial and prescribing some of patients to take vitamin c. Furthermore, patients were informed that they would be included in our trial to be prescribed vitamin c" (email 2015-4-22). Blinding of outcome assessment (detection bias) Low risk See above.
Inclusion: patients assigned for elective CABG surgery with cardiopulmonary bypass (CPB) were enrolled in the study.
Exclusion: renal or hepatic dysfunction, known hypersensitivity to the studied drugs, COPD, preoperative AF, pacemaker, class I and III antiarrhythmic agents or digoxin, any degree of AV block

Interventions
Vit C before the operation: Dose: 2 g Method: iv, but not explicitly stated; "after induction of anesthesia" implies iv administration Timing: "after induction of anesthesia" Vit C after operation: Dose: 3 g/d "1 g every 8 h daily until the fifth post-operative day" (p 60). Method: iv, but not explicitly stated. "Control group receiving saline infusion" suggests that vitamin C also was administered by infusion Duration: 5 d Placebo saline infusion

Notes
Very poorly reported study.

Random sequence generation (selection bias)
Unclear risk "divided into three equal groups" (p 60) (One group received magnesium). However, Ebade did not describe the method of allocation; alternative allocation?
Groups were balanced (

Interventions
Vit C dosage: Dose: "1.5 gm vitamin C" Method: iv Timing: "every 6 hours" (p 78) Thus, total vitamin C daily dose was 6 g/day. Placebo no explicit placebo. "Control group: 50 septic shock patients received conventional sepsis treatment only" (p 78). The patients received many drugs and it is unlikely that they identified vitamin C among all the other administered drugs.
We classify that all the other drugs serve as a functional placebo to vitamin C.

Notes
Quite poorly reported study.
We tried to contact Dr. Habib to ask for the details of their methods (2019-3-29 and 2019-4-1 and 2019-4-10), but we did not get any response.

Bias
Authors' judgement Support for judgement

Random sequence generation (selection bias)
Low risk Quasi-randomization: "All enrolled patients who met the inclusion criteria (n=100) were randomized (even odd randomization technique) into the study" (p 78).
Groups were balanced (Table 1) for age, sex, APACHE II score, SOFA score, mean arterial pressure, heart rate, respiratory rate, temperature, urine output, (Table 2) CRP and procalcitonin Allocation concealment (selection bias)

Not described
Blinding of participants and personnel (performance bias)

Not described
Blinding of outcome assessment (detection bias)

Not described
Incomplete outcome data (attrition bias)

Low risk
The report indicates that 100 participants were enrolled "All enrolled patients who met the inclusion criteria (n=100)", and there are 100 participants in their analysis Sadeghpour 2015 (not included in the statistical models for the 42% dropout rate)
Exclusion: who died within the 1st postoperative day and those who had not received adequate doses of drugs according to our protocol, severe complications (cardiac, respiratory or neurological) or emergency operation.

Interventions
Vit C before the operation: Dose: 2 g Method: iv Timing: "immediately before surgery" Vit C after operation: Dose: 1 g/d Method: po Duration: 4 d Placebo: "The patients in the placebo group received an equal number of identical tablets. The placebo tablets and ampoules were prepared in the same shape and size as the original" (p 2).

Length of mechanical ventilation
Notes "We paid it by ourselves besides getting help from the Rajaei cardiovascular research center." (email 2015-10-1) Additional information was received by email from Anita Sadeghpour (2015-5-12 and 2015-10-1), see below and above.
In 2019 we received a response describing that 42% of participants were excluded after randomization. The publication describes 113 vit C / 177 placebo participants which indicates that there is substantial difference in the exclusion rate between the vitamin C and placebo groups. This is such a severe violation of the ITT principle that we exclude the trial from our calculations, but we show the results in Fig. 1.

Authors' judgement Support for judgement
Random sequence generation (selection bias) Low risk "The study population was randomized one day before surgery to two groups (By using www.randomizer.org) ... The method of randomization was balanced block with an allocation sequence based on a block size of eight" (p 2). However, the sizes of the groups 113 vit C / 177 placebo are not consistent with block randomization. Therefore, we exclude this study in our sensitivity analysis.
Groups were balanced ( "Both the patients and the hospital staff were blind to the treatment allocation" (p 2). Blinding of participants and personnel (performance bias) Low risk "Both the patients and the hospital staff were blind to the treatment allocation" (p 2). "The Vit C was given in the operating room along with the other infusions by anesthesiologist technician" (email 2015-5-12). "The patients in the intervention group received 2 g of vitamin C ... intravenously, immediately before surgery in the operating theatre, followed by 1 g daily oral doses of the tablets for the first four postoperative days. The patients in the placebo group received an equal number of identical tablets. The placebo tablets and ampoules were prepared in the same shape and size as the original ones manufactured by the same pharmaceutical company." (p 2). Blinding of outcome assessment (detection bias) Low risk See above.

Incomplete outcome data (attrition bias)
High risk "we enrolled 500 patients but we excluded the patients who died on the first postoperative day, those who needed re operation due to technical problems and excessive bleeding, and those who had not received an adequate dose." (email 2019-1-24). Thus, data is published of 290 participants of 500 enrolled, which means that data of 42% of participants are missing. This is such a severe violation of the ITT principle that we exclude the trial from our nalysis. Inclusion: "older than 16 years; thermal injury within 2 hours before admission; burn covering greater than 30% of total body surface area" Exclusion: "preexisting hepatic, respiratory, cardiac, or renal dysfunction; and preexisting coagulopathy"

Interventions
Vit C dosage: Dose: "66 mg/kg per hour" corresponds to 110 g/day for a 70 kg patient Method: iv Timing: "24-hour study period" (p 327). Thus, vitamin C was administered for just 1 day. Placebo no specific placebo, but functional placebo: "The control group did not receive the ascorbic acid infusion", but "The administered volume of ascorbic acid was included in the 24-hour fluid intake calculations" (p 327). This indicates that patients could not distinguish vitamin C administration by difference in volumes Outcomes Length of mechanical ventilation Notes

Bias
Authors' judgement Support for judgement

Low risk
Quasi-randomization: "Randomization was performed according to the month of admission" (p 327). It is unlikely that randomization of burn patients according to the month of admission causes systematic bias.
Groups were balanced at baseline (Table  1 and Fig 1) for age, sex, weight, kinds of burns, areas of burns, thickness of burns, the number of inhalation injuries, heart rate, mean arterial pressure Allocation concealment (selection bias)

Low risk
Quasi-randomization: "Randomization was performed according to the month of admission" (p 327). It is unlikely that randomization of burn patients according to the month of admission causes systematic bias. There were no baseline difference in age, sex, weight, kinds of burns, areas of burns, thickness of burns, or in the rate of inhalation injuries (Table  1 and Fig 1). Blinding of participants and personnel (performance bias) Low risk "The administered volume of ascorbic acid was included in the 24-hour fluid intake calculations" (p 327). This means that patients had same amount of fluids in both groups. In addition, it is highly unlikely that patient with a burn covering greater than 30% TBSA would observe their treatments. Blinding of outcome assessment (detection bias)