Influence of early antioxidant supplements on clinical evolution and organ function in critically ill cardiac surgery, major trauma, and subarachnoid hemorrhage patients

We conducted a prospective, randomized, double-blind, placebo-controlled, single-center trial with the approval of the institutional ethics committee.

Two hundred consecutive patients admitted to the intensive care unit (ICU) at the University Hospital (Centre Hospitalier Universitaire Vaudois) of Lausanne were enrolled from January 2003 through September 2004. Three conditions admitted with organ failure deemed likely by the medical team to require at least 48 hours of ICU treatment were considered: cardiac valve or coronary bypass surgery with postoperative cardiac or respiratory failure, major trauma (with or without brain injury) with an Injury Severity Score (ISS) of greater than 9, and severe subarachnoid hemorrhage (SAH) (that is, World Federation of Neurological Surgeons grades 3, 4, and 5) [19]. These three pathologies were investigated based on the demonstration of oxidative stress-related damage [17, 18, 20]. Exclusion criteria were absence of consent, participation in another study, liver cirrhosis or major burns, and life expectancy of less than 48 hours or a lack of commitment to full aggressive care (anticipated withholding or withdrawing of treatments in the 48 hours).

Cardiac surgery patients' preoperative status was assessed using the Parsonnet score [21] and the Euroscore [22]. Trauma severity was based on the ISS [23] and the Glasgow Coma Scale (GCS) score on admission and at discharge. Severity of physiological condition was determined by the first 24 hours' Simplified Acute Physiology Score (SAPS II) [24] and the daily sequential organ failure assessment (SOFA) [25] scores. Pre-existing renal failure was defined as a preadmission creatinine clearance of less than 60 mL/minute (measured or calculated with the Cockcroft-Gault equation on preoperative creatinine value [26]).

After stratification for diagnosis, the patients were randomly assigned by the pharmacist to either AOX micronutrient or placebo group, using a random list with a four-block allocation. Patients, clinicians, and investigators were blinded to the treatment. Black plastic bags covered the solutions, and colored tubing was used for infusion. Trace elements and vitamins were prepared in separate bags. Labels carried the patients' name, study number, and whether the content was supposed to be vitamin or trace element. The ethics committee, considering the limited risks associated with the micronutrient supplements, delivered a waiver of consent for the study enabling the randomization, the initiation of the intervention, and the first blood sampling. Oral consent was requested within 48 hours and a written consent within 96 hours. The patient was asked first whether he/she was not competent at that time, and provisory consent was requested among the relatives and confirmed by the patient once his/her condition had recovered.

Intervention consisted of delivering either AOX supplement or placebo for 5 days starting within 24 hours of admission. The supplementation consisted of an initial loading dose (double dose for 2 days) followed by 3 days of therapeutic dose (Table 1). The intervention solutions were infused alternately intravenously (iv) over the course of 10 to 12 hours each (vitamins during the daytime and trace elements over night). Products were Decan^®, Sodium selenite^®, and Zinc gluconate^® (Laboratoires Aguettant, Lyon, France), Soluvit^® + Vitalipid^® (Fresenius Kabi AG, Bad Homburg, Germany), vitamin B₁ as Benerva^® and vitamin C (Streuli Pharma AG, Uznach, Switzerland), and α-tocopherol as Ephynal^® (Roche, Basel, Switzerland). Vitamin E was delivered by nasoenteric tube. Besides the intervention solution, all patients received the ICU's 'standard vitamin profile' consisting of 100 mg thiamine and 500 mg vitamin C iv per day. In case of suspicion of alcohol abuse, an additional 100 mg thiamine was delivered. The 11 patients requiring parenteral nutrition (PN) for a total of 82 days (3 in the placebo group and 8 in the AOX group; P = 0.13) received the recommended doses of micronutrients for PN in addition to the 'intervention solution' (1 ampule of Soluvit^® + 1 ampule of Vitalipid^® + 1 ampule of Decan^®) as part of standard care. In patients on full enteral nutrition (EN), one multivitamin and mineral tablet (Supradyn^®; Roche) were delivered per day with EN (these micronutrients are not included in Table 1).

The primary outcome variable was a change in the acute kidney injury (AKI) score. Changes in organ function monitored by the SOFA score [25] were considered as an important secondary endpoint but were not used to calculate the sample size. At the time of the trial initiation, the SOFA scores' capacity to detect changes in mainly cardiac and trauma patients was not known. Renal failure affects about 35% of critically ill patients [27] and remains a major determinant of length of hospital stay [28]. Three levels of severity were considered: (a) AKI based on acute alterations of urine output according to the AKI Network [29] (stage 1: urine output of less than 0.5 mL/kg for 6 hours; stage 2: urine output of less than 0.5 mL/kg for 12 hours; and stage 3: urine output of less than 0.3 mL/kg for 12 hours or anuria for 12 hours). Acute renal failure was further defined by a plasma creatinine increase of (b) 50 or (c) 90 μmol/L [30]. The SOFA score was further used to test global organ dysfunction: this score ascribes a value of severity of organ failure from 0 to 4 for cardiovascular, respiratory, renal, hepatic, nervous, and coagulation failure (maximum score of 24). The SOFA score was repeated daily until day 5 or until discharge from the ICU. Secondary outcome variables included the daily worst arterial partial pressure of oxygen/fraction of inspired oxygen (PaO₂/FiO₂) ratio and mechanical ventilator dependence. The number of ventilator-free days to day 30 was defined as the number of days of unassisted breathing to day 30 after randomization, assuming a patient survives and remains free of invasive or noninvasive assisted breathing for at least 2 consecutive calendar days after extubation, whatever the vital status at day 30. Infectious complications, duration of ICU stay (counted by quarter-days rounded to the closest 6 hours), and ICU, hospital, and 3-month mortality rates were recorded. All infectious complications were recorded using the Centers for Disease Control and Prevention definitions [31], with special emphasis on pulmonary infections: pneumonia was defined as the combination of systemic inflammatory response syndrome (SIRS) with a new infiltrate on the chest x-ray (or progression of an infiltrate), a new or persistent hypoxemia, and purulent sputum. A standardized questionnaire was used to assess quality of life at 3 months: the Medical Outcome Study Short Form 36-item health survey (MOS SF-36). The patients were contacted by telephone, and the questionnaire was limited to the physical components: physical activity (physical functioning [PF]: scores 10 to 30), limitation due to physical status (role functioning-physical [RP]: scores 4 to 8), pain (bodily pain [BP]: scores 2 to 12), perceived health (general health perception [GH]: scores 5 to 25), and summary physical score of the MOS SF-36 were analyzed [32]. Due to interview difficulties, psychological components were not elicited.

Plasma creatinine, C-reactive protein (CRP), glucose, albumin, leukocytes, and platelets were determined daily for clinical purposes, and aspartate amino transferase, alanine amino transferase, and urate were determined three times weekly using standard clinical laboratory methods. Blood samples were collected on admission (day 0) and on day 5 (end of supplementation) to determine plasma zinc and selenium: analysis was in duplicate by inductively coupled plasma mass spectrometry (Plasmaquad 3 ICP-MS; VG Elemental, Winsford, Cheshire, UK) using aqueous inorganic standards. All plasma specimens were diluted in 1% nitric acid/0.2% n-butanol/0.2% n-propanol and 10 parts per billion indium as internal standard [33]. Plasma GPX was determined by the RANSEL method (Randox Laboratories, Belfast, UK).

EN or PN was initiated on a clinical basis, according to the unit's clinical protocols with the standard ICU solutions. The energy target was calculated as 1.2 times the predicted resting energy from the Harris and Benedict equation in cardiac surgery and SAH patients and as 30 kcal per kg body weight in trauma patients. EN was considered first, and PN was used only when EN was contraindicated. Glucose-insulin infusions were delivered in those at risk of ischemic postoperative cardiac failure. The cumulated energy balance was calculated at 5 days and at the end of the ICU stay. For those patients discharged before day 5, the energy intakes were recorded until day 5.

The blood glucose target of 5 to 8 mmol/L was achieved by means of a continuous insulin infusion. For every patient, a mean of all blood glucose values per 24 hours was calculated during the first 5 days. The 24 hours' insulin doses were recorded. The variables were retrieved from the database of our clinical information system (MetaVision^®; i MD soft, Tel Aviv, Israel).

As there were no data available in the literature on the expected impact of supplementation on the SOFA score, the sample size was determined based on an expected reduction of acute renal failure of 20% defined as a creatinine increase of 50 μmol/L [34], an organ failure that has a significant impact on outcome. We realized an a priori power analysis, expecting an acute renal failure incidence of 30% and a 50% reduction, using an alpha level of 0.05 and a power of 0.9: these numbers resulted in a sample size of 186 (rounded to 200). A safety committee was formed to address safety issues, but it could not modify the sample size. After 60 and 120 patients, respectively, two meetings were conducted in order to detect over-mortality and adverse events: the two intermediate analyses did not detect any difference between groups.

Analysis was by intent to treat. Data are presented as mean ± standard deviation or as median and range when specified. Demographic data, energy balance, and baseline variables were analyzed by one-way analysis of variance (ANOVA) as they were normally distributed; two-way ANOVAs were used for variables repeated over time such as SOFA score, PaO₂/FiO₂ ratios, laboratory variables, and insulin dose. Post hoc comparisons were carried out by Dunnett test (effect of time versus baseline in each group) or Scheffe test (between-group comparisons at the same time point), where appropriate. Rank tests were used for nonparametric variables. Kaplan-Meier analysis was applied to length of hospital stay: nonsurvivors were considered as never achieving the event of interest (discharge) and censored at the end of evaluation period. Multiple and simple logistic regressions between variables were calculated. Significance was considered at a P value of less than 0.05; trends were considered up to a P value of less than 0.25. The statistical package was JMP^® version 5.1. (SAS Institute Inc., Cary, NC, USA).

There were 47 protocol violations (19 in the placebo group and 28 in the AOX group), evenly distributed among the three diagnostic categories (18 in cardiac, 22 in trauma, and 7 in SAH). Among these, 25 were considered nonsignificant as they reinforced the intervention effect (by increasing the doses of AOX), while 22 violations reduced the difference between groups (5 in placebo patients who received 1 or 2 doses of AOX, 17 in the AOX group by deletion of 1 to 3 doses of supplement). All patients were included in the intent-to-treat analysis.

Total energy delivery during the first 5 days was hypocaloric in most patients, and the progression of energy delivery was slower than recommended by our protocol. Eleven patients required PN for 3 to 11 days. The mean and median cumulated energy balances on day 5 were negative and did not differ significantly between groups (-5,415 kcal in placebo versus -5,680 kcal in AOX).

SF-36 could be retrieved in 140 (70%) of the 174 surviving patients (1 cardiac placebo patient died during the fourth month), including 68 AOX and 72 placebo patients (88 cardiac, 36 trauma, and 16 SAH). The 34 missing patients were not feeling well enough to answer (n = 11) or were having language problems (n = 7) or were lost to follow up (altogether, n = 17). Physical activity score (PF) tended to be higher in AOX (24.1 ± 4.9 versus 22.8 ± 5.7; P = 0.14). Physical limitation (RP: 5.8 ± 1.4 versus 5.5 ± 1.5; not significant), physical pain (BP: 8.9 ± 2.4 versus 9.0 ± 2.7; not significant), and perceived health (GH: 18.9 ± 4.5 versus 19.2 ± 4.1; not significant) did not differ. Perceived evolution of health after the hospital discharge (HT = Health Transition) was significantly better in the AOX patients, with significantly more frequent ratings 'better' and 'rather better' (P = 0.01).

Selenium may be the cornerstone of the AOX defense system in acute conditions [13], but other trace elements, and zinc particularly, are also important players [16]. In our study, the tested selenium dose (540 μg followed by 270 μg/day) can be considered low compared with other recent trials [15, 41, 42]. Nevertheless, it is a very substantial intake compared with the normal healthy subject requirements of 60 μg per day, and indeed it did correct the plasma selenium concentrations and normalize GPX activity. In the ICU, doses ranging between 350 and 1,000 μg for 10 to 15 days have been associated with clinical benefits [13], while chronic doses of greater than 450 μg/day in the general population have been associated with a reduction of the activity of the 5' triiodothyronine deiodinase (an indicator of upper safe intakes [43]). The reduction in renal failure in inflammatory patients from the first Angstwurm trial was achieved with doses ranging between 100 and 530 μg/day [34] but was not confirmed in subsequent studies with the same dose [44] or higher doses [15]. The latest trial of Forceville and colleagues [42] tested a loading dose of 4 mg followed by 1 mg/day for 2 weeks in 60 septic patients: it did not show any reduction in renal failure, while length of mechanical ventilation was nonsignificantly prolonged (14 days in the placebo group versus 19 days in the selenium group), possibly reflecting an incipient selenium toxicity.

On the other hand, the doses of zinc (2 days of 50 mg then 25 mg/day) and vitamins C, E, and B₁ were high compared with other trials. This combination was motivated by the inclusion of trauma patients, who develop early negative micronutrient balances and low plasma zinc levels. Furthermore, as neurological damage was present in two of our diagnostic groups, zinc was given based on trials in brain-injured patients in whom 30 to 40 mg zinc doses were associated with improved neurological outcome [45]. Finally, zinc is involved in tissue repair, and the improved wound healing in burns observed with 30 mg iv for 21 days [46] encouraged the addition of this rather large dose of zinc. Indeed, plasma levels normalized with the supplements and trauma patients appeared to benefit from the intervention.

Ascorbic acid, alpha-tocopherol, and thiamine were combined to reinforce the AOX cascade [16]. Vitamin C is strongly depressed in critical illness, and doses of 3 g per day are required to normalize plasma concentrations [47]. A loading dose of 2.7 g for 2 days was therefore adopted. Although chronic administration to healthy subjects of vitamin E of doses exceeding 150 mg/day is associated with increased mortality [48], several trials have delivered 3 g per day for shorter periods to trauma patients and other critically ill patients without any detectable side effects [38, 49]. Since the normal daily intake is of the order of 10 mg, our 300 to 600 mg/day dose can be considered at the low range of the high-dose spectrum. Finally, vitamin B₁, though not an AOX but the coenzyme of all carbohydrate decarboxylation reactions, was delivered to all patients as deficiency is a recognized problem in early critical illness [50], with recommendations to deliver 100 mg and up to 300 mg/day.

The supplementation did not achieve a significant reduction of the primary outcome (that is, early organ failure). There are different possible explanations. First, we did not collect the SOFA scores after discharge from the ICU – this might have been valuable. Indeed, the impact of a nutritional intervention may be detected only after 4 to 5 days. Then, despite the inclusion of 200 patients in three diagnostic categories, the study was still underpowered. The subgroup analysis is therefore confronted with an even poorer lack of power. Despite this problem, all the mean values were oriented in the direction of a clinical benefit, with trends in favor of the AOX intervention. These trends nevertheless can be considered only as hypothesis-generating: for example, the trauma patients might benefit from such an intervention but a larger trial is required. There was an a priori decision to analyze all patients and then by diagnostic category as the constitution of these categories was part of the randomization process. The small SAH group experienced very few complications and had less inflammation in agreement with other studies [36], contributing to the reduction of power. This was worsened by the lower-than-predicted mortality by either of the outcome scores [25]. Furthermore, despite stratification and randomization, we observed heterogeneity among the trauma patients. The more severe brain injuries in the AOX group, which directly impact on survival, explain the trend to higher mortality in this group and must be considered in the analysis [51]. This type of bias is unavoidable, even with strict randomization [35]. Balanced minimization is an efficient method of ensuring comparable groups [52]. It consists of determining, at the onset, outcome factors one would like to see evenly distributed in the groups. The person in charge of the randomization observes the progression of these factors during the trial, modifying group allocation if a difference develops. This method was not used in this study but probably would have avoided this unlucky uneven distribution of severity factors. Finally the characteristic of being a monocentric trial can be considered both a limitation but also an advantage due to reduction of variability of care.

Selenium-containing AOX micronutrient supplements did not significantly improve organ dysfunction as assessed by the SOFA score during the first 5 days. Nevertheless, the inflammatory response was significantly blunted as shown by a lower CRP in the AOX group. The diagnostic categories with an intense inflammatory response, such as patients with organ failure after cardiac surgery or major trauma patients, thus appear good candidates for an AOX intervention, while patients with limited inflammatory response are not. The optimal dose and combination of micronutrients still remain to be determined but should include selenium and zinc.