Vitamin E supplementation for prevention of morbidity and mortality in preterm infants

Criteria for considering studies for this review

Search methods for identification of studies

See Collaborative Review Group Strategy. 
 MEDLINE accessed using PubMed in October 2002 (1966 ‐ 2002), EMBASE in March 2002 (1974 ‐ 2002), and the Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library, Issue 1, 2003 and Issue 1, 2007. We examined reports in any language as long as the report included an abstract in English. The MEDLINE search was repeated in April 2003 and March 2007. 
 The MEDLINE search was done using the following search terms: (<vitamin E> OR <tocopherol>). The search was limited to <birth ‐23 months> AND (<randomized clinical trial> OR <controlled clinical trial>). Because this search yielded many studies conducted in children, it was re‐done using (<vitamin E> OR <tocopherol>) AND (prematurity OR preterm). 
 The EMBASE search was done using the following search terms: (Exp alpha tocopherol OR vitamin E.mp. OR *tocopherol/) AND (exp prematurity/ OR preterm.mp.). The search was limited to (human and infant < to one year>). 
 Selection was done independently by three investigators; any disagreement was resolved by discussion. 
 The CCTR does not accept searches using single‐letter words. Therefore the search was done using the following search terms: Tocopherol AND (prematurity OR preterm) AND infant. 
 Personal files were searched and reviewed relevant references cited in the other manuscripts.

Data collection and analysis

The standard methods of the Cochrane Neonatal Review Group (CNRG) were used. 
 http://neonatal.cochrane.org/ 
 "Organization and Methodology of a Systematic Review for CNRG (Guidelines for Reviewers & Editors)" 
 These included independent review of all candidate studies by all reviewers. The methodological quality of the trials was assessed by analyzing the risk of four types of bias: selection, performance, attrition and detection. 
 Standard methods of the CNRG were used for estimating treatment effects with 95% confidence interval (CI) and a fixed effect model. Dichotomous variables were assessed using relative risk (RR), risk difference (RD), and number needed to treat/harm. Data were extracted on as many randomized patients as possible, and analyzed on an intent‐to‐treat basis, using as denominator for dichotomous variables the total number of randomized patients. Since several studies have reported specific outcome variables only in selected subgroups subject to attrition bias (e.g., survivors) the names of the variables were modified accordingly. Exceptionally, the total number of patients in each arm of the study was not available because the authors have excluded patients from the study after randomization; in this case, the maximum number of available patients was used instead of the total number of randomized patients. Data reported in very low birth weight infants were treated as separate variable, to allow subgroup analyses in this important population. Continuous variables are presented as mean ± 1 standard deviation (SD). Where justified clinically, we merged two or more subgroups of a published study, using standard formulae to calculate the weighted mean and to estimate the weighted standard deviation (Altman 1991; Brion 1990; Zahr 1984). Continuous variables were assessed by the weighted mean difference (WMD) between the treatment group and the control group. 
 Additional data including those obtained after or pending publication were requested from the authors, if necessary.

I. Original plan:

To analyze studies separately according to the amount of vitamin E provided to the mother during pregnancy, and according to the amount, type or route of vitamin E supplementation of the infants in the control group: 
 1. Analysis excluding studies with documented prolonged (>1 week) high‐dose (>= 400 mg/day) vitamin E supplementation during pregnancy 
 1.1 Infant provided with placebo, with total daily dose of vitamin E < 10 mg/100 kcal 
 1.2. Infant provided with another dose of vitamin E (total intake >10 mg/100 kcal) 
 1.3. Infant provided with another type or route of vitamin E formulation 
 2. Analysis of studies with documented prolonged (> 1 week) high‐dose (>=400 mg/day) vitamin E supplementation during pregnancy 
 2.1. Infant provided with placebo, with total daily dose of vitamin E < 10 mg/100 kcal 
 2.2. Infant provided with another dose of vitamin E (total intake >10 mg/100 kcal) 
 2.3. Infant provided with another type or route of vitamin E formulation

Subgroup analyses planned: 
 1. Gestational age: 
 1.1. >= 29 weeks 
 1.2. < 29 weeks 
 If gestational age is not available or is not the cut‐off for the trials: 
 1.3. Birth weight > 1000 g 
 1.4. Birth weight <= 1000 g

2. Medication given in the treatment group: 
 2.1. Route of administration and formulation: 
 2.1.1. Enteral 
 2.1.1.1. Enteral lipid form ‐ isotonic to mildly hypertonic (< 400 mOsm/kg) 
 2.1.1.2. Enteral lipid form ‐ hypertonic (>= 400 mOsm/kg) 
 2.1.1.3. Enteral aqueous form 
 2.1.2. Parenteral 
 2.1.2.1. Parenteral, excluding polysorbate 
 2.1.2.2. Parenteral colloidal, excluding polysorbate 
 2.1.2.3. Parenteral colloidal with polysorbate 
 2.1.2.4. Parenteral olive oil

3. Total dose of vitamin E provided (iv + diet + supplement) 
 3.2.1. <25 mg/kg/day 
 3.2.2. >= 25 mg/kg/day

4. Time of onset of vitamin E supplementation in the treatment group 
 4.1. Within the first 48 hours of life 
 4.2. After the first 48 hours of life 
 4.3. After 4 weeks of life

5. Other nutritional components which may modify the effects of or the need for vitamin E 
 5.1 Iron supplementation 
 5.1.1. Iron supplementation > 2 mg/kg/day in both groups 
 5.1.2. Iron supplementation > 2 mg/kg/day in treatment group only 
 5.1.3. Iron supplementation <= 2 mg/kg/day

5.2. PUFAs 
 5.2.1. Formula with PUFAs >400 mg/100 ml formula in both groups 
 5.2.2. Formula with PUFAs <400 mg/100 ml formula in both groups 
 5.2.3. Formula with PUFAs >400 mg/100 ml formula in treatment group

5.3. Dietary or parenteral supplementation with one or more other antioxidants (in both groups) 
 5.3.1. Selenium fortification >= 3 mcg/kg/day 
 5.3.2. Vitamin A supplementation (>= 1500 U/day) 
 5.3.3. Selenium fortification >= 2 mg/100 ml formula and vitamin A supplementation (>=1500 U/day) 
 5.3.4. Neither

6. According to study design: 
 6.1. Randomized studies only 
 6.2. All studies

II. Modifications to the original plan:

Studies of vitamin E administration to preterm infants have been conducted for more than fifty years, during which standard of care including baseline vitamin E intake in pregnancy and in very low birth weight infants has changed considerably. Furthermore, the definitions of chronic lung disease of prematurity or bronchopulmonary dysplasia and of retinopathy or prematurity or retrolental fibroplasia have changed several times during the last decades. Several methods have been used to administer vitamin E, and some of the differences in design among studies became clear only during completion of the protocol.

Outcome variables in the available trials were often reported only is specific subgroups, e.g., very low birth weight infants. Because very low birth weight infants are the only ones at risk for some of the complications potentially treated or prevented by vitamin E (e.g., severe retinopathy of prematurity), separate outcome variables were created in that weight group to allow subgroup analyses. This was the only way to determine which (or whether any) specific method of vitamin E administration (dose, duration, level, etc) is both efficient and nontoxic in very low birth weight infants. Subgroups among very low birth weight infants are referred to as "all studies" when they included all studies irrespective of the weight cut‐off (either 1500 grams or 1000 grams), "equal to or less than 1500 grams" or "equal to or less than 1000 grams", as appropriate for each specific weight cut‐off. For many outcome variables the first subgroup includes "all infants in all studies."

An additional outcome variable that had not been anticipated was added as a post‐hoc analysis: retinal hemorrhage.

In order to limit the number of comparisons, the main planned comparisons (listed in above sections 1 and 2) were replaced by the following: 
 1. Vitamin E versus placebo 
 2. Vitamin E versus another form or route of vitamin E (intramuscular versus enteral, intravenous versus intramuscular, intramuscular in olive oil versus intramuscular in colloidal form)

Many trials used combination of enteral and parenteral vitamin E supplementation, so these entries were modified as follows: 
 2. Route of administration and formulation: 
 2.1. Enteral 
 2.1.1. Enteral hypertonic formulation at pharmacologic dose 
 2.2. Parenteral with or without enteral 
 2.2.1. Parenteral with hypertonic enteral formulation at pharmacologic dose 
 2.2.2. Parenteral colloidal with polysorbate 
 2.2.3. Intravenous (with or without other routes of administration) 
 2.2.4. Excluding intravenous administration

The cut‐off value for vitamin E intake were clarified as follows: 
 3. Total daily dose of vitamin E provided (iv + diet + supplement) to the treatment group, target serum level 
 3.1. <=20 mg d‐alpha‐tocopherol equivalents/kg/d OR 30 IU/kg/d, i.e., 30 mg/kg/d dl‐alpha‐tocopheryl acetate* 
 3.2. > 20 mg d‐a‐tocopherol equivalents/kg/d OR 30 IU/kg/d, i.e., 30 mg/kg/d dl‐alpha‐tocopheryl acetate* 
 * If the dose is expressed in IU/day or mg/day, we used mean birth weight (estimated, if necessary, to be 1000 g for studies on infants with birth weight < 1500 grams) in the group for the calculations.

Based on Johnson 1995 an entry based on serum tocopherol levels at the end of the period of vitamin supplementation was added: 
 3.3. Serum tocopherol level <=3.5 mg/dl (81 micromoles/L) 
 3.4. Serum tocopherol level >3.5 mg/dl

Subgroup analyses were added based on duration of the treatment: 
 4. Time of onset and duration of vitamin E supplementation in the treatment group 
 4.1. Within the first 48 hours of life 
 4.2. After the first 48 hours of life 
 4.3. Duration of treatment <= 1 week (7 daily doses) 
 4.4. Duration of treatment > 1 week (7 daily doses)

Since only randomized trials were included for this review, the subgroup analysis of randomized trials was eliminated.

Vitamin E administration during pregnancy and dose of vitamin E in the control group is addressed as subgroup analyses instead of main comparisons: 
 6. Vitamin E provided during pregnancy 
 6.1. Documented prolonged high‐dose (at least 400 IU/day for a week) vitamin E during pregnancy

7. Total dose of vitamin E provided to the control group 
 7.1. <= 10 mg vitamin E/100 kcal 
 7.2. > 10 mg vitamin E/100 kcal

Subgroup analyses based on other nutritional components that may modify the effects of or the need for vitamin E were assessed as planned, except comparisons in which vitamin E was a co‐intervention with one of those supplements were eliminated.

Because these studies have spanned more than 50 years of research, definitions of outcome variables such as retrolental fibroplasia (retinopathy of prematurity) and bronchopulmonary dysplasia have changed. Only entries for which we found relevant studies are presented in the Table of Comparisons.

Description of studies

The MEDLINE search retrieved 128 studies, the EMBASE search 35 studies and the CCTR search 14 studies. Review of personal files and reading of documents obtained by the initial searches and by screening all references listed in manuscripts yielded more than 50 additional studies. An updated search of the MEDLINE and the CCTR conducted in March 2007 yielded one additional retrospective study (Liu 2005). After excluding ineligible studies and after merging multiple publications of the same studies, a total of 62 trials remained. Twenty‐six randomized clinical trials (range of patients per trial: 10‐914) were included in this systematic review. Among these 26 trials, five were conducted in the 1970's, 17 in the 80's, three in the 90's and one in 2003.

Only those trials that prospectively compared a group of vitamin E‐supplemented neonates with a control group are reported in the Tables of Included and Excluded Studies. Thirty‐five studies are listed in the Table of Excluded Studies. Studies were placed in this group for the following reasons (some studies for more than one reason): (1) trials (n=18) that did not assess any of the predefined outcomes or did not provide analyzable outcomes because either mean and variability (standard error of the mean or standard deviation), frequencies or individual data were not available, (2) studies (n=7) in which the experimental group and the control group were not enrolled simultaneously, (3) quasi‐randomized trials (n=12, of which three had at least one additional criterion of exclusion), defined as trials in which patients were not randomly allocated to the treatment and the control arms of the study, (4) trials (n=3, two of which also fitted the first criterion) in which either no preterm infants had been enrolled or a mixture of full‐term and preterm infants had been enrolled and (5) trials (n=1) in which vitamin E was only a co‐intervention. A complete list of 51 additional publications (case series, reviews, retrospective chart reviews) that were considered but are not reported in this systematic review is available upon request.

Although the classical definition of very low birth infants is a birth weight less than 1500 grams, studies assessing the effect of vitamin E on outcomes in preterm infants used a birth weight equal to or less than 1500 grams. In order to shorten the text, the term "very low birth weight infants" is used to refer to this latter group.

1. Vitamin E versus placebo 
 This group included 25 randomized trials that assessed the effects of vitamin E supplementation by enteral route only (n=8), intramuscular with or without enteral (n=15), or intravenous with or without other routes (n=2). Vitamin E supplementation was initiated within the first 48 hours of life in all treated patients in 20 studies, and after 48 hours in all treated patients in four studies. The duration of vitamin E supplementation was up to one week in eleven studies and greater than one week in all patients in thirteen studies. The total dose of vitamin E intake in the treated group was equal to or less than 30 IU/kg/day in fifteen studies and greater than 30 IU/kg/day in nine studies. Serum tocopherol levels were up to 3.5 mg/dl in fifteen studies and greater than 3.5 mg/dl in eight studies. The total dose of vitamin E intake in the control group was up to 10 mg/100 kcal in 22 studies and greater than 10 mg/100 kcal in three studies.

Chiswick 1982 is a randomized trial assessing the effect of four daily intramuscular injections of 20 mg/kg/dose of dl‐alpha‐tocopheryl acetate as an aqueous colloidal solution (Ephynal®, Hoffmann‐La Roche, Basel, Switzerland) starting at 24 hours of age on mortality and intraventricular hemorrhage in preterm infants. Entry criteria were: preterm infants without major congenital malformations. The concentration of the solution was 1000 mg/2 ml, and individual doses greater than 0.5 ml were divided and given in two different injection sites. Controls received no injection of vitamin E. Only five infants, i.e., those without respiratory distress, received formula (SMA Gold Cap Milk®, Wyeth Laboratories, Slough, England) during the first four days of life. The total number of patients entered into the study was 35, including 14 in the treatment group and 21 in the control group. Gestational ages ranged from 25 to 36 weeks (average 29.1±3.3 weeks in the treated group and 29.3±3.3 weeks in the control group) and birth weights from 550 to 1750 g (average 1.21±0.29 kg in the treated group and 1.25±0.33 kg in the control group). Intraventricular hemorrhage was diagnosed by computerized tomography, clinical findings (uniformly blood‐stained cerebrospinal fluid, seizures, tense anterior fontanel) or autopsy. Because not all infants had a computerized tomography (number not provided) or an autopsy (n =1), it is possible that some cases of intraventricular hemorrhage may not have been diagnosed; for the calculations we only used demonstrated cases of intraventricular hemorrhage (in contrast with the authors who assumed that one patient had a hemorrhage). Outcome variables included mortality and intraventricular hemorrhage.

Chiswick 1983 is a randomized trial assessing the effect of three intramuscular injections of 20 mg/kg/dose of dl‐alpha‐tocopheryl acetate as an aqueous colloidal solution (Ephynal®, Hoffmann‐La Roche, Basel, Switzerland) given at 12, 36 and 60 hours of age on intraventricular hemorrhage in preterm infants. Entry criteria were gestational age less than 37 weeks and birth weight less than 1751 g. Controls received no injection of vitamin E. No information is provided about enteral intake of vitamin E. Intraventricular hemorrhage was diagnosed by ultrasonogram performed within the first 48 hours of life and at least once after 72 hours; repeat examinations were done where possible daily during the first week and less often afterwards. The total number of patients was 44, including 21 in the treatment group and 23 in the control group. Mean gestational ages and birth weights in the whole study population are not provided. Outcome variables included mortality, germinal matrix/intraventricular hemorrhage, and intraventricular hemorrhage.

Cruz 1983 is a randomized controlled trial assessing the effect of four intramuscular injections of 25 mg/kg of vitamin E (on days one, two, four and eight, respectively) on hematologic parameters in preterm infants. Entry criteria were: (1) mothers admitted with threatened premature delivery at 26‐34 weeks of gestation, (2) infants born to mothers admitted for threatened preterm delivery at 26‐34 weeks of gestation and delivered with a weight less than 2300 g. All mothers of infants in the randomized trial received 900 mg of vitamin E (Ephynal®, Roche) daily for three days, followed by 100 mg daily until delivery. A non‐randomized third group of infants, born to non‐supplemented mothers, received no intramuscular vitamin E supplementation. All infants were fed banked breast milk, followed by an infant formula (Allomin). All infants received oral multivitamin drops starting on day five of life; this provided two mg of vitamin E daily. Among 55 infants enrolled in this trial, 27 were randomized to the experimental group and 28 to the control group. Mean gestational ages were 32.8±1.9 weeks and 32.4±1.9 weeks, respectively, in the experimental and in the control group, and mean birth weights were, respectively, 1808±361 and 1817±357 g. Outcome variables included retrolental fibroplasia, and bronchopulmonary dysplasia.

Ehrenkranz 1982 is a double‐blind randomized trial assessing the effect of repeated intramuscular injections of 20 mg/kg of Vitamin E Injectable® (Hoffmann‐LaRoche) on neonatal outcomes. Entry criteria included respiratory distress syndrome and informed consent. Injections were given upon admission to the study and 24, 48, and 168 hours later. Additional doses were given twice weekly as long as infants received oxygen therapy and could not tolerate feedings and vitamin supplements. Patients in the control group received same volumes of placebo, i.e., the vehicle. All infants initially received an intravenous protein solution containing vitamin E 2.5 U/L. Enteral feedings provided 16 U/L (formula) or 3 U/L (human milk). Patients tolerating feeds regularly were given oral vitamin E supplementation (Aquasol E®, USV Pharmaceutical Corp.) at a dose of 50 IU/day if weight was less than 1000 g and 25 IU/day if weight was greater than 1000 g. Among 100 patients enrolled into the study, 47 were in the experimental group and 53 in the control group. Average gestational ages were 30.3±2.7 and 30.5±2.9 weeks , respectively, in the experimental group and in the control group, and average birth weights were, respectively, 1427±432 and 1425±466 g. Outcome variables included mortality (NY Acad Sciences 1982, table I, page 454; with one additional later death in the treated group and three in the control group [page 458]), bronchopulmonary dysplasia (diagnosis based on clinical, radiologic and/or pathologic evidence), patent ductus arteriosus, patent ductus arteriosus requiring therapy, intraventricular hemorrhage, and retrolental fibroplasia (numerator: Table 5A (revised), Ehrenkranz, Ophthalmol 1982; 89: page 988; denominator: Ehrenkranz, Ann NY Acad Sci 1982, table I, page 454; data for very low birth weight: Hittner, NEJM 1982 page 868, Figure 1). A stage II retrolental fibroplasia in this study corresponds to a stage III from the current classification (NEJM 1982).

Ferlin 1998 is a randomized trial assessing the effect of oral supplementation with 25 IU/day of vitamin E and 4 mg/kg/day of elemental iron (provided as iron sulfate) on hematologic parameters in preterm infants. Entry criteria included: birth weight less than or equal to 1600 g and gestational age less than or equal to 35 weeks, informed consent, stable clinical course, no blood transfusion, regular follow‐up visits after discharge, and absence of interfering events that would require discontinuation of medications. Patients were randomly allocated into one of four groups: group I: placebo started on the fifteenth day of life, iron starting at two months of age; group II: 4 mg/kg/day iron started on the fifteenth day of life; group III: 4 mg/kg/day iron and 25 IU/day alpha‐tocopherol (Roche) orally started on the fifteenth day of life; group IV: 25 IU/day vitamin E started on the fifteenth day of life. All infants received 1.5 IU vitamin E/day and 0.1 mg iron (in multivitamin preparation) starting on the seventh day of life. Iron supplementation was started at 2 months of age in patients who had not been supplemented with iron yet. The 40 patients enrolled in the study were distributed equally among the four groups. Median birth weights and quartiles were 1395 g (1210, 1510), 1315 g (1210,1430), 1310 g (1290,1420) and 1385 g (1260,1540), respectively, in groups I through IV. For this systematic review we compared group III (treatment) with group II (control) and group IV (treatment) with group I (control). The only analyzable outcome variable is hemoglobin concentration at two months of age; for this purpose, the individual data from Figure 2 was used.

Finer 1982 is a randomized trial assessing the effect of serial intramuscular injections of vitamin E on neonatal outcomes in preterm infants. Entry criteria included: birth weight between 750 and 1500 g, appropriate for gestational age, and informed consent. Infants in the treated group received 25 mg vitamin E, two doses, administered intramuscularly respectively within 12 hours of birth and 12 hours afterwards; then 20 mg intramuscularly daily for 14 days, then 20 mg intramuscularly every three days for five doses; then either 100 U daily orally or (if feeds not tolerated) 20 mg intramuscularly every three days. Infants in the control group did not receive any injections. The osmolality of the oral preparation, used as a 1:1 dilution, was 2,025 mOsm/kg. Among 126 infants entered into the trial, 62 were allocated to the treatment group and 64 to the control group. Average birth weights and ranges were 1197 g (820‐1460) in the treatment group and 1207 g (760‐1500) in the control group, and average gestational ages were, respectively, 29.3 weeks (26‐34) and 29.3 weeks (26‐35). Outcome variables included bronchopulmonary dysplasia, patent ductus arteriosus, and necrotizing enterocolitis. Outcomes described in some Infants only include: acute or cicatricial stage of retrolental fibroplasia [Lancet 1982, Table V, page 1089] (patients surviving at least one month) and blindness from retrolental fibroplasia (patients examined).

Fischer 1987 is a double‐blind randomized trial assessing the effect of oral administration of d‐alpha‐tocopherol polyethylene glycol‐1000‐succinate, 50 mg/day for three doses on bilirubin and hemoglobin in preterm infants. However, several infants received only 25 mg/day. Entry criteria included birth weight less than 1,500 g, normal size for gestational age, and informed consent. Control infants received placebo. The total number of patients entered into the study was 28, including seventeen in the treatment group and 11 in the control group. Average birth weights were 1220±170 g in the treatment group and 1250±150 g in the control group, and average gestational ages were, respectively, 30±1 weeks and 32±2 weeks. Outcome variables included hemoglobin and bilirubin concentrations.

Fish 1990 is a double‐blind randomized trial assessing the effect of dl‐alpha‐tocopherol (Ephynal®, Hoffmann‐La Roche) intramuscularly on neonatal outcomes in preterm infants. Entry criteria included birth weight equal to or less than 1000 g, postnatal age equal to or less than 24 hours, and informed consent. Exclusion criteria were: extremely poor condition upon admission or major congenital abnormalities. Vitamin E was provided in four intramuscular doses at 15, 10, 10 and 10 mg/kg on days one, two, four and six of life, respectively, in preterm infants. Doses of dl‐alpha‐tocopherol 10 mg/kg were administered intramuscularly every three days if the attending neonatologist felt that the infant could not tolerate oral vitamin E after 7 days of life. Control patients received a placebo. All neonates received oral dl‐alpha‐tocopheryl acetate (Aquasol E) 100 mg/kg/day beginning at admission; the dose was adjusted after day 7 to maintain total serum tocopherol level 0.5‐3.5 mg/dl. The total number of patients entered into the study was 149, of which two were excluded after entry, 73 were in the treatment group and 74 in the control group. Average birth weights in the treatment group and in the control group were, respectively, 782±70 g and 783±76 g, and average gestational ages were, respectively, 26±1 weeks and 26±1 weeks. Outcome variables included mortality, intracranial hemorrhage, necrotizing enterocolitis, sepsis, patent ductus arteriosus, and local reaction at injection site.

Graeber 1977 is a randomized trial assessing the effect of vitamin E supplementation (2 doses tested) and that of iron supplementation on neonatal outcomes in preterm infants. Entry criteria included birth weight less than 1500 g and gestational age less than 34 weeks. Exclusion criteria included ABO or Rh isoimmunization, hemoglobin value less than 13 g/dl at seven days of age, transfusion of whole blood or plasma at any age, transfusion of red blood cells after one week of age. Patients were divided into six groups according to the amount of vitamin E (none, 100 , 125 or 150 mg/kg intramuscularly) and the amount of iron‐dextran complex (none or 25 mg intramuscularly on days 7, 14, 21 and 28). Each vitamin E dose was divided into four doses on days one and two, and days seven and eight, respectively. For this systematic review we regrouped patients according to vitamin E allocation. Among 35 infants entered into the study twenty were randomized to vitamin E supplementation (treatment) and fifteen to no supplementation (control). Birth weight ranged from 940 to 1470 g (mean 1270 g). Outcomes included bronchopulmonary dysplasia and retrolental fibroplasia.

Gross 1977 is a randomized trial assessing the effect of intramuscular administration of a total dose of 125 mg/kg tocopheryl acetate (Roche) on hemoglobin concentration in preterm infants. Entry criteria included birth weight less than 2,500 g, gestational age less than 36 weeks, well, breathing room air, and hemoglobin concentration greater than 13 g/dl on the third day of life. Exclusion criterion was either Rh or ABO isoimmunization. Vitamin E was given in eight divided doses, an injection in each thigh, on days four, five, six and seven of life. Controls received no injection. Neither group received iron supplementation. The twenty patients entered into the study were equally divided into the treatment group and the control group. Mean gestational ages in the treatment group and in the control group were, respectively, 33.5±1.7 and 33.2±1.4 weeks, and mean birth weights were, respectively, 1717±176 and 1631±397 g. The outcome variable was hemoglobin concentration.

Gross 1979 is a randomized controlled trial assessing the effect of intramuscular administration of a total dose of 50 mg/kg dl‐alpha‐tocopheryl acetate (Roche) administered intramuscularly in 6 divided doses, an injection in each thigh on days one, two and three of life. Entry criteria were birth weight between 1000 and 2000 g, and appropriate for gestational age. Exclusion criteria included hemoglobin concentration less than 13 g/dl, and Rh or ABO isoimmunization. Infants in the control group received no injection. Neither group received iron supplementation. Among the 40 patients entered into the study, twenty were assigned to the treatment group and twenty to the control group. Eligible infants had a birth weight between 1000 and 2000 g and were appropriate for gestational age; average weight values are not provided. The outcome variable we used for this review was serum bilirubin concentration on day four of life.

Hittner 1981 is a randomized trial assessing the effects of enteral administration of 100 mg/kg/day of dl‐alpha‐tocopheryl acetate (Aquasol E®, preparation containing Tween 80, propylene glycol, sorbitol, oil of anise, sodium saccharin, imitation butterscotch flavor) from the first day of life until discharge. Entry criteria included birth weight less than or equal to 1500 g, requiring oxygen for respiratory distress, admitted to the unit within the first 24 hours of life, and informed consent. Exclusion criteria included necrotizing enterocolitis preventing oral administration of vitamin E (n = 1). Patients who died within four weeks (n = 48) were excluded from all other outcomes. Osmolality of the enteral vitamin E preparation was 3000 mOsm/L (Hittner 1984). Control infants received a placebo solution. The total number of patients entered into the study was 150, including 75 allocated to the treatment group and 74 to the control group. Mean gestational ages in the treatment group and in the control group were, respectively, 29.3±2.2 and 29.4±2.0 weeks, and mean birth weights 1093±245 and 1113±245 g. Outcome variables included (1) mortality (data from NEJM 1981, Figure 1, page 1367), and (2) among patients surviving at least four weeks and without necrotizing enterocolitis: bronchopulmonary dysplasia, patent ductus arteriosus, sepsis, periventricular‐intraventricular hemorrhage and retrolental fibroplasia.

Hittner 1984 is a randomized trial assessing the effect of intramuscular administration of dl‐alpha‐tocopherol, 55 mg/ml (Ephynal®, Hoffman‐La Roche, preparation containing benzyl alcohol, propylene glycol, ethyl alcohol, Emulphor E1‐620, and buffering salts) at doses of 15, 10, 10, and 10 mg/kg on days 1, 2, 4 and 6 of life, respectively, in preterm infants. Entry criteria included birth weight equal to or less than 1,500 g, oxygen requirement for respiratory distress, admission within 24 hours, and informed consent. Patients who died before 10 weeks of life (n = 33) were excluded from all other outcomes. Controls received placebo injections. Vitamin E 100 mg/kg/day was given orally to all infants starting within the first 24 hours of life and until vascularization of the retina was complete. Oral vitamin E was provided as an emulsion of dl‐alpha‐tocopheryl acetate in gelatin, silicon dioxide, polysorbate 80, and medium chain triglycerides; the osmolality was 150 mOsm/kg. Among the 168 patients entered into the study, 79 were allocated to the treatment group and 89 to the control group. Average gestational ages were 28.7±1.9 and 29.4±2.0 weeks, respectively, in the experimental group and in the control group, and average birth weights were, respectively, 1091±233 and 1113±245 g. Outcome variables included (1) mortality (Pediatrics 1984, Table 4 and Figure 1) and (2) among those surviving ten weeks: bronchopulmonary dysplasia, patent ductus arteriosus, sepsis, periventricular‐intraventricular hemorrhage, retrolental fibroplasia, necrotizing enterocolitis, number of transfusions and amount of blood transfused (primary reference).

Jansson 1978 is a randomized trial assessing the effect of iron and vitamin E on hematologic parameters in preterm infants. Entry criterion was a birth weight less than 2500 g. Patients who died before ten weeks (n =33) were excluded from all other outcomes. Patients were randomly allocated into three groups. Group I (control) received ferrous succinate starting at three weeks of age (elemental iron 2‐3 mg/kg/day); group II received dl‐alpha‐tocopheryl acetate (E‐vitamin®, AB ACO, dispersible in lipid solutions) 15 mg/day started at 10 days and ferrous succinate started at 3 weeks; and group III received tocopheryl acetate 15 mg/day started at 10 days and ferrous succinate started at 10 weeks. Among the 57 enrolled patients, 24 were allocated to Group I, 17 to group II and 16 to group III. For this systematic review we compared groups I and II, in which iron was initiated at the same postnatal age. Average birth weights in groups I (control) and II (treatment) were 1899 and 1767 g for those < 2 kg and 2135 and 2330 g for those 2.000‐2.499 kg, respectively (variability not provided). Mean gestational ages were, respectively, 34.6 and 35 weeks for those with a birth weight less than 2 kg, and 37.5 and 36.4 weeks for those with a birth weight between 2.000 and 2.499 kg. Outcome variables included hemoglobin concentration, reticulocyte count and platelet count.

Johnson 1989 is a double‐blind randomized trial assessing the effect of parenteral vitamin E (under a FDA‐approved IND) on retinopathy of prematurity in preterm infants. Entry criteria included: (1) birth weight equal to or less than 2000 g or (2) gestational age equal to or less than 36 weeks and oxygen treatment required; informed consent, and admission within 5 days of age. Treatment included: (1) for infants receiving intravenous feeding: 15 mg/kg vitamin E intramuscularly and 15 mg/kg intravenously (given at a 1:100 dilution over an 8‐hour period) at study admission, followed by 15 mg/kg intravenously the next day; (2) for infants receiving oral feedings: 15 mg/kg vitamin E intramuscularly and 100 mg/kg/day divided in aliquots in the feeds. Subsequent investigational vitamin E dosage was provided until retinal vascularization was complete or active ROP had subsided, to maintain serum tocopherol level of 5 mg/dl. Parenteral vitamin E was provided as free dl‐alpha‐tocopherol 50 mg/ml (55 IU/ml) emulsified with Emulfor E1‐620. Enteral vitamin E was provided as free dl‐alpha tocopherol 50 mg/ml with 180 mg/ml polysorbate 80, and 700 mg/ml propylene glycol 80. Controls received placebo both for oral and parenteral preparations, with dosage adjustments to mimic those in the control group. In infants in whom severe ROP had developed, the study medication was replaced with vitamin E. Among 914 infants enrolled into the study, 454 were allocated to the treatment group and 460 to the control group. Average birth weight was 1461±457 g in the treatment group and 1444±439 g in the controls. Average gestational age was 31.5±2.7 weeks in the treatment group and 31.6±2.8 weeks in the controls. Age at study entry was 1.6±1.2 days for patients in the treatment group and 1.5±1.0 days for controls. The only outcome variables available in the whole study population (Birth Defects 1988, page 226) was intraventricular hemorrhage. Examination for the presence of retinopathy of prematurity was done in 755 patients. Outcome variables available among the 545 infants with birth weight less than or equal to 1500 g (Pediatrics 1985 & 1989) included mortality, bronchopulmonary dysplasia (diagnosed radiographically), patent ductus arteriosus, patent ductus arteriosus requiring treatment, sepsis after study entry, intraventricular hemorrhage (Birth Defects page 226), retinopathy of prematurity, and necrotizing enterocolitis. Outcome variables available in subsets of the population included sepsis after study entry among very low birth weight infants treated for more than one week, bilirubin (excluding hemolytic anemia, polycythemia, and prior transfusion: Pediatrics 1980), and amount of blood transfusion (subset of very low birth weight infants: Pediatrics 1989).

Melhorn 1971 is a randomized trial assessing the effect of iron and vitamin E on hemoglobin concentration in infants admitted to the premature nursery. Exclusion criteria included: either small or large for gestational age, hemoglobin concentration < 14 g/dl in the first 24 hours of life, blood group incompatibility, hemoglobinopathy, red blood cell enzyme deficiency, infection, defects requiring surgical intervention. After stratification for gestational age and birth weight within the first 48 hours of life, patients were randomized to one of four groups: Group I: no supplement (control); Group II: vitamin E: alpha tocopheryl acetate 25 units/day orally from the eighth to the forty‐second day of life; group III: iron from the fifteenth to forty‐second day of life: 10 mg/day if weight < 1500 g, 15 mg/day if weight 1501‐2000 g, 20 mg/day if weight > 2001 g; group IV: iron and vitamin E. Among the 234 enrolled into the study, data are available in 186, including 45 randomized to group I, 47 to group II, 44 to group III and 50 to group IV. In category A, mean gestational age in each ranged from 30 to 31 weeks and mean birth weight ranged from 1170 grams to 1280 grams. In category B, mean gestational age was 34 weeks in each group, and mean birth weight ranged from 1700 grams to 1775 grams. In category C, mean gestational age in each group ranged from 37 to 38 weeks and mean birth weight from 2168 grams to 2243 grams. For this review, the effect of vitamin E was assessed by comparing group II with group I (no iron) and group IV with group III (iron). Outcome variables in categories A and B included hemoglobin at ten weeks of age and reticulocyte count.

Pathak 2003 is a double‐blind randomized trial assessing the effect of enteral administration of vitamin E 50 IU/day on hemoglobin concentration and reticulocyte count in very low birth weight infants receiving erythropoietin and iron. Entry criteria included: gestational age less than or equal to 32 weeks, birth weight less than or equal to 1250 grams, clinically stable (less than 7.5 ml/week phlebotomy, oxygen requirement less than or equal to 35%, intermittent mandatory ventilation less than or equal to 25 breaths per minute, and mean airway pressure less than or equal to 8 cm of water). Exclusion criteria included: a disease involving any major organ system, life‐threatening congenital malformations or sepsis, isoimmunization with clinically apparent hemolytic anemia, hemolytic disorder, and intraventricular hemorrhage grade III or greater. Controls received placebo. All patients entered into the study were receiving erythropoietin 100 units/kg/day five days a week for eight weeks or until discharge, whichever came first. Infants fed intravenously received a multivitamin preparation (MVI Pediatric®) containing 1.4 mg alpha‐tocopheryl acetate/ml at a dose of 1.5 ml/day and 3.25 ml/day to infants weighing < 1 kg and > 1 kg, respectively. All infants fed enterally at least 40 ml/kg/day received oral iron 6 mg/kg/day and 1 ml of a multivitamin preparation containing 5 U/ml vitamin E. Among 30 patients entered into the study, 15 were allocated to the treatment group and 15 to the control group. Average birth weights in the treatment group and in the control group were, respectively, 928±170 grams and 899±159 grams, average gestational ages were, respectively, 27.9±1.9 and 27.9±1.5 weeks, and postnatal ages at the time of study entry were, respectively, 24.4±15.5 days and 28.1±10.1 days. Outcome variables included mortality, patent ductus arteriosus requiring treatment, sepsis, necrotizing enterocolitis, hemoglobin concentration, reticulocyte count, number of transfusions, and total volume of blood transfused.

Phelps 1987a is a double‐blind randomized trial assessing the effect of intravenous administration of a dose of 20 mg/kg of an investigational drug [nonesterified dl‐alpha‐tocopherol in alcohol or its vehicle (10% ethyl alcohol, 10% propylene glycol, 10% Emulphor, 1% benzyl alcohol, and buffering salts)] provided within 24 hours of birth and repeated on day two (and sometimes day three). Entry criteria included: birth weight less than 1500 g or gestational age less than 33 weeks, less than 24 hours of age, free from recognized congenital anomalies of the eyes or congenital anomalies incompatible with survival, and informed consent. Doses were adjusted to achieve plasma tocopherol levels of 3‐3.5 mg/dl. Average daily intramuscular dose was 14.4±4.7 mg/kg/day during the first week, and ranged between 4.1 and 7.0 mg/kg/day during the subsequent weeks. Infants fed enterally received enteral vitamin E (free tocopherol), which was started at a dose of 100 mg/kg/day and adjusted according to the serum level. Supplemental intramuscular injections were given to infants with plasma levels lower than 2.5 mg/dl despite oral doses exceeding 200 mg/kg/day. Average oral dose (after the initial parenteral administration) was 69±66 mg/kg/day; 56 of 108 patients with oral vitamin E administration received intramuscular supplementation (one to four doses). Control infants received placebo. All infants fed intravenously received 1.2 mg/kg/day of tocopheryl acetate in standard multivitamin supplements. Controls received therapeutic vitamin E if they developed hemolytic anemia associated with a plasma level of tocopherol less than 0.5 mg/dl. Infants requiring surgery or chest tube placement were given vitamin A 2500 IU/day for a week. The total number of patients entered into the study was 287, including 140 in the treatment group and 147 in the control group. Average birth weights in the treatment group and in the control group were, respectively, 1181±307 grams and 1205±319 grams, and average gestational ages were, respectively, 29.5±2.5 weeks and 29.9±1.5 weeks. Outcome variables included (1) among all infants: mortality, sepsis after 48 hours, periventricular‐intraventricular hemorrhage and necrotizing enterocolitis, (2) among 232 infants with ophthalmologic examination: retinal hemorrhage, and (3) among 196 surviving infants: retinopathy of prematurity.

Rönnholm 1989 is a randomized trial assessing the effect of intramuscular administration of dU‐rac‐alpha‐tocopheryl acetate 20 mg/kg/day (Ephynal®, Hoffmann‐La Roche) during the first three days of life in preterm infants. The entry criterion was a birth weight less than or equal to 1520 g. Patients in both groups received oral administration of water soluble vitamin E dl‐alpha‐tocopheropolyethylene glycolol 1000‐succinate (part of a multivitamin preparation, Ekavitol®, Orion Corp Ltd) in increasing doses of 1.6 mg/day on day 3 to 10 mg/day at age 2‐12 weeks. Controls received oral vitamin E only. Iron supplementation was started gradually at two weeks of age, and increased until six weeks of age to reach four mg/kg/day if birth weight was <= 1000 grams and 3 mg/kg/day if birth weight was > 1000 grams. Vitamin A was supplemented as retinol palmitate (1500 IU/day), starting at twelve weeks of postnatal age, or approximately 42 weeks of postmenstrual age, i.e., after all outcomes of interest had already been assessed. Patients were also randomly supplemented with human milk protein, medium‐chain triglycerides, both or neither. The total number of patients entered into the study was 54, including 23 allocated to the treatment group and 31 to the control group. Absolutely no data are provided about the three patients who died from respiratory distress syndrome or from cerebral hemorrhage. Average birth weights in the treatment group and in the control group were, respectively, 1153±245 grams and 1222±228 grams, and average gestational ages were, respectively, 29.9±1.9 weeks and 30.4±2.1 weeks. Outcome variables among the 51 survivors included bronchopulmonary dysplasia, intraventricular hemorrhage, retrolental fibroplasia, and necrotizing enterocolitis.

Rudolph 1989 is a randomized controlled trial assessing the effect of intramuscular administration of dl‐alpha‐tocopherol (Hoffman‐La Roche) at dose of 25 mg/kg within 12 hours of birth, with repeated injections 24 hours later, and at 4, 5, 10, 20 and 30 days of age, yielding a total cumulative dose of 175 mg/kg on neonatal outcome. Entry criteria included: preterm delivery, birth weight 750‐1500 g, and intensive care. Controls received no vitamin E supplementation. All infants fed intravenously received 2.5 IU vitamin E per liter of parenteral solution. Among the 29 patients entered into the study, thirteen were allocated to the treatment group and sixteen to the control group. Average birth weights in the treatment group and in the control group were, respectively, 1202±198 grams and 1187±198 grams, and average gestational ages were, respectively, 30.7±2.3 weeks and 30.5±1.6 weeks. Outcome variables included mortality, patent ductus arteriosus (assessed clinically), severe retinopathy of prematurity and necrotizing enterocolitis.

Saldanha 1982 is a randomized controlled trial assessing the effects on intramuscular administration of 25 mg of dl‐alpha‐tocopherol (Hoffman‐La Roche) upon entry into the study and daily thereafter until serum tocopherol level rose above 2‐4 mg/dl or until oxygen was no longer required. Entry criteria included gestational age less than 37 weeks, hyaline membrane disease, oxygen requirement at least 60% in first 24 hours or 80% thereafter, and informed consent. Exclusion criteria included cyanotic heart disease, multiple malformations, and known infectious disease. Controls received no injection. The total number of patients entered into the study was 44, including 21 in the treatment group and 23 in the control group. Average birth weights in the treatment group and in the control group were, respectively, 1458±491 grams and 1554±557 grams. Average gestational ages among patients surviving ten days were, respectively, 32.4±2.1 weeks and 32.9±2.9 weeks. Outcome variables among patients surviving ten days or more included: bronchopulmonary dysplasia and patent ductus arteriosus (diagnosed clinically).

Schiller 1980 is a randomized trial assessing the effect of intramuscular administration of dl‐alpha‐tocopherol (175 mg/kg over a 30‐day period) on patent ductus arteriosus. Entry criteria were: preterm infant, and birth weight < 1750 g. Time of onset is not provided. Controls did not receive any injection. Among the 29 infants entered into the study, sixteen were allocated to the treatment group, and thirteen to the control group. The outcome variable was patent ductus arteriosus.

Sinha 1987 is a randomized trial assessing the effect of intramuscular administration of alpha‐tocopheryl acetate (Ephynal®, Hoffman‐La Roche) 20 mg/kg daily for three doses commencing within two hours after randomization (first day of life). Entry criteria included gestational age equal or less than 32 weeks, admitted from January 1984 to September 1985. No exclusion criteria were used before randomization; however, patients randomized before delivery and found after birth to have lethal malformations were excluded from the study. Controls received no placebo. The mothers of some infants were part of another randomized trial: they were allocated either to vitamin E 400 mg every 4‐6 hours or to placebo capsules during preterm labor. Among the 231 patients entered into the study, no data are provided on the three infants with lethal malformations, so that the total number of patients retained in the study is 228 (presumably 115 treated patients and 113 controls, based on totals in Table II plus numbers of excluded patients in the second paragraph of the discussion). Only partial data are provided in 18 patients with a hemorrhage on the initial ultrasonography obtained within two hours of life or (if out born) upon admission to the unit. In the manuscript, several data are presented only in the 210 patients with negative initial scan (102 patients in the treatment group and 108 in the control group), including birth weights in the treatment group and in the control group (respectively, 1273±315 grams and 1214±346 grams), gestational ages (respectively, 29.3±2.0 weeks and 28.8±2.2 weeks), and complete information about the classification of the intracranial hemorrhage. The authors used a classification in 3 grades; therefore, grade 3 was re‐coded into Papile's grade IV (intraparenchymal hemorrhage). Outcome variables available in all patients included mortality, periventricular‐intraventricular hemorrhage (adding numbers in Table II and those in the second paragraph of the discussion), necrotizing enterocolitis, and reaction at site of injection.

Smith 1985 is a double‐blind randomized trial assessing the effect of oral administration of dl‐alpha‐tocopheryl polyethylene glycol‐succinate (Mead Johnson) 50 mg/day once a day for three consecutive days; the first dose was given before 24 hours of life. Entry criteria included: gestational age 30‐36 weeks, birth weight 970‐2610 g, healthy, lack of pulmonary disease or oxygen requirement, no ABO or Rh isoimmunization, and no other sources of bilirubin production, such as hematoma or bruising. The only exclusion criterion was infection. Control patients received placebo. All infants were fed either breast milk or a proprietary formula without iron (vitamin E contents ranging from 12 to 25 IU/L). Decisions regarding phototherapy were made by staff physicians who were unaware of the study results. Among 30 patients entered into the study, seventeen were allocated to the treatment group and thirteen to the control group. Average birth weights were 1.79±0.38 kg in the treatment group and 1.83±0.33 kg in the control group; average gestational ages were 32.6±1.6 weeks and 33.4±1.5 weeks, respectively. Outcome variables included bilirubin concentration and hemoglobin concentration.

Zipursky 1987 is a double‐blind randomized trial assessing the effect of enteral (gavage) supplementation of 25 IU vitamin E daily provided as 16 mg dl‐alpha‐tocopherol (Hoffman‐La Roche) for six weeks in preterm infants smaller than 1500 g on anemia, coagulation tests and serum tocopherol levels. Entry criteria included birth weight < 1500 g, expected survival greater than 48 hours, and informed consent. Exclusion criteria included major congenital anomalies and Rhesus hemolytic disease. Infants in the control group received placebo (Tween 80). Age at entry into the trial was 2.7±3 days in treatment group and 2.9±2 days in control group. Infants in both groups also received recommended amounts of vitamin E, either in breast milk (5.4 IU/L and 0.8 mg/g of PUFA), SMA 20 (10 IU/L or 1.3 mg/g of PUFA) or SMA 24 (12 IU/L or 1.3 mg/g of PUFA). The range of PUFA content in milk used was 4‐6 g/L. No added iron was provided, although SMA 20 and SMA 24 contained, respectively, 12.7 and 15.2 mg/L of iron sulfate. Individual or group amounts of feeding are not provided, so that total amount of iron cannot be calculated. The total number of patients is not the same in the two follow‐up manuscripts (266 in Milner 1981 and 268 in Watts 1991). Assuming that these differences represent loss to follow‐up (two patients lost to follow‐up acknowledged in Milner 1981), the total number of patients is presumed to be 269 infants entered into the study, with 135 allocated to the treatment group and 134 to the control group. Average birth weights were 1149± 244 g and 1158±251 g in the treatment group and in the control group, respectively, and average gestational ages were 29.2±2.7 and 29.1±2.7 weeks, respectively. Mean age at entry into the trial was 2.7±3 days in the treatment group and 2.9±2 days in the control group. Baseline characteristics (number of infants < 1000 g, percentage of infants with severe illness) were similar in the two groups. Outcome variables included mortality and stages III or IV bronchopulmonary dysplasia (defined, respectively, by bilateral cystic lesions without or with fibrosis) and retrolental fibroplasia, and among specific subsets, hemoglobin concentration, reticulocyte count, platelet count, and coagulation tests.

2. Vitamin E versus another type or route of vitamin E formulation 
 Only one study was in this group. Bonati 1991 compared between two formulations of intramuscular vitamin E the amount of drug absorption and clinical outcomes in live born infants with gestational age less or equal to 32 weeks. Entry criteria included: live born infant, gestational age equal or less than 32 weeks, no malformations, born in participating centers over 3 months, and admitted to a neonatal intensive care within two hours of birth. Patients received three daily doses of 20 mg/kg of vitamin E starting at 8 hours of life. Infants were randomly allocated to the treatment group, receiving dl‐alpha‐tocopheryl acetate as an aqueous colloidal solution (Ephynal®, Hoffmann‐La Roche, Basel, Switzerland), or to the control group, receiving alpha‐tocopherol in olive oil solution (Evio Forte®, Bracco, Milan). The exact amounts of vitamin E administered enterally and by transfusion are not provided. Infants received 2‐98 ml/kg/day of enteral feeding, including breast milk (vitamin E content 1.11±0.53 mg/100 ml) or formula (vitamin E content 0.23±0.08 mg/100 ml). Among 50 enrolled patients, 44 (22 in each group) completed the study. Birth weights ranged from 670 g to 1800 g with an average of 1217±311 g. Average gestational age was 29.0±2.5 weeks.

Risk of bias in included studies

Among the 26 randomized clinical trials selected for this systematic review, ten were double‐blinded.

1. Vitamin E versus placebo 
 Chiswick 1982: No information is provided about the method of randomization or about blinding of allocation. Since patients in the control group did not receive intramuscular injection of vitamin E, intervention and outcome were not blinded. Although information is provided on all patients entered into the study, the exact number of patients with intraventricular hemorrhage is not provided, because not all patients had a computerized tomography or an autopsy. The authors assumed that all patients who died had an intraventricular hemorrhage; in one case, neither computerized tomography nor autopsy was available to confirm this assumption. For the purpose of this review, only confirmed cases of intraventricular hemorrhage were used.

Chiswick 1983: No information is provided about the method of randomization or about blinding of allocation. Since patients in the control group did not receive intramuscular injection of vitamin E, intervention and outcome were not blinded. The number of patients with intraventricular hemorrhage may have been underestimated because not all patients had ultrasonograms performed after 3 days of age.

Cruz 1983: No information is provided about the method of randomization or about blinding of allocation. Since patients in the control group did not receive intramuscular injection of vitamin E, intervention and outcome were not blinded. Follow‐up was complete.

Ehrenkranz 1982 is a double‐blind randomized controlled trial. The method of randomization is not provided. Allocation, intervention and outcome were blinded. Follow‐up was complete; however, assessment for intraventricular hemorrhage by ultrasonogram or computerized tomography was obtained in only 27 of 47 patients in the experimental group and 39 of 53 in the control group. Therefore, the rate of intraventricular hemorrhage may have been underestimated.

Ferlin 1998 is a randomized controlled trial with four parallel arms. The methods for randomization and blinding of allocation are not provided. Intervention and outcome were blinded. Completeness of follow‐up is unclear; by design, only patients with regular follow‐up visits after discharge were included.

Finer 1982: For randomization the authors used sealed envelopes, and stratification by birth weight and by severity of respiratory distress (requirement for oxygen and endotracheal intubation). Since controls did not receive placebo injections, intervention was not blinded. In contrast, the main outcome variable for the trial (retrolental fibroplasia) was blinded. Most data are available only for the 99 infants who completed the trial, i.e., survived for at least one month.

Fischer 1987 is a double‐blind randomized trial. The method of randomization is not provided. Intervention and outcome were blinded. Follow‐up was complete.

Fish 1990 is a double‐blind randomized trial. Randomization was performed by the pharmacy after stratification by birth weight. Although a placebo was given to the control group, the intervention was not entirely blinded because tocopherol levels were available to the clinicians, who were adjusting the dose accordingly after day seven of life. Two infants were excluded from statistical analysis because of initial inapparent structural anomalies (congenital heart disease and hydranencephaly). Outcome was blinded.

Graeber 1977: The method of randomization is not provided. Intervention and outcome were not blinded. Follow‐up was complete.

Gross 1977: The method of randomization is not provided. Intervention and outcome were not blinded. Follow‐up was complete.

Gross 1979: The method of randomization is not provided; however, randomization was stratified by weight (1000‐1500 g and 1501‐2000 g). Neither intervention, nor outcome were blinded. Follow‐up was complete.

Hittner 1981 is a double‐blind randomized trial. Randomization was performed using a random‐number table, after stratification by weight. Allocation, intervention and outcome were blinded. Follow‐up was not complete: information on 49 patients is incomplete. One patient with necrotizing enterocolitis was excluded (no information on treatment allocation), and one patient was excluded for necrotizing enterocolitis preventing oral administration of vitamin E (n=1). Patients who died before four weeks (n=48) were excluded from all other outcomes.

Hittner 1984 is a double‐masked randomized trial. Randomization was conducted using a random‐number table, after stratification by weight. Allocation, intervention and outcome were blinded. Mortality was reported only until ten weeks of life. None of the other outcomes are reported for the 33 patients who died before 10 weeks.

Jansson 1978: The method of randomization was not provided; randomization was stratified by weight (1000‐1999 and 2000‐2499 g). Neither treatment allocation, nor intervention, nor outcome was blinded. All patients were followed.

Johnson 1989 is a double‐blind randomized controlled trial. The method of randomization is not provided. Randomization was stratified by hospital and by birth weight. Allocation, intervention and outcome were blinded. Complete data are available in only 755 of 914 enrolled.

Melhorn 1971: The method of randomization is not provided. Randomization was stratified by weight and gestational age. Data are provided on only 186 patients. Whether allocation was blinded is unclear. Neither intervention nor outcome were blinded.

Pathak 2003 is a randomized double‐blind trial. Randomization was done using sealed envelopes in the pharmacy. Allocation, intervention, and outcome were blinded. Follow‐up was complete.

Phelps 1987a is a randomized double‐blind trial. Randomization was performed using sealed envelopes stratified by study center, sex, and birth weight, opened at the pharmacy (center 1) or drug control center (center 2). Numbers were balanced by groups of eight. Multiple births were randomized separately so that one of each pair received placebo and one tocopherol. Allocation, intervention and outcome were blinded. Follow‐up was complete for several neonatal outcomes. However retinal examinations were not obtained in 55 infants, including nine infants lost to follow‐up and 46 who had died. Among the 80 infants who developed retinopathy of prematurity, 53 completed the study, eighteen were lost to follow‐up and nine died with active retinopathy of prematurity.

Rönnholm 1989: The method of randomization is not provided, so that blinding of allocation cannot be assessed. Neither intervention, nor outcome was blinded. Three patients died of respiratory distress or cerebral hemorrhage and were excluded from the analysis.

Rudolph 1989: The method of randomization is not provided. Blinding of allocation cannot be assessed. The intervention was not blinded, but the outcome, assessed by a cardiologist, was blinded. Follow‐up was complete.

Saldanha 1982: Randomization was performed by drawing of a card from a randomized deck. Neither allocation, nor intervention, nor outcome was blinded. Incomplete data provided on nine babies who died before 10 days of age.

Schiller 1980: The method of randomization is not provided. Blinding of allocation cannot be assessed. The intervention was not blinded, whereas outcome was blinded. Whether follow‐up was complete is unclear.

Sinha 1987 is a randomized trial using sealed envelopes and thus allocation was blinded. However, controls did not receive a placebo, so that intervention was not blinded. Whether outcome was blinded is unclear. Three inborn infants randomized before birth had lethal major malformations and were excluded from the study. Eighteen infants had a periventricular‐intraventricular hemorrhage on the initial scan; for these infants, only partial information is provided about the severity of the bleeding during the study.

Smith 1985 is a randomized double‐blind trial, so that allocation, intervention and outcome should be blinded; nevertheless, no information is provided about the method of randomization. Follow‐up was complete.

Zipursky 1987 is a double‐blind randomized controlled trial in which the authors used a computerized randomization generated in the pharmacy and stratified by birth weight and by severity of illness. Allocation, intervention and outcome were blinded. Information on chronic lung disease was available in 266 of 269 enrolled infants, while only 225 infants had an eye examination. In the treatment group, hemoglobin levels, reticulocyte counts, and platelet counts were available in 44, 31, and 32 infants, respectively, at 6 weeks of age. In the control group, hemoglobin levels, reticulocyte counts, and platelet counts were available in 46, 33, and 36 infants, respectively, at 6 weeks.

2. Vitamin E versus another type or route of vitamin E formulation 
 Bonati 1991: No information is provided about method of randomization of the first baby within each center. Subsequent infants were given alternating medication, so that allocation, intervention and outcomes were not blinded. In each participating center, the first enrolled infant was randomly allocated to one of the two formulations; allocation was alternated in each subsequent infant. Among 50 patients in this study, only 44 completed the three‐day schedule of vitamin E; no data are provided on the other infants. Outcome variables included mortality, patent ductus arteriosus, bronchopulmonary dysplasia, necrotizing enterocolitis, intraventricular hemorrhage, sepsis, phototherapy, exchange transfusion and tocopherol levels.

Effects of interventions

No relevant data for the following subgroup analyses was found: gestational age >= 29 weeks, gestational age < 29 weeks, parenteral vitamin E in colloidal form with polysorbate, selenium fortification, and vitamin E supplementation started after four weeks of life (although controls in Pathak 2003 were started on placebo at an average age of 28 days).

VITAMIN E VERSUS PLACEBO OR NO TREATMENT (COMPARISON 01): 
 Primary outcomes: 
 Mortlaity:

Mortality until discharge (Outcome 01.01): 
 Mortality was assessed in twelve studies, involving a total of 994 treated patients and 1034 controls. Vitamin E supplementation did not significantly affect mortality (typical estimate, relative risk [RR] 0.97, confidence interval [CI] 0.83, 1.14; risk difference [RD] ‐0.01, CI ‐0.04, +0.03).

Subgroup analyses: data available for analysis by birth weight (see section 01.02), by route of administration, total dose, serum level, onset and duration of therapy in the treatment group, by dose of vitamin E in the control group, and by intake of iron and PUFA. Vitamin E supplementation did not significantly affect mortality in any of the analyses.

Mortality until discharge among very low birth weight infants (Outcome 01.02): 
 Mortality was assessed in seven studies, involving a total of 628 treated patients and 648 controls. Vitamin E supplementation did not significantly affect mortality (typical estimate RR 0.97, CI 0.80, 1.16; RD ‐0.01, CI ‐0.05, +0.04).

Subgroup analyses: data available for analysis by birth weight, by route of administration, total dose, serum level, onset and duration of therapy in the treatment group, by dose of vitamin E in the control group, and by intake of iron, and PUFA. Vitamin E supplementation did not significantly affect mortality in any of the analyses.

COMBINED OUTCOMES: 
 No trial assessed combined outcome at eighteen months including mortality [mortality, bronchopulmonary dysplasia, blindness, mental retardation or cerebral palsy], nor combined outcome at 18 months excluding mortality [bronchopulmonary dysplasia, blindness, mental retardation or cerebral palsy].

Secondary outcomes: 
 Bronchopulmonary dysplasia:

Bronchopulmonary dysplasia (Outcome 01.03): 
 Bronchopulmonary dysplasia was assessed in six studies, involving a total number of 558 treated patients and 569 controls. Vitamin E supplementation did not significantly affect the risk of bronchopulmonary dysplasia (typical estimate RR 0.91, CI 0.73, 1.14; RD ‐0.02, CI ‐0.07, +0.03).

Subgroup analyses: data available for analysis by birth weight (see section 01.04), by route of administration, total dose, serum level, onset and duration of therapy in the treatment group, by dose of vitamin E in the control group, and by intake of iron, PUFA and vitamin A. Vitamin E supplementation did not significantly affect this outcome in any of the analyses. 
 
 Bronchopulmonary dysplasia among very low birth weight infants (Outcome 01.04): 
 Bronchopulmonary dysplasia was assessed in four studies, involving a total number of 484 treated patients and 488 controls. Vitamin E supplementation did not significantly affect the risk of bronchopulmonary dysplasia (typical estimate RR 0.89, CI 0.71, 1.13; RD ‐0.02, CI ‐0.08, +0.03).

Subgroup analyses: data available for analysis by birth weight, by route of administration, total dose, serum level, onset and duration of therapy in the treatment group, by dose of vitamin E in the control group, and by intake of iron, PUFA and vitamin A. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Bronchopulmonary dysplasia among surviving patients (Outcome 01.05): 
 This outcome was assessed in four studies, involving a total of 159 treated patients and 163 controls. Vitamin E supplementation did not significantly affect the risk of bronchopulmonary dysplasia (typical estimate RR 1.15, CI 0.84, 1.58; RD 0.04, CI ‐0.05, +0.13).

Subgroup analyses: data available for analysis by birth weight (see section 01.06), by route of administration, total dose, serum level, onset and duration of therapy in the treatment group, by dose of vitamin E in the control group, and by intake of iron. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Bronchopulmonary dysplasia among surviving very low birth weight infants (Outcome 01.06): 
 This outcome was assessed in two studies, involving a total of 118 treated patients and 118 controls. Vitamin E supplementation did not significantly affect the risk of bronchopulmonary dysplasia (typical estimate RR 1.07, CI 0.70, 1.64; RD 0.02, CI ‐0.10, +0.13).

Subgroup analyses: data available for analysis by birth weight, by route of administration, total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Radiographic signs of bronchopulmonary dysplasia at six weeks to two months of age (Outcome 01.07): 
 This outcome was analyzed in four studies, involving a total of 176 treated patients and 188 controls. Vitamin E did not significantly affect the risk of radiographic signs of bronchopulmonary dysplasia (typical estimate RR 0.99, CI 0.67, 1.46; RD 0.00, CI ‐0.09, +0.08).

Subgroup analyses: data available for analysis by birth weight (see section 01.08), by route of administration, total dose, serum level, onset and duration of therapy in the treatment group, by dose of vitamin E in the control group, and by intake of iron and PUFA. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Radiographic signs of bronchopulmonary dysplasia at six weeks to two months of age among very low birth weight infants (Outcome 01.08): 
 This outcome was analyzed in two studies, involving a total of 129 treated patients and 135 controls. Vitamin E did not significantly affect the risk of radiographic signs of bronchopulmonary dysplasia (typical estimate RR 0.95, CI 0.61, 1.47; RD ‐0.01, CI ‐0.11, +0.09).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, by dose of vitamin E in the control group, and by intake of iron and PUFA. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Patent ductus arteriosus:

Patent ductus arteriosus (Outcome 01.09): 
 Patent ductus arteriosus was assessed in six studies, involving a total number of 481 treated patients and 495 controls. Vitamin E supplementation did not significantly affect the risk of patent ductus arteriosus (typical estimate RR 1.07, CI 0.93, 1.23; RD 0.03, CI ‐0.03, +0.09).

Subgroup analyses: data available for analysis by birth weight (see section 01.10), by route of administration, total dose, serum level, onset and duration of therapy in the treatment group, by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Patent ductus arteriosus among very low birth weight infants (Outcome 01.10): 
 Patent ductus arteriosus was assessed in four studies, involving a total number of 418 treated patients and 429 controls. Vitamin E supplementation did not significantly affect the risk of patent ductus arteriosus (typical estimate RR 1.09, CI 0.93, 1.28; RD 0.04, CI ‐0.03, +0.10).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Patent ductus arteriosus among surviving patients (at ten days to ten weeks of age) (Outcome 01.11): 
 This outcome was assessed in three, involving a total of 136 treated patients and 135 controls. Vitamin E supplementation did not significantly affect the risk of patent ductus arteriosus (typical estimate RR 0.98, CI 0.70, 1.38; RD ‐0.01, CI ‐0.11, +0.10).

Subgroup analyses: data available for analysis by birth weight (see section 01.12), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Patent ductus arteriosus among surviving very low birth weight infants (at ten days to ten weeks of age) (Outcome 01.12): 
 This outcome was assessed in two studies, involving a total of 118 treated patients and 118 controls. Vitamin E supplementation did not significantly affect the risk of patent ductus arteriosus (typical estimate RR 0.85, CI 0.57, 1.27; RD ‐0.04, CI ‐0.15, +0.07).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Patent ductus arteriosus requiring treatment (Outcome 01.13): 
 This outcome was assessed in three studies, involving a total number of 332 treated patients and 343 controls. Summary statistics showed no significant effect of vitamin E on patent ductus arteriosus (typical estimate RR 1.02, CI 0.79, 1.31; RD 0.00, CI ‐0.06, +0.07) with significant heterogeneity among studies both for RR (chi‐square 5.99, p = 0.05) and RD (chi‐square 6.78, p=0.034). 
 
 Subgroup analyses: data available for analysis by birth weight (see section 01.14), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by iron intake. Vitamin E supplementation significantly reduced the risk for patent ductus arteriosus requiring therapy in the following subgroup analyses: (1) excluding intravenous therapy (01.13.06) (typical estimate RR 0.56, CI 0.32, 0.97; RD ‐0.17, CI ‐0.31, ‐0.02), (2) Duration of treatment greater than one week (01.13.12) (typical estimate RR 0.56, CI 0.32, 0.97; RD ‐0.17, CI ‐0.31, ‐0.02); (3) vitamin E intake greater than 10 mg/100 Kcal in the control group (01.13.14) in a single study (estimate RR 0.54, CI 0.31, 0.95; RD ‐0.22, CI ‐0.40, ‐0.03). Ehrenkranz 1982 was the only study showing significant effect of vitamin E on this outcome; vitamin E was provided intramuscularly at high dose, starting within 48 hours and for more than one week in the treatment group, and vitamin E intake was greater than 10 mg/100 Kcal in the control group.

Patent ductus arteriosus requiring treatment among very low birth weight infants (Outcome 01.14): 
 This outcome was assessed in three studies, involving a total number of 285 treated patients and 290 controls. Summary statistics showed no significant effect of vitamin E on patent ductus arteriosus (typical estimate RR 1.20, CI 0.89, 1.60; RD 0.04, CI ‐0.03, +0.11).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by iron intake. Vitamin E did not significantly affect this outcome in any of the analyses.

Sepsis

Sepsis after study entry (Outcome 01.15): 
 Sepsis was assessed in four studies, involving a total number of 498 treated patients and 511 controls. Summary statistics showed a significant increase in risk of sepsis (typical estimate RR 1.52, CI 1.13, 2.04; RD 0.06, CI 0.02, 0.11). There was significant heterogeneity among studies for RD (chi‐square 8.15, p=0.04) but not for RR (chi‐square 2.89, p=0.41).

Subgroup analyses: data available for analysis by birth weight (see section 01.16), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by iron intake. Vitamin E significantly increased the risk of sepsis in the following subgroup analyses: (1) parenteral with or without enteral (01.13.03) or parenteral with hypertonic enteral formulation at pharmacologic dose (01.15.04) (typical estimate RR 1.57, CI 1.15, 2.14; RD 0.06, CI 0.02, 0.11); (2) Intravenous administration (01.15.05) (typical estimate RR 1.54, CI 1.07, 2.21; RD 0.05, CI 0.01, 0.10). However, there was significant heterogeneity among studies for RD (chi‐square 6.25, p=0.01) but not for RR (chi‐square 2.13, p=0.14); (3) Total dose of vitamin E in the treatment group greater than 30 IU/kg/day (01.15.07) (typical estimate RR 1.57, CI 1.15, 2.14; RD 0.06, CI 0.02, 0.11); (4) Serum tocopherol level in the treatment group greater than 3.5 mg/dl (01.15.09) (typical estimate RR 1.72, CI 1.24, 2.40; RD 0.10, CI 0.04, 0.15); (5) Onset of treatment within 48 hours of life (01.15.10) (typical estimate RR 1.57, CI 1.15, 2.14; RD 0.06, CI 0.02, 0.11); (6) Total dose of vitamin E in the control group equal to or less than 10 mg vit E/100 kcal (01.15.13) (typical estimate RR 1.48, CI 1.05, 2.08; RD 0.05, CI 0.01, 0.10).

Sepsis after study entry among very low birth weight infants (Outcome 01.16): 
 This outcome was assessed in four randomized trials, involving a total number of 400 treated patients and 407 controls. Summary statistics showed a significant increase in risk of sepsis (typical estimate RR 1.53, CI 1.13, 2.08; RD 0.07, CI 0.02, 0.13). 
 
 Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by iron intake. Vitamin E supplementation significantly increased the risk of sepsis in the following subgroups: (1) Birth weight less or equal to 1500 grams (01.16.02) (typical estimate RR 1.65, CI 1.13, 2.40; RD 0.08, CI 0.02, 0.14); (2) Subgroup characterized by parenteral with or without enteral (01.16.05), by total dose of vitamin E in the treatment group greater than 30 IU/kg/day (01.16.09) and by onset of therapy within 48 hours of life (01.16.12) (typical estimate RR 1.59, CI 1.15, 2.18; RD 0.08, CI 0.03, 0.13); (3) Intravenous vitamin E administration (01.16.07) (typical estimate RR 1.56, CI 1.07, 2.27; RD 0.07, CI 0.01, 0.12); (4) Serum level in the treatment group > 3.5 mg/dl (01.16.11) (typical estimate RR 1.72, CI 1.24,2.40; RD 0.10, CI 0.04,0.15); and (5) total dose of vitamin E in the control group <= 10 mg/100 kcal (01.16.15) (typical estimate RR 1.49, CI 1.04, 2.13; RD 0.06, CI 0.01, 0.12).

Sepsis after study entry among very low birth weight infants treated for more than one week (Outcome 01.17): 
 This outcome was assessed in four randomized trials, involving a total number of 362 treated patients and 364 controls. Summary statistics showed a significant increase in risk of sepsis (typical estimate RR 1.63, CI 1.17, 2.26; RD 0.08, CI 0.03, 0.13).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E and iron intake in the control group. Vitamin E supplementation significantly increased the risk of sepsis in the following subgroups: (1) Birth weight less or equal to 1500 grams (01.17.02) (typical estimate RR 1.79, CI 1.27, 2.53; RD 0.10, CI 0.04, 0.16); (2) Subgroup characterized by parenteral vitamin E with or without enteral administration (01.17.05), by total dose of vitamin E in the treatment group greater than 30 IU/kg/day (01.17.08) and by onset of therapy within 48 hours of life (01.17.11) (typical estimate RR 1.70, CI 1.20, 2.41; RD 0.08, CI 0.03, 0.14); (3) Intravenous vitamin E administration (01.17.06) (typical estimate RR 1.72, CI 1.11, 2.66; RD 0.07, CI 0.02, 0.13); (4) Serum level in the treatment group > 3.5 mg/dl (01.17.10) (typical estimate RR 1.90, CI 1.31, 2.75; RD 0.11, CI 0.05, 0.16); and (5) total dose of vitamin E in the control group <= 10 mg/100 kcal (01.17.13) (typical estimate RR 1.61, CI 1.08, 2.41; RD 0.07, CI 0.02, 0.13). 
 
 Sepsis among surviving very low birth weight infants (Outcome 01.18): 
 This outcome was assessed in two trials, involving 143 treated patients and 141 controls. Summary statistics showed no significant effect of vitamin E on sepsis (typical estimate RR 0.88, CI 0.48, 1.62; RD ‐0.02, CI ‐0.09, +0.06).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Germinal matrix/intraventricular hemorrhage

Germinal matrix/intraventricular hemorrhage (Papile's classification, grade I‐IV) (Outcome 01.19): 
 This outcome was assessed in seven studies involving 864 treated patients and 891 controls. Vitamin E supplementation significantly reduced the risk of germinal matrix‐intraventricular hemorrhage (typical estimate RR 0.85, CI 0.73, 0.99; RD ‐0.04, CI ‐0.08, 0.00). There was significant heterogeneity among studies for RD (chi‐square 17.81, p=0.0067) but not for RR (chi‐square 8.54, p=0.2). 
 
 Subgroup analyses: data available for analysis by birth weight (see section 01.20), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation significantly reduced the risk of hemorrhage in the following subgroups: (1) parenteral vitamin E administration with or without additional enteral vitamin E (01.19.03) (typical estimate RR 0.84, CI 0.72, 0.98; RD ‐0.04, CI ‐0.08, ‐0.01); (2) excluding intravenous vitamin E administration (01.19.06) (typical estimate RR 0.71, CI 0.58, 0.87; RD ‐0.14, CI ‐0.22, ‐0.06); (3) total dose of vitamin E in the treatment group less than or equal to 30 IU/kg/day (01.19.07) (typical estimate RR 0.65, CI 0.50, 0.85; RD ‐0.18, CI ‐0.29, ‐0.08); (4) onset of treatment within 48 hours of life (01.19.11), involving all patients in all studies for this outcome (for typical estimates see summary statistics); and (5) duration of vitamin E treatment less than or equal to one week (01.19.12) (typical estimate RR 0.65, CI 0.50, 0.85; RD ‐0.18, CI ‐0.29, ‐0.08).

Germinal matrix/intraventricular hemorrhage (Papile's classification, grade I‐IV) among very low birth weight infants (Outcome 01.20): 
 This outcome was assessed in three studies involving 385 treated patients and 392 controls. Vitamin E did not significantly affect the risk of hemorrhage (typical estimate RR 0.94, CI 0.75, 1.18; RD ‐0.02, CI ‐0.07, +0.04).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Germinal matrix/intraventricular hemorrhage (Papile's classification, grade I‐IV) among infants with negative initial ultrasonogram (Outcome 01.21): 
 This outcome was assessed in a single study involving 102 treated patients and 108 controls. Vitamin E supplementation significantly reduced the risk for hemorrhage (estimate RR 0.57, CI 0.40, 0.80; RD ‐0.23, CI ‐0.36, ‐0.10).

Germinal matrix/intraventricular hemorrhage (Papile's classification, grade I‐IV) among survivors (Outcome 01.22): 
 This outcome was assessed in three studies involving 166 treated patients and 169 controls. Vitamin E did not significantly affect the risk of hemorrhage (typical estimate RR 1.06, CI 0.70, 1.60; RD 0.01, CI ‐0.07, +0.10).

Subgroup analyses: data available for analysis by birth weight (see section 01.23), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Germinal matrix/intraventricular hemorrhage (Papile's classification, grade I‐IV) among surviving low birth weight infants (Outcome 01.23): 
 This outcome was assessed in two trials, involving 143 treated patients and 141 controls. Summary statistics showed no significant effect of vitamin E on germinal matrix/intraventricular hemorrhage (typical estimate RR 1.05, CI 0.69, 1.60; RD 0.01, CI ‐0.09, +0.11).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Severe intraventricular hemorrhage (grade III or IV) (Outcome 01.24): 
 This outcome was assessed in three studies, involving a total of 315 treated patients and 329 controls. Vitamin E supplementation did not significantly affect the risk of severe intraventricular hemorrhage (typical estimate RR 0.91, CI 0.60, 1.38; RD ‐0.01, CI ‐0.06, +0.04). There was significant heterogeneity among studies both for RR and RD (chi‐square 6.98, p=0.03).

Subgroup analyses: data available for analysis by birth weight (see section 01.25), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Severe intraventricular hemorrhage (grade III or IV) among very low birth weight infants (Outcome 01.25): 
 This outcome was assessed in two studies, involving a total of 213 treated patients and 221 controls. Vitamin E supplementation did not significantly affect the risk of severe intraventricular hemorrhage (typical estimate RR 1.03, CI 0.67, 1.60; RD +0.01, CI ‐0.06, +0.07). There was significant heterogeneity among studies both for RR (chi‐square 4.80, p = 0.029) and RD (chi‐square 4.89, p=0.027).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Severe intraventricular hemorrhage (grade III or IV) among surviving very low birth weight infants (Outcome 01.26): 
 This outcome was analyzed in three studies, involving a total of 158 treated patients and 162 controls. Vitamin E supplementation did not significantly affect the risk of hemorrhage (typical estimate RR 0.76, CI 0.41, 1.39; RD ‐0.03, CI ‐0.10, +0.04 with significant heterogeneity among the studies for RD (chi‐square 7.48, p=0.024).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Subgroup analysis showed that vitamin E supplementation significantly reduced the risk of hemorrhage (1) in the subgroup of patients with birth weight less than 1000 grams (01.26.03) (typical estimate RR 0.31, CI 0.10,0.92; RD ‐0.13, CI ‐0.24,‐0.02), and (2) in that with serum levels greater than 3.5 mg/dl (01.26.11) (estimate RR 0.20, CI 0.05, 0.85; RD 0.20, CI 0.05, 0.85).

Parenchymal brain hemorrhage

Parenchymal brain hemorrhage (grade IV) (Outcome 01.27): 
 This outcome was assessed in two studies, involving 242 treated patients and 255 controls. Meta‐analysis showed no significant overall effect of vitamin E on the risk of hemorrhage (typical estimate RR 1.35, CI 0.69, 2.67; RD 0.02, CI ‐0.02, +0.06) but showed substantial heterogeneity between the two studies both for RR (chi‐square 5.31, p = 0.021) and RD (chi‐square 7.73, p=0.0054). 
 
 Subgroup analysis: data available for analysis by birth weight (see section 01.28), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Subgroup analysis suggests that the significant difference in results between the two studies could be attributed to dose (01.27.06‐07), duration (01.27.10‐11) and type of administration (01.27.03‐05): Phelps 1987a used high‐dose, intravenous and prolonged administration of vitamin E, in contrast with Sinha 1987, which used low‐dose, intramuscular vitamin E for less than one week.

Parenchymal brain hemorrhage (grade IV) among very low birth weight infants (Outcome 01.28): 
 This outcome was assessed in a single study, involving 42 treated patients and 43 controls. Vitamin E supplementation in infants less than 1000 grams significantly increased the risk for hemorrhage (estimate RR 9.21, CI 1.22, 69.58; RD 0.19, CI 0.06, 0.032). This study used high‐dose, intravenous and prolonged administration of vitamin E.

Parenchymal brain hemorrhage (grade IV) among patients with negative initial ultrasonogram (Outcome 01.29): 
 This outcome was assessed in a single study, involving 102 treated patients and 108 controls. Vitamin E supplementation did not significantly affect the risk for hemorrhage (estimate RR 0.30, CI 0.06, 1.42; RD ‐0.05, CI ‐0.10, + 0.01).

Parenchymal brain hemorrhage (grade IV) among surviving very low birth weight infants (Outcome 01.30): 
 This outcome was assessed in two studies, involving 118 treated patients and 118 controls. Vitamin E supplementation did not significantly affect the risk of hemorrhage (typical estimate RR 1.46, CI 0.46, 4.66; RD 0.02, CI ‐0.04, +0.07).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Retinopathy of prematurity or retrolental fibroplasia

Retinopathy of prematurity or retrolental fibroplasia (Outcome 01.31): 
 This outcome was assessed in seven studies, involving 662 treated patients and 680 controls. Vitamin E supplementation did not significantly affect the risk of retinopathy of prematurity or retrolental fibroplasia (typical estimate RR 0.90, CI 0.75, 1.09; RD ‐0.02, CI ‐0.07, +0.02).

Subgroup analyses: data available for analysis by birth weight (see section 01.32), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E and intake of iron, vitamin A and PUFA in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Retinopathy of prematurity or retrolental fibroplasia among very low birth weight infants (Outcome 01.32): 
 This outcome was assessed in five studies, involving 489 treated patients and 486 controls. Vitamin E supplementation did not significantly affect the risk of retinopathy of prematurity or retrolental fibroplasia (typical estimate RR 0.88, CI 0.73, 1.07; RD ‐0.03, CI ‐0.09, +0.02).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron, vitamin A and PUFA. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Retinopathy of prematurity or retrolental fibroplasia among infants examined/survivors (Outcome 01.33): 
 This outcome was assessed in eight studies, involving 821 treated patients and 845 controls. Vitamin E did not significantly affect the risk of retinopathy or prematurity or retrolental fibroplasia (typical estimate RR 0.90, CI 0.78, 1.03; RD ‐0.03, CI ‐0.07, +0.01).

Subgroup analyses: data available for analysis by birth weight (see section 01.34), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron and PUFA. Vitamin E did not significantly affect the risk of retinopathy or prematurity or retrolental fibroplasia in any of the analyses.

Retinopathy of prematurity or retrolental fibroplasia among very low birth weight infants examined/survivors (Outcome 01.34): 
 This outcome was assessed in seven studies, involving 540 treated patients and 550 controls. Vitamin E did not significantly affect the risk of retinopathy or prematurity or retrolental fibroplasia (typical estimate RR 0.94, CI 0.82, 1.07; RD ‐0.03, CI ‐0.08, +0.03).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron and PUFA. Vitamin E did not significantly affect the risk of retinopathy or prematurity or retrolental fibroplasia in any of the analyses.

Severe retinopathy of prematurity or retrolental fibroplasia (grade III or worse) (Outcome 01.35): 
 This outcome was assessed in six studies, involving a total of 767 treated patients and 798 controls. Vitamin E did not significantly affect the risk of severe retinopathy of prematurity or retrolental fibroplasia in the treated group (typical estimate RR 0.72, CI 0.41, 1.25; RD ‐0.01, CI ‐0.03, +0.01).

Subgroup analyses: data available for analysis by birth weight (see section 01.36), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron and PUFA. Vitamin E significantly reduced the risk of retinopathy or prematurity or retrolental fibroplasia in the subgroup of studies with serum tocopherol level vitamin E in the treatment group greater than 3.5 mg/dl (01.35.10): typical estimate RR 0.39, CI 0.16, 1.00; RD ‐0.02, CI ‐0.04, 0.00.

Severe retinopathy of prematurity or retrolental fibroplasia (grade III or worse) among very low birth weight infants (Outcome 01.36): 
 This outcome was assessed in six studies, involving a total of 482 treated patients and 492 controls. Vitamin E did not significantly affect the risk of severe retinopathy of prematurity or retrolental fibroplasia in the treated group (typical estimate RR 0.59, CI 0.32, 1.08; RD ‐0.02, CI ‐0.05, 0.00).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron and PUFA. Vitamin E significantly reduced the risk of retinopathy or prematurity or retrolental fibroplasia (1) in the subgroup with birth weight equal to or less than 1500 grams (01.36.02) typical estimate RR 0.40, CI 0.18, 0.87; RD ‐0.03, CI ‐0.05, ‐0.01 and (2) in the subgroup with serum level in the treatment group greater than 3.5 mg/dl (01.36.12): typical estimate RR 0.34, CI 0.13, 0.87; RD ‐0.04, CI ‐0.07, 0.01.

Severe retinopathy of prematurity or retrolental fibroplasia among infants examined (Outcome 01.37): 
 This outcome was assessed in seven studies, involving a total of 783 treated patients and 804 controls. Vitamin E did not significantly affect the risk of severe retinopathy of prematurity or retrolental fibroplasia, typical estimate RR 0.67, CI 0.40, 1.12; RD ‐0.01, CI ‐0.03, 0.00.

Subgroup analyses: data available for analysis by birth weight (see section 01.38), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron and PUFA. Vitamin E significantly reduced the risk of retinopathy or prematurity or retrolental fibroplasia in the subgroup with serum tocopherol level vitamin greater than 3.5 mg/dl in the treatment group (01.37.11): typical estimate RR 0.39, CI 0.15, 0.98; RD ‐0.02, CI ‐0.04, 0.00.

Severe retinopathy of prematurity or retrolental fibroplasia among very low birth weight infants examined (Outcome 01.38): 
 This outcome was assessed in seven studies, involving a total of 525 treated patients and 537 controls. Vitamin E significantly reduced the risk of severe retinopathy of prematurity or retrolental fibroplasia, typical estimate RR 0.58, CI 0.34, 1.00; RD ‐0.03, CI ‐0.05, 0.00.

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron and PUFA. Vitamin E significantly reduced the risk of retinopathy or prematurity or retrolental fibroplasia in the following subgroups: (1) birth weight <= 1500 grams (01.38.02) (typical estimate RR 0.44, CI 0.22, 0.85; RD ‐0.03, CI ‐0.05, ‐0.01); (2) serum level in the treatment group greater than 3.5 mg/dl (01.38.14) (typical estimate RR 0.34, CI 0.13, 0.88; RD ‐0.04, CI ‐0.07, ‐0.01); and (3) total dose of vitamin E in the control group equal to or less than 10 mg/100 kcal (01.38.18) (typical estimate RR 0.56, CI 0.31, 1.00; RD ‐0.03, CI ‐0.06, 0.00). 
 
 Retinal detachment among surviving infants (Outcome 01.39): 
 This outcome was assessed in three studies, involving a total of 215 treated patients and 217 controls. Vitamin E supplementation did not significantly affect the risk of retinal detachment, typical estimate RR 1.02, CI 0.06, 16.09; RD 0.00, CI ‐0.02, +0.02. However, the incidence of that complication was too low to reach any conclusion: it developed in only one patient in each group in a single study.

Subgroup analyses: data available for analysis by birth weight (see section 01.40), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E did not significantly affect the risk of retinal detachment in any of the analyses.

Retinal detachment among surviving very low birth weight infants (Outcome 01.40): 
 This outcome was assessed in three studies, involving a total of 160 patients in the treatment group and 161 in the control group. Vitamin E did not significantly affect this outcome, typical estimate RR 1.02, CI 0.07, 15.84; RD 0.00, CI ‐0.02, +0.02. However, the incidence of that complication was too low to reach any conclusion: it developed in only one patient in each group in a single study.

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E did not significantly affect the risk of retinal detachment in any of the analyses.

Cicatricial retrolental fibroplasia, any stage, among patients examined after discharge (Outcome 01.41): 
 This outcome was assessed in five studies, involving 530 patients in the treatment group and 522 in the control group. Vitamin E did not significantly affected this outcome (typical estimate RR 0.82, CI 0.52, 1.27; RD ‐0.01, CI ‐0.04, +0.02).

Subgroup analyses: data available for analysis by birth weight (see section 01.42), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Cicatricial retrolental fibroplasia, any stage, among very low birth weight infants examined after discharge (Outcome 01.42): 
 This outcome was assessed in five studies, involving 436 patients in the treatment group and 422 in the control group. Vitamin E did not significantly affected this outcome (typical estimate RR 0.75, CI 0.48, 1.15; RD ‐0.02, CI ‐0.06, +0.01).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Blindness from retrolental fibroplasia among very low birth weight infants examined (Outcome 01.43): 
 This outcome was assessed in four studies, involving 231 treated patients and 236 controls. Vitamin E significantly decreased the risk of blindness, typical estimate, RR 0.29, CI 0.10, 0.88; RD ‐0.04, CI ‐0.08, ‐0.01). 
 
 Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron and PUFA. Vitamin E supplementation significantly reduced the risk of blindness (1) among four subgroups which involved all four studies: birth weight less or equal to 1500 grams (01.43.02), excluding intravenous vitamin E supplementation (01.43.08), onset of therapy within 48 hours of life (01.43.13) and duration of treatment greater than one week (01.43.14) (typical estimate identical to that for the main analysis), (2) in the subgroup with total dose of vitamin E in the control group equal to or less than 10 mg/100 kcal (01.43.15) (typical estimate RR 0.29, CI 0.10, 0.88; RD ‐0.05, CI ‐0.08, ‐0.01).

Necrotizing enterocolitis

Necrotizing enterocolitis (Outcome 01.44): 
 This outcome was assessed in eight studies, involving a total of 711 treated patients and 732 controls. Vitamin E supplementation did not significantly affect this outcome (typical estimate RR 1.37, CI 0.96, 1.95; RD 0.02, CI 0.00, +0.05).

Subgroup analyses: data available for analysis by birth weight (see section 01.45), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron. Vitamin E supplementation significantly increased the risk for necrotizing enterocolitis among patients whose controls received less than or equal to 10 mg vit E/100 cal (01.44.15) (typical estimate RR 1.49, CI 1.03, 2.16; RD 0.03, CI 0.00, 0.06).

Necrotizing enterocolitis among very low birth weight infants (Outcome 01.45): 
 This outcome was assessed in six studies, involving a total of 475 treated patients and 487 controls. Vitamin E supplementation did not significantly affect this outcome (typical estimate RR 1.29, CI 0.90, 1.87; RD 0.03, CI ‐0.01 +0.07).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Necrotizing enterocolitis among very low birth weight infants treated with vitamin E for more than one week (Outcome 01.46): 
 This outcome was assessed in five studies, involving a total of 364 treated patients and 370 controls. Vitamin E supplementation did not significantly affect this outcome (typical estimate RR 1.51, CI 0.99, 2.30; RD 0.04, CI 0.00 +0.09).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron. Vitamin E supplementation significantly increased the risk of necrotizing enterocolitis in the following subgroups: (1) birth weight less than or equal to 1500 grams (01.46.02): typical estimate RR 1.62, CI 1.04, 2.51; RD 0.05, CI 0.01, 0.10; and (2) serum tocopherol level in the treatment group greater than 3.5 mg/dl (01.46.12): typical estimate RR 1.60, CI 1.02, 2.52; RD 0.05, CI 0.00, 0.11.

Necrotizing enterocolitis among survivors (Outcome 01.47): 
 This outcome was assessed in two studies, involving a total of 91 treated patients and 95 controls. Vitamin E supplementation did not significantly affect this outcome (typical estimate RR 1.31, CI 0.31, 5.65, z = 0.37, p = 0.7; RD 0.01, CI ‐0.05, +0.07, z = 0.34, p = 0.7).

Subgroup analyses: data available for analysis by birth weight (see section 01.48), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Necrotizing enterocolitis among surviving very low birth weight infants (Outcome 01.48): 
 This outcome was assessed in a single study, involving a total of 68 treated patients and 67 controls. Vitamin E supplementation did not significantly affect this outcome (estimate RR 1.31, CI 0.31, 5.65; RD 0.01, CI ‐0.06, +0.09).

Serum bilirubin concentration

Serum bilirubin concentration on day 3‐5 (Outcome 01.49): 
 This outcome was assessed in three studies, involving a total of 54 treated patients and 44 controls. Meta‐analysis showed no significant effect of vitamin E on this outcome (weighted mean difference [WMD] ‐0.30, CI ‐1.17, +0.57) but substantial heterogeneity (chi‐square 9.90, p = 0.0071).

Subgroup analyses: data available for analysis by birth weight (see section 01.50), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron. Vitamin E supplementation significantly decreased serum bilirubin concentration in the subgroup with total vitamin E dose in the treatment group greater than 30 IU/kg/day (01.49.07) (mean difference [MD] ‐1.30, CI ‐2.60, 0.00).

Serum bilirubin concentration on day 3‐5 in very low birth weight infants (Outcome 01.50): 
 This outcome was assessed in two studies, involving a total of 27 treated patients and 21 controls. Meta‐analysis showed a significant decrease in serum bilirubin (WMD ‐1.46, CI ‐2.58, ‐0.33).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron. Vitamin E supplementation significantly decreased serum bilirubin concentration in the following subgroups: (1) Enteral (01.50.02) and total dose of vitamin E in the treatment group greater than 30 IU/kg/day (01.50.06) (MD ‐1.30, CI ‐2.60, 0.00); (2) Excluding intravenous (01.50.04), serum level in the treatment group equal to or less than 3.5 mg/dl (01.50.07), onset of treatment within 48 hours of life (01.49.08), duration of treatment less than or equal to one week (01.50.09) and total dose of vitamin E in the control group equal to or less than 10 mg/100 kcal (01.49.10), which included all patients from all studies (WMD ‐1.46, CI ‐2.58, ‐0.33).

Serum bilirubin concentration on day 3‐5 in a specific group (no hemolysis, no polycythemia, no prior transfusion) (Outcome 01.51): 
 This outcome was assessed in a single study involving 31 treated patients and 31 controls. Vitamin E significantly increased serum bilirubin (MD =1.30, CI 0.20, 2.40).

Hemoglobin concentration

Hemoglobin concentration (Outcome 01.52): 
 Hemoglobin concentration was assessed in eight studies, involving 211 patients in the treatment group and 205 controls. Vitamin E supplementation significantly increased hemoglobin concentration (WMD 0.46 g/dl, CI 0.24, 0.69).

Subgroup analyses: data available for analysis by birth weight (see section 01.53), by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron and PUFA. Vitamin E supplementation significantly increased hemoglobin concentration in the following subgroups: (1) infants with birth weight > 1000 grams (01.52.02) (WMD 0.49, CI 0.20, 0.77), (2) enteral vitamin E supplementation (01.52.03) (WMD 0.48, CI 0.25, 0.71), (3) excluding intravenous vitamin E supplementation (01.52.05) (involving all available studies; typical estimate identical to the summary statistics for this outcome), (4) vitamin E dose less than or equal to 30 IU/kg/day (01.52.06) (WMD 0.52, CI 0.29, 0.76), (5) serum tocopherol level in the treatment group less than or equal to 3.5 mg/dl (01.52.08) (WMD 0.47, CI 0.23, 0.71), (6) onset of treatment within 48 hours of life (01.52.09) (WMD 0.48, CI 0.21, 0.75), (7) onset of treatment after 48 hours of life (01.52.10) (WMD 0.44, CI 0.02, 0.85), (8) duration of treatment greater than one week (01.52.12) (WMD 0.54, CI 0.30, 0.77), (9) total dose of vitamin E in the control group equal to or less than 10 mg/100 kcal (01.52.13) (WMD 0.46, CI 0.24, 0.69), (10) iron supplementation in both groups (01.52.14) (WMD 0.57, CI 0.24, 0.91), and (11) iron supplementation in neither group (01.52.15) (WMD 0.47, CI 0.16, 0.78).

Hemoglobin concentration in very low birth weight infants (Outcome 01.53): 
 This outcome was analyzed in four studies, involving 102 treated infants and 94 controls. Meta‐analysis showed a significant increase in hemoglobin, WMD 0.43 g/dl, CI 0.09, 0.77, with significant heterogeneity among studies (chi‐square 10.38, p =0.02).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron and PUFA. Vitamin E supplementation significantly increased hemoglobin concentration in the following subgroups: (1) Enteral vitamin E supplementation (01.53.02), excluding intravenous supplementation (01.53.03), serum tocopherol level vitamin E in the treatment group lower than or equal to 3.5 mg/dl (01.53.06), total dose of vitamin E in the control group equal to or less than 10 mg/100 kcal (01.53.11), all involving all subjects from all studies; (2) total dose of vitamin E in the treatment group less than or equal to 30 IU/kg/day (01.53.04) (WMD 0.53, CI 0.16, 0.90); (3) Onset of treatment in the treatment group within 48 hours of life (01.53.07) (WMD 0.42, CI 0.07, 0.78); (4) Duration of treatment greater than one week (01.53.10) (WMD 0.53, CI 0.17, 0.89); (5) Iron supplementation in both groups (01.53.12) (WMD 0.67, CI 0.23, 1.10); (6) Iron supplementation in neither group (01.53.13) (WMD 0.51, CI 0.17, 0.86).

Reticulocyte count

Reticulocyte count (expressed in percent of total red blood cell count) (Outcome 01.54): 
 This outcome was assessed in two studies, involving 102 treated patients and 99 controls. Meta‐analysis showed that vitamin E supplementation significantly reduced reticulocyte count (WMD ‐1.66, CI ‐2.24, ‐1.07) and substantial heterogeneity among studies (chi‐square 6.44, p = 0.011).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron and PUFA. Vitamin E supplementation significantly reduced reticulocyte count in the following subgroups: (1) subgroups involving all patients from both studies: enteral (01.54.02), excluding intravenous vitamin E supplementation (01.54.03), total dose of vitamin E in the treatment group equal to or less than 30 IU/kg/day (01.54.04), serum tocopherol level in the treatment group equal to or less than 3.5 mg/dl (01.54.05), onset of treatment within 48 hours of life (01.54.06), duration of treatment greater than one week (01.54.07), total dose of vitamin E in the control group less than or equal to 10 mg/100 kcal (01.54.08) (typical estimates identical to the summary statistics for this outcome); (2) iron supplementation in both groups (01.54.09) (MD ‐1.96, CI ‐2.89, ‐1.03); and (3) iron supplementation in neither group (01.54.10) (MD ‐1.47, CI ‐2.22, ‐0.72). 
 
 Reticulocyte count (expressed in percent of total red blood cell count) in very low birth weight infants (Outcome 01.55): 
 This outcome was assessed in a single study, involving 31 treated patients and 33 controls. Vitamin E supplementation did not significantly affect reticulocyte counts (MD ‐0.30, CI ‐1.50, +0.90).

Absolute reticulocyte counts (expressed in million per liter) (Outcome 01.56): 
 This outcome was assessed in two studies, involving 32 treated patients and 39 controls. Meta‐analysis showed significant heterogeneity between the two studies (chi‐square analysis 13.03, p = 0.0003) and a significant reduction in reticulocyte count in the treated group (WMD ‐34.0, CI ‐46.2, ‐21.8, z = 5.48, p < 0.00001).

Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron. Vitamin E supplementation significantly reduced reticulocyte count in the following subgroups, which involved all patients in both studies: enteral (01.56.02), excluding intravenous vitamin E supplementation (01.56.03), serum tocopherol level less than or equal to 3.5 mg/dl (01.56.05), onset of treatment after 48 hours of life (01.56.06), duration of treatment greater than one week (01.56.07), total dose of vitamin E intake in the control group equal to or less than 10 mg/100 kcal (01.56.08), and iron supplementation in both groups (01.56.09) (typical estimates identical to summary statistics for this outcome).

Absolute reticulocyte counts (expressed in million per liter) in very low birth weight infants (Outcome 01.57): 
 This outcome was assessed in a single study, involving 31 treated patients and 33 controls. Vitamin E supplementation significantly reduced absolute reticulocyte counts (MD ‐54.3, CI ‐70.7, ‐37.9).

Blood transfusion

Number of transfusions (Outcome 01.58): 
 This outcome was assessed in a single study involving 15 treated patients and 15 controls. Vitamin E supplementation did not significantly affect the number of transfusions (MD ‐0.20, CI ‐1.17, +0.77).

Number of transfusions among survivors (Outcome 01.59): 
 This outcome was assessed in a single study involving 68 treated patients and 67 controls. Vitamin E supplementation did not significantly affect the number of transfusions (MD +1.20, CI ‐1.42, +3.82).

Amount of blood transfused (ml per kg of body weight) among very low birth weight infants (Outcome 01.60): 
 This outcome was assessed in two studies involving 223 treated patients and 231 controls. Vitamin E supplementation did not significantly affect the number of transfusions (WMD ‐0.5, CI ‐16.9, +15.9). 
 Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Amount of blood transfused among survivors (Outcome 01.61): 
 This outcome was assessed in a single study involving 68 treated patients and 67 controls. Vitamin E supplementation did not significantly affect the number of transfusions (MD +18.1, CI ‐17.1, +53.3).

Other outcomes

Platelet count (Outcome 01.62): 
 Platelet counts were analyzed in two studies involving a total of 49 treated patients and 60 controls. Vitamin E supplementation did not significantly affect the platelet count (WMD ‐29.52, CI ‐88.81, +29.76). 
 Subgroup analyses: data available for analysis by birth weight, by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group, and by dose of vitamin E in the control group and by intake of iron. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Coagulation tests (Outcomes 01.63 ‐ 01.65): 
 The effects of vitamin E on coagulation tests were analyzed in one study involving 23 treated patients and 24 controls (Zipursky 1987), which used low dose enteral administration Vitamin E administration did not significantly affect any of the coagulation tests, including prothrombin time, partial thromboplastin time, fibrinogen concentration and other factors not listed in this review.

Retinal hemorrhage among very low birth weight infants examined/survivors (Outcome 01.66): 
 This outcome was assessed in a single study, involving 111 treated patients and 121 controls, which used high‐dose, intravenous administration of vitamin E. Vitamin E did not significantly affect the risk of retinal hemorrhage in the whole study population (estimate RR 2.18, CI 0.97,4.89; RD 0.08, 0.16). Subgroup analysis among extremely low birth weight infants (n=85) showed that vitamin E significantly increased the risk of retinal hemorrhage (estimate RR 3.58, CI 1.28,10.00; RD 0.24, CI 0.07, 0.41).

Reaction at the site of injection (Outcome 01.67): 
 This outcome was reported in two studies, involving 175 treated patients and 182 controls. Meta‐analysis showed no significant effect of vitamin E on this outcome (typical estimate RR 4.13, CI 0.47, 36.69; RD 0.02, CI ‐0.01, +0.04). 
 Subgroup analyses: data available for analysis by birth weight (see section 01.67) and by route of administration, by total dose, serum level, onset and duration of therapy in the treatment group. Vitamin E supplementation did not significantly affect this outcome in any of the analyses.

Reaction at the site of injection in very low birth weight infants (Outcome 01.68): 
 This outcome was assessed in a single study involving 73 treated patients and 74 controls. Vitamin E supplementation did not significantly affect the risk of reaction, estimate RR 5.07, CI 0.25, 103.78; RD 0.03, CI ‐0.02, +0.074.

VITAMIN E VERSUS ANOTHER TYPE OR ROUTE OF VITAMIN E FORMULATION (COMPARISON 02):

Only one study was found in this category. Bonati 1991 did not show any significant difference in mortality (Outcome 02.01) and did not assess the other planned primary outcomes (long‐term combined outcomes). This study did not show any significant difference in any of the analyzed secondary outcomes, including bronchopulmonary dysplasia (Outcome 02.02), patent ductus arteriosus (Outcome 02.03), sepsis (Outcome 02.04), intracranial hemorrhage (Outcome 02.05), parenchymal hemorrhage (Outcome 02.06), necrotizing enterocolitis (Outcome 02.07), or exchange transfusion (Outcome 02.08). Plasma levels of free tocopherol peaked on day three, after three injections of vitamin E. In the treatment group, plasma levels of tocopheryl acetate were one third of levels of free tocopherol. In the control group receiving tocopherol in olive oil, plasma levels of free tocopherol were within deficient range and significantly lower than those receiving the colloidal solution of tocopheryl acetate.

Characteristics of included studies [ordered by study ID]

Characteristics of excluded studies [ordered by study ID]