Home > Health & Research > Health Education Campaigns & Programs > Cochrane Neonatal Review > Low versus high haemoglobin concentration threshold for blood transfusion for preventing morbidity and mortality in very low birth weight infants

Low versus high haemoglobin concentration threshold for blood transfusion for preventing morbidity and mortality in very low birth weight infants

Skip sharing on social media links
Share this:

Authors

Robin Whyte1, Haresh Kirpalani2

Background - Methods - Results - Characteristics of Included Studies - References - Data Tables & Graphs


1Department of Neonatal Pediatrics, IWK Health Centre - G2216, Halifax, Canada [top]
2Department of Pediatrics, University of Pennsylvania School of Medicine and Dept of Clinical Epidemiology and Biostatistics, McMaster University, Philadelphia, Pennsylvania, USA [top]

Citation example: Whyte R, Kirpalani H. Low versus high haemoglobin concentration threshold for blood transfusion for preventing morbidity and mortality in very low birth weight infants. Cochrane Database of Systematic Reviews 2011, Issue 11. Art. No.: CD000512. DOI: 10.1002/14651858.CD000512.pub2.

Contact person

Robin Whyte

Department of Neonatal Pediatrics
IWK Health Centre - G2216
5850/5980 University Avenue
Halifax Nova Scotia B3K 6R8
Canada

E-mail: Robin.Whyte@dal.ca

Dates

Assessed as Up-to-date: 02 September 2011
Date of Search: 31 August 2011
Next Stage Expected: 02 September 2013
Protocol First Published: Issue 4, 1997
Review First Published: Issue 11, 2011
Last Citation Issue: Issue 11, 2011

History

Date / Event Description
30 September 2010
New citation: conclusions not changed

Substantive amendment

20 August 2009
Amended

Converted to new review format.

Abstract

Background

Infants of very low birth weight often receive multiple transfusions of red blood cells, usually in response to predetermined haemoglobin or haematocrit thresholds. In the absence of better indices, haemoglobin levels are imperfect but necessary guides to the need for transfusion. Chronic anaemia in premature infants may, if severe, cause apnoea, poor neurodevelopmental outcomes or poor weight gain. On the other hand, red blood cell transfusion may result in transmission of infections, circulatory or iron overload, or dysfunctional oxygen carriage and delivery.

Objectives

To determine if erythrocyte transfusion administered to maintain low as compared to high haemoglobin thresholds reduces mortality or morbidity in very low birth weight infants enrolled within three days of birth.

Search methods

Two review authors independently searched the Cochrane Central Register of Controlled Trials (The Cochrane Library), MEDLINE, EMBASE, and conference proceedings through June 2010.

Selection criteria

We selected randomised controlled trials (RCTs) comparing the effects of early versus late, or restrictive versus liberal erythrocyte transfusion regimes in low birth weight infants applied within three days of birth, with mortality or major morbidity as outcomes.

Data collection and analysis

Two review authors independently selected the trials.

Results

Four trials, enrolling a total of 614 infants, compared low (restrictive) to high (liberal) haemoglobin thresholds. Restrictive thresholds tended to be similar, but one trial used liberal thresholds much higher than the other three. There were no statistically significant differences in the combined outcomes of death or serious morbidity at first hospital discharge (typical risk ratio (RR) 1.19; 95% confidence interval (CI) 0.95 to 1.49) or in component outcomes. Only the largest trial reported follow-up at 18 to 21 months corrected gestational age; in this study there was no statistically significant difference in a composite of death or adverse neurodevelopmental outcome (RR 1.06; 95% CI 0.95 to 1.19). One additional trial comparing transfusion for clinical signs of anaemia versus transfusion at a set level of haemoglobin or haematocrit, reported no deaths and did not address disability.

Authors' conclusions

The use of restrictive as compared to liberal haemoglobin thresholds in infants of very low birth weight results in modest reductions in exposure to transfusion and in haemoglobin levels. Restrictive practice does not appear to have a significant impact on death or major morbidities at first hospital discharge or at follow-up. However, given the uncertainties of these conclusions, it would be prudent to avoid haemoglobin levels below the lower limits tested here. Further trials are required to clarify the impact of transfusion practice on long term outcome.

Plain language summary

Low versus high haemoglobin concentration threshold for blood transfusion for preventing morbidity and mortality in very low birth weight infants

Very premature infants are extremely vulnerable and often require intensive care to survive. Anaemia is a condition in which the blood does not contain enough haemoglobin, the component of red blood cells which carries oxygen around the body. These babies become anaemic very quickly due to blood sampling and because they are unable to make blood cells quickly the haemoglobin level in the blood falls rapidly in the weeks after birth. Generally, the treatment for anaemia is blood transfusion, and many of these babies receive multiple transfusions of blood. The decision to give a transfusion usually depends on the measured amount of haemoglobin in the blood.

Physicians looking after very premature infants are unsure as to the level of haemoglobin at which they should give a transfusion. As transfusion is the introduction of another person's blood cells into the blood stream, there is a risk of infection and a risk of reaction to foreign blood components; the process requires careful monitoring and supervision to ensure safety. Some people find blood transfusion offensive or contrary to their religious values. Giving few or no transfusions reduces the risks of transfusion, but may result in low levels of haemoglobin and consequently a reduced supply of oxygen to the body which could have effects on survival, growth or development.

This review of five studies compares the effects of blood transfusion at low levels of haemoglobin to transfusion at high levels. Within the levels tested, there were no differences seen in survival, in the serious complications of prematurity, or in longer term development as measured at 18 to 21 months past the baby's due date. Allowing the baby to become a little more anaemic did not affect the baby's weight gain or breathing patterns. These conclusions are not firm, because too few babies have been studied. Our overall recommendation is not to exceed the higher levels of haemoglobin used in these trials, and thus diminish the risks of over-transfusion, but not to allow the level of haemoglobin to fall below the lower limits tested in these studies until further studies are completed.

[top]

Background

Description of the condition

The normal postnatal fall in haemoglobin concentration is greatly accelerated in infants of very low birth weight (Wardrop 1978; Stockman 1984). Anaemia of prematurity is caused by a combination of a suppressed postnatal response to erythropoietin and by rapid growth (Stockman 1984), and is greatly exacerbated by clinically prescribed blood sampling (Blanchette 1984). It is difficult to identify a single haemoglobin concentration level below which oxygen delivery may be inadequate. Oxygen consumption may be higher in the critically ill and when poor cardiorespiratory function reduces oxygen delivery, there may be a need for higher haemoglobin levels. Oxygen unloading to the tissues is greatly affected by the separate effects of acid-base balance, carbon dioxide tension and erythrocyte levels of 2, 3 diphosphoglycerate on haemoglobin (Sacks 1984). These effects are greatly amplified in adult rather than foetal haemoglobin, which is increased with transfusion as well as with postnatal age (Blank 1984; Barkemeyer 2000).

Failure of an adequate oxygen delivery to the tissues due to a low haemoglobin concentration may be compensated for by adjustments in cardiac output (limited in the newborn) (Lachance 1994; Alkalay 2003; Cambonie 2007) and in haemoglobin function (Sacks 1984; Meldon 1985). However beyond a certain undefined point, it is likely that growth and development will be impaired. Long term effects of anaemia of the low birth weight infant have not been documented, and indications for treatment for signs of anaemia (tachycardia) (Stockman 1986), poor weight gain (Stockman 1984a) and apnoea (Ross 1989) have not been defined.

Description of the intervention

Approaches to treating or preventing anaemia in very low birth weight infants include maximising placental transfusion at birth (Rabe 2004; Hosono 2008), strict limitation of blood sampling (Madan 2005; Venancio 2007; Mimica 2008) and red blood cell transfusion (Miyashiro 2005; Bell 2008; Crowley 2010). An alternative or additional strategy is to use erythropoietin, as recombinant human erythropoietin (rHuEPO); this is the subject of other Cochrane reviews (Aher 2006; Aher 2006a; Ohlsson 2006). Rates of blood transfusion for infants of very low birth weight are highest in the smallest infants and more than half of the transfusions are given in the first two weeks of life (Widness 1996; Beeram 2001). In the 1990s, about 40% of infants of < 1500 g and 95% of those < 1000 g birth weight received a blood transfusion. On average very low birth weight infants received between two and three transfusions each, and 30% of infants < 1000 g birth weight received in excess of 10 transfusions each (Widness 1996; Beeram 2001).

Transfusions may be given in emergencies for shock, or prior to surgery, but in critical care most transfusions are given to raise blood haemoglobin concentrations above ill-defined critical minima (Fetus and Newborn 1992; Shannon 1995; Fetus and Newborn 2002) or ‘thresholds’; the latter term is used in this review. Measurements other than haemoglobin level have been advocated, such as oxygen availability (Stockman 1986; Siggaard-Andersen 1990), fractional oxygen extraction (Wardle 2001) or the presence of clinical signs of anaemia (Blank 1984). Apart from the putative clinical signs, these measures are not easy to collect and have not improved prediction of outcome. Haemoglobin levels, on the other hand, are universally and quickly available. Measurements of haematocrit are generally interchangeable with haemoglobin levels (Christensen 2008).

Current recommendations advocate maintaining higher levels of haemoglobin in the early critical weeks. Clinical trials reflect this practice by comparing haemoglobin levels held apart by fixed amounts, but falling with postnatal age. While an essential outcome of trials of neonatal critical care must be survival, long term anaemia has the potential to affect both brain growth and other components of chronic disease of the premature infant (Andersen 2006). Therefore long term developmental, growth and health outcomes must be included and combined in both short term and long term follow-up studies.

Transfusion borne infections are rare but continue to emerge (Blajchman 2006; Stramer 2009). Potential and unproven adverse effects may be subtle and cumulative (Holman 1995), such as retinopathy of prematurity (Dani 2004), bronchopulmonary dysplasia (Silvers 1998), necrotising enterocolitis (Mally 2006), suppression of host resistance (Vamvakas 2007) or iron overload (Hirano 2001). Blood which has been stored, particularly if not leukoreduced, contains pro-inflammatory microparticles (Simak 2006), which may contribute to transfusion-related acute lung injury, multiple organ failure or infection (McFaul 2009). Transfusion borne infections are too rare to evaluate as outcomes in a clinical trial, though the number of donor exposures might be used as a proxy outcome.

How the intervention might work

Transfusion of packed red cells raises red cell volume and the haemoglobin level of the blood. The presence of additional adult haemoglobin may double the effect of haemoglobin on oxygen delivery to the tissues (Sacks 1984). However, it is not clear how well or when the transfused red cells function. Stored donor red cells are deficient in 2, 3 diphosphoglycerate (Delivoria-Papadopoulos 1971) and are less able to generate S-nitrosohaemoglobin (Bennett-Guerrero 2007) which assists in targeting red cell oxygen delivery to hypoxic tissues (Allen 2006). Some of these functions are known to recover in the recipient circulation (Zimrin 2009).Transfused allogeneic red cells have shorter life spans than the recipient cells (Bard 1997; Strauss 2004), and if the rise in haemoglobin from a donor source suppresses erythropoietin in the recipient (Keyes 1989; Frey 2001) the haematological benefits of transfusion could be short-lived or reversed. In the face of a falling haemoglobin level, use of a low rather than a high haemoglobin threshold for transfusion may result in later rather than fewer transfusions; therefore the effects of the different strategies may reflect transfusion practices as well as haemoglobin levels.

Why it is important to do this review

Both anaemia of prematurity and blood transfusion have their own morbidities. Clinical uncertainty about indications for transfusion is reflected in high variability in transfusion rates between treatment centres (Bednarek 1998; Ringer 1998). Restrictive transfusion (at low haemoglobin thresholds) led to equivalent outcomes when compared with conventional liberal transfusion (at high thresholds) in critically ill children (Lacroix 2007) and led to improved survival in subgroups of adults (Hébert 1999). A systematic review antedated the current generation of randomised evidence (Hume 1997). There have been no Cochrane reviews of blood transfusion of the newborn.

Objectives

  1. To determine if the use of lower versus higher haemoglobin concentration thresholds for blood transfusion affects mortality or adverse neurodevelopmental outcome of very low birth weight infants.
  2. To determine if blood transfusion prompted by clinical signs or other measures of anaemia versus transfusion at a fixed value or programme of haemoglobin thresholds affects the risk of mortality or adverse neurodevelopmental outcome of very low birth weight infants.

[top]

Methods

Criteria for considering studies for this review

Types of studies

We included randomised and quasi-randomised clinical trials.

Types of participants

We included very low birth weight infants, i.e. of birth weight less than or equal to 1500 g, or infants less than 32 weeks gestational age admitted to neonatal intensive care, at less than one week of age. We aimed specifically to include studies of infants receiving all levels of intensive care.

Types of interventions

Transfusions of packed red cells or whole blood, not including exchange transfusion or placental-infant transfusion.

We searched for trials in which the following comparisons were made.

  1. Transfusion at a low haemoglobin or haematocrit level (restrictive) versus transfusion at a high haemoglobin or haematocrit level (liberal), either at (a) fixed levels of haemoglobin or haematocrit or (b) at levels of haemoglobin or haematocrit varying with postnatal age or other conditions (e.g. severity of illness) affecting decisions to transfuse.
  2. No transfusion until clinical signs of anaemia (restrictive) versus transfusion at a set level of haemoglobin or haematocrit (liberal).

Types of outcome measures

Primary outcomes
  1. Death: before discharge from initial hospitalisation; or before a defined period of follow-up.
  2. A composite outcome of death or severe adverse outcomes; these competing outcomes are mutually exclusive. The event is present if either death or any severe adverse outcome occurs, or absent if none of the following occurs.
  • Death or severe morbidity (or its complement, survival without severe morbidity) at initial hospital discharge, where severe morbidity is defined as:
    • retinopathy of prematurity, grade 3 (ICCRP 2005) or more; or
    • severe adverse findings at ultrasound (grades 3-4 intraventricular haemorrhage (Papile 1983), hydrocephalus, cortical atrophy or periventricular leukomalacia) during first hospitalisation (Pinto-Martin 1995); or
    • bronchopulmonary dysplasia (chronic lung disease requiring additional oxygen at 36 weeks gestation) (Shennan 1988)
  • Death or severe adverse neurosensory outcome (or its complement, survival without serious adverse neurosensory outcome) at a defined period of follow-up at age 18 months adjusted gestational age or older, where severe adverse neurosensory outcome is defined as:
    • cerebral palsy by physician assessment; or
    • developmental delay (IQ > 2 standard deviations below the mean on validated assessment tool of cognitive function, e.g. Bayley Score of Infant Development); or
    • blindness (visual acuity < 20/200 in best eye); or
    • deafness (hearing loss requiring amplification of cochlear implantation).
Secondary outcomes

At initial hospital discharge:

  • severe morbidity, as defined components of primary outcome (2) above, each recorded as separate binary outcomes (present or not present);
  • moderate morbidity, as less severe components of primary outcome (2) above:
    • retinopathy of prematurity, grades 1-2 (ICCRP 2005);
    • intraventricular haemorrhage, grades 1-2 (Papile 1983);
    • bronchopulmonary dysplasia (chronic lung disease requiring additional oxygen at 28 days gestation);
  • haemoglobin or haematocrit level at discharge;
  • number of transfusions per infant;
  • number of donor exposures per infant, either from all blood products or from red blood cell transfusions alone;
  • measures of cost-effectiveness of blood transfusion;
  • postnatal acquisition of viral infection (cytomegalovirus, human immunodeficiency virus or hepatitis C);
  • weight gain as rate, or as weight at discharge; and
  • incidence of apnoea of prematurity.

Outcomes added after this review was conducted were:

  • the presence of persistent patency of the ductus arteriosus; and
  • the presence of necrotising enterocolitis.

At a defined period of follow-up at age 18 months adjusted gestational age or older:

  • severe adverse neurosensory outcome as defined components of the primary composite outcome (2) above, each recorded as separate binary outcomes (present or not present) in survivors;
  • moderate adverse neurosensory outcome defined as:
    • any neurologic disability; or
    • developmental delay (IQ > 1 standard deviation below the mean on validated assessment tool of cognitive function, e.g. Bayley Score of Infant Development).

Search methods for identification of studies

Electronic searches

We searched the Cochrane Central Register of Controlled Trials (The Cochrane Library 2010, Issue 5) using the terms BLOOD TRANSFUSION and INFANT. We searched MEDLINE (January 1966 to August 2011) using exploded MeSH terms BLOOD TRANSFUSION or the free text terms BLOOD AND TRANSFUS*, RED AND CELL AND TRANSFUS* or text terms ERYTHROCYTE AND TRANSFUSION in all languages. We combined the retrieved titles (AND) with the exploded MeSH term INFANT, NEWBORN. We searched EMBASE (January 1974 to August 2011) using the terms INFANT (with either "low birth weight" or "premature") with BLOOD TRANSFUSION. We searched the Science Citation Index (September 1984 to June 2011) for trials which cited the trials identified by our searches as MEDLINE and Embase. We sought registered details of selected trials in the U.S. National Institutes of Health resource ClinicalTrials.gov.

Searching other resources

We obtained information by personal communication, reviewing the reference lists of relevant articles, abstracts and conference proceedings (Society for Pediatric Research, European Society for Paediatric Research 1990 to 2010) and seeking results of unpublished trials.

Data collection and analysis

Selection of studies

Both review authors reviewed the search results independently; there were no disagreements in selection.

Data extraction and management

We used a data extraction form to extract the data together. We used standard methods of the Cochrane Collaboration (Deeks 2011) and the Cochrane Neonatal Review Group to assess trial quality, data extraction and synthesis of data, using risk ratio (RR), risk difference (RD) and weighted mean difference (WMD) where appropriate using Review Manager 5 software (RevMan 2011). We assessed each identified trial for methodological quality with respect to (a) inclusiveness of the population, (b) masking of allocation, (c) masking of intervention, (d) completeness of follow-up and (e) masking of outcome assessment. We sought clarification from at least one author of each trial considered for selection. Some unpublished data were available from the records of trials; we had access to data from the PINT 2006 study and communicated with the principal author of Bell 2005 et al for some details of outcomes. We also had access to the protocol, records and the results for the unpublished trial of Connelly 1999 et al.

Assessment of risk of bias in included studies

We assessed risk of bias in selection, performance, outcome and attrition. Our intention was to employ a fixed-effect model for data analysis unless there was clinical diversity in the question or statistical heterogeneity in the outcome, in which case we would use a random-effects model. We expressed all results with 95% confidence intervals (CIs) unless otherwise stated. All outcomes analysed as categories were binary (event or no event) and we recorded them once per participant. We recorded all primary outcomes as RRs with all randomised infants as denominator. For secondary analyses, we expressed outcomes as 'in survivors' where appropriate.

Dealing with missing data

We replaced missing data where the outcome could be determined (for example where infants who had died between randomisation and reaching the trial intervention had been excluded from analysis). Otherwise, where data were judged missing at random (for example where a local institutional review board refused permission to collect outcome data) we used only available data. Where it was likely that data were not missing at random, we did not include the outcome in the analysis. We judged clinical and methodological diversity between studies by comparing the subjects enrolled and the study interventions conducted (Deeks 2011).

Assessment of heterogeneity

We assessed statistical heterogeneity using the I2 statistic. We did not perform any subgroup analyses.There were insufficient studies available for evaluation of reporting bias.

[top]

Results

Description of studies

See Characteristics of Included Studies and Characteristics of excluded studies tables. From our searches of the literature we identified 10 randomised trials that evaluated the effects of transfusion at a given haemoglobin or haematocrit trigger in very low birth weight infants compared to either a different haemoglobin threshold or the occurrence of a clinical event. We have organised all studies in this review so that restrictive interventions are treated as interventions and liberal interventions as control (reflecting conventions at the time these studies were started). We expressed RRs as restrictive/liberal risks.

Included studies

Five studies qualified for inclusion. Four addressed comparison 1 (transfusion at a low haemoglobin or haematocrit level (restrictive) versus transfusion at a high haemoglobin or haematocrit level (liberal)) and one addressed comparison 2 (transfusion for clinical signs of anaemia (restrictive) versus transfusion at a set level of haemoglobin or haematocrit (liberal)).

1. Transfusion at a low haemoglobin or haematocrit level (restrictive) versus transfusion at a high haemoglobin or haematocrit level (liberal)

There were four included studies addressing this comparison. For ease of comparison we have used a mean corpuscular haemoglobin concentration of 340 g/l (Christensen 2008) to convert haematocrit to haemoglobin (haematocrit % x 34 = haemoglobin g/l) and have expressed transfusion thresholds in Table 1.

The (Premature Infants in Need of Transfusion (PINT)) study was published in two reports PINT 2006 , addressing outcomes by first hospital discharge and follow-up at 18 to 21 months corrected gestational age (Premature Infants in Need of Transfusion Outcome Study (PINTOS)). 451 newborn infants were enrolled from 10 neonatal intensive care units (ICUs) in Canada, the United States and Australia. Enrolled infants were less than 1000 g birth weight and 31 weeks gestational age, and less than 48 hours old at the time of enrolment. Infants could be enrolled if not transfused after six hours of age. Infants were randomly assigned to either the restrictive (low haemoglobin) threshold or the liberal (high haemoglobin) threshold for transfusion (15 ml/kg of packed red blood cells (PRBCs)). Transfusion thresholds fell with postnatal age over the first two weeks of age and were adjusted for the need for respiratory support (any positive pressure airway support or oxygen supplementation); see Characteristics of Included Studies. The capillary haemoglobin thresholds while receiving respiratory support in the first week were 115 g/l versus 135 g/l for restrictive and liberal groups, falling to 75 g/l versus 85 g/l at 15 days and later. If the infants no longer required respiratory support, these thresholds were 100 g/l versus 120 g/l respectively, falling to 75 g/l versus 85 g/l at or after 15 days. Transfusions outside the threshold algorithm were allowed for predefined emergencies or for surgery. Detailed outcomes describing the effect of the intervention on numbers of infants transfused, numbers of transfusions given per infant, haemoglobin levels and donor exposure were reported. The primary outcome of the primary report (PINT) was a composite outcome at first hospital discharge of death or survival with bronchopulmonary dysplasia, severe retinopathy of prematurity or brain injury on ultrasound (cystic periventricular leukomalacia, intraparenchymal echodensity, porencephalic cyst or ventriculomegaly). The secondary report (PINTOS) addressed a primary composite outcome at follow-up of mortality or severe adverse neurodevelopmental outcome (cerebral palsy, cognitive delay, severe visual or hearing impairment). Other clinical outcomes are described in Characteristics of Included Studies.

Bell 2005 et al conducted a single centre trial in the USA enrolling 103 infants of birth weight 500 to 1300 g. Infants were enrolled as soon after birth as possible but age at enrolment varied, being delayed in infants when there was no immediate indication to transfuse by high threshold criteria. Infants were permitted to be enrolled if they had received up to two perinatal transfusions prior to enrolment. Infants in the restrictive group were transfused at haematocrits of 34%, 28% or 22% respectively if they were, intubated, receiving continuous distending pressure or additional oxygen, or breathing room air. Infants in the liberal group were transfused at haematocrits of 46%, 38%, or 30% respectively. These values are equivalent to haemoglobin levels of 116 g/l, 95 g/l and 75 g/l for the restrictive group and 156 g/l, 129 g/l and 102 g/l for the liberal group. Infants were transfused with 15 ml/kg of PRBCs. The trial ended at discharge. The primary outcome was numbers of transfusions given per infant. Other measures of the intervention were numbers of infants transfused and rates of donor exposure. Clinical outcomes included survival, adverse ultrasound findings (periventricular leukomalacia and intraventricular haemorrhage), severe retinopathy of prematurity and bronchopulmonary dysplasia. Other outcomes are described in Characteristics of Included Studies. Partial follow-up of these infants has been recently published (McCoy 2011, Nopoulos 2011). These data are not included in this review and may be considered in future updates.

Chen 2009 et al conducted a single centre trial in Taiwan of 36 infants of less than 1500 g birth weight aged up to three days of age. Infants were randomised by computer allocation. Infants in the restrictive group were transfused at haematocrits of 35%, 30% or 22% respectively if they were intubated, receiving continuous distending pressure or in room air. Infants in the liberal group were transfused at haematocrits of 45%, 40%, or 30%. These values are equivalent to haemoglobin levels of 119 g/l, 102 g/l and 75 g/l for the restrictive group and 153 g/l, 136 g/l and 102 g/l for the liberal group. Infants were transfused with 15ml/kg of PRBCs. Haematologic outcomes were measured at 30 days of age and clinical outcomes were measured at 28 days of age and at 36 weeks postconceptional age. The primary outcome was number of transfusions given per infant. Clinical outcomes included survival, intraventricular haemorrhage, severe retinopathy of prematurity and bronchopulmonary dysplasia. Other outcomes are described in Characteristics of Included Studies. We obtained further outcomes and clarification from the author, including a composite outcome of death or survival with bronchopulmonary dysplasia, severe retinopathy of prematurity, or brain injury on ultrasound. Communication with the author of Chen 2009 indicated that there was a category of mild ventriculomegaly with high prevalence counted in their data which had a good outcome and which appeared qualitatively different from the more permanent condition used to define ventriculomegaly as severe brain injury in the other trials; we therefore asked the authors to report their outcome of brain injury excluding mild ventriculomegaly.

Connelly 1999 et al conducted a very small trial in a single Canadian centre recruiting 24 infants of less than 1500 g birth weight. The trial remains unpublished except in abstract form. Infants were randomised at birth to receive transfusions at 100 g/l versus 130 g/l for the first week. These thresholds were maintained in the second week in infants requiring more than 40% oxygen, while infants requiring less oxygen were transfused at thresholds of 90 g/l and 110 g/l. During and after the third week of life all infants were treated at threshold values of 80 g/l until discharge. Transfusions of PRBCs were given in amounts calculated to restore the haemoglobin level to 150 g/l. The primary outcome was the number of transfusions received per infant. Other outcomes describing the effect of the intervention on haemoglobin levels and donor exposure were reported. A composite outcome of death or survival with major disability (adverse ultrasound findings (periventricular leukomalacia and intraventricular haemorrhage), severe retinopathy of prematurity and bronchopulmonary dysplasia) was reported. Other clinical outcomes are described in Characteristics of Included Studies.

Table 1. Included trials of haemoglobin thresholds

Study

PINT 2006

Bell 2005

Chen 2009

Connelly 1999

Transfusion volume or targets

PRBC 15 ml/kg

PRBC 15 ml/kg

PRBC 10 ml/kg

To haemoglobin of 150 g/l

Haemoglobin threshold values in g/l

Highest Level of Respiratory Support

Definition

IPPV, CPAP or supplemental oxygen

Intubated (Phase 1)

Assisted ventilation

FiO 2 less than/or equal to 0.40

Age\group

Restricted

Liberal

Restricted

Liberal

Restricted

Liberal

Restricted

Liberal

Week 1

115

135

116

156

119

153

110

130

Week 2

100

120

116

156

119

153

110

130

Week 3

85

100

116

156

119

153

80

Week 4

85

100

116

156

119

153

80

Moderate Level of Respiratory Support

Definition

IPPR, CPAP or supplemental oxygen

CPAP or supplemental oxygen

(Phase 2)

CPAP

FiO 2 less than/or equal to 0.40

Age\group

Restricted

Liberal

Restricted

Liberal

Restricted

Liberal

Restricted

Liberal

Week 1

115

135

95

129

102

136

110

130

Week 2

100

120

95

129

102

136

110

130

Week 3

85

100

95

129

102

136

80

Week 4

85

100

95

129

102

136

80

Minimal or No Respiratory Support

Definition

No additional support

No additional support

(Phase 3)

Spontaneous breathing

FiO 2 < 0.40

Age\group

Restrictive

Liberal

Restrictive

Liberal

Restrictive

Liberal

Restrictive

Liberal

Week 1

100

120

75

102

75

102

110

130

Week 2

85

100

75

102

75

102

90

100

Week 3

75

85

75

102

75

102

80

Week 4

75

85

75

102

75

102

80

All values used are for capillary haemoglobin values (converted where appropriate from haematocrit). PINT 2006 and Connelly 1999 adjusted for postnatal age, and had two levels of respiratory support; Bell 2005 and Chen 2009 did not adjust for postnatal age, but characterised three 'phases' of respiratory disease.

Connelly 1999, Bell 2005, PINT 2006 and Chen 2009 have in common the intention to compare restrictive with liberal transfusion thresholds in all including critically ill low birth weight infants. However, Connelly 1999 limits the separation to the first two weeks of life, while Bell 2005, PINT 2006 and Chen 2009 continued the separation until discharge. All four studies allowed for higher haemoglobin transfusion thresholds in infants with respiratory disease. Connelly 1999 and PINT 2006 adjusted thresholds for postnatal age. The values used for the restrictive arms of the three trials are similar and at the time of the study could be considered the intervention representing change in practice. Bell 2005 and Chen 2009 used very similar thresholds, but these included much higher haemoglobin levels for the liberal group than either Connelly 1999 or PINT 2006.

Footnote: IPPV is Intermittent Positive Pressure Ventilation, CPAP is Continuous Positive Airways Pressure.

2. Transfusion for clinical signs of anaemia (restrictive) versus transfusion at a set level of haemoglobin or haematocrit (liberal)

There was one study addressing this comparison.

Blank 1984 et al studied 56 infants in a single centre in the USA. They compared a transfusion strategy for clinical signs alone (referred to in this review as 'restrictive') with a strategy of transfusing for a haemoglobin threshold of 100 g/l (referred to in this review as 'liberal') "regardless of their clinical status". The clinical signs used for transfusing infants were: a requirement for a preoperative surgery haemoglobin level of 100 g/l or greater; tachycardia greater than 170 beats per minute for four days without primary cardiac abnormality; no weight gain for seven days with caloric intake of more than 140 kcal/kg/day and clinically notable apnoea not responsive to theophylline therapy. It is implied from the designation of the restrictive group as the 'non-transfusion group' that clinical signs of anaemia would only be thus interpreted in the presence of a haemoglobin level of less than 100 g/l. Transfusions were permitted for all infants for replacement of blood loss, exchange transfusion or pulmonary haemorrhage. This haemoglobin threshold and these clinical indications were selected "to assess the benefits of 'booster' transfusions in formerly sick premature infants". Infants of less than 1500 g were enrolled, apparently from soon after birth, and the study was ended when infants reached 1600 g weight. No sample size calculation or statement of primary outcome was made. The outcomes reported were weight gain, infants with apnoea, cost of hospitalisation and a number of hematological outcomes. This was a much earlier study and the infants in this study are likely larger and more stable than in the later three included studies (above).

Studies awaiting classification

See Characteristics of studies awaiting classification. Mukhopadhyay 2004 et al conducted a randomised trial of 98 infants of birth weight 1 to 1.8 kg requiring respiratory support within 12 hours of birth. Infants were randomised to be transfused at haematocrit thresholds of less than/or equal to 30% or less than/or equal to 40% (haemoglobin levels of 102 g/l versus 136 g/l). The primary outcome measures were changes in vital signs and in oxygen requirement. Only summary statistics were available; we have been unable to obtain sufficient details of outcomes to include this study at present.

Excluded studies

We considered and excluded four additional studies from this review. Meyer 1993 et al recruited 24 infants of less than 33 weeks gestation with haematocrit values of less than/or equal to 35% to receive either no transfusion or transfusion. However, 14 infants initially randomised, were then subsequently removed because they received clinically directed blood transfusions. The allocation of the removed infants was not reported. No further interpretation was attempted. Ross 1989 randomised 16 preterm infants of < 32 weeks gestation at haematocrit less than/or equal to 29% to 'no transfusion' or 'transfusion', but required transfusions for the 'no transfusion' group three days later. All infants were older than 32 days at randomisation, and the outcomes were limited to changes over the three days. Ransome 1989 et al randomised infants of < 34 weeks gestation to haemoglobin transfusion thresholds of 70 g/l or specifically identified symptoms versus transfusion at 100 g/l, but these infants were "clinically well" and on average, 39 days old at trial entry. Wardle 2002 et al recruited 74 infants of < 1500 g birth weight to a randomised trial where the use of near infrared spectroscopy was used to determine fractional oxygen extraction (FOE). Infants were randomised to receive transfusions if FOE equalled or exceeded 0.47 or "if significant clinical concern" was expressed by clinicians. The primary hypothesis was that infants assessed for the need for transfusion by repeated FOE measurements would receive fewer transfusions than those assessed in the conventional way using haemoglobin. Infants were excluded if they required ventilation, and due to this selection, we excluded this study from this analysis.

Risk of bias in included studies

Allocation (selection bias)

PINT 2006 and Chen 2009 used a centralised computer random allocation system, and Connelly 1999 and Bell 2005 used sealed envelopes. There was therefore, blinded randomisation in these four trials. PINT 2006 with the largest sample size (451) appears to have achieved baseline equivalence. Bell 2005 randomised 103 infants. The distribution of sex is somewhat unequal between the cohorts (61% males in the restrictive group and 41% in the liberal group). The small number randomised by Chen 2009 (36) and Connelly 1999 (24) provides little guarantee against unequal allocation of confounding variables. Chen 2009 is relatively well balanced with respect to known variables, but in Connelly 1999 the restrictive group has more males (69% versus 55%) and larger infants (mean birth weight 1088 g versus 904 g) of more advanced gestational age (mean 28.5 versus 27.3 weeks) than the liberal group.

Blank 1984 does not give details of the methods used for randomisation. With only 56 infants randomised, there is again little guarantee that known or unknown confounding variables were equally distributed. Those that are known (gestational age, birth weight and initial haemoglobin) appear to be reasonably equally distributed.

Blinding (performance bias and detection bias)

None of the trials included could blind clinical caregivers to the allocation of infants. There is some evidence of co-intervention in the PINT 2006 trial as more transfusions (16% versus 7% of transfusions given, P < 0.01) were given by 'clinical decision' (i.e. transfusions permitted in addition to the transfusion algorithm). These were not protocol violations as the protocol specified that transfusions could be given for surgery and acute hypotension. These transfusions may also be considered outcomes, as the excess transfusions were given predominantly for bleeding, surgery and sepsis/shock. Bell 2005 also reported a difference in transfusions given in addition to the study thresholds, again more often in the restrictive group (13% versus 1% of transfusions given, P < 0.001). A further 3% of transfusions were not given according to protocol in the liberal group; all protocol transfusions were given to the restrictive group. In Connelly 1999 there were seven protocol violations (all resulting in transfusion above threshold levels) among the 65 transfusions given, but only two of these instances were recorded among the 25 transfusions given in the first two weeks of life, the period in which the two interventions were different. As both violations were in the high level group the separation was maintained. In Blank 1984 the decision to transfuse the restrictive group depended on unblinded clinical observation.

Bell 2005 reports nine different measures of apnoea, some taken from bedside nursing reports. In an environment where the unblinded intervention is administered by personnel selecting the events recorded as apnoea, there is opportunity for observer bias. We have therefore selected the outcome 'apnoea receiving methylxanthine treatment' as a measure requiring intervention and therefore more likely protected from bias.

Outcome data recorded at follow-up (PINT 2006) or from ultrasound (Connelly 1999; Bell 2005; PINT 2006; Chen 2009) evaluations were collected by observers blind to the intervention.

Overall, some evidence of co-intervention can be attributed to the lack of blinding of the intervention, and this has generally acted to reduce the difference in haemoglobin levels obtained by the two strategies.

Incomplete outcome data (attrition bias)

PINT 2006 reports no losses of data at discharge and 12/223 losses from the restrictive group and 9/228 from the liberal group at follow-up. The losses all occurred in one centre owing to local interpretation of privacy regulations. Incomplete assessment at follow-up (i.e. survivors in whom not all neurodevelopmental components could be assessed) was reported in three restrictive and six liberal-allocated infants. Full reporting was therefore available at follow-up in 93% of enrolled infants. No attempt has been made to impute data in this analysis. Bell 2005 withdrew one patient from the restrictive arm and two from the liberal arm from their reported analysis who died without transfusion prior to 48 hours of age. These infants have been replaced where possible in this review. Connelly 1999 reported outcomes on all infants enrolled. We were unable to establish comparable details for Blank 1984.

Selective reporting (reporting bias)

PINT 2006 reports a post hoc analysis where the definition of a pre-specified dichotomised secondary outcome (mental developmental index, or MDI) was changed from the planned outcome, MDI < 70 (two standard deviations below the mean), to MDI < 85 (one standard deviation below the mean). The justification for including this additional outcome was explicit; that the planned outcome (MDI < 70) had yielded a difference that was very close to statistical significance. As all data analysis was controlled by a coordination centre, this was the only post hoc analysis conducted.

Bell 2005 conducted a post hoc analysis combining two planned outcomes (grade 4 intraventricular haemorrhage and periventricular leukomalacia) which on their own reached differences close to statistical significance. The difference in combined outcome reached statistical significance Bell 2006.

Chen 2009 conducted a post hoc analysis of the effects of total volume of PRBC transfusion on the incidence of oxygen dependence at 28 days, which has not been included in this review. Connelly 1999 is unpublished but the data are available in its entirety. It is not possible to report further on selective reporting in Blank 1984.

Other potential sources of bias

The Connelly 1999 trial was closed early because of poor recruitment and poor compliance, which resulted in lack of power and a relative inability to detect large differences in outcomes.

Effects of interventions

Four trials (Connelly 1999; Bell 2005; PINT 2006; Chen 2009) compared transfusions given at low versus high haemoglobin threshold, which we have termed, respectively, restrictive and liberal practices. One trial (Blank 1984) compared using clinical signs alone to transfusion at a haemoglobin threshold, which could also be considered restrictive versus liberal practice. We have considered these two comparisons separately in this review.

Unless otherwise stated, all RRs are reported as ratios of risks in the restrictive (low threshold) group to those in the liberal (high threshold) group. Risk of an adverse outcome in excess of one indicates a greater risk in the restrictive arm. Likewise, for continuous outcomes, we reported differences in clinical outcomes as restrictive-liberal.

Transfusions given for low versus high haemoglobin thresholds (Comparison 1):

The analysis is predominantly influenced by the larger PINT 2006 trial.

These trials have in common the comparison of the use of low versus high haemoglobin thresholds for transfusion. However, there are sufficient differences in the thresholds used, the indications for transfusion, and in the transfusion targets sought to consider these trials clinically as well as methodologically diverse. Taken together, they make a similar type of comparison but their estimates of outcome may be expected to differ. Many of the analyses showed statistical heterogeneity. Therefore the analyses were conducted using a random-effects rather than a fixed-effect model. As the differences in transfusion intervention resulted in different haematological outcomes, we have reported these first as they report the success of the interventions in achieving separation between the two arms of the trials. We also reported two clinical outcomes which were not predefined in the selection criteria, patent ductus arteriosus and necrotising enterocolitis, as their clinical importance has become apparent during the preparation of this review.

(a) Transfusion and haematological outcomes
Infants transfused once or more (Outcome 1.1)

All four trials reported this outcome (Connelly 1999; Bell 2005; PINT 2006; Chen 2009). All infants, including those who died, were transfused in the study of Chen 2009. Ninety-one per cent of infants in all trials combined received at least one transfusion (typical RR 0.95; 95% CI 0.91 to 1.00).

Transfusions given from study start to discharge (Outcome 1.2)

All four trials reported this outcome (Connelly 1999; Bell 2005; PINT 2006; Chen 2009). In all trials infants in the restrictive group received fewer transfusions than those in the liberal group (typical mean difference (MD) -1.12; 95% CI -1.75 to -0.49) transfusions per infant.

Donor exposures from blood products (Outcome 1.3)

Two trials reported this outcome (Bell 2005; PINT 2006). The numbers of donor exposures from red cell transfusions only were significantly reduced in the restrictive group (typical MD -0.54; 95% CI -0.93 to -0.15) exposures per infant enrolled (Outcome 1.3.1). However, when donor exposures from all sources were compared in PINT 2006 the difference was no longer significant (typical MD -0.50; 95% CI -1.65 to 0.65) (Outcome 1.3.2).

Age at first transfusion, days of study (Outcome 1.4)

In all trials the restrictive groups received their first transfusion at a later age. Three trials reported median ages (PINT 2006 three versus four days (P < 0.01), for Bell 2005 three versus eight days and for Connelly 1999 four versus eighteen days (P < 0.01)). Chen 2009 reported mean age of transfusion, and direct enquiry from the author provided median ages of 11 days in both groups (P = 0.89). These results could not be combined.

Haemoglobin levels in survivors (Outcome 1.5)

The haemoglobin separations obtained were not as great as the separation in transfusion threshold values. Bell 2005 reported on a subset of 39 infants who remained in hospital at six weeks; these data have not been included. Likewise, the data for Connelly 1999 was not consistently reported and was not included. At four weeks the MD (restrictive-liberal) was -11.0; 95% CI -13.5 to -8.5 g/l for PINT 2006 (Outcome 1.5.1). Chen 2009 reported a non-significant difference MD of 1 (95% CI -0.30 to 2.30) g/l at 30 days. These results were not combined between studies. Only PINT 2006 and Connelly 1999 reported haemoglobin levels at discharge which were not significantly different between groups (Outcome 1.5.2).

(b) Primary Outcomes
Death (Outcome 1.6)

Death rates at first hospital discharge were highest in PINT 2006 (20%) which enrolled the smallest infants (Outcome 1.6.1). Bell 2005 reported only 6% deaths and Chen 2009 8%. There were no deaths in Connelly 1999 .Typical RR for this outcome was 1.23; 95% CI 0.86 to 1.76. Only PINT 2006 reported deaths at 18 to 21 months, of 22% (RR 1.09; 95% CI 0.76 to 1.56) (Outcome 1.6.2).

Death or severe morbidity (Outcome 1.7)

Three studies reported on this composite outcome when described as death, severe retinopathy, bronchopulmonary dysplasia or brain injury on ultrasound, which was present in 68% of infants (Connelly 1999; PINT 2006; Chen 2009). There was no significant difference in this outcome at first hospital discharge (typical RR 1.07; 95% CI 0.96 to 1.20) (Outcome 1.7.1). Only PINT 2006 reported outcomes at 18 to 21 months corrected gestational age (Outcome 1.7.2). The composite outcome, a composite of death or neurodevelopmental impairment, was not significantly different between groups (RR 1.17; 95% CI 0.94 to 1.47). If a less severe limit for MDI was used for defining cognitive primary outcome a statistically significant difference was obtained for this primary outcome (RR 1.21; CI 1.01 to 1.44) (Outcome 1.7.3).

Death or severe brain injury by first hospital discharge (Outcome 1.8)

Death or severe brain injury by first hospital discharge was reported by three trials as a composite of death with brain injury (i.e. with grade IV intraventricular haemorrhage, periventricular leukomalacia or ventriculomegaly) by first hospital discharge (Connelly 1999; Bell 2005; PINT 2006). This outcome was obtained separately from the author in Chen 2009. This outcome affected 27% of infants (typical RR 1.12; 95% CI 0.81 to 1.55).

(c) Secondary Outcomes
Retinopathy of prematurity in survivors (Outcome 1.9)

All four trials reported all grades of retinopathy (Connelly 1999; Bell 2005; PINT 2006; Chen 2009). For all grades 54% of infants were affected and typical RR was 0.98; 95% CI 0.84 to 1.14 (Outcome 1.9.1). There were no significant differences between groups for either milder grades (Outcome 1.9.2) which occurred in 42% of infants (typical RR 0.96; 95% CI 0.78 to 1.18) or for severe grades of retinopathy (Outcome 1.9.3) which occurred in 14% of infants (typical RR 1.04; 95% CI 0.68 to 1.58).

Brain Injury on ultrasound in survivors (Outcome 1.10)

All four trials reported on this outcome in various ways (Connelly 1999; Bell 2005; PINT 2006; Chen 2009). PINT 2006 reported intraparenchymal echodense lesions according to the classification used by Pinto-Martin 1995 which included grade IV intraventricular haemorrhage and non-cystic periventricular leukomalacia. Bell 2005, Chen 2009 and Connelly 1999 report periventricular leukomalacia or grade IV intraventricular haemorrhage. Further details of these outcomes were provided by the first author of Chen 2009. All trials have expressed a measure of severe brain injury including grade IV intraventricular haemorrhage, periventricular leukomalacia, and ventriculomegaly, and these have been combined in this analysis. Severe brain injury occurred in 16% of survivors (typical RR 1.07; 95% CI 0.50 to 2.27).

Bronchopulmonary dysplasia in survivors (Outcome 1.11)

All four trials reported infants requiring oxygen at 28 postnatal days and at 36 weeks postmenstrual age (Connelly 1999; Bell 2005; PINT 2006; Chen 2009). The second definition is more consistent with current definitions of severe bronchopulmonary dysplasia (Shennan 1988). One infant excluded by Chen 2009 in the liberal group with earlier pneumonia was included in this group. The typical RR for oxygen requirement at 28 days (Outcome 1.11.1) which was present in 74% of infants was 0.99; 95% CI 0.92 to 1.06. 49% of infants required supplementary oxygen at 36 weeks postmenstrual age (Outcome 1.11.2), typical RR 1.03; 95% CI 0.87 to 1.21.

Neurosensory impairment at 18-21 months follow-up among survivors (Outcome 1.12)

Only PINT 2006 reported this outcome, which is analysed by the numbers of surviving infants who could be evaluated. All of these components reported increased risks in the restrictive group which did not reach statistical significance. Cognitive delay was measured as the MDI of the Bayley score of < 70, which is two standard deviations below the mean. This outcome alone was analysed in the original report as an odds ratio (OR), adjusted for gestational age and study site. This analysis for cognitive delay reached borderline statistical significance favouring the liberal strategy (adjusted OR 1.74; 95% CI 0.98 to 3.11). Therefore a post hoc analysis was conducted redefining the criterion for cognitive delay as MDI < 85 (one standard deviation below the mean). With this interpretation the adjusted OR for cognitive delay was significantly increased with the restrictive strategy (adjusted OR 1.81; 95% CI 1.1 to 1.8. When expressed as unadjusted RR in this analysis, the RR for cognitive delay is 1.39; 95% CI 0.90 to 2.13 for MDI < 70, where 21% of surviving infants had this outcome (Outcome 1.12.1). With cognitive index defined as MDI < 85, which affected 39% of survivors, the RR is 1.32; 95% CI 1.00 to 1.74 (Outcome 1.12.2). Other components of this outcome were cerebral palsy (Outcome 1.12.3) (RR 1.29; 95% CI 0.55 to 3.03), severe visual impairment (Outcome 1.12.4) (RR 2.15; 95% CI 0.20 to 23.5) and severe hearing impairment (Outcome 1.12.5) (RR 1.43; 95% CI 0.33 to 6.30). Combining all neurosensory impairments (including cognitive delay as MDI < 70) as any neurosensory impairment (Outcome 1.12.6) yielded an event rate of 25% of survivors and a RR of 1.31; 95% CI 0.90 to1.90.

Apnoea requiring intervention (Outcome 1.13)

All four trials reported on the numbers of surviving infants requiring treatment for apnoea (Connelly 1999; Bell 2005; PINT 2006; Chen 2009). However PINT 2006 defined this as infants requiring intervention with bag-and-mask ventilation or intubation. Bell 2005 and Connelly 1999 defined this outcome as infants requiring methylxanthines for therapy of apnoea. The definition used by Chen 2009 (numbers of infants with absence of breathing for more than 20 seconds or a shorter pause (more than 10 seconds) associated with oxygen desaturation or bradycardia) was obtained from the author. Apnoea was reported in 76% of infants, with a typical RR of 1.01; 95% CI 0.95 to 1.08.

Weight gain in survivors was reported as different measures in each study. PINT 2006 reported a difference in absolute weight gain at 32 weeks gestation of -18 g; 95% CI -69 to 33 (Outcome 1.14). Chen 2009 reported weight gains in g for each postnatal week one to five and showed no significant difference between groups at any period. Connelly 1999 reported mean daily rate of weight gain to discharge as 15.1 g/d and 18.0 g/d in restrictive and liberal groups respectively, MD -2.90 (95% CI -7.34 to 1.54) g/d (Outcome 1.15). Bell 2005 reported time to double birth weight as median values of 60 and 59 days in restrictive and liberal groups respectively (P = 0.70) (Outcome 1.16); Chen 2009 also provided median values for this outcome of 61 and 72 days (P = 0.53). These results could not be combined, but none of the studies showed a significant effect of transfusion strategy on weight gain.

Length of hospitalisation (days) (Outcome 1.17)

Bell 2005, PINT 2006 and Chen 2009 reported this outcome. In restrictive and liberal groups respectively, days of hospitalisation were medians 74 and 73 days in PINT 2006 (P = 0.39) and 76 and 69 days in Bell 2005 (P = 0.45). Direct enquiry from Chen 2009 provided median data for this outcome of 67 and 65 days. These results could not be combined, but none of the studies showed a significant effect of transfusion strategy on length of stay.

(d) Outcomes that were not selection criteria but were reported in the included trials (post hoc analysis):
Patent ductus arteriosus (Outcome 1.18)

Patent ductus arteriosus requiring therapy with indomethacin was reported by Bell 2005, PINT 2006 and Chen 2009 reported the outcome patent ductus arteriosus as defined by echocardiography (information from author). Patent ductus arteriosus was reported in 54% of infants in the three trials combined and occurred with almost equal frequency in restrictive and liberal groups (typical RR 0.93; 95% CI 0.76 to 1.14) (Outcome 1.18.1). Sixteen per cent of all infants required surgery for patent ductus arteriosus in two trials (Bell 2005; PINT 2006) (Outcome 1.18.2): typical RR was 1.31; 95% CI 0.89 to 1.91.

Necrotising enterocolitis (Outcome 1.19) was reported by Bell 2005 and PINT 2006 as cases confirmed at surgery, autopsy or by finding pneumatosis intestinalis and hepatobiliary gas. Chen 2009 defined necrotising enterocolitis as meeting Bell's criteria (Bell 1978). Necrotising enterocolitis was reported in 6% of infants (typical RR 1.62; 95% CI 0.83 to 3.13).

Transfusions given for clinical signs only versus for haemoglobin threshold (Comparison 2)

The Blank 1984 report is the only included trial addressing this comparison. The published report is lacking in detail. The numbers of transfusions given with restrictive versus liberal strategies is not reported. However, there is graphical information which indicates that large differences in transfusion volumes resulted from the interventions; for example at six weeks postnatal age the percentage of haemoglobin A (most likely almost entirely from transfused blood) in the restrictive group was 58% compared with 89% in the liberal group. The only clinical indication used for transfusing the four infants transfused in the restrictive group was apnoea. The data are reported here in the order restrictive (transfusion for clinical signs) versus liberal (at a haemoglobin level of 100 g/l regardless of clinical signs), and the RRs are expressed as restrictive/liberal.

(a) Transfusion and hematological outcomes

The numbers of transfusions given were not reported. The effects of transfusion, manifested as decreased levels of haemoglobin F and of p50, are presented graphically and reported as statistically significant (P = < 0.05). From this we may conclude that the liberal group received significantly more transfused adult blood than the restrictive group. There were no reports of numbers of infants transfused once or more, of donor exposures from red cell transfusions or of age at first transfusion.

Discharge haemoglobin [g/l] (Outcome 2.1)

The haemoglobin at study discharge was statistically significantly lower in the restrictive group (91 versus 117 g/l, difference -26, 95% CI -35 to -17 g/l), which makes it likely that large differences in transfusion frequency were obtained between the groups.

(b) Primary Outcomes
Death prior to discharge (Outcome 2.2)

There were no deaths in either group.

Death or severe morbidity as a composite outcome was not reported.

Death or severe adverse neurosensory outcome at a defined period of follow-up as a composite outcome was not reported.
(c) Secondary outcomes
Presence of apnoea (Outcome 2.3)

There was no significant difference in the numbers of infants with apnoea (which was not defined) (RR 0.97; 95% CI 0.66 to 1.43).

Days to regain birth weight (Outcome 2.4)

Means were 27 versus 26 days; MD 1; 95% CI -5 to 6 days.

Length of hospitalisation (outcome 2.5)

Means were 49 and 51 days; MD -2; 95% CI-13 to 9 days.

Costs of hospitalisation (Outcome 2.5)

Means were $3642 versus $3430; MD $212; 95% CI -446 to 870.

The following secondary outcomes were not reported: retinopathy of prematurity; brain injury on ultrasound; bronchopulmonary dysplasia; neurosensory impairment at follow-up; and postnatal acquisition of viral infection.

Outcomes not reported in either comparison

We did not find any trials which reported measures of cost-effectiveness of blood transfusion in infants, or of postnatal acquisition of viral infection.

Discussion

Summary of main results

Trials comparing restrictive versus liberal haemoglobin thresholds

Four randomised trials were identified (the smallest unpublished) which compared the effects of giving blood transfusions to very low birth weight infants at low (restrictive) versus high (liberal) haemoglobin or haematocrit thresholds. Of 614 patients randomised, 73% were enrolled in the PINT 2006 trial. The trials differed in the birth weights of the patients enrolled and in the transfusion thresholds used. The restrictive thresholds used were similar, but the limits used for the liberal thresholds varied, as did the transfusion targets (see table in Included studies) . The four trials have in common the exploration of the effects of using a transfusion threshold lower than conventional, and it is reasonable to compare their outcomes. The trials compare the effects of different thresholds used for maintenance or 'top-up' transfusions and not indications used for emergency transfusion. Overall, a restrictive practice produced very little reduction of the numbers of babies receiving at least one transfusion (typical RR 0.95; 95% CI 0.91 to 1.00) but the effect varied greatly between studies, being greatest where the transfusion was made to a haemoglobin goal rather than as a set volume. The differences in mean haemoglobin levels resulting from the strategy were considerably smaller than the levels used as transfusion thresholds and were sometimes undetectable.

Restrictive versus liberal transfusion policy within the specific thresholds chosen had little effect on mortality at first hospital discharge (typical RR 1.23, 95% CI 0.86 to 1.76) or on the composite outcome, reported from two trials, of death or severe morbidity at discharge (typical RR 1.07; 95% CI 0.96 to 1.20). Three trials reported a composite outcome of death or brain injury which gave a typical RR of 1.12; 95% CI 0.81 to 1.55.

One follow-up study was conducted at 18 months corrected age in the PINT 2006 study. There was no significant difference in the primary outcome of death or adverse neurodevelopmental outcome (RR 1.17; 95% CI 0.94 to 1.47) or in the components of this composite outcome (death, cognitive delay, severe visual impairment or severe hearing impairment), but the confidence limits were large; significant risk attributable in particular to the use of low haemoglobin thresholds cannot be ruled out. The criterion for cognitive delay was a score of < 70 on the Bayley mental developmental index (MDI). Using a less severe definition (MDI < 85) gave a typical RR for the composite primary outcome of 1.21 (95% CI 1.01 to 1.44) favouring the liberal strategy. The argument for using this interpretation for the primary outcome is more clearly made in the original report, in which the analysis was adjusted for site and gestational age.

Specific disorders which have been thought to be related to transfusion policy have included necrotising enterocolitis, retinopathy of prematurity and bronchopulmonary dysplasia. No effect on these outcomes was attributable to restrictive or liberal policy.Within the limits of the transfusion strategies compared, and the differences in haemoglobin levels obtained, there was no reduction of weight gain attributable to a restrictive policy. The numbers of infants who required treatment for apnoea were not different between the strategies.

Transfusion for clinical signs only versus for transfusion at haemoglobin threshold

The single randomised trial of Blank 1984 compared the restrictive practice of waiting for symptoms attributable to anaemia to appear prior to transfusing versus the then current liberal practice of transfusing at a fixed level of haemoglobin (100 g/l). These infants were larger and healthier (there were no deaths) than those reviewed in the previous comparison. The effect of these practices on transfusion frequency was not documented, but the large differences in haemoglobin A levels, and a persistent difference of 26 g/l in haemoglobin seen at discharge from the study suggests very large differences in transfusion volumes received. The incidence of apnoea was high and no difference was found between groups (RR 0.97; 95% CI 0.66 to 1.43). This study, though underpowered, contributes to our understanding that rate of weight gain and incidence of apnoea are not outcomes affected by transfusion strategy.

Overall completeness and applicability of evidence

There is no clear benefit or risk attributable to the use of low versus high haemoglobin transfusion thresholds in very low birth weight infants at 18 months follow-up, but this conclusion is derived from a single trial powered for short term outcomes, and the possibility of undetected significant clinical effects cannot be ignored. For outcomes reported at hospital discharge, the conclusion of no benefit or risk from two trials is made with more confidence (RR 1.07, 95% CI 0.95 to 1.20). These uncertainties, combined with observations from post hoc analyses suggesting poorer outcomes in neurodevelopment PINT 2006 and at hospital discharge Bell 2005, make it difficult to reject a conclusion that the use of a high threshold may be beneficial. It can be said with more certainty that there is no justification for allowing haemoglobin transfusion thresholds to fall below the lower limits used in these studies. The single trial by Blank 1984 comparing transfusion for clinical signs only versus transfusion at a fixed haemoglobin level of 100 g/l is similarly inconclusive, as it is not possible to assess the completeness of the data or the numbers of infants lost to the analysis. There is no evidence from any trial in this review to support the association of transfusion strategies with apnoea, rate of weight gain or necrotising enterocolitis.

Quality of the evidence

This review consists of five randomised controlled trials in which there appears to be no allocation bias; the overall level of evidence is high. No attempts were made to blind the intervention in any trial, and there is some evidence of co-intervention. The outcomes included in the review are generally not susceptible to measurement or judgement bias (e.g. death) or measured by observers blind to the intervention(cranial ultrasound findings and follow-up evaluations).

Potential biases in the review process

Connelly 1999 is unpublished and has not been subjected to peer review. RKW is a co-author of two trials (PINT 2006 and Connelly 1999) reported in this review, and HK is a co-author of the PINT 2006 study. There is a potential for conflict of interest in the selection of trials for this review.

Agreements and disagreements with other studies or reviews

There are no previous systematic reviews including the Connelly 1999, Bell 2005 or PINT 2006 studies. An earlier systematic (non-Cochrane) review (Hume 1997) reviewed the literature up to 1996, which concluded that stable asymptomatic infants with haemoglobin levels greater than 7.0 g/l did not require red blood cell transfusions, which is in agreement in part with the conclusions of this review. Trials in which the haemoglobin has been increased with erythropoietin versus placebo, with restrictive transfusion thresholds similar to those described in this review have not shown differences in major outcomes and therefore appear to confirm the general findings of this review with respect to the safety of the lower limits for haemoglobin (Shannon 1995; Ohls 2001). Claims of uncontrolled studies (Stockman 1984a) which attributed poor weight gain to low haemoglobin levels in excess of the lower limits reviewed here have not been substantiated. The finding in this review that restrictive therapy has no effect on numbers of infants requiring treatment with apnoea is contrary to uncontrolled observations of individual apnoea frequency responding to transfusion (Joshi 1987; Sasidharan 1992); these effects are thought to be attributable to volume expansion and not specifically to blood transfusion (Bifano 1992). The overall conclusion that restrictive therapy does not affect survival is consistent with the findings of very similar randomised controlled trials (RCTs) conducted in children in paediatric intensive care (Lacroix 2007) and in adult intensive care (Hébert 1999).

Authors' conclusions

Implications for practice

The use of restrictive as compared to liberal haemoglobin transfusion thresholds in infants of very low birth weight results in modest reductions in exposure to transfusion and in haemoglobin levels. There is no evidence that using a lower haemoglobin transfusion threshold (using the limits tested in these trials) has an effect on mortality, major morbidities or on survival without major morbidity in very low birth weight infants. As the restrictive levels used were more similar among trials, a summarised approximation of the lower thresholds evaluated is presented in Table 2. Safety at haemoglobin levels below these lower limits has not been evaluated and these should be maintained until further evidence is available. As the trials differed greatly in high threshold values, we have not attempted such an approximation for the higher levels.

Table 2. Approximate lower limits for capillary haemoglobin and haematocrit thresholds evaluated in this review

Postnatal Age

Respiratory Support

No Respiratory Support

Haemoglobin g/l (Haematocrit %)

Week 1

115 (35%)

100 (30%)

Week 2

100 (30%)

85 (25%)

Week 3

85 (25%)

75 (23%)

Implications for research

The safe lower limits for haemoglobin transfusion thresholds remain undefined, and there is still uncertainty regarding the benefits of maintaining a higher level. These uncertainties affect decisions based on maintaining haemoglobin targets however these are achieved, be it by transfusion, erythropoietin therapy, placental transfusion or restrictions on blood sampling, although there are functional differences in the performance of haemoglobin to address when these interventions are compared. There are sufficiently provocative findings from reported secondary outcomes to justify a further larger clinical trial with primary outcomes which address the issues identified as secondary in these trials. Studies comparing the cost-effectiveness of strategies for maintaining haemoglobin levels in very low birth weight infants would also be helpful.

Acknowledgements

We thank Ms Darlene Chapman MLIS, Manager, Information Services, IWK Health Centre, who conducted and updated the searches used in this review and Dr Ellen M Bifano MD, Department of Newborn Medicine/Neonatology Crouse Irving Memorial Hospital, Syracuse New York, USA who contributed to the protocol phase of this review.

Contributions of authors

Both authors (HK and RKW) have contributed to the research and writing of this review. RKW is the guarantor of the review.

Declarations of interest

The authors have no conflicts of interest to declare.

Differences between protocol and review

  • None noted.

Additional tables

  • None noted.

Potential conflict of interest

  • None noted.

[top]

Characteristics of studies

Characteristics of Included Studies

PINT 2006

Methods

Block randomised into two groups stratified by birth weight ≦ 750 g or > 750g and study centre

Participants

Multicentre, Canada, USA, Australia. Viable infants < 1000 g birth weight up to 48 hours of age. Infants transfused within six hours of birth could be included

Interventions

Packed red blood cell transfusions were administered at threshold capillary blood haemoglobin levels: for infants receiving respiratory support (assisted ventilation, continuous positive airway pressure or supplemental oxygen) these were (restrictive versus liberal) 115 g/l versus 135 g/l in postnatal week one, (restrictive versus liberal) 100 versus 120 g/l in postnatal week two and 85 versus 100 g/l in postnatal week three until discharge. For infants not requiring respiratory support the triggers were 100 g/l versus 120 g/l in postnatal week 1, 85 g/l versus 100 g/l in postnatal week 2, and 75 g/l versus (restrictive versus liberal) 85 g/l in postnatal week 3 until discharge. When central rather than capillary blood was sampled, thresholds were 10% lower

Transfusion volume was 15 ml/kg.

Physicians were permitted to give additional transfusions in case of shock, severe sepsis, coagulation defects, surgery or unanticipated emergencies

Outcomes

Outcomes from the study that are considered in the review:

A. Before discharge from initial hospitalisation
Transfusion and hematological outcomes
  • infants ever transfused (i.e. transfused once or more);
  • transfusions per infant from study start to study end;
  • donor exposures (a) from red cell and (b) from all blood products;
  • postnatal age at first transfusion; and
  • haemoglobin levels at 4 weeks in survivors, and at discharge
Primary outcome (composite):
  • death or severe morbidity in survivors, as defined below; and
  • death or severe brain injury in survivors, as defined below
Secondary outcomes:
  • death; and
  • severe morbidity in survivors:
    • retinopathy of prematurity stages 3-5
    • bronchopulmonary dysplasia as oxygen requirement at 36 weeks gestation
    • severe brain injury: presence of cystic periventricular leukomalacia, intraparenchymal echodensity (which includes but is not limited to grade IV intraventricular haemorrhage), porencephalic cyst or ventriculomegaly (Pinto-Martin 1995) on the worst cranial ultrasound
Other outcomes:
  • retinopathy of prematurity stages 1-2;
  • cystic PVL and porencephalic cyst);
  • intraparenchymal echodense lesion;
  • bronchopulmonary dysplasia (oxygen dependence at 28 days): data obtained from authors;
  • apnoea requiring intervention (bag and mask resuscitation);
  • necrotising enterocolitis confirmed at surgery, autopsy, or by finding pneumatosis intestinalis, hepatobiliary gas;
  • patent ductus arteriosus requiring treatment with indomethacin or surgery;
  • weight gain at 32 weeks; and
  • head circumference at 32 weeks
B. At 18-21 months adjusted gestational age
Primary outcome (composite):
  • mortality or severe adverse neurodevelopmental outcome (as defined below)
Secondary outcomes:
  • mortality; and
  • severe adverse neurodevelopmental outcome:
    • cerebral palsy
    • cognitive delay (Bayley Mental Developmental Index (MDI) Score < 70)
    • severe visual impairment
    • severe hearing impairment.
Number Enrolled

451

Notes

No erythropoietin was given

Risk of bias table
Bias Authors' judgement Support for judgement
Allocation concealment (selection bias) Low risk

Adequate, by centralised concealed computer allocation.

Blinding (performance bias and detection bias) High risk

Caregivers could not be blinded to the intervention. There is evidence of co-intervention (performance bias) in that the mean numbers of transfusions per infant given for 'clinical reasons' i.e. in addition to the transfusion threshold, was 0.8, standard deviation (SD 2.0) in the restrictive group and 0.4 (SD 1.3) in the liberal group. (P = 0.007). This accounts for 16% of the transfusions given to the restrictive group and 7% of those given to the liberal group (P < 001). Transfusions given for clinical reasons were permitted and were therefore co-interventions rather than violations. Some additional transfusions could represent an outcome, such as an increased tendency for bleeding, surgery or sepsis resulting from allocation to the restrictive group

Incomplete outcome data (attrition bias) Low risk

One centre lost 12 (low threshold) and 9 (high threshold) infants from follow-up because of intervention by the local Institutional Review Board. A further 3 infants (low threshold) and 6 infants (high threshold) could not be categorised for neurodevelopmental outcomes because of poor compliance

Selective reporting (reporting bias) Unclear risk

A post hoc analysis was conducted where the incidence of cognitive delay was redefined as MDI Score < 85

Bell 2005

Methods

Block randomisation into two groups (restrictive and liberal transfusion thresholds), stratified by birth weight into 3 groups by birth weight 500 to750g, 751 to1000g and 1001 to 1300 g. Allocation was balanced by blocking every 4 enrolments within each stratum. Haematocrits were determined from arterial or capillary blood without distinction

Participants

Single centre, USA. 100 viable infants of 500 to 1300g birth weight. Infants who had been transfused > 2 times were excluded. Infants with alloimmune haemolytic disease, major birth defects or significant patent ductus arteriosus were excluded

Interventions

Packed red blood cell transfusions were administered at threshold capillary blood hematocrit levels. Threshold haematocrits depended on respiratory status defined as one of three levels: intubated (phase 1), receiving oxygen or continuous distending pressure (phase 2) or unsupported in room air (phase 3). Threshold haematocrits for phase 1 were (restrictive versus liberal) 46% versus 34%, phase 2 (restrictive versus liberal) 38% versus 28% and phase 3 (restrictive versus liberal) 30% versus 22% up to hospital discharge. Haematocrits falling below threshold were repeated by resampling; both values were required to be below threshold to qualify for transfusion. Additional transfusions were permitted at neonatologist's request for unexplained congestive heart failure assumed to be caused by anaemia, acute haemorrhage and presumed hypovolaemia, frequent or severe apnoea resistant to drug treatment, or request by a surgeon or anaesthesiologist for preoperative transfusion

Outcomes

There was no follow-up report.

Outcomes from the study that are considered in the review:

A. Prior to first discharge home.
Transfusion and hematological outcomes:
  • infants ever transfused (i.e. transfused once or more);
  • transfusions per infant from study start to study end (primary outcome);
  • donor exposures from red cell transfusions; and
  • postnatal age at first transfusion
Secondary outcomes:
  • death; and
  • death or severe brain injury in survivors, as defined below
Outcomes in survivors:
  • severe brain injury in survivors: (intraventricular haemorrhage grade IV or periventricular leukomalacia);
  • bronchopulmonary dysplasia (oxygen dependence at 28 days and at 36 weeks);
  • retinopathy of prematurity (total and stages 3-5);
  • intraventricular haemorrhage (all grades and grades III and IV);
  • periventricular leukomalacia;
  • number of infants with apnoea requiring methylxanthine treatment;
  • time to regain and to double birth weight; and
  • length of hospitalisation
Additional outcomes:
  • necrotising enterocolitis confirmed at surgery, autopsy, or by finding pneumatosis intestinalis, hepatobiliary gas (data obtained from author); and
  • patent ductus arteriosus requiring treatment with indomethacin and with surgery
Number Enrolled

103

Notes

There was no age limit to enrolment
Published analysis excludes three infants who died within 48 hours: these have been replaced in this analysis. No erythropoietin was given
Assuming MCHC of 340 g/l, equivalent haemoglobin concentrations are approximately (restrictive versus liberal) phase 1: 116 g/l versus 156 g/l, phase 2: 95 g/l versus 129 g/l and phase 3: 75 g/l versus 102 g/l.
Haematocrits were determined from arterial or capillary blood without distinction
Haematocrits were measured according to a regular schedule. Transfusion required two consecutive values below threshold
Intervention was not blinded

Risk of bias table
Bias Authors' judgement Support for judgement
Allocation concealment (selection bias) Low risk

Adequate from random number tables with concealed allocation by numbered, sealed envelopes

Blinding (performance bias and detection bias) High risk

Caregivers could not be blinded to allocation or the intervention. Clinical outcomes, especially of apneic events not requiring treatment, susceptible to bias. There was an excess of transfusions given in addition to the study thresholds, more often in the restrictive group (12.7% versus 0.8% of transfusions given in restrictive versus liberal groups ) these were co-interventions rather than protocol violations, as transfusions were permitted in this trial, notably for severe apnoea and congestive cardiac failure). In addition, 3% of transfusions prescribed by study protocol in the liberal group were not given; all prescribed transfusions were given to the restrictive group. Additional transfusions could represent an outcome, such as an increased tendency for bleeding, surgery or sepsis resulting from allocation to the restrictive group. Ultrasound evaluations were by blinded assessment

Incomplete outcome data (attrition bias) High risk

No data appear to be missing, but infants who died very early in the study were reported but excluded from analysis. These have been reintroduced in this analysis. A subset of 39 of the 103 infants enrolled were selected as (a) remaining in hospital and (b) suitable for physiologic measurements (either intubated or in room air, information from author) performed before the first transfusion given after the age of 6 weeks. These studies were haemoglobin, haematocrit, reticulocyte count, oxygen saturation, cardiac output (by echocardiography, and not measured if a significant shunt was present)), blood lactic acid, plasma erythropoietin and serum ferritin. The selection of infants and the timing of sampling could each have been affected by the study intervention and therefore could not be represented as randomly available. These data were therefore excluded from the analysis.

Ultrasound examinations were performed at 7 days of age in all infants and at 14 and 42 days of age in infants who were born at < 27 weeks’ gestation or who had an abnormal cranial ultrasound examination at 7 days of age. All surviving 97 subjects had day 7 ultrasound examinations and 52 (28 restrictive; 24 liberal) had day 42 examinations. Therefore it is assumed that the 45 infants who were both ≧ 27 weeks gestation and had normal 7 day ultrasounds would have had normal ultrasounds at day 42 if examined

Selective reporting (reporting bias) High risk

One major outcome (combined grade 4 IVH or periventricular leukomalacia) was added after the primary analysis

Chen 2009

Methods

Randomised into two groups (restrictive and liberal transfusion)

Participants

Single centre in Taiwan. Preterm infants with birth weight less than 1500 g of up to three days of age. Exclusions were any prior transfusion, major birth defects or chromosomal anomalies

Interventions

Transfusion when haematocrit fell below (restrictive) 35% in infants with assisted ventilation, 30% in infants with continuous positive airways pressure, or 22% in infants breathing spontaneously versus (liberal) 45% in infants with assisted ventilation, 40% in infants with continuous positive airways pressure or 30% in infants breathing spontaneously

Outcomes

Outcomes from the study that are considered in the review: there was no follow-up report

Transfusion and hematological outcomes at 30 days
  • infants ever transfused (i.e. transfused once or more);
  • transfusions per infant from study start to study end (primary outcome);
  • postnatal age at first transfusion; and
  • haemoglobin at 30 days
Primary clinical outcomes:
  • death or severe morbidity in survivors, as defined as retinopathy of prematurity Stages 3-5, bronchopulmonary dysplasia as oxygen requirement at 36 weeks gestation or severe brain Injury, defined as presence of cystic periventricular leukomalacia, intraparenchymal echodensity (which included lesions described in other studies as grade IV intraventricular haemorrhage), porencephalic cyst or ventriculomegaly on the worst cranial ultrasound; and
  • death or severe brain injury in survivors, as defined above
Secondary clinical outcomes:
  • death; and
  • outcomes in survivors (prior to first hospital discharge)
    • intraventricular haemorrhage all grades
    • bronchopulmonary dysplasia (oxygen dependence at 28 days and at 36 weeks)
    • retinopathy of Prematurity (all grades)
    • numbers of infants with apnoea (as reported by bedside nurse)
    • time to regain and to double birth weight
    • length of hospitalisation
Additional outcomes:
  • necrotising enterocolitis radiographically diagnosed using Bell criteria (Bell 1978); and
  • patent ductus arteriosus defined by echocardiography
Number Enrolled

36

Notes

No erythropoietin was given

Risk of bias table
Bias Authors' judgement Support for judgement
Allocation concealment (selection bias) Low risk

Allocation was by computer random allocation, concealed and committed (information from author)

Blinding (performance bias and detection bias) Unclear risk

Author reports that study nurse reviewed the chart and recorded the outcomes, and did not know the study allocation. It is not clear if there were opportunities for unblinding for all the outcomes by any of the staff applying the transfusion algorithms

Incomplete outcome data (attrition bias) Low risk

Author provided details on three infant deaths not originally included in clinical outcomes

Selective reporting (reporting bias) Unclear risk

There was no documented trial preregistration

Connelly 1999

Methods

Randomised by concealed allocation (envelope) into two groups

Participants

Single Centre, Canada. Viable infants < 1500 g birth weight up to 72 hours of age. Infants who had been transfused were not included

Interventions

Packed red blood cell transfusions were administered at threshold central blood haemoglobin levels: for infants in postnatal week one these were (restrictive versus liberal) 110 g/l versus 130 g/l. For infants in postnatal week two, the thresholds depended on respiratory status: those requiring > 40% oxygen maintained week one thresholds, while those not requiring this level of oxygen support used thresholds of (restrictive versus liberal) 90 g/l versus 100 g/l. After week two the trial ended, with all infants transfused at 80 g/l. When capillary rather than central blood was sampled, thresholds were 4% higher

Infants were transfused to the same targets, being 150 g/l in postnatal week one, 130 g/l in week two and 120 g/l in week three

Physicians were permitted to give additional transfusions in case of shock, severe sepsis, coagulation defects, surgery or unanticipated emergencies

Outcomes

There was no follow-up report

Outcomes from the study that are considered in the review:

Transfusion and hematological outcomes:
  • infants ever transfused (i.e. transfused once or more);
  • transfusions per infant from study start to study end (Primary outcome of this study);
  • postnatal age at first transfusion; and
  • haemoglobin levels at weeks 1 to 4 in survivors, and at discharge
Clinical outcomes addressing primary outcome of review:
  • death or severe morbidity in survivors, as defined below; and
  • death or severe brain injury in survivors, as defined below
Severe morbidity in survivors:
  • retinopathy of prematurity stages 3-5;
  • bronchopulmonary dysplasia as oxygen requirement at 36 weeks gestation; and
  • severe brain injury: presence of grade IV intraventricular haemorrhage or periventricular leukomalacia
Other outcomes:
  • death;
  • retinopathy of prematurity stages 1-2;
  • intraventricular leukomalacia;
  • grade IV intraventricular haemorrhage;
  • bronchopulmonary dysplasia (oxygen dependence at 28 days);
  • apnoea requiring intervention with methylxanthines;
  • weight gain at 32 weeks; and
  • head circumference at 32 weeks
Number Enrolled

24

Notes

Two transfusions in the liberal group and none in the restrictive group were given before the thresholds were reached and were therefore violations. Seven transfusions were given before the 80 g/l haemoglobin threshold was met in the liberal group versus four in the restrictive group in the later weeks when the threshold was the same for all infants. There were no violations of failure to transfuse at threshold

Risk of bias table
Bias Authors' judgement Support for judgement
Allocation concealment (selection bias) Low risk

Adequate, by sealed envelope and logging of eligible patients

Blinding (performance bias and detection bias) High risk

Caregivers could not be blinded to allocation or the intervention. Clinical outcomes, especially of apneic events not requiring treatment, susceptible to bias. There were 7 protocol violations (all resulting from transfusion before and therefore above the threshold levels) among the 65 transfusions given, but only 2 of these instances were recorded among the 25 transfusions given in the first two weeks of life, the period over which the two different regimes were compared; both of these involved transfusion at higher than prescribed levels, but as both were in the high level group the separation was maintained

Incomplete outcome data (attrition bias) Low risk

All outcomes were complete

Selective reporting (reporting bias) Low risk

These data are unpublished. All study data collected were accessible to the reviewers

Blank 1984

Methods

Randomised (no details available) into two groups

Participants

Single Centre, USA. Infants of < 1500 g birth weight, of no specified age

Interventions

Babies in the restrictive group, referred to in the report as group B, were given blood only for specific indications: requirement for a preoperative surgery haemoglobin level of 100 g/l or greater; tachycardia greater than 170 beats per minute for four days without primary cardiac abnormality; no weight gain for seven days with caloric intake of more than 140 calories/kg/day; or clinically notable apnoea not responsive to theophylline therapy

Babies in the liberal group, referred to in the report as group A, received blood only when their haemoglobin level fell below 100 g/l, ‘regardless of their clinical status’. Our interpretation here of ‘regardless of their clinical status’ meant that the liberal group would not receive transfusions for the clinical reasons specified for the restrictive group

All infants received packed RBCs for specific clinical indications, including iatrogenic blood loss due to laboratory determinations, exchange transfusion for hyperbilirubinaemia, and conditions associated with acute blood loss, such as pulmonary haemorrhage

Outcomes

There were no measures of transfusions numbers or volumes, or of donor exposure. There were no measures of ultrasound findings of brain injury, retinopathy of prematurity or of bronchopulmonary dysplasia. There was no follow-up report

Outcomes from the study that are considered in the review:

At study discharge when infants reached 1600 g weight, the following outcomes were reported:

  • discharge haemoglobin concentration;
  • death;
  • infants with apnoea;
  • days to gain birth weight; and
  • length of hospitalisation
Number Enrolled

56

Notes

This is an older trial published more than 25 years before this review and many details of the conduct of the trial are unavailable

Risk of bias table
Bias Authors' judgement Support for judgement
Allocation concealment (selection bias) Unclear risk

Unclear. Details of randomisation not available

Blinding (performance bias and detection bias) High risk

Allocation could not be concealed from caregivers. Clinical outcomes, especially of apneic events not requiring treatment, susceptible to bias

Incomplete outcome data (attrition bias) Unclear risk

No reference to missing data or infants removed from the study

Selective reporting (reporting bias) Unclear risk

There was no documented trial preregistration

Characteristics of excluded studies

Meyer 1993

Reason for exclusion

Although infants were randomised soon after birth, this study was primarily designed to test the effect of later transfusions for anaemia of prematurity, as evidenced by the rationale given for the choice of haematocrit level (to match the upper limit of the reported range for similar infants at eight weeks of age) and the mean time of transfusion (28 days, SD 4 days). Infants were, by design, removed from the randomised groups if transfusion was required on clinical grounds, leaving two groups of infants transfused or not transfused to maintain a haematocrit > 35%, provided they had not been removed from the trial on clinical grounds. This was not interpretable on an intention to treat (ITT) principle

Ransome 1989

Reason for exclusion

Infants were selected as "clinically well' and on average 39 days old at recruitment. This review addresses infants transfused soon after birth and includes infants receiving critical care

Ross 1989

Reason for exclusion

Infants ≦ 32 wk gestation 'scheduled to be transfused' at haematocrit < 29% were randomised to either immediate transfusion with 10ml/kg PRBC or a 3 day delayed transfusion with the same transfusion volume. Outcomes, all short term, were change in respiratory rate, heart rate or oxygen requirement, rate of weight gain, episodes of apnoea bradycardia, blood lactate concentration and measures of erythropoiesis. This was essentially a three day trial without long term outcomes

Wardle 2002

Reason for exclusion

In this randomised trial near infrared spectroscopy was used to determine peripheral fractional oxygen extraction (FOE). Infants were randomised to receive transfusions if FOE fell below 0.47 or "if significant clinical concern" was expressed by clinicians. The primary hypothesis was that infants assessed for the need for transfusion by repeated FOE measurements would receive fewer transfusions than those assessed in the conventional way using haemoglobin. Infants were excluded if they required ventilation, and this selection led to the exclusion of this study from this analysis. This was essentially a feasibility study

Characteristics of studies awaiting classification

Mukhopadhyay 2004

Methods

Randomisation by unspecified means

Participants

Preterm infants of birth weight 1 to 1.8 kg who required mechanical ventilation or continuous positive airway pressure for the first 12 hours

Interventions

Transfusion when haematocrit fell to less than/or equal to 30% versus less than/or equal to 40%

Outcomes

Primary outcomes were change in vital signs and in oxygen consumption. Secondary outcome measures were mortality, sepsis, intraventricular haemorrhage less than/or equal to grade 2 and retinopathy of prematurity

Notes

Published in abstract form where numbers were not available for inclusion in analysis. Attempts to contact authors failed to provide data consistent with original publication. We await full publication for clarification

Characteristics of ongoing studies

  • None noted.

[top]

References to studies

Included studies

Bell 2005

*Bell EF, Strauss RG, Widness JA, Mahoney LT, Mock DM, Seward VJ et al. Randomized trial of liberal versus restrictive guidelines for red blood cell transfusion in preterm infants. Pediatrics 2005;115(6):1685-91. [PubMed: 15930233]

McCoy TE, Conrad AL, Richman LC, Lindgren SD, Nopoulus PC, Bell EF. Neurocognitive profiles of preterm infants randomly assigned to lower or higher hematocrit thresholds for transfusion. Child Neuropsychology 2011;17(4):347-67.

Nopoulos PC, Conrad AL, Bell EF, Strauss RG, Widness JA, Magnotta VA et al. Long-term outcome of brain structure in premature infants: effects of liberal vs restricted red blood cell transfusions. Archives of Pediatrics and Adolescent Medicine 2011;165(5):443-50.

Blank 1984

Blank JP, Sheagren TG, Vajaria J, Mangurten HH, Benawra RS, Puppala BL. The role of RBC transfusion in the premature infant. American Journal of Diseases of Children 1984;138(9):831-3. [PubMed: 6206718]

Chen 2009

Chen HL, Tseng HI, Lu CC, Yang SN, Fan HC, Yang RC. Effect of blood transfusions on the outcome of very low body weight preterm infants under two different transfusion criteria. Pediatrics and Neonatology 2009;50(3):110-6. [PubMed: 19579757]

Connelly 1999

Unpublished data only

Connelly RJ, Stone SH, Whyte RK. Early vs. late red cell transfusion in low birth weight infants. In: Pediatric Research. Vol. 43. 1998:170A.

PINT 2006

* Kirpalani H, Whyte RK, Andersen C, Asztalos EV, Heddle N, Blajchman MA et al. The premature infants in need of transfusion (PINT) study: a randomized, controlled trial of a restrictive (low) versus liberal (high) transfusion threshold for extremely low birth weight infants. Journal of Pediatrics 2006;149(3):301-7. [PubMed: 16939737]

Whyte RK, Kirpalani H, Asztalos EV, Andersen C, Blajchman M, Heddle N et al. Neurodevelopmental outcome of extremely low birth weight infants randomly assigned to restrictive or liberal hemoglobin thresholds for blood transfusion. Pediatrics 2009;123(1):207-13. [PubMed: 19117884]

Excluded studies

Meyer 1993

Meyer J, Sive A, Jacobs P. Empiric red cell transfusion in asymptomatic preterm infants. Acta Paediatrica 1993;82(1):30-4. [PubMed: 8453217]

Ransome 1989

Ransome OJ, Moosa EA, Mothebe FM, Spector I. Are regular 'top-up' transfusions necessary in otherwise well, growing premature infants? South African Medical Journal 1989;75(4):165-6. [PubMed: 2645662]

Ross 1989

Ross MP, Christensen RD, Rothstein G, Koenig JM, Simmons MD, Noble NA et al. A randomized trial to develop criteria for administering erythrocyte transfusions to anemic preterm Infants 1 to 3 months of age. Journal of Perinatology 1989;9(3):246-53. [PubMed: 2681578]

Wardle 2002

[Other: ]

Wardle SP, Garr R, Yoxall CW, Weindling AM. A pilot randomised controlled trial of peripheral fractional oxygen extraction to guide blood transfusions in preterm infants. Archives of Disease in Childhood. Fetal and Neonatal Edition 2002;86(1):F22-7. [PubMed: 1181554]

Studies awaiting classification

Mukhopadhyay 2004

Mukhopadhyay K, Ghosh PS, Narang A, Dogra MR. Cut off level for RBC transfusion in sick preterm neonates. In: Pediatric Research. Vol. 55. 2004:289A.

Ongoing studies

  • None noted.

Other references

Additional references

Aher 2006

Aher S, Ohlsson A. Late erythropoietin for preventing red blood cell transfusion in preterm and/or low birth weight infants. Cochrane Database of Systematic Reviews 2006, Issue 3. Art. No.: CD004868. DOI: 10.1002/14651858.CD004868.pub2. [PubMed: 16856064]

Aher 2006a

Aher SM, Ohlsson A. Early versus late erythropoietin for preventing red blood cell transfusion in preterm and/or low birth weight infants. Cochrane Database of Systematic Reviews 2006, Issue 3. Art. No.: CD004865. DOI: 10.1002/14651858.CD004865.pub2. [PubMed: 16856063]

Alkalay 2003

Alkalay AL, Galvis S, Ferry DA, Simmons CF, Krueger RC Jr. Hemodynamic changes in anemic premature infants: are we allowing the hematocrits to fall too low? Pediatrics 2003;112(4):838-45. [PubMed: 14523175]

Allen 2006

Allen BW, Piantadosi CA. How do red blood cells cause hypoxic vasodilation? The SNO-hemoglobin paradigm. American Journal of Physiology. Heart and Circulatory Physiology 2006;291(4):h3507-12. [PubMed: 16751292]

Andersen 2006

Andersen CC, Collins CL. Poor circulation, early brain injury, and the potential role of red cell transfusion in premature newborns. Pediatrics 2006;117(4):1464-6. [PubMed: 16585360]

Bard 1997

Bard H, Widness JA. The life span of erythrocytes transfused to preterm infants. Pediatric Research 1997;42(1):9-11. [PubMed: 9212030]

Barkemeyer 2000

Barkemeyer BM, Hempe JM. Effect of transfusion on hemoglobin variants in preterm infants. Journal of Perinatology 2000;20(6):355-8. [PubMed: 11002873]

Bednarek 1998

Bednarek FJ, Weisberger S, Richardson DK, Frantz ID 3rd, Shah B, Rubin LP; SNAP II Study Group. Variations in blood transfusions among newborn intensive care units. The Journal of Pediatrics 1998;133(5):601-7. [PubMed: 9821414]

Beeram 2001

Beeram MR, Krauss DR, Riggs MW. Red blood cell transfusion practices in very low birth weight infants in 1990s postsurfactant era. Journal of the National Medical Association 2001;93(10):405-9. [PubMed: 11688921]

Bell 1978

Bell MJ, Ternberg JL, Feigin RD, Keating JP, Marshall R, Barton L et al. Neonatal necrotizing enterocolitis. Therapeutic decisions based upon clinical staging. Annals of Surgery 1978;187(1):1-7. [PubMed: 413500]

Bell 2006

Bell EF. Transfusion thresholds for preterm infants: how low should we go? The Journal of Pediatrics 2006;149(3):287-9. [PubMed: 16939732]

Bell 2008

Bell EF. When to transfuse preterm babies. Archives of Disease in Childhood. Fetal and Neonatal Edition 2008;93(6):F469-73. [PubMed: 18653585]

Bennett-Guerrero 2007

Bennett-Guerrero E, Veldman TH, Doctor A, Telen MJ, Ortel TL, Reid TS et al. Evolution of adverse changes in stored RBCs. Proceedings of the National Academy of Sciences of the United States of America 2007;104(43):17063-8. [PubMed: 17940021]

Bifano 1992

Bifano EM, Smith F, Borer J. Relationship between determinants of oxygen delivery and respiratory abnormalities in preterm infants with anemia. The Journal of Pediatrics 1992;120(2 Pt 1):292-6. [PubMed: 1735832]

Blajchman 2006

Blajchman MA, Vamvakas EC. The continuing risk of transfusion-transmitted infections. The New England Journal of Medicine 2006;355(13):1303-5. [PubMed: 17005947]

Blanchette 1984

Blanchette VS, Zipursky A. Assessment of anemia in newborn infants. Clinics in Perinatology 1984;11(2):489-510. [PubMed: 6378489]

Cambonie 2007

Cambonie G, Matecki S, Milesi C, Voisin M, Guillaumont S, Picaud JC. Myocardial adaptation to anemia and red blood cell transfusion in premature infants requiring ventilation support in the 1st postnatal week. Neonatology 2007;92(3):174-81. [PubMed: 17429222]

Christensen 2008

Christensen RD, Jopling J, Henry E, Wiedmeier SE. The erythrocyte indices of neonates, defined using data from over 12, 000 patients in a multihospital health care system. Journal of Perinatology 2008;28(1):24-8. [PubMed: 17972890]

Crowley 2010

Crowley M, Kirpalani H. A rational approach to red blood cell transfusion in the neonatal ICU. Current Opinion in Pediatrics 2010;22(2):151-7.

Dani 2004

Dani C, Martelli E, Bertini G, Pezzati M, Rossetti M, Buonocore G et al. Effect of blood transfusions on oxidative stress in preterm infants. Archives of Disease in Childhood. Fetal and Neonatal Edition 2004;89(5):F408-11. [PubMed: 15321958]

Deeks 2011

Deeks JJ, Higgins JPT, Altman DG (editors). Chapter 9: Analysing data and undertaking meta-analyses. Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March 2011]. The Cochrane Collaboration, 2011. Available from www.cochrane-handbook.org. Higgins JPT, Green S (editors).

Delivoria-Papadopoulos 1971

Delivoria-Papadopoulos M, Morrow G 3rd, Oski FA. Exchange transfusion in the newborn infant with fresh and "old" blood: the role of storage on 2, 3-diphosphoglycerate, hemoglobin-oxygen affinity, and oxygen release. The Journal of Pediatrics 1971;79(6):898-903. [PubMed: 5125406]

Fetus and Newborn 1992

Fetus and Newborn Committee, Canadian Paediatric Society. Guidelines for transfusion of erythrocytes to neonates and premature infants. Canadian Medical Association Journal 1992;147(12):1781-92. [PubMed: 1458420]

Fetus and Newborn 2002

Fetus and Newborn Committee, Canadian Paediatric Society. Red blood cell transfusions in newborn infants: Revised guidelines. Paediatrics & Child Health 2002;7(8):553-66. [PubMed: 20046468]

Frey 2001

Frey B, Losa M. The value of capillary whole blood lactate for blood transfusion requirements in anaemia of prematurity. Intensive Care Medicine 2001;27(1):222-7. [PubMed: 11280639]

Hirano 2001

Hirano K, Morinobu T, Kim H, Hiroi M, Ban R, Ogawa S et al. Blood transfusion increases radical promoting non-transferrin bound iron in preterm infants. Archives of Disease in Childhood. Fetal and Neonatal Edition 2001;84(3):F188-93. [PubMed: 11320046]

Holman 1995

Holman P, Blajchman MA, Heddle N. Noninfectious adverse effects of blood transfusion in the neonate. Transfusion Medicine Reviews 1995;9(3):277-87. [PubMed: 7549238]

Hosono 2008

Hosono S, Mugishima H, Fujita H, Hosono A, Minato M, Okada T et al. Umbilical cord milking reduces the need for red cell transfusions and improves neonatal adaptation in infants born at less than 29 weeks' gestation: a randomised controlled trial. Archives of Disease in Childhood. Fetal and Neonatal Edition 2008;93(1):F14-9. [PubMed: 17234653]

Hume 1997

Hume H. Red blood cell transfusions for preterm infants: the role of evidence-based medicine. Seminars in Perinatology 1997;21(1):8-19. [PubMed: 9190029]

Hébert 1999

Hébert PC, Wells G, Blajchman MA, Marshall J, Martin C, Pagliarello G et al. A multicentre, randomized, controlled clinical trial of transfusion requirements in critical care. New England Journal of Medicine 1999;340(6):409-17. [PubMed: 9971864]

ICCRP 2005

International Committee for the Classification of Retinopathy of Prematurity (ICCRP). The International Classification of Retinopathy of Prematurity revisited. Archives of Ophthalmology 2005;123(7):991-9. [PubMed: 16009843]

Joshi 1987

Joshi A, Gerhardt T, Shandloff P, Bancalari E. Blood transfusion effect on the respiratory pattern of preterm infants. Pediatrics 1987;80(1):79-84. [PubMed: 3601522]

Keyes 1989

Keyes WG, Donohue PK, Spivak JL, Jones MD Jr, Oski FA. Assessing the need for transfusion of premature infants and role of hematocrit, clinical signs, and erythropoietin level. Pediatrics 1989;84(3):412-7. [PubMed: 2771544]

Lachance 1994

Lachance C, Chessex P, Fouron JC, Widness JA, Bard H. Myocardial, erythropoietic, and metabolic adaptations to anemia of prematurity. The Journal of Pediatrics 1994;125(2):278-82. [PubMed: 8040778]

Lacroix 2007

Lacroix J, Hébert PC, Hutchison JS, Hume HA, Tucci M, Ducruet T et al. Transfusion strategies for patients in pediatric intensive care units. The New England Journal of Medicine 2007;356(16):1609-19. [PubMed: 17442904]

Madan 2005

Madan A, Kumar R, Adams MM, Benitz WE, Geaghan SM, Widness JA. Reduction in red blood cell transfusions using a bedside analyzer in extremely low birth weight infants. Journal of Perinatology 2005;25(1):21-5. [PubMed: 15496875]

Mally 2006

Mally P, Golombek SG, Mishra R, Nigam S, Mohandas K, Depalhma H et al. Association of necrotizing enterocolitis with elective packed red blood cell transfusions in stable, growing, premature neonates. American Journal of Perinatology 2006;23(8):451-8. [PubMed: 17009195]

McFaul 2009

McFaul SJ, Corley JB, Mester CW, Nath J. Packed blood cells stored in AS-5 become proinflammatory during storage. Transfusion 2009;49(7):1451-60. [PubMed: 19374730]

Meldon 1985

Meldon JH. Blood gas transport and 2, 3-DPG. Advances in Experimental Medicine and Biology 1985;191:63-73. [PubMed: 3938608]

Mimica 2008

Mimica AF, dos Santos AM, da Cunha DH, Guinsburg R, Bordin JO, Chiba A et al. A very strict guideline reduces the number of erythrocyte transfusions in preterm infants. Vox Sanguinis 2008;95(2):106-11. [PubMed: 18544122]

Miyashiro 2005

Miyashiro AM, Santos N, Guinsburg R, Kopelman BI, Peres Cde A, Taga MF et al. Strict red blood cell transfusion guideline reduces the need for transfusions in very-low-birthweight infants in the first 4 weeks of life: a multicentre trial. Vox Sanguinis 2005;88(2):107-13. [PubMed: 15720608]

Ohls 2001

Ohls RK, Ehrenkranz RA, Wright LL, Lemons JA, Korones SB, Stoll BJ et al. Effects of early erythropoietin therapy on the transfusion requirements of preterm infants below 1250 grams birth weight: a multicenter, randomized, controlled trial. Pediatrics 2001;108(4):934-42. [PubMed: 11581447]

Ohlsson 2006

Ohlsson A, Aher SM. Early erythropoietin for preventing red blood cell transfusion in preterm and/or low birth weight infants. Cochrane Database of Systematic Reviews 2006, Issue 3. Art. No.: CD004863. DOI: 10.1002/14651858.CD004863.pub2. [Other: ; PubMed: 16856062]

Papile 1983

Papile LA, Munsick-Bruno G, Schaefer A. Relationship of cerebral intraventricular hemorrhage and early childhood neurologic handicaps. The Journal of Pediatrics 1983;103(2):273-7. [PubMed: 6875724]

Pinto-Martin 1995

Pinto-Martin JA, Riolo S, Cnaan A, Holzman C, Susser MW, Paneth N. Cranial ultrasound prediction of disabling and nondisabling cerebral palsy at age two in a low birth weight population. Pediatrics 1995;95(2):249-54. [PubMed: 7838643]

Rabe 2004

Rabe H, Reynolds G, Diaz-Rossello J. Early versus delayed umbilical cord clamping in preterm infants. Cochrane Database of Systematic Reviews 2004, Issue 4. Art. No.: CD003248. DOI: 10.1002/14651858.CD003248.pub2. [PubMed: 15495045]

RevMan 2011

Review Manager (RevMan) [Computer program]. Version 5.1. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2011.

Ringer 1998

Ringer SA, Richardson DK, Sacher RA, Keszler M, Churchill WH. Variations in transfusion practice in neonatal intensive care. Pediatrics 1998;101(2):194-200. [PubMed: 9445491]

Sacks 1984

Sacks LM, Delivoria-Papadopoulos M. Hemoglobin-oxygen interactions. Seminars in Perinatology 1984;8(3):168-83. [PubMed: 6204387]

Sasidharan 1992

Sasidharan P, Heimler R. Transfusion-induced changes in the breathing pattern of healthy preterm anemic infants. Pediatric Pulmonology 1992;12(3):170-3. [PubMed: 1641274]

Shannon 1995

Shannon KM, Keith JF 3rd, Mentzer WC, Ehrenkranz RA, Brown MS, Widness JA et al. Recombinant human erythropoietin stimulates erythropoiesis and reduces erythrocyte transfusions in very low birth weight preterm infants. Pediatrics 1995;95(1):1-8. [PubMed: 7770284]

Shennan 1988

Shennan AT, Dunn MS, Ohlsson A, Lennox K, Hoskins EM. Abnormal pulmonary outcomes in premature infants: prediction from oxygen requirement in the neonatal period. Pediatrics 1988;82(4):527-32. [PubMed: 3174313]

Siggaard-Andersen 1990

Siggaard-Andersen O, Gothgen IH, Wimberley PD, Fogh-Andersen N. The oxygen status of the arterial blood revised: relevant oxygen parameters for monitoring the arterial oxygen availability. Scandinavian Journal of Clinical and Laboratory Investigation. Supplementum 1990;203:17-28. [PubMed: 2128557]

Silvers 1998

Silvers KM, Gibson AT, Russell JM, Powers HJ. Antioxidant activity, packed cell transfusions, and outcome in premature infants. Archives of Disease in Childhood. Fetal and Neonatal Edition 1998;78(3):F214-9. [PubMed: 9713035]

Simak 2006

Simak J, Gelderman MP. Cell membrane microparticles in blood and blood products: potentially pathogenic agents and diagnostic markers. Transfusion Medicine Reviews 2006;20(1):1-26. [PubMed: 16373184]

Stockman 1984

Stockman JA 3rd, Graeber JE, Clark DA, McClellan K, Garcia JF, Kavey RE. Anemia of prematurity: determinants of the erythropoietin response. The Journal of Pediatrics 1984;105(5):786-92. [PubMed: 6502312]

Stockman 1984a

Stockman JA 3rd, Clark DA. Weight gain: a response to transfusion in selected preterm infants. American Journal of Diseases of Children 1984;138(9):828-30. [PubMed: 6475871]

Stockman 1986

Stockman JA 3rd. Anemia of prematurity. Current concepts in the issue of when to transfuse. Pediatric Clinics of North America 1986;33(1):111-28. [PubMed: 3513096]

Stramer 2009

Stramer SL, Hollinger FB, Katz LM, Kleinman S, Metzel PS, Gregory KR et al. Emerging infectious disease agents and their potential threat to transfusion safety. Transfusion 2009;49(Suppl 2):1S-29S. [PubMed: 19686562]

Strauss 2004

Strauss RG, Mock DM, Widness JA, Johnson K, Cress G, Schmidt RL. Post-transfusion 24-hour recovery and subsequent survival of allogeneic red blood cells in the bloodstream of newborn infants. Transfusion 2004;44(6):871-6. [PubMed: 15157254]

Vamvakas 2007

Vamvakas EC, Blajchman MA. Transfusion-related immunomodulation (TRIM): an update. Blood Reviews 2007;21(6):327-48. [PubMed: 17804128]

Venancio 2007

Venancio JP, Santos AM, Guinsburg R, Peres Cde A, Shinzato AR, Lora MI. Strict guideline reduces the need for RBC transfusions in premature infants. Journal of Tropical Pediatrics 2007;53(2):78-82. [PubMed: 17166932]

Wardle 2001

Wardle SP, Weindling AM. Peripheral fractional oxygen extraction and other measures of tissue oxygenation to guide blood transfusions in preterm infants. Seminars in Perinatology 2001;25(2):60-4. [PubMed: 11339666]

Wardrop 1978

Wardrop CA, Holland BM, Veale KE, Jones JG, Gray OP. Nonphysiological anaemia of prematurity. Archives of Disease in Childhood 1978;53(11):855-60. [PubMed: 727808]

Widness 1996

Widness JA, Seward VJ, Kromer IJ, Burmeister LF, Bell EF, Strauss RG. Changing patterns of red blood cell transfusion in very low birth weight infants. Journal of Pediatrics 1996;129(5):680-7. [PubMed: 8917234 ]

Zimrin 2009

Zimrin AB, Hess JR. Current issues relating to the transfusion of stored red blood cells. Vox Sanguinis 2009;96(2):93-103. [PubMed: 19152602]

Other published versions of this review

  • None noted.

Classification pending references

  • None noted.

[top]

Data and analyses

1 Transfusion at a restrictive vs a liberal haemoglobin threshold

For graphical representations of the data/results in this table, please use link under "Outcome or Subgroup".

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate
1.1 Infants transfused once or more 4 614 Risk Ratio (M-H, Random, 95% CI) 0.95 [0.91, 1.00]
1.2 Transfusions per infant given from study start to study end 4 614 Mean Difference (IV, Random, 95% CI) -1.12 [-1.75, -0.49]
1.3 Donor exposures per infant from blood products 2 Mean Difference (IV, Random, 95% CI) Subtotals only
1.3.1 Donor exposures from red cell transfusions 2 554 Mean Difference (IV, Random, 95% CI) -0.54 [-0.93, -0.15]
1.3.2 Donor exposures from all sources 1 451 Mean Difference (IV, Random, 95% CI) -0.50 [-1.65, 0.65]
1.4 Age at first transfusion (of those transfused) 4 Other data No numeric data
1.5 Haemoglobin levels in survivors [g/l] 3 Mean Difference (IV, Random, 95% CI [g/l]) No totals
1.5.1 Haemoglobin level at 4 weeks of age 2 Mean Difference (IV, Random, 95% CI [g/l]) No totals
1.5.2 Haemoglobin at discharge 2 Mean Difference (IV, Random, 95% CI [g/l]) No totals
1.6 Death 4 Risk Ratio (M-H, Random, 95% CI) Subtotals only
1.6.1 Prior to first hospital discharge 4 614 Risk Ratio (M-H, Random, 95% CI) 1.23 [0.86, 1.76]
1.6.2 By 18-21 months follow-up 1 421 Risk Ratio (M-H, Random, 95% CI) 1.09 [0.76, 1.56]
1.7 Death or severe morbidity 3 Risk Ratio (M-H, Random, 95% CI) Subtotals only
1.7.1 By first hospital discharge 3 511 Risk Ratio (M-H, Random, 95% CI) 1.07 [0.96, 1.20]
1.7.2 At 18-21 months follow-up with MDI < 70 1 421 Risk Ratio (M-H, Random, 95% CI) 1.17 [0.94, 1.47]
1.7.3 At 18-21 months follow-up with MDI < 85 1 421 Risk Ratio (M-H, Random, 95% CI) 1.21 [1.01, 1.44]
1.8 Death or severe brain injury by first hospital discharge 4 614 Risk Ratio (M-H, Random, 95% CI) 1.12 [0.81, 1.55]
1.9 Retinopathy of prematurity in survivors 4 Risk Ratio (M-H, Random, 95% CI) Subtotals only
1.9.1 Retinopathy of prematurity in survivors, all cases 4 517 Risk Ratio (M-H, Random, 95% CI) 0.98 [0.84, 1.14]
1.9.2 Retinopathy of prematurity in survivors, grade 1 or 2 4 517 Risk Ratio (M-H, Random, 95% CI) 0.96 [0.78, 1.18]
1.9.3 Retinopathy of prematurity in survivors, > stage 3 4 517 Risk Ratio (M-H, Random, 95% CI) 1.04 [0.68, 1.58]
1.10 Brain Injury on ultrasound in survivors 4 517 Risk Ratio (M-H, Random, 95% CI) 1.07 [0.50, 2.27]
1.11 Bronchopulmonary dysplasia in survivors 4 Risk Ratio (M-H, Random, 95% CI) Subtotals only
1.11.1 Oxygen requirement at 28 days in survivors 4 544 Risk Ratio (M-H, Random, 95% CI) 0.99 [0.92, 1.06]
1.11.2 Oxygen requirement at 36 weeks gestation in survivors 4 524 Risk Ratio (M-H, Random, 95% CI) 1.03 [0.87, 1.21]
1.12 Neurosensory impairment at 18-21 months follow-up among survivors 1 Risk Ratio (M-H, Random, 95% CI) No totals
1.12.1 Cognitive delay MDI < 70 1 Risk Ratio (M-H, Random, 95% CI) No totals
1.12.2 Cognitive delay MDI < 85 1 Risk Ratio (M-H, Random, 95% CI) No totals
1.12.3 Cerebral palsy 1 Risk Ratio (M-H, Random, 95% CI) No totals
1.12.4 Severe visual impairment 1 Risk Ratio (M-H, Random, 95% CI) No totals
1.12.5 Severe hearing Impairment 1 Risk Ratio (M-H, Random, 95% CI) No totals
1.12.6 Any neurosensory impairment 1 Risk Ratio (M-H, Random, 95% CI) No totals
1.13 Apnoea requiring intervention in survivors 4 517 Risk Ratio (M-H, Random, 95% CI) 1.01 [0.95, 1.08]
1.14 Measures of weight gain in survivors [g] 2 Mean Difference (IV, Fixed, 95% CI [g]) No totals
1.14.1 Weight gain at 32 weeks adjusted gestation age 1 Mean Difference (IV, Fixed, 95% CI [g]) No totals
1.14.2 Weight gain at 35 days of age 1 Mean Difference (IV, Fixed, 95% CI [g]) No totals
1.15 Rate of weight gain from birth to discharge [g/d] 1 Mean Difference (IV, Random, 95% CI [g/d]) No totals
1.16 Time to double birth weight in survivors 2 Other data No numeric data
1.17 Length of hospitalisation 3 Other data No numeric data
1.18 Patent ductus arteriosus 3 Risk Ratio (M-H, Random, 95% CI) Subtotals only
1.18.1 Patent ductus arteriosus 3 590 Risk Ratio (M-H, Random, 95% CI) 0.93 [0.76, 1.14]
1.18.2 Patent ductus arteriosus requiring surgery 2 554 Risk Ratio (M-H, Random, 95% CI) 1.31 [0.89, 1.91]
1.19 Necrotising enterocolitis 3 590 Risk Ratio (M-H, Random, 95% CI) 1.62 [0.83, 3.13]

2 Transfusion for clinical signs only vs for transfusion at haemoglobin threshold

For graphical representations of the data/results in this table, please use link under "Outcome or Subgroup".

Outcome or Subgroup Studies Participants Statistical Method Effect Estimate
2.1 Discharge haemoglobin [g/l] 1 Mean Difference (IV, Fixed, 95% CI [g/l]) No totals
2.2 Death prior to discharge 1 Risk Ratio (M-H, Fixed, 95% CI) No totals
2.3 Infants with apnoea 1 Risk Ratio (M-H, Fixed, 95% CI) No totals
2.4 Days to regain birth weight [days] 1 Mean Difference (IV, Fixed, 95% CI [days]) No totals
2.5 Length of hospitalisation [days] 1 Mean Difference (IV, Fixed, 95% CI [days]) No totals
2.6 Costs of hospitalisation [$US] 1 Mean Difference (IV, Fixed, 95% CI [$US]) No totals

[top]

Sources of support

Internal sources

  • No sources of support provided.

External sources

  • No sources of support provided.

This review is published as a Cochrane review in The Cochrane Library, Issue 11, 2011 (see http://www.thecochranelibrary.com External Web Site Policy for information). Cochrane reviews are regularly updated as new evidence emerges and in response to feedback. The Cochrane Library should be consulted for the most recent version of the review.