Transfusion of red blood cells stored for shorter versus longer duration for all conditions

Summary of findings for the main comparison. Transfusion of RBCs of shorter vs longer storage duration

Transfusion of RBCs of shorter vs longer storage duration for all conditions
Patient or population: all conditions Setting: hospital Intervention: transfusion of RBCs of shorter storage duration Comparison: transfusion of RBCs of longer storage duration
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with transfusion of RBCs of longer storage duration	Risk with transfusion of RBCs of shorter storage duration
Mortality: in‐hospital mortality (within 24 hours)	Study population ‐ paediatric participants		RR 1.71 (0.41 to 7.10)	290 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
	21 per 1000	35 per 1000 (8 to 147)
Mortality: in‐hospital mortality (within 7 days)	Study population ‐ paediatric participants		RR 3.00 (0.13 to 71.34)	74 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 2 3}
	0 per 1000	0 per 1000 (0 to 0)
Mortality: in‐hospital mortality (within 7 days)	Study population ‐ adults		RR 1.42 (0.66 to 3.06)	1098 (1 RCT)	⊕⊕⊕⊝ MODERATE ⁴
	20 per 1000	28 per 1000 (13 to 60)
Mortality: time‐point not defined	Study population ‐ adults		RR 2.25 (0.55 to 9.17)	17 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{5 6 7}
	222 per 1000	500 per 1000 (122 to 1000)
Mortality: short term (up to 30 days)	Study population ‐ paediatric participants		RR 1.44 (0.46 to 4.48)	290 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
	34 per 1000	50 per 1000 (16 to 154)
Mortality: short term (up to 30 days)	Study population ‐ adults		RR 0.85 (0.50 to 1.45)	1121 (2 RCTs)	⊕⊕⊕⊝ MODERATE ⁸
	51 per 1000	43 per 1000 (25 to 74)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: we are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low certainty: we have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect
¹ Downgraded 1 level for indirectness, as the trial was not designed to assess the impact of transfusion on mortality, rather it was designed to assess if storage duration influenced the efficacy of RBCs in the particular participant population. ² Downgraded 1 level for imprecision, due to very small number of events. ³ Downgraded 1 level, as no information was reported to permit assessment of biases due to selection, performance and detection. However there were no concerns with regard to attrition or reporting bias. ⁴ Downgraded 1 level for indirectness, as the trial was not designed to assess the impact of transfusion on mortality, but was designed to assess the effect of storage duration on organ dysfunction in elective cardiac surgery patients. ⁵ Downgraded 1 level for imprecision, due to small sample size of 17 participants. ⁶ Downgraded 2 levels, as no information was reported to permit assessment of the risk of selection, performance, detection, attrition and reporting biases. The study was reported as part of a letter. ⁷ Downgraded 1 level for indirectness, as the trial was not designed to assess the impact of transfusion on mortality, but was designed to assess the feasibility of running a larger trial and whether the blood bank in the investigating hospital could provide an adequate inventory of RBCs. ⁸ Downgraded 1 level for indirectness, as the trials were not designed to assess the impact of transfusion on mortality, rather Steiner 2015 was designed to assess the effect of storage duration on organ dysfunction in elective cardiac surgery patients and Bennett‐Guerrero 2009 was a feasibility trial designed to assess recruitment and randomisation rates and to see if adequate separation of duration of storage between the two intervention arms was feasible (Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]).

Summary of findings 2. Transfusion of RBCs of shorter vs standard practice storage duration

Transfusion of RBCs of shorter vs standard practice storage duration for all conditions
Patient or population: all conditions Setting: hospital Intervention: transfusion of RBCs of shorter storage duration Comparison: transfusion of RBCs of standard practice storage duration
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with transfusion of RBCs of standard practice storage duration	Risk with transfusion of RBCs of shorter storage duration
Mortality: in‐hospital mortality (time points varied)	Study population ‐ adults		RR 1.05 (0.97 to 1.14)	25754 (4 RCTs)	⊕⊕⊕⊝ MODERATE ¹
	88 per 1000	92 per 1000 (85 to 100)
Mortality: in‐ICU mortality	Study population ‐ adults		RR 1.06 (0.98 to 1.15)	13066 (3 RCTs)	⊕⊕⊕⊝ MODERATE ²
	145 per 1000	153 per 1000 (142 to 167)
Mortality: short‐term mortality (up to 30 days) ‐ neonates	Study population ‐ neonates		RR 0.30 (0.01 to 7.02)	40 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{3 4 5}
	53 per 1000	16 per 1000 (1 to 369)
Mortality: short‐term mortality (up to 30 days) ‐ adults	Study population ‐ adults		RR 1.04 (0.96 to 1.13)	7510 (4 RCTs)	⊕⊕⊕⊝ MODERATE ⁶
	223 per 1000	232 per 1000 (214 to 252)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: we are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low certainty: we have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect
¹ Downgraded 1 level for high risk of bias in 2 trials; selection bias in Heddle 2012 and attrition bias in Hebert 2005. ² Downgraded 1 level for high risk of bias in 1 trial: attrition bias in Hebert 2005. ³ Downgraded 1 level for risk of bias, as information was not reported that would permit assessment of selection, detection and reporting biases. ⁴ Downgraded 1 level for indirectness, as the study was designed to measure safety of the duration of storage of RBCs. ⁵ Downgraded 1 level for imprecision due to small number of events. ⁶ Downgraded 1 level for high risk of bias in 2 trials; selection bias in Kor 2012 and attrition bias in Hebert 2005.

Background

available in

Description of the condition

The haemoglobin (Hb) contained within red blood cells (RBCs) is essential for oxygen transportation. Anaemia is defined by the World Health Organization (WHO) as a Hb concentration lower than 13 g/dL in men and lower than 12 g/dL in non‐pregnant women, and describes a clinical state in which oxygen transport is disturbed and tissue hypoxia may occur (Beutler 2006). Anaemia has no single cause; rather it is the consequence of a variety of factors. In high‐income countries, the overall prevalence of anaemia is estimated to be 10% (McLean 2009), however, this figure varies significantly with demographic profiles and patterns of co‐morbid diagnoses (McLean 2009; Tettamanti 2010). Children and older adults are most commonly affected by anaemia, for example, almost 90% of preterm infants with birth weights under 1.0 kg are anaemic (Martin 2010). In later life, rates rise again, largely because of the increasing incidence of co‐morbid diagnoses.

The aetiology of anaemia can be broadly divided into disease processes that impair RBC production and those in which the life span or distribution of RBCs is altered. In the former group, disorders such as acquired and iatrogenic marrow dysfunction, nutritional deficiency, and cytokine‐driven processes such as anaemia due to chronic disease are commonplace. In the latter group, examples include disease processes such as pathological bleeding and immune haemolysis.

When possible, reversing the primary cause of anaemia remains the treatment of choice, however this cannot always be achieved. Furthermore, when severe anaemia results in life‐threatening organ dysfunction, rapid correction is required. In these instances, RBC transfusion is the only viable treatment capable of restoring tissue oxygenation.

Description of the intervention

Red blood cell transfusion

Red blood cell transfusion has been a commonplace treatment for anaemia since the 1990s (Alter 2008), and is very widely practiced. Annually, in the UK, around 1.7 million RBC units are issued for transfusion (NICE 2015). This equates to transfusion of approximately 36 units per 1000 population per year, which is a figure not dissimilar to other high‐income countries (Cobain 2007; NICE 2015). It is surprising, for such a ubiquitous intervention, that a recent systematic overview concluded that rigorous clinical trial data were lacking to support the benefits of many current transfusion practices (Wilkinson 2011). Indeed, evidence obtained from randomised controlled trials (RCTs) indicates little or no benefit from RBC transfusion at higher recipient haemoglobin concentration thresholds (commonly termed 'liberal' policies for RBC transfusion) (Carson 2012; Carson 2016).

Red blood cell transfusions are also associated with some well‐described risks (Stainsby 2006); these are biological products and hazards such as bacterial and viral contamination and allergic reactions. The Serious Hazards of Transfusion (SHOT) scheme estimated that, in 2014, the risk of transfusion‐related morbidity was 63.5 per million blood components issued (SHOT 2015). Key amongst these risks is the potential long‐debated risk posed by RBCs with a prolonged storage duration (see below) (Schrier 1979). Therefore, practice guidelines now promote more restrictive policies for RBC transfusion in many clinical settings (Carson 2013; Carson 2016). Despite this, RBC transfusion remains a very common intervention; for example, while up to 60% of patients admitted to critical care units develop anaemia (Vincent 2002; Corwin 2004), only 10% to 15% have a history of chronic anaemia before admission to the intensive care unit (ICU). Unless modified by RBC transfusion, haemoglobin values typically decrease by about 0.5 g/dL/day during critical illness for reasons that include anaemia of inflammation associated with acute illness, haemodilution, co‐morbidities, bleeding and phlebotomy (Walsh 2010). As a result, 20% to 50% of critically ill patients receive a RBC transfusion, especially those with multiple organ failure. About 8% to 10% of the UK blood supply is transfused to patients in ICUs.

Red blood cell units and their storage

The uncertainty regarding the clinical consequences of transfusing RBC units that have been stored for longer periods before transfusion is a major concern. The debate about potential harm related to transfusion of a product that has been stored for a longer period was re‐ignited by the authors of the Transfusion Requirements In Critical Care (TRICC) trial (Hebert 1999). This landmark RCT compared liberal and restrictive transfusion practices in critically ill patients ‐ not length of product storage ‐ and investigators showed that restricting transfusions to maintain Hb concentration at 7 g/dL to 9 g/dL was safe, and superior to more liberal RBC use in some subgroups of patients. Crucially, the study authors suggested that the common practice of storing RBC units for prolonged periods might contribute to the unexpected adverse effects of liberal transfusion.

This suggestion is biologically plausible in view of the body of evidence that has demonstrated changes in many cellular and physiological properties of RBCs. These in vitro changes, which occur during RBC storage, are commonly known as the 'storage lesion' (D'Alessandro 2010; Glynn 2010). The storage lesion includes biochemical, metabolic and mechanical changes to RBCs, all of which may impair oxygen delivery. The term also encompasses changes that occur in the storage medium, which theoretically could mediate inflammatory or oxidative tissue damage (Sharifi 2000; Kucukakin 2011). The most commonly described biochemical and metabolic components of the storage lesion are impaired nitric oxide metabolism (Stapley 2012), depletion of cellular 2,3‐diphosphoglycerate (Vora 1989), and dysfunction of the membrane sodium‐potassium pump (D'Alessandro 2010). Nitric oxide depletion induces vasoconstriction, which, in turn impairs blood flow and oxygenation (Stapley 2012); depleted 2,3‐diphosphoglycerate reduces the oxygen affinity of haemoglobin (Sohmer 1979); while dysfunction of the membrane sodium‐potassium pump results in harmful potassium leakage from the RBCs into extracellular fluids (Hess 2010). Mechanical changes to the red cell membrane impair fluidity and red cell flow (Hess 2010), may reduce transit of RBCs through the microscopic vasculature of organs such as lungs and kidneys (Roback 2011a), and may impair oxygen uptake and delivery. Changes caused by the storage medium include the generation of inflammatory mediators such as the soluble CD40 ligand, interleukin‐6 (IL‐6) and interleukin‐8 (IL‐8) (Khan 2006; Kucukakin 2011). Potential oxidative damage may also arise from superoxide generation in the storage media (Kucukakin 2011).

Extended RBC storage, as described above, is fundamental to effective management of blood stocks. In the UK, due to stock rotation processes, the average duration of storage of a RBC unit at the time of transfusion is 18 to 21 days (NHSBT 2012). This is very similar to the situation throughout Europe and North America (Bennett‐Guerrero 2009; Heddle 2012). Changes in the storage lesion, as described above, may be well established by this time. Currently, it is biologically plausible that critically ill patients may be being denied the benefits of RBCs that have been stored for a shorter duration, and are being exposed to the additional clinical risks posed by RBC units that have been stored for a longer duration. Limited clinical data support this notion. Cohort studies have described associations between RBC storage duration and a wide range of clinically important adverse outcomes (including infection, organ failure, increased hospital stay and death) (Vamvakas 1999; Mynster 2000; Leal‐Noval 2003; Basran 2006; Koch 2008). However, these effects are not universally described (Vamvakas 2000; van de Watering 2006); although this important message is provided in the literature, many study authors point to the presence of significant confounding factors in the evidence (Steiner 2009). In particular, the strong linkage between the total volume transfused (which itself is strongly associated with the presence of co‐morbidities, severity of illness and worse prognosis) and the average duration of storage of RBC units issued makes inferring causality very difficult (Vamvakas 2010).

How the intervention might work

The goal of RBC transfusion is to improve tissue oxygenation through increasing the red cell mass. It is commonly presumed that lower Hb concentrations represent an accurate measure of diminished oxygen‐carrying capacity, which can be corrected in part by RBC transfusion. The processing methods for RBC collection and storage for transfusion have been studied for many years (Alter 2008). After blood collection, whole blood is centrifuged, plasma is depleted and RBCs are resuspended in an optimal additive solution ‐ a solution of additives that is designed to optimise and maintain the integrity of the RBCs ‐ for storage within specially designed bags. This process, in conjunction with effective refrigeration, has allowed the duration storage of RBC units to be significantly extended (D'Alessandro 2010). Many countries store RBC units routinely for up to 42 days. This period is defined by the arbitrary requirement that, after storage, more than 75% of RBCs should survive in the recipient's circulation at 24 hours (Roback 2011b). An extended shelf‐life facilitates stock management and is fundamental to effective blood banking. Amongst blood providers and blood banks, it is standard practice to issue the oldest stock first in preference to newer stock (Stanger 2012). This ensures that the demand for blood can be met and minimises wastage of a precious and financially costly resource (Stanger 2012).

Why it is important to do this review

If studies were to indicate that clinical outcomes are affected by storage duration, the implications for inventory management and clinical practice would be significant. Clinicians would expect to use RBC units stored for a shorter duration, as they would be safer and more efficacious than RBC units stored for longer periods. Implementation of such a strategy would place considerable strain on blood providers and blood banks. It might also result in increased wastage, higher financial costs and could threaten blood supplies (Glynn 2010). There is an urgent need to reconcile this issue ‐ keenly felt by clinicians, blood services and policy makers alike ‐ because of the potential harms to patients and massive logistical implications.

Although numerous published reviews have addressed this question (including Lelubre 2009; Zimrin 2009; Vamvakas 2010; Vamvakas 2011; Wang 2012; Lelubre 2013; Alexander 2016), some have included observational study data, or have not included recently published and ongoing trials in this area, or have only focused on one outcome (e.g. mortality in Chai‐Adisaksopha 2017). Since the previous iteration of this review (Brunskill 2015), several trials previously noted as ongoing have been completed and published, specifically: ABLE (Lacroix 2015), RECESS (Steiner 2015), TOTAL (Dhabangi 2015), INFORM (Heddle 2016), and TRANSFUSE (Cooper 2017). Incorporation of these within the review will increase the number of participants available for analysis significantly. Therefore, there is a need to update this Cochrane Review so that new guidelines and policies will be based on the most recent evidence pertaining to the effects of duration of storage on RBCs.

Objectives

available in

To assess the effects of using red blood cells (RBCs) stored for a shorter versus a longer duration, or versus RBCs stored for standard practice duration, in people requiring a RBC transfusion.

Methods

available in

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs) comparing transfusion of RBCs of shorter storage duration with transfusion of RBCs of longer storage duration.

Types of participants

People of any age (neonates, children and adults) requiring RBC transfusion for investigator‐diagnosed and ‐defined anaemia of any aetiology.

Types of interventions

No consensus has been reached about what defines shorter storage duration or longer storage duration RBC units. Arbitrarily defining shorter and longer storage durations for the purposes of this review is scientifically unsound, invites legitimate criticism of the clinical validity of the review's results and risks exclusion of a large proportion of available data. Consequently, we included all definitions of shorter and longer storage durations for RBCs. Studies comparing the following interventions were eligible for inclusion.

Transfusion of RBCs of shorter storage duration versus longer storage duration.
Transfusion of RBCs of shorter storage duration versus standard practice storage duration. Here, the duration of storage of the RBCs stored for longer was dictated by standard inventory management practice of each study site.

Types of outcome measures

Primary outcomes

Mortality measured at two time points: in hospital (with time points as defined by the participant group) and short term (up to 30 days).

Secondary outcomes

Long‐term mortality (more than 30 days)
Incidence of hospital‐acquired infection (as defined by the study authors)
Duration of organ support: respiratory (invasive and non‐invasive ventilation), haemodynamic (inotropic) and renal (haemofiltration)
Length of hospital and ICU stay
Adverse transfusion reactions
Economic or blood stock inventory outcomes (as reported by the study authors)

We were interested in data addressing any of the above outcomes, at any reported time points.

Search methods for identification of studies

In order to reduce publication and retrieval bias we did not restrict our search by language, date or publication status.

Electronic searches

In this 2018 update we searched the following electronic databases and ongoing trials databases on 20 November 2017:

Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2017, Issue 10);
MEDLINE (OvidSP, 1948 to 20 November 2017);
Embase (OvidSP, 1974 to 20 November 2017);
CINAHL (EBSCOHost, 1982 to 20 November 2017);
PubMed (epublications ahead of print only, 20 November 2017);
LILACS (Latin American Caribbean Health Sciences Literature) (Bireme, 1982 to 20 November 2017);
Transfusion Evidence Library (1980 to 20 November 2017);
Web of Science Conference Proceedings Citation Index‐Science (CPCI‐S, 1990 to 20 November 2017);
ClinicalTrials.gov (20 November 2017);
World Health Organization International Clinical Trials Registry Search Platform (ICTRP) (20 November 2017);
UMN‐CTR Japanese Clinical Trials Registry (20 November 2017);
Hong Kong Clinical Trials Registry (HKUCTR, 20 November 2017).

For the original version of this review we also searched the following three electronic databases up to 29 September 2014. However we did not need to search them for the update, as these databases have now been incorporated in the World Health Organization International Clinical Trials Registry Search Platform (ICTRP):

International Standard Randomised Controlled Trial Number Register (ISRCTN, 29 September 2014);
EU Clinical Trials Register (EU‐CTR, 29 September 2014);
Japan Primary Registries Network (29 September 2014).

All search strategies are reported in Appendix 1. We combined searches in MEDLINE, Embase and CINAHL with adaptations of the Cochrane RCT search filter, as detailed in Chapter 6 of the Cochrane Handbook for Systematic Reviews of Interventions (Lefebvre 2011).

Searching other resources

We checked references of all identified trials, relevant review articles and current treatment guidelines for further relevant literature. We limited these searches to 'first‐generation' reference lists.

Data collection and analysis

Selection of studies

One review author (CD) removed the duplicates and did the initial screening of titles and abstracts of references identified through electronic searches, excluding only those references which were clearly irrelevant (for example, when the intervention did not contain RBCs). These excluded references were not validated. For this update, two review authors (SB, AS) independently screened the remaining references and subsequently retrieved full texts for those requiring assessment for inclusion using a study‐specific eligibility form.

Data extraction and management

Two review authors (AS and MD) independently undertook data extraction for the studies identified for inclusion through the 2017 update search using a piloted, study‐specific data extraction form. Disagreements were resolved by consensus between the review authors. One review author (SB) entered the data into Review Manager 5 software (Review Manager 2014).

We undertook data extraction in accordance with guidance detailed in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). We anticipated potential issues with protocol non‐adherence. With regard to the intervention, we expected heterogeneity within the definitions of what constituted shorter and longer storage durations for the RBC units. Consequently, we specifically extracted intervention data for:

storage duration of RBC units transfused in the intervention group and in the control group; and
total volume of RBC transfusion in both groups.

In expectation of possible heterogeneity of RBC product specifications, we specifically extracted data on the use of:

irradiated blood;
whole blood; and
red cell leucoreduction ‐ leucoreduction is the filtering process by which white blood cells (leukocytes) are removed from whole blood before transfusion. White blood cells are removed because they confer no benefit to recipients, but can carry pathogens and cause adverse transfusion reactions.

In this update, SB undertook subcategorisation of studies by their definitions of shorter and longer storage durations for RBCs.

Assessment of risk of bias in included studies

Two review authors (AS and MD) assessed risk of bias for each trial using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b). We resolved disagreements by discussion. For each of the included trials, we assessed risk of bias (as low, high or unclear) for the following domains.

Generation of random sequence (selection bias)
Concealment of treatment allocation (selection bias)
Blinding of participants and personnel (person(s) delivering treatment) to treatment allocation (performance bias)
Blinding of outcome assessors to treatment allocation (detection bias)
Completeness of the outcome data (including checks for possible attrition bias through withdrawals, loss to follow‐up and protocol violations)
Selective reporting of outcomes (reporting bias)
Other sources of bias (other bias). We assessed whether each trial was free of problems, other than those listed above, that could put it at risk of bias

Measures of treatment effect

We carried out separate analyses according to the duration of follow‐up after treatment, this included: in hospital (with time points as defined by the individual studies), short term (up to 30 days), and long term (more than 30 days after receipt of study intervention). We expressed dichotomous data for each arm in individual studies as a proportion, or risk, and the treatment effect as an average risk ratio (RR) (i.e. using a random‐effects model) with 95% confidence intervals (CI), calculated using Mantel‐Haenszel methods. We expressed treatment effects for continuous data outcomes as mean differences (MD) and used a 95% CI in all instances. Where outcomes were measured in the same way across studies, we expressed continuous data for each arm in individual studies as a mean and standard deviation (SD), and the treatment effect as the MD.

With regard to the incidence of hospital‐acquired infection, we were going to estimate the incidence ratio as the ratio of new observed cases divided by expected numbers of cases for patients exposed to RBC transfusion. However, as we did not have sufficient homogenous data to permit meta‐analysis, we did not calculate incidence rates. Instead we reported incidences of, and calculated RRs for, different infections within each study.

We reported, but did not perform a formal analysis, of all reported data on length of hospital stay, length of stay in ICUs and duration of mechanical ventilation. Trials presented data on length of hospital stay and length of ICU stay as median values (with interquartile ranges), which are appropriate and robust ways to report these data, as these outcomes usually are not distributed normally; or as means (with SDs), which is usually an inappropriate way to report these particular outcomes (as length of stay is usually a skewed distribution). We have presented data for duration of mechanical ventilation as means (with SDs) and as number of participants requiring mechanical ventilation. When presented as means (with SDs), the same principles apply as for length of stay (i.e. that this is usually an inappropriate way to report an outcome of duration because usually it is skewed). We tabulated data for these outcomes and have reported them narratively within the text.

In the absence of appropriate skills in our facility to analyse economic and blood stock inventory outcomes fully at this time, we have presented these data in a narrative format. One review author (SB) entered all data into Review Manager 2014, and these were checked by a second review author (MT).

Unit of analysis issues

No unit of analysis issues arose while this review was performed; we identified no eligible cross‐over or cluster‐randomised trials. The unit of randomisation was the patient.

Dealing with missing data

We had no overall concern regarding missing data; but we did email the authors of four trials to seek clarification and request additional data (Dhabangi 2015; Steiner 2015; Cooper 2017; Spadaro 2017).

We are grateful to the authors of these four studies who responded to our emails as follows.

Cooper 2017: although the unadjusted hazard ratio for patient survival to follow‐up was reported as having been measured by the investigators, the data were not published in the trial report. The investigators sent us the hazard ratio data, but we have not used them in this version of the review because other comparable studies did not report this type of data.
Dhabangi 2015: we received clarification about the time points for mortality for the four participants who died after the first 24‐hour observation period.
Spadaro 2017: we received clarification regarding the time points at which inotropic and renal replacement therapy were reported.
Steiner 2015: we requested information about interquartile ranges for length of stay in hospital and in the ICU, and the median number of RBC units transfused per participant up to the seventh postoperative day, along with the number of transfusion‐related serious adverse events. The authors have responded indicating that they will send these data, and we hope to receive them soon.

When denominator data allowed, we converted reported percentages into actual numbers of participants for two studies (Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]), so we could calculate outcome RRs for these studies. We noted levels of attrition for all included trials.

Assessment of heterogeneity

With the addition of new trials in this 2018 update, we undertook assessments of statistical heterogeneity for three outcomes. Primarily, we explored clinical heterogeneity within studies. The clinical heterogeneity that we identified included age of the participants (neonates versus adult participants); underlying clinical diagnosis of participants across trials; duration of storage of the RBCs used for the intervention (we noted a range of storage duration for RBCs within each of the intervention arms, and in many instances, the range of the two intervention groups overlapped, see Table 1); and heterogeneity in the techniques used to measure outcomes (see Table 2, Table 3 and Table 4). We found no clinical heterogeneity due to trial design or risk of bias.

Table 1. Characteristics of the interventions

		Quantity of RBCs transfused		Duration of storage of RBCs as defined in the Methods		Duration of storage of RBCs received
Trial	Types of RBCs (RBC additive solutions)^a	Shorter storage duration	Longer storage duration/standard practice storage duration	Shorter storage duration	Longer storage duration/standard practice storage duration	Shorter storage duration	Longer storage duration/standard practice storage duration	Further information: shorter storage duration	Further information: longer storage duration/standard practice storage duration
Transfusion of shorter storage duration RBCs vs transfusion of longer storage duration RBCs
Neonate participant population
Fernandes 2005	Irradiated and leucoreduced (CPDA‐1)	Number of transfusions per infant transfused: mean 4.2 (SD 3.1); range 1 to 13 transfusions Mean RBC volume transfused: 62.9 (SD 45.1) mL/kg	Number of transfusions per infant transfused: mean 4.4 (SD 4.0); range 1 to 20 transfusions Mean RBC volume transfused: 65.3 (SD 58.3) mL/kg	RBCs stored for < 3 days	RBCs stored for ≤ 28 days	Mean: 1.6 (SD 0.6) days	Mean: 9.0 (SD 8.9) days	‐‐	‐‐
Liu 1994	Irradiated (CPDA‐1)	12 infants received 55 RBC transfusions during the study	13 infants received 73 RBC transfusions during the study	Quint packs stored for < 5 days When the need for multiple transfusions was anticipated, attempts were made to limit donor transfusions by reserving multiple quint packs from the same donor for a participant.	Assigned a single adult unit of PRBC, which was reserved for that infant up to the recommended storage time of 35 days	Mean: 3.5 (SD 1.2) days	Mean: 12.9 (SD 10.9) days	‐‐	‐‐
Paediatric participant population
Dhabangi 2013	Not stated (SAGM)	Volume of RBCs transfused, mean: 12.7 (SD 2.6) mL/kg	Volume of RBCs transfused, mean: 12.7 (SD 2.2) mL/kg	Stored for 1 to 10 days	Stored 21 to 35 days	Mean: 7.8 (SD 1.8) days	Mean: 27.2 (SD 3.9) days	‐‐	‐‐
Dhabangi 2015	Leukoreduced CP2D‐AS‐3	Not stated	Not stated	Stored for 1 to 10 days	Stored for 25 to 35 days	Median: 8 (IQR 7 to 9) days	Median: 32 (IQR 30 to 34) days	‐‐	‐‐
Adult participant population
Bennett‐Guerrero 2009 [1],	Leucoreduced (AS‐1 or AS‐3)	Number of units transfused per participant, median: 4 (IQR 2.5 to 6.5)	Number of units transfused per participant, median: 5 (IQR 4 to 10)	Mean: 7(SD 4) days	Mean: 21 (SD 4) days Transfusions in this arm were within the target window for only 5 of 10 transfused participants	Mean: 6 (V2) days	Mean: 18 (SD 7) days	Shortest stored RBC unit, mean: 5 (SD 1) days. Maximum storage duration, mean: 7 (SD 2) days	Shortest stored RBC unit, mean: 17 (SD 7) days Maximum storage duration, mean: 20 (SD 7) days
Neuman 2013	Not stated	Not stated	Not stated	Stored for < 10 days	Stored for > 21 days	Mean: 9.1 (SD 3.1) days	Mean 29.8 (SD 5.7) days	‐‐	‐‐
Schulman 2002	Not stated	Number of transfusions per infant transfused: mean 9.3 (SD 1.9)	Number of transfusions per infant transfused: mean 10.6 (SD 3.35)	Stored for < 7 days	Stored for > 20 days	Not stated	Not stated	‐‐	‐‐
Spadaro 2017	Non‐leucoreduced (SAGM)	Number of units transfused per participant, median: 2 (IQR 2 to 3)	Number of units transfused per participants, median: 2 (IQR 2 to 4)	Stored for ≤ 14 days	Stored for > 14 days	Median: 6 (IQR 5 to 10 days)	Median: 15 (IQR 11 to 20 days)	Per protocol analysis, median: 6 (IQR 5 to 10) days	Per protocol analysis, median: 18 (IQR 14 to 22) days
Steiner 2015	Leucoreduced (AS‐1, AS‐3 or AS‐5)	Number of units transfused per participant, median: 4 (IQR 2 to 6)	Number of units transfused per participants, median: 3 (IQR 2 to 6)	Stored for ≤ 10 days	Stored for > 21 days	Mean: 7.8 (SD 4.8) days	Mean: 28.3 (SD 6.7) days	‐‐	‐‐
Walsh 2004	Leucoreduced (SAGM)	Mean RBC volume transfused: 307 (SD 25) mL	Mean RBC volume transfused: 344 (SD 26) mL	RBC units stored for ≤ 5 days	RBC units stored for ≥ 20 days	Median: 2 (IQR 2 to 2.25) days	Median: 28 (IQR 26.75 to 31) days	Range = 2 to 3	Range = 22 to 32
Yuruk 2013	Leucoreduced (SAGM)	Number of RBC bags: median: 3 (IQR 2 to 3)	Number of RBC bags: median: 3 (IQR 3 to 3)	Storage duration of < 1 week	Storage duration of 3 to 4 weeks	Median: 7 (IQR 5 to 7) days	Median: 23 (IQR 22 to 28) days	‐‐	‐‐
Shorter storage duration RBCtransfusion vs standard practice storage duration
Neonate participant population
Fergusson 2012	Irradiated and leucoreduced (not stated)	Mean number of transfusion episodes: 5.01 (SD 4.00). Range 1 to 10 transfusion episodes	Mean number of transfusion episodes: 4.94 (SD 3.88) Range 1 to 10 transfusion episodes	Stored for ≤ 7 days	Range 2 to 42 days In this group, units were divided into aliquots to increase usage and reduce waste. Each aliquot was designated for use in a single infant up to its expiry date.	Mean: 5.10 (SD 2.05) days	Mean: 14.58 (SD 8.26) days	Median: 5.00 (IQR 4.00 to 6.00) days	Median: 13.00 (IQR 8.00 to 19.00) days
Strauss 1996	Irradiated (CPDA‐1 or AS‐1)	Mean number of transfusions per infant: 3.9 (SD 2.6)	Mean number of transfusions per infant: 4.7 (SD 2.8)	Stored for < 7 days	Stored for < 42 days From 1 dedicated donor who consented to donate a second unit when needed by the infant.	Stored for < 7 days = 100% of RBC transfusions (58 units transfused)	Stored for < 7 days = 17% of RBC transfusions (n = 11 units) Stored for 7 to 12 days = 36% of RBC transfusions (n = 24 units)	‐‐	The following constituted 47% of transfusions given 15 to 21 days = 16 units 22 to 28 days = 7 units 29 to 35 days = 3 units 36 to 42 days = 5 units
Strauss 2000	Irradiated, leucoreduced (CPDA‐1 or AS‐3)	Mean number of transfusions per infant: 6.7 (SD 3.9) Total number of transfusions given = 40	Mean number of transfusions per infant: 3.5 (SD 2.1) Total number of transfusions given = 28	Stored for < 7 days	Stored for< 42 days (RBC units from dedicated donors only)	All units < 7 days	28 units transfused in this arm: stored for < 7 days = 5 units; stored for 7 to 14 days = 9 units; stored for 15 to 42 days = 14 units	‐‐	‐‐
Adult participant population
Aubron 2012	Leucoreduced (SAGM)	Number of units transfused per participant, mean: 3.2 (SD 2.6)	Number of units transfused per participant, mean: 3.8 (SD 3.6)	"Freshest"	"Compatible, non expired units with the longest storage duration at the time of transfusion request”	Mean: 12.1 (SD 3.8) days Mean storage duration range of RBCs: 3 to 19 days	Mean: 23 (SD 8.4) days Mean storage duration range of RBCs: 7 to 41.5 days	Minimum storage duration, mean: 9.5 (SD 4.5) days Maximum storage duration, mean: 15 (SD 6.5) days	Minimum storage duration, mean: 20.5 (SD 8.9) days Maximum storage duration, mean: 26 (SD 9.2) days
Bennett‐Guerrero 2009 [2]	Leucoreduced (AS‐1 or AS‐3)	Number of units transfused per participants, median: 5 (IQR 2 to 7)	Number of units transfused per participants, median: 4 (IQR 2 to 5)	Less than 21 days	Unit standard of care = RBC unit stored for longest out first	Mean: 11 (4) days	Mean: 16 (SD 5) days	Shortest stored RBC unit, mean: 10 (SD 5) days Maximum storage duration, mean: 12 (SD 4) days	Shortest stored RBC unit, mean: 13 (SD 4) days Maximum storage duration, mean: 18 (SD 7) days No RBC unit was stored for > 31 days
Cooper 2017	Leucoreduced (SAGM)	Number of units transfused per participant, median: 2 (IQR 1 to 4)	Number of units transfused per participant, median: 2 (IQR 1 to 4)	"Freshest available"	"Oldest available"	Mean: 11.8 (SD 5.3) days	Mean: 22.4 (SD 7.5) days	Median: 10.7 (IQR 8.3 to 14.1) days	Median: 21.4 (IQR 16.7 to 27.4) days
Hebert 2005	Leucoreduced (CPD‐2 and AS‐3)	Median number of RBC units transfused: 3 (IQR 2 to 5)	Median number of RBC units transfused: 2 (IQR 2 to 4)	Storage duration < 8 days	This group received RBCs issued from the hospital blood bank with the longest storage time in accordance with standard blood bank procedure. To ensure the maximum separation in terms of duration of storage between groups, participants were allocated only on days when average duration of storage of RBCs in the blood bank exceeded 15 days.	Median: 4 days	Median: 19 days	Overall 73% of participants received RBCs with storage times that correspond to treatment allocation more than 90% of the time. Compliance target of 90% was attained by 91% of patients allocated to shorter storage duration arm compared with 59% in the standard duration arm.	Overall 73% of participants received RBCs with storage times that correspond to treatment allocation more than 90% of the time. Compliance target of 90% was attained by 91% of participants allocated to shorter storage duration arm compared with 59% in standard duration arm.
Heddle 2012	Leucoreduced (SAGM)	Total number of RBC units transfused: 1157 Median: 2 (IQR 2 to 4) Range 1 to 102	Total number of RBC units transfused: 2369 Median: 2 (IQR 2 to 5) Range 1 to 38	“Freshest available”	“Oldest in the inventory”	Maximum storage duration, mean: 12.0 (SD 6.8) days	Maximum storage duration, mean: 26.6 (SD 7.8) days	Range 2 to 27 days and 33 to 37 days	Range 3 to 42 days
Heddle 2016	Leucoreduced (SAGM)	Total number of RBC units transfused: 25,466 Number of units transfused per participant median: 1 (IQR 1 to 2) Range 1 to 87	Total number of RBC units transfused: 50,890 Number of units transfused per participant, median: 1 (IQR 1 to 2) Range 1 to 58	"Freshest available"	"Oldest available"	Mean: 13.0 (SD 7.6) days	Mean: 23.6 (SD 8.9) days	Median: 11 (IQR 8 to 16) days	Median: 23 (IQR 16 to 31) days
Kor 2012	Leucoreduced (not stated)	1 unit	1 unit	Single unit of RBCs stored for < 5 days Subsequent transfusions (after the first study transfusion) were standard issue	Single unit of standard issue RBCs (median = 21 days)	Median: 4.0 (IQR 3.0 to 5.0) days	Median: 26.5 (IQR 21.0 to 36.0) days	‐‐	4 participants received an RBC unit that had been stored for ≤ 14 days; 9 participants received an RBC unit that had been stored between 15 and 21 days, and 37 participants received an RBC unit that had been stored for > 21 days.
Lacroix 2015	Leucoreduced (SAGM)	Total number of RBC units transfused: 5198	Total number of RBC units transfused: 5210	"Freshest available" < 8 days	"Standard issue"	Mean: 6.1 (SD 4.9) days	Mean: 22.0 (SD 8.4) days	Mean number of RBC units per participant who received at least 1 transfusion: 4.3 (SD 5.2)	Mean number of RBC units per participant who received at least 1 transfusion: 4.3 (SD 5.5)

^aType of RBCs relates to details of whether trials report that RBCs were leucoreduced or irradiated, or whether whole blood rather than RBCs were transfused. The nature of the blood that was transfused is recorded here.

Abbreviations

AS‐1: additive solution‐1, a commercial additive solution containing sodium chloride, dextrose, adenine, mannitol
AS‐3: additive solution‐3, a commercial additive solution containing sodium chloride, dextrose, adenine, tri‐sodium citrate, citric acid, sodium phosphate
AS‐5: additive solutionl‐5, a commercial additive solution containing sodium chloride, mannitol, adenine, dextrose
CPDA‐1: citrate, phosphate, dextrose, adenine 1
CPD‐2: citrate, phosphate, dextrose‐2
CPD‐2‐AS‐3: citrate, phosphate, dextrose‐2 with additive solution‐3
IQR: interquartile range
PRBC = packed red blood cells
RBCs = red blood cells
SAGM = saline adenine glucose mannitol
SD: standard deviation

Table 2. Duration of organ support

Trial	Shorter storage duration	Longer storage duration/standard practice storage duration
Transfusion of RBCs of shorter vs longer storage duration
Respiratory support (invasive and non‐invasive ventilation)
Adult participant population
Bennett‐Guerrero 2009 [1]	Duration of mechanical ventilation, mean = 31 (SD 33) hours Percentage (number)^a of participants on any vasopressor > 48 hours after surgery = 17% (2)	Duration of mechanical ventilation, mean = 16 (SD 6) hours Percentage (number)^a of participants on any vasopressor > 48 hours after surgery = 25% (3)]
Transfusion of RBCs of shorter vs standard practice storage duration
Respiratory support (invasive and non‐invasive ventilation)
Adult participant population
Aubron 2012	Duration of mechanical ventilation, median 156 (IQR 6.1 to 253) days	Duration of mechanical ventilation, median 9.85 (IQR 0 to 198) days
Cooper 2017	Number of days alive and free of mechanical ventilation, median = 25 (IQR 11 to 28) days	Number of days alive and free of mechanical ventilation, median = 25 (IQR 13 to 28) days
Renal support (haemofiltration)
Adult participant population
Hebert 2005 Number of participants on dialysis over the 30 days of the study	0% (n = 0)	6% (n = 2)
Cooper 2017	Percentage (number)^a of participants requiring: renal replacement therapy = 13.9% (342) Days alive and free of renal replacement therapy, median: 28 (IQR 22 to 28) days	Percantage (number)^a of participants requiring: renal replacement therapy = 14.6% (360) Days alive and free of renal replacement therapy, median: 22 (IQR 22 to 28) days
Lacroix 2015	Duration of extrarenal epuration, mean: 2.5 (SD 10.1) days	Duration of extrarenal epuration, mean: 2.5 (SD 8.3) days
Haemodynamic support (vasopressors, inotropes)
Adult participant population
Lacroix 2015	Duration of cardiac or vasoactive drugs, mean: 7.1 (SD 10.2) days	Duration of cardiac or vasoactive drugs, mean: 7.5 (SD 11.2) days

^aAs the number of participants included in the analysis for this outcome was known, the number reported here was calculated for the purposes of this review.

Abbreviations

IQR: interquartile range
SD: standard deviation

Table 3. Length of stay: hospital and intensive care unit

Trial	Shorter storage duration	Longer storage duration/standard practice storage duration
Transfusion of RBCs of shorter vs longer storage duration
Neonate participant population
Fernandes 2005 Length of hospital stay: mean (SD)	60.8 (SD 37.3) days	62.6 (SD 48.3) days
Paediatric participant population
Dhabangi 2015 Length of hospital stay: median (IQR)	4 (IQR 2, 6) days*	4 (IQR 3 to 7) days*
Adult participant population
Bennett‐Guerrero 2009 [1] Length of hospital stay: mean (SD)	10 (SD 9) days	8.6 (SD 4) days
Steiner 2015 Length of hospital stay: median	8 days (IQR not reported)	8 days (IQR not reported)
Spadaro 2017 Length of hospital stay: median (IQR)	10 (IQR 6 to 17) days*	9 (IQR 7 to 17) days*
Bennett‐Guerrero 2009 [1] Length of ICU stay: mean (SD)	51 (SD 67) hours	51 (SD 56) hours
Steiner 2015 Length of ICU stay: median	3 days (IQR not reported)	3 days (IQR not reported)
Spadaro 2017 Length of ICU stay: median (IQR)	1 (IQR 1 to 6) days*	3(IQR 2 to 5) days*
Transfusion of RBCs of shorter vs standard practice storage duration
Neonate participant population
Fergusson 2012 Length of ICU stay: median (IQR)	84 (IQR 50 to 104) days*	77 (IQR 50 to 104) days*
Strauss 1996 Percentage discharge from hospital before 84 days	14%	10%
Adult participant population
Aubron 2012 Length of hospital stay: median (IQR)	21 (IQR 12 to 38) days	17 (IQR 8 to 27) days
Bennett‐Guerrero 2009 [2] Length of hospital stay: mean (SD)	10 (SD 7) days	13 (SD 14) days
Cooper 2017 Length of hospital stay: median (IQR)	14.5 (IQR 7.4 to 27.5) days	14.7 (IQR 7.4 to 28.3) days
Heddle 2016 Length of hospital stay: median (IQR)	10 (IQR 5 to 19) days	10 (IQR 5 to 20) days
Lacroix 2015 Length of hospital stay: mean (SD)	34.4 (SD 39.5) days	33.9 (SD 38.8) days
Aubron 2012 Length of ICU stay: median (IQR)	11 (IQR 5 to 15) days	7 (IQR 3 to 17) days
Bennett‐Guerrero 2009 [2] Length of ICU stay: mean (SD)	69 (SD 136) hours	47 (SD 51) hours
Cooper 2017 Length of ICU stay: median (IQR)	4.2 (IQR 2.0 to 9.3) days	4.2 (IQR 1.9 to 9.4) days
Lacroix 2015 Length of ICU stay: mean (SD)	15.3 (SD 15.4) days	15.3 (SD 14.8) days

Abbreviations

IQR: interquartile range
SD: standard deviation

See: Summary of findings for the main comparison Transfusion of RBCs of shorter vs longer storage duration; Summary of findings 2 Transfusion of RBCs of shorter vs standard practice storage duration

Table 4. Assessment of economic and blood stock inventory

Trial	Shorter storage duration	Longer storage duration/standard practice storage duration
Transfusion of RBCs of shorter vs longer storage duration
Adult participant population
Yuruk 2013 Number of bags of RBCs transfused to adult haematology participants: median (IQR)	3 (IQR 2 to 3) bags	3 (IQR 3 to 3) bags
Steiner 2015 Number of RBC units transfused per participant to postoperative day 7: median (need IQR) Number of RBC units transfused per participant "throughout the study period': median (IQR)	3 units (IQR not reported) 4 (IQR 2 to 6) units	3 units (IQR not reported) 3 units (IQR 2 to 6)
Spadaro 2017 Number of RBC units transfused per participant during surgery: median (IQR)	2 (IQR 2 to 3) units	2 (IQR 2 to 4) units
Transfusion of shorter storage duration RBCvs transfusion of standard practice storage duration
Adult participant population
Heddle 2016 Number of RBC units transfused per participant: median (IQR)	2 (IQR 2 to 4) units	2 (IQR 2 to 4) units
Cooper 2017 Number of RBC units transfused per participant: median (IQR)	2 (IQR 1 to 4) units	2 (IQR 1 to 4) units

Abbreviations

IQR: interquartile range

Assessment of reporting biases

We did not assess publication bias using funnel plots, as we had fewer than 10 studies in any one outcome analysis.

Data synthesis

The comparisons made within the primary studies were diverse because of the absence of uniform definitions of shorter and longer storage durations for RBCs. After data extraction, we assessed whether the included studies were suitable ‐ in terms of the definitions of shorter and longer storage durations they employed ‐ for inclusion in a single meta‐analysis. Unfortuantely, we found them not to be suitable (we noted a range of durations of storage of RBCs within each of the intervention arms, see Table 1), and there were also differences in clinical outcomes measured between trials (see also Assessment of heterogeneity).

We undertook meta‐analyses using Review Manager 5 on three occasions, employing a random‐effects model due to the anticipated heterogeneity arising from differences in participant characteristics and duration of storage of RBCs (Review Manager 2014).

Within each included trial, we analyzed all participants in the treatment groups to which they had been randomised.

We constructed 'Summary of findings' tables using GRADEpro GDT (GRADEpro GDT). We focused our summary of findings on the primary outcome of mortality up to 30 days. We made an assessment of the quality of the evidence based on study design limitations, inconsistency of results, indirectness of evidence, imprecision, and publication bias as described in the GRADE handbook (Schünemann 2011), with consideration of the optimal information size generated from trial sequential analysis (TSA).

Trial sequential analysis

We provided a sample size estimate showing how many participants would be needed to be included in a meta‐analysis for it to produce reliable results. We used TSA methods to explore treatment effects attained before the required sample size was reached (TSA 2011), by using TSA beta 0.9 software (TSA 2011). We applied TSA for the outcome of short‐term all‐cause mortality (up to 30 days).

We performed TSAs for this outcome on all trials regardless of risk of bias and also on trials with overall low risk of bias. We included all definitions of shorter and longer storage durations (or standard practice storage duration) for RBCs. This provided the required information size (the total number of participants) necessary to detect a statistically significant underlying effect. We estimated a mortality risk of 25% in the control group based on a mean of the observed ICU mortality of the control arms in the three recent large studies (Lacroix 2015; Heddle 2016; Cooper 2017). We calculated the information size necessary for a relative risk reduction (RRR) of 5%.

When we calculated cumulative Z‐curves that crossed trial sequential monitoring boundaries, we determined that statistical significance had been reached and the overall type I error rate had been maintained. We produced futility boundaries such that if the cumulative Z‐curve crossed the futility threshold, the evidence showed that the two treatments did not differ more than the anticipated effect size. We used the O’Brien Fleming alpha‐spending function with an overall type I error rate of 5% and with 80% statistical power to derive two‐sided sequential monitoring and futility boundaries.

Subgroup analysis and investigation of heterogeneity

If the data had been sufficient, we would have undertaken subgroup analyses based on age of participants (e.g. 'neonates', 'children' or 'adults'), and transfusion indications (long‐term transfusion dependence versus support during an acute illness), to look for differences in treatment effects.

Sensitivity analysis

If the data had been sufficient, we would have undertaken sensitivity analyses to explore aspects of methodology. These would have explored the effects of removing trials at high or unclear risk for the following bias domains: selection bias (reflecting lack of confirmation of random sequence generation and allocation concealment); detection bias (reflecting lack of assessor blinding); and attrition bias (reflecting high levels of missing data).

Results

Description of studies

See Characteristics of included studies and Characteristics of excluded studies.

Results of the search

Through the searches we ran to November 2017, we identified 4909 references, 1822 of which we excluded in the first screening because they were duplicates or were clearly irrelevant to the scope of this review. We screened the remaining 3087 references by title and abstract and excluded 3021 of them, because they did not meet the inclusion criteria for participants, or interventions, or both, or were not RCTs.

We obtained the full text of 66 references. We deemed 22 trials (reported in 50 of the 66 references) to be eligible for inclusion (Liu 1994; Strauss 1996; Strauss 2000; Schulman 2002; Walsh 2004; Fernandes 2005; Hebert 2005; Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]; Aubron 2012; Fergusson 2012; Heddle 2012; Kor 2012; Dhabangi 2013; Neuman 2013; Yuruk 2013; Dhabangi 2015; Lacroix 2015; Steiner 2015; Heddle 2016; Cooper 2017; Spadaro 2017); four trials are awaiting assessment (NCT00458783, NCT01534676; NCT02050230; NCT02724605), and there is one ongoing trial (NCT01977547). We excluded 11 studies because they did not meet the eligibility criteria of this review (Wasser 1989; Eshleman 1994; Hod 2011; Seitelbach 2011; Lebiedz 2012; Yamal 2015; Bao 2017; Rapido 2017; Rodrigues 2015; Chantepie 2015; Murphy 2017) (see Characteristics of excluded studies). We have reported full details in the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta‐Analyses) flow diagram (Figure 1).

Figure 1

A study flow diagram

Included studies

Participants in 21 trials were in‐patients (i.e. in hospital) (Liu 1994; Strauss 1996; Strauss 2000; Schulman 2002; Walsh 2004; Fernandes 2005; Hebert 2005; Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]; Aubron 2012; Fergusson 2012; Heddle 2012; Kor 2012; Dhabangi 2013; Neuman 2013; Dhabangi 2015; Lacroix 2015; Steiner 2015; Heddle 2016; Cooper 2017; Spadaro 2017). Participants in the remaining trial were outpatients of the haematology department (Yuruk 2013).

One manuscript reported data from two separate trials, which will be reported throughout as Bennett‐Guerrero 2009 [1] and Bennett‐Guerrero 2009 [2]. One trial was included in the content of a letter (Schulman 2002). Two trials were reported as conference abstracts (Schulman 2002; Neuman 2013).

Six of the 22 trials were feasibility trials for larger studies that were planned and have now been completed: Aubron 2012 for the TRANSFUSE trial (Cooper 2017); Dhabangi 2013 for the TOTAL trial (Dhabangi 2015); Hebert 2005 for the ABLE trial (Lacroix 2015); Bennett‐Guerrero 2009 [1] and Bennett‐Guerrero 2009 [2] for the RECESS trial (Steiner 2015); and Heddle 2012) for the INFORM trial (Heddle 2016). See Characteristics of included studies for details of these trials.

Sample sizes

A total of 42,635 participants was randomly assigned in the 22 trials. The number of participants assigned to each trial ranged from 17 in Schulman 2002, to 31,497 in Heddle 2016, with eight trials including more than 100 participants: Heddle 2012 (n = 910), Fergusson 2012 (n = 377), Dhabangi 2015 (n = 290), Lacroix 2015 (n = 2510), Steiner 2015 (n = 1481), Heddle 2016 (n = 31,497), Cooper 2017 (n = 4994), and Spadaro 2017 (n = 199). The total number of participants included per outcome is detailed in the Effects of interventions section and ranged from 23 to 23,281.

Setting

Eight trials were multi‐centred with participants from across Canada (Hebert 2005; Fergusson 2012, Lacroix 2015, Heddle 2016), the USA (Steiner 2015, Heddle 2016), Europe (Lacroix 2015, Cooper 2017), the Middle East (Heddle 2016, Cooper 2017), or Australia (Aubron 2012, Heddle 2016, Cooper 2017). The remaining 14 single‐centre trials were conducted in the USA (Liu 1994; Strauss 1996; Strauss 2000; Schulman 2002; Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]; Kor 2012; Neuman 2013), Brazil (Fernandes 2005), Canada (Heddle 2012), Scotland (Walsh 2004), the Netherlands (Yuruk 2013), Uganda (Dhabangi 2013; Dhabangi 2015), and Italy (Spadaro 2017).

Participants

The eligible trials included neonatal, paediatric or adult participants. Five trials enrolled very low birth weight premature neonates (Liu 1994; Strauss 1996; Strauss 2000; Fernandes 2005; Fergusson 2012), two trials focused on paediatric participants with severe malarial anaemia (Dhabangi 2013; Dhabangi 2015), and 15 trials included adult participants (Schulman 2002; Walsh 2004; Hebert 2005; Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]; Aubron 2012; Heddle 2012; Kor 2012; Neuman 2013; Yuruk 2013; Lacroix 2015; Steiner 2015; Heddle 2016; Cooper 2017; Spadaro 2017). In the trials with adult participants, the reasons for the hospital stay included cardiac surgery (Hebert 2005; Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]; Steiner 2015), haematological treatment for anaemia (Yuruk 2013), treatment for critical illness (Walsh 2004; Aubron 2012; Lacroix 2015; Cooper 2017), admission to a level 1 trauma centre (Schulman 2002), need for intubation and mechanical ventilation (Kor 2012), postoperative RBC transfusion after elective noncardiac surgery (Spadaro 2017), and the "requirement for a RBC transfusion" (with no further details given) (Heddle 2012; Neuman 2013; Heddle 2016).

Interventions

The definition (i.e. in the trial protocol or Methods section that defined the duration of storage of the RBCs) and the actual duration of storage of RBCs transfused to participants differed markedly across intervention arms and between trials (see Table 1). This was the case for trials that compared RBCs of shorter versus longer storage duration, or versus RBCs of standard practice storage duration.

Eleven trials compared transfusion of RBCs of shorter versus longer storage duration (Liu 1994; Schulman 2002; Walsh 2004; Fernandes 2005; Bennett‐Guerrero 2009 [1]; Dhabangi 2013; Neuman 2013; Yuruk 2013; Dhabangi 2015; Steiner 2015; Spadaro 2017).

Eleven trials compared transfusion of RBCs of shorter storage duration versus standard practice storage duration (Strauss 1996; Strauss 2000; Hebert 2005; Bennett‐Guerrero 2009 [2]; Aubron 2012; Fergusson 2012; Heddle 2012; Kor 2012; Lacroix 2015; Heddle 2016; Cooper 2017). The durations of storage for the standard practice storage duration RBCs was reported as: "a range of 2 to 42 days" in Fergusson 2012, "oldest in the inventory" in Bennett‐Guerrero 2009 [2], Heddle 2012, Lacroix 2015, Heddle 2016, and Cooper 2017, and less than 42 days old in Strauss 1996 and Strauss 2000.

Across all the trials, the defined durations for shorter storage of RBCs ranged from "freshest available" (Aubron 2012), to "less than 21 days old" (Bennett‐Guerrero 2009 [2]). The actual duration of storage for these RBCs ranged from a mean of 1.6 (SD 0.6) days in Fernandes 2005 to a mean of 13.0 (SD 7.6) days in Heddle 2016.

The additive solutions used to preserve RBCs ahead of transfusion and whether RBCs were leucoreduced and/or irradiated also differed between trials (see Table 1).

One trial reported instances of non‐compliance with allocated treatment (Hebert 2005). The other 21 trials either did not mention compliance with treatment allocation in their reports, or reported that there was no non‐compliance with treatment allocation.

Outcomes

No trial measured all outcomes of interest in this review.

Follow‐up periods for the primary outcomes measured ranged from within four hours of RBC transfusion to 180 days after randomisation. Details of the primary outcomes measured can be found in the Characteristics of included studies table. All studies reported measuring secondary outcomes, which are also detailed in the Characteristics of included studies table.

Excluded studies

For the full list of excluded studies, see Characteristics of excluded studies.

We excluded 11 studies from this review following assessment of their full text. We excluded two because the unit of randomisation was the intervention (Eshleman 1994; Seitelbach 2011), another because it was an observational, non‐randomised study (Lebiedz 2012), and another eight studies that assessed ineligible interventions (Wasser 1989; Hod 2011; Chantepie 2015; Rodrigues 2015; Yamal 2015; Bao 2017; Murphy 2017; Rapido 2017).

Ongoing studies and studies awaiting classification

We identified one ongoing RCT (NCT01977547; see Characteristics of ongoing studies table).

There are four studies awaiting classification (NCT00458783; NCT01534676; NCT02050230; NCT02724605), see Characteristics of studies awaiting classification table. According to ClinicalTrials.gov, NCT00458783 has completed (accessed 9 October 2014); NCT01534676 terminated after recruiting only three participants, and is also identified as completed (accessed 31 October 2018); NCT02050230 was also terminated due to slow recruitment of participants (last updated on ClinicalTrials.gov in October 2016, accessed 15 November 2018); and NCT02724605 completed in January 2018 (accessed 31 October 2018). To date, no further details have been reported for any of these studies.

Risk of bias in included studies

Allocation

Random sequence generation

We assessed 15 studies as being at low risk of bias for random sequence generation because they used either a web‐based system (Hebert 2005; Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]; Aubron 2012; Heddle 2012; Kor 2012; Lacroix 2015; Steiner 2015; Heddle 2016; Cooper 2017), or an interactive voice response system (and thereafter referral to a manual of unique random numbers generated by an independent statistician before study activation to determine trial arm allocation) (Fergusson 2012), or mixing of sealed envelopes for 20 minutes by three people (Dhabangi 2015), or a random numbers table (Liu 1994), or randomly permuted block sizes of four and six (Spadaro 2017), or random length block randomisation by an external, independent research unit (Walsh 2004).

We assessed seven studies as being at unclear risk of bias for random sequence generation, because there was insufficient information on which to base an assessment (Strauss 1996; Strauss 2000; Schulman 2002; Fernandes 2005; Dhabangi 2013; Neuman 2013; Yuruk 2013). We assessed no trials as having a high risk of bias for this domain (see Figure 2; Figure 3).

Figure 2

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies. Twenty‐two studies are included in this review.

Figure 3

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation concealment

We assessed 11 studies as being at low risk of bias for allocation concealment, because either participants' specific identification numbers were used with only the transfusion service scientist unblinded to treatment allocation (Aubron 2012; Steiner 2015), or a computerised randomisation schedule was maintained by non‐trial personnel (Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]), or only the study statistician at the co‐ordinating centre was aware of the randomisation codes (Lacroix 2015), or generation of random sequence occurred in a way that was outside the control of the treating clinician (Fergusson 2012; Heddle 2016; Cooper 2017); or allocation details were stored in sequentially numbered, sealed opaque envelopes (Hebert 2005; Dhabangi 2015; Spadaro 2017).

We assessed three studies as being at high risk of bias for allocation concealment, because either randomisation was performed using an unconcealed paper‐based sequence held in the blood bank (Heddle 2012), or the envelopes containing details of treatment allocation were not sequentially numbered (Kor 2012), or were not sequentially numbered and opaque (Walsh 2004).

We assessed eight studies as being at unclear risk of bias for allocation concealment, because there was insufficient information on which to base an assessment (Liu 1994; Strauss 1996; Strauss 2000; Schulman 2002; Fernandes 2005; Dhabangi 2013; Neuman 2013; Yuruk 2013).

Blinding

We have reported details regarding blinding to treatment allocation separately for participants, study personnel and outcome assessors.

Blinding of participants

We assessed 15 studies as being at low risk of bias for blinding of participants. In seven of these studies, participants received intensive care treatment (Walsh 2004; Hebert 2005; Aubron 2012; Kor 2012), or were neonates whose parents were blinded to treatment allocation (Strauss 1996; Strauss 2000; Fergusson 2012). For two of these studies, RBC units were provided by the hospital’s transfusion service in a blinded manner from induction of general anaesthesia to postoperative day 7, so participants were blinded to study group assignment (Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]). In Lacroix 2015 and Cooper 2017, opaque labels and/or obscuring stickers were used to cover the expiration and/or collection dates on blood units. In four studies participants were unblinded to treatment allocation, but as no outcomes were measured subjectively, we did not believe knowledge of treatment allocation would constitute a risk of bias in these trials (Heddle 2012; Dhabangi 2015; Steiner 2015; Heddle 2016).

We assessed seven studies as being at unclear risk of bias for this domain, because there was insufficient information on which to make an assessment of risk of bias (Liu 1994; Schulman 2002; Fernandes 2005; Dhabangi 2013; Neuman 2013; Yuruk 2013; Spadaro 2017). We assessed no trials as being at high risk of bias.

Blinding of study personnel

We assessed 14 trials as having low risk of bias for blinding of study personnel to treatment allocation. In 10 of these trials, study personnel were reported to be blinded to treatment allocation and it is unlikely that blinding could have been broken (Strauss 1996; Strauss 2000; Walsh 2004; Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]; Aubron 2012; Fergusson 2012; Kor 2012; Lacroix 2015; Cooper 2017). In Heddle 2012, Dhabangi 2015, Steiner 2015, and Heddle 2016, study personnel were unblinded to treatment allocation, but as no outcomes were measured subjectively, we did not believe that knowledge of treatment allocation constituted a risk of bias in these trials.

We assessed eight studies as being at unclear risk of bias for blinding of study personnel, because insufficient information was reported to permit assessment of risk of bias (Liu 1994; Schulman 2002; Fernandes 2005; Hebert 2005; Dhabangi 2013; Neuman 2013; Yuruk 2013; Spadaro 2017). We assessed no trials as being at high risk of bias.

Blinding of outcome assessors

We judged seven trials to be at low risk of detection bias because they reported blinding of outcome assessors (Fergusson 2012; Heddle 2012; Dhabangi 2015; Lacroix 2015; Steiner 2015; Heddle 2016; Cooper 2017). In Fergusson 2012, neonatologists blinded to study group allocation adjudicated composite outcomes independently. In Heddle 2012, Steiner 2015, Dhabangi 2015, and Heddle 2016, outcome assessors were not blinded, but as outcomes were objective, we believe that such knowledge would not have had an impact on outcome assessment.

We assessed 15 studies as being at unclear risk of bias for blinding of outcome assessors, because there was insufficient information reported about how outcomes were assessed (Liu 1994; Strauss 1996; Strauss 2000; Schulman 2002; Walsh 2004; Fernandes 2005; Hebert 2005; Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]; Aubron 2012; Kor 2012; Dhabangi 2013; Neuman 2013; Yuruk 2013; Spadaro 2017). We assessed no trials as being at a high risk of bias for this domain.

Incomplete outcome data

We assessed 20 trials as being at low risk of attrition bias, as all participants who were randomly assigned and received a RBC transfusion were included in the analysis of outcome data; there was minimal participant loss to follow‐up and reasons for loss to follow‐up and missing data were balanced in terms of numbers across intervention groups (Liu 1994; Strauss 1996; Strauss 2000; Walsh 2004; Fernandes 2005; Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]; Aubron 2012; Fergusson 2012; Heddle 2012; Kor 2012; Dhabangi 2013; Neuman 2013; Yuruk 2013; Lacroix 2015; Steiner 2015; Dhabangi 2015; Heddle 2016; Cooper 2017; Spadaro 2017). Information needed to assess attrition bias was insufficient in Schulman 2002.

One trial was at high risk of attrition bias (Hebert 2005). We made this assessment because outcome data for mortality outcomes at 30‐day and 90‐day follow‐up were incomplete.

Selective reporting

In 18 trials, investigators reported all prespecified outcomes in the Results section, and we deemed them to be at low risk of reporting bias (Liu 1994; Walsh 2004; Fernandes 2005; Hebert 2005; Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]; Aubron 2012; Fergusson 2012; Heddle 2012; Kor 2012; Dhabangi 2013; Neuman 2013; Yuruk 2013; Dhabangi 2015; Lacroix 2015; Steiner 2015; Heddle 2016; Cooper 2017).

Three trials did not define the outcomes they were interested in measuring in their Methods; therefore, it is impossible to identify whether reporting bias was present in these trials (Strauss 1996; Strauss 2000; Schulman 2002), and we have rated them as being at unclear risk of bias. We have rated one other trial as being at unclear risk of bias because only the primary outcome was prospectively registered (Spadaro 2017). We assessed no trials as having high risk of bias.

Other potential sources of bias

We have no concerns about other potential sources of bias.

Effects of interventions

We have reported outcome data in this review by comparison of interventions and then by outcome. We subdivided these sections further into trials of neonate, child and adult participants.

When denominator data allowed, we converted reported percentages into actual numbers of participants in two trials to enable us to calculate outcome RRs for these trials (Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]).

Diversity in the clinical setting; age of participants (see Characteristics of included studies); variability in what constituted shorter, longer and standard practice storage durations of RBCs within each of the trials; (see Table 1), and differences in measurement of outcomes, have meant that we had few opportunities to pool data for any of our primary and secondary outcomes. Therefore throughout this section, data are predominantly reported for each trial.

Transfusion of RBCs of shorter versus longer storage duration

Eleven trials, with 1864 participants, reported data for this comparison. Two trials included neonate participants (Liu 1994; Fernandes 2005), two trials included paediatric participants (Dhabangi 2013; Dhabangi 2015), and seven trials included adult participants (Schulman 2002; Walsh 2004; Bennett‐Guerrero 2009 [1]; Neuman 2013; Yuruk 2013; Steiner 2015; Spadaro 2017).

Primary outcome: Mortality: in hospital (with time points as defined by the participant group) and short term (up to 30 days)

Trials with neonate participants

No trial in this participant population reported data for this outcome.

Trials with paediatric participants

Dhabangi 2013 and Dhabangi 2015 provided data on mortality within 24 hours of the start of transfusion. Overall, there was no evidence of a difference in the risk of death between the shorter and longer storage duration intervention arms (RR 1.50, 95% CI 0.43 to 5.25; 2 trials, 364 participants, very low quality evidence; Analysis 1.1). Dhabangi 2015 also reported data on mortality at 30‐day follow‐up and found no evidence of a difference in the risk of death between the shorter and longer storage duration intervention arms (RR 1.40, 95% CI 0.45 to 4.31; 290 participants; low quality evidence; Analysis 1.1). The overall mortality event rate was low in both trials Analysis 1.1.

Trials with adult participants

Steiner 2015 reported mortality data at postoperative day 7. Investigators reported no difference in the risk of death between the shorter and longer storage duration intervention arms (RR 1.42, 95% CI 0.66 to 3.06; 1098 participants, moderate quality evidence; Analysis 1.1).

Bennett‐Guerrero 2009 [1] and Steiner 2015 reported data on mortality at up to 30 days after cardiac surgery. In these trials, there was no evidence of a difference in mortality between the shorter and longer storage duration intervention arms (RR 0.85, 95% CI 0.50 to 1.45; 1121 participants, moderate quality evidence; Analysis 1.1). This analysis was dominated by the Steiner 2015 study, which provided 96% of the participants in the analysis.

Schulman 2002 reported mortality in trauma patients, but did not report the time points of the deaths. Investigators reported no difference in the risk of death between the shorter and longer storage duration intervention arms (RR 2.25, 95% CI 0.55 to 9.17; 17 participants, very low quality evidence; Analysis 1.1).

Secondary outcome: Long‐term mortality (more than 30 days)

Trials with neonate participants

Fernandes 2005 reported mortality at up to 40 weeks' postconceptual age and observed no difference in the risk of death reported between the shorter and longer storage duration intervention arms (RR 0.90, 95% CI 0.44 to 1.85; 52 participants, very low quality evidence; Analysis 1.1).

Trials with paediatric participants

The trials investigating this participant population did not report data for this outcome (Dhabangi 2013; Dhabangi 2015).

Trials with adult participants

Spadaro 2017 reported all‐cause mortality at 90 days following non‐cardiac surgery. Investigators reported no difference in the risk of death between the shorter and longer storage duration intervention arms (RR 1.58, 95% CI 0.68 to 3.64; 199 participants, moderate quality evidence; Analysis 1.1).

Secondary outcome: Incidence of hospital‐acquired infection

Trials with neonate participants

Fernandes 2005 reported data for this outcome as the number of neonates developing clinical sepsis and necrotising enterocolitis. The trialists observed no differences between the shorter and longer storage duration intervention arms in incidence of clinical sepsis (RR 1.25, 95% CI 1.00 to 1.56; 52 participants; Analysis 1.2), or necrotising enterocolitis (RR 1.50, 95% CI 0.48 to 4.70; 52 participants; Analysis 1.2). The overall number of neonates developing clinical sepsis was high at 96% in the shorter storage arm and 77% in the longer storage arm. The overall number of neonates developing necrotising enterocolitis was considerably lower, at 23% in the shorter storage arm and 15% in the longer storage arm.

Trials with paediatric participants

The trials investigating this participant population did not report data for this outcome (Dhabangi 2013; Dhabangi 2015).

Trials with adult participants

Steiner 2015 reported data for this outcome as the number of participants who had developed "infections and infestations" 28 days after surgery, and reported no difference between the shorter and longer storage duration arms (RR 0.89, 95% CI 0.60 to 1.32; 1098 participants; Analysis 1.3). The percentage of participants who developed an infection by day 28 was 8% in the shorter storage arm and 9% in the longer storage arm (Analysis 1.3).

Spadaro 2017 reported the total number of participants who developed at least one infection, and the number of participants with a specific infection (sepsis, pulmonary, wound, peritonitis or urinary tract) by day 28. The trialists observed no differences between the shorter and longer storage duration arms for the risk of participants developing at least one infection (RR 0.85, 95% CI 0.52 to 1.41; 199 participants), sepsis (RR 0.73, 95% CI 0.26 to 2.02; 199 participants), pulmonary infection (RR 2.26, 95% CI 0.60 to 8.51; 199 participants), peritonitis (RR 1.94, 95% CI 0.36 to 10.35; 199 participants), or urinary tract infection (RR 1.94, 95% CI 0.18 to 21.06; 199 participants). Spadaro 2017 did observe a difference in the risk of participants developing a wound infection by day 28 between the two intervention arms, with fewer wound infections reported in the shorter storage duration arm (RR 0.32, 95% CI 0.12 to 0.86; 199 participants; Analysis 1.3).

Secondary outcome: Duration of respiratory support (invasive and non‐invasive ventilation), haemodynamic support (inotropic) and renal support (haemofiltration)

Respiratory support was reported as duration of mechanical ventilation and as the number of participants requiring mechanical ventilation. In terms of reporting of the duration of mechanical ventilation, this type of reporting requires the same considerations as length of stay, in that it usually has a skewed distribution and hence is better reported as a median value with interquartile ranges (IQR); none of the included trials reported the outcome in this way. For the trials that reported data on the duration of mechanical ventilation as means and SDs, we did not analyse the data, but reported them narratively and via a table (see Table 2).

Trials with neonate participants

In Fernandes 2005 researchers reported no difference between the shorter and longer storage arms in the mean duration (days) that mechanical ventilation was required (MD ‐6.40, 95% CI ‐18.83 to 6.03; 52 participants; Analysis 1.4). No trial reported data on haemodynamic or renal support in the neonate population.

Trials with paediatric participants

The trials investigating this participant population did not report data for this outcome (Dhabangi 2013; Dhabangi 2015).

Trials with adult participants

Bennett‐Guerrero 2009 [1] reported the mean duration of mechanical ventilation and measured the number of participants on any vasopressor for more than 48 hours after surgery. Investigators observed no difference between the shorter and longer storage arms in the number of participants on any vasopressor for more than 48 hours after surgery (RR 0.61, 95% CI 0.12 to 3.00; 23 participants (note: results not shown in analyses)). Adult participants in the longer storage arm required fewer hours of mechanical ventilation than those in the shorter storage arm (see Table 2).

In Spadaro 2017, investigators found no difference between the shorter and longer storage arms in the number of participants requiring inotropic support (RR 1.07; 95% CI 0.47 to 2.40; 199 participants), or renal support (RR 0.65, 95% CI 0.11 to 3.79, 199 participants) up to 28 days post surgery (Analysis 1.5).

Secondary outcome: Length of hospital and ICU stay

Trials with neonate participants

No trial reported median (IQR) data for this outcome. Fernandes 2005 reported mean (with SD) length of hospital stay. We have not reported these data: the actual data are provided in Table 3.

Trials with paediatric participants

Dhabangi 2015 reported median (IQR) length of hospital stay. We did not analyse these data: the actual data are presented in Table 3.

Trials with adult participants

Bennett‐Guerrero 2009 [1] reported mean (SD) length of hospital stay. Steiner 2015 and Spadaro 2017 reported median (IQR) ICU and hospital length of stay. We did not analyse these data: the actual data are presented in Table 3.

Secondary outcome: Adverse transfusion reactions

Trials with paediatric participants

Only one trial reported adverse transfusion reactions (Dhabangi 2015). One child in each group experienced a reaction that was related to transfusion within the first 24 hours of transfusion (facial oedema in the shorter storage arm and hives in the longer storage arm).

Trials with adult participants

Although Steiner 2015 reported serious adverse events that occurred in both intervention arms by day 28, the investigators did not specify those that were related to transfusion.

Secondary outcome: Assessment of economic or blood stock inventory outcomes

Trials with neonate participants

Fernandes 2005 and Liu 1994 reported mean number of donor transfusions per infant (i.e. mean number of donors to which each infant was exposed). This is an outcome of particular interest in very low birth weight, premature neonates. In practice, premature infants are routinely exposed to longer storage RBCs because of a 'dedicated paed pack' donor policy introduced in the 1980s to minimise the risk of transmission of viruses through transfusions from multiple donors. These packs are aliquots made from one batch of adult blood donation. These aliquots are stored and an aliquot is used each time the same infant requires an RBC transfusion. By design, this leads to higher rates of transfusion of longer storage duration RBCs. This creates a blood stock management activity. Investigators noted a difference in the mean number of donor transfusions per infant that favoured the longer storage arm (i.e. fewer donors exposed to) in both trials (MD 2.80, 95% CI 1.46 to 4.14; 52 participants in Fernandes 2005; and MD 1.70, 95% CI 0.07 to 3.33; 25 participants in Liu 1994: Analysis 1.6).

Trials with paediatric participants

The trials investigating this participant population did not report data for this outcome (Dhabangi 2013; Dhabangi 2015).

Trials with adult participants

Spadaro 2017, Steiner 2015, and Yuruk 2013 reported the median (IQR) number of RBC units transfused per participant during surgery (Spadaro 2017); number of RBC units transfused per participant up to postoperative day 7 and 'throughout the study period' (Steiner 2015); and number of bags of RBCs transfused to adult haematology participants (Yuruk 2013).

With the exception of the outcome 'throughout the study period' in Steiner 2015, overall the median number of units transfused in the shorter and longer storage duration arms were the same. The Steiner 2015 'throughout the study period' data showed that participants in the shorter storage arm received more RBC units than those in the longer storage arm (a median of 4 (IQR 2 to 6) units compared to 3 (IQR 2 to 6) units). We did not analyse these data: the actual data are presented in Table 4.

Transfusion of RBCs of shorter versus standard practice storage duration

Eleven studies that randomised 40,588 participants reported data for this comparison. Three studies included neonates (Strauss 1996; Strauss 2000; Fergusson 2012), and eight studies included adults (Hebert 2005; Bennett‐Guerrero 2009 [2]; Aubron 2012; Heddle 2012; Kor 2012; Cooper 2017; Heddle 2016; Lacroix 2015).

We use the term 'standard practice storage' when referring to the current standard practice storage duration intervention arm.

We calculated the number of participants included in the analyses of mortality at 30‐day and 90‐day follow‐ups in both intervention arms in Hebert 2005 from data provided in the study authors' Table 2 (page 1437 in Hebert 2005), as this value was not reported in the text. In this table Hebert 2005 reported the overall number of participants included in these analyses, as well as events and percentage event rates. From these figures, it was possible to calculate the actual number of participants included in these analyses per intervention.

Primary outcome: Mortality: in hospital (with time points as defined by the participant group) and short term (up to 30 days)

Trials with neonate participants

Strauss 1996 reported data on mortality up to 30 days. The investigators observed no differences in the number of reported deaths between the shorter and standard practice storage intervention arms (average RR 0.30, 95% CI 0.01 to 7.02; 40 participants, very low quality evidence; Analysis 2.1).

Trials with adult participants

Mortality data were reported by eight trials at a range of time points. Four trials reported in‐hospital mortality (Aubron 2012; Hebert 2005; Heddle 2012; Heddle 2016); three reported mortality in intensive care (Hebert 2005; Heddle 2016; Lacroix 2015); and five reported mortality at 30 days post intervention (Bennett‐Guerrero 2009 [2]; Hebert 2005; Kor 2012; Lacroix 2015; Cooper 2017) (see Characteristics of included studies table). In Bennett‐Guerrero 2009 [2], there were no reported deaths at 30‐day follow‐up.

Investigators reported no evidence of a difference in the number of reported deaths between the shorter and standard practice storage arms at any of the reported time points:

in hospital: four trials reported in‐hospital mortality with a median length of stay ranging from 10 to 21 days. When combined, there was no evidence of a difference in mortality between the two arms (average RR 1.05, 95% CI 0.97 to 1.14; 25,704 participants, moderate quality evidence; Analysis 2.1);
in ICU: three trials reported ICU mortality with a median length of stay ranging from 10 to 15 days. When combined, there was no evidence of a difference in ICU mortality between the two arms (average RR 1.06, 95% CI 0.98 to 1.15; 13,066 participants, moderate quality evidence; Analysis 2.1);
at 30‐day follow‐up across the combined data of five trials (average RR 1.04, 95% CI 0.96 to 1.13; I² = 0%; 7510 participants, moderate quality evidence; Analysis 2.1) (Bennett‐Guerrero 2009 [2]; Hebert 2005; Kor 2012; Lacroix 2015; Cooper 2017).

Secondary outcome: Long‐term mortality

Trials with neonate participants

Fergusson 2012 reported data on mortality at 90‐day follow‐up with no difference in the risk of death reported between the shorter and standard practice storage arms (average RR 0.97, 95% CI 0.61 to 1.54; 377 participants; moderate quality evidence; Analysis 2.1).

Trials with adult participants

Hebert 2005,Lacroix 2015 and Cooper 2017 reported data on mortality at 90‐day follow‐up and observed no difference in the risk of death reported between the shorter and standard practice storage arms (average RR 1.04, 95% CI 0.97 to 1.12; I² = 0%; 7398 participants, high quality evidence; Analysis 2.1).

Secondary outcome: Incidence of in‐hospital infection

Trials with neonate participants

Fergusson 2012 observed no differences between shorter and standard practice storage arms for the risk of premature neonates with necrotising enterocolitis (average RR 1.01, 95% CI 0.51 to 2.00; 377 participants; Analysis 2.2), clinically suspected infections (average RR 1.01, 95% CI 0.90 to 1.12; 377 participants; Analysis 2.2), or confirmed infections (average RR 1.06, 95% CI 0.91 to 1.22; 377 participants; Analysis 2.2).

Trials with adult participants

Lacroix 2015 reported data on hospital‐acquired infections (including hospital‐acquired pneumonia, peritonitis, mediastinitis, and bacteraemia) and observed no difference in the risk of acquiring a hospital infection between the shorter and standard practice storage arms (average RR 1.09, 95% CI 0.97 to 1.22; 2413 participants; Analysis 2.3).

Cooper 2017 reported data on 'new bloodstream infection' in the ICU and observed no difference in the risk of acquiring a new blood stream infection between the shorter and standard practice storage arms (average RR 0.90, 95% CI 0.57 to 1.41; 4919 participants; Analysis 2.3).

Secondary outcome: Duration of respiratory support (invasive and non‐invasive ventilation), haemodynamic support (inotropic) and renal support (haemofiltration)

Duration of respiratory support was reported as duration of mechanical ventilation and as the number of participants requiring mechanical ventilation. As mentioned previously, reporting of duration of mechanical ventilation requires the same considerations as length of stay, in that it is usually a skewed distribution and hence is better reported as a median IQR value. Since none of the included trials reported this outcome in this way, we did not analyse the data, but reported them narratively in a table (see Table 2).

Trials with neonate participants

Fergusson 2012 reported data on respiratory support in neonates, and noted no difference between the shorter and standard practice storage arms for the risk of premature neonates requiring mechanical ventilation (average RR 0.99, 95% CI 0.90 to 1.10; 377 participants; Analysis 2.4), or high‐frequency ventilation (average RR 1.02, 95% CI 0.79 to 1.31; 377 participants; Analysis 2.4). Full study data are reported in Table 2.

No trial reported data on haemodynamic or renal support in neonates.

Trials with adult participants

Five trials reported data on respiratory support in adults (Hebert 2005; Bennett‐Guerrero 2009 [2]; Aubron 2012; Lacroix 2015; Cooper 2017), and three reported data on renal support (Hebert 2005; Lacroix 2015; Cooper 2017). One trial reported data on haemodynamic support (Lacroix 2015). A variety of methods were used to measure this outcome. We did not undertake a meta‐analysis because of clinical diversity (ICU setting in Aubron 2012, and Cooper 2017; cardiac surgery in Hebert 2005), and because of the ways in which the outcomes were measured.

Respiratory support

When pooled, the Aubron 2012, Hebert 2005 and Cooper 2017 trials reported no differences between the shorter and standard practice storage arms for the risk of participants requiring mechanical ventilation (average RR 1.15, 95% CI 0.92 to 1.44; 3 studies, 5027 participants; I² = 72%; Analysis 2.4). Due to the high heterogeneity, we are uncertain of the true effect.

In Bennett‐Guerrero 2009 [2], investigators noted no differences between the shorter and standard practice storage arms for the risk of participants taking any vasopressor for more than 48 hours after surgery (average RR 3.67, 95% CI 0.46 to 29.49; 20 participants; Analysis 2.4).

In Hebert 2005, researchers reported no differences between the shorter and standard practice storage arms for the risk of participants requiring prolonged invasive mechanical ventilation (more than 48 hours) (average RR 1.67, 95% CI 0.60 to 4.64; 57 participants; Analysis 2.4), or prolonged mechanical ventilation (more than 48 hours) (average RR 1.34, 95% CI 0.60 to 2.98; 57 participants; Analysis 2.4), or vasoactive drugs, an aortic balloon pump or ventricular assist devices for more than 48 hours (average RR 1.49, 95% CI 0.45 to 4.98; 57 participants; Analysis 2.4).

The Lacroix 2015 trial reported no differences between the shorter and standard practice storage arms in the mean number of days for which mechanical ventilation was required (MD 0.3, 95% CI ‐1.01 to 1.61; 2413 participants; Analysis 2.5).

In Aubron 2012, researchers reported the median (with IQR) duration of mechanical ventilation and observed that it was shorter in the standard practice storage arm than in the shorter storage arm. Cooper 2017 reported no difference in the median days alive and free of invasive mechanical ventilation between arms (25 days). We did not analyse these median data: actual data are presented in Table 2.

Haemodynamic support

In Lacroix 2015, researchers reported no difference between the shorter and standard practice storage arms for the mean number of days for which cardiac or vasoactive drugs were required (MD ‐0.40, 95% CI ‐1.25 to 0.45; 2413 participants; Analysis 2.5).

Renal support

Hebert 2005 and Cooper 2017 reported the number of participants who required renal dialysis at 30 days and observed no difference in risk between the shorter and standard practice storage arms (average RR 0.95, 95% CI 0.83 to 1.09; I² = 0%; 4976 participants; Analysis 2.4).

In Lacroix 2015, researchers reported the number of days for which participants required renal support and observed no difference in risk between the shorter and standard practice storage arms (MD 0.20, 95% CI ‐0.54 to 0.94; 2413 participants; Analysis 2.5).

Cooper 2017 reported that participants in the shorter storage arm were alive and free of renal replacement therapy for more days than those in the standard practice storage arm (median (IQR)). We did not analyse these median data: the actual data are presented in Table 2.

Secondary outcome: Length of hospital and ICU stay

Trials with neonate participants

Strauss 1996 reported percentage data for length of hospital stay, and Fergusson 2012 reported median (IQR) data for the duration of intensive care unit stay. These data are available in Table 3.

Strauss 1996 reported the percentage of neonates discharged up to day 84: 67% in the shorter storage arm versus 53% in the standard practice storage arm. These data are available in Table 3.

Fergusson 2012 presented data as medians (IQR). Neonates in the shorter storage arm required a longer stay in the ICU. These data are available in Table 3.

Trials with adult participants

Aubron 2012, Cooper 2017 and Heddle 2016 presented length of stay data as medians (plus IQR). We did not analyse these data, but report them in Table 3. Bennett‐Guerrero 2009 [2] and Lacroix 2015 reported mean (with SD) length of hospital stay. We did not report these data but present them in Table 3.

In Aubron 2012, participants in the shorter storage arm required a longer stay in hospital and a longer stay in the ICU than those in the standard practice storage arm.

In Cooper 2017, participants in the shorter storage arm required a shorter stay in hospital than those in the standard practice storage arm, but there was no difference in the length of ICU stay between the two arms.

In Heddle 2016, there was no difference in the length of hospital stay between the shorter and standard practice storage arms.

Secondary outcome: Adverse transfusion reactions

Trials with neonate participants

Three trials (with 431 neonates) reported data for this outcome (Strauss 1996; Strauss 2000; Fergusson 2012). We did not undertake a meta‐analysis because event data were lacking. No adverse transfusion reactions were reported in Strauss 1996 or Strauss 2000, however, one diagnosis of cytomegalovirus (CMV) infection within a 90‐day ICU stay was reported in a neonate assigned to the standard practice storage arm in Fergusson 2012.

Trials with adult participants

Lacroix 2015 reported the number of participants who experienced an acute transfusion reaction and Cooper 2017 reported the number who experienced a febrile nonhaemolytic transfusion reaction. We did not meta‐analyse these data due to differences in the specifics of the transfusion reactions.

In Lacroix 2015, investigators observed no difference between the shorter and standard practice storage arms for risk of participants experiencing an acute transfusion reaction (average RR 0.67, 95% CI 0.19 to 2.36; 2413 participants; Analysis 2.6).

In Cooper 2017, investigators observed that participants in the standard practice storage arm experienced a lower risk of febrile nonhaemolytic transfusion reactions than those in the shorter storage arm (average RR 1.48, 95%CI 1.13 to 1.95; 4919 participants; Analysis 2.6).

Secondary outcome: Assessment of economic or blood stock inventory outcomes

Trials with neonate participants

Fergusson 2012, Strauss 1996 and Strauss 2000 reported the mean number of donor transfusions per infant (i.e. mean number of donors to which infant each infant was exposed). Differences in the mean numbers of donor transfusions per infant favoured the standard practice storage arm (i.e. exposed to fewer donors) in all trials (MD 1.62, 95% CI 1.17 to 2.07; 377 participants in Fergusson 2012; MD 2.10, 95% CI 0.82 to 3.38; 29 participants in Strauss 1996; MD 4.60, 95% CI 2.28 to 6.92; 21 participants in Strauss 2000; Analysis 2.7).

Trials with adult participants

Cooper 2017, Heddle 2016 and Lacroix 2015 reported the number of RBC units transfused per participant. In Lacroix 2015, investigators observed no difference between the shorter and standard practice storage arms in the mean number of RBC units transfused per participant (MD 0.00, 95% CI ‐0.43 to 0.43; 2413 participants; Analysis 2.7).

Heddle 2016 and Cooper 2017 reported the median (IQR) number of RBC units transfused per participant. In both trials, the median number was identical in both intervention arms. We did not analyse these data, but present them in Table 4.

Trial sequential analysis

The TSA for short‐term all‐cause mortality showed that the cumulative Z‐curve crossed the non‐inferiority boundaries after 80.9% of the required information size in the analysis with trials of low overall risk of bias and with 84.1% of the required information size in the analysis that included all trials. These results suggest that we may now have sufficient evidence to reject a 5% relative risk decrease or increase of death within 30 days when transfusing RBCs of shorter versus longer storage duration.

Discussion

available in

Summary of main results

This 2018 update of our 2015 review, has included six additional randomised controlled trials (RCTs) with an additional 40,771 participants, which represents a marked increase from the 1864 participants included in Brunskill 2015. In total, this 2018 update includes 22 RCTs that randomised 42,635 participants across a wide range of clinical settings.

Two main features were noted across these 22 RCTs:

There was marked clinical heterogeneity, as the trials were conducted in a variety of clinical settings, specifically critically ill neonates, children, hospitalised adults, cardiac surgery and adult intensive care.
There was wide diversity in the size, methodology and reporting of primary outcomes – from the highly pragmatic INFORM trial (Heddle 2016), which recruited over 30,000 participants and reported on in‐hospital mortality to smaller trials such as TOTAL (Dhabangi 2015), which closely evaluated physiology or efficacy end‐points, or both.

We have summarised the main findings for our two pre‐defined comparisons.

Transfusion of RBCs of shorter versus longer storage duration

Eleven trials, with 1864 participants, compared transfusion of RBCs of shorter versus longer storage duration. All trials contributed to mortality endpoints; and there was no evidence of difference in risk of mortality between the arms at any of our predefined time points (Analysis 1.1).

Not all studies provided data for our other predefined secondary clinical outcomes. For those that did, there were no significant differences reported between the arms for incidence of infection, requirements for organ support, or length of stay. Only one trial provided information on adverse reactions (Dhabangi 2015).

Assessment of economic and blood stock inventory outcomes were reported as the mean number of donor transfusions in two neonatal trials (i.e. the mean number of donors to which each infant was exposed).

Transfusion of RBCs of shorter versus standard practice storage duration

Eleven trials, with 40,588 participants, compared transfusion of RBCs of shorter versus standard practice storage duration. The large number of participants was due to one study, Heddle 2016, which provided qualitative data for 24,736 participants. All trials provided data on mortality. There was no evidence of a difference in risk of mortality between the arms at any of our predefined time points (Analysis 2.1). There were no clear differences reported between the arms for secondary clinical outcomes across all studies. We judged the quality of evidence to be high for long‐term mortality.

Only one study of critically ill participants contributed data to the outcome of adverse transfusion reactions (Cooper 2017). The trialists observed fewer febrile non‐haemolytic transfusion reactions in the standard practice storage arm than in the shorter storage arm. Otherwise, for studies that provided data on our secondary clinical outcomes, there were no significant differences reported between both arms for incidence of infection, requirements for organ support, or length of stay. In the three neonatal trials, participants who received RBCs of standard practice storage duration had a significantly fewer number of donor exposures than those who received RBCs of shorter storage duration.

Results from the Trial Sequential Analysis indicate that we may now have enough evidence to reject a 5% relative risk decrease/increase of death within 30 days when transfusing shorter storage blood versus blood of any longer storage duration across a range of clinical settings.

We identified four trials that are awaiting further assessment (NCT00458783; NCT01534676; NCT02050230; NCT02724605), and one ongoing RCT in critically ill paediatric patients that will be incorporated into future updates of this systematic review (NCT01977547). The ongoing RCT will evaluate changes in organ dysfunction as a primary outcome.

Overall completeness and applicability of evidence

This 2018 review update includes 22 studies that randomised a total of 42,635 participants, and now provides the most up‐to‐date evidence to assess the consequences of transfusion of RBCs of shorter versus longer storage duration. The trials were set across a wide range of clinical settings, and included the large INFORM trial that recruited across all hospitalised patients (Heddle 2016), which suggests broad applicability of the findings. However, certain subgroups of participants were less well represented as distinct groups, for example, those suffering from bleeding or trauma, in whom multiple transfusions are often required. The large number of participants, overall, was heavily influenced by one study (Heddle 2016).

In the previous version of this review (Brunskill 2015), we discussed several factors that affected applicability of the evidence. These included different baseline risks between different participant groups, differences in underlying physiology, small sample sizes (mainly due to pilot or feasibility studies), predominance of physiological efficacy outcome measures over clinically important outcomes, the effect of multiple units of transfusion and concerns about effects of extremes of storage duration. Some of these have now been addressed with the results of large RCTs, which were powered to detect differences in clinically important outcomes such as mortality and morbidity.

Twelve trials provided data on physiological markers of oxygen consumption or altered micro‐circulation for assessing the clinical effect of storage duration on the efficacy of storage duration. We were unable to undertake meta‐analysis of any participant population due to widespread differences in the measurements used to assess efficacy (see Table 5). In addition, such physiological markers can be altered by many other interventions in critically ill patients (e.g. fluid resuscitation, vasopressors/intotropes, renal replacement therapy), and therefore trying to disentangle the specific effect of storage duration of blood is likely to be very difficult.

Table 5. Physiological markers of oxygen consumption or alterations in microcirculation

Trial	Shorter storage duration	Longer storage duration/standard practice storage duration
Transfusion of RBCs of shorter vs longer storage duration
Neonatal participant population
Liu 1994	Mean change (before and after transfusion) in: potassium concentration (mmol/L) = 0.10 (SD 1.22) pH level = ‐0.02 (SD 0.08) base excess = ‐0.82 (SD 1.99)	Mean change (before and after transfusion) in: potassium concentration (mmol/L) = 0.10 (SD 1.22) pH level = ‐0.02 (SD 0.08) base excess = ‐0.82 (SD 1.99)
Paediatric participant population
Dhabangi 2013	Resolution of lactic acidosis within 4 hours of RBC transfusion = 92% (n = 34)	Resolution of lactic acidosis within 4 hours of RBC transfusion = 81% (n = 30)
Dhabangi 2015	Proportion of participants with a lactate of 3 mmol/L or lower at 8 hours after RBC transfusion = 58% (83/143) Cerebral tissue oxygen saturation (%) Beginning of transfusion: 72.8% End of transfusion: 77%	Proportion of participants with a lactate of 3 mmol/L or lower at 8 hours after RBC transfusion = 61% (87/143) Cerebral tissue oxygen saturation (%) Beginning of transfusion: 71.6% End of transfusion: 78.9%
Adult participant population
Bennett‐Guerrero 2009 [1]	PaO₂:FiO₂ ratio, mean at: Baseline = 420 (SD 158) 2 hours post operation = 256 (SD 73) Lactic acid, concentration 2 hours post operation: mean = 1.9 (SD 1.6)	PaO₂:FiO₂ ratio, mean at: Baseline = 386 (SD 128) 2 hours post‐operation = 360 (SD 64) Lactic acid, concentration 2 hours post operation: mean = 2.5 (SD 1.5)
Neuman 2013	"patients had a [mean (SD)] 0.191 (0.438) and 0.069 (0.242) µM increase in N0₂ at 1 and 24 hours after transfusion respectively" "Patients receiving fresh pRBC [PRBC] had no significant change in FMD" (i.e. flow‐mediated dilation, a measure of vascular endothelial function) No significant change in FHb (plasma‐free Hb) was noted before or after a RBC transfusion	"patients had a [mean (SD)] 0.191 (0.438) and 0.069 (0.242) µM increase in N0₂ at 1 and 24 hours after transfusion respectively" "Patients receiving fresh pRBC [PRBC] had no significant change in FMD" (flow‐mediated dilation, a measure of vascular endothelial function) No significant change in FHb (plasma‐free Hb) was noted before or after a RBC transfusion
Walsh 2004	Median (plus IQR) from baseline to post‐transfusion period Intragastric arterial difference in PCO₂ (Pg‐PaCO₂ gap) = ‐0.15 (IQR ‐0.31 to ‐0.01) PgCO₂,kPa = ‐0.46 (IQR ‐1.36 to ‐0.18) PgCO₂ ‐PaCO₂,kPa = ‐0.41 (IQR ‐0.82 to 0.10) pHi = 0.02 (IQR ‐0.01 to 0.05) Lactate = 0.10 (IQR ‐0.16 to 0.21) mmol/L Arterial pH = 0.00 (IQR ‐0.001 to 0.001) HCO₃ (actual) = 0.85 (IQR ‐1.44 to ‐0.53) mmol/L Arterial base excess = ‐0.70 (IQR ‐1.42 to ‐0.45)	Median (plus IQR) from baseline to post‐transfusion period Intragastric arterial difference in PCO2 (Pg‐PaCO2 gap) = 0.02 (IQR ‐0.29 to 0.11) PgCO₂,kPa = 0.16 (IQR ‐0.36 to 1.19) PgCO₂ ‐PaCO₂,kPa = 0.37 (IQR ‐0.17 to 1.66) pHi = ‐0.02 (IQR ‐0.06 to 0.01) Lactate = 0.03 (IQR ‐0.10 to 0.23) mmol/L Arterial pH = 0.00 (IQR ‐0.001 to 0.001) HCO₃ (actual) = ‐0.29 (IQR ‐0.90 to 0.07) mmol/L Arterial base excess = ‐0.71 (IQR ‐1.5 to ‐0.42)
Yuruk 2013	Mean changes in microcirculatory density (measured before and 30 minutes after transfusion) from 17 (SD 3) mm/mm² to 19 (SD 3) mm/mm².	Mean changes in microcirculatory density (measured before and 30 minutes after transfusion) from 17 (SD 3) mm/mm² to 19 (SD 2) mm/mm².
Transfusion of shorter storage duration RBCs vs transfusion of standard practice storage durationRBCs
Neonatal participant population
Fergusson 2012	Percentage (number) of participants requiring: supplemental oxygen = 96.8% (182) nasal continuous positive airway pressure = 79.3% (149)	Percentage (number) of participants requiring: supplemental oxygen = 94.7% (179) nasal continuous positive airway pressure = 80.4% (152)
Strauss 1996	Mean change from before to after transfusion: pH = 0.00 (SD 0.05) Sodium = 0.5 (SD 4.4) mEq/L Potassium = 0.3 (SD 0.9) mEq/L Calcium = ‐0.1 (SD 0.7) mg/dL Lactate = ‐0.3 (SD 0.4) mmol/L Glucose = ‐9 (SD 32) mg/dL	Mean change from before to after transfusion: pH = 0.00 (SD 0.13) Sodium = 0.6 (SD 5.1) mEq/L Potassium = ‐0.2 (SD 1.5) mEq/L Calcium = ‐0.1 (SD 0.5) mg/dL Lactate = ‐0.1 (SD 0.6) mmol/L Glucose = ‐7 (SD 25) mg/dL
Strauss 2000	Mean change from before to after transfusion: pH = ‐0.01 (SD 0.9) Sodium = ‐0.4 (SD 5.5) mEq/L Potassium = 0.26 (SD 1.16) mEq/L Calcium = ‐0.13 (SD 1.04) mg/dL Lactate = ‐0.16 (SD 0.30) mmol/L Glucose = ‐17.5 (SD 40.3) mg/dL	Mean change from before to after transfusion: pH = ‐0.03 (SD 0.6) Sodium = 1.9 (SD 4.7) mEq/L Potassium = 0.38 (SD 0.91) mEq/L Calcium = 0.13 (SD 0.64) mg/dL Lactate = ‐0.53 (SD 1.46) mmol/L Glucose = ‐16.7 (SD 24.4) mg/dL
Adult participant population
Bennett‐Guerrero 2009 [2]	PaO₂:FiO₂ ratio, mean at: Baseline = 414 (SD 97) 2 hours post operation = 267 (SD 81) Lactic acid, concentration 2 hours post operation: mean = 2.2 (SD 1.1)	PaO₂:FiO₂ ratio, mean at Baseline = 284 (SD 122) 2 hours post operation = 212 (SD 87) Lactic acid, concentration 2 hours post operation: mean = 2.4 (SD 1.0)
Kor 2012	Mean change from before to after transfusion PaO₂:FiO₂ = 2.5 (SD 49.3) mmHg	Mean change from before to after transfusion PaO₂:FiO₂ = ‐9.0 (SD 69.8) mmHg

Abbreviations

Hb: haemoglobin
IQR: interquartile range
Pg‐PaCO₂: intragastric minus arterial partial pressure of carbon dioxide
PgCO₂,kPa; intragastric partial pressure of carbon dioxide
pHi: intramucosal gastric pH
PRBC: packed red blood cells
RBC: red blood cell
SD: standard deviation

However, despite the high quality of some of the newer trials, certain factors were still present which limited our ability to perform a complete meta‐analysis. In particular, outcomes such as mortality were not standardised across the trials and were reported at multiple time points ranging from 10 days to 90 days in adult participants. Trial processes, such as separation strategies, were also not similar across all the trials. Therefore, even the addition of these recent large RCTs has not altered, but has confirmed, the overall conclusion of our previous review where we reported that no clear difference in mortality was observed between the different storage durations of RBCs.

Quality of the evidence

We summarised the GRADE quality of evidence in separate 'Summary of findings' tables for different comparisons (summary of findings Table for the main comparison; summary of findings Table 2). The overall quality of evidence for our mortality outcomes ranged from very low to moderate. Due to the potential underlying physiological differences between neonates, children and adults we subdivided our 'Summary of findings' tables into these groups. We downgraded studies for indirectness when the primary outcome of the trial(s) in that particular population was different to our chosen outcome of in‐hospital and short‐term (up to 30 days) mortality, and for indirectness when there were a very small number of deaths. Across the different subpopulations we found no issues with inconsistency.

The previous version of this review described difficulties in assessing risk of bias mainly due to lack of information (Brunskill 2015), which led to many domains being marked as being at an unclear risk of bias. In this 2018 update, we identified six new trials, four of which were large multicentre RCTs at low risk of bias across all domains. Some trials were unable to blind participants and clinicians due to regulatory restrictions concerning blood labelling, but we felt this was not an important issue as we used objective end‐points, such as mortality. The majority of trials were at low risk for attrition and reporting bias.

Potential biases in the review process

We identified no biases in the review process. The strengths of this review lie in the robust and comprehensive methods employed to find and assess all relevant trials. We have followed standard Cochrane methods for data extraction, analyzed our results with reference to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a; Higgins 2011b), and referred to a statistician when necessary. We have had a clinician and a methodologist working at all stages, independently of each other, to control for any bias that might ensue from knowledge of clinical or systematic review methods. Where necessary, we contacted trial authors to obtain missing data.

A potential weakness of this systematic review was our decision not to predefine the interventions of shorter, longer and standard practice storage duration, but rather to use the terminology used by the primary studies. This has meant that the distribution of RBC storage duration within the intervention groups is not always clear, and there is substantial cross‐over and overlap of storage durations between intervention groups.

Agreements and disagreements with other studies or reviews

This review is an update of a Cochrane Review published in 2015. The original review included 16 studies with 1864 participants. This updated review included six more studies with 40,771 more participants. In the previous review, we chose not to perform a meta‐analysis, but did perform it in this update, a result of the inclusion of large, high‐quality trials targeted towards our predefined clinically important outcomes (albeit at different time points). These trials also included large numbers of particularly vulnerable patient groups e.g. critically ill people.

The findings of this review are similar to those of other very recent systematic reviews. Rygard 2018 specifically focused on critically ill patients and found no evidence of benefit of transfusion of blood of shorter storage duration. The findings of their trial sequential analysis also agree with ours. They reported reaching the required information size to reject a 10% relative risk increase/decrease of death; we have enough evidence to reject an even smaller (5%) relative risk increase/decrease of death within 30 days, although we acknowledge that most of the data come from adult participants. Chai‐Adisaksopha 2017 performed a systematic review and meta‐analysis, which included 14 trials and 26,374 hospitalised patients of all ages, and also found no evidence of an effect of storage duration of RBCs on overall in‐hospital mortality.

Figure 1

A study flow diagram

Figure 2

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies. Twenty‐two studies are included in this review.

Figure 3

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Analysis 1.1

Comparison 1 Transfusion of shorter storage duration RBCs vs longer storage duration RBCs, Outcome 1 Mortality.

Analysis 1.2

Comparison 1 Transfusion of shorter storage duration RBCs vs longer storage duration RBCs, Outcome 2 Incidence of hospital‐acquired infection in neonates.

Analysis 1.3

Comparison 1 Transfusion of shorter storage duration RBCs vs longer storage duration RBCs, Outcome 3 Incidence of hospital‐acquired infection in an adults.

Analysis 1.4

Comparison 1 Transfusion of shorter storage duration RBCs vs longer storage duration RBCs, Outcome 4 Duration of organ support.

Analysis 1.5

Comparison 1 Transfusion of shorter storage duration RBCs vs longer storage duration RBCs, Outcome 5 Number of participants requiring organ support (adults).

Analysis 1.6

Comparison 1 Transfusion of shorter storage duration RBCs vs longer storage duration RBCs, Outcome 6 Economic or blood stock inventory outcomes.

Analysis 2.1

Comparison 2 Transfusion of shorter storage duration RBCs vs standard practice storage duration RBCs, Outcome 1 Mortality.

Analysis 2.2

Comparison 2 Transfusion of shorter storage duration RBCs vs standard practice storage duration RBCs, Outcome 2 Incidence of hospital‐acquired infection in a neonates.

Analysis 2.3

Comparison 2 Transfusion of shorter storage duration RBCs vs standard practice storage duration RBCs, Outcome 3 Incidence of hospital‐acquired infection in adults.

Analysis 2.4

Comparison 2 Transfusion of shorter storage duration RBCs vs standard practice storage duration RBCs, Outcome 4 Number of participants requiring organ support.

Analysis 2.5

Comparison 2 Transfusion of shorter storage duration RBCs vs standard practice storage duration RBCs, Outcome 5 Duration of organ support.

Analysis 2.6

Comparison 2 Transfusion of shorter storage duration RBCs vs standard practice storage duration RBCs, Outcome 6 Number of participants experiencing an adverse transfusion reaction.

Analysis 2.7

Comparison 2 Transfusion of shorter storage duration RBCs vs standard practice storage duration RBCs, Outcome 7 Economic or blood stock inventory outcomes.

Summary of findings for the main comparison. Transfusion of RBCs of shorter vs longer storage duration

Transfusion of RBCs of shorter vs longer storage duration for all conditions
Patient or population: all conditions Setting: hospital Intervention: transfusion of RBCs of shorter storage duration Comparison: transfusion of RBCs of longer storage duration
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with transfusion of RBCs of longer storage duration	Risk with transfusion of RBCs of shorter storage duration
Mortality: in‐hospital mortality (within 24 hours)	Study population ‐ paediatric participants		RR 1.71 (0.41 to 7.10)	290 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
	21 per 1000	35 per 1000 (8 to 147)
Mortality: in‐hospital mortality (within 7 days)	Study population ‐ paediatric participants		RR 3.00 (0.13 to 71.34)	74 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 2 3}
	0 per 1000	0 per 1000 (0 to 0)
Mortality: in‐hospital mortality (within 7 days)	Study population ‐ adults		RR 1.42 (0.66 to 3.06)	1098 (1 RCT)	⊕⊕⊕⊝ MODERATE ⁴
	20 per 1000	28 per 1000 (13 to 60)
Mortality: time‐point not defined	Study population ‐ adults		RR 2.25 (0.55 to 9.17)	17 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{5 6 7}
	222 per 1000	500 per 1000 (122 to 1000)
Mortality: short term (up to 30 days)	Study population ‐ paediatric participants		RR 1.44 (0.46 to 4.48)	290 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
	34 per 1000	50 per 1000 (16 to 154)
Mortality: short term (up to 30 days)	Study population ‐ adults		RR 0.85 (0.50 to 1.45)	1121 (2 RCTs)	⊕⊕⊕⊝ MODERATE ⁸
	51 per 1000	43 per 1000 (25 to 74)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: we are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low certainty: we have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect
¹ Downgraded 1 level for indirectness, as the trial was not designed to assess the impact of transfusion on mortality, rather it was designed to assess if storage duration influenced the efficacy of RBCs in the particular participant population. ² Downgraded 1 level for imprecision, due to very small number of events. ³ Downgraded 1 level, as no information was reported to permit assessment of biases due to selection, performance and detection. However there were no concerns with regard to attrition or reporting bias. ⁴ Downgraded 1 level for indirectness, as the trial was not designed to assess the impact of transfusion on mortality, but was designed to assess the effect of storage duration on organ dysfunction in elective cardiac surgery patients. ⁵ Downgraded 1 level for imprecision, due to small sample size of 17 participants. ⁶ Downgraded 2 levels, as no information was reported to permit assessment of the risk of selection, performance, detection, attrition and reporting biases. The study was reported as part of a letter. ⁷ Downgraded 1 level for indirectness, as the trial was not designed to assess the impact of transfusion on mortality, but was designed to assess the feasibility of running a larger trial and whether the blood bank in the investigating hospital could provide an adequate inventory of RBCs. ⁸ Downgraded 1 level for indirectness, as the trials were not designed to assess the impact of transfusion on mortality, rather Steiner 2015 was designed to assess the effect of storage duration on organ dysfunction in elective cardiac surgery patients and Bennett‐Guerrero 2009 was a feasibility trial designed to assess recruitment and randomisation rates and to see if adequate separation of duration of storage between the two intervention arms was feasible (Bennett‐Guerrero 2009 [1]; Bennett‐Guerrero 2009 [2]).

Summary of findings for the main comparison. Transfusion of RBCs of shorter vs longer storage duration

Summary of findings 2. Transfusion of RBCs of shorter vs standard practice storage duration

Transfusion of RBCs of shorter vs standard practice storage duration for all conditions
Patient or population: all conditions Setting: hospital Intervention: transfusion of RBCs of shorter storage duration Comparison: transfusion of RBCs of standard practice storage duration
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with transfusion of RBCs of standard practice storage duration	Risk with transfusion of RBCs of shorter storage duration
Mortality: in‐hospital mortality (time points varied)	Study population ‐ adults		RR 1.05 (0.97 to 1.14)	25754 (4 RCTs)	⊕⊕⊕⊝ MODERATE ¹
	88 per 1000	92 per 1000 (85 to 100)
Mortality: in‐ICU mortality	Study population ‐ adults		RR 1.06 (0.98 to 1.15)	13066 (3 RCTs)	⊕⊕⊕⊝ MODERATE ²
	145 per 1000	153 per 1000 (142 to 167)
Mortality: short‐term mortality (up to 30 days) ‐ neonates	Study population ‐ neonates		RR 0.30 (0.01 to 7.02)	40 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{3 4 5}
	53 per 1000	16 per 1000 (1 to 369)
Mortality: short‐term mortality (up to 30 days) ‐ adults	Study population ‐ adults		RR 1.04 (0.96 to 1.13)	7510 (4 RCTs)	⊕⊕⊕⊝ MODERATE ⁶
	223 per 1000	232 per 1000 (214 to 252)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: we are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low certainty: we have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect
¹ Downgraded 1 level for high risk of bias in 2 trials; selection bias in Heddle 2012 and attrition bias in Hebert 2005. ² Downgraded 1 level for high risk of bias in 1 trial: attrition bias in Hebert 2005. ³ Downgraded 1 level for risk of bias, as information was not reported that would permit assessment of selection, detection and reporting biases. ⁴ Downgraded 1 level for indirectness, as the study was designed to measure safety of the duration of storage of RBCs. ⁵ Downgraded 1 level for imprecision due to small number of events. ⁶ Downgraded 1 level for high risk of bias in 2 trials; selection bias in Kor 2012 and attrition bias in Hebert 2005.

Summary of findings 2. Transfusion of RBCs of shorter vs standard practice storage duration

Table 1. Characteristics of the interventions

		Quantity of RBCs transfused		Duration of storage of RBCs as defined in the Methods		Duration of storage of RBCs received
Trial	Types of RBCs (RBC additive solutions)^a	Shorter storage duration	Longer storage duration/standard practice storage duration	Shorter storage duration	Longer storage duration/standard practice storage duration	Shorter storage duration	Longer storage duration/standard practice storage duration	Further information: shorter storage duration	Further information: longer storage duration/standard practice storage duration
Transfusion of shorter storage duration RBCs vs transfusion of longer storage duration RBCs
Neonate participant population
Fernandes 2005	Irradiated and leucoreduced (CPDA‐1)	Number of transfusions per infant transfused: mean 4.2 (SD 3.1); range 1 to 13 transfusions Mean RBC volume transfused: 62.9 (SD 45.1) mL/kg	Number of transfusions per infant transfused: mean 4.4 (SD 4.0); range 1 to 20 transfusions Mean RBC volume transfused: 65.3 (SD 58.3) mL/kg	RBCs stored for < 3 days	RBCs stored for ≤ 28 days	Mean: 1.6 (SD 0.6) days	Mean: 9.0 (SD 8.9) days	‐‐	‐‐
Liu 1994	Irradiated (CPDA‐1)	12 infants received 55 RBC transfusions during the study	13 infants received 73 RBC transfusions during the study	Quint packs stored for < 5 days When the need for multiple transfusions was anticipated, attempts were made to limit donor transfusions by reserving multiple quint packs from the same donor for a participant.	Assigned a single adult unit of PRBC, which was reserved for that infant up to the recommended storage time of 35 days	Mean: 3.5 (SD 1.2) days	Mean: 12.9 (SD 10.9) days	‐‐	‐‐
Paediatric participant population
Dhabangi 2013	Not stated (SAGM)	Volume of RBCs transfused, mean: 12.7 (SD 2.6) mL/kg	Volume of RBCs transfused, mean: 12.7 (SD 2.2) mL/kg	Stored for 1 to 10 days	Stored 21 to 35 days	Mean: 7.8 (SD 1.8) days	Mean: 27.2 (SD 3.9) days	‐‐	‐‐
Dhabangi 2015	Leukoreduced CP2D‐AS‐3	Not stated	Not stated	Stored for 1 to 10 days	Stored for 25 to 35 days	Median: 8 (IQR 7 to 9) days	Median: 32 (IQR 30 to 34) days	‐‐	‐‐
Adult participant population
Bennett‐Guerrero 2009 [1],	Leucoreduced (AS‐1 or AS‐3)	Number of units transfused per participant, median: 4 (IQR 2.5 to 6.5)	Number of units transfused per participant, median: 5 (IQR 4 to 10)	Mean: 7(SD 4) days	Mean: 21 (SD 4) days Transfusions in this arm were within the target window for only 5 of 10 transfused participants	Mean: 6 (V2) days	Mean: 18 (SD 7) days	Shortest stored RBC unit, mean: 5 (SD 1) days. Maximum storage duration, mean: 7 (SD 2) days	Shortest stored RBC unit, mean: 17 (SD 7) days Maximum storage duration, mean: 20 (SD 7) days
Neuman 2013	Not stated	Not stated	Not stated	Stored for < 10 days	Stored for > 21 days	Mean: 9.1 (SD 3.1) days	Mean 29.8 (SD 5.7) days	‐‐	‐‐
Schulman 2002	Not stated	Number of transfusions per infant transfused: mean 9.3 (SD 1.9)	Number of transfusions per infant transfused: mean 10.6 (SD 3.35)	Stored for < 7 days	Stored for > 20 days	Not stated	Not stated	‐‐	‐‐
Spadaro 2017	Non‐leucoreduced (SAGM)	Number of units transfused per participant, median: 2 (IQR 2 to 3)	Number of units transfused per participants, median: 2 (IQR 2 to 4)	Stored for ≤ 14 days	Stored for > 14 days	Median: 6 (IQR 5 to 10 days)	Median: 15 (IQR 11 to 20 days)	Per protocol analysis, median: 6 (IQR 5 to 10) days	Per protocol analysis, median: 18 (IQR 14 to 22) days
Steiner 2015	Leucoreduced (AS‐1, AS‐3 or AS‐5)	Number of units transfused per participant, median: 4 (IQR 2 to 6)	Number of units transfused per participants, median: 3 (IQR 2 to 6)	Stored for ≤ 10 days	Stored for > 21 days	Mean: 7.8 (SD 4.8) days	Mean: 28.3 (SD 6.7) days	‐‐	‐‐
Walsh 2004	Leucoreduced (SAGM)	Mean RBC volume transfused: 307 (SD 25) mL	Mean RBC volume transfused: 344 (SD 26) mL	RBC units stored for ≤ 5 days	RBC units stored for ≥ 20 days	Median: 2 (IQR 2 to 2.25) days	Median: 28 (IQR 26.75 to 31) days	Range = 2 to 3	Range = 22 to 32
Yuruk 2013	Leucoreduced (SAGM)	Number of RBC bags: median: 3 (IQR 2 to 3)	Number of RBC bags: median: 3 (IQR 3 to 3)	Storage duration of < 1 week	Storage duration of 3 to 4 weeks	Median: 7 (IQR 5 to 7) days	Median: 23 (IQR 22 to 28) days	‐‐	‐‐
Shorter storage duration RBCtransfusion vs standard practice storage duration
Neonate participant population
Fergusson 2012	Irradiated and leucoreduced (not stated)	Mean number of transfusion episodes: 5.01 (SD 4.00). Range 1 to 10 transfusion episodes	Mean number of transfusion episodes: 4.94 (SD 3.88) Range 1 to 10 transfusion episodes	Stored for ≤ 7 days	Range 2 to 42 days In this group, units were divided into aliquots to increase usage and reduce waste. Each aliquot was designated for use in a single infant up to its expiry date.	Mean: 5.10 (SD 2.05) days	Mean: 14.58 (SD 8.26) days	Median: 5.00 (IQR 4.00 to 6.00) days	Median: 13.00 (IQR 8.00 to 19.00) days
Strauss 1996	Irradiated (CPDA‐1 or AS‐1)	Mean number of transfusions per infant: 3.9 (SD 2.6)	Mean number of transfusions per infant: 4.7 (SD 2.8)	Stored for < 7 days	Stored for < 42 days From 1 dedicated donor who consented to donate a second unit when needed by the infant.	Stored for < 7 days = 100% of RBC transfusions (58 units transfused)	Stored for < 7 days = 17% of RBC transfusions (n = 11 units) Stored for 7 to 12 days = 36% of RBC transfusions (n = 24 units)	‐‐	The following constituted 47% of transfusions given 15 to 21 days = 16 units 22 to 28 days = 7 units 29 to 35 days = 3 units 36 to 42 days = 5 units
Strauss 2000	Irradiated, leucoreduced (CPDA‐1 or AS‐3)	Mean number of transfusions per infant: 6.7 (SD 3.9) Total number of transfusions given = 40	Mean number of transfusions per infant: 3.5 (SD 2.1) Total number of transfusions given = 28	Stored for < 7 days	Stored for< 42 days (RBC units from dedicated donors only)	All units < 7 days	28 units transfused in this arm: stored for < 7 days = 5 units; stored for 7 to 14 days = 9 units; stored for 15 to 42 days = 14 units	‐‐	‐‐
Adult participant population
Aubron 2012	Leucoreduced (SAGM)	Number of units transfused per participant, mean: 3.2 (SD 2.6)	Number of units transfused per participant, mean: 3.8 (SD 3.6)	"Freshest"	"Compatible, non expired units with the longest storage duration at the time of transfusion request”	Mean: 12.1 (SD 3.8) days Mean storage duration range of RBCs: 3 to 19 days	Mean: 23 (SD 8.4) days Mean storage duration range of RBCs: 7 to 41.5 days	Minimum storage duration, mean: 9.5 (SD 4.5) days Maximum storage duration, mean: 15 (SD 6.5) days	Minimum storage duration, mean: 20.5 (SD 8.9) days Maximum storage duration, mean: 26 (SD 9.2) days
Bennett‐Guerrero 2009 [2]	Leucoreduced (AS‐1 or AS‐3)	Number of units transfused per participants, median: 5 (IQR 2 to 7)	Number of units transfused per participants, median: 4 (IQR 2 to 5)	Less than 21 days	Unit standard of care = RBC unit stored for longest out first	Mean: 11 (4) days	Mean: 16 (SD 5) days	Shortest stored RBC unit, mean: 10 (SD 5) days Maximum storage duration, mean: 12 (SD 4) days	Shortest stored RBC unit, mean: 13 (SD 4) days Maximum storage duration, mean: 18 (SD 7) days No RBC unit was stored for > 31 days
Cooper 2017	Leucoreduced (SAGM)	Number of units transfused per participant, median: 2 (IQR 1 to 4)	Number of units transfused per participant, median: 2 (IQR 1 to 4)	"Freshest available"	"Oldest available"	Mean: 11.8 (SD 5.3) days	Mean: 22.4 (SD 7.5) days	Median: 10.7 (IQR 8.3 to 14.1) days	Median: 21.4 (IQR 16.7 to 27.4) days
Hebert 2005	Leucoreduced (CPD‐2 and AS‐3)	Median number of RBC units transfused: 3 (IQR 2 to 5)	Median number of RBC units transfused: 2 (IQR 2 to 4)	Storage duration < 8 days	This group received RBCs issued from the hospital blood bank with the longest storage time in accordance with standard blood bank procedure. To ensure the maximum separation in terms of duration of storage between groups, participants were allocated only on days when average duration of storage of RBCs in the blood bank exceeded 15 days.	Median: 4 days	Median: 19 days	Overall 73% of participants received RBCs with storage times that correspond to treatment allocation more than 90% of the time. Compliance target of 90% was attained by 91% of patients allocated to shorter storage duration arm compared with 59% in the standard duration arm.	Overall 73% of participants received RBCs with storage times that correspond to treatment allocation more than 90% of the time. Compliance target of 90% was attained by 91% of participants allocated to shorter storage duration arm compared with 59% in standard duration arm.
Heddle 2012	Leucoreduced (SAGM)	Total number of RBC units transfused: 1157 Median: 2 (IQR 2 to 4) Range 1 to 102	Total number of RBC units transfused: 2369 Median: 2 (IQR 2 to 5) Range 1 to 38	“Freshest available”	“Oldest in the inventory”	Maximum storage duration, mean: 12.0 (SD 6.8) days	Maximum storage duration, mean: 26.6 (SD 7.8) days	Range 2 to 27 days and 33 to 37 days	Range 3 to 42 days
Heddle 2016	Leucoreduced (SAGM)	Total number of RBC units transfused: 25,466 Number of units transfused per participant median: 1 (IQR 1 to 2) Range 1 to 87	Total number of RBC units transfused: 50,890 Number of units transfused per participant, median: 1 (IQR 1 to 2) Range 1 to 58	"Freshest available"	"Oldest available"	Mean: 13.0 (SD 7.6) days	Mean: 23.6 (SD 8.9) days	Median: 11 (IQR 8 to 16) days	Median: 23 (IQR 16 to 31) days
Kor 2012	Leucoreduced (not stated)	1 unit	1 unit	Single unit of RBCs stored for < 5 days Subsequent transfusions (after the first study transfusion) were standard issue	Single unit of standard issue RBCs (median = 21 days)	Median: 4.0 (IQR 3.0 to 5.0) days	Median: 26.5 (IQR 21.0 to 36.0) days	‐‐	4 participants received an RBC unit that had been stored for ≤ 14 days; 9 participants received an RBC unit that had been stored between 15 and 21 days, and 37 participants received an RBC unit that had been stored for > 21 days.
Lacroix 2015	Leucoreduced (SAGM)	Total number of RBC units transfused: 5198	Total number of RBC units transfused: 5210	"Freshest available" < 8 days	"Standard issue"	Mean: 6.1 (SD 4.9) days	Mean: 22.0 (SD 8.4) days	Mean number of RBC units per participant who received at least 1 transfusion: 4.3 (SD 5.2)	Mean number of RBC units per participant who received at least 1 transfusion: 4.3 (SD 5.5)
^aType of RBCs relates to details of whether trials report that RBCs were leucoreduced or irradiated, or whether whole blood rather than RBCs were transfused. The nature of the blood that was transfused is recorded here. Abbreviations AS‐1: additive solution‐1, a commercial additive solution containing sodium chloride, dextrose, adenine, mannitol AS‐3: additive solution‐3, a commercial additive solution containing sodium chloride, dextrose, adenine, tri‐sodium citrate, citric acid, sodium phosphate AS‐5: additive solutionl‐5, a commercial additive solution containing sodium chloride, mannitol, adenine, dextrose CPDA‐1: citrate, phosphate, dextrose, adenine 1 CPD‐2: citrate, phosphate, dextrose‐2 CPD‐2‐AS‐3: citrate, phosphate, dextrose‐2 with additive solution‐3 IQR: interquartile range PRBC = packed red blood cells RBCs = red blood cells SAGM = saline adenine glucose mannitol SD: standard deviation

Table 1. Characteristics of the interventions

Table 2. Duration of organ support

Trial	Shorter storage duration	Longer storage duration/standard practice storage duration
Transfusion of RBCs of shorter vs longer storage duration
Respiratory support (invasive and non‐invasive ventilation)
Adult participant population
Bennett‐Guerrero 2009 [1]	Duration of mechanical ventilation, mean = 31 (SD 33) hours Percentage (number)^a of participants on any vasopressor > 48 hours after surgery = 17% (2)	Duration of mechanical ventilation, mean = 16 (SD 6) hours Percentage (number)^a of participants on any vasopressor > 48 hours after surgery = 25% (3)]
Transfusion of RBCs of shorter vs standard practice storage duration
Respiratory support (invasive and non‐invasive ventilation)
Adult participant population
Aubron 2012	Duration of mechanical ventilation, median 156 (IQR 6.1 to 253) days	Duration of mechanical ventilation, median 9.85 (IQR 0 to 198) days
Cooper 2017	Number of days alive and free of mechanical ventilation, median = 25 (IQR 11 to 28) days	Number of days alive and free of mechanical ventilation, median = 25 (IQR 13 to 28) days
Renal support (haemofiltration)
Adult participant population
Hebert 2005 Number of participants on dialysis over the 30 days of the study	0% (n = 0)	6% (n = 2)
Cooper 2017	Percentage (number)^a of participants requiring: renal replacement therapy = 13.9% (342) Days alive and free of renal replacement therapy, median: 28 (IQR 22 to 28) days	Percantage (number)^a of participants requiring: renal replacement therapy = 14.6% (360) Days alive and free of renal replacement therapy, median: 22 (IQR 22 to 28) days
Lacroix 2015	Duration of extrarenal epuration, mean: 2.5 (SD 10.1) days	Duration of extrarenal epuration, mean: 2.5 (SD 8.3) days
Haemodynamic support (vasopressors, inotropes)
Adult participant population
Lacroix 2015	Duration of cardiac or vasoactive drugs, mean: 7.1 (SD 10.2) days	Duration of cardiac or vasoactive drugs, mean: 7.5 (SD 11.2) days
^aAs the number of participants included in the analysis for this outcome was known, the number reported here was calculated for the purposes of this review. Abbreviations IQR: interquartile range SD: standard deviation

Table 2. Duration of organ support

Table 3. Length of stay: hospital and intensive care unit

Trial	Shorter storage duration	Longer storage duration/standard practice storage duration
Transfusion of RBCs of shorter vs longer storage duration
Neonate participant population
Fernandes 2005 Length of hospital stay: mean (SD)	60.8 (SD 37.3) days	62.6 (SD 48.3) days
Paediatric participant population
Dhabangi 2015 Length of hospital stay: median (IQR)	4 (IQR 2, 6) days*	4 (IQR 3 to 7) days*
Adult participant population
Bennett‐Guerrero 2009 [1] Length of hospital stay: mean (SD)	10 (SD 9) days	8.6 (SD 4) days
Steiner 2015 Length of hospital stay: median	8 days (IQR not reported)	8 days (IQR not reported)
Spadaro 2017 Length of hospital stay: median (IQR)	10 (IQR 6 to 17) days*	9 (IQR 7 to 17) days*
Bennett‐Guerrero 2009 [1] Length of ICU stay: mean (SD)	51 (SD 67) hours	51 (SD 56) hours
Steiner 2015 Length of ICU stay: median	3 days (IQR not reported)	3 days (IQR not reported)
Spadaro 2017 Length of ICU stay: median (IQR)	1 (IQR 1 to 6) days*	3(IQR 2 to 5) days*
Transfusion of RBCs of shorter vs standard practice storage duration
Neonate participant population
Fergusson 2012 Length of ICU stay: median (IQR)	84 (IQR 50 to 104) days*	77 (IQR 50 to 104) days*
Strauss 1996 Percentage discharge from hospital before 84 days	14%	10%
Adult participant population
Aubron 2012 Length of hospital stay: median (IQR)	21 (IQR 12 to 38) days	17 (IQR 8 to 27) days
Bennett‐Guerrero 2009 [2] Length of hospital stay: mean (SD)	10 (SD 7) days	13 (SD 14) days
Cooper 2017 Length of hospital stay: median (IQR)	14.5 (IQR 7.4 to 27.5) days	14.7 (IQR 7.4 to 28.3) days
Heddle 2016 Length of hospital stay: median (IQR)	10 (IQR 5 to 19) days	10 (IQR 5 to 20) days
Lacroix 2015 Length of hospital stay: mean (SD)	34.4 (SD 39.5) days	33.9 (SD 38.8) days
Aubron 2012 Length of ICU stay: median (IQR)	11 (IQR 5 to 15) days	7 (IQR 3 to 17) days
Bennett‐Guerrero 2009 [2] Length of ICU stay: mean (SD)	69 (SD 136) hours	47 (SD 51) hours
Cooper 2017 Length of ICU stay: median (IQR)	4.2 (IQR 2.0 to 9.3) days	4.2 (IQR 1.9 to 9.4) days
Lacroix 2015 Length of ICU stay: mean (SD)	15.3 (SD 15.4) days	15.3 (SD 14.8) days
Abbreviations IQR: interquartile range SD: standard deviation

Table 3. Length of stay: hospital and intensive care unit

Table 4. Assessment of economic and blood stock inventory

Trial	Shorter storage duration	Longer storage duration/standard practice storage duration
Transfusion of RBCs of shorter vs longer storage duration
Adult participant population
Yuruk 2013 Number of bags of RBCs transfused to adult haematology participants: median (IQR)	3 (IQR 2 to 3) bags	3 (IQR 3 to 3) bags
Steiner 2015 Number of RBC units transfused per participant to postoperative day 7: median (need IQR) Number of RBC units transfused per participant "throughout the study period': median (IQR)	3 units (IQR not reported) 4 (IQR 2 to 6) units	3 units (IQR not reported) 3 units (IQR 2 to 6)
Spadaro 2017 Number of RBC units transfused per participant during surgery: median (IQR)	2 (IQR 2 to 3) units	2 (IQR 2 to 4) units
Transfusion of shorter storage duration RBCvs transfusion of standard practice storage duration
Adult participant population
Heddle 2016 Number of RBC units transfused per participant: median (IQR)	2 (IQR 2 to 4) units	2 (IQR 2 to 4) units
Cooper 2017 Number of RBC units transfused per participant: median (IQR)	2 (IQR 1 to 4) units	2 (IQR 1 to 4) units
Abbreviations IQR: interquartile range

Table 4. Assessment of economic and blood stock inventory

Table 5. Physiological markers of oxygen consumption or alterations in microcirculation

Trial	Shorter storage duration	Longer storage duration/standard practice storage duration
Transfusion of RBCs of shorter vs longer storage duration
Neonatal participant population
Liu 1994	Mean change (before and after transfusion) in: potassium concentration (mmol/L) = 0.10 (SD 1.22) pH level = ‐0.02 (SD 0.08) base excess = ‐0.82 (SD 1.99)	Mean change (before and after transfusion) in: potassium concentration (mmol/L) = 0.10 (SD 1.22) pH level = ‐0.02 (SD 0.08) base excess = ‐0.82 (SD 1.99)
Paediatric participant population
Dhabangi 2013	Resolution of lactic acidosis within 4 hours of RBC transfusion = 92% (n = 34)	Resolution of lactic acidosis within 4 hours of RBC transfusion = 81% (n = 30)
Dhabangi 2015	Proportion of participants with a lactate of 3 mmol/L or lower at 8 hours after RBC transfusion = 58% (83/143) Cerebral tissue oxygen saturation (%) Beginning of transfusion: 72.8% End of transfusion: 77%	Proportion of participants with a lactate of 3 mmol/L or lower at 8 hours after RBC transfusion = 61% (87/143) Cerebral tissue oxygen saturation (%) Beginning of transfusion: 71.6% End of transfusion: 78.9%
Adult participant population
Bennett‐Guerrero 2009 [1]	PaO₂:FiO₂ ratio, mean at: Baseline = 420 (SD 158) 2 hours post operation = 256 (SD 73) Lactic acid, concentration 2 hours post operation: mean = 1.9 (SD 1.6)	PaO₂:FiO₂ ratio, mean at: Baseline = 386 (SD 128) 2 hours post‐operation = 360 (SD 64) Lactic acid, concentration 2 hours post operation: mean = 2.5 (SD 1.5)
Neuman 2013	"patients had a [mean (SD)] 0.191 (0.438) and 0.069 (0.242) µM increase in N0₂ at 1 and 24 hours after transfusion respectively" "Patients receiving fresh pRBC [PRBC] had no significant change in FMD" (i.e. flow‐mediated dilation, a measure of vascular endothelial function) No significant change in FHb (plasma‐free Hb) was noted before or after a RBC transfusion	"patients had a [mean (SD)] 0.191 (0.438) and 0.069 (0.242) µM increase in N0₂ at 1 and 24 hours after transfusion respectively" "Patients receiving fresh pRBC [PRBC] had no significant change in FMD" (flow‐mediated dilation, a measure of vascular endothelial function) No significant change in FHb (plasma‐free Hb) was noted before or after a RBC transfusion
Walsh 2004	Median (plus IQR) from baseline to post‐transfusion period Intragastric arterial difference in PCO₂ (Pg‐PaCO₂ gap) = ‐0.15 (IQR ‐0.31 to ‐0.01) PgCO₂,kPa = ‐0.46 (IQR ‐1.36 to ‐0.18) PgCO₂ ‐PaCO₂,kPa = ‐0.41 (IQR ‐0.82 to 0.10) pHi = 0.02 (IQR ‐0.01 to 0.05) Lactate = 0.10 (IQR ‐0.16 to 0.21) mmol/L Arterial pH = 0.00 (IQR ‐0.001 to 0.001) HCO₃ (actual) = 0.85 (IQR ‐1.44 to ‐0.53) mmol/L Arterial base excess = ‐0.70 (IQR ‐1.42 to ‐0.45)	Median (plus IQR) from baseline to post‐transfusion period Intragastric arterial difference in PCO2 (Pg‐PaCO2 gap) = 0.02 (IQR ‐0.29 to 0.11) PgCO₂,kPa = 0.16 (IQR ‐0.36 to 1.19) PgCO₂ ‐PaCO₂,kPa = 0.37 (IQR ‐0.17 to 1.66) pHi = ‐0.02 (IQR ‐0.06 to 0.01) Lactate = 0.03 (IQR ‐0.10 to 0.23) mmol/L Arterial pH = 0.00 (IQR ‐0.001 to 0.001) HCO₃ (actual) = ‐0.29 (IQR ‐0.90 to 0.07) mmol/L Arterial base excess = ‐0.71 (IQR ‐1.5 to ‐0.42)
Yuruk 2013	Mean changes in microcirculatory density (measured before and 30 minutes after transfusion) from 17 (SD 3) mm/mm² to 19 (SD 3) mm/mm².	Mean changes in microcirculatory density (measured before and 30 minutes after transfusion) from 17 (SD 3) mm/mm² to 19 (SD 2) mm/mm².
Transfusion of shorter storage duration RBCs vs transfusion of standard practice storage durationRBCs
Neonatal participant population
Fergusson 2012	Percentage (number) of participants requiring: supplemental oxygen = 96.8% (182) nasal continuous positive airway pressure = 79.3% (149)	Percentage (number) of participants requiring: supplemental oxygen = 94.7% (179) nasal continuous positive airway pressure = 80.4% (152)
Strauss 1996	Mean change from before to after transfusion: pH = 0.00 (SD 0.05) Sodium = 0.5 (SD 4.4) mEq/L Potassium = 0.3 (SD 0.9) mEq/L Calcium = ‐0.1 (SD 0.7) mg/dL Lactate = ‐0.3 (SD 0.4) mmol/L Glucose = ‐9 (SD 32) mg/dL	Mean change from before to after transfusion: pH = 0.00 (SD 0.13) Sodium = 0.6 (SD 5.1) mEq/L Potassium = ‐0.2 (SD 1.5) mEq/L Calcium = ‐0.1 (SD 0.5) mg/dL Lactate = ‐0.1 (SD 0.6) mmol/L Glucose = ‐7 (SD 25) mg/dL
Strauss 2000	Mean change from before to after transfusion: pH = ‐0.01 (SD 0.9) Sodium = ‐0.4 (SD 5.5) mEq/L Potassium = 0.26 (SD 1.16) mEq/L Calcium = ‐0.13 (SD 1.04) mg/dL Lactate = ‐0.16 (SD 0.30) mmol/L Glucose = ‐17.5 (SD 40.3) mg/dL	Mean change from before to after transfusion: pH = ‐0.03 (SD 0.6) Sodium = 1.9 (SD 4.7) mEq/L Potassium = 0.38 (SD 0.91) mEq/L Calcium = 0.13 (SD 0.64) mg/dL Lactate = ‐0.53 (SD 1.46) mmol/L Glucose = ‐16.7 (SD 24.4) mg/dL
Adult participant population
Bennett‐Guerrero 2009 [2]	PaO₂:FiO₂ ratio, mean at: Baseline = 414 (SD 97) 2 hours post operation = 267 (SD 81) Lactic acid, concentration 2 hours post operation: mean = 2.2 (SD 1.1)	PaO₂:FiO₂ ratio, mean at Baseline = 284 (SD 122) 2 hours post operation = 212 (SD 87) Lactic acid, concentration 2 hours post operation: mean = 2.4 (SD 1.0)
Kor 2012	Mean change from before to after transfusion PaO₂:FiO₂ = 2.5 (SD 49.3) mmHg	Mean change from before to after transfusion PaO₂:FiO₂ = ‐9.0 (SD 69.8) mmHg
Abbreviations Hb: haemoglobin IQR: interquartile range Pg‐PaCO₂: intragastric minus arterial partial pressure of carbon dioxide PgCO₂,kPa; intragastric partial pressure of carbon dioxide pHi: intramucosal gastric pH PRBC: packed red blood cells RBC: red blood cell SD: standard deviation

Table 5. Physiological markers of oxygen consumption or alterations in microcirculation

Comparison 1. Transfusion of shorter storage duration RBCs vs longer storage duration RBCs

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality Show forest plot	7		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

1.1 In‐hospital mortality: within 24 hours (children)	2	364	Risk Ratio (M‐H, Random, 95% CI)	1.50 [0.43, 5.25]
1.2 In‐hospital: within 7 days (children)	1	74	Risk Ratio (M‐H, Random, 95% CI)	3.00 [0.13, 71.34]
1.3 In‐hospital mortality: within 7 days (adults)	1	1098	Risk Ratio (M‐H, Random, 95% CI)	1.42 [0.66, 3.06]
1.4 Mortality: time‐point not defined (adults)	1	17	Risk Ratio (M‐H, Random, 95% CI)	2.25 [0.55, 9.17]
1.5 Short‐term mortality: up to 30 days (children)	1	290	Risk Ratio (M‐H, Random, 95% CI)	1.40 [0.45, 4.31]
1.6 Short‐term mortality: up to 30 days (adults)	2	1121	Risk Ratio (M‐H, Random, 95% CI)	0.85 [0.50, 1.45]
1.7 Long‐term mortality: > 30 days (neonates)	1	52	Risk Ratio (M‐H, Random, 95% CI)	0.9 [0.44, 1.85]
1.8 Long‐term mortality: > 30 days (adults)	1	199	Risk Ratio (M‐H, Random, 95% CI)	1.58 [0.68, 3.64]
2 Incidence of hospital‐acquired infection in neonates Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

2.1 Number of clinical sepsis events	1	52	Risk Ratio (M‐H, Random, 95% CI)	1.25 [1.00, 1.56]
2.2 Number of necrotising enterocolitis events	1	52	Risk Ratio (M‐H, Random, 95% CI)	1.5 [0.48, 4.70]
3 Incidence of hospital‐acquired infection in an adults Show forest plot	2		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

3.1 Number of participants developing infections and infestations (as classified under serious advere events) to day 28	1	1098	Risk Ratio (M‐H, Random, 95% CI)	0.89 [0.60, 1.32]
3.2 Number of participants developing at least 1 postoperative infection by day 28	1	199	Risk Ratio (M‐H, Random, 95% CI)	0.85 [0.52, 1.41]
3.3 Number of participants developing sepsis by day 28	1	199	Risk Ratio (M‐H, Random, 95% CI)	0.73 [0.26, 2.02]
3.4 Number of participants developing a pulmonary infection by day 28	1	199	Risk Ratio (M‐H, Random, 95% CI)	2.26 [0.60, 8.51]
3.5 Number of participants developing a wound infection by day 28	1	199	Risk Ratio (M‐H, Random, 95% CI)	0.32 [0.12, 0.86]
3.6 Number of participants developing peritonitis by day 28	1	199	Risk Ratio (M‐H, Random, 95% CI)	1.94 [0.36, 10.35]
3.7 Number of participants develoiping a urinary tract infection by day 28	1	199	Risk Ratio (M‐H, Random, 95% CI)	1.94 [0.18, 21.06]
4 Duration of organ support Show forest plot	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

4.1 Duration of mechanical ventialation (neonates)	1		Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
5 Number of participants requiring organ support (adults) Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

5.1 Number of adult participants requiring renal support up to day 28 post surgery	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
5.2 Number of patients requiring inotropic drug support up to day 28 post surgery	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
6 Economic or blood stock inventory outcomes Show forest plot	2		Mean Difference (IV, Random, 95% CI)	Totals not selected

6.1 Mean donor exposure (neonate participant population)	2		Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]

Comparison 1. Transfusion of shorter storage duration RBCs vs longer storage duration RBCs

Comparison 2. Transfusion of shorter storage duration RBCs vs standard practice storage duration RBCs

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality Show forest plot	10		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

1.1 In‐hospital mortality: time‐points varied (adults)	4	25754	Risk Ratio (M‐H, Random, 95% CI)	1.05 [0.97, 1.14]
1.2 In‐ICU mortality (adults)	3	13066	Risk Ratio (M‐H, Random, 95% CI)	1.06 [0.98, 1.15]
1.3 Short‐term mortality: up to 30 days (neonates)	1	40	Risk Ratio (M‐H, Random, 95% CI)	0.30 [0.01, 7.02]
1.4 Short‐term mortality: up to 30 days (adults)	5	7530	Risk Ratio (M‐H, Random, 95% CI)	1.04 [0.96, 1.13]
1.5 Long‐term mortality: up to 90 days (neonates)	1	377	Risk Ratio (M‐H, Random, 95% CI)	0.97 [0.61, 1.54]
1.6 Long‐term mortality: at 90 days (adults)	3	7398	Risk Ratio (M‐H, Random, 95% CI)	1.04 [0.97, 1.12]
2 Incidence of hospital‐acquired infection in a neonates Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

2.1 Necrotising enterocolitis (Bell criteria ≥ 2)	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
2.2 Clinically suspected infection	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
2.3 Confirmed infection	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
3 Incidence of hospital‐acquired infection in adults Show forest plot	2		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

3.1 Hospital‐acquired infections (adults)	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
3.2 New blood stream infection whilst in ICU (adults)	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
4 Number of participants requiring organ support Show forest plot	5		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

4.1 Number of neonates requiring mechanical ventilation	1	377	Risk Ratio (M‐H, Random, 95% CI)	0.99 [0.90, 1.10]
4.2 Number of neonates requiring high‐frequency ventilation	1	377	Risk Ratio (M‐H, Random, 95% CI)	1.02 [0.79, 1.31]
4.3 Number of adults requiring mechanical ventilation	3	5027	Risk Ratio (M‐H, Random, 95% CI)	1.15 [0.92, 1.44]
4.4 Number of adults requiring mechanical ventilation for > 48 hours	1	57	Risk Ratio (M‐H, Random, 95% CI)	1.34 [0.60, 2.98]
4.5 Number of adults requiring prolonged invasive ventilation for > 48 hours	1	57	Risk Ratio (M‐H, Random, 95% CI)	1.67 [0.60, 4.64]
4.6 Number of adults requiring vasoreactive drugs or an aortic balloon pump or a ventricular assist device for > 48 hours	1	57	Risk Ratio (M‐H, Random, 95% CI)	1.49 [0.45, 4.98]
4.7 Number of adults on any vasopressor > 48 hours after surgery	1	20	Risk Ratio (M‐H, Random, 95% CI)	3.67 [0.46, 29.49]
4.8 Number of adults requiring renal support	2	4976	Risk Ratio (M‐H, Random, 95% CI)	0.95 [0.83, 1.09]
5 Duration of organ support Show forest plot	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

5.1 Duration (days) of mechanical ventilation (adults)	1		Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
5.2 Duration (days) of cardiac or vasoreactive drugs (adults)	1		Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
5.3 Duration (days) of renal support (adults)	1		Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
6 Number of participants experiencing an adverse transfusion reaction Show forest plot	2		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

6.1 Adults	2		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
7 Economic or blood stock inventory outcomes Show forest plot	4		Mean Difference (IV, Random, 95% CI)	Totals not selected

7.1 Mean donor exposure (neonates)	3		Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
7.2 Mean number of RBC units transfused per participant (adults)	1		Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]

Comparison 2. Transfusion of shorter storage duration RBCs vs standard practice storage duration RBCs