Publication

• Title: Prehospital Whole Blood in Traumatic Hemorrhage — a Randomized Controlled Trial
• Acronym: SWiFT
• Year: 2026
• Journal published in: New England Journal of Medicine
• Citation: Smith JE, Cardigan R, Sanderson E, et al. Prehospital whole blood in traumatic hemorrhage—a randomized controlled trial. N Engl J Med. 2026 Mar 17.

Context & Rationale

• Background

  Major traumatic haemorrhage remains a leading preventable cause of early trauma death.
  Prehospital resuscitation had already shifted away from crystalloid-heavy care towards earlier blood-based haemostatic resuscitation.
  Whole blood offered an intuitively attractive package: red cells, plasma, and platelets in one bag, with simpler logistics, less donor exposure, and potentially faster delivery in austere prehospital conditions.
  Before SWiFT, however, civilian evidence was dominated by observational studies and small-trial data, and in England the more immediate comparator question was not red cells versus blood products, but whether whole blood was superior to an already balanced red-cell-plus-plasma strategy. ¹²
• Research Question/Hypothesis

  In patients with life-threatening traumatic haemorrhage requiring prehospital transfusion, does transfusion of up to 2 units of leukocyte-depleted whole blood reduce 24-hour death or massive transfusion, as compared with standard care using red cells plus plasma? ¹
• Why This Matters

  The answer had immediate service-level implications for UK civilian trauma systems because whole blood requires a scarce donor pool, different manufacturing pathways, and careful stewardship to avoid wastage.
  At the same time, practice guidance and position statements were already becoming more favourable to whole blood, despite the absence of a definitive civilian prehospital superiority trial. ¹³⁴

Design & Methods

• Research Question: Was prehospital whole-blood transfusion superior to balanced component therapy in reducing the composite of death from any cause or massive transfusion within 24 hours in traumatic haemorrhage?
• Study Type: Pragmatic, phase 3, multicentre, open-label, randomised, controlled, superiority trial conducted in prehospital emergency medicine across 10 air ambulance services in England, with patients received by 19 hospitals; enrolment occurred under emergency waiver of consent.
• Population:
  • Patients of any age with traumatic injury who were attended by a participating air ambulance service clinical team and judged to require prehospital blood transfusion for major traumatic haemorrhage.
  • Key exclusions: no intravenous or intraosseous access, known objection to blood transfusion, or blood already administered before arrival of the participating air ambulance service.
  • Pragmatic field enrolment allowed non-traumatic haemorrhage and traumatic cardiac arrest cases to be entered if a trial box was opened; these groups were prespecified for exclusion from the primary modified intention-to-treat analysis.
• Intervention:
  • Up to 2 units of leukocyte-depleted, platelet-replete whole blood before hospital arrival.
  • The UK product was group O, RhD negative, low-titre anti-A and anti-B, and Kell negative; total volume per unit was approximately 470 mL, stored at 2–6°C for up to 21 days.
  • Administration could be intravenous or intraosseous; additional blood components after hospital arrival were given according to routine care.
• Comparison:
  • Up to 2 units of red cells and 2 units of plasma before hospital arrival.
  • Plasma was thawed fresh-frozen plasma or lyophilised plasma, according to local air ambulance practice.
  • No other aspects of prehospital transport care or in-hospital haemorrhage management were protocolised.
• Blinding: Allocation was concealed at the level of identical sealed trial boxes prepared by transfusion laboratories; air ambulance teams were blinded until the box was opened, after which bedside blinding was not feasible. Receiving hospitals were unblinded. The primary endpoint components were objective.
• Statistics: A total of 602 analysable participants were required to detect a 12% absolute reduction in death or massive transfusion within 24 hours (from 68% to 56%) with 85% power at a two-sided 5% significance level, with one interim analysis using O’Brien–Fleming boundaries. After inflating for approximately 25% traumatic cardiac arrest exclusions and 5% dropout, the planned total sample size was 848; after a blinded reassessment it was increased to 900. The primary analysis was modified intention-to-treat; a per-protocol primary analysis was prespecified.
• Follow-Up Period: 24 hours for the primary endpoint, with additional outcomes through 30 and 90 days.

Key Results

This trial was not stopped early. One planned interim analysis was performed after 400 non-traumatic-cardiac-arrest participants, and a blinded sample-size reassessment increased the target sample from 848 to 900; the trial then continued to completion. Overall, 942 patients were randomised, 641 comprised the modified intention-to-treat population, and primary-outcome data were available for 616 participants.

• The primary endpoint was neutral: whole blood was not superior to standard red-cell-plus-plasma care for the composite of 24-hour death or massive transfusion.
• No mortality signal emerged at 24 hours, 30 days, or 90 days; if anything, point estimates were numerically but not statistically against whole blood at later time points.
• Prespecified and exploratory subgroup analyses did not reveal a convincing effect-modification signal. By injury type, the primary-outcome odds ratio was 1.06 (95% CI 0.54 to 2.05) in blunt trauma and 1.09 (95% CI 0.58 to 2.03) in penetrating trauma. Confidence intervals for secondary outcomes were not adjusted for multiplicity.

Internal Validity

• Randomisation and Allocation: Allocation concealment was good. Randomisation was central, web-based, stratified by air ambulance service, and implemented through identical sealed trial boxes packed by transfusion laboratories.
• Drop out or exclusions: This is the major internal-validity challenge. Of 942 randomised participants, only 641 entered the modified intention-to-treat analysis. Post-randomisation exclusions were substantial and included traumatic cardiac arrest on air ambulance arrival (225 participants), non-traumatic haemorrhage (33, plus 1 participant with both non-trauma and cardiac arrest), missing cardiac-arrest status (4), and data withdrawal (38). A serious consent-process breach led to deletion of data from 21 participants.
• Performance/Detection Bias: The trial was open-label after the box was opened. The primary endpoint was objective, which limits classic ascertainment bias, but the massive-transfusion component can still be influenced by downstream clinician behaviour in an unblinded environment.
• Protocol Adherence: Adherence was reasonably good for a prehospital trauma trial. Protocol deviations occurred in 6.9% of the whole-blood group and 6.8% of the standard-care group. Randomisation errors occurred in 5.1% and 5.7%, respectively. Correct box contents were not administered in 0.9% of whole-blood participants and 1.0% of standard-care participants.
• Baseline Characteristics: Groups were well matched. Median Injury Severity Score was 33 (IQR 18 to 48) with whole blood and 34 (IQR 18 to 45) with standard care. Severe head injury (AIS ≥3) was present in 42.0% versus 46.9%, median systolic blood pressure was 95 versus 96 mm Hg, and median heart rate was 110 versus 110 beats/min.
• Heterogeneity: Clinical heterogeneity was real. The trial included adults and children, blunt and penetrating trauma, and depended on clinician judgement for transfusion initiation. That pragmatism increases relevance but dilutes biological specificity. Missing injury-registry data were also important: Injury Severity Score was missing in 31.4% and AIS head/neck/c-spine data in 32.8%, which prevented formal statistical subgroup analyses for some prespecified strata.
• Timing: Timing was appropriate for a prehospital haemorrhage question, but the window for treatment separation was short. Median total prehospital time from air ambulance arrival to emergency department arrival was 68 minutes with standard care and 67 minutes with whole blood; median transport time was 29 versus 28 minutes.
• Dose: The intervention ceiling was modest: up to 2 units of whole blood. This may have been too small an exposure to yield a detectable clinical advantage before in-hospital haemorrhage control and further transfusion diluted group separation.
• Separation of the Variable of Interest: Separation was incomplete by design. Whole-blood units were younger than control red cells (median 13.5 vs 18 days), but 220 participants in the whole-blood group subsequently received red cells and 191 received plasma within 24 hours. In addition, 7 standard-care participants received whole blood prehospital because of mispacked boxes or stock limitations.
• Key Delivery Aspects: The comparator was not weak usual care but an already balanced red-cell-plus-plasma strategy. That makes the question clinically sharp, but it also means SWiFT tested incremental superiority over a mature haemostatic approach rather than over red cells alone or crystalloid-heavy care.
• Crossover: Crossover was uncommon but relevant. Before hospitalisation, 7 participants randomised to standard care received whole blood. This would, if anything, bias toward the null.
• Adjunctive therapy use: Adjunctive haemostatic therapy was similar across groups. Tranexamic acid was given to 97.2% with whole blood and 98.1% with standard care; calcium to 73.1% versus 75.1%; fibrinogen concentrate to 6.0% versus 5.7%; and prothrombin complex concentrate to 1.6% versus 1.7%.
• Outcome Assessment: The primary endpoint was clinically important and objective. Follow-up for the primary outcome was high, with primary-outcome data available for 616 of 641 modified intention-to-treat participants (96.1%).
• Statistical Rigor: The analysis broadly followed the protocol. The study was powered for a large 12% absolute treatment effect, not for smaller but potentially meaningful differences. The blinded sample-size increase to 900 was appropriate given higher-than-expected attrition, and sensitivity analyses for missing primary data did not materially alter the conclusion.

Conclusion on Internal Validity: Internal validity is moderate. Allocation and follow-up for the primary endpoint were good, but extensive post-randomisation exclusions, open-label care after box opening, and limited treatment separation materially constrain confidence that a smaller true treatment effect would have been detected.

External Validity

• Population Representativeness: The cohort was representative of severely injured civilian patients selected for physician-led prehospital transfusion in a mature trauma system. Most were male (75.5%), most had blunt trauma (71.3%), and median Injury Severity Score was approximately 33 to 34.
• Applicability to Similar Systems: Applicability is strongest for UK-style or comparable advanced prehospital systems with physician-paramedic teams, established blood-product logistics, and access to both whole blood and balanced component therapy.
• Applicability to Dissimilar Systems: Generalisability is weaker in systems using red cells alone, no plasma, no physician response model, or substantially longer rural transport times, because SWiFT compared whole blood against a strong and balanced comparator.
• Important Exclusions: The main analytic findings do not apply to non-traumatic haemorrhage or to patients in traumatic cardiac arrest on air ambulance arrival, both of whom were excluded from the primary modified intention-to-treat analysis.
• Paediatric Generalisability: Paediatric inference is limited. Only 21 paediatric participants were included in the modified intention-to-treat population, so meaningful age-specific conclusions cannot be drawn.
• Resource Requirements: Whole-blood implementation depends on a specific donor base, cold-chain management, transfusion-laboratory support, traceability, and mechanisms for recycling near-expiry stock. These constraints will limit direct transferability to less mature networks.

Conclusion on External Validity: External validity is good for mature civilian trauma systems already capable of prehospital blood-product resuscitation, but more limited for resource-constrained settings, non-traumatic major haemorrhage, traumatic cardiac arrest, and paediatric practice.

Strengths & Limitations

• Strengths: Largest randomised civilian prehospital whole-blood trial to date.
  Pragmatic delivery across 10 air ambulance services and 19 hospitals.
  Clinically important, objective primary endpoint.
  Robust allocation concealment before box opening.
  Independent data and safety monitoring and trial steering oversight.
  Meaningful safety dataset in real-world UK whole-blood logistics.
• Limitations: Open-label treatment after box opening.
  Extensive post-randomisation exclusions from the primary analytic population.
  Composite endpoint mixes mortality with a transfusion-threshold outcome that depends partly on trial accounting conventions.
  Whole-blood exposure capped at 2 units.
  Comparator was already balanced and strong.
  Secondary outcomes were not multiplicity-adjusted and the trial was not powered for smaller effect sizes.
  Missing registry-derived injury data, worsened by the 2023 cyberattack on the trauma database, limited subgroup inference.

Interpretation & Why It Matters

• Clinical signal

  SWiFT does not support superiority of up to 2 prehospital whole-blood units over balanced red-cell-plus-plasma transfusion for life-threatening traumatic haemorrhage.
• What it refutes

  The trial challenges the assumption that the logistical elegance and physiological completeness of whole blood necessarily translate into better hard outcomes when the alternative is already balanced component therapy. ³⁴
• What it confirms

  Whole blood appears feasible and acceptably safe in a civilian prehospital network, but the justification for adoption becomes logistical and operational rather than superiority-based.
• Practical implication

  For mature systems already delivering red cells and plasma prehospital, SWiFT supports equipoise rather than wholesale replacement of component therapy by whole blood. The next question is likely patient selection, not universal substitution. ⁶⁷

Controversies & Other Evidence

• The first methodological controversy is the analytic distance from randomisation. Trial entry occurred when the box was opened, yet 301 of 942 randomised participants were excluded from the modified intention-to-treat analysis, mainly because they were in traumatic cardiac arrest on air ambulance arrival or had non-traumatic haemorrhage. That preserves the prespecified target population but weakens the simplicity of an all-randomised estimate.
• The second controversy is the composite endpoint itself. SWiFT combined death with massive transfusion and used a trial-specific convention in which 2 units of whole blood were counted as equivalent to 4 component units. That approach is physiologically arguable, but it means the massive-transfusion component is partly shaped by design assumptions rather than raw bag count alone.
• The third controversy is dose and comparator strength. SWiFT tested a maximum of 2 prehospital whole-blood units against an already balanced red-cell-plus-plasma strategy, not against red cells alone or crystalloids. A neutral result in this context does not mean whole blood is ineffective in all settings; it means superiority over a mature balanced-component programme was not shown.
• The pre-SWiFT guideline landscape was already moving towards whole blood. The 2023 European trauma bleeding guideline discussed prehospital blood-product use but stopped short of recommending for or against it, reflecting continuing uncertainty. EAST nevertheless conditionally recommended whole blood for adult civilian trauma patients receiving transfusion, and the NAEMSP position statement later recommended low-titre group O whole blood as the first-choice prehospital product in systems able to support a high-quality blood programme. ⁵³⁴
• More recent pooled evidence still tends to favour whole blood, but it remains heterogeneous. An updated 2026 JAMA Surgery meta-analysis found lower 24-hour and 30-day mortality in civilian studies of whole blood versus component therapy, yet the prediction intervals were wide and between-study heterogeneity was substantial. An invited commentary accompanying that analysis argued that the field may need better patient selection and more personalised resuscitation rather than a universal one-product strategy. SWiFT therefore remains pivotal because it is a large neutral randomised comparison against a balanced prehospital component regimen. ⁶⁷

Summary

• SWiFT randomised 942 patients in UK civilian prehospital trauma care and analysed 641 in a modified intention-to-treat population.
• Prehospital transfusion of up to 2 whole-blood units was not superior to red cells plus plasma for the composite of 24-hour death or massive transfusion: 48.7% versus 47.7%; RR 1.02; 95% CI 0.80 to 1.31; P=0.84.
• No mortality benefit emerged at 24 hours, 30 days, or 90 days, and there was no clear safety advantage or penalty for whole blood in major clinical outcomes.
• Whole blood was associated with more prolonged prothrombin time on hospital arrival, but not with more thrombotic events or more serious adverse events.
• The key clinical message is not that whole blood is unsafe or useless, but that superiority over a strong balanced-component prehospital strategy was not shown in this trial.

Overall Takeaway

SWiFT is a landmark negative trauma transfusion trial. It substantially narrows the claim that prehospital whole blood should be expected to outperform balanced red-cell-plus-plasma care in civilian traumatic haemorrhage, while still supporting the feasibility and apparent safety of whole-blood delivery in mature prehospital systems. Its enduring importance is methodological and practical: it shifts the field from enthusiasm based mainly on observational data towards a more conditional, selection-focused, system-specific use of whole blood.

Bibliography

• 2023Smith JE, Barnard EBG, Brown-O’Sullivan C, et al. The SWiFT trial (Study of Whole Blood in Frontline Trauma)—the clinical and cost effectiveness of pre-hospital whole blood versus standard care in patients with life-threatening traumatic haemorrhage: study protocol for a multi-centre randomised controlled trial. Trials. 2023;24:725.
• 2023Tucker H, Brohi K, Tan J, et al. Association of red blood cells and plasma transfusion versus red blood cell transfusion only with survival for treatment of major traumatic hemorrhage in prehospital setting in England: a multicenter study. Crit Care. 2023;27:25.
• 2024Meizoso JP, Cotton BA, Lawless RA, et al. Whole blood resuscitation for injured patients requiring transfusion: a systematic review, meta-analysis, and practice management guideline from the Eastern Association for the Surgery of Trauma. J Trauma Acute Care Surg. 2024;97(3):460-470.
• 2025Brown JB, Yazer MH, Kelly J, et al. Prehospital Trauma Compendium: Transfusion of Blood Products in Trauma - A Position Statement and Resource Document of NAEMSP. Prehosp Emerg Care. 2025 Apr 1:1-10.
• 2023Rossaint R, Afshari A, Bouillon B, et al. The European guideline on management of major bleeding and coagulopathy following trauma: sixth edition. Crit Care. 2023;27:80.
• 2026Ibrahim W, Meza Monge K, Menzel J, et al. Whole-Blood vs Component Therapy in Adult Trauma: An Updated Systematic Review and Meta-Analysis. JAMA Surg. 2026 Mar 11:e260197.
• 2026Wallen TE, Holcomb JB. From Balanced Blood Products to Personalized Medicine in Acute Resuscitation of Trauma Patients. JAMA Surg. Published online March 11, 2026.