What the Evidence Can Actually Tell Us
For civilian prehospital blood transfusion programs, the central evidence question is not simply whether blood “works.” It is what a particular study can support about a particular product, comparator, patient group, time point, and EMS system. This chapter builds the reading discipline needed to make that judgment. The payoff is practical: a program can respond to an urgent clinical need without turning a limited association into a promise of lower mortality.
A result is only as clear as the question around it
The AHRQ review examined prehospital hemorrhagic-shock resuscitation as a set of related questions. It considered whole blood, packed red blood cells (PRBC), plasma, crystalloid fluids, and combinations of treatments. It also examined barriers and facilitators to implementation and identified research needed to close the evidence gaps. That scope matters because “the evidence on prehospital blood” is not one body of observations answering one question. A study of whole blood versus no prehospital transfusion is not interchangeable with a study of PRBC versus plasma, and neither is the same as a study of how a blood program maintains its cold chain.
The review separated the clinical effectiveness questions from the implementation question. For the main product and strategy comparisons, it included randomized controlled trials and comparative nonrandomized studies. A randomized trial assigns treatment according to a study plan. A comparative nonrandomized study observes what happened when groups received different interventions, often in routine care. The latter can be valuable in a rare, time-critical setting, but the patients receiving blood may differ from those who did not: they may be more severely injured, treated by a different service, transported by air, or located closer to a capable trauma center. Statistical adjustment can reduce some differences; it cannot guarantee that every relevant difference was measured.
The review also drew boundaries around the setting. It focused on treatment delivered by EMS clinicians before hospital care, rather than assuming that results from an emergency department transfer directly to an ambulance or helicopter. It limited follow-up to 30 days because later outcomes are increasingly shaped by surgery, intensive care, complications, rehabilitation, and other factors after the prehospital intervention. It also excluded studies with more than a small proportion of interfacility transfers, because those patients may already have received products that cannot be accounted for. These choices make the question narrower, but they make the answer more relevant to the program being studied.
A useful way to read the evidence is to separate four kinds of outcome. Patient-centered outcomes include survival or mortality at hospital arrival, 24 hours, discharge, or up to 30 days, as well as morbidity after discharge. These are closest to the question families and policymakers care about: did patients live, and what happened to their function or recovery? Physiologic outcomes include blood pressure, heart rate, level of consciousness, lactate, and measures of clotting. They can reveal whether a patient’s condition changed during the short prehospital window, but a better blood pressure or lactate value is not itself proof of longer survival. Process outcomes include time to treatment, amount transfused, and amount of crystalloid infused. They tell us how a program behaves. Harms include transfusion reactions, coagulopathy, acute respiratory distress syndrome, infection, thrombotic events, and other complications.
The implementation and future-research questions add still another layer. A program may have a workable supplier relationship, trained clinicians, validated storage, and a reliable handoff even when the comparative mortality evidence is uncertain. Conversely, a product with a plausible physiologic rationale may not be useful if it cannot be kept cold, administered safely, or replaced without unacceptable waste. Treating implementation evidence as if it were a mortality result is a category error. Treating it as irrelevant because it is not a randomized outcome estimate is another one.
The review’s strength-of-evidence framework is intended to prevent those errors. It considered risk of bias, directness, consistency, precision, and reporting bias. Risk of bias asks whether design or missing data could systematically distort a result. Directness asks whether the population, intervention, comparison, and outcome match the question of interest. Consistency asks whether studies point in roughly the same direction. Precision asks how tightly the estimate is bounded, often reflected in the width of a 95% confidence interval. Reporting bias concerns whether the available publications may give an incomplete picture of the outcomes that were measured.
The labels then carry different meanings. High strength of evidence means the estimate is likely close to the true effect and stable. Moderate means the estimate is reasonably reliable but some doubt remains. Low means the evidence has major deficiencies and additional research could change the conclusion. Insufficient means an effect cannot be estimated with confidence because evidence is absent or too defective. That last label does not mean that the intervention has no effect. It means the evidence has not earned a dependable answer.
The difference becomes visible in a pooled estimate. The review found no difference in PRBC and usual care for mortality within 24 hours, with a pooled risk ratio of 0.86 and a 95% confidence interval from 0.49 to 1.49. A risk ratio of 1 would mean equal observed risk; the interval here remains compatible with a meaningful reduction or increase. Heterogeneity was also high, meaning the studies did not behave as though they were estimating one simple common effect. Calling that result “blood has no benefit” would overstate the data. Calling it “blood reduces mortality” would do the same in the other direction. The defensible statement is that the available studies did not demonstrate a reliable difference, and the uncertainty remains clinically important.
The diagram is a reading sequence, not a claim that every study must contain every step. It shows why the path from mechanism to policy has gates. A product can make physiologic sense, change a measured vital sign, and still lack evidence of a mortality advantage. A program can be operationally ready, yet need to describe its decision as conditional because its patients or setting differ from the published studies.
Why the review does not produce a simple ranking
Across the clinical questions, the review included 118 studies and more than 40,000 patients, but it did not find one blood or fluid intervention that was consistently superior. That conclusion is easy to misread because it compresses several different patterns. Some comparisons showed no clear difference with low strength of evidence. Some had insufficient evidence because studies were too few or too imprecise. Some individual studies suggested benefit while others pointed toward harm. The shared conclusion is not that all treatments are interchangeable. It is that the evidence base cannot reliably rank them across patients and EMS systems.
Whole blood illustrates the problem. The relevant clinical evidence came from four retrospective nonrandomized studies. Two studies comparing whole blood with no prehospital transfusion or crystalloid suggested lower mortality rates, but their estimates were imprecise and not statistically clear. Two studies comparing whole blood with blood products or component therapy pointed toward higher mortality, also without a statistically reliable difference. The groups differed at baseline, and two studies used potentially overlapping registry populations. The review therefore rated the strength of evidence for mortality as insufficient. The result leaves the physiologic and operational rationale for whole blood intact, but it does not establish that whole blood is better than components in civilian prehospital care.
PRBC has a larger set of comparative observations, but more observations do not automatically remove confounding. Eight nonrandomized studies found no significant mortality difference versus usual care or active interventions. The review rated some discharge comparisons at moderate strength and most 24-hour and 30-day comparisons at low strength. Results varied with the comparator, treatment combination, transport setting, and timing of measurement. Harms and morbidity were reported inconsistently; no eligible studies reported morbidity outcomes in the main PRBC question. The correct summary is therefore “no consistent mortality advantage demonstrated,” not “PRBC is equivalent to every alternative” or “PRBC does not work.”
Plasma shows why a promising subgroup can coexist with a cautious overall conclusion. Two randomized trials comparing plasma with crystalloid found no clear mortality effect, with wide confidence intervals. Secondary analyses of a larger plasma trial suggested improved survival in selected groups, including some patients transported directly from the scene or meeting particular hemorrhage thresholds. Those analyses can be informative, but they were not all randomized comparisons preserved in their original balance, and several publications shared patient populations. The review rated plasma versus usual care as low strength for mortality and plasma versus active interventions as insufficient. A signal in a selected analysis should invite better testing, not become a universal claim about plasma.
Crystalloid comparisons are even more heterogeneous. Low-volume strategies sometimes appeared to favor lower 24-hour mortality, but pooled evidence was insufficient and 30-day mortality showed no clear effect. Comparisons with PRBC showed no difference in pooled 30-day mortality with low strength of evidence; 24-hour results were inconsistent. Comparisons with plasma were also inconclusive. Several studies linked higher crystalloid volumes with greater later blood-product requirements, and some reported associations with acute respiratory distress syndrome or mortality. Other studies found no difference. Those observations support concern about dilution and volume, especially when blood is unavailable, but they do not turn one threshold into a universal rule.
The review did identify a stronger signal for a coordinated Advanced Resuscitative Care Bundle combining PRBC, tranexamic acid, and calcium. Three nonrandomized studies found a moderate-to-large reduction in mortality at hospital discharge compared with usual care, and the review rated that outcome at moderate strength of evidence. One study also associated each minute of delay to blood administration with higher odds of mortality. Yet this is bundle-level evidence from nonrandomized comparisons. It does not tell us that PRBC alone, tranexamic acid alone, calcium alone, or a particular rapid-infusion arrangement caused the entire observed association. It also does not erase the inconsistent results from other strategies.
Applicability is part of every one of these conclusions. The clinical literature is concentrated in adults with trauma and often underrepresents women, pregnant patients, children, non-White populations, people with nontraumatic hemorrhage, and patients in rural or frontier systems. Studies rarely report enough detail to show whether a helicopter program, a ground ambulance service, an intercept model, or a system with different clinician qualifications would obtain the same result. The evidence may be credible for the patients and systems actually studied while remaining uncertain for those left out.
For a reader evaluating a new result, four questions keep the interpretation anchored to the program. First, who received the intervention and who was the comparator? “Usual care” differs between systems, and a blood group may have been selected because clinicians recognized greater severity. Second, when was the outcome measured? Survival at arrival, 24 hours, discharge, and 30 days are related but not identical endpoints. Third, what kind of outcome changed? A process improvement or physiologic response can support feasibility and mechanism without establishing mortality benefit. Fourth, how much confidence should the estimate receive in this population and EMS environment? Risk of bias, imprecision, inconsistency, and missing modifiers are not ceremonial caveats; they determine whether a result should change policy.
The answer that emerges is deliberately modest but useful. The AHRQ review does not support a definitive product ranking, and it does not supply a universal clinical protocol. It does support a more precise judgment: the field has enough rationale, experience, and selected signals to justify careful program development in systems that can govern it, but not enough comparative certainty to claim that one intervention will improve mortality for every civilian patient. New evidence would change that judgment if it directly compared active alternatives in adequately powered, multicenter studies; used consistent mortality and harm definitions; reported patient and EMS-system characteristics; and included populations currently missing from the literature.
That distinction is the foundation for comparing the products themselves. The next question is not which label wins in the abstract. It is what whole blood, components, plasma, and crystalloid each contribute—and what each asks the program to make reliable.