Observational studies on transfusion in trauma comparing high versus low plasma:erythrocyte ratio were prone to survivor bias because plasma administration typically started later than erythrocytes. Therefore, early deaths were categorized in the low plasma:erythrocyte group, whereas early survivors had a higher chance of receiving a higher ratio. When early deaths were excluded, however, a bias against higher ratio can be created. Survivor bias could be reduced by performing before-and-after studies or treating the plasma:erythrocyte ratio as a time-dependent covariate.We reviewed 26 studies on blood ratios in trauma. Fifteen of the studies were survivor bias-unlikely or biased against higher ratio; among them, 10 showed an association between higher ratio and improved survival, and five did not. Eleven studies that were judged survivor bias-prone favoring higher ratio suggested that a higher ratio was superior.Without randomized controlled trials controlling for survivor bias, the current available evidence supporting higher plasma:erythrocyte resuscitation is inconclusive.
THERE is consensus in trauma management that fresh frozen plasma (FFP) should be given when continuous hemorrhage and coagulopathy are present.1,–,4In recent years, however, the best way of giving FFP in the 2% of civilian and 7% of military trauma patients who require massive transfusion (MT) has been controversial.5,–,8This controversy is between use of conventional fluid and blood management guidelines versus the so-called “1:1” strategy in the subset of trauma patients who require MT.6,–,8The definition of MT in the literature varies, but is typically equal to or more than 10 units of packed erythrocytes within 6–24 h of hospital admission.1,–,5
Conventional fluid management in hemorrhagic shock has been to give crystalloid and erythrocytes initially.9According to current guidelines, FFP should be considered after 1–1.5 blood volumes have been lost or coagulation tests are raised (international normalized ratio more than 2 or activated partial thromboplastin time more than 1.5×normal), in the presence of excessive microvascular bleeding.3,4The decision to give FFP may therefore be made at 0.5–4 h following hospital admission. Furthermore, FFP requires thawing, and is sometimes ordered in small quantities (e.g. , 10–15 ml/kg or equal to or less than 4 units).3,4Whereas group O erythrocytes for severe shock has been readily available for decades, use of universal donor plasma (AB, a rather rare commodity) is only a recent development. The number of units of plasma given compared to that of erythrocytes (P:E) tends therefore to have been far less than one during the initial hours of resuscitation. It just so happens also that hemorrhage and coagulopathy are important causes of death in the first few hours of resuscitation.7,10In recent years, many practitioners have argued that conventional transfusion guidelines do not sufficiently address the needs of trauma patients with massive and ongoing bleeding. Based only on anatomic and physiologic parameters identified within minutes of arrival in hospital and on initial response to resuscitation, clinicians can reasonably identify that small subset of patients whose transfusion requirement will likely be substantial,11,12and administer prethawed FFP and blood products such that P:E approaches 1:1–2 early in the course of resuscitation.5,–,8This nearly unity ratio is continued but is terminated as soon as hemorrhage is controlled. The emphasis of this so-called “1:1” philosophy is very much on the early (within less than 0.5–1 h of admission) attainment of a P:E ratio of 1:1–2, hence achieving such a ratio at the 24thhour after hospital admission (playing catch-up) does not constitute “1:1.”7
There has not been a single randomized controlled trial (RCT) to date validating either traditional transfusion guidelines or “1:1” in MT.7Proponents of the “1:1” philosophy cite supportive observational studies.6,7Opponents counter that observational studies are prone to biases.6Without data from RCTs, a 17-member multidisciplinary panel recently would not recommend either for or against transfusion of FFP at a P:E ratio of more than or equal to 1:3 during MT.2
There is in particular one type of bias, survivor bias (SB), or immortal time bias, that is universally cited as the flaw that puts the validity of P:E studies in doubt (SB is actually not uncommon in observational studies).6,8,13It arises because patients with massive and ongoing hemorrhage often die during the early hours of hospital admission before receiving substantial quantities of FFP (when using traditional transfusion guidelines), and thus are categorized (without randomization) in the low P:E cohort in observational studies, whereas patients surviving long enough to receive sufficient FFP (finally caught up) are categorized in the high P:E cohort.13Hence observational studies using data from centers using conventional (as opposed to “1:1”) guidelines have a built-in SB favoring the high P:E cohort.13In logistic regression analysis, low FFP may similarly emerge as an associative factor in mortality.
To avoid this form of SB, some investigators have excluded those patients who died within half to several hours after hospital admission. Unfortunately, since hemorrhage and coagulopathy are such important causes of mortality during the first few hours in a significant proportion of trauma patients (other causes of early deaths include severe head injury with or without associated hemorrhage, but within the first 6 h, exsanguination is dominant14), the practice of including only patients who survive until intensive care unit admission 4–7 h later, for example, excludes a significant portion of patients who had died from exsanguination who potentially could have benefitted from increased FFP therapy. In the process, a much less discussed, albeit important, SB against high P:E is created.8,10
As an increasing number of trauma centers appear to be aiming for a “1:1” philosophy based on its effectiveness as reported in observational studies, there still remains concern that many of these studies may be flawed because of SB. Thus we need to pause and ask the important question: How widespread is SB in P:E studies, and are we being too hasty or too cautious on this important resuscitation issue? To answer this we have appraised published P:E studies for the presence of SB. Randomized trials and observational studies are prone to other bias, but SB is our focus here. Also, only studies on P:E ratios in trauma resuscitation were examined; other hemostatic therapies and adjuncts are outside the scope of this review.
It is important to note that for the purpose of this review, high P:E at 24 h and “1:1” are not exactly interchangeable. The former applies to studies in centers that adhered to conventional guidelines (in which FFP typically is started much later than erthryocytes), and P:E was calculated using the cumulative FFP and erythrocytes at the end of 24 h. The latter applies to studies in centers where there was an early aggressive FFP protocol, in which “1:1” means a high P:E that was achieved shortly after hospital admission.
Materials and Methods
The MEDLINE database was searched using OVID interface from 1966 to July 2011, combining the keywords “massive transfusion” and “trauma.” All 210 abstracts thus found were reviewed, and the full texts of 75 studies, case series, and reviews were examined. Abstracts that have not been published as full papers were excluded. Also the January 2010-July 2011 issues of J Trauma, Injury, Crit Care Med, Intens Care Med, Crit Care (London), Surgery, Am J Surg, Br J Surg, Can J Surg, J Am Coll Surg, Ann Surg, World J Surg, Anesthesiology, Anesth Analg, Br J Anaesth, Can J Anesth, Transfusion, Vox Sang, and Resuscitation were reviewed. Finally, a search on Google was made (“trauma+ coagulopathy”) and the first 100 hits vetted. Bibliographies of all reviewed papers were then searched for more articles.
Any study was included for analysis if it consisted of a comparison between high FFP:packed erythrocyte ratio versus low ratio in trauma resuscitation involving MT. MT was defined as equal to or more than 10U erythrocytes over less than or equal to 24 h, or any average of equal to or more than 1U erythrycytes/h within the first 12 h of resuscitation. Case series and reports, nonhuman studies, reviews, commentary articles, and nontrauma papers were also excluded, although their references were reviewed.
For studies involving the use of warm fresh whole blood (WFWB), the equivalent blood ratios were used, meaning 1 unit of WFWB was considered the same as 1 unit each of FFP, platelets, and packed erythrocytes.
Use of recombinant activated factor VII (rFVIIa) and other prohemostatic nonblood products was noted but was not used as an inclusion or exclusion criterion.
Two of the authors (Drs. Ho and Dion) independently examined the final list of papers chosen for review and drew consensual conclusions. There was no blinding of the names of authors, their affiliations, or the journal titles. RCTs, if there were any, would be judged SB-free.
A study was considered SB-unlikely:
if cohorts after and before the implementation of a MT protocol that called for the early attainment of P:E = 1:1–2 were compared. The cohorts were independent. (Such studies fulfill the objective of comparing “1:1” with conventional fluid and blood product management as the MT protocol typically calls for the early/earlier attainment of equivalent units of FFP and erythrocytes);
if P:E was analyzed as a time-dependent covariate.
A study was considered SB-prone:
if patients were drawn from the same pool and were categorized in the low or high P:E cohort depending on how much blood products they had received up to a fixed time point (e.g. , at 24 h or time of death). All patients were included from time of hospital admission or shortly thereafter to that time point. This bias is in favor of high P:E because early deaths were categorized in the low P:E cohort. Such studies are high vs. low P:E and do not truly compare “1:1” and conventional fluid/blood management as there is no stipulation on an early attainment of a P:E of 1:1–2; only the summative P:E at the end of 24 house or time of death is considered;
if patients were drawn from the same pool, but only those surviving beyond the first few hours or long enough to be admitted to an intensive care unit were studied. This bias is against high P:E because hemorrhage is a dominant cause of death within the first few hours14and patients who had survived that long either had less severe coagulopathy or might have already benefitted from having high P:E management. Here again, such studies do not truly compare “1:1” and conventional management as only the summative P:E at the end of 24 h is considered.
No RCTs were found. Thirty eight (fig. 1) uncontrolled observational trauma studies comparing high versus low P:E in MT were identified. The definitions of high and low P:E vary with the studies.15,–,52In general, high P:E is loosely defined as equal to or more than 1:1–2 and low FFP is defined as less than 1:1–2.15,–,50From that list, 12 studies were excluded, three because trauma patients made up only a fraction of the patient population studied,41,–,43three because they involved only patients with ruptured abdominal aortic aneurysm,44,–,46one (involving patients with vascular or extremities injuries) because massively and nonmassively transfused patients were included,47one because the patient population was similar to another study already included,48two because both the study and control groups were given low FFP,49,50and two because only platelet:erythrocyte ratios were examined.51,52
A total of 6,655 patients were studied in the 26 reports included in the final analysis (table 1, fig. 2).15,–,40Because overlapping of patients likely occurred in some studies,19,20,27,29the total number of patients is probably slightly less. In figure 2we provide a breakdown of the studies based on civilian or military, effect of high P:E/“1:1,” and whether they were SB-prone (favoring high P:E or against high P:E) or SB-unlikely. Mortality is typically reported as in-hospital or 30-day mortality.
In terms of the crucial result of interest the following was found:
Among 22 civilian studies,19,–,40seven studies29,–,35(n = 1,259) judged SB-unlikely showed “1:1” as superior, four studies37,–,40(n = 668) judged SB-unlikely showed “1:1”/high P:E as not superior, seven studies22,–,28(n = 2,603) judged SB-prone (favoring high P:E) showed high P:E as superior, three studies19,–,21(n = 945) judged SB-prone (against high P:E) showed high P:E as superior, and one study36(n = 81) judged SB-prone (against “1:1”) showed “1:1” as not superior;
Studies with SB favoring high P:E and showing an association between high P:E and increased survival accounted for 11 of the 26 studies analyzed.
Overall, of 26 studies on high versus low P:E ratios, 21 found an association between high P:E (i.e. , 1:1–2)/“1:1” and improved survival but only 10 (all civilian) of them were either SB-unlikely (7) or had a bias against high P:E (3). Five (all civilian) of the 26 studies found no such favorable association but one of them had a built-in bias against high P:E and one had a control group whose transfusion management was closer to “1:1” than to conventional practice (Dr. Jesper Dirks, Clinical Associate Professor, Department of Anesthesia, Centre of Head and Orthopedics, Copenhagen University Hospital, Copenhagen, Denmark, personal e-mail communication, June 2011). In other words, only 10 of the 26 studies showed high P:E as superior and were SB-free or had SB against high P:E. No association between high P:E and reduced short-term survival has been reported.
Our review discovered that SB is rather prevalent in studies on blood product ratios in trauma requiring MT. The positive association between high P:E and improved survival in SB-prone (favoring high P:E) studies may be partly or wholly factitious. On the other hand, the SB-stigma should not apply to all P:E studies. In fact, 10 studies that found a positive association between “1:1”/high P:E and improved survival were SB-free or had SB against high P:E,19,–,21,29,–,35one study that did not find such a positive association had a bias against “1:1,”35and no study found “1:1” or high P:E to have been associated with decreased survival.
Study Design Strategies to Prevent SB
Including only patients who have survived the initial few hours when FFP administration typically lags is one technique of avoiding SB in favor of “1:1.”53,54The exact number of hours to exclude is highly variable. For example, mean time to first FFP (not necessarily P:E equals 1:1–2 status) before and after implementation of a “1:1” policy at Stanford Medical Center was 254 and 169 min, respectively.31On the other hand, at the R. Adam Cowley Shock Trauma Center, where “1:1” has been standard practice for a longer time,8,35,36the mean time to achieving nearly 1:1–2 blood product ratios would be shorter. One group excluded deaths within the first 30 min of admission to hospital but only one patient was excluded.23This was probably insufficient to significantly reduce SB that favors high P:E.
Including only patients who have survived to reach the intensive care unit may eliminate SB that favors high P:E. The price to pay, unfortunately, is that any benefits of high P:E become harder to observe. About 10–20% of trauma deaths are possibly preventable and 10–80% of them are due to hemorrhage, mostly occurring within 6 h of admission and with coagulopathy playing a major role.10,14Including only those patients who survive surgery to be admitted to the intensive care unit (some 4.5–7 h after admission20,21,36,49), as in Scalea et al. 's,36Gonzalez et al. 's21and Duschene et al. 's20studies, would exclude patients who might have serious hemostatic issues (and have succumbed), and in whom “1:1” might have helped. Instead, such a study would have focused only on those who might have less serious hemostatic issues (and have survived), and in whom the benefits of “1:1,” if there had been any, would not have been uncovered.8,10Furthermore, if a center already had “1:1” in place, as might have been the case at the A. Adams Cowley Shock Trauma Center,35,36studying only survivors beyond the first few hours of admission meant that some of the studied patients might have already enjoyed the benefits of “1:1,” had there been any.
Clearly, the longer the period of exclusion since admission in such observational studies, the less SB against conventional plasma management (in centers without a “1:1” policy) there is, but the higher the potential bias against high P:E or “1:1” (regardless of whether the transfusion protocol is conventional or “1:1”).
Another strategy for avoiding SB is to populate the low and high FFP cohorts with independent subjects. This has been done by taking data collected after implementation of a “1:1” protocol and comparing them with a historical cohort. This strategy avoids a Hawthorne effect that can be associated with RCTs. However, comparing 2-yr cohorts before and after implementing “1:1” ignores secular trends8and possibly allows the benefits of other advances in resuscitation to be credited to “1:1.” For instance, many centers have seen a dramatic decrease in 24-h crystalloid use as they have transitioned to a “1:1” policy. It is unclear if the potential benefit of “1:1” is from more FFP or less crystalloid, or both.
Another way of avoiding SB is to model the relationship between mortality and P:E ratio over time, and treating the P:E ratio as a time-dependent covariate.35,37,38This is a recommended technique of avoiding SB in observational studies.53,54Snyder et al. compared mortality of patients who had had high P:E at the end of 24 h and those who had had low P:E and found the former to be associated with increased survival.37This advantage became statistically insignificant when they divided the 24-h study period into 0.5–6 h subintervals and calculated the P:E and the mortality rates of the high P:E and low P:E groups within each subinterval.37deBiasi et al. also did not find the mortality of high P:E and low P:E to be different but found that a high FFP deficit (not a low P:E) in relationship to erythrocytes was associated with higher mortality.35This study has been put into the category of studies that have found a positive association between high P:E and improved survival (see table 1and fig. 2) because the study's conclusion calls for equal number of FFP and erythrocytes units.
In “1:1,” Is It the Ratio, or the Timing, That Matters?
In one before-and-after study in which the cohort after the implementation of a “1:1” MT protocol was found to have significantly better survival, the cumulative P:E at 24 h after admission was 1:1.8 both before and after adoption of “1:1.”31The only difference was that “1:1” (the “after” group) resulted in earlier administration of the first unit of FFP [169 (95% CI: 130–209) min], whereas clinicians using traditional resuscitation principles (the “before” group) started late with FFP [254 (185–323) min; P = 0.04] but played “catch-up.” In that study, platelets were also started significantly earlier in the post-MT group.31The early use of WFWB was associated with a survival advantage in one study.16In Dirks' comparison of trauma patients before and after institution of a “1:1” protocol, there was no difference in mortality.39They used international normalized ratio more than 1.2 or activated partial thromboplastin time more than 35 s as a FFP-transfusion trigger, and had on standby prethawed FFP during both the pre- and post-MT periods (Dr. Jesper Dirks, Clinical Associate Professor, Department of Anesthesia, Centre of Head and Orthopedics, Copenhagen University Hospital, Copenhagen, Denmark, personal e-mail communication, June 2011). First exposure to FFP was 28 min after admission even in the pre-“1:1” cohort, possibly accounting for the lack of clinical outcomes difference. These studies16,31,39suggest that if “1:1” is indeed superior, it might be due to the earlier, not just the increased, use of FFP. As such, it is worth emphasizing that studies examining the cumulative P:E at 24 h after admission do not shed enough light on the “1:1” paradigm as they do not factor in the timing at which P:E equal to 1:1–2 was reached. It is possible that in future studies the 6-h period/endpoint after admission will be more revealing.
The reviewed studies typically used 24 h-, 30 day-, and in-hospital mortality as end-points. Since uncontrolled hemorrhage is an important cause of early death, and “1:1” emphasizes early aggressive use of FFP, it makes sense to measure mortality at 6 h. In Holcomb et al. 's (SB-prone favoring high P:E) study, the Kaplan–Meier survival curve diverges mainly during the initial 6 h of admission, after which the curves were largely parallel and much flatter.23Later end-points are also important to determine if early and increased use of FFP actually leads to increased or decreased overall exposure to blood products, which could be reflected in the incidences of transfusion-related acute lung injury, respiratory distress syndrome, and multiorgan failure. The issue of competing mortality risks has received less attention in transfusion studies. All clinicians recognize that trauma patients largely die from hemorrhage, head injury, and multiorgan failure/sepsis, and some patients have multiple causes of death. These events, although related to the initial injury, occur at distinctly different time points, and the effect of competing mortality risks must be accounted for.
Notable Studies Not Included in Our Analysis
Several MT studies on P:E that are of interest were excluded from this review because nontrauma patients were included. In Johannson et al. 's. (SB-unlikely) study, P:E at 24 h after and before implementation of “1:1” were 1:1.3 and 1:1.6, respectively, but the post-“1:1” group received FFP earlier and had better survival.41Rose et al. found in their (SB-prone favoring high P:E) study an association of P:E ratio more than 1.1 with improved survival in elective and emergent nontrauma and trauma patients requiring MT.42Johannson et al. found in a (SB-prone favoring high P:E) study that MT patients whose mean P:E was 1:1.25 had a survival rate of 50%, whereas those whose PE was 1:2.5 had a survival rate of 7.7%.43A further three MT observational (SB-unlikely) studies involved patients with ruptured abdominal aortic aneurysm showed that after implementation of “1:1” survival improved.44,–,46
Military versus Civilian
Military and civilian injuries share many similarities, but there are some notable differences. A civilian must wait for an ambulance but a soldier injured on the battlefield usually receives immediate basic care from a fellow soldier and then a combat medic55before embarking on an arduous journey to the nearest combat support hospital. Blood products at the combat hospital are usually available in adequate amounts, and not infrequently augmented by WFWB. Since 2004 many deployed hospitals have used thawed AB plasma and erythrocytes as the primary initial resuscitation fluids. With planning and routine use, WFWB is available within 30 min of a request but can sometimes take up to 2 h. Some civilian trauma centers have followed this practice, and thawed AB plasma has become available in recent years. Because of the higher energy transfer and predominance of multiple penetrating injures, combat casualties have a higher chance of requiring MT, placing greater burdens on the blood bank and operating room logistics and personnel. Despite these differences, survival after combat injury and massive hemorrhage is at least equivalent and in many cases superior to those seen in civilian centers. So far, all four military studies15,–,18have all included patients from the time of hospital admission or shortly thereafter, and are therefore SB-prone, favoring high P:E. Future observational military studies should use time-dependent analysis to avoid SB.
In summary, we have outlined criteria for identifying SB when appraising observational studies on the use of FFP in trauma requiring MT and have applied them to published studies to discover that SB is common. Trauma management is complex and it takes the combined merit of many interventions to bring about measurable improvements. The presence of SB in part may explain the rather big reduction in mortality associated with high P:E in some studies. On the other hand, the bias cannot be used to dismiss all of the available data to date. Doing so is cherry-picking and spoils the debate. Uncontrolled observational studies are prone to other biases and confounders that diminish their validity. In closing, therefore, we emphasize the need for good RCTs to answer this question. To this end, the multicenter Prospective Randomized Optimum Platelet and Plasma Ratio trial comparing blood product ratios in trauma patients predicted to require MT will start enrolling patients in 2012,‡‡and the Trauma Formula-driven versus Lab-guided study (ClinicalTrials.gov Identifier: NCT00945542) comparing “1:1”versus conventional resuscitation in patients with hemorrhagic shock is currently enrolling patients, as is the Early Whole Blood in Patients Requiring Transfusion After Major Trauma trial (ClinicalTrials.gov Identifier: NCT01227005).§§