Time-limited association between plasma transfusion and mortality in pediatric traumatic brain injury

Abstract

Background:

Traumatic brain injury (TBI) is a predominant cause of pediatric mortality. While prehospital plasma administration has been associated with lower mortality in adults with TBI, the impact of early plasma in children is unknown.

Methods:

In this retrospective cohort study, we examined the impact of plasma transfusion within 4 hours of arrival on 4-hour, 24-hour, and 30-day mortality in children aged 1-17 years with severe TBI (head Abbreviated Injury Scale 4-5) using the National Trauma Data Bank from 2020-2022. We excluded subjects with mild-moderate or nonsurvivable TBI or missing plasma or weight data. Cox proportional hazard models, clustered by facility, assessed the effect of early plasma on mortality, adjusting for age, sex, trauma mechanism, interfacility transfer, shock, total Glasgow Coma Scale (GCS), Injury Severity Score (ISS), trauma center level, insurance, binary whole blood, red blood cell, and platelet administration, and weight-adjusted total 4-hour transfusion volumes.

Results:

Of 367,065 children in the National Trauma Data Bank from 2020-2022, 14,691 met inclusion criteria, of whom 1,594 (10.9%) received early plasma. Subjects were mostly male (67.8%), with median(IQR) age 12 years (5-15), GCS 11(3-15), ISS 25(17-29), and 28.7% presenting in shock. The adjusted hazard ratio (HR) for the effect of plasma administration on mortality was HR = 0.610 (95% CI: 0.430-0.864; p = 0.005) at 4 hours, HR = 0.894 (95% CI: 0.706-1.131; p = 0.350) at 24 hours, and HR = 1.132 (95% CI: 0.961-1.334, p = 0.138) at 30 days.

Conclusions:

This study reports a significant association between early plasma administration and a lower risk of 4-hour mortality among children with severe TBI that does not persist at or beyond 24 hours. While these data suggest that plasma resuscitation may extend the window for lifesaving intervention, additional prospective data are needed.

Level of Evidence:

Level III, Therapeutic/Care Management

Background

Traumatic brain injury (TBI) is a primary cause of pediatric death and disability.(1-3) Children with severe TBI have high rates of trauma-induced coagulopathy (TIC), endogenous changes in the hemostatic, endothelial, immune, and inflammatory system, that can occur as soon as one hour post-injury.(4-7) The mechanisms of TBI-related TIC have not been fully elucidated, although potential mechanisms include release of coagulopathic brain-specific molecules after an injury breaches the blood-brain barrier, or shock and subsequent hypoperfusion of the brain (4,7) Despite a lack of mechanistic clarity, TBI-related TIC has been shown to be a distinct entity that is clinically relevant; in children with intracranial bleeding, some data suggest that coagulopathy is a predictor of progression of hemorrhage.(5,7-15) Currently, there are no known effective therapies to mitigate or prevent TIC in children with severe TBI.

In adults, both the Prehospital Air-Medical Plasma (PAMPer) and Control of Major Bleeding After Trauma (COMBAT) trials investigated the use of prehospital plasma as a first-line resuscitative fluid in severely injury patients.(16,17) While only the PAMPer trial demonstrated a mortality benefit for plasma administration in the initial analysis of the whole cohort, a combined post-hoc analysis of both trials suggested a 30-day mortality benefit based on transport time.(16-18) Subsequent secondary analyses of the PAMPer trial have since demonstrated that prehospital plasma was associated with significantly lower 30-day mortality among adult trauma patients with TBI as compared to those without TBI.(19,20)

Although promising in the adults with TBI, application of these findings to pediatric trauma patients warrants caution. First, children demonstrate unique physiologic responses to trauma in comparison to adults, and their coagulation pathways continue to develop throughout childhood.(4,21) Additionally, young children are at risk for abusive head trauma (AHT), a particularly morbid, complex mechanism of TBI frequently associated with an acute-on-chronic injury pattern that is not seen in the adult population and may be associated with mechanistically distinct coagulopathy.(1,22-24)

The primary purpose of this study is to identify any mortality differences that may be associated with early (within 4 hours of admission) transfusion of plasma in a large, multicenter national database of children with severe TBI; we hypothesized that early receipt of plasma would be associated with reduced mortality.

Methods

Study design, setting, and participants

A retrospective cohort study was conducted using data from the National Trauma Data Bank (NTDB) for the years 2020-2022. The database was queried for children aged 1-17 years with severe TBI, defined as head Abbreviated Injury Scale (AIS) equal to 4 or 5. Children with head AIS 1-3 (minor-moderate TBI), head AIS 6 (not survivable), head AIS 9 (unknown severity), or missing data for the key variables of binary receipt of plasma within 4 hours of hospital admission and weight were excluded. Study size was arrived at by utilizing all eligible records in the NTDB database from 2020-2022 to maximize power. This study was exempt from the Institutional Review Board due to the use of a national deidentified database. The STROBE guidelines were used to ensure proper reporting of methods, results, and discussion (Supplemental Digital Content 2).

Variables and definitions

The primary study outcome measures were mortality at 4 hours, 24 hours, and 30 days. Cox proportional hazard ratios for the effect of early plasma administration on mortality were generated. Demographic and other clinical variables of interest included age, sex (male/female), trauma mechanism (blunt/penetrating/both/other), prehospital interfacility transfer status, binary shock on admission, total Glasgow Coma Scale (GCS) score on admission, initial Injury Severity Score (ISS), pediatric trauma center verification level (1-3), insurance status (governmental/private/other), binary receipt of platelets within 4 hours of hospital admission, binary receipt of red blood cells (RBCs) within 4 hours of hospital admission, binary receipt of whole blood within 4 hours of hospital admission, and total blood product quantity transfused (mL/kg) within 4 hours of hospital admission. Presence of shock was defined as shock index, pediatric age-adjusted (SIPA) score greater than 1.2 in children ages 1-6 years, SIPA score greater than 1 in children ages 7-12 years, and SIPA score greater than 0.9 in children 13 years of age and older.(25,26) These covariates were included for to account for differences in injury type and severity, prehospital and in-hospital treatment, and individual-specific factors which could otherwise impact survival.

Statistical methods

Descriptive statistics were used to describe cohort demographics and clinical characteristics, including number (%) for categorical variables, and median (Interquartile Range; IQR) for continuous variables. Unadjusted Kaplan-Meier curves and log-rank tests for comparison were used to preliminarily estimate survival by plasma group at 4 hours, 24 hours, and 30 days. Plasma administration may be done as part of empiric hemostatic resuscitation in severely injured trauma patients.(27-35) As such, in an unadjusted model, plasma transfusion could be merely a surrogate marker for critical illness. To account for this, an adjusted model was used for primary survival analysis. Adjusted HRs and 95% confidence intervals (CIs) were then estimated with Cox proportional hazards models controlling for all potential confounders described above in addition to our intervention of interest, clustered by facility, for the same time intervals. Covariate-adjusted survival curves were generated for comparison. The proportional hazards assumption was assessed using Schoenfeld residuals. Statistical significance was defined as p-value < 0.05. All statistical analysis was conducted with STATA version 18 (Stata Corp, College Station, TX).

Adjusting for plasma volume and ratios

For the purposes of examining the effect of weight-adjusted plasma volume on survival, adjusted HRs and 95% CIs were estimated with a Cox model of patients who received plasma using the following categories: 1) low-dose plasma (> 0 to ≤ 20 mL/kg plasma received); and 2) high-dose plasma (> 20 mL/kg plasma received). For the purposes of this adjusted model, all other previously described covariates were included; however, to avoid collinearity, the variable of total transfusion volume excluded plasma volume. This analysis was only performed for the time point(s) with a significant association between mortality and binary plasma.

Similarly, to examine the effect of blood product ratios on survival, the plasma: total transfusion volume (TTV) ratio was generated. The TTV included whole blood, RBCs, plasma, and platelets. This was treated as a continuous variable, and adjusted HRs and 95% CIs were estimated with a Cox model of patients who received plasma with all other previously described covariates included. This analysis was only performed for the time point(s) with a significant association between mortality and binary plasma.

Sensitivity analysis

Sensitivity analysis was conducted on the excluded patients missing binary plasma and weight values to compare mortality outcomes to the non-missing cohort using the Chi-square test. Sensitivity analysis was also completed using multiple imputation to account for missing outcome data to verify the validity of the model. A second sensitivity analysis was additionally performed after excluding children with suspected abusive mechanism of injury to omit those with possible AHT, which may differ from the TBI mechanisms seen in accidental trauma. Children aged 1-3 years with ICD 10 codes corresponding to initial or unspecified episode of confirmed or suspected child abuse or shaken infant syndrome were identified and excluded from the sensitivity analysis cohort. Lastly, sensitivity analyses were performed for children with isolated severe TBI (maximum AIS for abdomen, thorax, and extremity < 3) and for children with severe TBI and polytrauma (at least one maximum AIS abdomen, AIS thorax, or AIS extremity ≥ 3). Adjusted HRs and 95% CIs were then estimated with Cox proportional hazards models for these groups for each time interval as above.

Results

Participants

Of 367,065 children included in the NTDB from 2020-2022, 14,691 met inclusion and exclusion criteria. Baseline demographics and clinical characteristics of these children are shown in Table 1. In this analysis cohort, 1,594 (10.9%) children received plasma within 4 hours of admission. Subjects who received plasma had higher rates of penetrating injury (27.4% vs 9.5%) and shock on admission (62.1% vs 24.7%), lower median (IQR) initial GCS scores (3 (3-4) vs 14 (3-15)), higher median (IQR) ISS (33 (26-43) vs 24 (17-29)), and higher rates of transfusion of other blood products. Overall, 12,911 children had complete data for all outcome measures and covariates (Figure 1).

Table 1: Cohort baseline demographics and clinical characteristics.

Some participants may be missing data, and percentages for each group were calculated using total participants with complete variable data only. Percentages may not add to 100.0% due to rounding. GCS = Glasgow Coma Scale; ISS = Injury Severity Score.

Variable	Plasma in 4 hours (n = 1,594)	No plasma in 4 hours (n = 13,097)	Total (n = 14,691)
Age (years), median (IQR)	14 (7-16)	12 (5-15)	12 (5-15)
Sex, no. (%)
Male	1,078 (68.0%)	8,825 (67.8%%)	9,903 (67.8%%)
Female	508 (32.0%%)	4188 (32.2%%)	4,696 (32.2%%)
Mechanism of injury, no. (%)
Blunt	1,062 (66.6%%)	10,673 (81.6%%)	11,735 (80.0%%)
Penetrating	436 (27.4%%)	1,241 (9.5%%)	1,677 (11.4%%)
Mixed	11 (0.7%%)	27 (0.2%%)	38 (0.3%%)
Other	85 (5.3%%)	1,141 (8.7%%)	1,226 (8.3%%)
Interfacility transfer, no. (%)	400 (25.1%%)	5,834 (44.6%%)	6,234 (42.5%%)
Shock on admission, no. (%)	919 (62.1%%)	3,039 (24.7%%)	3,958 (28.7%%)
Initial total GCS, median (IQR)	3 (3-4)	14 (3-15)	11 (3-15)
ISS, median (IQR)	33 (26-43)	24 (17-29)	25 (17-29)
Pediatric trauma center level verification, no. (%)
Level 1	582 (36.5%%)	5,191 (39.6%%)	5,773 (39.3%%)
Level 2	217 (13.6%%)	2,086 (15.9%%)	2,303 (15.7%%)
Level 3	795 (49.9%%)	5,820 (44.4%%)	6,615 (45.0%%)
Insurance Status, no. (%)
Governmental	788 (50.6%%)	6,123 (47.6%%)	6,911 (47.9%%)
Private	583 (37.4%%)	5,580 (43.4%%)	6,163 (42.7%%)
Other	187 (12.0%%)	1,159 (9.0%%)	1,346 (9.3%%)
Received blood product transfusion in 4 hours, no. (%)
Red blood cells	1,469 (92.2%%)	964 (7.4%%)	2,433 (16.6%%)
Platelets	707 (44.4%%)	72 (0.5%%)	779 (5.3%%)
Whole blood	157 (9.8%%)	144 (1.1%%)	301 (2.0%%)
Transfusion volume by product, if received (mL/kg), median (IQR)
Plasma	11.4 (6.3-21.6)	n/a	11.4 (6.3-21.6)
Red blood cells	20 (10.0-38.0)	10.7 (5.8-19.1)	15.6 (8.3-28.8)
Platelets	6.4 (3.7-11.2)	9.1 (3.7-13.5)	6.5 (3.7-11.9)
Whole blood	16.7 (8.7-25.0)	12.5 (7.4-20.0)	15 (7.8-23.8)
Total blood product transfusion requirement in 4 hours (mL/kg), median (IQR)	34.12 (17.81-64.88)	0 (0-0)	0 (0-0)

Figure 1: Subject flow diagram.

Participants were initially screened for eligibility. Only subjects with complete data were included in the final analysis.

Outcome data & main results

In total, there were 568 mortalities (3.9%) within 4 hours, 1,056 (7.2%) mortalities within 24 hours, and 2,936 (16.3%) mortalities within 30 days. Unadjusted mortality rates in the plasma group were higher than in the general cohort at each time point (142 (8.9%) at 4 hours, 363 (22.8%) at 24 hours, and 847 (53.1%) at 30 days). Unadjusted Kaplan-Meyer survival curves corroborated significantly increased unadjusted risk of mortality in the plasma recipient group compared to the no-plasma group on log-rank tests across all 3 time points (p < 0.001).

An adjusted Cox proportional hazards model demonstrated a lower HR for the cohort that received plasma when examining 4-hour mortality (HR = 0.610 (95% CI: 0.430-0.864; p = 0.005)). There was no statistically significant difference at 24 hours (HR = 0.894 (95% CI: 0.706-1.131; p = 0.350)) or at 30 days (HR = 1.132 (95% CI: 0.961-1.334, p = 0.138)). Table 2 lists HR and p-values for all variables included in the model. A covariate-adjusted survival curve for 4-hour mortality illustrates the difference in survival between plasma groups (Figure 2).

Table 2: Cox proportional hazard ratios, 95% confidence intervals, and p-values for all variables included in the models of mortality at 4 hours, 24 hours, and 30 days.

HR = Hazard Ratio; GCS = Glasgow Coma Scale; ISS = Injury Severity Score. For variables with groups in parentheses, the former was compared with the latter (the reference).

Variable	HR (95% CI), 4-hour mortality	P-value, 4-hour mortality	HR (95% CI), 24-hour mortality	P-value, 24-hour mortality	HR (95% CI), 30-day mortality	P-value, 30-day mortality
Received Plasma in 4 hours	0.610 (0.430-0.864)	0.005	0.894 (0.706-1.131)	0.350	1.132 (0.961 - 1.334)	0.138
Received blood product transfusion in 4 hours
Red blood cells	2.190 (1.524-3.147)	< 0.001	2.068 (1.634-2.617)	< 0.001	1.458 (1.255 - 1.693)	< 0.001
Platelets	0.892 (0.598-1.330)	0.575	1.004 (0.789-1.279)	0.972	1.149 (0.980 - 1.347)	0.088
Whole blood	0.999 (0.629-1.586)	0.995	0.917 (0.662-1.270)	0.600	1.117 (0.900 - 1.387)	0.317
Total transfusion requirement in 4 hours	1.000 (1.000-1.000)	0.696	1.000 (1.000-1.000)	0.718	1.000 (1.000 - 1.000)	0.821
Age	0.936 (0.908-0.965)	< 0.001	0.976 (0.958-0.995)	0.015	0.977 (0.966 - 0.988)	< 0.001
Sex (Female vs. Male)	0.937 (0.683-1.287)	0.688	0.891 (0.742-1.071)	0.218	0.929 (0.828 - 1.043)	0.215
Mechanism of injury
Blunt	Reference	n/a	Reference	n/a	Reference	n/a
Penetrating	6.873 (5.044-9.366)	< 0.001	4.477 (3.620-5.537)	< 0.001	3.779 (3.307 - 4.318)	< 0.001
Mixed	3.754 (0.754-18.697)	0.106	2.983 (1.178-7.552)	0.021	1.744 (0.805 - 3.778)	0.158
Other	0.987 (0.534-1.823)	0.967	1.334 (0.991-1.798)	0.058	1.293 (1.061 - 1.576)	0.011
Interfacility (Transferred vs. Not transferred)	0.685 (0.499-0.940)	0.019	0.823 (0.685-0.989)	0.037	0.866 (0.777 - 0.964)	0.009
Shock (Present vs. Absent)	1.222 (0.911-1.640)	0.181	1.513 (1.258-1.820)	< 0.001	1.468 (1.310 - 1.645)	< 0.001
Total GCS	0.628 (0.549-0.720)	< 0.001	0.667 (0.628-0.707)	< 0.001	0.686 (0.666 - 0.707)	< 0.001
ISS	1.022 (1.011-1.034)	< 0.001	1.018 (1.010-1.027)	< 0.001	1.017 (1.013 - 1.022)	< 0.001
Pediatric trauma center level verification
Level 1	Reference	n/a	Reference	n/a	Reference	n/a
Level 2	0.947 (0.624-1.438)	0.798	0.979 (0.716-1.337)	0.892	1.015 (0.864 - 1.193)	0.856
Level 3	1.286 (0.964-1.717)	0.087	1.505 (1.227-1.847)	< 0.001	1.304 (1.154 - 1.472)	< 0.001
Insurance Status, no (%)
Governmental	Reference	n/a	Reference	n/a	Reference	n/a
Private	1.162 (0.846-1.597)	0.354	1.180 (0.991-1.404)	0.063	1.021 (0.915 - 1.139)	0.707
Other	2.134 (1.554-2.930)	<0.001	1.956 (1.569-2.438)	< 0.001	1.749 (1.501 - 2.037)	< 0.001

Figure 2: Covariate-adjusted 4-hour survival curves comparing children who received plasma within 4 hours of admission vs children who did not receive plasma.

shows a significantly lower probability of mortality in the plasma group compared to no-plasma (HR = 0.610 (95% CI: 0.430-0.864); p = 0.005). Time is represented on the x-axis, and the survival probability is shown of the y-axis.

Effects of plasma volume and ratios on survival

In an adjusted Cox model of patients who received plasma that included weight-adjusted plasma volumes (low-dose plasma n = 1,152; high-dose plasma n = 442), there was no significant difference in association between plasma volume and mortality (HR for high-dose plasma in comparison to reference low-dose plasma: 1.207 (95% CI: 0.664 – 2.194), p = 0.537). When examining plasma:TTV as a continuous variable, there was an approximate 4.3% decrease (95% CI: 19.0% decrease to 13.0% increase) in hazards for every 10% increase in plasma:TTV for 4-hour mortality; however, this did not meet statistical significance (p = 0.605). As there was no significant association with 24-hour or 30-day mortality and binary plasma in the adjusted model, these time points were not examined for either plasma volume or plasma ratio effects.

Sensitivity analysis: Missing data

Among patients with head AIS 4 or 5 (n = 15,656), 450 patients were excluded due to missing binary plasma, 468 due to missing weight, and 47 due to missing both (Figure 1). Compared to patients with complete plasma and weight data, those missing plasma only had similar rates of 4-hour mortality (5.3% vs 3.9%, p = 0.114) and 24-hour mortality (6.7% vs 7.2%, p = 0.673), but lower 30-day mortality (12.7% vs 16.3%, p = 0.039). By contrast, patients missing weight only had higher rates of 4-hour mortality (45.9% vs 3.9%, p < 0.001), 24-hour mortality (52.8% vs 7.2%, p < 0.001), and 30-day mortality (57.3% vs 16.3%, p < 0.001). Sensitivity analysis using multiple imputation to account for patients missing outcome data to verify the validity of the model was also completed. After multiple imputation, there were no significant differences for any survival outcomes.

Sensitivity analysis: Excluding abusive head trauma

Sensitivity analysis was conducted after excluding young children with suspected or confirmed child abuse or shaken infant syndrome, to omit subjects with likely AHT. Overall, 439 (2.99%) subjects were identified as likely AHT and excluded. "Likely AHT” was identified by International Classification of Diseases, 10^th Revision (ICD-10) codes for suspected or confirmed physical child abuse and shaken infant syndrome in children less than 4 years of age. Adjusted Cox analyses demonstrated similar HR to the general cohort, with plasma HR for 4-hour mortality 0.612 (95% CI: 0.429-0.873; p = 0.007), plasma HR for 24-hour mortality 0.888 (95% CI: 0.697-1.131; p = 0.335), and plasma HR for 30-day mortality 1.148 (95% CI: 0.968-1.362; p =0.111).

Sensitivity analysis: Isolated severe TBI vs Polytrauma

Overall, 11,500 children (78.3%) in the cohort experienced an isolated severe TBI, with no major injuries to the chest, abdomen, or extremities; by contrast, 3,191 (21.7%) of children had at least one major injury to these body regions in addition to their severe TBI. Children with isolated severe TBI had similar HRs to the general cohort, with plasma HR for 4-hour mortality 0.630 (95% CI: 0.391-1.016; p = 0.058), plasma HR for 24-hour mortality 0.893 (95% CI: 0.654-1.220; p = 0.477), and plasma HR for 30-day mortality 1.097 (95% CI: 0.865-1.390; p =0.445). Children with polytrauma had a plasma HR for 4-hour mortality of 0.482 (95% CI: 0.263-0.886; p = 0.019), plasma HR for 24-hour mortality of 0.809 (95% CI: 0.555-1.180; p = 0.272), and plasma HR for 30-day mortality of 1.219 (95% CI: 0.960-1.547; p =0.104).

Discussion

In this retrospective cohort analysis of pediatric trauma patients with severe TBI, we report a significant association between early plasma administration and lower 4-hour mortality in a Cox proportional hazards model. Although this benefit was not present at or beyond 24 hours post-admission, these findings suggest that plasma resuscitation may extend the potential window for further lifesaving intervention in this cohort. Due to the limitations of a retrospective database analysis, additional confirmatory prospective pediatric data are needed to confirm these results and further test this hypothesis.

Crystalloid fluids (e.g. normal saline) are the current standard for resuscitation of children with TBI. However, substantial data show harm with crystalloid resuscitation in this population: dilutional coagulopathy, acute respiratory distress syndrome, abdominal compartment syndrome, and multiorgan failure.(21,30,36-39) As a potential mechanism, increased shedding of the endothelial glycocalyx, loss of membrane integrity, and increased inflammation are described with the use of crystalloids in a shock state, which can worsen edema, tissue hypoxia and organ failure.(7,40-44) Due to increasing recognition of these harms, adult trauma trials like PAMPer were designed with early plasma transfusion as an alternative resuscitative fluid to show a mortality benefit, specifically in subjects with TBI.(17,19,41)

However, these results should not be extrapolated to children without pediatric-specific data, as there is a need for clear and deliberate timing of plasma intervention relative to injury. A prior pediatric study has shown a relationship between TBI and coagulopathy, as well as harm with the reactive and delayed use of plasma. This single-center retrospective study reported that both TBI and plasma transfusion within 24 hours of admission were predictors of sustained fibrinolysis shutdown on thromboelastography and mortality in severely injured children.(45) A study examining changes in TEG following major trauma in children found that changes in coagulopathy occurred within one hour of injury, and tended towards the development of fibrinolysis shutdown over time.(5) These studies emphasize the importance of early intervention along the pathway to TIC, given the rapidity of worsening coagulopathy over time. With this study, we therefore investigate the use of early plasma within 4 hours of hospital admission and report an early mortality benefit for children with the use of early plasma in severe TBI.

In addition to timing of transfusion, volume of transfusion may also have significant effects. Stratifying our analysis by volume in the subgroup of plasma recipients did not show a volume-dependent relationship between plasma and mortality, which is in contrast to a prior study though their reference group was no plasma which likely is a reflection of injury severity versus the isolated impact of plasma itself.(45) Our analysis of plasma transfusions ratios also showed no significant relationship between increasing plasma ratios and mortality. However, both of these analyses are limited by sample size and lack of an appropriately matched comparison group; further studies are warranted.

The finding that plasma resuscitation may have a time-limited association with mortality, as evidenced by a lack of difference at later mortality timepoints (24-hour and 30-day), has both biologic explanation and precedence in the literature. Another example of time-limited interventional benefit is found in the adult Pragmatic, Randomized Optimal Platelet and Plasma Ratios (PROPPR) trial, which found that transfusion of platelets:plasma:pRBCs of 1:1:1 as compared to 1:1:2 demonstrated decreased death from exsanguination at 24 hours and achieved hemostasis faster; however, there were no significant differences in 24 hour or 30-day mortality.(31) If early plasma resuscitation does provide benefit to patients experiencing TIC in the setting of severe TBI, those patients that survive the acute, hemorrhagic coagulopathy still face potentially devastating longer-term sequelae that are ultimately not impacted by the initial resuscitation but by the whole bundle of care and intervening in-hospital complications.(2,3) Ultimately, careful consideration must be made as to the benefits of interventions which do not persist to affect long-term outcomes.

Limitations

Our study has several key limitations. First, the retrospective design inherently limits the ability to show more than associations between administration of early plasma and mortality at any time point. Though we adjusted for anticipated confounders, there may be additional variables that were not included in our analysis or available in the database. The NTDB is limited in scope and does not include granular information regarding precise reasons for or timing of transfusions or prehospital transfusion, nor are data provided for infants, limiting generalizability to this age category. Causes of death are not captured in this database either, and limit conclusions regarding drivers of long-term mortality in this population that appear to dilute the effects of early plasma over time. Further, transfusion data are limited to the time periods documented in the database, therefore the ability to account definitively for changing risk of death over time within those time periods is not possible and limits applicability and generalizability of the proportional hazards model. There is a potential for survival bias, wherein subjects must have lived long enough to receive plasma, and early deaths may be poorly accounted for. This may also be reflected in the mortality discrepancies between our excluded cohort missing weight data compared to our cohort with complete data, as subjects who did not have a recorded weight had significantly higher mortality rates across all three time points, suggesting that many of these children likely died too quickly for any intervention. However, these children were excluded, which somewhat limits bias from this source. Finally, the outcome of mortality is not the most optimal for studies of TBI; while some expert groups recommend a later post-discharge neurologic outcome, this was not available in the dataset and is often limited by loss to follow up.(46)

Conclusion

Early plasma administration is significantly associated with lower risk of 4-hour mortality among children with severe TBI; however, this association is not significant at 24 hours or 30 days. Although there are limitations of the database and study design, this work suggests that early plasma likely does not lead to harm in this population and there may be some time-limited benefit. However, high-quality, large prospective pediatric data are still needed in this high-risk population.

Supplementary Material

Supplemental Data File (.doc, .tif, pdf, etc.)

Supplemental Digital Content 2: STROBE Statement Checklist. STOBE (STrengthening the Reporting of OBservational studies in Epidemiology) has provided a list of required components for all cohort studies. This list has been attached with relevant corresponding page numbers.

Conflict of Interest Statement (can upload 1 file with ALL AUTHOR forms in PDF)

Supplemental Digital Content 1: Conflict of Interest Disclosure Forms. All authors have provided JTACS COI Disclosure Forms attached herein.