Utility of point-of-care platelet aggregation testing for transfusion prediction

Abstract

Early identification and initiation of therapy for life-threatening hemorrhage is essential to minimize patient morbidity and mortality. In primary hemostasis, platelet function is integral to reach this goal, but major hemorrhage leads to impaired platelet mechanical activation and aggregation. Current devices for measuring platelet function are cumbersome or not promptly available for clinical decision making in this setting. Within this manuscript we prospectively evaluate a novel, rapid assay utilizing measures of platelet aggregation to predict hemorrhage. In this prospective cohort study at an academic regional Level I trauma center, we included adult (> 16 years old) participants who were triaged as level I or II trauma activations. The primary exposure studied was platelet aggregation analyzed on a prototype device. The primary and secondary outcomes measured were life-threatening hemorrhage (death from hemorrhage or need for hemorrhage control procedure) and transfusion requirements of > 2 units of blood components, respectively. Standard descriptive statistics were used to characterize the cohort. Predictive outcomes were analyzed using multivariable regression to compare: (1) the platelet aggregation assay; (2) clinical parameters (systolic blood pressure, heart rate, and injury mechanism); and (3) a combined model. Of 761 patients, 482 patients met inclusion criteria for our study, 36 (7.5%) had life-threatening hemorrhage and 43 (8.9%) patients required > 2 units of blood transfusion. For life-threatening hemorrhage, platelet aggregation had an area under the curve (AUC): 0.61 (95% confidence interval [CI] 0.53–0.69); clinical parameters AUC: 0.83 (CI 0.75–0.91); and the combined model AUC: 0.85 (CI 0.79–0.92) which was not significantly improved when compared to clinical parameters alone (p = 0.32). For transfusion of > 2 units, the platelet aggregation model had AUC: 0.68 (CI 0.61–0.76); clinical parameters AUC: 0.84 (CI 0.79–0.90); and combined model AUC: 0.88 (CI 0.83–0.93), improving transfusion prediction over clinical parameters alone (p = 0.013). In a cohort of traumatically injured patients, a novel, rapid measure of platelet aggregation enhanced well-established clinical parameters to predict the need for blood transfusion but not life-threatening hemorrhage. Future work should validate the clinical utility of this technology in a larger cohort and patients with significant non-traumatic hemorrhage.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-51134-5.

Introduction

Hemorrhage and resulting hemorrhagic shock are common life-threatening problems in the emergency department (ED) and hospital¹, representing a potentially preventable cause of death, especially amongst traumatically injured patients^2–5. During hemorrhage, circulating platelets and their subsequent adhesion to the subendothelial matrix through the damaged endothelium are the driver of primary hemostasis⁶. When platelets encounter exposed subendothelial collagen at sites of vascular injury, von Willebrand Factor (VWF) binds to collagen and facilitates platelet tethering and rolling via the GPIb–IX–V axis, initiating platelet plug formation⁶. During life-threatening hemorrhage, failure of this platelet plug is frequently due to platelet dysfunction rather than depletion^7,8, and thus blood components are given empirically in patients with clinical evidence of major hemorrhage for resuscitation and replacement of lost thrombotic elements⁹.

There are currently multiple clinically available assays to help diagnose coagulopathy by characterizing patient clotting and lysis. However, these assays have limitations. Thromboelastography (TEG) and rotational thromboelastometry (ROTEM)^10,11 do not detect the most affected pathways implicated in platelet dysfunction such as vWF–GPIb–IX–V Axis or ADP–P2Y12 Signaling Pathway^12–14. Additionally, TEG-platelet mapping is complex, difficult to rapidly interpret, and requires normal fibrin function for reliable results^15,16. Further, impedance platelet aggregometry has not been shown to be useful in informing real-time clinical decision making^15,16. Thus, there is an unmet need to develop rapid and accurate diagnostic tests identifying platelet dysfunction during injury- or illness-induced coagulopathy.

The ATLAS microfluidic test platform shears whole blood at activating regimes (> 16,000 s⁻¹) to induce a platelet aggregation, activation, and contractile densifying action. The ATLAS does not require the aid of activating reagents to elucidate the native whole blood response, and produces initial results in a timespan that is relevant to clinical care (2 min)¹⁷. Platelet force, which is a direct output of the contractile actin-myosin mechanism, has been previously shown to be an integrator of the function of the total activation pathway and subsequent signaling cascade central to platelet function¹⁷. Previous evidence that inhibition of GPIb or αIIbβ3 platelet sites independently reduce platelet force suggests that different forms of platelet dysfunction can all be detected through a loss of platelet force^17,18. These recent developments in measuring platelet contractile force can potentially bridge the gap that exists in clinical evaluation for coagulopathy, and more specifically for platelet activity or function¹⁹. Current assays lack this ability and rely on agonists to probe specific receptors as part of characterizing aggregometry (ex: VerifyNow™), and have not been shown to be predictive of outcomes²⁰. The ATLAS platform’s reagent-free operation and 2-min initial result time position it as a candidate for filling this gap in acute trauma care.

Utilizing this novel ATLAS microfluidic platform, we aimed to evaluate the clinical predictive ability of a bedside assay assessing platelet force aggregation in a prospective study of traumatically injured patients at risk for hemorrhagic shock and mortality.

Methods

Study design and participants

This was a prospective cohort study performed at an academic regional Level I trauma center from August 2022 to July 2024. Participants were eligible for inclusion if they were older than 16 years of age and arrived at the trauma center as a level I or level II trauma activation (Supplemental Fig. 1—Trauma Activation Criteria). Participants were excluded if they had been administered greater than 3 units of blood components of any kind prior to study blood draw at the time of inclusion, were known to be pregnant, or were incarcerated (Fig. 1). Participants were further excluded if they were found to have no injuries after trauma team evaluation based on physical exam and radiographic evaluation, if they did not have enough blood drawn for the assay, or if they had platelet force assays with a low confidence level or extreme values (as defined below).

Fig. 1.

CONSORT diagram—Study population, inclusion, exclusion, and final analyzed cohort.

The study protocol was approved by the local Institutional Review Board with a waiver of informed consent given the minimal risk of the study, which required 2 mL of blood for analysis and the emergency nature of the studied condition and study procedures, precluding detailed informed consent procedures. The study was conducted in accordance with the Declaration of Helsinki.

The clinical treatment team in the ED was blinded to platelet assay results.

Exposures and the stasys ATLAS system

The platelet contractile force assay (ATLAS system, Stasys Medical) is a microfluidic platelet force testing system based upon flowing whole blood across flexible micron scale force sensors that then aggregate and activate platelets (Supplemental Fig. 2—Microfluidics Platform). In engineered microfluidic channels, platelets in whole blood are shear activated with microscale sensors to reflect their activation during primary hemostasis at a vascular defect. The assay can quantify the contractile force of a platelet plug, a direct measurement of mechanical adherence and densification, and measure the rate of platelet aggregation.

The mechanism by which platelets adhere to the sensors rapidly occurs when whole blood contacts the un-coated exposed silicone sensors themselves. Von Willebrand factor and an array of other available proteins immediately bind to the sensors, providing binding sites for platelet GP(Ib)-IX receptor complexes, which forms catch bonds with vWF linking the membrane with the internal actin cytoskeleton¹⁸. Under physiological shear stress, this then initiates membrane mechanoreceptor triggering sequences and intracellular signaling and further platelet recruitment under shear²¹. Thus, without specimen pre-treatment such as chemical activation or processing, a direct whole blood sample can begin resulting within 2 min, with activation starting within seconds of exposure to high shear and extending for a desired length of the test to observe platelet mechanics in real time. This assay has been previously validated against healthy donor controls, who demonstrated an average platelet force of 158 nN and maximum force of 328 nN, establishing reference values for an uninjured population¹⁹. Further, this assay has previously shown predictive ability in vitro, among patients taking aspirin in a cross-sectional pilot study of traumatically injured patients^19–21.

During clinical blood drawings at the time of ED presentation, each participant had 2 mL of whole blood collected and placed into a Sodium Citrate 2.7 mL, BD vacutainer. A 1 ml syringe was loaded onto the platelet contractile force assay where it was exposed to an activating shear rate of 16,000 s⁻¹ for 45 seconds¹⁷. The shear rate of 16,000 s⁻¹ was selected for the current vertically-aligned sensor configuration and has been validated to produce equivalent platelet activation, aggregation, and microclot build rates to those observed at 12,000 s⁻¹ in the previously reported horizontally-aligned configuration¹⁷. It was then exposed to a non-activating shear rate of 500 s⁻¹ for the remaining 4 min and 15 s of the test to provide a supply of bloodborne proteins and platelets without further encouraging activation and aggregation. The first high shear rate is above the critical activation level for platelets, and they aggregate in this phase due to glycoprotein driven interactions between flowing platelets and proteins deposited from the whole blood onto the silicone surface. There is no coating in the cartridge, but from confocal immunofluorescent staining of the platelet plugs on the sensors platelets completely bind to the surface through vWF-GP(Ib) complex¹⁷. The second phase provides a steady stream of platelets and blood constituents without shear-aggregating additional platelets onto the nucleated platelet plug. The silicone sensors are fluorescently labeled and are viewed by a 5 MP image sensor (Ximea Corp™), with an image taken every 2 s.

Tests are considered valid if at least 7 measurable sensors are detectable at the start of the test and are subsequently detectable through the 5-min time course of the entire test and report platelet forces in the possible sensor force range. The sensors have a physical bending range of 0 nN to 1000 nN based on their physical dimensions (micropost diameter = 7.5 µm, height = 26 µm), material elasticity coefficients of (PDMS = 3.8 MPa), and distance between the surfaces of the cylindrical sensor micropost and the much larger and rigid sensor block. Erroneous readings outside this range, such as when a gas bubble or thrombus detachment causes sensors to rapidly move, are flagged and excluded from the data aggregation done by the ATLAS automatically.

Platelet force tests were validated by quantifying the number of sensors that were individually detectable per image frame. If there were at least 7 sensors that were in focus, analyzable in every frame, and did not drop out of detection for the duration of the test, this was a valid test. Low confidence, in the context of the ATLAS, is a function of how many of the total number of visible sensors can be analyzed by machine vision. There are 28 sensors in view of the camera, and we used a cutoff confidence of at least 30% of the sensors being fully analyzable from start to finish. Thus, if they were below this cutoff they were designated low confidence. Platelet force assays were defined as having extreme values if they reported a force beyond the maximal range and were automatically dropped from the readout of each individual test. Causes for either low confidence intervals or extreme values are typically loss of optical focus, sample contamination such as debris from needle punctures during draw or handling, or pre-clotting in the blood collection tube or access line leading to large platelet masses lodging against the sensors instead of forming due to shear activation.

These validated data result in a direct measurement of platelet adherence strength (in nN) and platelet aggregation rate (which both can be plotted against time)¹⁸. The two platelet measurements used in the ROC analysis are: (1) platelet adherence force, measured in nano-Newtons (range 0–1000 nN), derived from mechanical deflection of the silicone microposts; and (2) platelet aggregation rate, a unitless ratio calculated as the relative decrease in light intensity in the sensor area from baseline (range 0–1, where 0 = no aggregation and 1 = complete light blockade by platelet mass). The average platelet adherence measurement (0–1 unitless) was multiplied by the average strength measurement (0–1000 nN) for each sensor to create a combined platelet force aggregation measurement for a single sensor. These combined platelet force aggregations measurements were averaged for all fully analyzable sensors to generate a single score for a patient. These measurements served as our two primary predictors.

Outcomes

The primary outcome was life-threatening hemorrhage, a binary outcome defined as one or more of the following: (1) death after massive transfusion (greater than 10 units blood transfused in the first 24 h after hospital arrival); (2) death from hemorrhagic shock; or (3) surgery or interventional radiology for hemorrhage control, defined as surgery or angiography within the first 24 h of ED presentation for all participants needing any transfusion during this 24-h period. Death after massive transfusion or from hemorrhagic shock was determined by chart review by two physicians blinded to platelet measurements, with any discrepancy adjudicated by a third physician. Our secondary outcome was receiving greater than 2 units of packed red blood cells in the 24 h after ED arrival.

Statistical analysis

Standard descriptive statistics were used to characterize the cohort, including baseline demographics (age, sex, race/ethnicity), medical history (antiplatelet and anticoagulation usage), injury characteristics (mechanism of injury, Injury Severity Score²², pre-hospital blood component and crystalloid administration), and vitals (systolic blood pressure, heart rate, temperature) and labs (lactate, hematocrit, platelets, International Normalized Ratio, Partial Thromboplastin Time) at ED arrival. We examined the ability of platelet force aggregation via measurements of platelet adherence strength and platelet aggregation rate to predict both life-threatening hemorrhage and transfusion requirement, respectively, with multivariable logistic regression. Three predictive logistic regression models were compared: (1) a model of platelet force aggregation measures (platelet adherence strength and platelet aggregation rate) as described by the ATLAS microfluidic platform; (2) a composite of clinical parameters based on the Assessment of Blood Consumption (ABC) score; and (3) a combined model of both the platelet force aggregation and clinical parameters. The ABC score is a validated model of clinical measures utilizing systolic blood pressure and heart rate at presentation, positive Focused Assessment with Sonography for Trauma (FAST), and penetrating vs blunt mechanism to determine likelihood of requiring massive transfusion protocol²³. Within our study we utilized those portions of the ABC score which were readily available to us: systolic blood pressure, heart rate, and mechanism of injury (penetrating vs. blunt). Receiver operating characteristics (ROC) curves and the area under the curve (AUC) for life-threatening hemorrhage (primary) and significant transfusion volume (secondary) were generated from these models, including 95% confidence intervals (CI). Likelihood ratio tests were run to compare the nested models using the likelihood chi-square test statistic. All analyses were conducted using R version 4.4.0, including the pROC and rms extension packages.

Results

Cohort characteristics

From August 2022 to July 31, 2024, 761 participants were screened and deemed eligible for inclusion. A total of 19 participants were excluded for incarceration, pregnancy, or receipt of greater than 3 units of blood prior to arrival (Fig. 1). Of the remaining 742 participants, 93 were excluded due to having no injuries after trauma team evaluation, 108 for not enough blood drawn for the assay, 47 for having low confidence levels, and 12 for having extreme values on the assay (Fig. 1). Of those who were excluded for greater than 3 units of blood prior to arrival (N = 7), low confidence assays (N = 47), and extreme values on assay (N = 12) we found no difference in baseline characteristics (Supplemental Table 1). Those without enough blood drawn for the assay (N = 108) did not have demographics recorded and thus could not be compared. The 482 participants, 63% of the original cohort, who were included in our analysis had a median age of 47 years and were more frequently white (78.4%) males (71.8%) reflecting our trauma patient population. A total of 40 participants (8.3%) were on either single or dual antiplatelet therapy, and 25 participants (5.2%) were on oral anticoagulation prior to injury (Table 1). The cohort had a median systolic blood pressure of 129 mmHg and median heart rate of 91 bpm. A penetrating mechanism was present in 22.6% of participants and the overall median Injury Severity Score was 10 (Inter Quartile Range [IQR] 4–20). The participants were not hypothermic (36.6 Celsius, IQR 36.3–36.9) and median lactate at the time of arrival was within normal limits (1.9 mmol/L, IQR 1.2–2.8). Prior to blood draw for platelet force aggregation testing, 6.8% had packed red blood cells (pRBCs) transfused, 4.6% had plasma transfused, 0.4% had platelets transfused, 1.5% had whole blood transfused, and 22.2% had crystalloid fluid administered. Participant lab results at the time of study blood draw showed normal platelet counts (237.5 cells/μL, IQR 197.0–285.3) and coagulation studies that were within normal limits (International Normalized Ratio 1.1, IQR 1.0–1.2; Partial Thromboplastin Time 26.1, IQR 24.3–28.7). The non-transfused participants in the current study showed closely corresponding values to previously reported healthy donor controls (average force 160 nN, maximum force 310 nN), supporting the assay’s validity in an injured population¹⁹. Of participants receiving blood components after inclusion, a median of 3 units of pRBCs, 3 units of plasma, 1 unit pack of platelets, 1.5 five-packs of cryoprecipitate, and 2 units of whole blood were administered (Table 2).

Table 1.

Cohort demographics and clinical characteristics.

	N = 482 *Median (interquartile range) unless specified
Age (years)	47 (30–65)
Sex, N (%)
Female	136 (28.2%)
Male	346 (71.8%)
Race/Ethnicity
American Indian/Alaska Native	2 (0.4%)
Asian	2 (0.4%)
Native Hawaiian or Other Pacific Islander	1 (0.2%)
Black or African American	88 (18.3%)
White	378 (78.4%)
Hispanic or Latino (not exclusive of race)	48 (10.0%)
Antiplatelets
Aspirin	34 (7.1%)
Clopidogrel/Ticagrelor	15 (3.1%)
Anticoagulants
Warfarin	7 (1.5%)
Direct Oral Anticoagulant	18 (3.7%)
Mechanism of injury
Blunt	373 (77.4%)
Penetrating	109 (22.6%)
Injury Severity Score	10 (4–20)
Pre-hospital blood component administration
Packed red blood cells	33 (6.8%)
Plasma	22 (4.6%)
Platelets	2 (0.4%)
Cryoprecipitate	0 (0.0%)
Whole blood	7 (1.5%)
Pre-hospital crystalloid administered	107 (22.2%)
Initial vital signs
Systolic blood pressure (mmHg)	129 (110–141)
Heart rate (beats/min)	90.5 (77–111)
Temperature (°C)	36.6 (36.3–36.9)
Initial laboratory values
Lactate (mmol/L)	1.9 (1.2–2.8)
Hematocrit (%)	40.0 (35.0–43.0)
Platelets (cells/μL)	237.5 (197.0–285.3)
International normalized ratio	1.1 (1.0–1.2)
Partial thromboplastin time	26.1 (24.3–28.7)

Table 2.

Cohort outcomes.

	N = 482 *Median (interquartile range) unless specified
Primary outcome: Life-threatening hemorrhage	36 (7.5%)
Death from hemorrhagic shock	9 (1.9%)
Emergency procedure for hemorrhage control	33 (6.8%)
Secondary outcome: > 2 units transfused	43 (8.9%)
Packed red blood cells (units)	3.0 (2.0–4.5)
Plasma (units)	3.0 (2.0–4.0)
Platelets (packs)	1.0 (1.0–2.0)
Cryoprecipitate (five-pack units)	1.5 (1.0–3.0)
Whole blood (units)	2.0 (1.5–2.0)

Primary outcome

Our primary outcome of life-threatening hemorrhage was experienced by 36 participants (7.5%) with 9 participants dying from hemorrhagic shock and 33 participants undergoing an emergent procedure for hemorrhage control with overlap between categories (Table 2). Of the nine patients that died, the median time to death from hospital arrival was one day (IQR 0.08–2.30 days). The platelet force aggregation prediction model had an AUC of 0.61 (95% confidence interval [CI] 0.53–0.69), the clinical parameter model had an AUC of 0.83 (CI 0.75–0.91), and the combined model had an AUC of 0.85 (CI 0.79–0.92) (Fig. 2 panels A, B, and C, respectively). The optimal Youden cutoff for the platelet force aggregation model was 0.075 (sensitivity 0.67, specificity 0.55), clinical measures was 0.106 (sensitivity 0.69, specificity 0.84), and combined platelet force aggregation was 0.073 (sensitivity 0.86, specificity 0.75). There was no significant difference between the combined predictive model (platelet force aggregation and clinical parameters of the ABC score) and the model of clinical measures alone (P = 0.32).

Fig. 2.

Receiver operating characteristics for prediction of life-threatening hemorrhage (death from hemorrhagic shock or need for emergent procedure for hemorrhage control) by (A) platelet force aggregation measures (platelet aggregation strength and rate); (B) clinical measures (systolic blood pressure, heart rate, and penetrating mechanism); and (C) both platelet force aggregation and clinical measures. Shown are areas under the curve (AUC) with respective 95% confidence intervals. Likelihood chi-square test statistic comparisons: Model A v. B (p-value < 0.000), Model A v. C (p-value < 0.000), Model B v. C (p = 0.32).

Secondary outcome

The requirement of greater than 2 units of transfusion was experienced by 43 participants (8.9%). The platelet force aggregation model had an AUC of 0.68 (CI 0.61–0.76), the clinical parameters score model had an AUC of 0.84 (CI 0.79–0.90), and the combined model had an AUC of 0.88 (CI 0.83–0.93) (Fig. 3 panels A, B, and C, respectively). The optimal Youden cutoff for the platelet force aggregation model was 0.074 (sensitivity 0.84, specificity 0.48), clinical measures was 0.062 (sensitivity 0.93, specificity 0.65), and combined platelet force aggregation was 0.092 (sensitivity 0.86, specificity 0.80). The predictive model with both platelet force aggregation and clinical parameters of the ABC score was significantly better in its ability to predict transfusion compared to the model of clinical measures alone (P = 0.01).

Fig. 3.

Receiver operating characteristics for prediction of major transfusion (greater than 2 units of blood transfused from time of injury to first 24 h of hospitalization) by (A) platelet force aggregation measures (platelet aggregation strength and rate); (B) clinical measures (systolic blood pressure, heart rate, and penetrating mechanism); and (C) both platelet force aggregation and clinical measures. Shown are areas under the curve (AUC) with respective 95% confidence intervals. Likelihood chi-square test statistic comparisons: Model A v. B (p-value < 0.000), Model A v. C (p-value < 0.000), Model B v. C (p = 0.01).

Discussion

In this prospective cohort study of patients at high-risk for hemorrhage after injury, we evaluated the ability of platelet force aggregation to predict life-threatening hemorrhage and need for transfusion of blood components when compared to clinical assessments alone. Our cohort included primarily younger males who were not on previous antithrombotic agents and had a relatively low level of injury severity after sustaining most often a blunt mechanism of trauma. In this cohort, the addition of platelet force aggregation measures to established clinical assessments did not improve detection of life-threatening hemorrhage but did improve detection of patients requiring transfusion of greater than two units of blood components.

Hemorrhagic shock remains a clinical diagnosis, but early signs and symptoms of hemorrhage may be subtle and in need of additional means to identify these minor manifestations²⁴. Thus, a rapid, simple bedside test identifying early coagulopathy along with clinical measures is a current unmet need in clinical care. Prior research suggests using thromboelastography in a similar approach has been shown to improve mortality in injured patients with hemorrhagic shock, but has been cited as cumbersome and difficult to interpret^15,16. Our study’s findings add to this knowledge base by characterizing the specific contribution of platelet dysfunction to hemorrhage in traumatically injured patients using this novel platform. Platelet force aggregation, as discussed in this manuscript, is not meant to be used by itself but instead to complement clinical decision making using established clinical assessment tools such as the ABC score. Thus, it was expected the clinical parameters from the ABC score outperformed measures of platelet force aggregation alone in prediction of both our primary and secondary outcomes.

The key finding within the study is the ability of platelet force aggregation to improve the ability of clinical parameters alone to predict the need for transfusion of greater than two units of blood. While those suffering from hemorrhagic shock remain a rather straightforward clinical diagnosis, those who are in earlier stages of hemorrhage show less vital sign abnormalities thus making it harder to discriminate on clinical parameters alone²⁴. The usage of this adjunct assay to improve the detection of these early signs and symptoms of hemorrhage is clinically important. It is possible the discriminative ability of this test is better in less severe shock (e.g., Class I or II rather than Class III or IV) where vital signs can be relatively normal and FAST exams are falsely reassuring. Using the Youden-optimal cutoff, the combined model achieved a sensitivity of 86% and specificity of 80%—compared to 93% sensitivity and 65% specificity for clinical parameters alone. This specificity gain is potentially meaningful in any level of practice, as it would reduce unnecessary early activation of massive transfusion protocols in our population where the majority (> 91%) do not ultimately require significant transfusion, while preserving sensitivity for those who do. Further, this may be especially important in community or rural settings where traumatically injured patients of any acuity level (level I/II/III) may present before the vital sign changes of significant hemorrhage manifest. Results could inform a clinician in this setting in assisting diagnosis of early hemorrhage, prompting rapid transfer to a higher level of care for definitive hemorrhage control, and utilizing blood components judiciously. The relative value of this specificity gain will vary by practice setting and resource availability, as the cost of a false positive (unnecessary massive transfusion protocol activation) and a false negative (missed hemorrhage) are not equivalent and differ substantially between rural and urban trauma centers. The Youden index, while a useful general metric, assumes equal misclassification costs and a 50% outcome prevalence—assumptions that do not hold in this cohort where life-threatening hemorrhage occurred in fewer than 10% of participants. A cost-approach framework, which explicitly models the asymmetric costs of false positive and false negative results against local prevalence and resource constraints, would better optimize thresholds for context-specific clinical implementation and could be considered in future prospective implementations of this assay^25,26.

While we studied traumatically injured patients due to their inherent substantial risk of hemorrhage and relatively large numbers of eligible individuals, we believe these results should be further evaluated in the non-injured population at risk for hemorrhage. Current management paradigms for damage control resuscitation have not been extensively tested in the non-injured population regardless of their similar manifestations²⁷. Use of TEG to guide blood component resuscitation has been investigated in other populations, such as patients with cirrhosis undergoing invasive procedures²⁸, those undergoing cardiac surgery²⁹, and those with postpartum hemorrhage³⁰. Future studies of the ATLAS platform should aim to test the utility of platelet force aggregation measures as a complementary tool to clinical assessment in other patient populations at risk for life-threatening hemorrhage. In the longer term, if validated across broader populations, the platform’s reagent-free operation and rapid result time may position it as a more user-friendly alternative to existing viscoelastic technologies.

While there were several strengths of this study including its size, well-defined clinically relevant outcomes, and prospective inclusion of patients at risk for life-threatening hemorrhage at a level I trauma center, there were also limitations. This was a feasibility study of our ability to identify critically injured participants, collect blood samples, and perform ATLAS testing in an expedited fashion. As such, a formal sample size assessment was not performed and thus there is a possibility of underpowering. The clinical significance of the combined model, while statistically significant, was only a small incremental gain in clinical accuracy; however, a larger sample size with a greater proportion of clinical outcomes would provide better discrimination of the accuracy and clinical utility of platelet dysfunction in traumatically injured patients at risk for hemorrhage requiring transfusion. Additionally, a substantial proportion of screened patients (N = 108, 14%) were excluded due to insufficient blood volume for the assay, reflecting the practical challenges of conducting research during acute trauma resuscitation, which may limit the generalizability of these findings to real-world implementation. Other limitations include the lack of a validation cohort to ensure reproducibility of our findings and no comparison with other contemporary measures of platelet function assays or TEG.

Conclusions

In a prospective cohort study of traumatically injured patients at high-risk for hemorrhage, a novel, rapid platelet force aggregation assay increased the accuracy of bedside clinical measures to predict the need for transfusion but did not increase the accuracy of the same clinical measures for identifying life-threatening hemorrhage. We believe this indicates more utility of this assay in early hemorrhagic shock when vital signs may be normal and subclinical hemorrhage is occurring. Future work should validate the clinical utility of this technology in a larger cohort of traumatically injured patients as well as in other populations at risk for non-traumatic hemorrhage.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ED

Emergency Department

vWF

Von Willebrand factor

TEG

Thromboelastography

ROTEM

Rotational thromboelastometry

ABC score

Assessment of blood consumption (ABC) score

FAST

Focused Assessment with Sonography for Trauma

CI

95% Confidence interval

ROC Curve

Receiver operating characteristics curve

AUC

Area under the curve

IQR

Inter quartile range

pRBC

Packed red blood cell

Author contributions

MBP, LT, and SC conceptualized the project. LT provisioned all resources and software for the project. SR, TM, KM, PW, AT, SP, KH, SD, AL, and JL assisted in data acquisition and curation. CH and MS performed formal statistical analysis. SR and TM wrote the original draft. All authors performed critical review and editing of the manuscript.

Funding

This work is supported by NHLBI 1R61HL156508.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

TM and SR are supported by NIH T32GM135094. SPC reports funding from NIH, PCORI, DOD and Becton Dickinson and consulting with Reprieve Cardiovascular, Prenosis, Corteria, and Sequana. MBP is a part-time employee of the US Government (Veterans Affairs) and receives grant support from NIH (T32GM135094, R01AG058639, R01 GM120484), Veterans Health Administration, Rehabilitation Research and Development Service (I01 RX002992), DOD (W81XWH-16-D-0024-0001, DoD W81XWH-21-PRMRP-CTA), honoraria from NIH CSR for study section participation, Elsevier, and Behring and serves on the Scientific Advisory Board for Liberate Medical. MBP has received unrestricted institutional research support provided by Vanderbilt University Ingram Chair in Surgical Sciences. LT, AL, JL are employees of Stasys Medical. MS, SR, TM, CH, SP, AT, KM, SD and KH report no COI. The authors declare no competing interests.

Ethics approval and consent to participate

We confirm that the work covered by the manuscript has been conducted with the ethical approval of all relevant bodies. This was approved by the local IRB with a waiver of informed consent.