Effect of PCC on thrombin generation in patients on FXaI with bleeding or needing urgent surgery (GAUGE)

Key Points

• •FXaI bleeding is common, and reversal options are limited. PCC is used off-label, but in vivo effects on thrombin generation are unclear.
• •PCC increased thrombin generation but did not affect LT; higher baseline FXaI levels blunted the effect of PCC on thrombin generation.

Visual Abstract

Abstract

Prothrombin complex concentrates (PCC) are used off-label to treat direct factor Xa inhibitor (FXaI)–associated bleeding or to optimize hemostasis before urgent surgery. The Global Assessment of Hemostatic TGA Indices Following the Use of Prothrombin Complex Concentrates for Major Bleeding or Urgent Surgery in Patients Treated with Factor Xa Inhibitors (GAUGE) study is a prospective observational cohort study of the effects of PCC on thrombin generation and hemostasis in FXaI-treated patients. FXaI-treated patients who presented with major bleeding or needed urgent surgery received PCC (50 IU/kg). Platelet-poor plasma was collected before and after PCC administration. Thrombin generation assay parameters, including lag time (LT), time to peak (TTP), peak thrombin generation (peak), endogenous thrombin potential (ETP), and mean velocity rate index (mVRI) were measured using calibrated automated thrombography. Hemostatic efficacy, thromboembolism, and mortality were adjudicated in duplicate. Multivariable log-linear regression was used to evaluate PCC effects on FXaI levels. Unsupervised k-means clustering was used to identify patients with the largest ETP increment. GAUGE included 101 episodes of FXaI-associated PCC use (FXaI-associated bleeding = 71; urgent surgery = 30). The median PCC dose was 48.0 IU/kg (interquartile range [IQR], 42.0-50.0). PCC did not affect FXaI levels (P > .05). PCC increased ETP (Δ868.9 ± 766.6 nM/min), peak (Δ85.7nM ± 108.2nM), and mVRI (Δ16.5 nM/min [IQR, 4.4-45.0]). PCC did not affect the LT or TTP (P > .05). Patients in cluster 1 had significantly lower pre-PCC FXaI levels (79 ng/mL [IQR, 55-93] vs 165 ng/mL [IQR, 77-219]; P = .009) and a greater ΔETP increment (P < .0001) than patients in cluster 2. Effective hemostasis was achieved in 50.0% of major bleeding events (95% confidence interval [CI], 38.4-61.6), and procedural hemostasis was normal in 76.7% of urgent surgeries (95% CI, 59.1-88.2). The 30-day risks of thromboembolism and mortality were 6.9% (95% CI, 3.4-13.6) and 12.9% (95% CI, 7.7-20.8), respectively. PCC increased quantitative thrombin generation without shortening LT or TTP, supporting a prohemostatic effect rather than true FXaI reversal.

Introduction

The oral direct factor Xa inhibitors (FXaIs), including apixaban, rivaroxaban, and edoxaban, are commonly used for stroke prevention in atrial fibrillation and for the treatment and secondary prevention of venous thromboembolism.^1,2 Although FXaIs have a more favorable bleeding profile than vitamin K antagonists, major bleeding events still occur in 2% to 4% of FXaI-treated patients every year.³ Anticoagulation-associated bleeding events lead to substantial morbidity and mortality, with case-fatality rates ranging from 8% to 15%.³ Management is challenging in patients receiving FXaIs who need emergent surgery because surgery often cannot be delayed until drug clearance, and there is no widely available, approved reversal strategy specifically indicated for this scenario.

Andexanet alfa is a recombinant modified FXa decoy protein that binds FXaIs and rapidly reduces FXaI levels. In the ANNEXA-4 study on patients with FXaI-associated bleeding, andexanet resulted in marked reductions in anti-Xa activity and was associated with hemostatic efficacy in 82% of the patients.⁴ More recently, the randomized ANNEXA-I trial on patients with FXaI-associated intracerebral hemorrhage showed that andexanet improved control of hematoma expansion compared with usual care.⁵ However, a growing body of research supports that andexanet is associated with a consistent and clinically important risk of ischemic stroke.4, 5, 6, 7 High cost, limited availability, and the absence of a regulatory indication for urgent surgery have limited the use of andexanet and bolstered the need for alternative, widely accessible strategies for FXaI reversal.⁸ Andexanet was withdrawn from the United States market after a Food and Drug Administration safety communication.⁹

Four-factor prothrombin complex concentrates (PCC) are widely used off-label as nonspecific prohemostatic agents for FXaI-associated bleeding or to optimize hemostasis before urgent surgery.^10,11 Prospective and retrospective cohort studies have reported hemostatic efficacy ranging from 60% to 80% in FXaI-treated patients with bleeding, with thromboembolic event risk estimates of 6% to 8%.12, 13, 14 The PROXY study suggested that a fixed 2000 IU dose of PCC administered before urgent surgery in FXaI-treated patients with FXaI levels ≥75 ng/mL yields near-normal surgical hemostasis and a low risk of postoperative thrombosis.¹⁵

Quantifying thrombin generation can provide a comprehensive measure of the coagulation phenotype.¹⁶ Thrombin generation assays (TGA) continuously measure thrombin production using a fluorogenic substrate and provide information on both kinetic and quantitative parameters of thrombin generation. Kinetic parameters include lag time (LT; time for thrombin generation to first occur in minutes) and the time to peak (TTP; time from start to maximal thrombin generation in minutes), whereas quantitative parameters include peak thrombin generation (peak; maximal thrombin generation in nanomolar) and endogenous thrombin potential (ETP; area under the thrombin generation curve; nanomolar per minute). The mean velocity rate index (mVRI) is a composite parameter whereby the peak is divided by the time difference between TTP and LT (peak/[TTP − LT]).^17,18 FXaIs lead to progressive prolongation of TGA kinetic parameters and suppression of peak, whereas ETP remains relatively preserved at lower FXaI levels.¹⁷ Eerenberg et al conducted a randomized crossover trial in healthy volunteers showing that PCC normalized ETP and prothrombin time (PT) in rivaroxaban-treated participants, establishing proof of concept that PCC can restore thrombin generation in the presence of an FXaI.¹⁹ In vitro data support that PCC increases quantitative thrombin generation parameters in FXaI-anticoagulated plasma but often incompletely corrects LT and TTP and frequently raises ETP above the normal range.¹⁷ However, there are limited data on the in vivo effects of PCC on thrombin generation in patients with FXaI-associated bleeding or those needing urgent surgery.

The Global Assessment of Hemostatic TGA Indices Following the Use of Prothrombin Complex Concentrates for Major Bleeding or Urgent Surgery in Patients Treated with Factor Xa Inhibitors (GAUGE study) was designed to address this knowledge gap. GAUGE is a prospective, single-center cohort study that enrolled FXaI-treated patients with major bleeding or those needing urgent surgery who received PCC as part of routine care. The objectives of the GAUGE study were to characterize the effects of PCC on TGA parameters and measure clinical hemostatic efficacy and thromboembolic complications in FXaI-treated patients.

Methods

Study design and population

We conducted a single-center, prospective observational cohort study (GAUGE) at The Ottawa Hospital, a tertiary academic hospital in Ottawa, Canada. The study design, rationale, and analysis plan are described elsewhere.²⁰ Adult patients, aged ≥18 years, were eligible if they were prescribed an oral FXaI (apixaban, rivaroxaban, or edoxaban) for any indication and presented with bleeding warranting administration of PCC or required urgent surgery. In all cases, the treating physician or surgeon had decided to administer PCC (Octaplex; Octapharma) to optimize hemostasis as part of standard management. Patients with dropping hemoglobin levels without an identifiable bleeding source or patients who had received activated PCC/recombinant factor VIIa were excluded. Institutional protocols recommended a PCC dose of 50 IU/kg based on actual body weight, with no maximum dose. Dosing was at the discretion of the treating physician. Potentially eligible episodes were identified through a centralized transfusion medicine database, which was populated whenever PCC was issued. An electronic PCC order set implemented across The Ottawa Hospital standardized PCC dosing and recommended blood draws within 4 hours before and 4 hours after PCC administration. Research staff screened PCC releases within 24 hours to confirm FXaI exposure, indication, and eligibility. Deferred written consent was obtained for ongoing data collection and use of stored plasma, followed by prospective baseline data collection and 30-day follow-up.

Ethics approval

This study was approved by the Ottawa Health Science Network Research Ethics Board. Deferred written informed consent for participation, data collection, and use of stored plasma was obtained from all participants or their substitute decision-makers in accordance with the Declaration of Helsinki.

Data, sample collection, and laboratory testing

Baseline clinical data included demographics, comorbidities, FXaI type and dose, timing of last FXaI intake, and renal function (creatinine clearance, milliliter per minute; calculated using the Cockcroft-Gault equation). For bleeding presentations, we recorded the bleed site and whether the bleeding event met the International Society on Thrombosis and Haemostasis (ISTH) definition of major bleeding.²¹ Pre-PCC and post-PCC intracranial hemorrhage (ICH) hematoma volumes were quantified using Quantomo software.²² For urgent surgeries or procedures, we documented the procedure type and the timing of PCC administration relative to the start of the procedure. Data related to clinical course included hemodynamic parameters, hemoglobin trends, blood products administered, and details of hemostatic procedural interventions. Participants were followed for 30 days after PCC administration. Follow-up was conducted by telephone interviews and chart reviews.

Platelet-poor plasma (PPP) for thrombin generation measurements was obtained from residual plasma samples collected as part of routine care. Blood was collected by venipuncture into 2.7-mL, 3.2% sodium citrate tubes (0.109 mol/L; Becton Dickinson). Samples were centrifuged at 1500g for 15 minutes at room temperature to generate PPP, then aliquoted and stored at −80°C until they were shipped to a central research laboratory (Diagnostica Stago, Gennevilliers, France). Institutional centrifugation protocols at The Ottawa Hospital undergo regular quality assurance to ensure PPP production (target residual platelet count <10 × 10⁹/L) and are reviewed monthly by transfusion medicine laboratory charge technologists with routine auditing and quality checks. All samples used to generate PPP were centrifuged and aliquoted within 4 hours of collection.

Laboratory data included hemoglobin (gram per liter) and platelet count (×10⁹/L), PT (reference range, 10.0-14.0 seconds; Thromborel S), activated partial thromboplastin time (19.0-28.0 seconds; Actin FS), fibrinogen level (Clauss method; reference range, 1.9-4.5 g/L; Dade Thrombin Reagent), D-dimer (reference range, <500 ug/L; Innovance D-dimer kit), and FXaI levels. FXaI levels (nanogram per milliliter) were measured using Innovance Anti-Xa (Siemens Healthcare Diagnostics, Marburg, Germany) with STA-apixaban/rivaroxaban calibrators (Diagnostica Stago, Asnière-sur-Seine, France).

Thrombin generation

Thrombin generation was measured using calibrated automated thrombography/thrombinoscope on PPP samples (Diagnostica Stago, Gennevilliers, France), with personnel being blinded to participant characteristics and pre-PCC/post-PCC sample status. Fluorescence was measured using a Fluoroskan Ascent fluorometer (Thermo Fisher Scientific, Helsinki, Finland). TGA was measured using 5pM tissue factor (PPP Reagent) and 4μM phospholipids. Reactions were initiated by adding 20 μL of FluCa buffer to wells containing 80 μL of PPP. Reactions were carried out in duplicate, and averaged values from duplicate runs were recorded. TGA parameters measured included LT (minutes [min]), TTP (minutes), peak (nanomolar; nM), ETP (nM•min), and mVRI (nM/min). Reference ranges for each parameter were derived from a panel of 30 healthy volunteers (aged 24-55 years) without known bleeding or thrombotic disorders, yielding nonparametric 95% reference intervals for LT (2.7-3.3 minutes), TTP (4.5-6.0 minutes), peak (220nM-394nM), ETP (981-1850 nM•min), and mVRI (94-205 nM/min).²⁰ For the primary analyses, post-PCC normalization of thrombin generation was defined as a TGA parameter value within the respective reference intervals specified above.

Because Octaplex contains heparin, which can suppress thrombin generation, we conducted a prespecified sensitivity analysis in which the PPP reagent was reconstituted with polybrene solution (0.025 mg/mL) to neutralize heparin. TGA was repeated on polybrene-treated pre- and post-PCC samples.

Composite thrombograms with corresponding 95% confidence interval (CI) bands for pre-PCC and post-PCC measurements were generated by reexpressing individual thrombogram values on a standardized 0- to 60-minute grid with 0.1-minute resolution using linear interpolation between observed measurement points. Interpolated curves were then averaged pointwise to generate composite pre-PCC and post-PCC thrombograms, with the number of contributing patients varying by time point according to observed signal coverage. At each time point, 95% CI were calculated using cross-sectional standard errors derived from all available subject-level curves.

Hemostatic and thrombotic outcomes

Hemostatic efficacy, thromboembolic events, and death were adjudicated in duplicate by 2 independent hematologists. For patients receiving PCC for major bleeding events, hemostasis was adjudicated using both ISTH criteria for effective hemostasis (effective vs ineffective) and a modified version of the criteria used by Sarode et al (good/moderate vs poor; supplemental Appendix).^23,24 Adjudicators were blinded to thrombin generation results. For patients receiving PCC before urgent surgery or invasive procedures, surgical hemostasis was assessed by the treating surgeon using a binary normal/abnormal scale, with abnormal defined as a concern that bleeding exceeded expectations for the procedure. Procedures with abnormal hemostasis meeting the ISTH definition for surgical major bleeding were recorded.²⁵

Thromboembolic events that occurred within 30 days of PCC administration included deep vein thrombosis, pulmonary embolism, ischemic stroke, transient ischemic attack, systemic arterial embolism, myocardial infarction, and valvular or cardiac-chamber thrombosis, defined using standard criteria (supplemental Appendix), with each event classified as definitely, likely, possibly, or unlikely to be related to PCC. All-cause mortality at 30 days was recorded and categorized by adjudicators as caused by fatal bleeding, fatal thromboembolism, or other causes.

Statistical analysis

Continuous data were summarized using means and standard deviations for normally distributed data, or medians and interquartile ranges (IQR) for nonnormally distributed data. The normality of distributions was assessed with the Shapiro-Wilk test. Categorical data were summarized using counts and percentages. Clinical outcomes are reported as proportions with corresponding 95% CI (Wilson score method for binomial proportions). FXaI levels, D-dimer, and fibrinogen values below the limit of detection (LoD) were substituted with LoD/2. Values greater than the upper LoD were substituted with values equal to the LoD.

The primary end point was the mean paired difference in LT between pre-PCC and post-PCC measurements. LT was selected a priori because it reflects the initiation phase of coagulation, correlates with FXaI levels, and addresses the key unanswered mechanistic question of whether PCC can shorten FXaI-prolonged LT and TTP rather than simply augmenting peak and ETP, as previous in vitro research has shown that PCC increases peak and ETP more consistently.²⁰ Sample size assumed a 2-sided paired t test with α of 0.05, 100 paired LT measurements, and a standard deviation of the paired differences of 2.0 minutes, providing 90% power (1 − β) to detect a mean change (detectable alternative) in LT of 0.65 minutes.²⁰ Secondary laboratory end points included paired differences for TTP, peak, ETP, and mVRI as well as the proportion of participants whose post-PCC values for each parameter fell within predefined reference ranges. For normally distributed data, pre- vs post-PCC differences were analyzed using paired t tests; for nonnormal data, the Wilcoxon signed-rank test was used. Correlations between pre-PCC FXaI levels and TGA parameters were assessed using Spearman correlation coefficients.

Exploratory analyses were undertaken to evaluate the effects of PCC on FXaI levels and determine whether the impact of PCC on TGA parameters was influenced by baseline FXaI levels. No formal adjustment for multiple comparisons was applied because of the exploratory nature of these analyses. All P values <.05 were considered statistically significant. First, we used multivariable log-linear regression (restricted to participants with paired pre- and post-PCC FXaI levels) to determine whether PCC exerted an effect on FXaI levels independent of normal time-dependent FXaI clearance. Δ-time, defined as the total elapsed time between the pre- and post-PCC draws, was included as a covariate to account for expected pharmacokinetic elimination of FXaIs over time. The model was also adjusted for baseline FXaI level. Second, an unsupervised machine learning approach using k-means clustering was used to identify clinically interpretable patterns in thrombin generation response to PCC. Clustering was conducted on 3 ETP features (pre-PCC ETP, post-PCC ETP, and ΔETP) using z score standardization to ensure equal weighting. Clustering was carried out using the Lloyd-Forgy algorithm and Euclidean distance (k = 3; random initializations = 10, fixed random seed). Only patients with complete pre-PCC ETP, post-PCC ETP, ΔETP, and pre-PCC FXaI levels were included in the clustering data set (n = 35). For clinical interpretability, clusters were subsequently ordered according to cluster-specific mean ΔETP. Patients in the cluster with the largest positive ΔETP (cluster 1) were designated the high-ΔETP group (larger increase in thrombin generation after PCC). Patients in the cluster with smaller ΔETP values (cluster 2) were designated the low-ΔETP group. A third cluster (n = 2), characterized by very high baseline pre-PCC ETP and a paradoxical decrease in ETP after PCC, was not evaluated further because its small size limited interpretability and was of uncertain clinical significance. Baseline pre-PCC anti-Xa were summarized for each cluster to contextualize anticoagulant intensity. All statistical analyses were performed using R Statistical Software (R Core Team 2025 version 4.5.2; R Foundation for Statistical Computing, Vienna, Austria).

Results

Patient characteristics

Between June 2021 and October 2024, 533 PCC release records were screened. A total of 101 FXaI-associated episodes in 100 patients met eligibility criteria and were included in GAUGE (Table 1), comprising 71 major bleeding episodes and 30 episodes in which PCC was given before urgent surgery (supplemental Figure 1). The most common indication for anticoagulation was atrial fibrillation (n = 75 [74.3%]; median CHADS₂ score, 2.0 [IQR 2.0-3.0]), followed by venous thromboembolism (n = 22). Creatinine clearance was estimated to be <30 mL/min at presentation in 20 patients. Among major bleeding episodes, 29 of the 71 episodes (41%) involved ICH, 15 gastrointestinal bleeding (21%), 3 genitourinary bleeding (4%), and 24 bleeding events at other sites (34%). The median Glasgow Coma Scale score among patients with ICH was 15 (IQR, 14-15). Overall, 69 of the 71 bleeding episodes (97%) met ISTH criteria for major bleeding, and 31 of the 71 episodes (44%) required a surgery or procedure to control bleeding. General surgery was the most common (50.0%) procedure type in patients receiving PCC for urgent surgery, followed by orthopedic surgery (13.3%) and interventional radiology (10.0%).

Table 1.

Patient characteristics

Parameter	Major bleeding (n = 71)	Urgent surgery (n = 30)
Age, y	80 (72-85)	71 (64-81)
Male, n (%)	34 (47.9)	17 (56.7)
Weight, kg	75 (65-93)	87 (77-103)
Creatinine clearance,∗ mL/min	59 (40-80)	78 (37-112)
Antiplatelet therapy, n (%)
ASA	5 (7.0)	1 (3.3)
Clopidogrel	2 (2.8)	1 (3.3)
FXaI, n (%)
Apixaban	45 (63.0)	22 (73.3)
Rivaroxaban	23 (32.0)	6 (20.0)
Edoxaban	3 (4.0)	2 (6.7)
Bleeding sites, n (%)
ICH	29 (40.8)
Non-ICH	42 (59.2)
Procedure type, n (%)
General surgery		15 (50.0)
Orthopedic surgery		4 (13.3)
IR		3 (10.0)
Dental		1 (3.3)
Cardiac		1 (3.3)
Endoscopic		1 (3.3)
Vascular surgery		1 (3.3)
Gynecology		1 (3.3)
Thoracic surgery		1 (3.3)
Urological		1 (3.3)
Neurosurgical		1 (3.3)

Continuous data are summarized using median (IQR), and categorical data are summarized as counts (n) and proportions (%).

ASA, acetylsalicylic acid; IR, interventional radiology.

^∗Calculated using the Cockcroft-Gault equation.

PCC dosing and blood products

The median initial PCC dose was 3500 IU (IQR, 3000-4000), corresponding to a weight-based dose of 48 IU/kg (IQR, 42-50; Table 2). The median approximate time between last FXaI intake and PCC administration for the overall cohort was 19 hours (IQR, 14-27) among patients for whom timing could be ascertained (n = 91).

Table 2.

PCC dosing and timing of administration

Parameter	Overall (N = 101)	Major bleeding (n = 71)	Urgent surgery (n = 30)
PCC dosing, IU	3500.0 (3000-4000)	3500.0 (3000-4000)	4000.0 (3125-4500)
Weight-based PCC dosing, IU/kg	48.0 (42.0-50.0)	48.2 (41.1-50.0)	46.7 (42.7-49.3)
Time from last FXaI dose to PCC∗, h	19.4 (13.8-27.4)	18.6 (13.7-26.1)	23.3 (15.1-28.8)
Time from presentation to PCC, h		5.5 (3.2-10.1)
Time from PCC to procedure start, h			1.2 (0.1-2.0)

Values are reported as median (IQR).

^∗Timing of the last FXaI dose was approximated to the nearest hour based on patient/substitute decision-makers-reported timing of the last intake. Time from hospital presentation to PCC was approximated based on the date/time of presentation to triage at the emergency department (outpatients) or the approximate date/time of bleed onset for patients admitted to the hospital.

Packed red blood cells were transfused in 38 episodes. Among non-ICH bleeding episodes, 30 of the 42 patients (71%) received packed red blood cells, with a median of 3 units (IQR, 2-5) among patients who received transfusions. Platelet transfusions were administered in 8 episodes, including 5 episodes in patients who were experiencing thrombocytopenia at the time of their bleeding event (platelet count, <100 × 10⁹/L). Plasma was administered in 11 episodes. Tranexamic acid was used in 31 episodes, most commonly as 1 to 2 1 g IV boluses.

Conventional coagulation testing, D-dimer, fibrinogen, and FXaI levels

PCC administration was associated with a statistically significant shortening of the PT (P < .0001) and activated partial thromboplastin time (P = .004). PCC had no effect on D-dimer (P = .67) or fibrinogen levels (P = .25; supplemental Figure 2).

Pre-PCC–calibrated anti-Xa levels (n = 63) had a median of 91 ng/mL (IQR, 66-180). Post-PCC anti-Xa levels (n = 66) had a median of 70 ng/mL (IQR, 49-132). Paired pre- and post-PCC FXaI levels were available for 48 episodes (Figure 1). After adjusting for the effect of Δ-time between sampling and baseline FXaI level using multivariable log-linear regression, PCC did not affect FXaI levels (P = .60). Longer duration between testing and PCC administration was associated with a greater proportional decrease in FXaI levels (P = .0002).

Figure 1.

Effect of PCC on FXaI levels. Paired pre- and post-PCC FXaI levels were available for 48 episodes. The median total time between pre- and post-PCC samples was 5 hours (IQR, 3-7). In this paired subset, the median pre-PCC anti-Xa level was 93 ng/mL (IQR, 70-167) and the median post-PCC level was 72 ng/mL (IQR, 52-135). The median within-patient change was −21 ng/mL (IQR, −58 to −12; P < .0001).

Thrombin generation

Baseline FXaI levels were positively correlated with LT and TTP and inversely correlated with peak thrombin, ETP, and mVRI (Figure 2). In the overall cohort, 63 participants had evaluable pre-PCC TGA measurements, and 57 participants had post-PCC measurements (Figure 3), yielding 55 paired samples for evaluation (Table 3). Participants with paired TGA samples were broadly comparable to those without paired TGA testing, with only small between-group differences in baseline clinical characteristics (supplemental Table 1). PCC significantly increased ETP, peak, and mVRI but did not affect LT or TTP (Figure 4A). Among patients with post-PCC TGA measurements, normalization of TGA parameters was most frequent for peak and ETP (Figure 3). Results following the addition of polybrene to samples were unchanged (supplemental Table 2).

Figure 2.

FXaI levels and TGA parameters. Scatterplots depicting the relationship between baseline (before PCC) FXaI level and each TGA parameter (n = 37 for all pairs). (A) LT (minutes); (B) TTP (minutes); (C) peak (nanomolar); (D) ETP (nM·min); (E) mVRI (nanomolar per minute). FXaI levels were strongly correlated with LT (ρ, 0.52; P = .0009), TTP (ρ, 0.52; P = .0008), peak (ρ, −0.60; P < .0001), and mVRI (ρ. −0.58; P = .0002), and moderately correlated with ETP (ρ, −0.48; P = .002).

Figure 3.

Impact of PCC on TGA parameters. Box and whisker plots depicting the impact of PCC on each TGA parameter. The central box line represents the median, lower and upper edges represent the first and third quartiles, whereas box plot whiskers represent the minimum and maximum values. The yellow dashed lines represent the upper limit of normal and lower limit of normal boundaries based on a reference interval established in healthy volunteers.²⁰ Using prespecified reference ranges, post-PCC LT was within the normal range in 12% of the patients, TTP in 4%, peak in 39%, ETP in 39%, and mVRI in 21%. Post-PCC ETP exceeded the upper limit of normal in 26 of the 57 patients (46%), indicating overcorrection in nearly half of those patients with ETP measurements.

Table 3.

Effects of PCC on thrombin generation

Parameter	Before PCC	After PCC	Δ (before and after)	P value
LT, min	4.9 (4.0-6.4)	5.1 ± 1.5	–0.3 (–1.0 to 0.5)	.19
TTP, min	7.7 (6.7-9.3)	8.9 ± 2.3	0.3 (–0.8 to 1.2)	.31
Peak, nM	140 (82-190)	219 ± 113	86 ± 108	<.0001
ETP, nM·min	914 (748-1101)	1719 (1174-2493)	869 ± 767	<.0001
mVRI, nM/min	47 (25-64)	60 (29-93)	17 (4-45)	<.0001

Thrombin generation parameters among patients with available pre-PCC (n = 63), post-PCC (n = 57), and paired TGA measurements (after PCC − before PCC; n = 55). Pre-PCC to post-PCC sampling for thrombin generation measurements occurred over a median of 5.0 hours (IQR, 3.0-7.0). Values are reported as mean ± standard deviation for normally distributed variables and median (IQR) for nonnormally distributed variables. Δ denotes (after PCC) − (before PCC). P values are from paired t tests or Wilcoxon signed-rank tests, as appropriate.

Figure 4.

Impact of PCC on thrombin generation. (A) Overall effects of PCC on thrombin generation. Composite thrombograms were generated using all available pre-PCC (n = 63) and post-PCC (n = 57) records for which TGA data were available. (B) Cluster 1, high-ΔETP group (n = 12) and (C) cluster 2, low-ΔETP group patients (n = 21) were identified using k-means clustering of patients for whom baseline FXaI levels and both pre-PCC and post-PCC ETP values were available. Patients in cluster 1 had lower baseline FXaI levels (79 ng/mL [55-93]) than cluster 2 (165 ng/mL [77-219]; P = .009), and correspondingly higher pre-PCC ETP values (1140 nM·min [1086-1311] vs 737 nM·min [657-835]; P < .0001). PCC administration was associated with a larger ΔETP among patients in cluster 1 compared with patients in cluster 2 (+1715 nM·min [1418-1905] vs +454 nM·min [325-813]; P < .0001), along with higher median post-PCC ETP values (2849 nM·min [2687-3018] vs 1185 nM·min [958-1644]; P < .0001).

Among the 35 patients with complete data on TGA parameters and FXaI levels included in the k-means clustering analysis, 2 meaningful response groups were identified (Figure 4B-C), including cluster 1 participants (n = 12; high-ΔETP group) and cluster 2 participants (n = 21; low-ΔETP group). Cluster 1 participants had lower baseline FXaI levels than cluster 2 participants and correspondingly higher pre-PCC ETP values. PCC administration was associated with a larger ΔETP among cluster 1 participants and higher median post-PCC ETP values.

Clinical outcomes

Using ISTH criteria, hemostasis was assessable in 68 of the 71 major bleeding episodes and was adjudicated as effective in 34 of the 68 episodes (50.0%; 95% CI, 38.4-61.6). Hemostasis was deemed assessable in 66 of the 71 cases using the Sarode criteria and was adjudicated as good/moderate hemostasis in 44 of the 66 episodes (66.7%; 95% CI, 54.7-76.8). Among the 30 cases in which PCC was used to optimize hemostasis before urgent surgery, procedural hemostasis was rated as normal in 23 (76.7%; 95% CI, 59.1-88.2). Among the 7 procedures with abnormal hemostasis, 4 met the ISTH definition of surgical major bleeding. Pre-PCC FXaI levels were available in 6 of these 7 cases and were comparable to those in the entire study cohort (median, 79 ng/mL [IQR, 74-148]; range, 39-233). These cases included percutaneous nephrostomy tube insertion, endoscopic retrograde cholangiopancreatography with sphincterotomy, hip revision with hardware removal/ open reduction and internal fixation, below-knee amputation, femur open reduction and internal fixation, cervical spine (C2-C6) posterior decompression and fusion, and cystoscopy with clot evacuation/fulguration.

The 30-day risk of thromboembolism was 6.9% (95% CI, 3.4-13.6). Event types included 2 deep vein thrombosis, 1 pulmonary embolism, 3 strokes, and 1 patient who experienced a multifocal embolic event with both stroke and systemic embolism, occurring at a median of 17 days after PCC administration (IQR, 8-24), with events occurring on days 0, 1, 14, 17, 23, 24, and 28 (supplemental Table 3). Therapeutic anticoagulation had not been restarted in any of these cases. The relationship between all thromboembolic events and PCC administration was rated as unlikely.

A total of 13 patients died, corresponding to a 30-day risk of mortality of 12.9% (95% CI, 7.7-20.8). Deaths occurred at a median of 7 days (IQR, 3-11) after PCC administration. Of these, 5 patients died because of fatal bleeding, and the remainder were adjudicated as having died because of other causes. None of the deaths were attributed to fatal thromboembolism.

Discussion

The increasing use of FXaIs means that FXaI-associated major bleeding and the need for urgent surgery in patients treated with anticoagulants are now frequent clinical problems, yet readily available, affordable, and effective reversal strategies remain limited. Our findings indicate that PCC increases thrombin generation without changing FXaI levels or correcting FXaI-induced delays in thrombin generation of LT or TTP. These results build on existing in vitro research by characterizing PCC’s biological effects in vivo in acutely ill FXaI-treated patients. Collectively, findings from the GAUGE study support the hypothesis that the prohemostatic effect of PCC in FXaI-associated bleeding is mediated by increasing thrombin generation, particularly the ETP, rather than by directly reversing FXaI activity or normalizing the thrombin generation curve.

In keeping with this mechanism, PCC did not decrease FXaI levels. This aligns with pharmacokinetic data from healthy volunteers, where PCC corrected the ETP and PT but did not affect apixaban concentrations or anti-FXa activity.²⁶ In vitro work similarly shows that PCC incompletely restores thrombin generation at therapeutic FXaI concentrations and only normalizes ETP and peak at low (<75 ng/mL) FXaI levels.^27,28 The GAUGE study extends these data by showing that baseline FXaI levels strongly modulate the magnitude of the PCC-induced increase in ETP. Patients with lower FXaI levels exhibited greater increases in ETP and higher post-PCC ETP than patients with higher FXaI levels. Together, these results support a model in which PCC does not alter FXaI pharmacokinetics or drug levels but instead provides an excess of coagulation factors that partially restores thrombin generation.

The GAUGE study was designed as a translational study powered to detect changes in thrombin generation kinetics, particularly LT. FXaIs characteristically prolong LT and TTP and suppress peak and mVRI while often leaving ETP relatively preserved, resulting in a flattened, protracted, or what has been described as a camel-back thrombin generation curve.^17,29,30 In the GAUGE study, PCC increased ETP, peak, and mVRI but did not shorten LT or TTP, and thrombin generation remained markedly abnormal in most patients compared with healthy volunteers. The clinical relevance of normalizing ETP while LT and TTP remain prolonged is uncertain. Still, emerging data suggest that both ETP and LT carry prognostic information regarding bleeding risk in anticoagulated patients.^17,31, 32, 33, 34, 35 In particular, work with TIX-5 (a specific inhibitor of FXa-mediated FV activation) has highlighted the importance of LT in hemostasis. Schuijt et al used TIX-5 to demonstrate that FXa-dependent conversion of FV to FVa is crucial for coagulation initiation.³⁶ Building on this mechanistic foundation, Maag et al³⁷ and Sol-Maag et al^37,38, using data from the BLEEDS cohort (>16 000 patients treated with vitamin K antagonists), demonstrated that prolonged thrombin generation LT in the presence of TIX-5 was independently associated with major bleeding. The persistence of a markedly prolonged LT after PCC administration in the GAUGE study may therefore reflect ongoing FXa inhibition at the level of FXa-mediated FV activation. However, the clinical significance of persistent LT/TTP prolongation despite improvement in quantitative thrombin generation parameters (including ETP) remains uncertain and warrants further study.

Clinically, hemostatic efficacy according to the ISTH definition was achieved in approximately half of assessable major bleeding episodes, and perioperative hemostasis was normal in three-quarters of urgent procedures. These proportions are lower than those reported in previous PCC cohorts.12, 13, 14, 15 Comparisons across studies are challenging and should be made with caution because of the variable implementation of hemostasis definitions, small sample sizes, and heterogeneous patient populations. In the GAUGE study, hemostatic outcomes were independently adjudicated in duplicate by experts, using prespecified criteria that allowed hemostasis to be rated as ineffective when there was clear clinical deterioration, even when follow-up imaging was unavailable. Nonetheless, the observed in vivo increases in peak, ETP, and mVRI after PCC administration provide biological evidence of a prohemostatic effect in the continued presence of FXa inhibition. The variability in clinical hemostatic outcomes highlights the need for further studies integrating standardized clinical end points with mechanistic biomarkers to clarify which thrombin generation parameters (kinetic vs quantitative) best reflect net hemostatic benefit in this setting.

Thromboembolic events occurred in 7 of the 101 episodes (6.9%) in the GAUGE study, with a median time to event of 17 days. All events were adjudicated as unlikely to be directly attributable to PCC. This 30-day risk is comparable to that observed in previous PCC cohorts of FXaI-associated bleeding and appears lower than thrombotic risk estimates reported with andexanet alfa in ANNEXA-4 and ANNEXA-I.^4,5,39 In ANNEXA I, adjudicated thrombotic events within 30 days occurred in 10.3% of the patients randomized to andexanet and 5.6% of those assigned to usual care, although the timing of these events was not reported.⁵ Food and Drug Administration briefing materials for ANNEXA-I indicate that the 30-day risk of thrombosis was reestimated at 14.6% with andexanet vs 6.9% with usual care, and thrombotic events in the andexanet arm were front-loaded, with a median time to first event of 3.5 days vs 16 days in the usual-care arm.⁴⁰ Thrombotic events in the GAUGE study were temporally dispersed, and none occurred in patients who had been restarted on therapeutic-dose anticoagulation at the time of the event. Although our observational design precludes causal inference, these findings suggest that standard-dose PCC for FXaI-associated bleeding carries a thrombotic risk broadly consistent with the high-risk baseline profile of this population.

This study has several limitations. First, the observational nature of the study meant that requests for laboratory testing and imaging were not fully standardized, leading to missing FXaI levels and thrombin generation data in some patients and the potential for confounding in between-patient comparisons. Primary analyses relied on within-patient comparisons from before PCC to after PCC. Although paired TGA measurements were not available for all participants, thereby reducing precision, comparison of episodes with and without paired TGA testing (supplemental Table 1) showed only modest differences for some variables, which in clinical terms were unlikely to have substantially affected the conclusions of the thrombin generation analyses. Second, the modest sample sizes of the major bleeding and urgent surgery subgroups limit the precision of clinical estimates. Third, thrombin generation was measured using residual clinical citrated plasma rather than dedicated research specimens. Although technologists handled samples using a standardized protocol, reliance on leftover samples may have introduced unmeasured pre-analytical variability. Finally, the logistical challenges of plasma collection in this critically ill, multimorbid population inevitably constrained the breadth of biological data that could be obtained.

PCC increases thrombin generation but does not reduce FXaI levels or correct FXaI-induced prolongation of TGA LT or TTP. These findings suggest that PCC exerts a measurable prohemostatic effect on thrombin generation in the continued presence of FXa inhibition, although the clinical significance of improving quantitative parameters without shortening LT or TTP remains uncertain, and further studies are needed to determine which thrombin generation measures best reflect net hemostatic benefit. Baseline FXaI level modulated the response to PCC, where higher pre-PCC FXaI levels were associated with an attenuated effect of PCC on thrombin generation. Future studies should integrate clinical hemostatic outcomes with selected mechanistic biomarkers such as thrombin generation to complement clinical end points and guide more rational, individualized approaches to FXaI reversal.

Conflict-of-interest disclosure: J.R.S. reports in-kind laboratory support from Diagnostica Stago; and honoraria (paid to his research institution) from Octapharma. G.L.G. reports honoraria for lectures and advisory boards from Emcure, Pfizer, Sanofi, and Takeda Pharmaceuticals. D.M.S. reports honoraria from AstraZeneca and Servier; and consulting fees from VarmX and Perosphere, paid indirectly to her research institute. Research institute of L.A.C. received honoraria from Bayer, Bristol Myers Squibb–Pfizer Alliance, The Academy for Continued Advancement in Healthcare Education, Amag Pharmaceutical, LEO Pharma, Sanofi, Valeo Pharma, and Servier. The remaining authors declare no competing financial interests.