Association of pre-endoscopic fresh frozen plasma transfusion with clinical outcomes in patients with acute upper gastrointestinal bleeding and mild coagulopathy: a two-center retrospective cohort study

Abstract

The clinical benefit of fresh frozen plasma (FFP) transfusion prior to endoscopy in patients with acute upper gastrointestinal bleeding (AUGIB) and mild coagulopathy remains uncertain. We evaluated the association between pre-endoscopic FFP transfusion and clinical outcomes in patients with AUGIB and an international normalized ratio (INR) of 1.5–2.5. We conducted a retrospective two-center cohort study including adult patients admitted with AUGIB and INR 1.5–2.5 at two tertiary referral hospitals in Thailand between 2016 and 2020. Patients were categorized according to receipt of pre-endoscopic FFP transfusion. Multivariable logistic regression analyses were performed using baseline covariates and bleeding severity scores. An exploratory composite in-hospital major adverse event endpoint was evaluated to improve model stability. Among 244 patients (158 received FFP; 86 did not), those receiving pre-endoscopic FFP had higher crude rates of 30-day all-cause mortality (23.4% vs. 11.6%), in-hospital mortality (24.1% vs. 7.0%), pulmonary edema (23.4% vs. 4.7%), and the composite in-hospital major adverse event endpoint (40.5% vs. 10.5%) (all p < 0.01). After multivariable adjustment, pre-endoscopic FFP transfusion remained associated with the composite endpoint (adjusted odds ratio [aOR] 5.28; 95% confidence interval [CI], 2.17–12.82), in-hospital mortality (aOR 5.36; 95% CI, 1.87–15.37), pulmonary edema (aOR 3.85; 95% CI, 1.21–12.26), and 30-day mortality (aOR 2.69; 95% CI, 1.09–6.66). In subgroup analyses, these associations were more consistent among patients with variceal bleeding. In patients with AUGIB and mildly elevated INR, pre-endoscopic FFP transfusion was associated with higher mortality and pulmonary complications, particularly in those with variceal bleeding. Given the retrospective design and potential for residual confounding, these findings should be interpreted with caution. Nevertheless, these findings support consideration of a more selective, context-based approach to plasma transfusion and highlight the need for prospective studies to inform evidence-based transfusion strategies.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-41863-y.

Introduction

Acute upper gastrointestinal bleeding (AUGIB) is a common and potentially life-threatening medical emergency worldwide, affecting approximately 0.8%–1.5% of the general population¹. Despite advances in endoscopic therapy and supportive care, mortality remains between 2% and 5%, particularly among older adults and patients with significant comorbidities or hemodynamic instability at presentation². Although outcomes have improved over recent decades^3–5, important uncertainty persists regarding optimal pre-endoscopic resuscitation strategies, especially with respect to blood product transfusion.

Initial management of AUGIB focuses on hemodynamic stabilization through intravenous fluid resuscitation and blood product administration as clinically indicated^6–8. Red blood cell transfusion is generally recommended when hemoglobin levels fall below 7–8 g/dL, and platelet transfusion is advised for platelet counts below 50,000/mm³ in the setting of active bleeding. In contrast, the role of fresh frozen plasma (FFP) remains less clearly defined. In clinical practice, FFP is frequently administered to patients with coagulopathy to correct an elevated international normalized ratio (INR) prior to endoscopic intervention, based on the assumption that normalization of coagulation parameters may reduce ongoing or recurrent bleeding. However, this approach is largely empirical, and high-quality evidence demonstrating clinical benefit is limited.

Current guideline recommendations regarding FFP transfusion thresholds are inconsistent. The National Institute for Health and Care Excellence (NICE) recommends FFP administration when the INR exceeds 1.5⁹, whereas several international guidelines advocate more conservative thresholds, typically an INR greater than 2.5^10–12. Notably, the Baveno VII consensus guidelines for variceal bleeding discourage routine FFP use, citing both the absence of proven benefit and the potential for harm, including volume overload and delay in endoscopic intervention¹³.

Recent observational studies have further raised concerns regarding the safety and effectiveness of FFP transfusion in AUGIB. Emerging evidence suggests that FFP use may be associated with adverse outcomes, including pulmonary edema, nosocomial infections, prolonged hospitalization, and increased mortality, without clear evidence of hemostatic benefit^14–18. Despite these concerns, FFP transfusion remains common in real-world practice, reflecting ongoing clinical uncertainty and the absence of definitive evidence to guide decision-making^19–21.

To address this gap, we conducted a two-center retrospective cohort study to evaluate the association between pre-endoscopic FFP transfusion and clinical outcomes in patients with AUGIB and mildly elevated INR values (1.5–2.5). We compared short-term and intermediate-term outcomes, including in-hospital mortality, rebleeding, pulmonary complications, 30-day mortality, and medical resource utilization, between patients who received pre-endoscopic FFP and those who did not. By examining a well-characterized cohort from two tertiary care hospitals in Thailand, this study aims to provide clinically relevant evidence to inform transfusion practices and guideline development in the management of acute upper gastrointestinal bleeding.

Objectives

The primary outcome was to investigate the association between pre-endoscopic FFP transfusion and 30-day all-cause mortality in patients presenting with AUGIB and mild coagulopathy, defined as an INR between 1.5 and 2.5.

Secondary outcomes included the evaluation of other clinical outcomes, such as in-hospital mortality, rebleeding, pulmonary edema, as well as medical resource utilization during hospitalization including length of hospital stay.

Methods

Study design and setting

This two-center retrospective cohort study was conducted at two tertiary referral hospitals in Thailand: Hatyai Hospital in southern Thailand and Ayutthaya Hospital in central Thailand. The study period spanned from January 2016 to December 2020.

Study population

Eligible participants were adults (aged ≥ 18 years) presenting with AUGIB, defined by clinical manifestations such as hematemesis, coffee-ground emesis, or melena, who underwent confirmatory upper gastrointestinal endoscopy during hospitalization. Inclusion required an INR between 1.5 and 2.5 measured at presentation. Exclusion criteria included prior upper endoscopy before admission, a diagnosis of AUGIB within the preceding three months, normal endoscopic findings, or incomplete medical records.

Data collection and variables

Clinical data were retrospectively extracted from electronic medical records using a standardized case report form. Collected variables included demographic characteristics, comorbidities, medication history, initial vital signs, laboratory parameters, and pre-endoscopic management (e.g., intravenous fluids, vasoactive agents, antibiotics, and transfusions). Endoscopic findings, time to endoscopy, and therapeutic interventions were also documented. Time to endoscopy was defined as the interval from hospital presentation to the start of upper gastrointestinal endoscopy.

All covariates were defined a priori as values obtained at initial hospital presentation, prior to endoscopic evaluation. Vital signs were recorded from the initial triage or first physician assessment, and laboratory parameters were obtained from the first blood draw on arrival. Repeat laboratory or vital sign measurements obtained before endoscopy were not used for covariate adjustment, as these values may reflect evolving illness severity and could be influenced by early management decisions.

Patients were categorized into two groups according to receipt of pre-endoscopic FFP transfusion. The volume and number of FFP units administered before endoscopy were recorded as part of exposure characterization. Use of other blood components during hospitalization was recorded for descriptive and outcome-related analyses. Packed red blood cell (PRBC) transfusion was recorded separately as units administered prior to endoscopy and as total units administered during hospitalization. Only pre-endoscopic PRBC transfusions were considered baseline covariates for multivariable adjustment.

During the study period, no standardized institutional protocol mandated FFP transfusion for patients with AUGIB and an INR of 1.5–2.5. Decisions regarding pre-endoscopic FFP transfusion were made at the discretion of the treating physicians, typically with the intent to correct perceived coagulopathy prior to endoscopy.

Protocol and interventions

At both participating centers, patients with AUGIB were managed according to routine guideline-based clinical practice rather than a rigidly standardized institutional protocol. Initial management focused on hemodynamic stabilization, including crystalloid fluid resuscitation, airway protection when indicated, and early initiation of appropriate pharmacologic therapy.

Pre-endoscopic medical management commonly included proton pump inhibitors, vasoactive agents for suspected variceal bleeding, empirical antibiotics, and PRBC transfusion using a restrictive strategy targeting a hemoglobin threshold of approximately 7 g/dL, unless clinically contraindicated. The choice and timing of specific agents were determined by the treating physician based on clinical judgment.

FFP transfusion was not governed by a formal institutional protocol and was administered at the discretion of the treating physician based on an integrated assessment of coagulation parameters, bleeding severity, hemodynamic status, and response to initial resuscitation. Vitamin K was administered on a case-by-case basis, primarily in patients receiving warfarin or those with suspected deficiency.

All patients underwent upper gastrointestinal endoscopy during the index hospitalization, generally within 24 h of presentation. For variceal bleeding, endoscopic variceal ligation was the primary hemostatic technique for esophageal varices, and cyanoacrylate injection was used for gastric varices when indicated. For non-variceal bleeding with high-risk stigmata, standard endoscopic hemostatic modalities—including thermal coagulation, mechanical clipping, and combination therapy with or without injection—were used at the discretion of the endoscopist.

Escalation strategies, including repeat endoscopy and consultation for interventional radiology–guided transcatheter arterial embolization or surgery, were available as part of routine care for patients with persistent or recurrent bleeding. Post-endoscopic management followed standard practice, with close clinical monitoring for signs of rebleeding and repeat endoscopy performed when clinically indicated.

Outcome measures and definitions

The primary outcome was 30-day all-cause mortality, defined as death from any cause occurring within 30 days of presentation and ascertained using records from the Thai civil registration system.

Secondary outcomes included in-hospital rebleeding, in-hospital pulmonary edema, in-hospital mortality, and length of hospital stay. In-hospital rebleeding was defined as recurrence of clinical signs of upper gastrointestinal bleeding (hematemesis, melena, or hematochezia) after initial stabilization, accompanied by a hemoglobin decrease of > 2 g/dL or the need for additional blood transfusion. Repeat endoscopy or imaging was performed at the discretion of the treating physician and was not required for outcome definition.

In-hospital pulmonary edema was defined as new-onset respiratory deterioration requiring medical intervention in conjunction with compatible radiographic findings on chest imaging (e.g., pulmonary congestion or interstitial/alveolar edema). The diagnosis was based on clinical assessment documented by the treating physician. Because of the retrospective design, formal differentiation between transfusion-associated circulatory overload and transfusion-related acute lung injury was not consistently feasible; however, identified cases represented clinically significant pulmonary complications temporally associated with hospitalization and transfusion exposure.

In-hospital mortality was defined as death from any cause during the index hospitalization. Length of hospital stay was calculated as the number of days from admission to discharge or death.

An exploratory composite outcome of in-hospital major adverse events was prespecified for sensitivity analysis, defined as the occurrence of any of the following during the index hospitalization: in-hospital mortality, pulmonary edema, or in-hospital rebleeding.

Statistical analysis

Descriptive statistics were used to summarize baseline characteristics. Continuous variables were expressed as mean ± standard deviation or median with interquartile range, as appropriate. Categorical variables were presented as counts and percentages. Between-group comparisons were performed using the Student’s t test or Wilcoxon rank-sum test for continuous variables and the chi-square test or Fisher’s exact test for categorical variables.

Univariate and multivariable logistic regression analyses were performed to evaluate associations between pre-endoscopic FFP transfusion and clinical outcomes. In the overall cohort, covariates included age, sex, chronic kidney disease, Glasgow–Blatchford score, AIMS65 score, presence of shock at presentation, fresh blood via nasogastric tube, pre-endoscopic packed red blood cell transfusion units, bleeding etiology (variceal vs. non-variceal), hospital center, and time to endoscopy. Subgroup models included the same covariates, except that bleeding etiology was omitted in the non-variceal analysis, and Model for End-Stage Liver Disease–Sodium (MELD-Na) score was additionally included in the variceal analysis.

Time to endoscopy was included in the primary multivariable models as a marker of early care processes and potential system-level variation between centers. Because pre-endoscopic FFP transfusion could theoretically influence procedural timing, we performed sensitivity analyses excluding time to endoscopy to evaluate whether it functioned as a potential mediator. The direction and magnitude of associations were not materially changed.

To minimize multicollinearity, composite severity scores (Glasgow–Blatchford score and AIMS65) were used as summary indicators of bleeding severity, and their individual component variables were not simultaneously included as separate covariates.

Multicollinearity was assessed using variance inflation factors (VIFs). A VIF value greater than 5 was considered indicative of significant multicollinearity.

Propensity score methods were considered but not performed because of the modest sample size and risk of overfitting, particularly in subgroup analyses. Results were reported as adjusted odds ratios or regression coefficients with 95% confidence intervals. Because pre-endoscopic FFP transfusion could occur after hospital arrival, baseline (presentation-time) covariates were used for adjustment to reduce bias related to time-updated illness severity. To avoid adjustment for potential mediators, post-endoscopy PRBC transfusions were not included in multivariable models.

Missing data were assessed for all baseline variables. The proportion of missing data was low, with all variables having less than 5% missing values (Supplementary Table S1). Multivariable analyses were therefore conducted using a complete-case approach, assuming data were missing at random. Given the minimal extent of missing data, multiple imputation was not performed.

All statistical analyses were performed using STATA version 15.1 (StataCorp LLC, College Station, TX, USA). A two-sided p value < 0.05 was considered statistically significant.

Sample size justification

Given the retrospective design, a total population sampling approach was employed, including all eligible patients during the study period. Although no formal a priori sample size calculation was performed, the final cohort of 244 patients provided sufficient event rates to allow multivariable adjustment for key outcomes.

Ethical considerations

This study was conducted in accordance with the Declaration of Helsinki. The protocol was reviewed and approved by the Institutional Review Boards of both participating hospitals (Hatyai Hospital: HYH EC 092-66−01; Ayutthaya Hospital: COA014/2568). The requirement for informed consent was waived due to the retrospective design. All data were anonymized prior to analysis and stored in secure, password-protected databases accessible only to authorized investigators. The study adhered to the STROBE guidelines for observational studies.

Patient and public involvement

Patients and the public were not involved in the design, conduct, reporting, or dissemination of this research.

Results

Baseline characteristics

A total of 244 patients with acute upper gastrointestinal bleeding and an INR between 1.5 and 2.5 were included in the analysis (Table 1). Of these, 158 patients (64.7%) received FFP transfusion prior to endoscopic evaluation, while 86 patients (35.2%) did not.

Table 1.

Baseline characteristics.

Baseline characteristics n (%)	Total Enroll (n = 244)
	No FFP Group n = 86 (35.2%)	FFP Group n = 158 (64.7%)	p–value
Demographic Data
Age (years)*	56.0 ± 13.4	53.3 ± 10.0	0.098
Male sex	79 (91.9%)	138 (87.3%)	0.393
Body mass index (kg/m2)*	22.8 ± 3.7	22.7 ± 4.3	0.804
Underlying diseases
Hypertension	17 (19.8)	18 (11.4)	0.087
Diabetes mellitus	12 (14.0)	17 (10.8)	0.535
Cirrhosis	51 (59.3)	97 (61.4)	0.785
-MELD score	19.2 ± 5.3	20.0 ± 4.9	0.358
Chronic kidney disease	5 (5.8)	7 (4.7)	0.758
Congestive heart failure	1 (1.2)	2 (2.3)	1.000
Dyslipidemia	6 (7.0)	5 (3.2)	0.202
Cerebrovascular disease	3 (3.5)	4 (2.5)	0.700
Chronic obstructive pulmonary disease	2 (2.3)	2 (1.3)	0.615
History of syncope	62 (37.2)	83 (52.5)	0.023
Presence of shock at initial assessment	14 (16.5)	29 (18.4)	0.860
Mean arterial pressure at presentation (mmHg)*	85.1 ± 16.5	80.7 ± 17.2	0.055
Medications
Antiplatelet	5 (5.9)	8 (5.1)	0.547
Warfarin	4 (4.7)	3 (1.9)	0.246
Nonsteroidal anti–inflammatory drugs	6 (7.0)	10 (6.3)	1.000
Steroid	0 (0.0)	3 (1.9)	0.554
Proton pump inhibitor	11 (12.8)	12 (7.6)	0.251
Initial Laboratory Results*
Hemoglobin (g/dL)	8.1 ± 2.6	7.3 ± 2.5	0.016
Platelet count (⋅103/µL) **	99,000 (60,000–161,000)	91,500 (64,000–152,000)	0.768
Creatinine (mg/dL) **	0.9 (0.7–1.5)	0.9 (0.7–1.4)	0.855
Sodium (mmol/L)	138.1 ± 5.7	137.8 ± 12.0	0.838
Albumin (mg/dL)	2.6 ± 0.6	2.5 ± 0.6	0.096
International normalized ratio	1.7 ± 0.2	1.8 ± 0.2	0.013
Scoring Systems*
Glasgow–Blatchford score	10.8 ± 4	12.5 ± 3.6	0.040
AIMS65 score	2.2 ± 1.0	2.4 ± 1.9	0.244
Rockall score	3.7 ± 2.3	4.5 ± 2.4	0.023
Non–Endoscopic Managements
Fresh blood on NG	31 (36.0)	73 (46.2)	0.138
Pre endoscopic proton pump inhibitor	86 (100)	153 (96.8)	0.165
Pre endoscopic prokinetic	0 (0)	2 (1.3)	0.542
Pre endoscopic antibiotic	37 (84.9)	141 (89.2)	0.581
Vasopressor	67 (77.9)	132 (83.5)	0.302
Pre Endoscopic Blood Product Used
Need for Packed red blood cell transfusion	46 (53.5%)	143 (90.5)	< 0.001
Packed red blood cell transfusion (unit) *	1.1 ± 1.0	3.1 ± 2.0	< 0.001
Platelet transfusion (unit) **	6 (1–10)	6 (6–10)	0.258
Fresh frozen plasma transfusion (mL) *	-	1000 ± 630	N/A
Endoscopic Managements
Pre–endoscopic hemoglobin (g/dL)*	8.5 ± 2.2	7.8 ± 2.0	0.013
Time to endoscopy (hours)**	19.5 (12.0–34.0)	18.0 (10.0–36.0)	0.762
Endoscopy within 24 h	56 (65.1)	107 (67.7)	0.770
Endoscopic Findings
Variceal bleeding	55 (64.0)	112 (70.9)	0.313
Peptic related ulcer	21 (24.4)	38 (24.1)	0.898
Gastritis/Duodenitis/esophagitis	10 (11.6)	7 (4.4)	0.186
Tumor	0 (0)	1 (0.6)	1.000
Helicobacter pylori infection	10 (11.6)	9 (5.4)	0.132
Total Blood Product Used
Packed red blood cell transfusion received	59 (68.6)	144 (91.1)	< 0.001
Packed red blood cell transfusion, (unit) *	2 ± 1.5	4.5 ± 3.0	< 0.001
Platelet transfusion, (unit) **	0 (0–0)	0 (0–6)	< 0.001
Fresh frozen plasma transfusion, (mL) **	180 ± 490	1300 ± 1000	< 0.001

Data were expressed as number (%) unless specific, *Mean ± SD, **Median (IQR; Interquartile range), FFP; Fresh frozen plasma, MELD; Model for End-Stage Liver Disease.

Patients who received pre-endoscopic FFP were clinically more severe at presentation, with higher bleeding severity and greater resuscitative requirements. Baseline demographic characteristics—including age, sex, body mass index, and the prevalence of major comorbidities such as cirrhosis, diabetes mellitus, chronic kidney disease, and cardiovascular disease—were generally comparable between groups. In contrast, marked differences were observed in indicators of bleeding severity and early resuscitation.

The FFP group presented with significantly lower hemoglobin levels (7.3 ± 2.5 vs. 8.1 ± 2.6 g/dL; p = 0.016) and higher INR values (1.8 ± 0.2 vs. 1.7 ± 0.2; p = 0.013). Bleeding severity scores were also higher among patients receiving FFP, as reflected by increased Glasgow–Blatchford scores (12.5 ± 3.6 vs. 10.8 ± 4.0; p = 0.040) and Rockall scores (4.5 ± 2.4 vs. 3.7 ± 2.3; p = 0.023).

Consistent with greater clinical severity, patients in the FFP group required more intensive resuscitation prior to endoscopy. Pre-endoscopic packed red blood cell transfusion was more frequently administered in the FFP group (90.5% vs. 53.5%; p < 0.001), and the mean number of packed red blood cell units transfused before endoscopy was substantially higher (3.1 ± 2.0 vs. 1.1 ± 1.0 units; p < 0.001). Total blood component utilization during hospitalization—including packed red blood cells, platelets, and FFP—was also greater among patients who received pre-endoscopic FFP.

Other laboratory parameters, including platelet count, serum creatinine, and serum albumin, did not differ significantly between groups. Among patients with cirrhosis, baseline liver disease severity was similar, as reflected by comparable Model for End-Stage Liver Disease scores.

Clinical outcomes

All patients underwent diagnostic upper gastrointestinal endoscopy. Endoscopic hemostatic interventions were performed more frequently in the FFP group, consistent with greater baseline bleeding severity (Supplementary Table S2). Rescue interventions, including transcatheter arterial embolization or surgery, were rare in both groups.

Clinical outcomes are summarized in Table 2. Patients who received pre-endoscopic FFP transfusion experienced a significantly higher incidence of the composite in-hospital major adverse event endpoint compared with those who did not receive FFP (40.5% vs. 10.5%; p < 0.001). Individual adverse events were also more frequent in the FFP group, including in-hospital pulmonary edema (23.4% vs. 4.7%; p < 0.001) and in-hospital mortality (24.1% vs. 7.0%; p = 0.001). Although in-hospital rebleeding occurred more often among patients receiving FFP (7.0% vs. 1.2%), this difference did not reach statistical significance (p = 0.061). Patients in the FFP group also had a longer length of hospital stay (8.1 ± 6.8 vs. 5.8 ± 3.2 days; p < 0.001). Thirty-day all-cause mortality was higher in the FFP group (23.4% vs. 11.6%; p = 0.003).

Table 2.

Clinical outcomes and adjusted outcome of patients who received FFP prior to endoscopy and those who did not.

Outcomes	Total enroll (n = 244)			Adjusted Odds Ratio (95% CI) †α	p–value
	No FFP Group n = 86	FFP Group n = 158	p–value
Composite in-hospital major adverse event***	9 (10.5)	64 (40.5)	< 0.001	5.277 (2.171–12.823)	< 0.001
In-hospital rebleeding	1 (1.2)	11 (7)	0.061	4.926 (0.503–48.248)	0.171
In-hospital pulmonary edema	4 (4.7)	37 (23.4)	< 0.001	3.851 (1.209–12.261)	0.023
In-hospital mortality	6 (7.0)	38 (24.1)	0.001	5.359 (1.868–15.373)	0.002
Length of hospital stay, days*	5.8 ± 3.2	8.12 ± 6.8	< 0.001	0.145 (−1.508–1.798)	0.863
30-day all-cause mortality	10 (11.6)	37 (23.4)	0.003	2.691 (1.088–6.655)	0.032

Data are presented as n (%) unless otherwise specified.

*Mean ± SD.

***Composite endpoint defined as in-hospital mortality, pulmonary edema, or in-hospital rebleeding.

^†Adjusted odds ratios or regression coefficients, as appropriate.

^αAdjusted for age, sex, chronic kidney disease, baseline severity scores (Glasgow–Blatchford score and AIMS65), time to endoscopy, presence of fresh blood via nasogastric tube, presence of shock at presentation, number of packed red blood cell transfusions prior to endoscopy, bleeding etiology (variceal vs. non-variceal), and hospital center.

Abbreviations: FFP, fresh frozen plasma; CI, confidence interval.

Complete-case multivariable analyses for the primary and secondary outcomes included 242 patients. After adjustment for baseline covariates and bleeding severity scores, pre-endoscopic FFP transfusion remained independently associated with a higher risk of the composite in-hospital major adverse event endpoint (adjusted odds ratio [aOR], 5.277; 95% confidence interval [CI], 2.171–12.823; p < 0.001). Significant associations were also observed for in-hospital pulmonary edema (aOR, 3.851; 95% CI, 1.209–12.261; p = 0.023) and in-hospital mortality (aOR, 5.359; 95% CI, 1.868–15.373; p = 0.002). Pre-endoscopic FFP transfusion was additionally associated with increased 30-day all-cause mortality after adjustment (aOR, 2.691; 95% CI, 1.088–6.655; p = 0.032). In contrast, adjusted associations with in-hospital rebleeding and length of hospital stay were not statistically significant.

Variance inflation factor analysis demonstrated low multicollinearity across all models (maximum VIF 1.65 in the overall cohort; Supplementary Table S3).

Subgroup analyses

Non-variceal bleeding

Among patients with non-variceal upper gastrointestinal bleeding (n = 77), 46 received pre-endoscopic FFP and 31 did not (Table 3). The crude incidence of the composite in-hospital major adverse event endpoint was higher in the FFP group (32.6% vs. 9.7%; p = 0.020). However, after multivariable adjustment, this association was attenuated and did not reach statistical significance (aOR, 4.297; 95% CI, 0.730–25.293; p = 0.107).

Table 3.

Subgroup analysis of non–variceal group.

Outcomes n (%)	Non–variceal group (n = 77)
	No FFP Group n = 31 (40.2%)	FFP Group n = 46 (59.7%)	p–value	Adjusted Odds Ratio (95% CI) †α	p–value
Composite in-hospital major adverse event*	3 (9.7)	15 (32.6)	0.020	4.297 (0.730–25.293)	0.107
In-hospital rebleeding	0 (0)	6 (13.0)	0.076	-	-
In-hospital pulmonary edema	1 (3.2)	7 (15.2)	0.134	0.562 (0.026–12.184)	0.714
In-hospital mortality	2 (6.5)	9 (19.6)	0.183	39.327 (1.628–949.982)	0.024
Length of hospital stay, days*	5.7 ± 3.0	9.5 ± 8.9	0.009	0.466 (–3.386–4.319)	0.810
30-day all-cause mortality	4 (12.9)	9 (19.6)	0.544	11.012 (0.810–149.694)	0.072

Data are presented as n (%) unless otherwise specified.

*Mean ± SD.

***Composite endpoint defined as in-hospital mortality, pulmonary edema, or in-hospital rebleeding.

^†Adjusted odds ratios or regression coefficients, as appropriate.

^αAdjusted for age, sex, chronic kidney disease, baseline severity scores (Glasgow–Blatchford score and AIMS65), time to endoscopy, presence of fresh blood via nasogastric tube, presence of shock at presentation, number of packed red blood cell transfusions prior to endoscopy, and hospital center.

Abbreviations: FFP, fresh frozen plasma; CI, confidence interval.

No statistically significant adjusted associations were observed for pulmonary edema, rebleeding, or 30-day mortality in this subgroup. Although in-hospital mortality demonstrated a statistically significant adjusted association (aOR, 39.327; 95% CI, 1.628–949.982; p = 0.024), the estimate was imprecise with a wide confidence interval, reflecting the small number of events. Length of hospital stay was longer in the FFP group in unadjusted analysis but was not significant after adjustment.

Variceal bleeding

In contrast, among patients with variceal bleeding (n = 167), adverse associations with pre-endoscopic FFP transfusion were more consistent and pronounced (Table 4). Patients receiving FFP had higher crude rates of the composite in-hospital major adverse event endpoint (43.8% vs. 10.9%; p < 0.001), pulmonary edema (26.8% vs. 5.5%; p = 0.001), in-hospital mortality (25.9% vs. 7.3%; p = 0.004), and 30-day mortality (25.0% vs. 10.9%; p = 0.041).

Table 4.

Subgroup analysis of variceal group.

Outcomes n (%)	Variceal group (n = 167)
	No FFP Group n = 55 (33.0%)	FFP Group n = 112 (67.0%)	p–value	Adjusted Odds Ratio (95% CI) †α	p–value
Composite in-hospital major adverse event*	6 (10.9)	49 (43.8)	< 0.001	5.193 (1.998–17.502)	0.001
In-hospital rebleeding	1 (1.8)	5 (4.5)	0.665	3.765 (0.340–41.686)	0.280
In-hospital pulmonary edema	3 (5.5)	30 (26.8)	0.001	4.379 (1.142–16.792)	0.031
In-hospital mortality	4 (7.3)	29 (25.9)	0.004	5.625 (1.575–20.095)	0.008
Length of hospital stay, days*	3.8 ± 3.3	7.4 ± 5.6	0.018	0.003 (−1.725–1.732)	0.997
30-day all-cause mortality	6 (10.9)	28 (25.0)	0.041	3.132 (1.042–9.418)	0.042

Data are presented as n (%) unless otherwise specified.

*Mean ± SD.

***Composite endpoint defined as in-hospital mortality, pulmonary edema, or in-hospital rebleeding.

^†Adjusted odds ratios or regression coefficients, as appropriate.

^αAdjusted for age, sex, chronic kidney disease, baseline severity scores (Glasgow–Blatchford score and AIMS65), time to endoscopy, presence of fresh blood via nasogastric tube, presence of shock at presentation, Model for End-Stage Liver Disease–Sodium Score, number of packed red blood cell transfusions prior to endoscopy, and hospital center.

Abbreviations: FFP, fresh frozen plasma; CI, confidence interval.

After multivariable adjustment, pre-endoscopic FFP transfusion remained independently associated with the composite in-hospital major adverse event endpoint (aOR, 5.193; 95% CI, 1.998–17.502; p = 0.001), in-hospital pulmonary edema (aOR, 4.379; 95% CI, 1.142–16.792; p = 0.031), in-hospital mortality (aOR, 5.625; 95% CI, 1.575–20.095; p = 0.008), and 30-day all-cause mortality (aOR, 3.132; 95% CI, 1.042–9.418; p = 0.042). No significant adjusted association was observed for in-hospital rebleeding or length of hospital stay.

Sensitivity analysis

Sensitivity analyses excluding time to endoscopy from the multivariable models yielded similar effect estimates in the overall cohort as well as in both non-variceal and variceal subgroups. The direction and magnitude of the associations between pre-endoscopic FFP transfusion and clinical outcomes were not materially changed (Supplementary Tables S4–S6).

Discussion

This two-center retrospective cohort study provides insight into the safety and clinical implications of fresh frozen plasma (FFP) transfusion in patients with acute upper gastrointestinal bleeding (AUGIB) and mild coagulopathy, defined by an international normalized ratio (INR) of 1.5–2.5. Our findings indicate that pre-endoscopic FFP transfusion was associated with higher mortality and pulmonary complications. To address concerns regarding potential overfitting in models with limited event counts, we additionally evaluated a composite in-hospital major adverse event endpoint, which increased event numbers and yielded more stable estimates; the direction of associations remained consistent in this analysis.

These findings may reflect the complex changes in blood clotting and circulation seen in patients with cirrhosis, rather than a direct harmful effect of FFP itself. In patients with cirrhosis, INR does not reliably predict bleeding risk because both pro-clotting and anti-clotting factors are reduced, creating a fragile but rebalanced system. Giving plasma may increase blood volume, which could raise portal pressure or place additional stress on the heart and lungs in vulnerable patients. However, these possible mechanisms cannot be confirmed from this retrospective study. It is also possible that the observed associations are due to residual confounding by indication, meaning that patients who appeared more severely ill or at higher bleeding risk were more likely to receive FFP and were also more likely to experience poor outcomes despite statistical adjustment.

Our results are consistent with and extend prior observational studies. Subramaniam et al. reported a dose-dependent association between FFP transfusion and mortality in non-variceal bleeding, while Mohanty et al. demonstrated increased mortality in variceal bleeding, particularly among patients with higher MELD scores^22,23. Liu et al. similarly observed adverse associations between FFP use and outcomes in patients with INR values below 1.5²⁴. Collectively, these studies, together with our findings, raise concern regarding liberal FFP use in the setting of gastrointestinal bleeding and liver disease.

Growing evidence suggests that INR is an unreliable marker of bleeding risk in cirrhosis. INR primarily reflects reductions in procoagulant factors but fails to capture the concomitant decline in endogenous anticoagulants that characterize the so-called rebalanced hemostatic state in chronic liver disease^10,13. Our findings further support a move away from INR-based transfusion thresholds toward clinical context–driven decision-making.

Importantly, coagulopathy in cirrhosis is pathophysiologically distinct from anticoagulation due to pharmacologic agents. In cirrhosis, decreases in procoagulant factors occur in parallel with reductions in anticoagulant proteins such as protein C, protein S, and antithrombin, resulting in a tenuous equilibrium rather than a true hypocoagulable state. Seminal work by Tripodi and colleagues demonstrated that thrombin generation in cirrhotic patients may be preserved or even increased despite prolonged INR values, underscoring the limited utility of INR as a predictor of bleeding risk²⁵. More recent studies have further emphasized that cirrhosis may be associated with a prothrombotic tendency, including increased risk of portal vein and microvascular thrombosis, particularly in advanced disease or during acute illness²⁶. In this context, routine “correction” of INR with FFP may be unnecessary in many cirrhotic patients, and its clinical benefit remains unproven. This paradigm contrasts with drug-induced anticoagulation, in which coagulation abnormalities directly reflect pharmacologic inhibition and targeted reversal strategies are often appropriate.

Current guideline recommendations reflect these pathophysiologic distinctions. The NICE guideline suggests FFP transfusion for INR values greater than 1.5⁹, whereas international consensus guidelines recommend more conservative thresholds, typically INR > 2.5²⁰. The Baveno VII consensus specifically advises against routine FFP use prior to endoscopy in variceal bleeding because of potential harm¹³. Our findings, particularly among cirrhotic patients, support the latter recommendations.

The increased transfusion burden observed in the FFP group likely reflects clinical judgment in response to more severe bleeding presentations. Patients who received FFP had lower hemoglobin levels, higher INR values, and higher bleeding severity scores at presentation. Pulmonary edema in this cohort may therefore reflect the cumulative effects of aggressive resuscitation—including crystalloid infusion and red blood cell transfusion—in patients with more severe bleeding rather than a direct adverse effect of FFP alone. Although our multivariable models adjusted for baseline severity and pre-endoscopic red blood cell transfusion, residual confounding remains an important limitation. Confounding by indication may have influenced both the decision to administer FFP and the likelihood of adverse outcomes, and unmeasured factors such as fluid balance, cardiac function, or provider preference may have further contributed to bias. In addition, outcomes were not stratified by FFP dose or timing; thus, potential dose–response relationships or time-dependent effects could not be assessed. Future studies incorporating dose-response analyses and time-resolved modeling are warranted, as are investigations into alternative strategies such as prothrombin complex concentrates or viscoelastic test–guided transfusion.

In subgroup analyses, no statistically significant adjusted associations between FFP and adverse outcomes were observed among patients with non-variceal bleeding, although estimates were imprecise because of small sample size and limited event counts. In contrast, among patients with variceal bleeding, pre-endoscopic FFP transfusion remained associated with increased mortality and pulmonary complications after adjustment. Although baseline MELD scores were similar between groups, residual confounding related to bleeding severity, resuscitation intensity, and cardiopulmonary reserve cannot be excluded. Nonetheless, these associations are biologically plausible given the susceptibility of cirrhotic patients to volume overload and portal pressure augmentation.

Several limitations merit consideration. The retrospective observational design precludes causal inference. Given the retrospective design, residual confounding—including confounding by indication, time-dependent severity, and differences in resuscitation strategies—cannot be fully excluded. Therefore, the observed associations should not be interpreted as evidence that FFP directly caused worse outcomes. Data on fluid balance, diuretic use, and detailed cardiopulmonary status were unavailable, and alternative hemostatic interventions were not systematically evaluated. Rebleeding was defined using clinical criteria and hemoglobin decline, and repeat endoscopy was not uniformly performed, introducing potential misclassification. Most patients in this cohort had cirrhosis-related coagulopathy, while only a small minority were receiving vitamin K antagonists and none were on direct oral anticoagulants; therefore, these findings should not be extrapolated to anticoagulated populations without caution. Additionally, the absence of detailed time-stamp data limited assessment of time-dependent confounding.

In summary, among patients with AUGIB and mild coagulopathy, pre-endoscopic FFP transfusion was associated with higher mortality and pulmonary complications, particularly among those with variceal bleeding. These findings should be interpreted cautiously, as they may reflect residual confounding related to baseline severity and resuscitation intensity rather than a direct adverse effect of plasma transfusion. Nevertheless, the results support a more selective and context-based approach to plasma use and highlight the need for prospective studies to better define optimal transfusion strategies.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

KB designed and conceptualized the study, acquired, interpreted the data and drafted the manuscript. AC designed and conceptualized the study, acquired, formally analyzed, interpreted the data, and drafted the manuscript. TN, NS, NR and KC acquired, formally analyzed, interpreted the data. MR, SC, and VP formally analyzed and critically revised the manuscript for important intellectual content. All authors have reviewed and approved the final version of the manuscript. AC is the guarantor and accepts full responsibility for the work and the conduct of the study, had access to the data, and controlled the decision to publish.

Data availability

Data are available upon reasonable request. Deidentified participant data are available from Arunchai Chang (busmdcu58@gmail.com) on reasonable request, subject to approval by the participating institutions and data privacy regulations.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

The study protocol was reviewed and approved by the Institutional Review Boards of both participating hospitals (Hatyai Hospital: HYH EC 092-66-01; Ayutthaya Hospital: COA014/2568). This study was conducted in accordance with the principles of the Declaration of Helsinki. The requirement for informed consent was waived by both ethics committees because of the retrospective nature of the study.