Predictive biomarkers of early allograft dysfunction after liver transplantation: A prospective pilot study

Abstract

BACKGROUND

Early allograft dysfunction (EAD) contributes to significant morbidity and mortality. Although EAD can occur after any preservation method, the risk is higher with static cold storage (SCS) - still the predominant modality worldwide - given its longer ischemia time. We hypothesize specific peri-implantation alterations in serum cytokine and transcription factor levels are associated with EAD in recipients of allografts preserved using SCS.

AIM

To identify peri-implantation predictive biomarkers of EAD in SCS recipients.

METHODS

We conducted a prospective single-center pilot study of adult deceased donor SCS-preserved liver transplant recipients from August 2023 to July 2024. EAD was defined by the Olthoff criteria. Arterial serum was obtained pre-implantation (timepoint, T1) and 2 (T2) and 48 hours (T3) post-implantation to measure biomarker levels by multiplex immunoassay. Biomarker levels between non-EAD and EAD groups, and their associations with clinical outcomes, were compared using Mann-Whitney U, χ² or Fisher’s exact, Friedman, Wilcoxon Z, and Spearman correlation tests.

RESULTS

EAD occurred in 8 of 24 SCS recipients (33.3%). Compared with non-EAD recipients, those with EAD had lower interleukin-6 (IL-6) levels at T1 [6.3 pg/mL interquartile ranges (IQR): 3.7, 12.9 vs 15.4 pg/mL IQR: 8.7, 24.9, P = 0.0433], greater peri-implantation increases in induced protein 10 (IP-10) (T2-T1) [386.7 pg/mL (IQR: -26.7, 1280.7) vs -181.5 pg/mL (IQR: -452.4, 145.9), P = 0.02], and higher 48-hour post-implantation hypoxia inducible factor-1 alpha (HIF-1α) levels at T3 [611.2 pg/mL (IQR: 384.7, 869.9) vs 157.9 pg/mL (IQR: 110, 273), P = 0.0095]. Discriminatory performance was moderate-to-strong for pre-implantation IL-6 [area under the curve (AUC) = 0.7578], peri-implantation IP-10 (AUC = 0.7969), and 48-hour post-implantation HIF-1α (AUC = 0.8889).

CONCLUSION

When integrated with established risk factors and pending external validation, these biomarkers may predict EAD and enhance early risk stratification in SCS recipients.

Key Words: Early allograft dysfunction; Ischemia reperfusion injury; Oxidative stress; Biomarkers; Liver transplantation

Core Tip: Early allograft dysfunction (EAD) remains common in liver transplantation, affecting roughly 20%-25% of recipients of allografts preserved via static cold storage, the traditional mode of organ preservation. Dynamic changes in serum cytokine and transcription factor levels - lower baseline interleukin-6, rising interferon-gamma-induced protein 10 peri-implantation, and elevated post-implantation hypoxia inducible factor-1 alpha are associated with EAD, suggesting potential early biomarkers. After validation in a larger study, peri-implantation biomarker profiling may help identify high-risk allografts early, potentially guiding management.

INTRODUCTION

Early allograft insufficiency remains a significant challenge following orthotopic liver transplantation (OLT). More than 10000 deceased donor liver transplant are performed each year in the United States[1], and roughly one in four recipients develops evidence of allograft impairment within the first postoperative week, a clinical syndrome termed early allograft dysfunction (EAD)[2,3]. The onset of EAD portends worse short- and long-term postoperative outcomes, including refractory coagulopathy, metabolic derangements, infectious complications, vasoplegia, allograft loss, multiorgan failure, and mortality[4,5].

Susceptibility to EAD reflects a convergence of donor, recipient, and surgical factors that influence the severity of ischemia-reperfusion injury (IRI), the central pathophysiologic driver of EAD, by modulating the degree of immune dysregulation and cellular injury following liver transplantation[6-9]. Well established donor-related risks include advanced donor age, donation after circulatory death (DCD), prolonged ischemic intervals, and hepatic steatosis, while recipient characteristics such as older age, higher body mass index (BMI), and greater severity of liver disease further increase vulnerability[2,3]. At the cellular level, IRI begins with transient interruption of oxygen delivery followed by abrupt reoxygenation, triggering oxidative stress, immune activation, and endothelial dysfunction that culminate in hepatocellular injury[6-8,10].

Persistent donor shortages have expanded the donor pool to include higher-risk extended criteria donors (ECDs) and marginal quality allografts, which are more susceptible to IRI and may contribute to higher rates of EAD[10-12]. Although ex vivo machine perfusion (MP) can reduce ischemic injury compared with conventional static cold storage (SCS)[13-16], SCS remains the predominant preservation method worldwide - including 84.5% of liver transplants in the United States in 2023 - due to its simplicity, accessibility, and cost-effectiveness[11,17-19]. As higher-risk allografts are increasingly utilized, identifying reliable predictive biomarkers of EAD in SCS recipients is critical for risk stratification and optimizing post-transplant management.

Small single center studies have examined perioperative cytokine dysregulation in SCS recipients[20,21]. Although results have been heterogenous, collectively they suggest the preoperative and perioperative inflammatory milieu may contribute to IRI and the development of EAD. One study linked biopsy-proven IRI to elevated preoperative levels of tumor necrosis factor-α (TNF-α), interleukin (IL)-2, IL-5, IL-7, IL-8, IL-13, and IL-1 receptor antagonist (IL-1Ra), along with persistent intraoperative increases in IL-8[20]. Another small study demonstrated lower preoperative levels of IL-6 and IL-2R were associated with EAD, while postoperative elevations in monocyte chemoattractant protein-1 (MCP-1), IL-8, regulated upon activation, normal T-cell expressed and secreted (RANTES), monokine induced by gamma interferon (MIG), induced protein 10 (IP-10), and IL-2R during the first 90 days suggested ongoing nuclear factor kappa B activation and T cell involvement in the broader postoperative period[22].

Biomarker profiles during the narrower peri-implantation period and their relationship to EAD remain largely unexplored. Based on prior data, we hypothesized specific alterations in cytokines and transcription factors during this period correlate with EAD in SCS recipients, potentially revealing predictive biomarkers for early detection and intervention.

MATERIALS AND METHODS

Study design and participant selection

This prospective, observational pilot cohort study was conducted between August 2023 to July 2024 at Emory University Hospital in Atlanta, GA, United States, a tertiary referral center performing more than 125 deceased donor OLTs annually[23]. Institutional review board approval was obtained from Emory University (protocol number 5728), and the detailed study methodology has been described previously[24]. The study cohort comprised adult recipients (≥ 18 years) undergoing OLT using SCS-preserved deceased donor allografts, including both donation after brain death and DCD donors. Exclusion criteria included allografts preserved via machine perfusion, living donor liver transplantation, early re-transplantation within 7 days of the index procedure (original case retained), multiorgan transplantation, and pregnancy or breastfeeding.

Study measures and outcomes

Serum cytokines and transcription factors were selected as candidate biomarkers of EAD based on prior publications and biologic plausibility (Supplementary Table 1)[6,20,22,25]. These included interleukin-1 beta, TNF-α, IL-2Ra, interferon-gamma IP-10, MIG, IL-8, MCP-1, RANTES, pentraxin 3, interferon-gamma, IL-6, and hypoxia inducible factor-1 alpha (HIF-1α). The study’s primary outcome was EAD, defined according to the Olthoff criteria as the occurrence of at least one of the following: Total bilirubin ≥ 10 mg/dL on postoperative day (POD) 7, international normalized ratio ≥ 1.6 on POD 7, or peak aspartate transaminase (AST) or alanine transaminase (ALT) exceeding 2000 IU/L within the first 7 days after transplantation[2]. Secondary outcomes were peak AST level within the first postoperative week (an established surrogate for IRI and postoperative allograft function)[26,27], postoperative acute kidney injury (AKI) per the Kidney Disease Improving Global Outcomes criteria, extubation in the operating room (OR), intensive care unit (ICU) and hospital length of stay (LOS), and all-cause mortality at 30-day.

Sample size calculation and human subject enrollment

A power analysis was performed using exploratory data from Friedman et al[22] Postoperative day 1 IL-8 levels in patients with and without EAD were used to estimate effect size. As only medians and interquartile ranges (IQRs) were reported, means were approximated using the medians, and standard deviations were estimated by dividing the IQR by 1.35. A post-hoc, two-tailed power analysis with an effect size of 0.717, a = 0.1, and group sizes of 29 (EAD) and 44 (non-EAD) indicated 84% power. Based on the same parameters, achieving 70% power a priori would require a minimum of 40 patients.

Study eligibility was assessed by a thorough review of electronic medical records. Patients were approached for enrollment while in the hospital ward, ICU, or preoperative area, with no restrictions based on demographic characteristics. Confidential written informed consent was obtained from all participants in accordance with the ethical principles outlined in the Declarations of Helsinki and Istanbul. Participation in the study had no impact on the patients’ standard medical care.

Procedures

The protocol for collection and measurement of biomarkers and standard of care (SOC) laboratory tests has been previously described[24]. Blood samples were obtained via an arterial line at three designated timepoints (T): (T1) after induction of general anesthesia before surgery (baseline, pre-implantation), (T2) two hours after allograft implantation (i.e., in situ reperfusion), and (T3) 48 hours after allograft implantation. At each timepoint, roughly 12 mL of blood was drawn into two EDTA tubes (approximately 6 mL each), then processed within 30 minutes by centrifugation at 2000 × g for 20 minutes at 4 °C. Plasma was aliquoted and stored at -80 °C for subsequent batch analysis according to institutional biorepository protocols. Biomarker concentrations were determined using Meso Scale Discovery electrochemiluminescence immunoassays (Meso Scale Diagnostics LLC, Rockville, MD, United States) per manufacturer guidelines. Comparisons of serum biomarker levels were made between patients who developed EAD and those who did not. Postoperatively, patients were monitored for 30 days, with weekly chart reviews documenting relevant exposures, covariates, and outcomes, including established risk factors for EAD[2,3] and perioperative immunosuppressive therapy. SOC laboratory tests, including metabolic panels, coagulation profiles, arterial blood gases, bilirubin, and liver enzymes, were collected at admission, peri-implantation, and post-operatively in the ICU. Artificial intelligence was used solely for language editing and refinement of this manuscript. It did not contribute to the conception, design, data acquisition, analysis, interpretation, or generation/drafting of the manuscript content.

Statistical analysis

Baseline demographic and clinical characteristics were summarized using descriptive statistics. Continuous variables were reported as mean ± SD or median with IQR, depending on their distribution as determined by the Shapiro-Wilk test, while categorical variables were expressed as frequencies and percentages. Comparisons between the cytokine and transcription factor levels of patients who developed EAD and those who did not were made using the following statistical tests: Student’s t-test or Mann-Whitney U test for continuous measures and χ² or Fisher’s exact test for categorical measures. Temporal changes in serum biomarker levels across the three study timepoints were examined using repeated measures ANOVA or the Friedman test, with post hoc pairwise comparisons when indicated. Relationships between biomarker levels and clinical or laboratory parameters were explored using Pearson or Spearman correlation coefficients. Analyses were performed on available data, with cases containing missing values excluded. Given the sample size and exploratory nature of the study, multivariable adjustment was not undertaken to avoid overfitting and maintain clarity of interpretation. Statistical significance was set at a two-sided P value < 0.05. All analyses were performed using SAS 9.4 (SAS institute Inc., Cary, NC, United States) or Python.

RESULTS

Patient characteristics

A total of 135 patients were eligible for enrollment. Thirty-one were excluded, including 11 undergoing multiorgan transplantation, 1 with recent re-transplantation, and 20 involving machine perfusion. Although MP cases were included in the broader OXIDATIVE cohort, where this was not an exclusion criterion, they were excluded from this analysis. One such case in the OXIDATIVE cohort was excluded due to aborted surgery prior to allograft implantation and is not relevant to this study. Enrollment was non-consecutive due to limited availability of the principal investigator and/or study team, with patients enrolled at varying times through the day and night. All patients who were approached consented to participate. The final cohort included 24 patients, of whom 1 (4.2%) received an allograft procured with normothermic regional perfusion and subsequently preserved with SCS prior to implantation.

Blood samples were successfully collected at T1 and T2 for all participants; at T3, 7 samples (29.2%) were unavailable due to arterial line removal per the SOC. Among the 24 included patients, 18 (75%) were from standard criteria donors and 6 (25%) were from ECDs, including 3 from DCD donors. EAD was observed in 8 of 24 patients (33.3%), a rate modestly higher than the 20%-25% typically reported in prior studies[2,3]. Of the 8 patients who developed EAD, 7 (87.5%) met criteria based on AST levels (one of whom also met criteria based on ALT levels), while 1 patient (12.5%) met criteria based on total bilirubin level. Patient characteristics by outcome (EAD vs non-EAD) are shown in Table 1.

Table 1 Baseline characteristics of static cold storage recipients by early allograft dysfunction status, n (%).

Variable	Subset	Non-EAD (n = 16)	EAD (n = 8)
Donor age	Median (IQR)	41.5 (30.5, 52)	55.5 (38, 61.5)
Donor mode of death	DCD	2 (13)	1 (13)
	ECD	2 (13)	1 (13)
	SCD	12 (74)	6 (74)
Recipient age	Median (IQR)	58 (43, 64)	59 (51.5, 63.5)
Recipient gender	Female	8 (50)	3 (38)
	Male	8 (50)	6 (62)
BMI	Median (IQR)	25.5 (23.5, 30)	34 (32.5, 36)
MELD score	Median (IQR)	24.5 (18, 35)	28 (27, 35)
Primary ESLD etiology	Acute liver failure	2 (12)	1 (12)
	Alcohol-related cirrhosis	5 (32)	0
	Autoimmune	0 (0)	1 (12)
	Graft failure	1 (6)	0
	HCC	1 (6)	0
	HCV	0	3 (38)
	MASH	4 (25)	3 (38)
	PBC	1 (6)	0
	PSC	2 (12)	0
Comorbidities	No	2 (12)	2 (25)
Hepatic encephalopathy	Yes	14 (88)	6 (75)
Hepatopulmonary syndrome	No	12 (75)	7 (88)
	Yes	4 (25)	1 (12)
Non-obstructive CAD	No	11 (69)	5 (63)
	Yes	5 (31)	3 (37)
Obstructive CAD	No	16 (100)	7 (88)
	Yes	0	1 (12)
Hepatorenal syndrome	No	10 (63)	5 (63)
	Yes	6 (37)	3 (37)
Esophageal varices	No	5 (31)	4 (50)
	Yes	11 (69)	4 (50)
Ascites	No	4 (25)	3 (38)
	Yes	12 (75)	5 (62)
Hypertension	No	0	0
	Yes	16 (100)	8 (100)
Pre-operative laboratory values:
Creatinine (mg/dL)	Median (IQR)	0.95 (0.72, 1.33)	1.11 (0.84, 1.53)
INR	Median (IQR)	1.72 (1.48, 2.76)	1.36 (1.1, 2.25)
Hemoglobin (g/dL)	Median (IQR)	9.45 (8.9, 12.25)	10.9 (8.7, 13.05)
Platelets (109/L)	Median (IQR)	82 (62, 118)	88 (62.5, 127.5)
Fibrinogen (mg/dL)	Median (IQR)	169 (137, 138)	204.5 (122.5, 301.5)
Total ischemia time (hours)	Median (IQR)	6.42 (5.17, 7.18)	6.17 (4.64, 8.07)
Cold ischemia (hours)	Median (IQR)	5.81 (4.59, 6.69)	5.71 (3.9, 7.49)
Warm ischemia (minutes)	Median (IQR)	32.5 (29, 37.5)	35 (28, 48)
Inferior vena cava clamp type	Piggyback	16 (100)	6 (75)
	Total cross clamp	0	2 (25)
Day 7 INR	Median (IQR)	1.03 (0.94, 1.1)	1.17 (1.05, 1.47)
Day 7 total bilirubin (mg/dL)	Median (IQR)	2.05 (0.85, 4.25)	2.55 (2.1, 5.1)
Peak 7-day AST (U/L)	Median (IQR)	997 (609, 1509.5)	3215 (2413, 6318)
Peak 7-day ALT (U/L)	Median (IQR)	482 (230, 957.5)	1304 (543, 1948)
Methylprednisolone exposure (relative to T2)	Before	6 (38)	3 (38)
	After	10 (62)	5 (62)
Received basiliximab		8 (50)	4 (50)
Basiliximab exposure (relative to T2)	Before	3 (37)	1 (25)
	After	5 (62)	3 (75)

AST: Aspartate transaminase; ALT: Alanine transaminase; INR: International normalized ratio; SCDs: Standard criteria donors; ESLD: End-stage liver disease; MASH: Metabolic dysfunction-associated steatohepatitis; HCC: Hepatocellular carcinoma; HCV: Hepatitis C virus; MELD: Model for end-stage liver disease; EAD: Early allograft dysfunction; IQR: Interquartile ranges; DCD: Donation after circulatory death; ECD: Extended criteria donor; BMI: Body mass index; PBC: Primary biliary cirrhosis; PSC: Primary sclerosing cholangitis; CAD: Coronary artery disease.

Overall, women accounted for just under half of participants (45.8%). The most frequent etiologies of end-stage liver disease were metabolic dysfunction-associated steatohepatitis, observed in 7 patients (29.2%), and alcohol-related cirrhosis, present in 5 patients (20.8%). Acute liver failure was an indication for liver transplantation in 3 of 24 patients (12.5%), while hepatocellular carcinoma contributed as a primary or secondary indication in 1 patient (4.2%). No alcohol-related cirrhosis cases developed EAD, whereas all autoimmune and hepatitis C virus (HCV) cases did, with HCV representing 37.5% of EAD cases. The median donor age was 44.5 years (IQR: 31.5, 56.5) and the median recipient age was 58 years (IQR, 45, 64.3). Recipients had a median BMI of 30 kg/m² (IQR: 24, 34.5) and a median model for end-stage liver disease (MELD) score of 27 (IQR: 19.3, 36).

Recipients who developed EAD were more often male (62% vs 50%), received allografts from older donors [55.5 years (IQR: 38, 61.5) vs 41.5 years (IQR: 30.5, 52)], had higher BMI [34 kg/m² (IQR: 32.5, 36) vs 25.5 kg/m² (IQR: 23.5, 30)], and exhibited higher MELD scores [28 (IQR: 27, 35) vs 24.5 (IQR: 18, 35)] compared with those who did not develop EAD. EAD patients also demonstrated a higher median AST level [3215 U/L (IQR: 2413, 6318) vs 997 U/L (IQR: 609, 1509.5)] during the first postoperative week. Methylprednisolone administration timing (before vs after T2) did not significantly differ between EAD and non-EAD groups (P = 1.00).

Biomarker levels and EAD

Among SCS recipients, EAD was significantly associated with a lower baseline IL-6 level at T1 [6.3 pg/mL (IQR: 3.7, 12.9) vs 15.4 pg/mL (IQR: 8.7, 24.9), P = 0.0433], a peri-implantation increase in IP-10 (T2-T1) [386.7 pg/mL (IQR: -26.7, 1280.7) vs -181.5 pg/mL (IQR: -452.4, 145.9), P = 0.02], and an elevated HIF-1α level at T3, 48 hours post-implantation [611.2 pg/mL (IQR: 384.7, 869.9) vs 157.9 pg/mL (IQR: 110, 273), P = 0.0095] (Table 2). T3 biomarker levels were missing for 7 recipients (29.2%), the majority of whom belonged to the non-EAD group (6 of 7, 85.7%), compared with a single recipient in the EAD group (1 of 7, 14.3%). Boxplots comparing IL-6 at T1, peri-implantation changes in IP-10 (T2-T1), and HIF-1α at T3 between non-EAD and EAD groups are presented in Figure 1 (remaining timepoints are shown in Supplementary Figure 1).

Figure 1 Boxplots comparing levels of each statistically significant biomarker at its corresponding timepoint. A: Interleukin-6 at T1 (baseline), P = 0.0433; B: Peri-implantation interferon-gamma-inducible protein 10 (T2-T1), P = 0.02; C: Hypoxia inducible factor-1 alpha at T3 (48 hours post-implantation), P = 0.0095, between non-early allograft dysfunction (EAD) and EAD groups. HIF-1α: Hypoxia inducible factor-1 alpha; IL-6: Interleukin-6; IP-10: Interferon-gamma-inducible protein 10; SCS: Static cold storage.

Table 2 Serum immunomodulatory protein levels at each timepoint for early allograft dysfunction and non-early allograft dysfunction groups among static cold storage recipients.

Biomarker	Timepoint	No EAD (n = 16)	EAD (n = 8)	P value
HIF-1α	T1	341.2 (88.1, 855.8)	588.3 (76.4, 1285.7)	0.7133
	T2	520.7 (415.6, 1007.6)	1087 (608.7, 1758)	0.2207
	T3	157.9 (110, 273)	611.2 (384.7, 869.9)	0.0095
	T2-T1	171.2 (-25.8, 423.8)	430.6 (86.3, 739.7)	0.1590
IL-2Rα	T1	2492 (1684, 4685)	2080 (1357, 2781)	0.1984
	T2	1158 (803, 2226)	1402 (737.7, 1674)	0.5403
	T3	3614 (3426, 4495)	3068 (2719, 5631)	0.3798
	T2-T1	-888.9 (-3419.4, 227.7)	-654.9 (-1659.3, -267.4)	0.7133
IP-10	T1	946.8 (369.2, 1304.5)	616.6 (381.2, 716.8)	0.3272
	T2	469.5 (383.7, 686.1)	871.1 (491.2, 1897)	0.1779
	T3	920.6 (644.5, 2371)	976.8 (449.4, 1212)	0.4945
	T2-T1	-181.5 (-452.4, 145.9)	386.7 (-26.7, 1280.7)	0.0200
MCP-1	T1	476.9 (330.1, 644.9)	390.6 (352.1, 497.5)	0.4624
	T2	3251 (1276, 5205)	3317 (2136, 5470)	0.6242
	T3	451.3 (386.2, 628.9)	725.3 (517.4, 1089)	0.2046
	T2-T1	2742.4 (939.7, 4605.5)	2967 (1473, 5074)	0.5815
RANTES	T1	2258 (1108, 4669)	4557 (1407, 13462)	0.2207
	T2	3002 (765, 4140)	2209 (1284, 7100)	0.4624
	T3	525.8 (417.7, 1124.5)	349.5 (222.4, 613.8)	0.4945
	T2-T1	-486.5 (-2029.5, 1915.7)	-188.2 (-3805, 2191)	0.9512
IL-1β	T1	0.59 (0.43, 0.82)	0.42 (0.3, 0.58)	0.2724
	T2	2.6 (1.1, 4.2)	2.3 (1.2, 5.5)	0.5815
	T3	0.5 (0.47, 1.02)	0.59 (0.43, 0.82)	0.8453
	T2-T1	2.14 (0.57, 3.61)	1.74 (0.75, 4.96)	0.6056
TNF-α	T1	2.21 (1.48, 4.07)	2.3 (1.8, 2.8)	1.000
	T2	4.1 (2.6, 17.1)	5.4 (3.3, 8.2)	0.6682
	T3	3.8 (2.2, 8.7)	3.9 (2.3, 6.1)	0.9223
	T2-T1	1.3 (0.35, 11.3)	2.59 (0.96, 5.6)	0.8065
MIG	T1	162.6 (95.5, 224.5)	127.1 (90.5, 142.4)	0.3272
	T2	320.5 (221.4, 501.5)	511.2 (348.1, 1022.5)	0.1258
	T3	473.3 (193.7, 593.4)	578.9 (194.1, 1269)	0.5582
	T2-T1	171.5 (83.6, 313.4)	414.7 (239.9, 898)	0.0662
IL-8	T1	28.1 (19.7, 123.9)	75.3 (11.5, 126.4)	0.8065
	T2	375.7 (79.9, 731.8)	457.9 (274.5, 1708)	0.2446
	T3	40.9 (29, 58.5)	117.7 (46.4, 164.8)	0.1432
	T2-T1	348.2 (60.5, 677.5)	387.7 (164.4, 1613.3)	0.2207
IL-6	T1	15.4 (8.7, 24.9)	6.3 (3.7, 12.9)	0.0433
	T2	95.1 (57.7, 168.3)	130.6 (96, 434.1)	0.1258
	T3	6.8 (5.9, 12.2)	8.4 (7.9, 12)	0.1432
	T2-T1	82.7 (52.1, 152.5)	121.7 (92.8, 430.3)	0.0982
PTX3	T1	26350 (16282, 46843)	19700 (13195, 74614)	0.5815
	T2	93788 (74749, 118562)	95021 (42785, 121719)	0.6242
	T3	98274 (56929, 138102)	139127 (85375, 263959)	0.0971
	T2-T1	52611 (42493, 73971)	58756.5 (16182.5, 78433.9)	0.6682
IFN-γ	T1	9.4 (5.4, 26.1)	9.6 (4.2, 11.8)	0.3745
	T2	20.2 (8.8, 38.2)	23.9 (7.3, 44.4)	0.9512
	T3	8.7 (3.4, 52.8)	9.1 (4.5, 11.1)	1.000
	T2-T1	2.8 (-1.9, 11.3)	13.14 (3.2, 29.6)	0.0982

HIF-1α: Hypoxia inducible factor-1 alpha; IL: Interleukin; IP-10: Interferon-gamma-inducible protein 10; MCP-1: Monocyte chemoattractant protein-1; RANTES: Regulated upon activation, normal T-cell expressed and secreted; IL-1β: Interleukin-1 beta; TNF-α: Tumor necrosis factor-α; MIG: Monokine induced by gamma interferon; PTX3: Pentraxin 3; IFN-γ: Interferon-gamma.

Serum IL-6, IP-10, and HIF-1α as predictive biomarkers of EAD

Receiver operating characteristic (ROC) curve analysis demonstrated baseline IL-6 levels had moderate discriminatory ability for predicting EAD [area under the curve (AUC) = 0.7578; 95%CI: 0.5195-0.9961]. The optimal IL-6 cutoff, determined by a Youden index of 0.188, was 43.3 pg/mL, yielding a sensitivity of 62.5% and specificity of 56.3%. Peri-implantation IP-10 exhibited good discriminatory performance (AUC = 0.7969; 95%CI: 0.6132-0.9806). The optimal IP-10 cutoff, determined by a Youden index of 0.5, was -191.8 pg/mL, corresponding to a sensitivity of 100% and specificity of 50%. HIF-1α levels at 48 hours post-implantation showed strong discriminatory ability (AUC = 0.8889; 95%CI: 0.7131-1.0000). The optimal HIF-1α cutoff, determined by a Youden index of 0.714, was 551.6 pg/mL, demonstrating a sensitivity of 71.4% and specificity of 100%. ROC curve analyses are shown in Figure 2.

Figure 2 Receiver operating characteristic curve analysis for the discriminatory ability of each statistically significant biomarker at its corresponding timepoint. A: Baseline interleukin-6; B: Peri-implantation interferon-gamma-inducible protein 10; C: 48-hour post-implantation hypoxia inducible factor-1 alpha as predictive biomarkers of early allograft dysfunction. HIF-1α: Hypoxia inducible factor-1 alpha; IL-6: Interleukin-6; IP-10: Interferon-gamma-inducible protein 10; ROC: Receiver operating curve.

EAD biomarkers and secondary outcomes in SCS recipients

Baseline IL-6 levels at T1, peri-implantation (T2-T1) IP-10 levels, and 48-hour post-implantation HIF-1α levels at T3 were evaluated for associations with peak AST level within 7 days postoperatively, post-operative AKI, extubation in the OR, ICU LOS, hospital LOS, and 30-day all-cause mortality (Supplementary Table 2).

Baseline IL-6 was moderately correlated with peak AST level within 7 days postoperatively (r = -0.51043, P = 0.0108), a well-established surrogate marker of allograft injury[28] and component of the EAD definition[2]. No significant associations were observed between baseline IL-6 and the other secondary outcomes. Similarly, peri-implantation changes in IP-10 were weakly-to-moderately correlated with peak AST level within 7 days (r = 0.48174, P = 0.0171) and ICU LOS (r = 0.52649, P = 0.0082) but was not associated with the remaining secondary outcomes. Post-implantation HIF-1α at 48 hours was moderately correlated with peak AST level within 7 days postoperatively (r = 0.55588, P = 0.0254), without significant relationships to other outcomes.

DISCUSSION

In our cohort of SCS recipients, a distinct peri-implantation pattern of circulating cytokines and transcription factors was associated with the development of EAD. Specifically, EAD was characterized by a lower baseline IL-6 level, greater peri-implantation increase in IP-10, and higher HIF-1α level 48 hours post-implantation. Pending external validation, these cytokines and transcription factors may have utility as predictive biomarkers of EAD.

We acknowledge baseline clinical differences between EAD and non-EAD groups. As an observational pilot study, the design was not randomized, and such imbalances are anticipated. EAD is multifactorial, reflecting contributions from donor, recipient, and surgical factors, with IRI as the central unifying mechanism. Established EAD risk factors, such as older donor age, higher recipient BMI, and greater macrosteatosis, likely modulate this process through related cellular and immunologic pathways.

Exploratory analysis of biomarker dynamics in SCS recipients with EAD

In our study, lower baseline IL-6 levels were associated with EAD, consistent with prior studies[22]. The role of IL-6 in the development of EAD is complex, reflecting its tightly regulated, time-dependent pro- and anti-inflammatory effects during hepatic IRI. During ischemia, IL-6 signaling appears to be hepatoprotective by attenuating hepatocyte apoptosis, limiting innate immune activation, and promoting hepatocyte survival and regeneration[29,30]. Conversely, excessive or sustained IL-6 signaling following implantation and in situ reperfusion may amplify inflammatory cascades and contribute to allograft dysfunction[31,32].

We speculate this finding may reflect inadequate pre-implantation immune conditioning or preparedness, thereby limiting early adaptive responses and predisposing the allograft to exaggerated inflammatory injury upon implantation and reperfusion. Furthermore, in alignment with previous studies[31,33], IL-6 levels increased peri-implantation and remained higher post-implantation among those who developed EAD compared with those who did not; however, these differences did not reach statistical significance in our study, which could be due to our small sample size.

Consistent with prior studies[22,34] we observed elevated perioperative circulating IP-10 levels were associated with EAD; notably, our study uniquely demonstrates this increase within the more narrowly defined peri-implantation period. Our data is consistent with enhanced chemotactic recruitment of activated T cells, B cells, natural killer cells, and dendritic cells to the allograft, a process well described in hepatic IRI[35]. Elevated IP-10 levels in other studies have correlated with markers of proinflammatory activation and hepatic IRI, including increased MCP-1 and IL-8[22], which were also higher peri-implantation in recipients who developed EAD in our cohort, although these differences did not reach statistical significance.

Unexpectedly, higher circulating HIF-1α levels 48 hours post-implantation were associated with EAD. However, missing data at this timepoint may introduce bias and limit interpretability. Accordingly, this observation should be viewed cautiously, though it underscores the context-dependent nature of hypoxia signaling in liver transplantation. Hypoxia inducible factors (HIFs) are key regulators of cellular adaptation to hypoxia. Under normoxic conditions, HIF subunits undergo oxygen-dependent degradation, whereas hypoxic stress suppresses this degradation and allows HIF accumulation and transcription of genes involved in cellular survival[36]. Accordingly, HIF-1α is traditionally regarded as hepatoprotective rather than directly injurious. The observed post-implantation increase in HIF-1α may represent a compensatory response to oxidative stress aimed at promoting metabolic recovery. Alternatively, persistent or excessive activation following reperfusion may signal maladaptive regeneration or unresolved tissue injury. It is thus plausible both insufficient and excessive HIF-1α activity is deleterious, depending on timing and biological context.

Interpretation of circulating HIF-1α is further complicated by its primarily intracellular function[24], as it is traditionally studied using tissue-based techniques such as Western blot or immunohistochemistry. However, HIF-1α may enter the bloodstream through passive leakage from damaged or dying cells[37], as occurs during IRI and EAD. Consequently, circulating HIF-1α may serve as an indirect indicator of ongoing tissue-level processes, rather than baseline hypoxia signaling. Despite these challenges, circulating HIF-1α has demonstrated clinical relevance as a biomarker in other disease contexts[38-41], and our findings suggest it may similarly hold value in the setting of EAD. Hypoxia signaling represents a promising avenue for future research in OLT, as elucidating the role of HIFs may improve risk stratification for EAD and inform the development of targeted protective interventions[6].

Another implication of our findings is select cytokines and transcription factors demonstrate variable but potentially complementary discriminatory performance for EAD. Baseline IL-6 was modestly predictive of EAD with balanced but moderate sensitivity and specificity. Peri-implantation changes in IP-10 showed good discriminatory ability, characterized by high sensitivity but limited specificity. Notably, failure to exhibit a substantial peri-implantation decrease in IP-10, or a relative increase, was associated with EAD, though the relatively low specificity indicates this marker may be better suited for risk stratification or screening than definitive prediction. Conversely, HIF-1α measured 48 hours post-implantation exhibited strong discriminatory capacity, with high specificity and moderate sensitivity, supporting its potential role as a diagnostic biomarker.

Together, these findings suggest complementary roles across timepoints and support the potential utility of a multi-marker approach for EAD risk stratification, prediction, and diagnosis. However, these results should be interpreted cautiously given the exploratory design, small sample size, and missing T3 data. These associations reflect statistical discrimination within this cohort and do not establish causality. External validation is needed to assess reproducibility, calibration, and clinical applicability.

When considered alongside established risk factors, such as donor age, total ischemia time, and recipient BMI and MELD score, the addition of baseline IL-6, peri-implantation IP-10, and 48-hour post-implantation HIF-1α may predict the development of EAD and incrementally improve risk stratification among SCS recipients. Nevertheless, prospective validation and assessment of clinical utility are required before these biomarkers can be incorporated into routine clinical frameworks.

Study limitations

Several limitations should be considered when interpreting our findings. Enrollment was limited by principal investigator and/or study team availability, resulting in non-consecutive patient inclusion and potential selection bias. However, the lack of targeted enrollment (patients were recruited at all hours and across a broad range of MELD scores) likely yielded a reasonably representative sample under the circumstances.

To reduce heterogeneity, we excluded all patients whose allografts were preserved via machine perfusion, as this approach decreases total ischemia time compared with SCS. One SCS patient received an allograft procured using NRP, which reduces donor warm ischemia time. Although both MP and NRP mitigate ischemic injury, they do so via distinct mechanisms. Because NRP is not a form of allograft machine perfusion, this SCS recipient was retained in the analysis. We acknowledge this patient likely experienced reduced donor warm ischemia time relative to other SCS recipients; however, this variable was not recorded. Given the overall small sample size, we elected to include this case in the analysis.

Overall, the modest sample size and cohort heterogeneity limited statistical power to detect definitive biomarker differences between EAD and non-EAD groups, and missing measurements at the T3 48-hour post-implantation timepoint further constrained analyses. The small number of events also precluded stable multivariable modeling of all significant biomarkers due to risk of overfitting. As an exploratory pilot study, these results should therefore be interpreted cautiously and viewed as hypothesis-generating, requiring validation in larger cohorts.

Measured serum cytokine and transcription factor levels may have been affected by differences in perioperative immunosuppression. At Emory University Hospital, patients receive intraoperative methylprednisolone during the neohepatic phase once hemostasis is achieved, followed by a brief postoperative course. Mycophenolate mofetil is typically started on the first postoperative night, and tacrolimus on postoperative day 1 or 2, depending on renal function. Patients with reduced creatinine clearance (< 60 mL/minute) receive intraoperative basiliximab to delay tacrolimus initiation and reduce the risk of kidney injury. Accordingly, T1 samples were collected before any immunosuppressive therapy, T2 samples primarily reflect exposure to methylprednisolone and, in a few cases basiliximab, and T3 samples reflect exposure to multiple agents.

This variability is most likely to influence measurements at T3, the timepoint from which we draw the least inference due to missing samples. We speculate the timing of T2 sampling likely limited the influence of methylprednisolone on measured biomarkers. Detectable cytokine modulation from methylprednisolone generally occurs no earlier than 4 hours after administration, including suppression of proinflammatory mediators such as TNF-α and IL-8[42,43]. In our cohort, the relative timing of methylprednisolone administration with respect to T2 did not differ between EAD and non-EAD groups. Even in cases where methylprednisolone was given before T2, samples were collected within one hour of dosing, further limiting potential effects. Taken together, these factors suggest methylprednisolone exposure had minimal impact on peri-implantation biomarker levels in this study. Basiliximab, on the other hand, has a fast onset and may have influenced IL-2Ra levels.

Another potential source of bias is intraoperative blood product transfusion. Large-volume transfusion could theoretically dilute circulating biomarker levels, leading to artificially lower measured serum concentrations. If this effect were substantial, reduced biomarker levels would be expected at T2 and T3; however, we did not observe a consistent pattern suggesting a major dilution effect.

CONCLUSION

In this cohort of SCS recipients, EAD occurred in about one-third of patients and was associated with lower baseline IL-6, greater peri-implantation increases in IP-10, and higher 48-hour post-implantation HIF-1α. These biomarkers demonstrated moderate-to-strong discriminatory performance for EAD and when considered alongside established risk factors, including donor age, total ischemia time, recipient BMI, and MELD score, may enhance early risk stratification. However, given the small sample size and exploratory design, these findings require external validation before clinical application.

ACKNOWLEDGEMENTS

The authors would like to thank many staff for their contribution to this project, including: (1) The Georgia Clinical & Translational Science Alliance (CTSA) and its Clinical Research Center for providing laboratory resources; (2) The Emory Research Hemostasis & Coagulation Core Laboratory for storing samples and helping perform the serum biomarker analyses; and (3) The Biostatistics Collaboration Core at Emory University’s Rollins School of Public Health for assisting in the statistical analysis. Preliminary data was not previously presented. Members of the OXIDATIVE study group include the following: (1) Shagun Mathur, MD, Assistant Professor, Department of Anesthesiology, Emory University School of Medicine, Atlanta, GA, United States; (2) Devin Weinberg, MD, PhD, Assistant Professor, Department of Anesthesiology, Emory University School of Medicine, Atlanta, GA, United States; (3) Kati Running, MD, Assistant Professor, Department of Anesthesiology, Emory University School of Medicine, Atlanta, GA, United States; (4) Gaurav P Patel, MD, Associate Professor, Department of Anesthesiology, Emory University School of Medicine, Atlanta, GA, United States; (5) Susan Smith, MD, Assistant Professor, Department of Anesthesiology, Emory University School of Medicine, Atlanta, GA, United States; (6) Ricky Matkins, MD, Assistant Professor, Department of Anesthesiology, Emory University School of Medicine, Atlanta, GA, United States; and (7) Cinnamon Sullivan, MD, Associate Professor, Department of Anesthesiology, Emory University School of Medicine, Atlanta, GA, United States.