Early Allograft Dysfunction After Live Donor Liver Transplantation: It's Time to Redefine?

Abstract

Background

An ideal definition of early allograft dysfunction (EAD) after live donor liver transplantation (LDLT) remains elusive. The aim of the present study was to compare the diagnostic accuracies of existing EAD definitions, identify the predictors of early graft loss due to EAD, and formulate a new definition, estimating EAD-related mortality in LDLT recipients.

Methods

Consecutive adult patients undergoing elective LDLT were analyzed. Patients with technical (vascular, biliary) complications and biopsy-proven rejections were excluded.

Results

There were 19 deaths due to EAD of a total of 304 patients. On applying the existing definitions of EAD, we revealed their limitations of being either too broad with low specificity or too restrictive with low sensitivity in patients with LDLT. A new definition of EAD-LDLT (total bilirubin >10 mg/dL, international normalized ratio [INR] > 1.6 and serum urea >100 mg/dL, for five consecutive days after day 7) was derived after doing a multivariate analysis. In receiver operator characteristics analysis, an AUC for EAD-LDLT was 0.86. The calibration and internal cross-validation of the new model confirmed its predictability.

Conclusion

The new model of EAD-LDLT, based on total bilirubin >10 mg/dL, INR >1.6 and serum urea >100 mg/dL, for five consecutive days after day 7, has a better predictive value for mortality due to EAD in LDLT recipients.

Liver transplantation has become an established treatment for the patients with decompensated end-stage liver disease. Deceased donor liver transplantation (DDLT) is preferred, but in parts of the world having low organ donation rates, the live donor liver transplantation (LDLT) has evolved as an important therapeutic alternative, having comparable survival. The postoperative course of these two types of grafts is entirely different. In LDLT, as the grafts are of optimal quality and the cold ischemia time (CIT) is limited, the primary nonfunction (PNF) which is usually caused by ischemia-reperfusion (IR) injury is rare.^1,2 Live donor liver grafts are more likely to present with marginal function, also called as early allograft dysfunction (EAD), which usually sets in after the first week of transplantation. The lower graft and patient survival rates and increased postoperative morbidity have been reported in association with EAD.³ The available definitions for describing EAD are mostly derived from DDLT experiences with low positive predictive value (PPV) of 19%–68%.3, 4, 5, 6 The aim of the present study was to compare the diagnostic accuracies of existing definitions of EAD, identify the predictors of early graft loss, and formulate a new definition which can estimate the EAD-related mortality in LDLT recipients.

Materials and Methods

The study was approved by institutional ethics and scientific review board and is in accordance with the Declaration of Helsinki principles (2000) for medical research involving human subjects. Prospectively maintained an electronic database of the consecutive adult patients undergoing LDLT at the Institute of Liver and Biliary Sciences, New Delhi, from June 2011 to August 2018 was reviewed. A total of 470 LDLTs were performed during the study period. To obtain a uniform study population, we excluded pediatric recipients, adult patients transplanted for acute liver failure, patients receiving unusual graft types, and those having technical complications and immunological graft injury (Figure 1). The remaining 304 recipients were analyzed. We did not exclude sepsis as a cause of EGD, as the cause–effect relationship can not be ascertained. Sepsis can be secondary to EAD and vice versa. The clinical timing of both also coincides. After exclusion of all causes other than graft dysfunction, of 304, 19 patients had EAD-related mortality (Figure 1).

Figure 1.

Derivation of uniform study population. LDLT = living donor liver transplantation; ALF = acute liver failure; CLD = chronic liver disease; PNF = primary nonfunction; AMR = antibody-mediated rejection; HAT = hepatic artery thrombosis; PVT = portal vein thrombosis; SAE = splenic artery embolization; LHV = left hepatic vein; ACR = acute cellular rejection; EAD = early allograft dysfunction.

Surgical Technique and Perioperative Management

Our donor evaluation and selection protocol have been published previously.⁷ The surgical technique of donor hepatectomy, recipient operation, and postoperative care have been described previously.^8,9 Middle hepatic vein (MHV) is preserved in the donor and routinely reconstructed in the recipient. From September 2015, during the bench procedure, both portal and antegrade arterial flush of the liver grafts is routinely performed.¹⁰ We have never performed the inflow modulation. Our postoperative care and immunosuppression protocol have been published previously.¹¹

Definitions

Post-transplant complications were graded in accordance with Clavein-Dindo classification (CDC).¹² Patients having in-hospital mortality with the element of graft dysfunction, leading to early graft loss, and no other obvious cause were considered as EAD-related mortality. The existing definitions of EAD by Olthoff et al,⁵ A2ALL study group,³ Ikegami et al,⁴ and Okamura et al⁶ are depicted in Supplementary Table S1. In-hospital mortality was defined as death within 90 days after surgery.

Statistical Analysis

Statistical analyses were performed with SPSS Statistics Version 21 (IBM Corp. Armonk, NY). Continuous variables were expressed as medians (IQR) and compared with the Student t-test or the Mann–Whitney test as appropriate. The categorical data were compared with the chi-square or Fisher exact test as appropriate. Univariate and multivariate analyses were performed for the variables predicting early graft loss, using logistic regression. A two-tailed P-value < 0.05 was considered statistically significant. Calibration and internal cross-validation of the new EAD-LDLT definition was performed by the Hosmer–Lemeshow goodness of fit test and the K-fold cross-validation method. Finally, the receiver operator characteristics analysis was performed to identify the predictive accuracy and discriminative ability of each definition. Areas under curves (AUCs) were compared with the test of proportions.

Results

The baseline characteristics of the patients are depicted in Table 1. In the early graft loss group, the median (IQR) donor age was significantly higher (41{25–44} vs. 29 {21–37} years, P = 0.004) as compared with that in the group without any early graft loss. The median (IQR) absolute graft weight was significantly lower (572{460–726} vs. 684{604–760}; P = 0.03) in the early graft loss group (Table S2).

Table 1.

Baseline Characteristics of the Study Population.

Demographics	n = 304
Recipient age (years)	48 (41–53)
Recipient sex (%)	Male,87.2%
Donor age (years)	29 (23–38)
Donor sex (%)	Male,41.4%
Donor liver attenuation index	+10.46 (+8.6 to +12.8)
Chronic liver disease/ACLF (%)	89.5%/10.5%
Child-Pugh status (A/B/C) (%)	0.3%/10.5%/89.1%
MELD-Na score	23 (19–27)
Creatinine clearance (mL/min)	78.4 (63.5–95.7)
Serum urea level (mg/dL)	44 (36–51)
Etiology (%)
Alcohol	140 (44.7%)
Cryptogenic	43 (13.7%)
HBV	30 (9.5%)
HCV	24 (7.6%)
NASH	44 (14%)
Others	32
Intra operative parameters
Graft weight (grams)	681 (588–760)
GRWR	0.93 (0.82–1.1)
Cold ischemia time (mins.)	95 (79–114)
Warm ischemia time (mins.)	30 (22–38)
Anhepatic phase (mins.)	136 (112–172)
PVT (%)	11.2 (34/304)
Portal venous flow (L/min)	2.7 (1.0–6.5)
Surgery duration (mins.)	710 (620–820)
Blood loss (ml)	2400 (1500–3800)
Post-operative parameters
Ascites > 1L on day 14 (%)	31.3
Severe sepsis (%)	36.5
Hospital stay (days)	22 (18–29)
ICU/HDU Stay (days)	9 (7–13)
Retransfer to ICU (%)	12.5
Major morbidity (CDC ≥ IIIa) (%)	42.8

All the continuous variables are expressed as medians (IQR).

ACLF = acute on chronic liver failure; MELD = model for end-stage liver disease; HBV = hepatitis B virus; HCV = hepatitis C virus; NASH = nonalcoholic steatohepatitis; GRWR = graft-to-recipient weight ratio; PVT = portal vein thrombosis; AST = aspirate transaminase; ALT = alanine transaminase; INR = international normalized ratio; ICU = intensive care unit; HDU = high dependency unit.

Postoperative Outcomes

The overall in-hospital mortality was 7.8% (29/372). After eliminating other causes, of 304 patients, 19 patients (6.25%) had mortality due to graft insufficiency with early graft loss. Other causes of mortality are illustrated in Figure 1. As expected, in the early graft loss population, there was a significant rise in the median of peak total bilirubin, aspirate aminotransferase, alanine aminotransferase, international normalized ratio (INR), and serum urea and creatinine levels. However, there was a dramatic fall in the median platelet count. The major morbidity was 42.8%, with the rate of severe sepsis being 36.5%. As compared with patients without early graft loss, the rate of severe sepsis was also significantly higher in the early graft loss group (84.2% vs. 33.3%, P < 0.001) as well as the rate of major (CDC ≥ IIIa) complications (88.5 vs.39.6%, P < 0.001). The length of intensive care unit (ICU)/high dependency unit (HDU) stay, the rate of retransfer to ICU/HDU, and the overall duration of hospital stay were significantly higher in early graft loss group (Table S2). Interestingly, there was no association between the GRWR or the graft type and the proportion of recipients dying because of EAD (Figure S1).

Diagnostic Accuracies of the Existing Definitions of EAD

We applied the existing commonly used definitions of EAD on our study population. The Olthoff's, A2ALL study group and Okamura's definitions performed poorly with respect to PPV (17.5%, 17.9% and 21.1, respectively), whereas the definitions by Okamura et al and Ikegami et al showed very low sensitivity (26.7% and 21.1%, respectively) for predicting EAD-related mortality (Table 2).

Table 2.

Diagnostic Accuracies of the Existing Definitions of EAD.

Definition	Total population (n = 304)	EAD mortality (n = 19)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Olthoff et al	99 (32.5%)	17	89.7	71.7	17.5	99.0
A2ALL study	98 (32.2%)	17	88.7	70.5	17.9	98.6
Okamura et al	15 (4.9%)	4	26.7	96.1	21.1	94.8
Ikegami et al	6 (1.9%)	4	21.1	99.6	80.0	95.0

EAD = early allograft dysfunction; LDLT = live donor liver transplantation; PPV = positive predictive value; NPV = negative predictive value.

Derivation of the New EAD-LDLT Definition

In LDLT, the graft dysfunction usually evolves after the first postoperative week when the regenerating partial graft fails to meet the metabolic demands of the recipient.⁴ Thus, we concentrated on the chronological changes that occur in the laboratory parameters, after postoperative day (POD) 7. All 19 cases with early graft loss had a maximum total bilirubin>10 mg/dL (P < 0.001) and peak INR>1.6 (P < 0.001) after POD 7. In addition, there was a significant association of rise in the serum urea level after POD 7, with the maximum serum level being >100 mg/dL (P < 0.001, AUC = 0.906). As the EAD is a dynamic process, and also to account for the graft dysfunction due to other treatable causes, mainly sepsis and possibility of spontaneous recovery, we considered bilirubin >10 mg/dL, INR >1.6 and urea >100 mg/dL for 5 consecutive days after POD 7, individually, in the univariate and multivariate analysis. Univariate analysis of the predictors for early graft loss identified the donor age and the absolute graft weight as significant clinical factors. Among laboratory variables, total bilirubin >10 mg/dL, INR >1.6 and serum urea >100 mg/dL, each for 5 consecutive days after POD 7 were the predictors, which were significant in multivariate analysis as well (Table 3). All of the three abovementioned parameters defining EAD were evaluated individually for sensitivity and specificity. Individually also, all the three parameters predicted EAD-related mortality having better PPV than the Olthoff's definition and better sensitivity than the Ikegami's criteria. To increase the specificity and PPV, the bilirubin and INR parameters were accounted together for the analysis. Eighteen (5.7%) of 304 recipients satisfied these criteria (total bilirubin> 10 mg/dL + INR >1.6 for 5 consecutive days after day 7) with a high PPV rate of 77.8%. To make it further specific and stringent, all the three parameters were considered together, which increased the PPV to 87.5% (vs. that of Ikegami, 80%) with specificity of 99.3%, maintaining the same level of sensitivity of 73.7% as that of INR and bilirubin combined together (Table 4).

Table 3.

Univariate and Multivariate Analysis of the Risk Factors for Early Graft Loss.

Variables	Univariate analysis		Multivariate analysis
	OR (95% CI)	P-value	OR (95% CI)	P-value
Recipient age	0.98 (0.94–1.03)	0.65
Donor age	1.08 (1.02–1.13)	0.006	–	–
Donor gender (M)	0.36 (0.11–1.10)	0.07	–	–
Donor LAI	1.0 (1.02–1.23)	0.56
MELD-Na score	1.03 (0.96–1.11)	0.37
PVT	2.27 (0.71–7.27)	0.17
Graft weight	0.99 (0.99–1.02)	0.004	–	–
Graft type (left)	1.96 (0.67–5.71)	0.22
GRWR	0.29 (0.3–2.9)	0.29
Blood loss	1.0 (1.01–1.12)	0.68
Duration of surgery	1.0 (0.99–1.00)	0.49
Anhepatic phase	0.99 (0.98–1.00)	0.78
Cold ischemia time	0.99 (0.98–1.01)	0.68
Warm ischemia time	0.97 (0.92–1.01)	0.20
Portal venous flow	0.86 (0.93–1.12)	0.34
Peak AST	1.04 (0.95–1.14)	0.021	–	–
Peak ALT	2.02 (1.08–3.21)	0.012	–	–
Lowest platelet count	1.98 (1.01–2.13)	0.016	–	–
INR > 1.6a	29.6 (3.03–222.5)	<0.001	5.4 (2.1–66.3)	0.04
Total bilirubina > 10 mg/dL	41.6 (9.2–188.5)	<0.001	16.4 (3.2–88.5)	0.001
Ureaa > 100 mg/dL	51.6 (11.3–232.5)	<0.001	16.3 (3.1–80.5)	0.001

EAD = early allograft dysfunction; LAI = liver attenuation index; MELD = model for end-stage liver disease score; PVT = portal vein thrombosis; GRWR = graft-to-recipient weight ratio; INR = international normalized ratio.

^aFor five consecutive days after day 7.

Table 4.

Evolution of New Model for EAD in LDLT Setting.

Definition	Total population (n = 304)	EAD mortality (n = 19)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Total bilirubin > 10 mg/dLa	56 (18.4)	15	78.9	85.6	26.8	98.4
INR > 1.6a	38 (12.5)	16	84.2	92.3	42.1	98.9
Urea > 100 mg/dLa	38 (12.5)	16	84.2	90.9	38.1	98.9
Total bilirubin > 10 mg/dL + INR>1.6a	18 (5.7)	14	73.7	98.6	77.8	98.3
EAD-LDLT: Total bilirubin > 10 mg/dL + INR>1.6 + urea>100 mg/dLa	16(5.1)	14	73.7	99.3	87.5	98.3

EAD = early allograft dysfunction; LDLT = live donor liver transplantation; INR = international normalized ratio; PPV = positive predictive value; NPV = negative predictive value.

^aFor five consecutive days after day 7.

Calibration and Validation of the New EAD-LDLT Model

The calibration of this new EAD-LDLT model was performed by the Hosmer–Lemeshow test. It revealed that our new EAD-LDLT model is a good fit (P = 0.09) (Figure 2). In addition, the internal cross-validation of the new definition was carried out by K-fold cross-validation method. The mean absolute errors of the K-fold (K = 5) random samples were 0.034, 0.035, 0.039, 0.043, and 0.047, respectively; which proved its validity. The new EAD-LDLT model using only 5 days of three post-LTvariables after POD 7 yielded an AUC of 0.86, which is better than that obtained by previous definitions (Figure 3). In addition, the predictive accuracy and discriminative ability measured by AUC were significantly superior to that of Olthoff's model (AUC, 0.86 vs. 0.80; P = 0.04).

Figure 2.

The scatter diagram for new EAD-LDLT model. A plot would indicate a good calibration model showing acceptable agreement between the predicted cases and the actual outcomes. EAD = early allograft dysfunction; LDLT = live donor liver transplantation.

Figure 3.

The receiver operator characteristics (ROC) analysis of all the definitions.

Discussion

To the best of our knowledge, this is the first study comparing the diagnostic accuracies of the commonly described definitions of EAD in the literature in LDLT recipients. The existing definitions of EAD have limitations of being either too broad with low specificity or too restrictive with low sensitivity in recipients of LDLT. The new model of EAD-LDLT, based on total bilirubin >10 mg/dL, INR >1.6, and serum urea >100 mg/dL, for five consecutive days after day 7, has a better predictive value for mortality due to EAD in LDLT recipients.

In LDLT, the graft is obtained from a highly selected healthy donor and implanted with a reasonably shorter CIT. As a result, PNF is rare and more often, transient graft malfunction also called EAD develops because of the inability of a partial graft to meet the metabolic demands of the recipient.^1,2 It typically evolves after the first week which coincides with ischemic strain on the regenerating liver parenchyma. This injury is further compounded by the portal hyperperfusion on a partial graft. It usually manifests as hyperbilirubinemia, coagulopathy, and sometimes ascites with much lesser transaminitis as there is no parenchymal necrosis.^4,13 In contrast to PNF, most of EAD in LDLT are reversible because of the regenerative capacity of partial grafts with full functional recovery. However, EAD has the potential to cause an adverse impact on graft and patient survival. Early recognition of EAD is important to prognosticate the patients to aggressively manage the complications related to it and, importantly, to timely consider for retransplantation.

An ideal definition of EAD should include the simple and objectively measured parameters. As of now, there is no uniformity in defining the criteria for EAD in LDLT across the world. Most of the existing definitions are extrapolated from DDLT. Furthermore, different studies chose different arbitrary cutoff levels.¹⁴ The most commonly used definition for EAD was described by Olthoff et al⁵ in 2010. They reported the incidence of EAD to be 23.2% (69/297) with a PPV of only 18.8%. Thirteen of 69 patients who satisfied EAD criteria died within 6 months. In our analysis, 32.5% (n = 99) of patients satisfied the Olthoff's definition. But, only 17.5% of this cohort had early graft loss leading to mortality. The reason for this low PPV is it being a broad definition, diagnosing EAD based on a single-day (day 7) value satisfying any one of the three parameters. More so, in our analysis, there was only one patient who satisfied the transaminitis component (≥2000 IU/L) of the Olthoff's criteria, which indicates that the degree of transaminitis does not correlate well with the graft dysfunction in LDLT, the reason being a minimal IR injury than in DDLT. This model was further validated in LDLT by Pomposelli et al³ in A2ALL study, involving 613 patients after excluding the transaminitis component. A total of 110 patients (17.9%) developed EAD, of which 24% had graft failure at 90 days, PPV being low. Similar to Olthoff's model, the definition derived by A2ALL study group also have low PPV of 17.9%, when applied to our study cohort. Recently, Okamura et al⁶ validated Olthoff's definition in 260 LDLT patients. About 84 (32.3%) patients full filled the original Olthoff's criteria. About 59 (22.7%) and 46 (17.7%) patients had total bilirubin ≥10 mg/dL and INR ≥1.6 on POD 7, respectively, and 22 (8.5%) subjects fulfilled both criteria. They observed that the new model combining TB of 10 mg/dL or greater and INR of 1.6 or greater on POD 7 was strongly associated with early graft loss (59.1%, RR, 6.97 at 90 days; 68.2%; RR, 5.75 at 180 days). On validating the Okamura's modified Olthoff definition in our study population, we revealed its low sensitivity (26.7%) and PPV (21.1%), despite having high specificity of 96.1%. The major caveat of abovementioned studies is that all causes of mortality (technical, immunological, and non–graft-related) were taken into account while defining EAD. Ideally, EAD-related mortality should exclude all other causes while defining early graft loss primarily due to graft dysfunction.

Ikegami et al⁴ described the concept of delayed functional hyperbilirubinemia (DFH-20). They analyzed 210 patients of LDLT, after excluding thegraft dysfunction secondary to technical, immunological, or recurrent hepatitis-related causes within 1 month of surgery. About 22 (10.4%) had DFH-20 and 59% (13/22) of these patients had graft-related mortality. To date, the model of DFH-20 has never been validated. In the present study, only 6 patients (1.9%) fulfilled the DFH-20 criterion, out of which 4 patients had EAD-related mortality, which shows a high specificity rate (99.6%) and PPV (95%). However, the sensitivity was very low (21.1%). Thus, DFH-20 is too restrictive and many patients with EAD could not be stratified.

In contrast to previous studies in the present study, a homogenous cohort of recipients who underwent LDLT for chronic liver disease was selected for analysis. In addition, most of our recipients had advanced liver disease, the median MELD-Na score being 23 (19–27). Univariate and multivariate analysis revealed total bilirubin, INR, and urea as the significant factors predicting early graft loss. The optimum cutoff values of these parameters to predict the early graft loss were derived. Most of the previous definitions have taken a single cross-sectional value of the studied parameters on day 7 after liver transplantation. But, as the EAD evolves dynamically, a continuous evaluation of the parameters for 5 days (after 7th postoperative day) is a true reflector of the graft dysfunction. Moreover, hyperbilirubinemia and coagulopathy in the immediate postoperative period (first 7 days) may be a reflection of the preoperative status of the recipient and perioperative insults. Thus, to account for the dynamicity of the graft function and exclude the instances of recovery (either spontaneous or after management of any treatable causes) of the dysfunction, continuous evaluations of these significant parameters for 5 consecutive days were taken into account. Furthermore, we studied the individual ability of these 3 significant variables in predicting the EAD-related mortality and then compared it with combined capability for the same. We found that when all of these 3 factors were combined, it predicted the early graft loss in the best possible way. It is a much more accurate definition which is neither too accommodative like that of Olthoff nor too narrow as Ikegami. As both the sensitivity and PPV are high (73.7% and 87.5%, respectively), it can predict the early graft loss due to EAD and also the need of retransplantation much more precisely than all the existing definitions.

The bilirubin and INR have been known as prominent predictors of early graft loss and mortality after LDLT.^6,15 In the present analysis, apart from bilirubin and INR, we observed that the increased serum urea level (>100 mg/dL) after POD7 for the five consecutive days was independently associated with early graft loss both in univariate and multivariate analysis. Conventionally, urea, as a common biochemical metabolite, reflects the kidney function but its level also rises in many instances of liver dysfunction.¹⁶ Urea synthesis takes place exclusively in the liver. Because of the high metabolic needs of the recipient, the functional incapacitation of the regenerating partial graft occurs in EAD. Furthermore, owing to the high levels of circulating stress hormones such as glucocorticoids and glucagon on the background of the graft dysfunction, and also owing to the accumulation of by-products of the urea cycle, urea synthesis gets overstimulated in the graft.¹⁷ As a result, very high levels of serum urea suggest graft dysfunction rather than functional recovery.

Previously, the EAD after LDLT has been studied mainly from the viewpoint of small for size syndrome (SFSS).¹³ The concept of SFSS is based on GRWR. However, graft dysfunction in the partial grafts is multifactorial such as recipient's underlying disease severity, the extent of portal hypertension, donor age, liver steatosis, absolute graft weight, ischemic injury, technical factors, mechanical and functional outflow obstruction¹⁸ rather than just GRWR and portal inflow. As the surgical techniques and the perioperative management strategies are evolving, many centers have published good results in recipients receiving the grafts with GRWR <0.8.^4,19, 20, 21 With this, the suspicion of SFSS causing EAD is fading due to a better understanding of its pathophysiology. Our study seconds this revelation. Since the beginning of our transplant program, our technique of outflow reconstruction has evolved. We have reported an excellent 28 days patency rate of 90% of the reconstructed neo-MHV.⁹ As a result, in our study cohort; both the low GRWR and the portal venous flow did not have a significant impact on EAD.

In our study, donor age was found to be the significant risk factors for early graft loss after extrapolating the deduced definition on our study cohort. The donor age was also strongly associated with EAD after LDLT in the studies by Ikegami et al⁴ and Pompeselli et al.³ Although, in the present era of extended criteria donors, the acceptable donor age is increasing, the effect of age on graft quality cannot be denied. The chances of encountering steatosis and subclinical inflammation in the grafts from the older donors increase with age. As a result, grafts from the older subjects are less resilient. In addition, these grafts are more prone to the IR injury and also the shear stress injury caused by portal hyperperfusion, leading to increased probability of having EAD in the early post-transplant period. Furthermore, our study has brought attention to the absolute graft weight as an important factor in causing graft loss due to EAD. However, GRWR did not turn out to be significant. This disparity can be accounted by recipient body weight being affected by conditions such as ascites, tissue edema, etc.

Recipient's metabolic demand as reflected by MELD score has been shown to be an important determinant of the outcomes after LDLT. Ikegami et al⁴ reported a MELD score >15 as a risk factor for DFH. Similarly, in the A2ALL study experience,³ recipients with EAD had a higher mean preoperative MELD score (17.3 vs. 15.1, P < 0.001). The overall morbidity and mortality in our present study were comparable with the published literature, in spite of having a high median MELD-Na score of 23 (19–27). Similarly, early graft function was not affected by a high MELD score in the series by Yoshizumi et al.²²

Limitation of this study is single-center experience. Although an optimal definition has been derived by us, which is also internally validated, it needs external validation from other centers. In this study, patients with mortality due to potential graft dysfunction were considered as patients with EAD. Patients with EAD who might have recovered from the dysfunction would be fallaciously included in a group without EAD. As stated previously, sepsis could not be excluded as a primary cause of graft loss and mortality because sepsis can be both a cause and resultant of EAD, although known causes of sepsis such as vascular and biliary complications and infected intra-abdominal collections were excluded. In addition, we could not rule out the immunological insult causing the graft loss as the histological support to establish or refute the same is lacking.

Despite the existence of multiple definitions for EAD, an optimal definition in LDLT is lacking. Our efforts in comparing existing definitions and giving a balanced version materialized into a new EAD-LDLT model based on serum total bilirubin, INR, and serum urea. Total bilirubin >10 mg/dL, INR >1.6, and serum urea >100 mg/dL, for five consecutive days after day 7 had better sensitivity, specificity, PPV, and NPV than the previous calculators. It has an AUC of 0.86, predicting the EAD-related mortality better than the existing definitions. The calibration and the internal cross-validation of the new model confirmed its predictability. GRWR and MELD did not turn out to be of significance in causing graft loss due to EAD. However, the donor age and the absolute graft weight predicted the mortality related to EAD.

CREDIT AUTHORSHIP CONTRIBUTION STATEMENT

Study concept: Viniyendra Pamecha; Study design: Viniyendra Pamecha, Bramhadatta Pattnaik, Piyush Kumar Sinha; Data collection: Bramhadatta Pattnaik, Piyush Kumar Sinha, Shridhar Vasantrao Sasturkar, Nihar Mohapatra, Ashok Choudhury; Analysis and interpretation of data: Guresh Kumar, Viniyendra Pamecha, Bramhadatta Pattnaik, Piyush Kumar Sinha; Manuscript drafting: Bramhatta Pattnaik, Nilesh Patil, Viniyendra Pamecha; Critical revision of the manuscript for important intellectual content: Viniyendra Pamecha, Nilesh Patil, Shiv Kumar Sarin.

Conflicts of interest

The authors have none to declare.

Ethical statement

• 1.The present study is compliant with the ethical standards.
• 2.There is no source of funding.
• 3.There is no conflict of interest.
• 4.The study is approved from the institute ethics committee (no.: IEC/2017/48/MA10).

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Figure S1.

Impact of GRWR (A) and graft type (B) on the mortality due to early graft dysfunction (EGD).