EXPERIMENTAL AND CLINICAL TRANSPLANTATION

Objectives: For liver transplant candidates, trans-plantation or death on wait lists can be competing risks. This study used competing risk analysis to estimate the probability of death among patients on wait lists for liver transplant. Materials and Methods: We retrospectively analyzed liver transplant candidates registered at the Mon-taseriyeh Transplant Center Registry of Mashhad University of Medical Sciences (Iran) from 2013 to 2024. We followed patients from listing through transplant, death, or end of study. We collected demographic, clinical, laboratory, and follow-up details. We used Gray’s test to assess cumulative incidence of death across different listing periods, orthotopic liver transplant groups, and Model for End-Stage Liver Disease severity levels. To estimate the effects of various covariates while accounting for transplantation as a competing event, we conducted a competing risk regression using the Fine and Gray subdistribution hazard model. We used R software (cmprsk and cuminc packages) for statistical analyses. Results: Average age of patients was 50.84 ± 13.78 years. Over the follow-up period, 503 patients (60.0%) received transplants, 233 (27.8%) died while waiting for transplant, and 102 (12.2%) were administratively censored. Among transplant patients, 65.9% had Model for End-Stage Liver Disease scores between 10 and 20, with mortality increasing with increased scores. The hazard model showed no significant differences in death risk by age, sex, marital status, year of transplant, or etiology group. However, patients with higher Model for End-Stage Liver Disease scores had significantly greater risk of death than those with lower scores (P < .001). Conclusions: Increased Model for End-Stage Liver Disease score emerged as the most significant pre-dictor of mortality among patients waiting for liver transplant. Focusing on candidates with high scores and tackling socioeconomic barriers could improve survival outcomes. These insights can inform future approaches to optimize patient prioritization and transplant allocation.

Key words : Model for End-Stage Liver Disease, Orthotopic liver transplantation

Orthotopic liver transplantation (OLT) remains the gold standard of therapy for patients with end-stage liver disease; however, increased demand for transplants has exceeded the static supply of donor organs from deceased donors, leading to longer waiting lists, longer wait times, and death.1 The effect of waiting 12 months versus 2 months corresponded with a drop in overall survival rate of 5.07% and 8.33% at 5- and 10-years posttransplant, respectively.² In an effort to bridge the gap, the United Network for Organ Sharing (UNOS) introduced the Model for End-Stage Liver Disease (MELD) score in 2002 to optimize organ allocation.^3,4 This system predicts disease severity and mortality risk using serum creatinine, bilirubin, and INR. The original equation was as follows: MELD = 3.78 × ln[serum bilirubin (mg/dL)] + 11.2 × ln[INR] + 9.57 × ln[serum Cr (mg/dL)] + 6.43.⁵
Silberhumer and colleagues have shown that mortality on the waiting list increases linearly with higher MELD scores at listing. Patients with a MELD score below 11 had a 0.0% mortality rate, whereas those with a score above 24 faced a 61.9% mortality rate, regardless of whether they had compensated or decompensated cirrhosis. The study confirmed that the MELD score accurately predicts 3-month mor-tality in patients with chronic liver disease awaiting transplant and could be helpful for donor liver allocation.⁶
The MELD-based liver graft allocation policy has resulted in fewer new registrations and deaths on wait lists, reduced wait times, and increased the number of transplant procedures, all while maintaining stable overall graft and patient survival rates after transplant.
In a study from Merion and colleagues, recipients of deceased donor transplants experienced a 79% reduction in mortality risk compared with candi-dates on wait lists. Specifically, at a MELD score of 18 to 20, recipients had a 38% lower risk of death than candidates. In addition, at the 1-year follow-up, patients who were at lower risk of dying before the transplant did not gain a significant survival advan-tage from undergoing a liver transplant.⁷
A patient waiting for a liver transplant has 2 potential outcomes: getting the transplant or dying. These are known as competing risks because expe-riencing one outcome either prevents or affects the likelihood of the other.^8,9 Moreover, Kim and colleagues demonstrated that the Kaplan-Meier met-hod tends to overestimate the risk of death during the waiting period, whereas competing risk analysis estimated a 3-year mortality of 10%.¹⁰ Gray (1988) developed a modified χ² test to compare cumulative incidence curves across 2 groups, specifically asses-sing differences in cumulative incidence rates. Unlike the log-rank test that overlooks competing events, Gray’s method considers both the primary and competing events.¹¹
Of note, 1 minus the Kaplan-Meier estimate can be greater than the estimated cumulative incidence function (CIF). The log-rank test, based on cause-specific hazards, can yield significantly different results than Gray’s test, which is based on the subdistribution hazard (subhazard).^12,13 The CIF for a specific cause “k” depends not only on the hazard of cause k but also on the hazards of all other causes, and it is derived indirectly from the model parameters.^14,15 The Fine-Gray model assesses how covariates inf-luence the CIF through subhazard ratios and directly relates the CIF to the subhazard function.¹⁶
In this study, we used competing risk analysis to identify factors linked to the cumulative incidence of mortality among patients with liver disease waiting for a transplant.

Materials and Methods

Participants and study design
This historical cohort study examined liver transp-lant candidates registered at the Montaseriyeh Transplant Center, Mashhad University of Medical Sciences, Mashhad, Iran, from 2013 to 2024. The study focused on adult patients with end-stage chronic liver disease, excluding pediatric candidates under aged 17 years and those with acute liver conditions. Patients were tracked from the time of registration until they received a transplant, died, or the study concluded. This research received approval from the Ethics Committee of MUMS under the Code of Ethics IR.MUMS.FHMPM.REC.1402.187.
Data and sample size determination
We collected data from 925 patients on the liver transplant waiting list at Montaseriyeh Hospital in Mashhad, Iran, from 2013 through 2024. After excluding 40 patients under 17 years old, 5 with acute liver disease, and 42 with incomplete records, we had a final sample size of 838 patients.
We divided the study period into 4 eras: 2013-2015, 2016-2018, 2019-2021, and 2022-2024. Etiologies included hepatitis B virus (HBV)-related or hepatitis C virus (HCV)-related hepatocellular carcinoma (HCC), cryptogenic cirrhosis, autoimmune hepatitis, primary sclerosing cholangitis, Wilson disease, and other. Disease severity was assessed using MELD scores.
This study was approved by the Ethics Com-mittee of Mashhad University of Medical Sciences (IR.MUMS.FHMPM.REC.1402.187), and all partici-pants gave written informed consent.
Statistical analyses
We presented continuous variables as mean ± SD for data that followed a normal distribution and as median with interquartile range (IQR) for data that did not. We presented categorical variables as frequencies and percentages. We used the χ² test to assess associations between categorical variables and study eras. We compared continuous variables across eras with the Kruskal-Wallis test. We used competing risk methodology to account for events that prevent the primary outcome from occurring. We assessed group differences in the cumulative incidence of the event of interest by using Gray’s test, which directly compares cumulative incidence functions across groups. To further examine the impact of covariates, including sex, age, calendar year of listing, underlying cause of liver disease, and baseline disease severity, on the risk of wait list mortality while considering competing events, we used the Fine-Gray proportional subdistribution hazards model.12 This model assumes proportional subhazards and generates subhazard ratios that indicate how covariates affect the risk.17,18 We conducted the analysis by using R software with the cmprsk and cuminc packages.

The study cohort consisted of 838 patients with a mean age of 50.84 ± 13.78 years. Over a median follow-up of 6.7 months, 503 patients (60.0%) underwent liver transplant, 233 (27.8%) died while on the waiting list, and 102 (12.2%) were administratively censored. Table 1 lists the baseline demographic and clinical features of these patients, categorized into 4 eras from 2013 to 2024. The cohorts were largely comparable across different eras regarding characteristics such as age, competing risks, marital status, etiology, labo-ratory values, and MELD scores (P > .05). The only characteristic showing a significant difference between eras was sex distribution (P = .025), with increased rate of male patients from 56.8% in the earliest period (2013-2015) to 70.8% in 2016-2018. Patients with MELD scores above 20 were more common than those in other categories throughout all periods. The distribution of liver disease causes remained stable over time, with viral and crypto-genic causes being the most frequent in each era (P = .14) (Table 1).
Figure 1 shows how patient outcomes varied across different MELD score groups. Most patients had MELD scores between 10 and 20, with 65.9% receiving transplants, 21.3% dying, and 12.9% censored. Patients with MELD scores >30 expe-rienced a significantly higher mortality rate than those with lower scores (P < .001) (Figure 1).
After patients were added to the transplant list, the likelihood of receiving a transplant and the risk of death began to diverge (Figure 2). Over time, the chance of transplant steadily increased, whereas the risk of death grew more slowly, particularly after the first few months. By 60 months, approximately 60% of patients had received a transplant, surpassing the 29% who died, and this gap continued to widen over time (Figure 2).
Figure 3 compares the cumulative incidence of death by MELD diagnosis groups. The cumulative incidence of death increased progressively with increasing MELD score (Gray test P < .001). Patients with a MELD score >30 had the highest cumulative incidence of death, reaching 57% at 40 months, compared with 42% in the 21-30 score group, 22% in the 10-20 score group, and 12% in the <10 score group (Figure 3).
Figure 4 illustrates the cumulative incidence of transplant across various MELD diagnosis groups. The incidence showed significant variation depending on baseline MELD scores (Gray test P = .004). The highest transplant rates were observed in the intermediate MELD score groups (10-20 and 21-30), with 75% at 100 days. In contrast, the rates were substantially lower in both the low MELD score (<10) and very high MELD score (>30) groups (Figure 4).
Patients with only an elementary education consistently exhibited the highest cumulative risk of death throughout the study period (Figure 5). By 80 months, the probability of death was approximately 47% in the elementary group compared with 32% in the diploma group and 15% in the university group (Gray test P < .001).
The cumulative incidence of transplant was lowest among patients with elementary education and highest among those with university education (Figure 6). By 80 months, about 81% of university-educated patients had received transplants compared with 61% of diploma holders and just 50% of elementary-educated individuals (Gray test P < .001). No significant differences in wait list mortality or transplantation rates were shown across various demographic and clinical factors, such as age (divided at 50 years), sex, treatment period, and underlying disease cause (Gray test P > 0.05 for all factors).
The Fine-Gray subdistribution hazard model identified MELD score and educational level as significant predictors of mortality, considering transplant receipt as a competing risk (Table 2). A higher baseline MELD score was strongly associated with an increased death risk. Compared to the reference group (MELD score <10), patients with MELD scores of 20 to 30 had more than three times the risk (subdistribution hazard ratio = 3.15; 95% CI, 1.63–6.11; P < .001), whereas scores >30 carried over 5 times the risk (subdistribution hazard ratio = 5.27; 95% CI, 2.45–11.35; P < .001). A clear dose–response pattern showed risk increasing with MELD severity. Conversely, higher educational levels appeared protective. Patients with a diploma (subdistribution hazard ratio = 0.71; 95% CI, 0.52-0.98; P = .037) or a university degree (sub-distribution hazard ratio = 0.51; 95% CI, 0.29-0.88; P = .017) had a significantly lower risk of death com-pared with those with only elemen-tary education.
Other variables, including age group, sex, marital status, era of inclusion, and cause of underlying liver disease, showed no significant association with the cumulative incidence of the outcome in this model (P > .05 for all variables).

Liver transplantation is the last line of treatment in patients with severe liver disorders and concerns all patients with end-stage liver disease when other medical therapies have failed.¹⁹ Mortality rates among patients waiting for a liver transplant are a key indicator of how well deceased donor organ allocation and distribution systems work.¹⁰ Analysis is needed on the subsequent outcomes to account for competing risks of transplantation and death.¹²
Our study included 293 women (35%) and 545 men (65%) among patients on wait lists. Of these, 163 women (32.4%) and 340 men (67.6%) received transplants. Furthermore, 87 women (37.3%) and 146 men (62.7%) died. Censoring occurred for 43 women (42.2%) and 59 men (57.8%). Our study noted no significant link between sex and the competing risk groups.
Several studies have shown that women account for only 40% of patients on wait lists, probably as the result of lower occurrence of viral-related liver disease and HCC in this population.^20,21 A recent study showed that women with cirrhosis had lower transplant rates despite showing no difference in their all-cause or liver-related mortality compared with men.²² Nephew and Serper recognized the presence of sex-based disparities in wait list mortality and allocation of liver transplants.²³ In the study from Pose and colleagues, women were more likely to be delisted following improvement in patients with alcohol-related decompensated cirrhosis.²⁴
Our study showed that the cumulative mortality rate rose among patients on the wait list, consistent with the survey by Bagheri and colleagues.²⁵
Our research showed that viral infections, such as HBV, HCV, and HCC, along with cryptogenic causes, are primary etiologies. In a study from Saeedi and colleagues, hepatitis B accounted for 23% and cryptogenic cirrhosis for 22.6% of cases, making these the main causes of cirrhosis. The use of the Fine-Gray model revealed that age, serum albumin levels, and the presence of encephalopathy are significant predictors of the cumulative incidence of mortality.²⁶
In our study, the Fine-Gray model showed no significant differences in the cumulative incidence of death across sex, etiology group, or time period. However, education level and MELD score were signi-ficantly associated with the subdistribution hazard of mortality. In Bittermann and colleagues, low health literacy was independently linked to physical frailty and the likelihood of not being placed on the wait list.²⁷ Gruttadauria and colleagues showed a higher survival rate among patients with a higher education level.²⁸ Yoo and Thuluvath showed that individuals with a high school diploma had lower survival rates than those with a graduate degree.²⁹ A prospective cohort study at Shiraz Transplant Center used machine learning models to identify important predictors of mortality, such as posttransplant aspar-tate aminotransferase, creatinine levels, recipient age, posttransplant ALT levels, and tacrolimus use.³⁰
The optimal allocation system should aim to reduce wait list mortality and maximize posttransplant benefits while maintaining transparency, objectivity, and reproducibility, all the while ensuring equal access for all patients. Because these goals can conflict, policies need regular review and adjustment as the patient population evolves. The MELD score, which predicts short-term prognosis, has become a common tool for prioritizing liver transplant candidates, leading to a notable reduction in wait list mortality and removal from wait lists.³¹ Among our patients, 66 patients (7.9%) had MELD scores below 10, 536 (64.0%) had scores that ranged between 10 and 20, 200 (23.9%) had scores between 20 and 30, and 36 (4.3%) had scores of 30 or higher. Mortality risk rose steadily with increasing MELD scores, reaching 61.1% in the group with scores of 30 or above. Our study also showed that transplant rates were 65.9% for patients with MELD scores from 10 to 20 and 54% for those from 20 to 30. Moreover, Chaib and colleagues indicated that a MELD score of 26 correla-ted with the best outcomes for patients on the liver transplant wait list.³²
Most European nations have adopted MELD as their main system for allocating liver transplants. Salvalaggio and colleagues showed that, following MELD’s implementation in São Paulo, there was a decline in both wait times and mortality rates among listed patients.³³ Those listed after MELD’s intro-duction faced a notably lower risk of death while waiting. Nonetheless, posttransplant results remained consistent. In general, MELD proves to be an effective and practical tool for liver transplant allocation, particularly in developing countries.

Our study demonstrated that the most accurate way to evaluate mortality risk for liver transplant candidates is through competing risk analysis, which accounts for both death and transplantation simul-taneously. The traditional Kaplan-Meier approach estimates survival by assuming transplant never occurs, resulting in an overestimation of death risk among patients on wait lists. In addition, our study indicated a strong association between higher edu-cation levels and reduced mortality risk, suggesting that better-educated patients may have improved health literacy, easier access to health care, or better adherence to medical guidance, all of which can enhance survival prospects. Future studies should focus on designing patient-specific interventions for those with lower educational backgrounds to help close the outcome gap on the wait list.
Most notably, with high MELD scores linked to increased mortality risk, the prognostic importance of this scoring system has been emphasized. Thus, early intervention for high-risk patients and use of MELD scores are needed to guide risk stratification in clinical decision-making.

Volume: 24 Issue: 3 March 2026

FULL TEXT