• Allocation;
  • hepatocellular carcinoma;
  • liver transplantation;
  • MELD score


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgement
  7. References

Patients with hepatocellular carcinoma (HCC) within Milan criteria receive priority on the liver transplant waiting list (WL) and compete with non-HCC patients. Dropout from the WL is an indirect measure of transplant access. Competing risks (CR) evaluation of dropout for HCC and non-HCC patients has not previously been reported. Patients listed between 16 March 2005 and 30 June 2008 were included. Probability of dropout was estimated using a CR technique as well as a Cox model for time to dropout. Overall, non-HCC patients had a higher dropout rate from the WL than HCC patients (p < 0.0001). This was reproducible throughout all regions. In Cox regression, tumor size, model for end-stage liver disease (MELD) score and alpha fetoprotein (AFP) were associated with increased dropout risk. Multivariable analysis with CR showed that MELD score and AFP, were most influential in predicting dropout for HCC patients. The index of concordance for predicting dropout with the CR was 0.70. HCC patients appear to be advantaged in the current allocation scheme based on lower dropout rates without regard to geography. A continuous score incorporating MELD, AFP and tumor size may help to prioritize HCC patients to better equate dropout rates with non-HCC patients and equalize access.

Implementation of the Model for End-stage Liver Disease (MELD) scoring system for the prioritization of patients for liver transplantation (LT) has been widely viewed as successful. This continuous scoring system is most accurate when the liver failure itself is likely to cause death in the near future. However, there are a number of medical conditions for which mortality from intrinsic liver disease is relatively low, but other disease symptoms or risks pose a more legitimate need for transplant than the MELD-calculated mortality risk. Policymakers foresaw the need to include some mechanism to adequately assign priority for these conditions where the liver disease is not necessarily life threatening, but the need for liver transplant deserves some priority nonetheless (1). The most common of these conditions is hepatocellular carcinoma (HCC), where the threat to mortality from cancer is greater than that of near-term liver failure. The original MELD plan included a systematic approach for allowing patients with HCC meeting Milan Criteria (MC) (2) extra priority in the allocation scheme. Originally, patients meeting MC (single tumors ≥2 cm and ≤5 cm in size or up to 3 tumors the largest of which is ≤3 cm; T2 lesions) were assigned a MELD score of 29 in an attempt to estimate the risk of HCC progression beyond MC. Since patients with more advanced HCC (beyond MC) do not receive additional priority under this system, the risk of advancing beyond MC can be considered the risk of ‘dropping off’ the list. After initial experience with the new plan, it became evident that the risk for HCC patients of dropping off the list due to HCC progression was much less than originally estimated. Single center reports indicated that the dropout rate for HCC patients was 11.0%, 57.4% and 68.7% at 6, 12 and 18 months for patients meeting MC (3). Subsequent policy adjustments have reduced the MELD score assigned to HCC patients meeting the MC to 24 in April of 2003 and 22 in January of 2005. The initial HCC priority score of 22 is meant to equate the drop out risk for HCC patients to the risk of dying from progressive liver disease defined by MELD score of 22. These adjustments have all been aimed at maintaining equity in access to deceased donor livers for both HCC and non-HCC liver transplant patients while maintaining respectable outcomes for all recipients. In the only other report examining access to LT for HCC compared with non-HCC patients, Freeman et al. (4), using Cox models, found that dropout rates were similar for the two groups 3 months after listing. However, this report utilized data from the very early stages of the MELD policy when HCC patients received the higher MELD exception priority and it did not account for the competing risks (CR) faced by waiting patients (i.e. transplant, continued waiting or drop out). Since Cox estimation may overestimate dropout rates (in particular, at later time points) by censoring for transplantation rather than including them in a CR assessment, a reexamination of dropout risks using more recent data is likely to be more reflective of the actual waiting list (WL) dynamics and more helpful in designing future policy. Moreover, there are no studies with large enough sample sizes to adequately assess potential predictive factors such as tumor characteristics, locoregional ablative treatments (AT) or other demographics for dropout for HCC patients.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgement
  7. References

Study population

All analyses were based on Organ Procurement and Transplant Network (OPTN) data as of 16 January 2010. Adult (age ≥18) WL registrations for which an initial approved HCC exception application was submitted between 16 March 2005 and 30 June 2008 (dates encompassing most recent HCC policy change) were classified as being in the HCC group. All other adult registrations (non-Status 1) without a diagnosis of cancer and without an approved non-HCC exception occurring in the same time period were assigned to the non-HCC group. In addition, registrations removed for any of the following reasons were excluded: living donor transplant, transferred to another center, transplanted at another center and removed in error.


For the purposes of this analysis, a WL ‘dropout’ for an HCC registration occurred for any of the following removals: candidate removed by an OPTN member for death; candidate not transplanted but identified as having died in the Social Security Death Master File; candidate removed for ‘too sick’; and candidate removed for ‘other’ reasons that were determined to be related to the diagnosis of HCC. A WL dropout for a non-HCC registration was defined similarly, except that removals for ‘other’ reasons were grouped together and not defined as only those related to liver disease. For HCC registrations, time to dropout was measured from the date of the initial HCC application to removal. For non-HCC registrations, time to dropout was measured from the date of initial listing to removal. Registrations still waiting were censored as of the date of the analysis. Factors associated with dropout for HCC patients were based on data collected on the initial HCC exception application.

Statistical analysis

Univariate statistical analysis of dropout rates was based on CR. Probabilities were estimated using the method of Kalbfleisch and Prentice (5) and variance estimates of the probabilities were derived using the method of Aalen (6). Multivariable analyses of dropout rates were based on the Cox proportional hazards model (non-CR) as well as the CR method of Fine and Gray (7) The CR methods allows for all patients to be placed into a category; transplanted, dropped out or still waiting. In the Cox analysis, patients are censored for any removal other than death. All statistical calculations were performed using SAS (version 9.2, Cary, NC) or R for Windows (Version 2.5.1).


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgement
  7. References

Between 16 March 2005 and 630 June 2008 there were 4197 initial applications approved for HCC exception priority. The demographics of this population are displayed in Table 1. All of these applications were for T2 lesions. Of this total, there were 575 (13.7%) dropouts. There were 1783 (42.5%) initial applications reporting some form of AT. Trans arterial chemoembolization (TACE) was reported for 1116 (62.6%) of applications citing ablative treatment; radiofrequency ablation (RFA) on 477 (26.8%) applications; and other or a combination of TACE/RFA on 190 (10.6%) applications. The number of tumors reported on the initial applications was 1 n = 2893 (69%); 2 n = 925 (22.0%); 3 n = 379 (9.0%). Median tumor size was 2.6 cm (diameter of the largest lesion reported on the initial HCC application), mean 2.75 cm and a range of 0.1–5 cm (1.5–4.6 cm:5th–95th percentile). Median alpha fetoprotein (AFP) was 14 ng/mL, mean 310 ng/mL and range 1.0 to 58 820 ng/mL. For the same time period, there were 18 068 patients added to the WL with no HCC exception application. Of these, 5385 had a MELD score >21 (the range of MELD scores at which HCC patients are listed) and 12 683 had a MELD score <21.

Table 1.  Demographics of patients listed for OLT
Factors at listing/initial HCC applicationHCCNon-HCC
  1. 1Indicates p < 0.001 for testing if the distribution of the factor is the same across the 2 groups.

  2. 2Patients with an initial status of inactive excluded.

 White276365.812 99972.0
 Male325677.611 42163.0
 Region 12085.06563.6
 Region 252312.5243813.5
 Region 350912.1218112.1
 Region 44149.9187710.4
 Region 582219.6343419.0
 Region 61583.85232.9
 Region 73408.115378.5
 Region 82536.010535.8
 Region 942510.117739.8
 Region 102947.012697.0
 Region 112516.013277.3
Total4197100.018 068100.0
Mean age in years1419756.418 06852.6
Mean Lab MELD1419711.918 06818.1
Mean Match MELD1419722.217 87718.0


Patients without HCC consistently had a higher dropout rate from the waitlist compared to HCC patients at every time point evaluated after listing (Table 2). The use of any type of ablation had a minimal effect on waitlist dropout decreasing this from 10.1% for no ablation to 8.0% for ablation at 180 days after listing. Intrinsic MELD score in addition to HCC had a moderate impact on dropout increasing from 4.4% for MELD <8 + HCC, to 14.5% for MELD >14 + HCC at 180 days after listing. Likewise, tumor size had a modest impact on dropout increasing from 7.8% for those 2.0–2.6 cm to 10.1% for those >3.3 cm at 180 days after listing. AFP had a dramatic impact on dropout increasing from 7.4% for AFP <500 to 24.9% for AFP >1000 after listing.

Table 2.  % Dropout from the waitlist at 30, 60, 90, 180 and 365 days
Days after listing306090180365
Non-HCC All MELD (number at risk) 6.0 (13 863) 8.4 (12 454)10.2 (11 571)13.6 (9958) 17.7 (8228)
All HCC exceptions (number at risk)1.8 (3070)3.6 (2421)5.1 (1976)8.6 (974)11.5 (361)
No Ablation2.44.36.310.1 11.8
HCC + MELD < 7.2
HCC + MELD 8– 8.0
HCC + MELD 11– 9.3
HCC + MELD > 18.4
Tumor <2 cm0.
 2–2.6 cm1.
 2.6–3.3 cm1.
 >3.3 cm2.64.76.310.1 13.4
AFP <5001.
 500–10002. 19.9
 >10005.610.7 14.7 24.9 29.9

Waitlist dropout at 12 months varied by region increasing from 6.8% in region 11 to 17.2% in region 5 for HCC patients (Figure 1). Dropout for non-HCC patients increased from 14.3% in region 8 to 24.5% in region 1. Nationally, dropout for HCC patients at 12 months was 11.5%, for non-HCC (MELD <21) 13.3% and for all non-HCC 17.7%. In all 11 regions, dropout for non-HCC patients was consistently higher than that of HCC patients.


Figure 1. % Dropout within 12 months: HCC and Non-HCC Patients by Region.

Download figure to PowerPoint

Multivariable waitlist analysis

A Cox regression model of time to dropout and CR model for HCC patients were developed and compared to analyze the waitlist. Factors considered in the model included tumor stage, number of tumors, maximum tumor size, ablation type, MELD score, log AFP level and year of listing. Table 3 shows the CR and Cox proportional hazard models hazard ratios (HR) for the effect of tumor size on dropout. Increasing tumor size results in an increased HR for dropout for both models. The Cox model consistently results in a larger HR than the CR model. Table 4 shows the results for the two models for the impact of recipient MELD on dropout for HCC patients with the Cox model producing a consistently higher HR for each MELD score. Table 5 shows the HR for the affect of AFP on dropout. Again, the Cox model produces a consistently larger HR than the CR model. Multivariable analysis results for the Cox and CR models are shown in Table 6. In the Cox regression model, MELD, AFP and tumor were predictive of dropout, and in the CR regression model, MELD and AFP were predictive of HCC dropout. The index of concordance for the ability of the Cox model to predict dropout using MELD alone is 0.68; and for Cox model (log (AFP), max tumor size, MELD, 3 tumors) = 0.70; and for the CR model (log (AFP), max tumor size, MELD) = 0.70.

Table 3.  Hazard ratio and 95% confidence interval (CI) for impact of maximum tumor size on dropout
Tumor size (cm)CoxCR
  1. Cox = Cox proportional hazard method; CR = Competing risk method.

11.08 (1.03,1.13)1.04 (0.99,1.10)
1.51.16 (1.06,1.27)1.09 (0.98,1.20)
21.25 (1.09,1.43)1.13 (0.98,1.31)
2.51.34 (1.12,1.61)1.18 (0.97,1.43)
31.44 (1.15,1.81)1.23 (0.96,1.57)
3.51.55 (1.19,2.04)1.28 (0.96,1.71)
Table 4.  Hazard ratio and 95% confidence interval (CI) for impact of MELD score on dropout for HCC patients
  1. Cox = Cox proportional hazard method; CR = Competing risk method.

 71.14 (1.12,1.16)1.09 (1.07,1.11)
 81.29 (1.25,1.33)1.19 (1.15,1.23)
 91.46 (1.39,1.54)1.30 (1.23,1.37)
101.66 (1.55,1.78)1.42 (1.32,1.52)
111.89 (1.73,2.06)1.54 (1.41,1.69)
122.14 (1.93,2.38)1.68 (1.51,1.87)
132.43 (2.16,2.74)1.84 (1.62,2.08)
142.76 (2.41,3.17)2.00 (1.74,2.31)
153.14 (2.69, 3.66)2.18 (1.86,2.56)
163.57 (3.00, 4.23)2.38 (1.99,2.84)
Table 5.  Hazard ratio and 95% confidence interval (CI) for impact of AFP on dropout
Log AFP (ng/mL)CoxCR
  1. Cox = Cox proportional hazard method; CR = Competing risk method.

11.01 (0.91, 1.14)0.95 (0.85,1.06)
21.08 (0.88, 1.32)0.95 (0.78,1.16)
31.19 (0.91, 1.56)1.00 (0.77,1.29)
41.39 (1.02, 1.89)1.10 (0.81,1.49)
51.68 (1.21, 2.34)1.28 (0.92,1.77)
62.13 (1.52,2.98)1.56 (1.12,2.18)
72.83 (2.04, 3.92)2.00 (1.44,2.79)
Table 6.  Multivariable analysis of time to dropout using Cox and CR methods with 95% CI
FactorCoxCompeting risks
  1. Cox = Cox proportional hazard method; CR = Competing risk method.

MELD0.1271.14 (1.12, 1.16)0.0871.09 (1.07, 1.11)
Log AFP−0.007 0.99 (0.88, 1.13)−0.077 0.93 (0.82, 1.05)
Log AFP20.0221.02 (1.01, 1.04)0.0251.03 (1.01, 1.04)
Max size0.1471.16 (1.06, 1.27)0.0821.09 (0.98, 1.20)
3 Tumors (vs. 1 or 2)0.4341.54 (1.16, 2.05)0.2431.27 (0.96, 1.69)


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgement
  7. References

This study attempts to look at equity of access to liver transplant by defining the access endpoint as dropout (i.e. removal) from the WL for HCC or non-HCC patients. Reasons for dropout are frequently due to death or progression of disease severity that might translate into an unacceptable outcome. Parity in the system would imply that this dropout rate would be similar across different patient groups. Results from this study show that the dropout rate for HCC patients is consistently lower than that of non-HCC patients. This finding is striking considering that multiple downward adjustments have already been made to HCC prioritization. In addition, this advantage occurs in every region, albeit to varying degrees, indicating that HCC patients continue to have enhanced access to transplantation relative to the non-HCC patients in throughout the United States.

Patients on the waitlist for LT can have one of three different outcomes: transplant, death or removal. Each of these outcomes is a CR with the others as the occurrence of one outcome precludes or impairs the others. In the current analysis, we use the CR statistical method and compare results with those from the Cox method to evaluate the outcome of HCC and non-HCC patients on the LT waitlist and more specifically the removal or dropout of HCC patients. This technique (CR) has been used previously to evaluate deaths on the waitlist at a single institution (8). In this study, the authors clearly described that censoring patients at the time of transplant for Kaplan–Meier or Cox outcomes results in a reduction in the size of the at risk population and tends to overestimate events rates. In the only previous study to look at waitlist removals for malignant and chronic disease patients, the Cox method was used (9) and consequently the rates for removal for the chronic disease and HCC patients may have been overestimated. For this reason, we compared the CR and Cox methods using OPTN data to evaluate waitlist removals. Consistent with other reports comparing Cox and CR for removals for liver disease patients (8) and for heart and renal transplant patients (10,11) our Cox analysis was associated with a larger hazard ratio than the CR method. Thus, the Cox method would appear to overestimate the risk of waitlist dropout for HCC patients compared with the CR method. However, the Cox and CR methods demonstrate that MELD score, tumor size and AFP are the most important variables, regardless of the statistical method used, influencing waitlist dropout for HCC patients. These three variables were also the most significant predictors for HCC drop out in our previous analyses of the OPTN database (9). The ability of these factors to predict dropout as measured by the index of concordance (0.66–0.69) indicates reasonable accuracy for predicting dropout but not nearly as accurate as MELD in predicting death for the non-HCC patients (12).

Tumor size has been consistently identified as important in predicting HCC progression in most studies (3,13). Tumor size is one of two primary factors (the other being number) used in the MC and in UCSF system (3) and there are reports suggesting that maximum tumor size may be more associated with vascular invasion than tumor number (14). We again identified AFP as a significant predictor of dropout. AFP has been frequently reported as an important predictor for HCC tumor progression (15,16). We chose to analyze AFP as a continuous variable in keeping with the tumor size and MELD score variables and did not select a ‘cut-off’ value to avoid the invariable specificity/sensitivity trade-offs when naturally continuous variables are dichotomized. Our results clearly indicate that AFP treated as a continuous variable (logarithmically transformed) is strongly associated with HCC dropout. Last, laboratory MELD score was a strong predictor of HCC patient dropout and interestingly, MELD is a stronger predictor than either tumor size or AFP for dropout in the HCC patients. Again, MELD score was identified as an important predictor for HCC drop out in our previous study (9) and several other studies have also found intrinsic liver disease severity as important in predicting HCC drop out. In fact, one of the Italian liver allocation systems incorporates recipient MELD into the HCC allocation sequence (17). Presumably, the underlying hepatitis that most of the HCC patients have is also progressing while they wait.

A number of previous reports have shown that loco-regional therapy (RFA, TACE) decreases waitlist dropout (3,18–20) and improves posttransplant survival, especially in patients who need to wait more than a few months for LT. In this analysis, ablative therapy was not associated with drop out, consistent with our previous analysis of an earlier MELD era (9). This finding might be due to the fact that, under the current U.S. system, most patients with an HCC priority score receive a transplant in less than a year of listing. Thus, any effect in arresting or slowing tumor progression caused by the ablative treatment may be obscured by the fact that the ablated tumors are removed with the transplant sooner than the ablated tumors could progress. Under the present system, given the variation in practice and regional organ availability, it would make it challenging to mandate that ablation be part of any policy regarding HCC prioritization.

The current system utilizes a continuous MELD score for non-HCC patients who may vary with time based on changing clinical conditions of the candidate. In contrast, the system that assigns a fixed priority for HCC patients meeting MC and subsequently awards fixed incremental points every 3 months is not a continuous system. Our data clearly show that HCC patients have varying degrees of drop out risk based on tumor size, AFP and MELD score variables. These results would suggest that HCC candidates would be better prioritized by using a continuous scoring system that ranks patients according to their risk of dropout. Importantly, and contrary to the current system, our data suggest that MELD score should be included as one component of any new HCC allocation policy. In addition, tumor size and log AFP have also been repeatedly identified as other important variables in the OPTN system that should be incorporated in a continuous fashion. Development of a continuous HCC priority score that is predictive of dropout would more closely resemble the system for patients without HCC and would help to equalize access to the donor pool for the HCC and non-HCC candidates. In order to allow for time to determine the rapidity of HCC progression and to avoid possibility that rapidly progressive tumors with very elevated AFP values are observed before transplant, some period of time of observation could be incorporated into such a policy. An HCC priority score based solely on tumor size, AFP and MELD might give undue priority to the patients at highest risk for recurrence with the potential for diminished outcomes. A time element would need to be incorporated into the HCC priority score that would hopefully mitigate the risk of increased posttransplant HCC recurrence.

In conclusion, our data suggest that HCC patients have better access to the liver transplant donor pool compared with the non-HCC patients as evidenced by their lower dropout rate from the waitlist. We suggest that the assignment of a static exception score that does not incorporate the underlying MELD score may be one explanation for this observation. Development of a continuous HCC priority score based on factors predictive of dropout like the ones we have identified here and others such as waiting time would be more equitable for HCC and non-HCC in regard to access to OLT.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgement
  7. References

This work was supported wholly or in part by Health Resources and Services Administration contract 234-2005-370011C. The content is the responsibility of the authors alone and does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products or organizations imply endorsement by the U.S. Government.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgement
  7. References