Increasing disparity in waitlist mortality rates with increased model for end-stage liver disease scores for candidates with hepatocellular carcinoma versus candidates without hepatocellular carcinoma†‡
This study was supported by the National Institutes of Health (grant 1-F32-DK-089694-01 from the National Institute of Diabetes and Digestive and Kidney Diseases to David Goldberg) and by the Health Resources and Services Administration (contract 234-2005-370011C). The contents of this article are the responsibility of the authors alone and do 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 US Government.
Since February 27, 2002, the United Network for Organ Sharing (UNOS) has relied on the Model for End-Stage Liver Disease (MELD) to prioritize candidates listed for liver transplantation in the United States. Initially developed and validated to predict 90-day waitlist mortality from the time of the initial listing, the MELD score has subsequently been shown to accurately predict 90-day waitlist mortality for candidates throughout their time on the wait list.1-6
There are candidates for whom the risk of waitlist removal for death or clinical deterioration is not adequately captured by the MELD score; these candidates include patients with extrahepatic manifestations of liver disease such as hepatopulmonary syndrome and metabolic liver disease and, most notably, hepatocellular carcinoma (HCC). For patients whose perceived and/or actual clinical risk the MELD score does not address, policymakers have instituted standardized criteria and algorithms by which candidates receive MELD exception points.7
Candidates with HCC whose tumors are within the Milan criteria (MC; either 1 tumor < 5 cm in diameter or 3 tumors with each < 3 cm in diameter) receive standardized MELD exception points. When MELD allocation was initiated, it was agreed that mortality was not the appropriate endpoint for HCC candidates, but instead tumor expansion beyond the MC would necessitate dropout from the wait list. As a result, candidates with HCC within the MC and with T2 lesions were assigned 29 points at the time of the initial listing with incremental points granted every 3 months (corresponding to a 10% increase in waitlist dropout). The initial point allocation was decreased to 24 in April 2003 and to 22 in January 2005. Both decreases occurred because it was noted that the allocated MELD score did not correlate with the risk of waitlist dropout for these candidates. There have, however, been no further changes in the allocation of incremental exception points over time. The current policy states that an HCC candidate with a tumor within the MC may receive an exception MELD score (“equivalent to a 15% probability of candidate death within 3 months”) and 3 additional points every 3 months (“equivalent to a 10% increase in candidate mortality”).7
Two recent publications have addressed the issue of differences in waitlist mortality between HCC candidates and non-HCC candidates, and they have demonstrated that non-HCC candidates have a greater risk of waitlist dropout.8, 9 However, these studies did not specifically address the systematic upgrades in exception points based on the waiting time; this issue remains unresolved.
The goal of this study was to explore the appropriate allocation of exception points for HCC candidates by (1) comparing the 90-day waitlist outcomes of candidates with an initial HCC exception score of 22 to non-HCC candidates with similar initial MELD scores; (2) determining whether HCC candidates with a MELD exception score of 22 have a 15% risk of waitlist dropout; (3) determining of whether HCC candidates with MELD exception scores of 25 and 28 in fact have 25% and 35% risks, respectively, of waitlist dropout; and (4) the comparing of the 90-day waitlist outcomes of HCC candidates with MELD exception scores of 25 and 28 to non-HCC candidates with similar MELD scores.
CI, confidence interval; HCC, hepatocellular carcinoma; MC, Milan criteria; MELD, Model for End-Stage Liver Disease; OPTN, Organ Procurement and Transplantation Network; OR, odds ratio; UNOS, United Network for Organ Sharing.
PATIENTS AND METHODS
Analyses were based on data from the Organ Procurement and Transplantation Network (OPTN)/UNOS database as of May 31, 2009. We included all adults (≥18 years old) who were listed for initial liver transplantation between January 1, 2005 and May 31, 2009. January 1, 2005 marked the implementation date for the current MELD allocation policy. The HCC group included all candidates who received exception points for T2 stage HCC (within the MC) according to the current exception policy. The non-HCC cohort included all candidates without a diagnosis of HCC after the exclusion of candidates with any other approved non-HCC exception. This enabled a comparison of HCC candidates and candidates with comparable laboratory MELD scores. All status 1 candidates were excluded.
The primary outcome was waitlist removal for death or clinical deterioration within 90 days. This endpoint included all candidates who were removed with the “died” code as well as candidates who were removed because they were too sick or medically unsuitable. We considered candidates who were removed from the wait list because of clinical deterioration as equivalent to candidates who died because chronic liver disease is almost uniformly fatal without transplantation. This grouping is consistent with previous research.10, 11 After the exclusion of clinical deterioration, death on the wait list was also evaluated as a secondary outcome.
For the cohort of candidates who were evaluated from the time of the initial listing, the 90-day outcomes were determined from the date of receipt of 22 MELD exception points for HCC candidates or from the date of listing for non-HCC candidates. For analyses of HCC candidates and non-HCC candidates who had already been listed, the period of 90 days was determined from the receipt of additional HCC exception points (MELD scores of 25 and 28) for HCC candidates or from the updating of the MELD score (24-26 or 27-29; see the next section) for non-HCC candidates.
The current analysis used the binary outcome of waitlist removal (yes/no) within 90 days for death or clinical deterioration rather than the time to waitlist removal because the MELD score is meant to predict 90-day waitlist mortality. Also, because HCC candidates are granted additional MELD exception points every 3 months, we wanted to determine the 90-day risk of waitlist removal for HCC candidates after the receipt of 22, 25, or 28 exception points.
All HCC candidates listed with 22 MELD exception points were compared to all non-HCC candidates listed with an initial laboratory MELD score of 21 to 23. A comparison group of non-HCC candidates with a calculated MELD score of 21 to 23 at listing was chosen because if the HCC exception MELD score accurately predicts 90-day waitlist survival from the initial listing, these 2 groups would be expected to have similar 90-day risks of waitlist removal for death or clinical deterioration. A MELD score range of 21 to 23 was chosen because limiting the analysis solely to non-HCC candidates with a calculated MELD score of 22 resulted in a sample size substantially smaller than the HCC cohort (n = 877). To ensure a lack of selection bias in the non-HCC group, a secondary analysis was conducted to compare candidates with 22 HCC MELD exception points to all non-HCC candidates with a MELD score of 21 to 23, regardless of the initial MELD score. This demonstrated a 90-day risk of waitlist removal or death similar to that demonstrated by the analysis of non-HCC candidates with an initial laboratory MELD score of 21 to 23 (data not shown).
The 90-day outcomes for HCC candidates with MELD scores of 25 and 28 were also compared to non-HCC candidates with MELD scores of 24 to 26 and 27 to 29, respectively. The non-HCC cohorts included all candidates with MELD scores of 24 to 26 and 27 to 29, regardless of the initial MELD score (see the flowchart in Fig. 1). This can be exemplified by a candidate who first enters the wait list with a MELD score of 14 but whose score subsequently escalates to 25. This candidate would have the same expected 90-day mortality as a patient who first enters the wait list with a different MELD score but also subsequently arrives at a MELD score of 25.1 Candidates with laboratory MELD scores of 24 to 26 and 27 to 29 at any point in time on the wait list were chosen to enable a comparison of non-HCC candidates and HCC candidates with equivalent MELD scores and thus similar predicted risks of waitlist removal for death or clinical deterioration. Non-HCC candidates could have a calculated MELD score of 24 to 26 or 27 to 29 at more than 1 point in time. Therefore, within each MELD category (24-26 and 27-29), we included only the first time point at which the MELD score met the criteria for that stratum.
To ensure that candidate outcomes were not included in more than 1 MELD category (eg, a candidate who was listed with a MELD score of 21 on January 1, 2006, whose MELD score increased to 25 on February 1, 2006, and who died on March 1, 2006, ie, within 90 days of MELD scores of 21 and 25), we included non-HCC candidates in a subsequent category only if they remained on the wait list for more than 90 days after their inclusion in the preceding category (eg, a candidate listed with a MELD score of 21 that was subsequently increased to 24 was included in the cohort of patients with MELD scores of 24 to 26 only if the listing date for the MELD score of 24 was at least 90 days after the listing date for the MELD score of 21).
The characteristics of HCC and non-HCC candidates were compared with Fisher's exact test and chi-square tests for dichotomous variables and with 2-sample t tests or Wilcoxon rank-sum tests for continuous variables. Chi-square tests were used to compare the unadjusted 90-day risk of removal for death or clinical deterioration between HCC and non-HCC candidates within each MELD category and between HCC candidates in the 3 different MELD categories. Student t tests were used to compare the MELD scores at the time of waitlist removal for HCC and non-HCC candidates. For comparisons of outcomes after waitlist removal, we used Social Security Death Master File data included in the OPTN/UNOS data set (updated as of March 14, 2011). For candidates removed from the wait list for clinical deterioration, Wilcoxon rank-sum tests were used to compare the average times between waitlist removal and actual death. As a supplementary analysis, chi-square tests were used to determine whether the 90-day risk of waitlist removal for death or clinical deterioration for HCC candidates listed with a MELD score of 22 was similar to the risk for non-HCC candidates with initial MELD scores of less than 21.
Logistic regression models were used to determine whether the odds of waitlist removal for death or clinical deterioration within 90 days of listing differed between HCC candidates and non-HCC candidates. Separate models were fitted to compare HCC candidates with exception MELD scores of 22, 25, and 28 to non-HCC candidates with MELD scores of 21 to 23, 24 to 26, and 27 to 29, respectively. Adjustment variables were selected for inclusion if they were independently associated with the outcome (P < 0.05), if their removal from the model changed the coefficient for HCC by ≥10%, or if they were needed for clinical validity. The variables tested with these criteria included the recipient's age at listing, sex, race/ethnicity, blood type, and insurance type (private or public). We included fixed effects for the UNOS region to account for unobserved heterogeneity in the risk of waitlist removal across regions and for any correlations due to the clustering of candidates within regions.12
The unadjusted posttransplant survival of HCC liver recipients was compared with log-rank testing. All analyses were conducted with Stata 11.13
This study received an exempt review from the institutional review board of the University of Pennsylvania.
Between January 1, 2005 and May 31, 2009, 6246 candidates were listed with T2 HCC exception points. The demographics and primary diagnoses of the HCC and non-HCC cohorts are displayed in Table 1. Compared to non-HCC candidates, HCC candidates were older (P < 0.001), more likely to be male (P < 0.001), and less likely to be white (P < 0.001) or Asian (P < 0.001). The primary diagnoses of the HCC cohorts differed from the diagnoses of each of the non-HCC cohorts. According to the UNOS primary diagnosis codes, HCC candidates were less likely to have alcoholic cirrhosis, nonalcoholic steatohepatitis/cryptogenic cirrhosis, or cholestatic cirrhosis. Although 28.7% (1793/6246) of the HCC candidates were listed with HCC as the primary diagnosis without a clear specification of the background liver disease etiology, 863 (48.1%) were positive for hepatitis C virus antibody. The inclusion of these candidates with HCC patients listed with hepatitis C virus as the primary diagnosis yielded 3526 HCC patients (56.4%) with hepatitis C virus (ie, a significantly greater proportion than the proportion of the non-HCC cohort).
Table 1. Patient Demographics and Primary Diagnoses
HCC Candidates With a MELD Score of 22 (n = 6246)
MELD Score of 21-23 (n = 2564)
MELD Score of 24-26 (n = 4655)
MELD Score of 27-29 (n = 2737)
The data are presented as means and standard deviations.
Other includes patients whose ethnicity was defined as Hawaiian, Native American, or multiracial.
Figure 2 depicts the rates of waitlist removal for death or clinical deterioration within 90 days for candidates with HCC and candidates without HCC in the 3 MELD categories. Two hundred fifty-nine of 6246 HCC candidates (4.1%) who were listed with 22 exception points were removed from the wait list because of death or clinical deterioration within 90 days of listing, whereas 283 of 2564 non-HCC candidates (11.0%) with an initial MELD score of 21 to 23 were removed. Similar data were obtained when all non-HCC candidates with a MELD score of 21 to 23, regardless of the initial MELD score, were analyzed (data not shown). Ninety-three of the 2009 HCC candidates (4.6%) with a MELD score of 25 were removed from the wait list for death or clinical deterioration within 90 days, whereas 805 of the 4655 non-HCC candidates (17.3%) with a MELD score of 24 to 26 were removed. Lastly, 20 of the 675 HCC candidates (3.0%) with a MELD score of 28 were removed from the wait list for death or clinical deterioration within 90 days, whereas 646 of the 2737 non-HCC candidates (23.6%) with a MELD score of 27 to 29 were removed.
The risk of death alone differed between HCC and non-HCC candidates (P < 0.001) within each MELD stratum (2.0% for a MELD score of 22 versus 7.6% for a MELD score of 21-23, P < 0.001; 2.1% for a MELD score of 25 versus 11.4% for a MELD score of 24-26, P < 0.001; and 1.3% for a MELD score of 28 versus 15.6% for a MELD score of 27-29, P < 0.001). However, the risk of death alone for HCC candidates with exception MELD scores of 22, 25, and 28 were statistically similar (2.0% versus 2.1% versus 1.3%, P = 0.42).
In the HCC cohorts, the proportion of waitlist removals for death equaled the proportion for clinical deterioration (2.0% versus 2.1% for a MELD score of 22, P = 0.49; 2.1% versus 2.5% for a MELD score of 25, P = 0.46; and 1.3% versus 1.7% for a MELD score of 28, P = 0.65). Conversely, in each of the non-HCC MELD cohorts, the proportion of waitlist removals for death was approximately twice the proportion for clinical deterioration (7.6% versus 3.4% for a MELD score of 21-23, P < 0.001; 11.4% versus 5.9% for a MELD score of 24-26, P < 0.001; and 15.6% versus 8.0% for a MELD score of 27-29, P < 0.001).
There was regional variability in the 90-day rates of waitlist removal for death or clinical deterioration among the HCC and non-HCC cohorts (Supporting Fig. 1). The variability was similar among the HCC cohorts with MELD scores of 22 and 25, but it was much more pronounced among the non-HCC cohorts.
Among the HCC candidates who were removed for death or clinical deterioration, the mean laboratory MELD score at removal was 22.4 ± 10.2, and the scores were similar across all 3 MELD categories. However, among the non-HCC candidates, the MELD score at removal increased from 29.2 ± 7.7 for the cohort with MELD scores of 21 to 23 to 31.9 ± 9.0 for the cohort with MELD scores of 24 to 26 and to 32.2 ± 6.5 for the cohort with MELD scores of 27 to 29. Among the HCC candidates removed for death or clinical deterioration, the laboratory MELD score at removal was higher than the exception MELD score for 106 of the 259 candidates (40.9%) with a MELD score of 22, for 22 of the 93 candidates (23.7%) with a MELD score of 25, and for 5 of the 20 candidates (25.0%) with a MELD score of 28.
Among the HCC and non-HCC waitlist candidates in each MELD stratum who were removed for clinical deterioration, similar proportions subsequently died during follow-up. However, in each MELD stratum, HCC candidates survived longer after delisting than non-HCC candidates (Table 2). The median survival times were as follows: 27 days (interquartile range = 3-187 days) for patients with a MELD score of 22 versus 2 days (interquartile range = 1-14 days) for patients with a MELD score of 21 to 23 (P < 0.001), 87 days (interquartile range = 8-202 days) for patients with a MELD score of 25 versus 3 days (interquartile range = 1-13 days) for patients with a MELD score of 24 to 26 (P < 0.001), and 32.5 days (interquartile range = 1-175 days) for patients with a MELD score of 28 versus 1 day (interquartile range = 0-6 days) for patients with a MELD score of 27 to 29 (P = 0.02).
Table 2. Removal Due to Clinical Deterioration During Follow-Up and Time to Death
Candidates were removed because they were too sick or were medically unsuitable.
The data are presented as medians and interquartile ranges.
Comparison of the number of days from delisting to death for HCC and non-HCC candidates within each MELD category (Wilcoxon rank-sum test).
HCC MELD score of 22
Non-HCC MELD score of 21-23
HCC MELD score of 25
Non-HCC MELD score of 24-26
HCC MELD score of 28
Non-HCC MELD score of 27-29
To explore whether the risk of waitlist dropout for HCC candidates varied according to the initial listing laboratory MELD scores, we stratified HCC candidates into 2 groups: laboratory MELD scores ≥ 15 and laboratory MELD scores < 15.14 Candidates with a laboratory MELD score ≥ 15 at the initial listing had a higher 90-day risk of removal for death or clinical deterioration (8.4% versus 2.5% for the cohort with a MELD score of 22, P < 0.001; 7.8% versus 3.6% for the cohort with a MELD score of 25, P < 0.001; and 5.2% versus 2.3% for the cohort with a MELD score of 28, P = 0.06). However, according to a comparison of all candidates with an initial laboratory MELD score ≥ 15, the risk of waitlist dropout remained higher for non-HCC candidates versus HCC candidates.
In univariate regression models, all tested recipient variables (except for the blood type) were significantly associated with the risk of 90-day waitlist removal for death or clinical deterioration. Nevertheless, the blood type was included in multivariate models because of the potential association with the outcome of interest.15 In the multivariate regression models assessing the 90-day survival of HCC candidates, each of the other tested independent variables (age at listing, sex, race/ethnicity, and insurance) was significantly associated with the odds of waitlist removal for death or clinical deterioration within 90 days.
Table 3 provides unadjusted and adjusted odds ratios (ORs) of 90-day waitlist removal due to death or clinical deterioration for HCC and non-HCC candidates within each MELD stratum. The adjusted odds of waitlist removal were substantially and consistently lower for HCC candidates versus non-HCC candidates, and they decreased with increasing MELD scores [for a MELD score of 22, OR = 0.32, 95% confidence interval (CI) = 0.27-0.39; for a MELD score of 25, OR = 0.21, 95% CI = 0.17-0.27; and for a MELD score of 28, OR = 0.09, 95% CI = 0.05-0.14]. We obtained similar results when we examined 90-day waitlist removal for death alone.
Table 3. ORs Comparing 90-Day Wait List Mortality and Removal Due to Clinical Deterioration for HCC Candidates and Non-HCC Candidates Within Each MELD Category
The multivariate model was adjusted for the recipient's age at listing, sex, race/ethnicity, blood type, and insurance status (private versus public); the UNOS region was treated as a fixed effect.
Non-HCC MELD score of 21-23
Non-HCC MELD score of 24-26
Non-HCC MELD score of 27-29
Transplantation for HCC Candidates and Non-HCC Candidates
Transplantation rates within 90 days were higher for each of the non-HCC cohorts versus the HCC cohorts with similar MELD scores, with increasing rates correlating with increasing MELD scores (data not shown). The median donor risk index, the mean donor age, and the use of donation after cardiac death organs were similar for all 6 groups of patients. There was increased use of regionally shared organs in the non-HCC cohorts with MELD scores of 24 to 26 and 27 to 29 versus their respective HCC cohorts (data not shown). Across the 3 HCC cohorts, similar proportions of patients underwent transplantation with laboratory MELD scores of <15 (65.0%-70.5%), 15 to 21 (23.7%-28.3%), and >22 (5.8%-7.7%).
Posttransplant Outcomes for HCC Candidates With MELD Exception Scores of 22, 25, and 28
For HCC candidates who underwent transplantation, posttransplant patient survival did not differ between the 3 MELD strata (log-rank P = 0.20; Table 4 and Fig. 3). By the end of the follow-up, 465, 161, and 48 HCC transplant recipients (16.5%, 17.1%, and 13.1%) with exception MELD scores of 22, 25, and 28, respectively, had died. The causes of graft failure are not well coded for determining the risk of recurrent HCC in transplant patients. Although the risk of posttransplant mortality was higher for non-HCC recipients versus HCC transplant recipients in the corresponding MELD category, the time to death was longer for non-HCC recipients (data not shown).
Table 4. Posttransplant Outcomes of Recipients With HCC
Exception MELD Score for HCC at the Time of Transplantation
Risk of Wait List Removal for HCC Candidates With a MELD Score of 22 Versus Non-HCC Candidates With a MELD Score < 21
We compared the 90-day risk of waitlist removal for death or clinical deterioration for HCC candidates with an initial MELD score of 22 and 4 groups of non-HCC candidates with initial MELD scores less than 21 (Table 5). According to a comparison with non-HCC candidates with MELD scores of <15, 15 to 16, 17 to 18, and 19 to 20, the unadjusted 90-day risk of waitlist removal for death or clinical deterioration for patients with an HCC MELD score of 22 was most comparable to the risk for non-HCC candidates with an initial MELD score of 15 to 16 (4.2% versus 4.1%).
Table 5. Risk of Death or Removal Due to Clinical Deterioration for HCC Candidates and Non-HCC Candidates With MELD Scores of 6 to 20 at the Time of Listing
We have shown that under the current allocation policy, the odds of waitlist removal for death or clinical deterioration are significantly lower for HCC candidates versus non-HCC candidates. The observed differences increase steeply as the analysis progresses from low MELD strata to high MELD strata. These results not only reinforce previous work that has questioned the initial MELD exception points bestowed upon HCC candidates within the MC, but also newly challenge the incremental MELD exception points granted at 3-month intervals. Our data suggest that neither the initially awarded MELD exception points for HCC nor the incrementally awarded points accurately reflect the risk of waitlist removal for HCC candidates, particularly when they are juxtaposed against non-HCC candidates.
The current MELD allocation system is based on a principle set forth by the Department of Health and Human Services in 1998: an organ allocation policy should be based on medical urgency.16 The Institute of Medicine resoundingly echoed this sentiment 1 year later with its recommendation to eliminate waiting time as an allocation criterion. An allocation policy should ensure equitable access to organs based on candidates' medical characteristics and disease prognosis rather than waiting times.3, 17 Our analyses show that with respect to HCC candidates versus non-HCC candidates, the current allocation policy fails in both ways.
Although OPTN policy states that the initial 22 MELD exception points allocated to HCC candidates are meant to be equivalent to a 15% probability of waitlist dropout within 3 months, the data do not support this contention.7 Only 4.1% of the HCC candidates with T2 stage HCC tumors who were listed with a MELD score of 22 were removed from the wait list within 90 days because of death or clinical deterioration, whereas 11.0% of the non-HCC candidates with MELD scores of 21 to 23 were removed. Although the data presented here are similar to the data reported by Washburn et al.,8 the 2 studies differ in that Washburn et al. compared HCC candidates to all non-HCC candidates, regardless of the MELD scores. Massie et al.9 recently reported similar data as well. However, their HCC comparator group included all HCC candidates with 22 MELD exception points and not only those candidates within the MC. From January 1, 2005 to March 23, 2011, 7360 HCC candidates within the MC were accepted with 22 MELD exception points. Another 1034 candidates (14.0%) were categorized according to the UNOS coding criteria as “HCC not meeting criteria” but received 22 MELD exception points through a petition to a regional review board [ie, they received the exception diagnosis code of 3 (“HCC not meeting criteria”), but the exception HCC code was listed as yes]. These candidates, who were presumably outside the MC, were not included in our analysis. This difference may explain the higher waitlist dropout rate reported by Massie et al.
We extended our analyses of HCC candidates and non-HCC candidates to provide a novel assessment of the differences in the risk of waitlist removal between HCC candidates with 25 MELD exception points and HCC candidates with 28 MELD exception points and between these HCC candidates and non-HCC candidates with comparable laboratory MELD scores.
The magnitude of the difference in the risk of death between HCC candidates and non-HCC candidates was staggering, especially at the higher MELD levels. With increasing MELD scores, the risk of waitlist removal was stable for HCC candidates (4.6% for patients with a MELD score of 25 versus 3.0% for patients with a MELD score of 28), whereas it increased significantly for non-HCC candidates (17.3% for patients with MELD scores of 24-26 versus 23.6% for patients with MELD scores of 27-29). The additional exception points granted to HCC candidates every 3 months are meant to reflect the increasing risk of tumor extension beyond acceptable MC boundaries, with the number of points meant to reflect a 10% increase in the risk of waitlist dropout.7 The data that we have presented clearly contradict the notion that waitlist dropout secondary to tumor progression increases over time. Point increases in the calculated MELD score directly correspond to a natural log increase in the risk of death. The fact that waitlist dropout for HCC does not escalate in the same manner calls into question the entire premise of granting exception MELD points to HCC candidates. Survival after waitlist removal was significantly longer for HCC candidates who were delisted for clinical deterioration rather than death versus non-HCC candidates.
Despite similar MELD scores at listing, the candidates with HCC consistently had lower laboratory MELD scores (13.2 for the patients with a MELD score of 22 versus 22.0 for the patients with a MELD score of 21-23, 13.1 for the patients with a MELD score of 25 versus 24.8 for the patients with a MELD score of 24-26, and 13.2 for the patients with a MELD score of 28 versus 27.8 for the patients with a MELD score of 27-29). The fact that the proportion of waitlist removals was stable across the 3 cohorts of HCC candidates but increased significantly for the non-HCC candidates suggests that the risk of waitlist removal is directly related to the severity of the underlying liver disease rather than tumor progression. Further support for this concept is the fact that the 90-day risk of waitlist removal for death or clinical deterioration was similar for HCC candidates listed with 22 exception points and non-HCC candidates listed with a laboratory MELD score of 15 to 16.
Although additional exception points are given to HCC candidates according to the waiting time because of concerns about tumor progression, the current MELD exception point policies were created in an era when locoregional therapies for HCC were not as widespread, and they have not been altered despite evidence that these therapies have decreased waitlist dropout rates.18-20 Current locoregional strategies have been shown to improve the survival of patients with HCC (both within and beyond the MC). Among patients with HCC who are not transplant candidates and receive only locoregional therapy, those with a radiographic response have a median survival of 2 years.21 Furthermore, recent Italian data demonstrated that candidates with T2 stage tumors who responded to locoregional therapy had a lower risk of waitlist dropout and decreased posttransplant HCC recurrence.20 This correlation between the response to therapy and posttransplant HCC recurrence likely explains our finding that the posttransplant survival rates of HCC recipients with 22, 25, and 28 exception points were similar, with the recurrence risk related more to the therapy response than the waiting time. The waitlist mortality for HCC candidates decreased with rising exception scores, which correlated with longer wait times. The available data do not support a definitive rationale for this finding. Although this observation may be due to a more favorable tumor biology, an alternative explanation is that these candidates received additional locoregional therapies as they accrued exception points. Altogether, there is compelling evidence that current policies of automatic, incremental upgrades in exception points must be thoroughly re-evaluated.
Our study has several limitations. The use of the OPTN/UNOS database limited the data elements available for covariate adjustments. However, the OPTN/UNOS database not only has been used for previous research evaluating this question but also is used in the design and alteration of allocation policy. For our outcome definition, we relied on the outcomes of HCC candidates under the current allocation system. We would have been unable to determine their risk of death or clinical deterioration if they had not been given the same exception points. Although it is theoretically possible that fewer HCC candidates would have undergone transplantation and more HCC candidates would have been removed from the wait list for death or clinical deterioration if they had not received exception points, it is unlikely that this explains the results that we see. Because of the magnitude of the difference in the risk of death for HCC candidates versus non-HCC candidates, the risk of death for HCC candidates who underwent transplantation would have needed to be substantially larger than the risk for non-HCC candidates who underwent transplantation to alter our current results. Our analyses focused only on candidates' 90-day outcomes from a single point in time. Although this has the potential for introducing a time-dependent misclassification bias, a candidate's current MELD score is the single most important determinant of mortality risk on the waiting list.1, 9 Moreover, our goal was to evaluate the 90-day risk of waitlist dropout for HCC candidates from the time of receiving exception points to determine the appropriateness of current policies.1, 9 We do not have data on the impact of locoregional therapies, which directly correlate with survival and could help us to stratify candidates with HCC for exception points.21 UNOS provides some data on whether candidates have received locoregional therapy, but detailed data on the number of treatments and the success of treatments are limited. Lastly, although we could compare posttransplant survival across the 3 HCC MELD cohorts, we could not evaluate differences in recurrent HCC because these data are not reliably coded in the OPTN/UNOS database.
In conclusion, our data suggest that HCC candidates have substantially lower odds of waitlist removal for death or clinical deterioration than non-HCC candidates across a broad range of MELD scores. The initial allocation of 22 MELD exception points to HCC candidates is too high according to their actual risk of 90-day waitlist dropout from listing. The steep increase in the risk differential between HCC patients and non-HCC patients within higher MELD strata (25 and 28) challenges the notion that an increased waiting time for HCC candidates is associated with an increased risk of waitlist dropout justifying the systematic addition of acceptance points every 3 months. These data highlight the urgent need to reconsider the equitable treatment of HCC patients and non-HCC patients by our national allocation policy. Although the intention of bestowing HCC exception points to accurately reflect the risk of waitlist removal for tumor expansion beyond acceptable listing criteria appears eminently reasonable, practice has demonstrated that the risk has been grossly overestimated. Without question, certain HCC candidates merit exception points, but other factors, including their response to locoregional therapy, the severity of the underlying liver disease, and regional access to transplantation, must be taken into account. Although future research is planned to determine optimal allocation policies for HCC candidates with respect to non-HCC candidates, our data strongly indicate that the current number of exception points allocated for HCC should be lowered.