OPTN Policy Regarding Prioritization of Patients with Hepatopulmonary Syndrome: Does It Provide Equitable Organ Allocation?

Authors


Abstract

United Network for Organ Transplantation (UNOS) policy 3.6.4.5.1 provides exception points to patients diagnosed with hepatopulmonary syndrome (HPS) to compensate for their reported increased mortality risk. We compared pre- and posttransplant and overall outcomes in 255 patients receiving exception points under this policy (HPS policy patients) with 32 358 nonexception control patients listed in the model for end-stage liver disease (MELD) era to determine whether the intent of the policy is being met.

Overall, 92.5% of HPS policy patients versus 45.5% of controls had been transplanted, 5.1% versus 31.2% remained on waiting list and 1.5% versus 14.1% had died while awaiting transplant (p < 0.0001 for each comparison). Relative risk (RR) of death for HPS policy patients compared to controls was 0.158 (confidence interval [CI]: 0.059–0.420, p = 0.0002) pretransplant, and 0.827 (CI: 0.587–1.170, p = 0.28) posttransplant. Overall (combined waitlist and posttransplant) RR of death was 0.514 (CI: 0.374–0.707, p = 0.00004) compared with controls. After adjustment for laboratory MELD, overall RR was 0.807 (CI: 0.587–1.110, p = 0.19), indicating that HPS policy patients' mortality risk would be similar to that of controls had they been listed with their laboratory MELD score.

HPS policy patients have a significant pretransplant survival advantage over standard liver transplant candidates because of the exception points awarded, and have similar posttransplant survival. Better criteria for diagnosing and grading of HPS are required.

Abbreviations: 
HPS

hepatopulmonary syndrome

MELD

model for end-stage liver disease

K-M

Kaplan–Meier

RR

relative risk

CI

confidence interval

UNOS

United Network for Organ Transplantation

OPTN

Organ Procurement Transplant Network

SRTR

Scientific Registry of Transplant Recipients

A-a gradient

alveolar arterial oxygen gradient

PaO2

arterial oxygen partial pressure

Introduction

Most (1–4), but not all (5), studies suggest that cirrhotic patients with hepatopulmonary syndrome (HPS) have a worse prognosis while waiting for liver transplantation than that in similar cirrhotic patients without HPS. As mortality risk in cirrhosis is modeled accurately by the end-stage liver disease (MELD) score (6,7), this implies that HPS patients would be expected to have a higher waiting list mortality risk than that predicted by their laboratory MELD score.

Posttransplant mortality is also reported to be higher in HPS patients (2,8–13). Arterial oxygen partial pressure (PaO2) of <300 mmHg on breathing 100% oxygen (10), PaO2 of <50 mmHg on room air (2), high macroaggregated albumin (MAA) shunt fraction (10), or both (8), were reported to predict poor posttransplant outcomes by some, but not all, studies (13). Several recent series, however, have reported excellent posttransplant survival in HPS patients (14–16). Even if there is not consensus about pre- or posttransplant survival, almost all studies have shown liver transplantation to be highly effective in reversing HPS (8,9,11,13–24), although subclinical abnormalities may persist (25).

Since the inception of MELD, the United Network for Organ Sharing (UNOS) policy 3.6.4.5.1 has allocated exception points to HPS patients because their prognosis is thought to be worse than that predicted by MELD (26). The policy criteria for HPS are quoted: ‘Candidates with clinical evidence of portal hypertension, evidence of a shunt, and a PaO2 <60 (mmHg) on room air may be referred to the regional review board (RRB). Candidates should have no significant clinical evidence of underlying primary pulmonary disease’ (26). The policy does not specify the number of exception points to be allocated. It states ‘sufficient MELD points are allocated that would provide candidates with a reasonable probability of being transplanted within 3 months’ (26). Questions were raised about the diagnostic accuracy of these criteria at the recent conference of the MELD Exception Study Group and Conference (MESSAGE) (27).

The policy criteria may not accurately distinguish HPS from other causes of hypoxemia (28), or define the severity of the HPS. The diagnostic criteria for HPS are recognized to be imprecise, and there is no confirmatory test (29–31). Hence, some subjects who meet all UNOS policy 3.6.4.5.1 diagnostic criteria may not have HPS (29–31). On the other hand, some patients with HPS may not be referred because of variable screening for HPS from center to center. These problems inherent to the UNOS criteria were addressed at the recent MELD Exception Study Group and Conference (MESSAGE), and specific actions to correct these problems have been proposed (27).

Because HPS policy criteria may not be sufficiently precise, we sought to establish whether the policy was achieving its intention of allocating organs equitably between HPS and non-HPS patients. Given the limitations of the data available, and the inherent difficulties in establishing a diagnosis of HPS, we did not attempt to verify whether HPS policy patients approved for exception points truly had HPS. This is a study of whether policy 3.6.4.5.1 leads to equitable organ allocation, not a study of the HPS as such. Equitable organ allocation is understood to mean that mortality of patients receiving points under policy 3.6.4.5.1 should be similar to mortality of control patients not receiving exception points.

Patients and Methods

Data reported here were supplied by the UNOS (the contractor for the Organ Procurement and Transplantation Network [OPTN]) in the form of the liver Standard Transplant Analysis and Research (STAR) file of May 1, 2007. This data set includes death information from the Social Security Death Master File (SSDMF), permitting crosschecking of deaths with that source. The interpretation and reporting of these data are the responsibility of the authors, and in no way should be seen as an official policy of or interpretation by the OPTN or the U.S. government. The University of Iowa Institutional Review Board approved performance of this study.

All patients listed for liver transplantation between the inception of MELD on February 27, 2002, and the end of the study on March 1, 2007, were included. This end date was chosen because ascertainment of deaths from the SSDMF appeared to be complete up to that date. Patients listed pre-MELD, but who were transplanted after February 27, 2002, were excluded from the analysis as these patients had already been listed for highly variable times before the change to MELD.

The HPS policy group included all patients in whom exception points were granted under UNOS policy 3.6.4.5.1, and for whom the exception points were still active at the time of removal from the list or end of the study. We excluded from the analysis 26 patients for whom exception points were requested, but not granted, and 30 patients initially granted points, but whose exception points lapsed or were rescinded prior to removal from the waiting list. UNOS does not track pulmonary function, blood gas or shunt data in patients with HPS (26), so we could not establish whether the diagnosis of HPS was correct, or the severity of the pulmonary shunt.

The control group consisted of all patients who were listed for transplant on or after February 27, 2002, and who did not meet criteria for any form of exception (HPS, hepatocellular carcinoma, amyloidosis, portopulmonary hypertension, primary oxaluria, other metabolic diseases, nonmetastatic hepatoblastoma, etc.). Primary outcome was overall mortality from the time of listing till death or the end of the study, including both deaths on the waiting list and peri- and posttransplant deaths. A patient was identified as dead if so indicated in either UNOS data or the SSDMF. In the occasional cases where these two files gave different dates of death, the earlier date was arbitrarily accepted. Secondary outcomes included time on waiting list, death while awaiting transplant, posttransplant deaths and posttransplant graft failure.

Only UNOS removal code 8 ‘died’ was used as death outcome for the analysis of death on the waiting list. Codes 5 (unsuitable), 6 (refused), 7 (transferred), 9 (other), 12 (condition improved), 13 (too sick) and 17 (removed in error) were censored. However, for analysis of overall death (all deaths from initial listing to end of study), the SSDMF was used; this captured all deaths irrespective of how they were coded at the time of removal from the list. The following removal codes were treated as transplanted: 3, 14 (transplanted at another center), 4 (deceased donor transplant), 15 (living donor transplant), 19 (multiorgan transplant), 21 (died during transplant) and 18 (emergency deceased donor transplant). Baseline characteristics of the policy 3.6.4.5.1 patients and the nonexception control patients were compared by t-tests (continuous variables) or the chi-square test (categorical variables), as appropriate. Outcomes of patients (or grafts in patients) meeting HPS policy criteria were compared to control patients (grafts) using the Kaplan–Meier (K-M) methods and the log-rank test. We also used the Cox proportional hazard modeling to determine relative risks (RR) for the various outcomes controlling for certain variables. We investigated whether differences in pretransplant outcomes were related to the regional differences in frequency of granting the HPS exception using a mixed survival model, accomplished by ‘clustering’ for region using the Cox proportional hazard function in S-Plus (Insightful Corp., Seattle, WA) (32). Since by UNOS policy, only the MELD score is used to allocate livers, in analyses of waiting time on the transplant list, death on the waiting list and overall death, we adjusted only for MELD scores (either laboratory or final) at the time of removal from the list. In the analysis of posttransplant patient and graft survival, we controlled for all variables shown by the Scientific Registry for Transplant Recipients (SRTR) to affect liver transplant outcomes (33). These factors included donor age, status 1, split liver or DCD donor liver, multiorgan transplant, cause of donor death, recipient age, whether the recipient was on mechanical support, recipient serum creatinine, presence of ascites, distance of donor from recipient center, recipient final laboratory MELD and recipient etiology of liver disease. As there were significantly more high-risk (status 1, split liver, multiorgan transplant) patients in control group, we also did a separate sensitivity analysis after excluding these patients.

Data management, statistical analysis and preparation of graphics were done using SAS for Windows, version 9.1 (SAS Institute, Inc., Cary, NC) and S-Plus for Windows, version 8.0. Exact permutational p-values for contingency tables were computed using StatXact-8 (Cytel Software Corp., Cambridge, MA).

Results

Patients and controls

The distribution of patients and their outcomes is shown in Figure 1. There were 255 HPS policy patients of whom 236 (92.5%) were transplanted, 4 (1.5%) died while waiting and 13 (5.1%) remained on the waiting list at the end of the study. The control group consisted of 32 358 patients of whom 14 719 (45.5%) had been transplanted, 4561 (14.1%) died while waiting and 10 091 (31.2%) remained listed. Table 1 gives the reasons for removal from the list in the HPS policy and control groups. There was only one death in HPS policy patients in the censored categories, as opposed to 1318 deaths among censored patients in the control group. Therefore, exclusion of deaths among the censored patients in this analysis minimizes the survival advantage for the HPS policy patients.

Figure 1.

Distribution of patients. HPS = HPS policy patients; WL-Deaths = deaths on waiting list; Post-TP = posttransplant. Other = see Table 1 for details.

Table 1. Reasons for removal from transplant list
 Reason for removal (UNOS code)HPS policyControl
TransplantedTransplanted at another center (3, 14)  0   43
Deceased donor (4)23413 718
Living donor (15)  2  885
Multiorgan transplant (19)  0    2
Died during transplant (21)  0   61
Deceased donor emergency TP (18)  0   10
Total transplanted 23614 719
DeathDied on list (8)  4 4546
Still waitingNo removal code 1310 091
AliveDeadAliveDead
 
CensoredUnsuitable (5) 00   73  66
Refused transplant (6) 00  108  46
Transferred (7) 00  161  58
Other (9) 10  744 285
Improved (12) 00  266  24
Too sick (13) 00  326 835
Removed in error (16) 01    6   4
Total (censored groups) 11 16841318

There were no significant baseline differences between HPS policy patients and controls with regard to the following: level of education, primary payer status (private, government, etc.), history of variceal bleeding, transjugular intrahepatic portal systemic shunt, hepatic encephalopathy, prior abdominal surgery or prior malignancy. There were also no significant differences between groups with regard to the number of dialyses before transplant, the presence of peripheral vascular disease, muscle wasting, coronary artery disease, portal vein thrombosis or diabetes.

Significant differences in baseline characteristics of listed HPS policy patients and nonexception controls are given in Table 2. There were significantly more females, White and Hispanic patients and patients with alcoholic or viral hepatitis in the HPS policy group than those among controls. HPS policy patients, in general, had less advanced disease than control patients, as demonstrated by a lower frequency of hepatic encephalopathy, ascites, being on mechanical support or mechanical ventilation or having a condition requiring intensive care unit (ICU) or hospital admission. More HPS policy patients required some assistance (due to mild disability), possibly due to their pulmonary abnormalities, but fewer were disabled, reflecting their less advanced liver disease. However, more HPS policy patients had a history of cerebrovascular disease.

Table 2. Pretransplant variables
Pretransplant variableHPS N (%)Control N (%)p
  1. Analysis of baseline characteristics of listed HPS policy and control patients. Number (N) and percentages (%) shown. Continuous data shown as mean and standard deviation (SD). Under etiology, category ‘other’ includes acute hepatic necrosis, metabolic diseases, nonalcoholic steatohepatitis, tumor and other. ‘Biliary’ includes biliary atresia, primary biliary cirrhosis (PBC) and primary sclerosing cholangitis (PSC).

  2. ICU = intensive care unit; FU = follow-up; COPD = chronic obstructive pulmonary disease; BMI = body mass index.

Gender
 Female107 (42.00)11 587 (35.80)0.04
 Male148 (52.00)20 771 (64.20) 
 Total25532 358 
Age52.40 (9.53)52.18 (9.64)0.69
Race
 Asian4 (1.57)1010 (3.12)0.022
 Black8 (3.14)2415 (7.46) 
 Hispanic44 (17.25)4568 (14.12) 
 Multiracial1 (0.39)325 (1.00) 
 White198 (77.65)24 037 (74.29) 
BMI28.00 (5.49)30.94 (6.34)0.17
Etiology
 Etoh96 (40.68)4822 (32.76)0.0012
 Viral102 (43.22)5504 (37.39) 
 Biliary8 (3.10)3041 (9.40) 
 Other35 (13.70)5913 (18.27) 
Encephalopathy
 None98 (41.53)4042 (27.46)<0.0001
 Stage 1–2135 (57.20)8979 (61.00) 
 Stage 3–43 (1.27)1698 (11.54) 
Ascites
 Present59 (60.2)4957 (76.97)<0.0005
Cerebrovascular disease5 (1.99)211 (0.66)0.0373
Functional status
 Normal156 (66.38)20 397 (74.2)0.001
 Slight assist68 (28.94)5338 (19.42) 
 Disabled11 (4.68)1748 (6.36) 
Mechanical sopport0 (0)633 (1.96)0.024
Medical condition
 In ICU1 (0.40)1416 (4.45)<0.0001
 Hospital6 (2.39)3159 (9.93) 
Outpatient244 (97.21)27 242 (85.62) 
Outcomes at end of studyHPS N (%)Control N (%)p
 
Outcome
 Dead4 (1.57)4546 (14.05)<0.0001
 Transplant236 (92.55)14 719 (45.49) 
 Waiting13 (5.10)10 091 (31.19) 
 Other2 (0.78)3002 (9.28) 
Death by end of FU (includes post-TP deaths)
 Dead38 (14.90)8313 (25.69)<0.0001

Significantly more subjects in the HPS policy group had a history of pulmonary embolus (p < 0.0001), and were on treatment for chronic obstructive pulmonary disease (COPD) (p < 0.0001) (Table 3). This suggests that at least in some patients in the HPS policy group, hypoxemia may have been due to or aggravated by COPD or pulmonary embolus rather than by true HPS. There was striking regional variation in the proportion of patients granted exception points under HPS policy, with regions 3, 5 and 11 accounting for more than half of all cases, each of these regions having a 2.5- to 3-fold higher proportion of HPS cases than that in region 7, which had the lowest rate, and almost twice the average proportion of all centers.

Table 3. Pulmonary disease and regional variation of HPS policy and control patients.
COPD treatment Pulmonary embolusHPS N (%) 18 (7.17) 4 (1.59)Control N (%) 441 (1.39) 69 (0.22)  p <0.0001 <0.0001
N% All HPSRegional %N% All ControlRegional %
  1. HPS = hepatopulmonary syndrome policy patients; % HPS = HPS policy patients transplanted as percentage of all HPS policy patients; % control = control patients transplanted as percentage of all control patients; Regional %= percentage of patients within region being transplanted for HPS policy or control.

Region
  111 4.310.821331 4.1199.18<0.0001
  224 9.410.54443713.7199.46<0.0001
  35923.141.52381411.7998.48<0.0001
  422 8.630.703121 9.6599.30<0.0001
  54216.470.70598518.599.30<0.0001
  6 6 2.350.581023 3.1699.42<0.0001
  711 4.310.3828808.999.62<0.0001
  812 4.710.641873 5.7999.36<0.0001
  917 6.670.592884 8.9199.41<0.0001
 1022 8.630.952304 7.1299.05<0.0001
 112911.371.062706 8.3698.94<0.0001

Waiting list outcomes

Overall outcomes of patients are summarized at the bottom of Table 2. The fraction of patients dying on the waiting list was almost one-ninth as frequent in HPS policy patients as in control patients (1.57% vs. 14.05%, p < 0.0001). HPS policy patients were more than twice as likely than control patients to get a transplant (92.55% vs. 45.49%, p < 0.0001), and consequently significantly fewer remained on the waiting on the list at the end of the study (5.10% vs. 14.05%, p < 0.0001) (Figure 1 and Table 2).

Survival analysis of waiting list outcomes

As intended, HPS policy patients got transplanted significantly earlier than controls, with median waiting time of 199.53 (278) days on the list versus 362.02 (439) days (p < 0.009) (Figure 2A), and significantly fewer HPS policy patients remained on the waiting list at the end of the study than controls (Table 2). Unadjusted Cox analysis (Table 4) showed that the likelihood of receiving a transplant in HPS policy patients was more than 2.5 times that in controls (RR: 2.527, confidence interval [CI]: 2.223–2.870, p < 0.0001). Adjusting for laboratory MELD further increased the RR of receiving a transplant to 3.351 (CI: 2.945–3.81, p < 0.0001), consistent with the fact that HPS policy patients were listed with lower laboratory MELD scores compared with those in the controls. When the Cox analysis was adjusted for final MELD (including the exception points), the RR was reduced to only 1.26, indicating that the major part of the advantage enjoyed by the HPS policy patients was attributable to the added exception MELD points.

Figure 2.

(A) Fraction of patients remaining on waiting list. (B) Fraction of patients alive on waiting list. (C) Fraction of transplanted patients alive after transplantation. (D) Fraction of patients alive from time of listing (overall survival).

Table 4. Covariate-adjusted Cox models for patient and graft survival
ModelRR HPS95% CIp
  1. 1Relative risk of HPS policy compared to control patients on the list getting a transplant.

  2. 2Relative risk of HPS policy compared to control patients dying on the waiting list.

  3. 3Relative risk of HPS policy compared to control patients dying after transplant.

  4. 4Relative risk of HPS policy compared to control patient graft failure posttransplant.

  5. 5Relative risk of HPS policy compared to control patients dying at any time from initial listing till end of study, including pre-, peri and posttransplant deaths.

  6. RR = relative risk; CI = confidence interval; HPS = HPS policy patients; Lab MELD = calculated, laboratory MELD without exception points at removal from list or censoring; Final MELD = MELD incorporating exception points at removal from list or censoring.

PretransplantTransplant1Unadjusted2.5272.223–2.870<0.0001
Lab MELD3.3512.945–3.810<0.0001
Final MELD1.2601.107–1.4300.0005
Death: waiting list2Unadjusted0.1580.0592–0.42000.0002
Lab MELD0.2850.107–0.757<0.0001
Final MELD0.1340.0502–0.3560<0.0001
PosttransplantPost-TP death3Unadjusted0.8270.587–1.1700.28
Adjusted0.9780.684–1.4000.9
Graft failure4Unadjusted0.7230.520–1.0000.05
Adjusted0.7890.561–1.1100.17
OverallDeath: any time5Unadjusted0.5140.374–0.7070.00004
Lab MELD0.8070.587–1.1100.190

Analyses of the data after exclusion of patients with pulmonary embolus or COPD (as these patients did not actually meet policy criteria) did not meaningfully affect the results (unadjusted RR of death on waiting list after exclusion of these patients was 0.187, CI: 0.07–0.498, p = 0.0008). The RR for waitlist mortality adjusted for laboratory MELD was 0.134 (CI: 0.05–0.35, p < 0.0001).

Clearly, the chance of death on the waiting list was not equalized by granting exception points to the HPS policy patients, as was intended by the HPS policy. Indeed, the K-M analysis showed that HPS policy patients had a significantly reduced risk of death while awaiting transplant compared with the controls (Figure 2B).

Cox analysis showed that unadjusted risk of death on the waiting list in HPS policy patients was about 16% of the risk of death among controls (Table 4). Differences remained highly significant even after adjusting for the laboratory and final (including exception points) MELDs. Thus, contrary to the assumptions of the HPS policy, patients listed with this exception actually had a lower (uncorrected) risk of death at the time of listing, which remained even after correction of their laboratory MELD score. Granting the exception points accentuates, rather than reduces, this difference in the risk of death in HPS policy patients versus controls. By design, pretransplant factors, other than MELD score, were not accounted for in these Cox models. As a sensitivity analysis, we examined whether regional differences in the fraction of patients awarded HPS policy points affected our conclusions. The results of this analysis did not meaningfully change our conclusions (results not shown). There was also no difference in waiting list mortality when analyzed by region (p = 0.66, data not shown).

Posttransplant outcomes

In the period under study, 236 HPS policy and 14 719 control patients were transplanted. There were no differences between transplanted patients in the HPS policy group and controls with regard to blood group match, presence of angina, diabetes, donor race, donor stroke, donor gender, recipient educational level, muscle wasting, peripheral vascular disease, portal hypertensive bleeding, previous abdominal surgery, prior malignancy or payer type. There were no significant differences in proportion of HPS policy and control patients done by year.

Analysis of factors found to differ significantly between the HPS policy group and control group are given in Table 5. These differences largely reflect that the HPS policy group had less severe liver disease than the control group. Thus, HPS policy patients had significantly lower laboratory MELD at the time of transplant, and included fewer status 1 recipients. Fewer had ascites, peritonitis, variceal bleeds, encephalopathy, received high-risk liver transplants (multiorgan transplants deceased after cardiac death [DCD] donors, split livers) or were in ICU, hospital or were disabled prior to their transplant. The ready access to organs in the HPS policy group was reflected in smaller number of HPS policy patients receiving living donor organs, and fewer regionally or nationally shared DCD donor organs. In the transplanted group, there was also a higher proportion of HPS policy patients who had COPD or history of pulmonary embolus than that in the control group (Table 3), suggesting some inaccuracy in the diagnosis of HPS in HPS policy patients.

Table 5. Analysis of baseline characteristics in transplanted patients
Parameter HPS, Mean (SD) (n = 236)Control, Mean (SD) (n = 14 719)p
  1. N = number; SD = standard deviation; HPS = HPS policy group; Lab MELD = laboratory (calculated) MELD at transplant; TIPS = transjugular intrahepatic portal systemic shunt; COPD = chronic obstructive pulmonary disease; CVA = cerebrovascular event; High risk = multiorgan transplant, split liver, donation after cardiac death donor, status 1 recipients.

GenderMale136 (57.63)9926 (67.44)0.0014
Female100 (42.37)4793 (32.56) 
Age 52.4 (9.53)52.18 (9.64)0.69
Initial MELD 13.51 (3.82)17.12 (8.2)<0.0001
MELD at transplant 25.18 (3.94)22.77 (8.65)<0.0001
Lab MELD at transplant 14.16 (3.98)22.59 (9.10)<0.0001
Recipient raceAsian4 (1.69)381 (2.59)0.0010
Black5 (2.12)1244 (8.45) 
Hispanic40 (16.95)1695 (11.52) 
Multiracial1 (0.42)136 (0.92) 
White186 (78.81)11 262 
 (76.52) 
Creatinine at transplant 0.87 (0.31)1.65 (1.41)<0.0001
AscitesMinimal67 (28.39)2420 (16.18)<0.0001
Moderate139 (58.90)7777 (52.84) 
Severe30 (12.71)4522 (30.72) 
Bacterial peritonitis 7 (3.23)1172 (6.63)0.0132
EncephalopathyNone98 (41.53)4042 (27.46)<0.0001
Stage 1–2135 (57.2)8979 (61.0) 
Stage 3–43 (1.27)1698 (11.54) 
Portal vein thrombosis 6 (2.74)535 (3.87)0.0497
TIPS 12 (5.48)1259 (9.10)0.0479
COPD 18 (7.69)205 (1.40)<0.0001
Pulmonary embolus 3 (1.28)35 (0.24)0.0072
Medical condition at transplantICU7 (3.20)1368 (9.88)<0.0001
Hospital19 (8.68)2560 (18.49) 
Out193 (88.13)9917 (71.63) 
Functional state at transplantNo limitation6616 (60.84)111 (61.67)0.0029
Some restriction2658 (24.44)57 (31.67) 
Disabled1601 (14.72)12 (6.67) 
High risk 12 (5.08)2523 (17.14)0.0001
Recipient CVA 5 (2.14)81 (0.55)0.0061
Donor typeDeceased donor218 (99.09)13 106 (94.05)0.0016
Living2 (0.91)829 (5.95) 
ShareLocal162 (73.64)10 100 (72.48)0.0063
Regional53 (24.09)2774 (19.91) 
National5 (2.27)1061 (7.61) 
Malignancy since listing 1 (0.46)233 (1.68)0.0012
Graft failure 36 (15.25)2996 (20.35)0.0532
Death since transplant 33 (13.98)2449 (16.64)0.2767
List time 199.53 (278.0)362.02 (439.0)<0.009

Survival analysis: posttransplant mortality and graft failure

K-M analysis revealed that posttransplant survival (Figure 2C) (p = 0.28) was similar in HPS policy and control patients; there was marginally better graft survival in the HPS policy group (p = 0.05, figure not shown). Cox analysis (Table 4) showed that the RR for death in the HPS policy group was similar, both when unadjusted and when adjusted, for all factors associated with increased risk, including accounting for high-risk surgery. Similarly, graft failure was similar in the HPS policy group and controls whether or not adjusted for other risk factors. Because high-risk transplants occurred almost exclusively among the control patients, analyses were repeated after exclusion of high-risk patients (status 1, multiorgan, split and DCD liver transplants). There were no meaningful differences from the main analysis shown (data not shown).

Analysis of the posttransplant death after exclusion of patients with pulmonary embolus or COPD did not meaningfully affect the results (unadjusted RR of death following transplant after exclusion of these patients was 0.773, CI: 0.525–1.140, p = 0.19). The RR for posttransplant mortality adjusted for all factors known to be associated with increased risk was 0.978 (CI: 0.68–1.40, p = 0.9). There was no difference in posttransplant survival by region (p = 0.18).

Overall patient survival

Ultimately, the most important outcome measure of equity between HPS policy and other patients is the overall survival from the time of listing onward, incorporating survival after transplantation or other removal from the waiting list. Overall mortality in the HPS policy group for the entire study period (including waiting list and peri- and posttransplant) was approximately half that of the control group (14.90% vs. 25.69%, p < 0.0001) (Table 2). This is confirmed by the K-M analysis of patient survival (Figure 2D), which was significantly better in HPS policy patients compared to that in controls (p < 0.0001). Cox analysis of overall mortality risk in HPS policy and control patients is given in Table 4. The unadjusted RR of death overall was about half that of controls (RR: 0.514, CI: 0.374–0.707, p = 0.000042). After adjusting for the listing laboratory MELD score, the mortality risk in the HPS policy patients was still numerically lower compared to that in control patients (RR: 0.807, CI: 0.587–1.110, p = 0.190), but the difference was no longer statistically different. This indicates that the overall mortality risk of patients in the HPS policy group would not have been significantly different from controls if they had been listed with their laboratory MELD score, and had not received additional exception points in their final MELD score.

Discussion

HPS patients may have an increased mortality risk not accounted for by their MELD score (4,34). The intention of UNOS policy 3.6.4.5.1 is to provide exception MELD points to patients with HPS to compensate for this expected excess mortality, and make their survival equal to that of liver transplant candidates without HPS. Although the final rule explicitly states that the medically most urgent cases should be transplanted first, it has become clear that equity may include not only equal access to transplant (i.e. equal survival on the transplant list), but also equal outcomes posttransplant. For this reason, we incorporated both measures in our analysis. We did this study to establish if the policy was functioning as intended. Our analysis has shown that, contrary to the intention of the policy, overall death after listing (pre- and posttransplant) in HPS policy patients was only half that in controls. (Figure 2, Tables 2 and 5). The risk of death of HPS policy patients while awaiting transplant was only 15.8% of that in controls (Tables 4). Contrary to what was intended by the policy, HPS policy patients have been given a substantial survival advantage compared with the control patients.

This unexpectedly good outcome for HPS policy patients may result from either the failure of the current HPS policy criteria to identify accurately patients with HPS, or the failure to characterize the severity of their condition. Consistent with first of these possibilities, significantly more patients in the HPS policy group had a history of pulmonary embolus or COPD (Table 3), which suggests that some patients may have been misclassified. Consistent with the second of these possibilities, our analysis shows that, after adjusting for liver disease severity (as assessed by laboratory MELD), HPS policy patients have an overall survival similar to that in control patients (RR: 0.827, CI: 0.587–1.170, p = 0.19, Table 4). That is, HPS policy patients are predicted to have survival similar to that in control patients had they received no exception points, but were listed with their laboratory MELD score. In fact, univariate analysis of all available pretransplant variables showed that HPS policy patients had less severe liver disease than the control patients (Table 2). The less advanced liver disease of the HPS policy patients is reflected in their significantly lower calculated (laboratory) MELD scores at transplant (14.16 ± 3.98 vs. 22.59 ± 9.10, p < 0.0001). No factor other than the milder severity of the liver disease could be identified that accounted for the improved survival of HPS policy patients on the waiting list.

These data would not necessarily negate the justice of providing exception points to HPS patients if (analogous to patients with hepatocellular cancer) their posttransplant survival would be adversely affected by a delay in receiving a liver transplant. While it is possible that early transplantation protected HPS patients from worse posttransplant outcomes, the UNOS data do not support a worse posttransplant outcome in HPS policy patients than that in controls. K-M analysis showed no difference in patient survival after transplant. Cox modeling, to adjust for all factors known to affect posttransplant survival, showed that posttransplant patient and graft survival was similar in HPS policy patients and nonexception control patients. Thus, there is no data-driven evidence to show that they would have done worse if they were transplanted with their laboratory MELD.

That posttransplant survival was similar in HPS and control patients was surprising, given the prior publications showing worse posttransplant survival in HPS patients (2,8–13). There are several possible explanations for this. Since Krowka et al. (2,10) and Arguedas et al. (8) showed that PaO2 of <50 mmHg or high MAA shunt fraction, or both, predicted a poor posttransplant outcome, it is possible that patients with these poor prognostic markers were not referred for transplant. Indeed, Martinez-Palli et al. stated in a recent review that when these high-risk factors are present, they exclude the patient from orthotopic liver transplantation (OLT) (35). It is unknown how widespread this practice is, and it is impossible to assess this from the data available through UNOS. However, in a recent study of 40 HPS patients of which eight (20%) were denied transplant, denial was for nonpulmonary reasons in all cases (11), although the presence of severe hypoxemia (which was slightly, but statistically, significantly worse in excluded patients) may have influenced the decision not to transplant. We have no way of assessing whether the most advanced HPS patients were excluded without ever being referred to the UNOS RRB, which represents a potential fault in the system.

The similar posttransplant outcomes in HPS patients and controls could also possibly be explained by improved medical and surgical management in transplanting HPS patients in the MELD era. There are several recent reports of similar posttransplant outcomes in HPS patients and controls (14–16). The excellent posttransplant outcomes for HPS is a very important finding that emphasizes that these patients should not be excluded from transplant because of concerns that transplant may be futile.

Ultimately, the survival advantage among the HPS policy patients occurred predominantly because of earlier transplantation due to extra HPS policy exception points. In the pretransplant period, HPS policy patients were more than 2.5 times more likely than controls to get a transplant. After adjusting for the exception points they had received, their likelihood of getting a transplant was still 26% higher than that of controls. As a consequence of the earlier transplantation, their risk of death while on the waiting list was strikingly reduced to only 15.8% of that of the control patients.

There were marked regional variations (Table 3) in the proportion of patients transplanted under policy 3.6.4.5.1, suggesting that there are regional differences in the rate of diagnosis of HPS or the application for exception points.

This, and the significantly higher proportion of HPS policy patients with pulmonary embolus and COPD, underscores the difficulty and variability in diagnosing HPS. Hypoxemia, the key criterion of policy 3.6.4.5.1, may also be affected by a variety of other parameters in cirrhotic patients, including impaired ventilation from tense ascites or hepatic hydrothorax (28) or intrinsic pulmonary diseases associated with specific liver disease (36). Up to 80% of nonhypoxemic cirrhotic patients may have nonspecific intrapulmonary shunting (29), and etiologies of hypoxemia other than pulmonary shunts may be present; hence, current policy criteria would give an incorrect diagnosis of HPS in some patients.

The wide regional variation in HPS policy patients transplanted suggests HPS patients are inconsistently identified despite there being a variety of simple screening methods to identify patients at risk (37–40). It is not possible to assess from the data available whether patients with severe HPS were not identified, but the fraction of HPS patients among controls would be expected to be small, and unlikely to meaningfully affect the outcomes of the analysis.

Although we have suggested that patients meeting UNOS criteria for HPS in this study had improved outcomes because they were incorrectly diagnosed or staged with HPS, we cannot rule out, based on our data, the possibility that prognosis is not worse in HPS. Natural history studies show that patients seldom die of nonhepatic causes (1–4), implying that the MELD score should accurately predict survival probability. Hopkins et al. reported significantly less waiting list mortality (12% in HPS patients vs. 36% in non-HPS group) in patients awaiting transplantation, which is similar to the results of this study (5). However, it is important to emphasize that this is a study of policy 3.6.4.5.1, and not of HPS patients per se, because the criteria specified by UNOS for HPS and the data collected are inadequate to accurately identify HPS patients, or to characterize the severity of HPS. The study group at the recent MESSAGE conference has highlighted these problems with current UNOS criteria for HPS, and has proposed potential solutions (27), including that more detailed data should be collected on pre- and post-liver transplant outcomes, that measures of HPS progression should be incorporated and that granting of exception points should be made contingent on the provision of data. These measures would facilitate in optimizing outcomes in patients with HPS, and avoid biasing allocation to or away from them (27).

In summary, patients receiving exception MELD points under policy 3.6.4.5.1 criteria have a substantial survival advantage on the waiting list, have similar posttransplant survival and substantially improved overall survival when compared to nonexception control patients. Hence, policy 3.6.4.5.1 is not achieving its intent. Our study raises important questions about the accuracy of the diagnosis of HPS in patients receiving exception points, and the need for more standardized criteria of severity and progression of HPS. It has raised the question of whether the good outcomes posttransplant are the result of the early transplantation, or because of improvements in management. It also raises questions about the liberal provision of exception points under the policy, screening and standardization of workup for HPS. Improved criteria are clearly needed to diagnose HPS and characterize its severity. Better tracking of data on HPS patients would help to improve equity in allocation, and better define the natural history of HPS.

Ancillary