Liver transplantation for children with biliary atresia in the pediatric end-stage liver disease era: The role of insurance status

Authors


  • See Editorial on Page 470

  • This work was presented in part at the 2012 American Transplant Conference in Boston, MA.

Address reprint requests to Ronen Arnon, M.D., Department of Pediatrics, Mount Sinai Medical Center, 1 Gustave L. Levy Place, New York, NY 11029. E-mail: ronen.arnon@mountsinai.org

Abstract

Socioeconomic status influences health outcomes, although its impact on liver transplantation (LT) in children with biliary atresia (BA) is unknown. We hypothesized that governmental insurance [public insurance (PU)], rather than private insurance (PR), would be associated with poorer outcomes for children with BA. Children with BA who underwent first isolated LT between January 2003 and June 2011 were identified from United Network for Organ Sharing Standard Transplant Analysis and Research files. We identified 757 patients with PR and 761 patients with PU. The race/ethnicity distribution was significantly different between the groups (65% white, 12% black, and 10% Hispanic in the PR group and 33% white, 26% black, and 29% Hispanic in the PU group, P < 0.01). Wait-list mortality was higher for the PU group versus the PR group [46/1654 (2.7%) versus 29/1895 (1.5%), P < 0.01]. PR patients were older than PU patients at transplant (2.4 ± 4.5 versus 1.5 ± 3.0 years, P < 0.01). The donor types differed between the groups: 165 children (21.8%) in the PR group received living donor grafts, whereas 79 children (10.4%) in the PU group did (P < 0.01). The 1- and 5-year posttransplant patient survival rates were greater for the PR group versus the PU group (98.0% versus 94.1% at 1 year, P < 0.01; 97.8% versus 92.2% at 5 years, P < 0.01). Cox proportional hazards models revealed that the insurance type (PU), the donor type (deceased), and life support were significant risk factors for death. A separate analysis of deceased donor LT revealed that the PU group still had significantly worse patient and graft survival. In conclusion, PU coverage is an independent risk factor for significantly increased wait-list and posttransplant mortality in children with BA. Further studies are needed to unearth the reasons for these important differences in outcomes. Liver Transpl 19:543–550, 2013. © 2013 AASLD.

Abbreviations
BA

biliary atresia

BMI

body mass index

FILT

first isolated liver transplantation

LT

liver transplantation

MELD

Model for End-Stage Liver Disease

PELD

Pediatric End-Stage Liver Disease

PR

private insurance

PU

public insurance

SES

socioeconomic status

SPLIT

Studies of Pediatric Liver Transplantation

UNOS

United Network for Organ Sharing.

Biliary atresia (BA) is the most common indication for liver transplantation (LT) in children. Analyses of risk factors for patient death and graft loss have been previously summarized.[1] Infants are at greater risk for death than older children according to both US Studies of Pediatric Liver Transplantation (SPLIT) and United Network for Organ Sharing (UNOS) reports.[2, 3] Factors linked to graft loss include immunological factors such as the type of calcineurin inhibition used and the impact of rejection.[2, 4] Technical variant grafts were associated with graft loss in the SPLIT study but were not associated with increased patient mortality.[2] Factors that contribute to inequitable access to the transplantation network (UNOS/Organ Procurement and Transplantation Network) for adults with liver disease that might influence outcomes after LT include socioeconomic status (SES), geographic location, and delayed referral.[5, 6] Many studies have found a relationship between lower SES (as defined by income or insurance) and the survival of adults with malignancies or other chronic diseases.[7-10] Yoo and Thuluvath[11] studied the effects of SES on posttransplant survival in adult liver recipients. They found that patients with Medicare and Medicaid had lower survival rates than patients with private insurance (PR). They suggested that an inability to obtain medication was a factor in the poor survival of patients receiving governmental insurance and recommended further examination in prospective studies. We hypothesized that governmental insurance [public insurance (PU)], rather than PR, is associated with poorer outcomes after first isolated liver transplantation (FILT) for children with BA in the Pediatric End-Stage Liver Disease (PELD) era.

PATIENTS AND METHODS

BA was chosen for this analysis because it is the most common isolated indication for LT in children and is a relatively uniform disease. Patients with BA who were younger than 19 years at the time of transplantation were identified from the UNOS Standard Transplant Analysis and Research files for LT recipients between January 2003 and June 2011. Data from 2003 onward were used because this time period captures the Model for End-Stage Liver Disease (MELD)/PELD era. The following variables were examined.

Recipients

The following recipient data were collected: age, sex, race/ethnicity, body mass index (BMI), bilirubin, creatinine, albumin, dialysis or life support status, peritonitis, calculated PELD/MELD score at transplant, and submission/approval of PELD/MELD scoring exceptions.

Donor and Transplant Characteristics

Data on donors and surgical procedures, including the donor age, sex, race, ethnicity, and BMI, the graft type, and the transplanted graft cold and warm ischemia times, were analyzed.

Insurance Status

In accordance with UNOS categories, the type of insurance was categorized as PR or PU. PR is defined by UNOS as the funds of agencies or any worker's compensation covered by a private insurer. PU refers to state Medicaid funds, the Children's Health Insurance Program, or funds from another governmental agency.

Race/Ethnicity

UNOS lists the following race/ethnicity options:

  • White: European descent, Arabic, Middle Eastern, or North African (not black).
  • Black or African American: African descent [eg, African American, African (continental), West Indian, or Haitian].
  • Asian: Asian descent (eg, Chinese, Filipino, Japanese, or Korean).
  • Hispanic/Latino: Central or South American descent (eg, Mexican, Puerto Rican, or Cuban).
  • Other: American Indian or Alaska native, native Hawaiian, or other Pacific Islander.

Transplant Center Activity

In order to address a possible center effect on outcomes, we divided the centers according to the volume of transplants that they performed during the study period. We defined a low-volume pediatric transplant center as a center that performed up to 5 pediatric LT procedures per year (average during the study period) and a medium- to high-volume center as a center that performed an average of 6 or more per year. These data were provided by direct communication with a UNOS statistician and were not part of the original Standard Transplant Analysis and Research file.

Available data were 100% complete except for the BMI values and the non–heart-beating status of donors. For recipients, BMI data were 90% complete because of missing values as well as invalid entries, and data on the non–heart-beating status of donors were missing for 20% of the patients. The circumstances regarding the BMI data were also verified with UNOS staff via personal correspondence with the study statistician. The primary outcome measures were graft loss and death. Data on waiting-list deaths of patients with BA according to the insurance type were obtained by direct contact with UNOS representatives.

Statistical Analysis

All statistical analyses were performed with IBM SPSS Statistics version 19th. A 2-tailed P value (α level) of 0.05 or less was chosen as the level of statistical significance. Categorical variables were compared with χ2 analysis. Independent sample t tests were used to compare outcomes for continuous measures. Kaplan-Meier analysis was used to estimate posttransplant patient and graft survival. Risk factors for death and graft loss were analyzed with Cox proportional hazards models.

RESULTS

Characteristics of the Sample

During the 8.5-year period from January 2003 to June 2011, 1579 children underwent FILT for BA. The mean follow-up time was 3.51 ± 2.76 years. Seven hundred fifty-eight patients (48%) were white, 335 (21.2%) were Hispanic, 302 (19.1%) were black, 111 (7%) were Asian, and 73 (4.6%) were other ethnicities. Seven hundred fifty-seven of the 1579 patients (47.9%) had PR, and 761 (48.2%) had PU. The rest (59 patients or 3.9%) had foreign insurance or another type or self-paid and were excluded from the analysis.

Whites and Asians had significantly higher PR rates than blacks and Hispanics (P < 0.05). There was no statistical difference between blacks and Hispanics in the distribution of insurance types (Table 1).The mean PELD/MELD scores at listing were similar for whites (12.1 ± 10.2), blacks (14.3 ± 9.2), and Hispanics (12.8 ± 11.4). The mean PELD/MELD scores for PR (12.6 ± 12.6) and PU (13.0 ± 9.9) were also similar at listing.

Table 1. Race/Ethnicity and Insurance Type
 PR (n = 757)PU (n = 761)Other (n = 61)
  1. NOTE: The data are presented as numbers and percentages.

White (n = 758)493 (65.1)252 (33.1)13 (21.3)
Black (n = 302)93 (12.3)199 (26.1)10 (16.4)
Hispanic (n = 335)78 (10.3)226 (29.7)31 (50.8)
Asian (n = 111)69 (9.1)36 (4.7)6 (9.8)
Other (n = 73)24 (3.1)48 (6.3)1 (1.6)

The characteristics of patients with BA who underwent FILT are presented by the insurance type in Table 2. Patients with BA who had PR at transplant were significantly older than patients with PU. There were no significant differences in the calculated PELD/MELD scores at transplant, in exceptions submitted or approved, or in the waiting times until transplantation between the PU and PR groups. There was no difference in the prevalence of prior abdominal surgery between the PR and PU groups, although the UNOS database does not specify the type of surgery (ie, the Kasai operation or another surgical procedure). The cold ischemia time was significantly shorter in the PR group, their donors were older, and 21.8% of them underwent living donor LT, whereas 10.4% of the patients in the PU group did (P < 0.01).

Table 2. Characteristics According to the Insurance Type
 PR (n = 757)PU (n = 761)P Value
  1. a

    The data are presented as means and standard deviations.

  2. b

    Data were missing for 10%.

  3. c

    The data are presented as medians and 25th and 75th percentiles.

Recipient characteristics at transplant   
Age (years)a2.4 ± 4.51.5 ± 3.0<0.01
Female sex (%)55.762.5<0.01
BMI z scoreb0.12 ± 1.04−0.00 ± 0.980.76
Albumin (g/dL)3.0 ± 0.43.0 ± 0.80.10
Total bilirubin (mg/dL)11.7 ± 10.212.0 ± 10.00.67
International normalized ratio1.70 ± 1.601.74 ± 3.300.95
Prior abdominal surgery (%)79.778.30.52
PELD/MELD scorec20 (12–29)21 (12–30)0.28
PELD score > 20 (%)52.454.40.47
Status 1 (%)12.511.40.50
Exception submitted (%)32.731.80.70
Exception approved (%)90.793.40.28
Time between listing and transplant (days)a194.5 ± 425.0165.0 ± 305.60.12
Transplant characteristics   
Cold ischemia time (hours)a6.7 ± 4.57.4 ± 4.1<0.01
Procedure type (%)  <0.01
Cadaveric: whole50.353.6 
Cadaveric: technical variant27.936.0 
Living21.810.4 
Donor characteristics   
Age (years)a14.4 ± 13.910.9 ± 11.7<0.01
BMI z scorea0.08 ± 1.03−0.11 ± 0.96<0.01

Primary Outcomes

Death on the Waiting List (2003–2011)

Data on deaths on the waiting list for patients with BA between 2003 and 2011 were analyzed by the insurance type (PR/PU). χ2 without Yates' correction revealed a significant difference in the proportions who died on the wait list in the PR [29/1895 (1.5%)] and PU groups [46/1654 (2.7%), P < 0.01, odds ratio = 0.54, 95% confidence interval = 0.34–0.87]. The UNOS database does not contain information on the causes of death on the waiting list.

The outcomes after LT by the insurance type are presented in Table 3. The 1- and 5-year patient and graft survival rates were significantly better for the PR group versus the PU group. Figure 1A,B shows a Kaplan-Meier analysis of patient and graft survival. Inferior patient and graft survival rates were evident for patients with PU as early as 30 days. Five years after transplantation, there were further decreases in patient and graft survival (a widening of the gaps between the survival curves; Fig. 1A,B) for patients with PU.

Figure 1.

Kaplan-Meier analysis of (A) patient survival and (B) graft survival by the insurance type.

Table 3. Outcomes After LT by Insurance Type
 PRPUP Value
Graft survival (%)   
30 days92.388.2<0.01
1 year91.786.5<0.01
5 years90.282.7<0.01
Patient survival (%)   
30 days98.795.8<0.01
1 year98.094.1<0.01
5 years97.892.2<0.01

Causes of graft loss by the insurance type are shown in Fig. 2. Vascular thrombosis (hepatic artery thrombosis or portal vein thrombosis) and primary nonfunction were the main etiologies of graft loss, and there was no statistical difference between the PR and PU groups. Patients with PU had a higher rate of graft loss due to chronic rejection than patients with PR, although the difference did not quite reach statistical significance (P = 0.05). There were no differences in 1- and 5-year patient and graft survival rates by race/ethnicity.

Figure 2.

Causes of graft loss by insurance type.

Transplant Center Activity

In order to analyze the potential effects of transplant centers on LT outcomes, we preliminarily assessed the utilization of low-volume centers, which we defined as centers with up to 5 cases per year. The proportions of children with PU and PR who underwent transplantation at low-volume centers were similar (425/759 versus 351/655, χ2 P > 0.05). Interestingly, the outcomes were worse for the PU group versus the PR group independently of this division of center transplant activity. (Data on transplantation by center volume were available for 86.5% of the patients with PR). Outcomes after LT by the insurance type and the transplant center volume are shown in Table 4.

Table 4. Outcomes After LT by Insurance Type and Transplant Center Volume
 PR/Low-Volume Center (n = 351)PU/Low-Volume Center (n = 425)PR/Medium- to High-Volume Center (n = 304)PU/Medium- to High-Volume Center (n = 334)
Graft survival (%)    
1 year90.086.391.488.6
5 years88.978.987.180.6
Patient survival (%)    
1 year97.393.297.995.6
5 years97.589.896.391.3

Multiple Listing

We explored whether PU and PR patients were listed in more than 1 region; we assumed that this might reflect an ability to travel to centers with more experience and better outcomes. During the study period, only 1% to 4% of the patients were listed in more than 1 region, and there was no significant difference between the PU and PR groups.

Living Donors and Outcomes

The PELD score at transplant was significantly lower for patients who underwent living donor LT versus patients who underwent deceased donor LT (17.7 ± 12.8 versus 20.8 ± 12.7, P < 0.01). To assess whether the higher percentage of living donors in the PR group was associated with the better posttransplant outcomes of these patients, we separately analyzed all patients who underwent deceased donor LT. This analysis revealed that children with PU still had significantly worse patient and graft survival than children with PR (Table 5).

Table 5. Outcomes After LT by Insurance Type: Deceased Donors Only
 PR (n = 592)PU (n = 682)P Value
Graft survival (%)   
30 days92.187.70.01
1 year91.285.8<0.01
5 years90.081.7<0.01
Patient survival (%)   
30 days98.695.9<0.01
1 year97.894.0<0.01
5 years97.592.1<0.01

Risk Factor Analyses

Risk factors for death were analyzed with Cox proportional hazards models (Table 6). The overall model was significant (χ2 = 53.61, P < 0.01). The insurance type (PU), life support at transplant, and donor type (deceased donor) were found to be risk factors for death among the patients who underwent LT. Risk factors for graft survival were analyzed with Cox analyses as well (Table 7). The overall model was significant (χ2 = 47.99, P < 0.01). The insurance type (PU), life support, and donor type (deceased donor) were found to be risk factors for graft loss among patients with BA who underwent LT. Next, these analyses were repeated for those cases who did have a numerical value for the PELD/MELD score (n = 1383). There were some changes to the results. For death, the model remained significant (χ2 = 67.12, P < 0.01). When the PELD/MELD score was included, the donor type no longer emerged as a significant predictor, but the PELD/MELD score did. For graft loss, the model was also still significant (χ2 = 45.44, P < 0.01). The predictors were unchanged in this model.

Table 6. Cox Proportional Hazards Model for Predicting Death After FILT for BA
Predictor VariableOdds RatioConfidence IntervalP Value
  1. NOTE: Significant predictors are bolded.

  2. a

    Race/ethnicity was overall a nonsignificant predictor; therefore, comparisons between groups with corresponding odds ratios and confidence intervals are not reported.

Recipient age (years)1.020.92–1.120.77
Recipient sex (female)0.810.48–1.380.44
Insurance type0.340.17–0.65<0.01
Race/ethnicitya0.75
Bilirubin at transplant (mg/dL)1.020.99–1.040.26
Creatinine at transplant (mg/dL)0.400.08–1.890.25
Albumin at transplant (mg/dL)1.100.77–1.580.60
International normalized ratio at transplant1.020.85–1.210.90
Life support0.300.13–0.720.01
Donor age (years)1.020.99–1.050.21
Donor sex0.870.52–1.450.59
Status 11.730.74–4.040.21
Cold ischemia time (hours)1.041.00–1.080.06
Donor type3.701.19–11.480.02
Transplant type0.710.34–1.440.34
Days between listing and transplant1.000.99–1.000.37
Table 7. Cox Proportional Hazards Model for Predicting Graft Loss after FILT for BA
Predictor VariableOdds RatioConfidence IntervalP Value
  1. NOTE: Significant predictors are bolded.

  2. a

    Race/ethnicity was overall a nonsignificant predictor; therefore, comparisons between groups with corresponding odds ratios and confidence intervals are not reported.

Recipient age (years)0.960.90–1.020.15
Recipient sex (female)0.850.62–1.170.32
Insurance type0.570.41–0.81<0.01
Race/ethnicitya0.46
Bilirubin at transplant (mg/dL)1.000.98–1.010.56
Creatinine at transplant (mg/dL)0.610.27–1.380.24
Albumin at transplant (mg/dL)1.090.88–1.340.43
International normalized ratio at transplant0.980.84–1.140.79
Life support0.350.20–0.61<0.01
Donor age (years)1.010.99–1.030.23
Donor sex0.960.71–1.310.82
Status 11.420.84–2.410.19
Cold ischemia time (hours)1.010.98–1.040.52
Donor type2.121.11–4.070.02
Transplant type1.130.72–1.780.59
Days between listing and transplant1.001.00–1.000.63

DISCUSSION

In this study, we have shown that PU is significantly associated with greater wait-list mortality and decreases in both early and late posttransplant patient and graft survival for pediatric LT recipients with BA in the United States. This analysis is focused on BA patients because BA is the largest single diagnostic category for pediatric LT and represents a relatively homogeneous disease process. The clinical heterogeneity of the diagnoses for the rest of the children who undergo LT may prevent a meaningful analysis of this nature for children overall. Wait-list mortality is likely related to the medical management of complications of BA, including infections and sequelae of end-stage liver disease, and to access to both a center and a liver for transplantation. There are no UNOS data on the causes of death on the waiting list, so a detailed analysis of wait-list mortality is not feasible. The times on the wait list were similar for the PR and PU groups, although living donor transplantation was more frequent in the PR group. Rates of petitioning for and receiving exceptions were similar for the PR and PU groups, and this suggests that access to transplantation was not the cause of increased wait-list mortality.

Early posttransplant effects likely reflect peritransplant issues, as reported previously,[3] whereas late effects reflect issues likely related to ongoing follow-up after transplantation. It is notable that the differences in patient and graft survival increased with time, and this suggests that there are potential problems in both segments of posttransplant processes.

The insurance type is a proxy for SES; unfortunately, other SES measures such as family income and parents' education status are not collected in the UNOS database. There are a number of ways in which lower SES could affect posttransplant outcomes for children with BA; these include the degree of illness at transplant, the type of transplant available, access to medications and medical care, the type of center (a center of excellence versus other centers), and subsequent adherence to the posttransplant medical regimen. One possible explanation for the differences in outcomes is that the children with PU were sicker at the time of transplant. A UNOS-based analysis of adults awaiting LT listing demonstrated that ethnic minorities and patients with Medicaid were more likely to have a MELD score greater than 20 at listing.[12] In this study, there were no significant differences in the PELD/MELD scores of patients with PR and patients with PU at listing and at transplant. Moreover, there was no significant difference in the number of patients who were listed as status 1 for LT, nor was there a difference in exceptions that were sought or granted. Technical issues related to transplantation could have contributed to worse outcomes in the PU group. In particular, the age/size at transplantation and the use of either technical variants or living donor transplants could have influenced outcomes. Children with PR were significantly older at transplant than children with PU. Although younger patients might have less favorable outcomes after LT because of technical complexities, a multivariate analysis and follow-up univariate tests did not reveal that the age at transplant was a risk factor for death and graft loss in our study.

Living donor LT was more common in the PR group. A higher proportion of patients undergoing living donor LT in the PR group may in part explain the significantly lower cold ischemia time and the older donor age. It is not clear why the rate of living donor transplantation was lower in the PU group. Potential explanations might include concerns about the financial consequences of serving as a donor (eg, coverage, copays, and lost income) and health issues of parents with low SES (eg, obesity and type 2 diabetes).[13] It is notable that a difference in outcomes after LT was still observed when the analysis was restricted to deceased donor transplants.

A strong consideration for a potential impact of SES on outcomes both before and after transplantation is adherence to the medical regimen. The relevance of this issue has been clearly demonstrated in adults. The finding of a nearly significant increase in the prevalence of chronic rejection related to graft loss in PU supports this potential explanation. Chronic rejection is one of the leading causes of graft loss occurring more than 1 year after LT.[14] Poor outcomes for patients with PU were described by Yoo and Thuluvath[11] in adult recipients after LT. An inability to obtain immunosuppressive medications and differences in health care access were proposed as explanations for the poorer outcomes of these patients. With other medical conditions, it has been shown that Medicaid patients are less likely to receive optimal care and are more likely to have poor outcomes for medical conditions (eg, asthma, cystic fibrosis, and myocardial infarction) in comparison with privately insured patients.[10, 15] The underuse of controller medications among Medicaid-insured children has been found to be widespread, and racial minorities and children whose parents are less educated are at higher risk for underuse.[15]

We explored a possible center effect on the outcomes of the different study groups. This is an exceptionally complex and controversial issue.[16, 17] Tracy et al.[18] studied the association between center volume and outcomes in pediatric LT. They found a significant center volume/outcome relationship among pediatric LT centers but not among freestanding children's hospitals or children's hospitals within adult hospitals. Low-volume centers that were not children's hospitals or children's hospitals within adult hospitals had the least favorable aggregate 1-year observed-to-expected patient death ratios. Our initial UNOS query did not address this issue for this cohort, and a full analysis of a center effect is beyond the scope of this investigation. We proposed that centers with lower volumes were ones at which transplants occurred on average less than every 2 months (ie, <6 cases per year). The proportions of transplants occurring in lower volume centers were the same for the PU and PR groups. Interestingly, poorer outcomes with PU persisted independently of transplantation at a lower volume center.

In comparison with relatively small single-center studies, the use of the UNOS database has provided data on the SES and outcomes of the largest number of pediatric patients with BA who required LT. Notwithstanding this, one of the limitations of the UNOS transplant database is that the information is acquired from transplant centers with the objective of ensuring the proper allocation of organs, so some data elements that are key to this analysis are not collected. Although the UNOS database collects certain covariates such as the income of recipients and their education (parameters that are important in analyzing SES), there are no data when the recipient is a child. Therefore, we could not appropriately analyze the SES of each recipient's family. Variables such as medications at transplant, adherence, and types of vascular complications (hepatic artery thrombosis or portal vein thrombosis) are not captured in the UNOS database. It is hoped that such parameters will be collected in ongoing pediatric studies such as SPLIT and the Childhood Liver Disease Research and Education Network. Another limitation of this analysis of the UNOS database is related to the lack of data on the patients before transplantation, including the age at the diagnosis of BA, the age at the time of the Kasai operation, and details of wait-list morbidity and mortality.

In summary, children with BA who had PU had significantly worse pretransplant and posttransplant survival than children with PR. The insurance status was an independent and strong risk factor related to survival. The current analysis does not permit a detailed understanding of the mechanism by which PU increases mortality risk. It does not appear to be simply related to the degree of illness at transplant. Future studies of transplant outcomes need to account for the insurance status. A detailed investigation of the underlying causes of increased mortality in those with PU is needed in order to try to prevent future deaths among these children.

ACKNOWLEDGMENT

The authors thank UNOS for graciously allowing access to its database and Mr. Or Arnon for his technical support.

Ancillary