| |
Hepatitis C Virus Seropositivity in Organ Donors Reduced Survival in Heart Transplant Recipients
|
| |
| |
JAMA. Oct 18, 2006;296:1843-1850.
Leanne B. Gasink, MD, MSCE; Emily A. Blumberg, MD; A. Russell Localio, PhD; Shashank S. Desai, MD; Ajay K. Israni, MD, MS; Ebbing Lautenbach, MD, MPH
Division of Infectious Diseases of the Department of Medicine (Drs Gasink, Blumberg, and Lautenbach), Department of Biostatistics and Epidemiology (Drs Localio and Lautenbach), and Center for Clinical Epidemiology and Biostatistics (Drs Gasink, Localio, and Lautenbach), University of Pennsylvania School of Medicine, Philadelphia; Heart Failure and Transplant Program, Inova Fairfax Hospital Department of Medicine, Falls Church, Va (Dr Desai); and Hennepin County Medical Center and Department of Epidemiology and Community Health, University of Minnesota, Minneapolis (Dr Israni).
".....our study demonstrates a survival disadvantage among heart transplant recipients with HCV-positive donor hearts. This association appears to be independent of the recipient's HCV status and recipient age. Preferential allocation of HCV-positive donors to older cardiac transplant candidates is not appropriate.... A shortage of cardiac organ donors results in a substantial number of deaths among persons awaiting cardiac transplantation.1 One potential approach to increasing the availability of donors is to broaden the criteria used to identify appropriate donors. For example, the cardiac donor pool could be expanded by using donors with hepatitis C virus (HCV) infection...."
ABSTRACT
Context Although liberalization of donor criteria could expand the donor pool, the use of certain "marginal donors," such as those who are hepatitis C virus (HCV) positive, is controversial. Little is known about the effect of donor HCV positivity on survival in cardiac transplantation.
Objectives To examine the association between donor HCV positivity and survival among heart transplant recipients and to determine the effects of recipient age and recipient HCV status on this association.
Design, Setting, and Participants A multicenter cohort study was performed using the US Scientific Registry of Transplant Recipients. Adult heart transplant patients who received their transplants between April 1, 1994, and July 31, 2003, were eligible for inclusion.
Main Outcome Measure All-cause mortality.
Results
Of 10 915 patients meeting entry criteria, 261 received an HCV-positive donor heart. Mortality was higher among recipients of HCV-positive donor hearts at 1 year (16.9% vs 8.2%; P<.001), 5 years (41.8% vs 18.5%; P<.001), and 10 years (50.6% vs 24.3%; P<.001).
Using Kaplan-Meier methods, 1-, 5-, and 10-year survival rates were 83%, 53%, and 25%, and 92%, 77%, and 53% for recipients of HCV-positive and HCV-negative donor hearts, respectively (P<.001, log-rank test).
Recipients of HCV-positive donor hearts were more likely to die of liver disease and coronary vasculopathy.
After propensity matching, the overall hazard ratio (HR) associated with receipt of an HCV-positive donor heart was 2.10 (95% confidence interval [CI], 1.60-2.75).
Stratified analyses showed that HRs did not vary by recipient HCV status or by recipient age (for recipients aged 18-39 years: HR, 1.75 [95% CI, 0.70-4.40]; for recipients aged 40-59 years: HR, 2.23 [95% CI, 1.42-3.52]; and for recipients aged 60 years and older: HR, 2.07 [95% CI, 1.32-3.27]; overall P value for interaction, >.10).
Conclusions Receipt of a heart from an HCV-positive donor is associated with decreased survival in heart transplant recipients. This association appears to be independent of recipient HCV status and age. Preferential allocation of HCV-positive donors to HCV-positive recipients and/or older recipients is not warranted.
COMMENT
We found that receipt of a heart from an HCV-positive donor is associated with increased mortality and that this risk is present early after transplantation. Neither recipient HCV status nor recipient age altered this relationship.
Several previous studies have reported no difference in survival among persons who received HCV-positive donor hearts compared with those who received HCV-negative donor hearts.6-7 Recent guidelines from the American Society of Transplantation stated that the use of HCV-positive donor grafts "does not appear to be a major influence on graft and patient survival."3 However, mortality among 34 recipients of HCV-positive donor hearts was 2.8-fold (95% CI, 1.3-5.9) greater than controls at the Cleveland Clinic.8 Our study, the largest and only multicenter comparative study to address potential confounding, provides stronger evidence that donor HCV antibody positivity confers a survival disadvantage. Furthermore, the nearly 35-fold higher percentage of exposed vs unexposed patients whose deaths were caused by liver disease suggests that this survival disadvantage is a direct result of HCV infection. However, non-hepatic-related mortality is also possible. Hepatitis C virus seropositivity has been associated with coronary artery plaque, and donor HCV positivity has been linked to accelerated coronary vasculopathy in heart transplant recipients.8, 20
Contrary to previous speculation, we did not find that older recipient age mitigated the effect of an HCV-positive donor on recipient survival.5 After stratifying on category of recipient age, HCV donor positivity was significantly associated with increased mortality only in recipients older than 39 years. Whether young recipients are resistant to the adverse effects of HCV-positive donors is unclear. Interaction by age was not present statistically in our main model, but was at most marginally present in subanalyses using only patients with complete data. That liver disease and viral hepatitis were frequently implicated as a cause of death in older recipients with HCV-positive donor hearts, but not seen among exposed recipients younger than 40 years, suggests that younger age offers some protection against the unfavorable effects of exposure. However, fewer numbers of exposed individuals between 18 and 39 years old (n = 28) could result in inadequate power to detect an association in that age group. Alternatively, the processes by which decisions are made to use HCV-positive donors in particular candidates may differ across age groups.
The significance of a recipient's HCV status remains unclear. Although a recipient's HCV status was neither a confounder nor an effect modifier of the association between donor HCV status and survival, a clear assessment is limited by the high percentage of recipients for whom recipient HCV status was unknown and the rarity with which HCV RNA confirmed chronic infection in HCV antibody-positive candidates. Prior comparative studies offer no additional insight, as they did not include HCV-positive recipients.7-8
Donor HCV antibody results were missing in only 1% of eligible patients, making the possibility of biases related to excluding patients with unknown exposure status unlikely. In the reclassification of these individuals as all exposed or all unexposed, associations with survival still existed. Notably, because donor HCV RNA measurements are not routinely performed, there is no way to verify chronic HCV infection in these individuals. A proportion of antibody-positive donors, therefore, were presumably not actively infected with HCV. This potential misclassification would be expected to bias our results toward the null. Thus, the true risk associated with using an HCV-infected donor heart may be even greater than estimated by our study.
Our study had several additional limitations. As in all nonrandomized studies, the possibility for unknown and unmeasured confounding exists. However, we used a propensity analysis to enable a more rigorous adjustment for confounding than would be possible with standard multivariable analysis.15 Like other studies using large databases, data accuracy cannot be directly validated and missing data can be problematic. These issues might be less of a concern in studies using the SRTR because participation is mandatory for all US transplant centers and a system of electronic edits is in place to minimize inappropriate data entries.21-22 Furthermore, outcomes were supplemented by the Social Security National Death Index and few data were missing on important variables such as donor HCV status and recipient age, although some other data had higher rates of missingness. Although we have included every heart transplant recipient in the United States with an HCV-positive donor heart during the study period, these results might not be generalizable to centers outside the United States where different HCV genotypes predominate, or to centers excluded from our study because they had not used an HCV-positive donor. Finally, we were not able to determine whether survival among patients who received a heart from an HCV-positive donor might well have been even worse had they not undergone transplantation at all.
In summary, our study demonstrates a survival disadvantage among heart transplant recipients with HCV-positive donor hearts. This association appears to be independent of the recipient's HCV status and recipient age. Preferential allocation of HCV-positive donors to older cardiac transplant candidates is not appropriate.
INTRODUCTION
A shortage of cardiac organ donors results in a substantial number of deaths among persons awaiting cardiac transplantation.1 One potential approach to increasing the availability of donors is to broaden the criteria used to identify appropriate donors. For example, the cardiac donor pool could be expanded by using donors with hepatitis C virus (HCV) infection.
Hearts from donors infected with HCV carry a substantial risk of transmission of HCV to the recipient, and high rates of subsequent liver enzyme abnormalities have been observed.2-3 Nonetheless, the effect of donor HCV antibody positivity on patient survival is unclear. The use of HCV donors for cardiac transplantation is controversial and varies.3-4 In 2001, an American Heart Association consensus conference report stated that HCV-positive donors "may be appropriate in selected higher-risk recipients."1 Others have advocated the preferential allocation of HCV-positive donors to older candidates.5
Few data exist regarding the effect of HCV donor positivity on survival in heart transplantation. Reports are contradictory and limited by small sample sizes, short follow-up, and difficulties in controlling for nonrandom organ allocation.5-8 We aimed to define the relationship between donor HCV status and survival in cardiac transplant recipients using a large cohort. We also sought to determine if this relationship was confounded or modified by 2 factors of interest: recipient HCV status and recipient age.
METHODS
We used the Scientific Registry of Transplant Recipients (SRTR),which includes data on all donors, wait-listed candidates, and transplant recipients in the United States, submitted by members of the Organ Procurement and Transplantation Network (OPTN), and has been described elsewhere.9 The Health Resources and Services Administration, US Department of Health and Human Services, provides oversight to activities of the OPTN and SRTR contractors. The study was approved by the University of Pennsylvania Institutional Review Board. The requirement for written consent was waived by the institutional review board.
Study Cohort
All heart transplant patients at least 18 years old who received their transplants between April 1, 1994, and July 31, 2003, were eligible and were followed up until August 1, 2004. This period coincides with the initiation of routine screening for donor HCV antibodies and ensures at least 1 year of follow-up for all patients. Donors were considered HCV antibody positive if the variables "donor HCV status" and/or "donor anti-HCV" were reported as "positive." Recipients receiving hearts from these individuals were defined as "exposed." Donors who were not HCV positive and with "negative" results recorded for either variable were considered HCV negative. Recipients receiving hearts from these individuals were defined as "unexposed." Recipients for which donor HCV status was unknown were excluded.
Patients who received transplants at centers that did not use at least 1 HCV-positive donor during the study period were excluded to prevent inclusion of controls that had no chance of receiving an HCV-positive donor heart (eg, center policy precluded use). Additional exclusion criteria included persons undergoing simultaneous, combined organ transplants and those dying, undergoing another cardiac transplant, or lost to follow-up within 30 days.
End Points
Patient follow-up was defined as time from transplantation until date of death or last known follow-up. Death date was determined using the date of recorded death reported by the OPTN and supplemented by data from the Social Security Death Master File. Because follow-up forms are due annually on the day after the transplant anniversary, deaths reported after the most recent date of expected follow-up were ignored. Biases related to counting extra time for patients who have follow-up forms turned in early were therefore avoided. Patients alive at maximum follow-up date or lost to follow-up were censored.
Data Collection
Recipient characteristics included age, sex, race, cytomegalovirus (CMV) antibody status, blood type, body mass index, location immediately prior to transplant (at home or hospitalized), days on waiting list, and cardiac diagnosis. Race determination was based on data submitted to the SRTR by individual transplant centers. The presence of diabetes, cerebrovascular disease, renal insufficiency (creatinine >1.5 mg/dL [132.6 ƒÊmol/L]), hepatitis B surface antigen status, and history of prior transplant were recorded. Life-sustaining measures collected included inotropes, intra-aortic balloon pump, ventilator, ventricular assist device, and extracorporeal membrane oxygenation. Recipient HCV infection was defined as positive if "recipient anti-HCV," "recipient RIBA [recombinant immunoblot assay]," or "recipient RNA" was positive. Recipients were defined as HCV negative if none of these was positive and at least one was reported as negative. Owing to changes in the United Network for Organ Sharing criteria during the study period, status was recorded as status 1 (status 1A and status 1B) or status 2.10
Donor variables included age, sex, race, CMV antibody status, blood type, and cause of death. Donor history included previous myocardial infarction, history of diabetes, history of hypertension, renal insufficiency, hepatitis B surface antigen status, tobacco use, and heavy alcohol use.
Peritransplant variables included transplant date, ischemic time (minutes), and human leukocyte antigen (HLA) mismatches. Encrypted center identification and total volume of adult heart transplants performed at each center during the study period were also recorded.
Statistical Analysis
The t test or Wilcoxon rank sum and 2 or Fisher exact test were used to compare continuous and categorical variables, respectively, between exposed (HCV-positive donor) and unexposed (HCV-negative donor) transplant recipients depending on the distribution of the data. Unadjusted patient survival was estimated using Kaplan-Meier methods. The assumption of proportional hazards was confirmed by graphical methods.
To address incomplete and missing data, multiple imputation was performed on the data to generate 5 data sets.11 Analyses were then performed for each data set. Results were combined to arrive at appropriate variances that accounted for missing data.12-13
Owing to the large number of variables with potential to confound the association between donor HCV status and survival, a propensity score was developed. Using logistic regression, a conditional probability of receiving an HCV-positive donor heart, given a set of covariates and without regard to survival, was assigned to each patient.14 Propensity scores have been shown to be less biased, more robust, and more precise when there are few outcomes relative to confounders.15-16 Recipient HCV status and recipient age were not used in the development of the propensity score so that their effects on the association between exposure and survival could be assessed in the main model. Each exposed patient was matched on the basis of propensity score to 4 unexposed patients using nearest neighbor matching as implemented in the program "nnmatch" in Stata version 9.1.17
Cox proportional hazards analysis was performed using the propensity-matched cohort, stratified by matched set. The presence of confounding was determined by a change in the effect size of the relationship between donor HCV status and survival by 15% or more.18 The presence of interaction was determined using the likelihood ratio test, and a P value of <.10 was considered significant. Analyses were also stratified by center to demonstrate that confounding by center was not present. Analyses including the entire cohort of eligible patients that used subclassification rather than matching were used to confirm our results.16, 19 In these models, multivariable Cox regression was performed and adjusted by quintile of propensity score. Covariates not well distributed between exposure groups within each quintile and other factors of interest (eg, recipient HCV status, recipient age) were examined further to identify confounding and interaction after adjustment for quintile of propensity score.
Subanalyses excluded individuals for whom no imputation was required (ie, no variables were missing). Two additional analyses were performed to evaluate potential biases based on exclusion of persons with unknown donor HCV status. First, all patients with unknown donor HCV status were labeled "unexposed," and then all were labeled "exposed."
Based on estimates of sample size and median survival provided by the SRTR, we estimated that we were well powered to detect a hazard ratio (HR) associated with receipt of an HCV-positive donor heart as low as 1.5.
Statistical analyses were performed with STATA 9.0 software (Stata Corporation, College Station, Tex). P values are 2-sided.
RESULTS
Study Cohort
Of 18 618 adult heart transplants performed during the study period, 10 915 met inclusion criteria (Figure 1) and 261 had an HCV-positive donor. Information on every variable of interest was available for 6094 patients (55.8%). Variables with the highest percentage of missingness included HLA mismatches (15.2%), recipient HCV status (14.6%), recipient CMV antibody status (12.4%), ischemic time (11.4%), candidate cerebrovascular disease (9.7%), and candidate diabetes (9.0%). All other variables used had rates of missingness between 0% and 2.4%. Pretransplant HCV RNA measurements were recorded in only 262 patients (2.4%). Among the 248 recipients with positive HCV antibody and/or HCV RIBA results, only 31 (12.5%) had confirmatory HCV RNA testing reported and 14 demonstrated viremia. Notably, recipient age was recorded for all patients.
Donor HCV Status and Survival
In the eligible cohort, overall mortality was higher among recipients with HCV-positive donor hearts throughout the transplant period. A total of 2723 deaths (132 exposed [50.6%] and 2591 unexposed [24.3%]; P<.001) were reported. Differences in reported mortality between exposed and unexposed patients were evident at 1 year (16.9% [n = 44] vs 8.2% [n = 869]; P<.001), 5 years (41.8% [n = 109] vs 18.5% [n = 1971]; P<.001), and 10 years (50.6% [n = 132] vs 24.3% [n = 2591]; P<.001). Using Kaplan-Meier methods, 1-, 5-, and 10-year survival rates were 82.9%, 52.9%, and 24.7%, for recipients of HCV-positive donor hearts and 91.7%, 77.4%, and 53.3% for recipients of HCV-negative donor hearts, respectively (P<.001, log-rank test). The crude HR associated with donor HCV positivity was 2.36 (95% confidence interval [CI], 1.98-2.81). Kaplan-Meier curves for the entire cohort are illustrated in Figure 2. One exposed and 73 unexposed patients received another cardiac transplant.
Cause of death was reported in 102 (77.3%) and 1870 (72.2%) of deceased exposed and unexposed patients, respectively. Recipients with HCV-positive donor hearts were more likely to die secondary to viral hepatitis or liver failure (13.7% [n=14] vs 0.4% [n = 8], P<.001) and coronary artery disease (8.8% [n = 9] vs 3.6% [n = 69], P = .007), but less likely to die secondary to graft failure (7.8% [n = 8] vs 16.5% [n = 308], P = .02).
Donor HCV Status and Survival in Propensity-Matched Patients
The variables in Table 3 and Table 4 were used to develop a propensity score and propensity-matched cohort with several notable exceptions. Recipient age and recipient HCV status were not included so that their effects could be examined in the main model; transplant center volume and the interaction between donor and recipient CMV status were also included in the development of the propensity score. Other variables initially of interest, namely recipient and donor hepatitis B surface antigen, donor diabetes, and donor previous myocardial infarction, were left out because they occurred so rarely, either overall or among exposed patients, that inclusion was problematic.
A comparison between propensity-matched patients (n = 1305) in the first imputed data set is given in Table 3 and Table 4. In contrast to the entire cohort (n = 10 915), the propensity-matched patients were well matched for factors included in the development of the propensity score except for year of transplantation.
As in the original cohort, over a median follow-up of 4 years, reported mortality was higher among recipients of hearts from HCV antibody-positive donors compared with recipients of hearts from HCV antibody-negative donors at 1 year (16.9% [n = 44] vs 11.1% [n = 116]; P = .01), 5 years (41.8% [n = 109] vs 23.5% [n = 245]; P<.001), and 10 years (50.6% [n = 132] vs 31.4% [n = 328]; P<.001). Using Kaplan-Meier methods, receipt of an HCV-positive donor heart was associated with decreased survival. At 1, 5, and 10 years, survival rates were 82.9%, 52.9%, and 24.7% for recipients of HCV-positive donor hearts and 88.7%, 73.1%, 44.7% for recipients of HCV-negative donor hearts, respectively (P<.001, log-rank test).
Figure 3 illustrates Kaplan-Meier curves for the propensity-matched cohort. Results of analyses in all 5 imputed data sets were quantitatively and qualitatively similar. After combining all 5 data sets, donor HCV status was significantly associated with decreased survival (Table 5). Adjustments for propensity score (as a continuous variable), baseline variables associated with mortality, transplant year, and transplant center did not substantially change this HR. Analyses investigating the effect of recipient HCV status and recipient age on the relationship between donor HCV status and survival are shown in Table 5.
Additional Analyses
Propensity-matched analyses including only patients with complete data available yielded results similar to the primary analysis except that interaction by recipient age reached statistical significance (Table 5). Equal proportions of exposed (55.9% [n = 146]) and unexposed (55.8% [n = 5948]) patients had complete data available.
Classification of all patients with unknown donor HCV status as unexposed yielded results identical to those obtained in our primary analysis. Classifying these patients as exposed produced lower HRs, but an association was still present (HR for combined data sets = 1.66 [95% CI, 1.30-2.13]).
Finally, analyses using subclassification by propensity score, rather than matching, yielded similar results to the propensity-matched analysis when both the entire cohort and only patients with complete data available were included.
To further assess potential differences between recipient age groups, age-specific findings were examined in detail. Mortality was similar across recipient age categories in the propensity-matched cohort (n = 1305) (38.0%, 33.7%, and 37.0% in ages 18-39 years, 40-59 years, and 60 years, respectively) as well as the original cohort (n = 10 915) (24.7%, 24.3%, and 26.4%). In the original cohort, no deaths were attributed to viral hepatitis or liver failure in the youngest cohort. However, death was frequently attributed to these causes in exposed persons compared with unexposed persons who were 40 to 59 years old, (10.4% [n = 5] vs 0.5% [n = 5] of reported deaths) and persons 60 years of age and older (20.9% [n = 9] vs 0.5% [n = 3] of reported deaths).
|
|
| |
| |
|
|
|
