Deceased donor factors associated with poor graft outcome are well known, but how often these factors lead to livers left untransplanted is poorly defined. A nested, case-control study was conducted using the United Network for Organ Sharing (UNOS) database from 1987 to 2005. Only those donating ≥1 solid organ were included. Primary outcome was livers not transplanted (LNT, cases) versus transplanted (LT, controls). Primary variables for multivariate analysis were donor age and obesity. Covariates included donation after cardiac death (DCD), cerebral vascular accident death, viral serologies, cancer, ALT and bilirubin. There were 23 373 (26%) LNT's from 91 362 donors who donated at least one organ. Percent LNT fell over time (1987–1990: 48%; 1991–1995: 29%; 1996–2000: 21%; 2000–2005: 16%; p < 0.01). Increased age (odds ratio: 4.2, 95% confidence interval 3.6–4.9, p < 0.01) and obesity (2.1, 1.9–2.3, p < 0.01) were significantly associated with LNT across all time periods. Other significant factors included DCD and elevated ALT. For 2001–2005, population attributable risk indicate that age >40, abnormal ALT and obesity account for 32.6%, 25.3% and 9.2% of untransplanted livers, respectively. Use of expanded criteria livers has pushed LNT lower in spite of an aging and heavier donor population. Nevertheless, age and obesity still account for a significant portion of untransplanted livers.
Despite rising demand, a considerable number of livers from deceased donors are not transplanted. Therefore, it is important to understand how often and why deceased donor livers are left unused. In 2002, 1273 (10.3%) of all deceased donor livers recovered were ultimately not transplanted into a recipient (1). Presumably these livers were unused due to unacceptable characteristics such as advanced donor age or donation after cardiac death (2–5). However, United Network for Organ Sharing (UNOS) reporting of nontransplanted livers often labels the organ as ‘poor quality’ without further specification. Therefore, little is known about which factors most commonly account for nontransplanted livers.
We were particularly interested in nontransplanted livers from donors who successfully donate another solid organ (i.e., kidney, lung, heart or pancreas). This subgroup is of interest since they met other minimum criteria for successful donation (e.g., HIV negative, lack of widespread cancer). We hypothesized that obesity and increased age drive much of the liver nontransplantability in this donor subgroup. Both could dissuade a surgeon from transplanting the liver, obesity through its association with graft steatosis (2,4,6). Both characteristics are also on the rise in the general population (7,8). Therefore, we examined the UNOS database to assess deceased donor characteristics for associations with liver nontransplantability and relative importance based on prevalence in the donor population.
We conducted a nested, case-control study of utilization of livers from deceased donors in the United States. We used the UNOS database to identify deceased donors from October 1, 1987 to July 16, 2005. The database includes individual organ dispositions. The Saint Louis University Institutional Review Board (IRB) approved of this study prior to initiation.
The study population was limited to deceased, solid organ donors who successfully underwent procurement, followed by transplantation of at least one solid organ from October 1, 1987 to July 16, 2005.
The primary outcome of interest was liver not transplanted (LNT = case) versus transplanted (LT = control). Split liver donation was excluded given the likely differences in liver quality required for split livers compared to whole grafts. The proportion of LNT was also examined separately for 1987–1990, 1991–1995, 1996–2000 and 2001–2005 to assess change over time.
The primary variables of interest were donor age and obesity. Donor ages were calculated from recorded birth date, organ recovery date and age entries. Donors were grouped by age increments (i.e., <9 years, 9–20, 20–30, 30–40, 40–50, 50–60, 60–70 and ≥70). Donors with calculated ages of ≤0 or >100 years were excluded. We used obesity as a surrogate marker for fatty liver disease, since donor biopsy data are recorded in a small minority of cases making steatosis difficult to assess in the UNOS database (4). Obesity was defined as a body mass index (BMI) of ≥30 kg/m2 for adults (≥20 years old) and ‘BMI-for-age’ in the 95th percentile, for pediatric and adolescent donors 9–20 years old (9,10). No donor less than 9 years of age was classified as obese. Obesity is a surrogate marker for fatty liver disease in our study, and data suggest that fatty liver disease is rare in the under 9 age group (11). Donors with a calculated BMI <14 or >55 were excluded since these outliers were likely representative of data entry errors.
Factors cited as important to the decision to transplant a liver were examined. These included death by cerebral vascular accident, donation after cardiac death (DCD), inotropic blood pressure support, elevated aminotransferases (ALT and AST), elevated bilirubin, infection, positive viral serologies (human immunodeficiency virus [HIV], human T-lymphotropic virus [HTLV], hepatitis B and hepatitis C) and history of malignancy (2,4,12–19). Due to high colinearity, between ALT and AST, using either or both did not alter our results significantly. Therefore, we included only ALT in our analysis. HIV and HTLV status variables were dropped because positive cases were too few. Serum sodium was not assessed as greater than 92% of donors had no sodium value recorded. Similarly, liver biopsy data were not included in the analysis because only 8.8% of donors were recorded as having a liver biopsy, and specific histologic findings were not available. The presence or absence of diabetes and hypertension were included given their established associations with insulin resistance and fatty liver disease. We also examined three insulin resistance covariates, defined as donors who had one or more of the following: obesity, diabetes and/or hypertension. UNOS does not capture data on donor hyperlipidemia. Donor blood type was included since the ease of matching a graft to a recipient varies across ABO blood groups. Finally, relevant sociodemographic factors including gender, race, and history of drug use were included in the analysis.
We computed the numbers and percentages of donors with LT and LNT for the entire study period (1987–2005) and for 4–5 year periods (1987–1990, 1991–1995, 1996–2000, 2001–2005). We used histograms to determine trends in age and obesity over time.
For univariate analysis, the percentages of donors with specific demographic and clinical characteristics were computed for the LNT cases and LT controls. The association between each variable and LNT versus LT was evaluated using Mantel-Haenzsel stratified analysis. The crude odds ratio and 95% confidence interval estimated the strength and precision of associations with LNT. On multivariate analysis, the odds ratios were adjusted for potential confounding using logistic regression. We entered age, obesity and covariates in four blocks. We also looked for possible interactions between donor age, obesity, diabetes and hypertension and their effect on LNT versus LT status since the presence of more than one of these variables may serve as a better surrogate marker of fatty liver disease. Modeling was also done by 4–5 year period. Analyses by region and organ procurement organization (OPO) were also done for 2001–2005.
Finally, adjusted population attributable risk percents (PAR%) were calculated ([(adjusted odds ratio −1)/adjusted odds ratio]× prevalence of the exposure × 100%) to determine which variables account for the most untransplanted livers (20).
Analyses were done using SPSS, version 13 (SPSS, Inc., Chicago, IL) and SAS version 9.1 for Windows (SAS Institute Inc., Cary, NC).
There were 174 582 deceased donors entered in the UNOS database from October 1, 1987 to July 16, 2005. 106 231 had data indicating successful donation of at least one solid organ (i.e., heart, pancreas or pancreas segment, right and/or left lung, right and/or left kidney, liver or liver segment, intestine or intestinal segment). Those donors unable to successfully donate even one organ (68 351 or 39%) were significantly older (mean age 45.3 ± 21.8 vs. 34.9 ± 1.4 years, p < 0.001) with higher proportions over 70 (14% vs. 2.3%, p < 0.001) and over 60 (29.2% vs. 9.7%, p < 0.001). Unsuccessful donors were also more likely to be obese (OR 1.2, p < 0.001), have diabetes (3.0, p < 0.001) or hypertension (2.5, p < 0.001). Mean ALT was higher (162.5 vs. 74.6, p < 0.001). As expected, they were more likely to have positive viral serologies (odds ratios [OR's] for HBsAg: 52.7, p < 0.001, HCV: 7.3, p < 0.001, HIV: 424.0, p < 0.001), have DCD status (OR 2.5, p < 0.001) and have a history of cancer (OR 2.3, p < 0.001). Death by cerebro-vascular accident (CVA) only mildly increased the risk of unsuccessful donation of all organs (OR 1.1, p < 0.001). We also excluded 770 split liver donations, 8222 with extreme BMI's (<14 or >55 kg/m2), and 3552 cases with calculated ages ≤0 or >100. Of the remaining donors, 91 362 with liver disposition data were included in our study.
Cases and controls
There were 23 373 (26%) LNT cases and 67 989 (74%) LT controls of the 91 362 deceased donors who successfully donated at least one solid organ. The percentage of LNT cases has declined significantly over time: 48% for 1987–1990, 29% for 1991–1995, 21% for 1996–2000, 16% 2000–2005 (p < 0.01) (Figure 1). Overall, 12 788 (54.7%) of the 23 373 LNT cases had livers that were never recovered. In the 2000–2005 era, 2324 (51.0%) of 4550 LNT cases had livers not recovered. Amongst all LNT's, 93.1% successfully donated kidneys, 24.6% heart, 8.4% lungs, 4% pancreas and 0.2% intestine.
Table 1 shows the demographic and clinical characteristics of cases and controls. Donors less than 30 years of age or in their 70s had increased odds of having their liver transplanted, whereas donors between 30 and 69 had increased odds of having their LNT. Obese donors were more likely to have their LNT (OR 1.90, 95% CI 1.91–1.99). Diabetes and hypertension were associated with LNT but to a lesser degree than obesity and age.
Table 1. Univariate analysis of selected demographic and clinical characteristics of LNT versus LT donors
1Liver not transplanted.
3Donation after cardiac death.
Adjusted odds ratios for our final logistic regression model are shown in Figure 2 for the total study population and for 2001–2005. Those variables with adjusted odds ratios greater than 1.0 represent an increased risk of LNT and were statistically significant (p < 0.05) if the 95% confidence intervals do not cross 1.0. Age and obesity remained statistically significant and were more strongly associated with nontransplantation after controlling for covariates. The adjusted odds ratios were 4.2, 95% CI 3.6–4.9 (p < 0.01) for donors aged 50–60 and 2.1, 95% CI 1.9–2.3 (p < 0.01) for obese donors. All donor age groupings were compared to the referent age group of 9–20 years old. Similarly, ALT and bilirubin groups were compared to the normal referent range groups of ALT <46 U/L and bilirubin <1.3 mg/dL, respectively. Blood groups were compared to Blood Group O as the referent. Insulin resistance (IR) variables (1, 2 or 3 insulin resistance characteristics) and interaction terms between age, obesity, diabetes and hypertension did not alter the model significantly (data not shown). Odds ratios for age, obesity and other covariates were similar across time intervals. DCD and elevated ALT had the highest odds ratios. Only those for 2001–2005 are shown in Figure 2 since they are most pertinent to current practice.
Population attributable risk percents—2001–2005
Population attributable risk percents for variables with significant adjusted odds ratios are shown for the 2001–2005 era (Figure 3). Age >40, elevated ALT, and obesity had the highest PAR%'s. While DCD status had a high adjusted odds ratio (Figure 2), its PAR% was relatively low due to low population prevalence (Figure 3).
Obesity and age odds ratios by region and OPO
We determined age and obesity OR's by region for 2001–2005 (Figure 4A,B). These OR's were adjusted for DCD status, ALT, bilirubin and anti-HCV, the significant variables determined in our main modeling (Figure 2). Only the age group 60–70 is shown, but age group OR values in each region showed a peak in the 50–60 age group similar to our main model results. We also determined unadjusted odds ratios for obesity and age by OPO for 2001–2005 (Figure 4C,D). The variation in donor livers handled at each OPO was large (58–1447) as was the variation in OR's. The smaller numbers per OPO precluded robust multivariate analysis and age was dichotomized at donor age over 50. While we could have increased the numbers by including prior eras, we were most interested in recent utilization practice.
Age and obesity trends
Because age and obesity had relatively high adjusted PAR%, we examined donor age and obesity prevalence across time to look for trends. While mean age of our total cohort has increased significantly (29.0 years for 1987–1990, 32.7 years for 1991–1995, 36.4 years for 1996–2000, 38.7 years for 2001–2005, p < 0.001) the nature of this change is best captured with age histograms. Histograms show the age distributions for LNT and LT donors by specific 4–5 year periods (Figure 5A,D). For the LT donors in particular, a second peak centered on age 50 has appeared. The number of donors in the more desirable 15–25 age bracket has increased as well, but to a lesser degree. Age data was largely complete but did show improvement with less missing values through the years (1987–1990: 3.3% missing; 1991–1995: 0.5% missing; 1996–2000: <1% missing; 2001–2005: <1% missing; p < 0.05).
Percent obesity has also increased significantly for both LNT and LT donors (Figure 6). However missing BMI data was more common than for age data (1987–1990: 99% missing; 1991–1995: 69% missing; 1996–2000: 17% missing; 2001–2005: 1.6% missing; p < 0.05). We included a 1994–1995 era histogram because the missing data rate fell from nearly 100% in 1993 to 30% in 1994 and to 10% in 1995, likely reflecting a policy change in collecting height and weight data. While the 1994–1995 histogram represents a shorter time frame than the other histograms, it still gives a fair representation of the proportion of donors with obesity.
As the demand for liver transplantation rises, it is important to understand how often and why deceased donor livers are left unused. Such nontransplantability is of particular interest in deceased persons who are able to donate at least one organ since they have met other minimum criteria for donation. Ojo et al. demonstrated that 10% of livers recovered are not ultimately transplanted, while Feng and colleagues showed that these recovered but discarded livers were associated with donor age, DCD and decreased height (1,5). However, no studies have looked at overall nontransplantation of livers (i.e. recovered and not recovered).
We found that 39% of those registered did not have even one organ procured and transplanted. These unsuccessful ‘donors’ were significantly older and had more diabetes, hypertension, obesity, positive viral serologies, and cancer. Fully 80% had no recovery date recorded, so organs likely were not even removed. Many may have had consent issues, but we did not analyze this group extensively as we were primarily interested in those successfully donating at least one organ. Of these ‘successful’ donors, the percent of nontransplanted livers has been substantial (26%) over the past 18 years, but has fallen significantly from a high of 48% in the late 1980s to just 16% in the last 5 years (p < 0.01) (Figure 1). This percentage fall is largely due to the dramatic rise in transplanted livers, while the absolute numbers of LNT cases has fallen. Such changes have likely been driven by increased demand and use of expanded criteria livers.
While the proportion of LNT has fallen, it remains significant. In 2001–2005, 4550 livers from persons donating at least one other organ were not transplanted. However, the precise reasons for LNT are not well detailed in the UNOS database. Nonspecific terms such as ‘poor organ function’ and ‘diseased organ’ are often codified and used.
Among the factors associated with poor graft survival, we hypothesized that increased age and obesity would explain much of LNT (6,15). Both are common and increasing in the general population (7,8). Increased age is an established risk for graft failure per the Workgroup on Expanded Criteria Organs for Liver Transplantation and would weigh in on the decision to not transplant (4). While obesity was not considered a risk factor, the Workgroup deemed steatosis as a ‘potential’ risk. Lack of donor biopsy data did not allow adequate analysis. In lieu of biopsy data, we used donor obesity as a surrogate marker for steatosis, since the prevalence of fatty liver in the obese may be as high as 75% (6). We also examined diabetes and hypertension alone and in combination with obesity.
Our study confirmed age as significantly associated with nontransplantability after controlling for other covariates. The 9- to 20-year-old donors served as the referent (OR = 1.0) with all other age groups being compared to this group of donors (Figure 2). Inadequate liver volume probably explains the increased OR's for less than 9-year-old donors. Donors 20–30 had only mildly elevated aOR and lost significance in the 2001–2005 era. The OR's peaked in the 50–70 range but then declined. The decline in OR's for 70 year-olds is likely due to selection bias with only the very fittest even being considered for donation.
The reasons for donors aged 30–50 to have such increased OR's are less clear, but may be due to the high desirability of livers in the reference group and unaccounted confounders. The reference group of 9- to 20-year-old donors consists of children and young adults who probably had fewer comorbidities compared to 30–50 year olds. Unlike the other age groups, some in the referent group could have been considered for either pediatric or adult recipients perhaps increasing the chance of transplantation. Certain confounders such as nontrauma cause of death which has been associated with graft failure were not examined (4). Nationwide, the leading causes of death in 35–55 year olds are nontraumatic (e.g., heart disease), while the leading cause in 9–20 year olds is trauma (21). Also risk of transmissible infections due to recent high risk behavior (e.g., recent high risk sexual contact or incarceration) was not examined and would be more prevalent in the over 30 group compared to the reference group. Thus, compared to the highly sought livers in 9–20 year olds, perhaps livers from 30 to 50 year olds, with a confounder, would be less likely to be transplanted.
Obesity was also independently associated with liver nontransplantation (2.1, 95% CI 1.9–2.3 for obese donors) in our model (Figure 2). Odds ratios for obesity were similar across all periods, except for the 1987–1992 period where BMI data was so sparse as to not yield valid OR's. Decreasing donor height which would increase BMI has been associated with liver discard after procurement (5). Obesity's apparent independence in our model is likely due to our inability to include fatty liver disease. There is no empiric reason that a liver would go unused purely because the donor is obese. Indeed, nontransplantability due to steatosis would explain why obesity was associated with LNT in our study, but not graft failure in others (4). Livers transplanted from obese donors are likely selected for absence of significant steatosis, while those not transplanted are culled due to the presence of significant steatosis.
Other factors strongly associated with LNT included DCD and abnormal ALT (Figure 2). Diabetes and hypertension had much weaker associations. Interaction terms between diabetes, hypertension and obesity were not significant (data not shown). In 2001–2005, obesity and age OR's varied widely by region and OPO (Figure 4). The differences likely reflect local transplant center practices. Yet firm conclusions are difficult to draw since the confidence intervals overlap greatly and the OR's for OPO's were unadjusted. We also did not account for exported organs.
PAR% expresses liver nontransplantability attributable to a particular risk factor based on its prevalence in the donor population. Age >40, abnormal ALT and obesity carried the highest adjusted PAR%'s (32.6%, 25.3% and 9.2%, respectively) (Figure 3). Therefore, these three factors may account for up to two-thirds of nontransplanted livers. Though centers are more accepting of livers from older and heavier donors, age and probably fatty liver disease remain major determinants for leaving livers unused, while accepting other organs from the same donor. DCD and positive viral serologies are also strong deterrents to transplanting a liver. However, they are less commonly encountered and therefore account for fewer livers left untransplanted.
Since age, elevated ALT and obesity through its association with NAFL may account for up to 67% of LNT, we looked for trends in these variables over time. Age histograms show a growing second peak around age 50 (Figure 5). This second peak is remarkable for its height and width with many livers coming from donors now over 60 years. Obesity prevalence has also increased (Figure 6). These changes are likely due to our increasing acceptance of older livers and livers with some fatty infiltration, but also reflect trends in the general population. There were no similar trends in ALT levels for the same time periods.
Weaknesses in our study stem from incomplete data. Only 8.4% of donors had sodium values, which were primarily recorded in the last few years. Thus we did not include this variable in the main modeling. However, separate modeling on the subgroup with sodium values yielded a PAR% of only 0.6% for sodium >170 mEq/L. Results for age, obesity and ALT were not significantly altered (data not shown). Cold ischemia time is not captured when the liver is discarded. The reasons for turning down a liver locally and extending cold ischemia time once procured are likely amongst the variables we examined. However, cold ischemia time's affect on these variables could not be determined. Donor hospital length of stay is not recorded, but has been linked to poor graft outcome and may be an important covariate (22). Finally, erroneous and missing entries are inevitable, but should have occurred equally across cases and controls.
Despite these shortcomings, our study identifies the relative weights of donor factors that account for nontransplanted livers, both recovered and nonrecovered, on a national level. The prevalence of elevated ALT has been stable over time, while age and obesity are on the rise. Despite the remarkable increases in donor age and BMI (and presumably nonalcoholic fatty liver), the proportion and number of nontransplanted livers has fallen in the last 18 years, reflecting acceptable outcomes for older livers and perhaps livers with more fatty infiltration. Such outcomes will be critical as the donor pool continues to age and have more fatty liver disease.
This study was supported by American Gastroenterological Association 2004 AGA, Roche Junior Faculty Clinical Research Award in Hepatology; Saint Louis University Young Investigators Award, Saint Louis University, Department of Internal Medicine.