With increasing short-term survival, the transplant community has turned its focus to delineating the impact of medical comorbidities on long-term outcomes. Unfortunately, conditions such as diabetes, hypertension and hyperlipidemia are difficult to track and often managed outside of the transplant center by primary care providers. We collaborated with Kaiser Permanente Northern California to create a database of 598 liver transplant recipients, which incorporates diagnostic codes along with laboratory and pharmacy data. Specifically, we determined the prevalence of diabetes, hypertension and hyperlipidemia both before and after transplant and evaluated the influence of disease duration as a time-dependent covariate on posttransplant survival. The prevalence of these comorbidities increased steadily from the time of transplant to 7 years after transplant. The estimated risk for all-cause mortality (hazard ratio = 1.07 per year increment, 95% CI 1.01–1.13, p < 0.02) and mortality secondary to cardiovascular events, infection/multisystem organ failure and allograft failure (hazard ratio = 1.08 per year increment, 95% CI 1.00–1.16, p = 0.05) increased for each additional year of diabetes. No associations were found for duration of hypertension and hyperlipidemia. Greater attention to management of diabetes may mitigate its negative impact on long-term survival in liver transplant recipients.
Liver transplantation has become extremely successful in the short term with 1- and 3-year patient survival now exceeding 85% and 75%, respectively (1). In this context, long-term outcomes and quality of life have taken center stage in both the clinical and the research arenas. With the decreasing risk of early graft failure and similar short-term causes of mortality, the transplant community's focus has shifted toward maximizing long-term survival after transplantation. This includes minimizing the impact of comorbidities such as the metabolic syndrome and its components: diabetes, hypertension and hyperlipidemia and obesity. The reported prevalence of these conditions after liver transplantation has varied greatly with estimates ranging between 14–61% for diabetes, 53–69% for hypertension and 31–51% for hyperlipidemia (2–4). The upper ranges certainly justify the proclamation that the metabolic syndrome is “an epidemic waiting to happen” (5).
There is substantial agreement that these comorbidities have a negative impact on long-term recipient and/or graft survival. Posttransplant diabetes, the most widely studied of these conditions, has been associated with increased rejection, infection, neuropsychiatric problems and death (6,7). Moreover, for hepatitis C virus (HCV) recipients, posttransplant diabetes has been associated with more aggressive recurrent disease and allograft fibrosis resulting in increased mortality (2,8). The individual impact of hypertension and hyperlipidemia on posttransplant survival has not been studied in depth. A recent analysis of long-term follow-up identified hypertension as a risk factor for death, but this association did not persist in the final multivariate model (7). Despite this lack of data about hypertension and hyperlipidemia, the metabolic syndrome as a whole has been linked to an increased frequency of vascular events and de novo fatty liver disease after transplantation (3,9).
Although these findings are intriguing, a critical assessment shows that the data supporting these reports suffer important shortcomings. Studies have been limited by small populations, cross-sectional designs and limited follow-up. Without sufficient patient numbers and appropriate longitudinal follow-up beginning before and extending beyond transplantation, it is impossible to accurately define the impact of these comorbidities on long-term, posttransplant mortality. Not only is there a lack of data on the incidence of diabetes, hypertension and hyperlipidemia, there is a paucity of data about the duration of disease. Heretofore, it has been impossible to analyze the cumulative effect of these conditions over time. This limitation is particularly apparent, as the posttransplant care for these conditions has increasingly shifted to local health care providers.
Given the need for high quality longitudinal data that spans the pre- and posttransplant timeframes on a large cohort of liver transplant recipients, we established a collaboration with colleagues at Kaiser Permanente Northern California (KPNC). KPNC has a comprehensive electronic medical record since 1996 that includes pharmacy, laboratory and encounter data. This integrated information system provides detailed data regarding repeated clinical encounters between liver transplant recipients and not only their primary care physicians but also all other nontransplant providers. The resultant dataset was merged with transplant data to assess the impact of medical comorbidities as time-dependent covariates on long-term mortality.
We performed a retrospective cohort study of all adult (>18 years of age) KPNC members who received a liver transplant or simultaneous liver–kidney transplant between January 1, 1997 and June 30, 2009. KPNC is a staff model integrated health care delivery organization with a comprehensive electronic medical record system covering both inpatient and outpatient care. The KPNC membership provides a unique opportunity to conduct longitudinal studies in a large cohort that is generally reflective of the underlying population. The KPNC membership accounts for one-third of the insured population in Northern California, servicing approximately 3 million members residing in 14 counties. The KPNC membership, compared to the United States Census Bay Area Metropolitan Statistical Area, has a higher percentage of Caucasians (71% vs. 64%, respectively), but fairly comparable percentages of Asian/Pacific Islanders (10% vs. 11%), African Americans (7% vs. 9%) and Hispanics (10% vs. 14%). The KPNC membership includes members of the United States Medicare and Medicaid populations, although, compared with census data, it has lower proportions of people at the extremes of socioeconomic class (i.e. extreme poverty and wealth). The University of California San Francisco (UCSF) Liver Transplant Service performs liver transplant procedures and provides posttransplant follow-up care for KPNC members. Both the UCSF Committee on Human Research and the KPNC Institutional Review Board approved the study and its methods.
Predictors and definition of terms
UCSF records were used to collect data on multiple donor (age, gender, race, cause of death and body mass index), recipient (etiology of liver disease, presence of hepatocellular carcinoma, serum creatinine, total bilirubin, international normalized ratio and pretransplant dialysis) and transplant (cold ischemia time) factors. Model for end-stage liver disease scores were calculated based purely on laboratory values. KPNC records were used to collect data on recipient demographics (age, gender and race) and pre- and posttransplant data regarding the diagnosis and duration of diabetes, hypertension and hyperlipidemia. Finally, the potential interaction between HCV liver disease and recipient duration of diabetes was analyzed.
Several different approaches to the diagnosis of diabetes, hypertension and hyperlipidemia were explored. Our primary definition for diabetes, hypertension and hyperlipidemia was based on physician diagnostic codes (Appendix A). Our secondary definition was more stringent, requiring not only a physician diagnostic code but also a prescription appropriate to the specific medical condition (e.g. an antihypertensive medication for a person with a diagnosis of hypertension) recorded 6 months or more after transplantation. Insulin therapy was also assessed based on whether any form of insulin was prescribed before transplantation or more than 6 months after transplantation. The latter criterion ensured that short-lived, posttransplant diabetes mellitus was not included. Total duration of disease was defined as the time elapsed from the first diagnosis to the end of the follow-up period.
Mortality data was compiled from institutional records at KPNC and UCSF as well as the National Social Security Death Index. All deaths recorded at either KPNC or UCSF were compared against the Social Security Death Index to ensure the death was not falsely attributed. There was complete concordance between these data sources. Specific causes of death were also compiled from KPNC and UCSF records.
Descriptive data are reported as means ± standard deviations or numbers and percentages, unless otherwise specified. Duration of recipient diabetes, hypertension and hyperlipidemia were treated as time-dependent covariates. All other predictors were treated as categorical or continuous variables as appropriate. Survival analysis was accomplished via Cox proportional hazard models. Any predictor with a p value <0.10 in univariate analysis was incorporated into the final multivariate model. P values ≤0.05 were considered statistically significant. All calculations were performed using STATA 10.0 statistical software (College Station, TX, USA).
Recipient and donor characteristics
A total of 598 adult KPNC patients underwent either primary liver (n = 542) or simultaneous liver–kidney (n = 56) transplantation between January 1, 1997 and June 30, 2009. Donors were 40.4 ± 16.1 years old, 90% deceased (brain and cardiac death), more likely to be male (54%) and Caucasian (63%). Cerebrovascular accident was the most frequent cause of donor death (43%), followed by trauma (39%) (Table 1).
Table 1. Recipient and donor characteristics for UCSF / KPNC liver transplant recipients (n = 598)
53.4 ± 9.5
MELD score at transplant
23.6 ± 12.6
Hepatitis C virus disease
40.4 ± 16.1
Cause of death
Recipients were 53.4 ± 9.5 years old, predominantly male (62%) and Caucasian (43%). HCV (49%) was the most common etiology of liver disease; approximately one-third of recipients (32%) had hepatocellular carcinoma (Table 1). A total of 117 deaths were recorded, 71 of which occurred more than 1-year after transplant. Allograft failure due to recurrent HCV was the most common cause of death (31%) followed by malignancy (28%), cardiovascular events (10%) and infection (8%; Table 2). Overall 5-year survival was 80% (Figure 1).
Table 2. Cause of death
Cause of death
N = 71
1Death due to trauma, small bowel obstruction or complications of an operation unrelated to transplant.
De novo malignancy
Prevalence, duration and treatment of diabetes, hypertension and hyperlipidemia over time
On the basis of physician-entered diagnostic codes alone, diabetes (22%), hypertension (30%) and hyperlipidemia (12%) were common before transplantation in our population (Table 3). The prevalence of each condition increased steadily until 7 years after transplant when 35% had diabetes, 56% had hypertension and 22% had hyperlipidemia. The prevalence of all three conditions decreased between 7 and 10 years after transplant. It should, however, be noted that the 10-year data is based on only 43 recipients.
Table 3. Prevalence of diabetes, hypertension and hyperlipidemia after liver transplantation
Pre-Tx n = 598
Year 1 n = 518
Year 3 n = 362
Year 5 n = 225
Year 7 n = 134
Year 10 n = 43
Recipients with diabetes had a mean duration of 6.98 ± 4.79 years with a maximum duration of 15.2 years. Sixty percent of diabetic recipients had pretransplant diabetes whereas 40% developed diabetes after transplantation. We did not encounter any patients with pretransplant diabetes that did not have persistent diabetes after transplant. Similarly, recipients with hypertension had a mean duration of 6.34 ± 4.35 years with a maximum duration of 15.44 years. Finally, the duration of hyperlipidemia was shorter; 4.52 ± 4.07 years with a maximum duration of 14.76 years.
Treatment of these conditions also varied before and after transplant. Overall, 79% of diabetic recipients received insulin at some point either before or after liver transplantation. Among recipients with pretransplant diabetes, 27% required insulin before transplantation whereas 67% required insulin 6 months or more after transplant. Among recipients who developed diabetes after transplantation, 56% required insulin. Antihypertensive medications, excluding propranolol and nadolol, were administered to 17% of all recipients before transplant. Six months or more after transplant, 48% of all recipients were treated with antihypertensive medications. Finally, 8% of all recipients received antilipemic medication before transplant whereas 20% were treated 6 months or more after transplant.
Predictors of long-term survival
We examined donor, recipient, transplant and posttransplant factors to identify predictors of long-term mortality in recipients who survived at least 1 year. For the three conditions of interest, we examined the duration of each using both our primary and secondary definitions (see Methods) as well as simple presence or absence. Univariate analysis identified recipient age at transplant (hazard ratio [HR] 1.03 per year increment; 95% confidence interval [CI] 1.00–1.06; p = 0.02) and HCV liver disease (HR 2.86; 95% CI 1.72–4.77; p < 0.01; Table 4) as strongly associated with long-term mortality for liver transplant recipients. As for diabetes, hypertension and hyperlipidemia, the duration of diabetes was a highly potent predictor of long-term mortality. When diabetes was defined by diagnostic code alone, the HR was 1.08 per year increment (95% CI 1.02–1.13; p < 0.01); when diabetes was defined by diagnostic code plus prescription requirement, the HR was 1.10 per year (95% CI 1.00–1.21; p = 0.05). Notably, the less precise definition of diabetes, simply absence or presence, did not demonstrate a significant association (HR 1.18; 95% CI 0.74–1.88; p = 0.50). Neither hypertension nor hyperlipidemia, assessed as duration or presence/absence, were statistically significant predictors of long-term survival (Table 4).
Table 4. Univariate analysis of risk factors for long-term survival after liver transplantation
Donor, recipient and transplant factors
Recipient age at tx (per year)
Female recipient gender
HCV liver disease
Recipient race (compared to white)
Recipient obesity at transplant (BMI > 30)
Donor age (per 10 years)
Donor male gender
Donor race (compared to Caucasian)
Donor body mass index
Donor cause of death (compared to anoxia)
Cold ischemia time (per hour)
Recipient diabetes, hypertension and hyperlipidemia
Recipient diabetes mellitus (yes/no)
Duration of diabetes mellitus (per year)
Dx Code alone
Dx Code and prescription
Recipient hypertension (yes/no)
Duration of hypertension (per year)
Dx Code alone
Dx Code and prescription
Recipient hyperlidemia (yes/no)
Duration of hyperlipidemia (per year)
Dx Code alone
Dx Code and prescription
To further explore the impact of recipient diabetes, we analyzed whether the need for insulin had an impact on long-term survival (Table 5). Neither insulin use at any point during the study period (HR 1.06; 95% CI 0.66–1.72; p = 0.80) nor after transplant was a predictor of long-term survival (HR 1.33; 95% CI 0.81–2.18; p = 0.25). Furthermore, when insulin therapy was incorporated into bivariate models with duration of diabetes, duration of diabetes remained a significant predictor of mortality while insulin therapy was actually protective, though this relationship did not reach statistical significance (insulin use ever HR 0.54; 95% CI 0.28–1.06; p = 0.08; insulin use after transplant HR 0.82; 95% CI 0.43–1.57; p = 0.55).
Table 5. Impact of insulin therapy on long-term survival after liver transplantation
Insulin use pre- and/or posttransplant
Insulin use posttransplant
Bivariate model 1
Insulin use pre- and/or posttransplant
Duration of diabetes (per year)
Bivariate model 2
Insulin use posttransplant
Duration of diabetes (per year)
When all predictors with a p value of <0.10 in univariate analysis were entered into a multivariate model, recipient age (HR 1.03 per year increment; 95% CI 1.00–1.06; p = 0.02), HCV liver disease (HR 2.85; 95% CI 1.71–4.75; p < 0.01) and diabetes duration (HR 1.07 per year increment; 95% CI 1.01–1.13; p = 0.02) again emerged as significant negative predictors of long-term survival (Table 6). The interaction between HCV liver disease and diabetes duration was explored by adding it to the above multivariate model. The interaction was not statistically significant (HR 0.93, 95% CI 0.82–1.03, p = 0.14); recipient age at transplant (HR 1.03 per year increment, 95% CI 1.00–1.06, p = 0.02), HCV liver disease (HR 3.81, 95% CI 1.97–7.36, p = 0.01) and duration of diabetes (HR 1.13 per year increment, 95% CI 1.04–1.24, p = 0.01) remained statistically significant. Therefore, this interaction was not presented in our final multivariate model.
Table 6. Multivariate analysis of long-term survival after liver transplantation
Age at transplant (per year)
HCV liver disease
Duration of diabetes (per year)
Finally, we explored the association between diabetes, hypertension and hyperlipidemia and specific causes of death. There were few late deaths due to cardiovascular disease and duration of diabetes (HR 1.09, 95% CI 0.93–1.28, p = 0.29), hypertension (HR 1.01, 95% CI 0.84–1.22, p = 0.92), or hyperlipidemia (HR 1.12, 95% CI 0.92–1.38, p = 0.25) were not statistically significant predictors on univariate analysis. However, duration of diabetes was a significant predictor of long-term mortality due to the combination of recurrent HCV, cardiovascular events and infection (HR 1.08, 95% CI 1.00–1.16, p = 0.05). Duration of hypertension (HR 1.02, 95% CI 0.92–1.11, p = 0.68) and hyperlipidemia (HR 1.06, 95% CI 0.95–1.20, p = 0.31) were not significant predictors.
Our review of 598 liver transplant recipients showed that a significant proportion of liver transplant recipients suffered from diabetes, hypertension and/or hyperlipidemia before transplant. This substantial disease burden may reflect our center's acceptance criteria for transplantation in this relatively modern cohort or may be related to the dominance of HCV and the escalating prevalence of nonalcoholic steatohepatitis as indications for liver transplantation. We also showed that the prevalence of diabetes, hypertension and hyperlipidemia increased steadily after transplantation such that, 7 years after transplant, more than one-third had diabetes and more than half had hypertension. Interestingly, prevalence decreased between 7 and 10 years after transplantation, possibly reflecting an insufficient number of recipients with adequate follow-up. Alternative explanations include more stringent selection criteria for the earliest cohort members or that recipients suffering from one or more of these conditions may be less likely to survive. Larger studies with longer follow-up are needed to fully elucidate these trends and their determining factors.
Despite the increasing prevalence of diabetes, hypertension and hyperlipidemia among liver transplant recipients, diabetes duration, defined by diagnostic code alone or diagnostic code plus prescription requirement, was the only condition that had a negative impact on long-term survival. Notably, neither the presence of diagnosis as a categorical variable nor the need for insulin was associated with increased mortality risk. Similarly, hypertension and hyperlipidemia, whether treated as a categorical or a time-dependent variable, were not predictors of long-term mortality. Although the detrimental effects of hypertension and hyperlipidemia on cardiovascular health are clear in the general population, their lack of impact in the liver transplant population may well reflect the latter's highly selected nature. Transplant centers thoroughly assess candidates to ensure appropriate cardiovascular health to tolerate the liver transplant procedure. It is also possible that hypertension and hyperlipidemia were not present for long enough or were not severe enough, either at baseline or secondary to effective treatment, to affect posttransplant mortality.
We have clearly shown a discrepant impact of diabetes, compared to hypertension and hyperlipidemia, on long-term survival after liver transplantation. Unlike hypertension and hyperlipidemia that predominantly predispose to cardiovascular disease, diabetes exerts a more pervasive physiologic effect, thereby increasing mortality through multiple mechanisms. Diabetes is a potent risk factor not only for cardiovascular events (10), but also for progressive renal dysfunction (11,12) and infection (6,13). Diabetes has been strongly associated with aggressive recurrent hepatitis C after transplantation, accelerated fibrosis and cirrhosis (8). Therefore, as a comorbid condition that can compromise multiple organ systems simultaneously, diabetes may disproportionately increase the risk of death for liver transplant recipients. This is supported by our subanalysis that demonstrated the negative impact of diabetes duration on the combined endpoint of long-term mortality from allograft failure secondary to recurrent HCV, cardiovascular events and infection.
Despite the large sample size and the unique long-term information available from the integrated KPNC database, our study has limitations. First, we relied primarily on physician diagnostic codes for the assignment of disease status; to increase specificity, we also conducted analyses that required a prescription for an appropriate medication for each condition. As a retrospective study, we were limited by the type of primary data available for assigning diagnoses. Although the KPNC database includes a comprehensive laboratory component, we did not collect and analyze glucose or lipid panels as there is no method to confidently evaluate the conditions of collection (e.g. fasting, concurrent pulse of steroids, relationship to medication use, etc.) for the extremely large number of samples available. Similarly, standardized blood pressure measurements were not systematically available until 2005 and may be heavily influenced by the use of therapeutic medications. Incorporation of such primary data for diabetes, hypertension and hyperlipidemia to establish diagnoses would have required consideration as to whether medical treatment was being administered; such treatments, in turn, may confound the measurements. A treated patient with well-controlled hypertension, for example, may have normal blood pressures; the blood pressures in this circumstance could not be used as part of the definition of hypertension. Thus, we determined that the use of diagnostic codes, supplemented by appropriate medication use, provided the most accurate assignment of disease status. Similar to blood pressure measurements, weights were not incorporated in the electronic medical record until 2005 rendering it impossible to accurately describe and analyze obesity in this population over time. Finally, our analyses included an in-depth analysis of all-cause mortality but only univariate analysis of more disease-specific mortality. Although we described specific causes of death, we were unable to perform multivariate analysis of each individual cause of death due to a limited number of events.
Our study strongly emphasizes the importance that future work in this area must be supported by comprehensive, multidimensional and longitudinal data to capture the dynamic impact of medical comorbidities on long-term survival of liver transplant recipients. Our collaboration with KPNC, an integrated health system that delivers primary care to a large proportion of UCSF transplant recipients, has enabled us, for the first time, to consider these conditions as covariates, which change over time, rather than as simple binary predictors. This methodological advance highlights many important avenues for future studies. Simply categorizing a medical condition as present or absent is woefully inadequate and should no longer be acceptable going forward. First, future analyses should acknowledge that these comorbidities likely exert their negative biologic impact proportional to severity as well as duration. Greater emphasis on hemoglobin A1c levels, serial blood pressures and lipid panels in future studies, ideally using prospectively collected data, may allow us to analyze disease severity. Second, analyses to delineate the relative contribution of preexisting recipient factors versus immunosuppression drugs on the development and the severity of these conditions are needed. This information will provide perspective as to whether immunosuppression regimens targeted to minimize toxicity are likely to improve long-term survival. Third, critical assessments of treatment efficacy (e.g. diabetic control) may inform our ability to modify risk. Fundamentally, the question is whether aggressive diagnosis and management of these conditions will yield survival benefits. This information may substantially alter the focus of long-term care for the liver transplant recipient.
The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.
Table 7. Appendix A
*Denotes all codes beginning with these numbers included.