Pre-transplant depression, antidepressant use, and outcomes of orthotopic liver transplantation

Abstract

Depression is a common problem among patients awaiting organ transplantation, but little is known about the impact of depression and its treatment on outcomes of liver transplantation. In this retrospective cohort analysis, we studied all patients over 18 years of age who underwent liver transplantation during a 5-year period (2004-2008) at a single center. Among 179 recipients, 65 patients had depression, as defined by health care provider assessment, prior to transplantation. Depression was defined as past or active depression or adjustment disorder. The association between pretransplant depression and outcomes of time to death, graft failure, first acute cellular rejection episode, first infection and first rehospitalization were assessed. Of the total sample, more patients with depression required post-transplant psychiatric care (37% vs. 18%), with an adjusted hazard ratio of 2.28 (1.27, 4.11). Other outcomes, including hospital readmission, acute cellular rejection, graft failure, mortality, and infection were similar for patients with and without depression. Among those with depression, patients on antidepressants at the time of transplantation had less frequent acute cellular rejection (13% vs. 40%) than those not taking antidepressants, with an adjusted hazard ratio of 0.14 (0.03, 0.62). Other outcomes were similar between these two groups. These data indicate that depression affects post-transplant psychiatric morbidity but not other medical outcomes of liver transplantation. Pharmacologic treatment of depression may significantly reduce the incidence of acute cellular rejection in patients undergoing liver transplantation. However, future prospective studies of mental health and liver transplantation are required to definitively assess the effects of antidepressant medications on medical outcomes.

Introduction

The prevalence of cirrhosis in the United States is estimated to be 400,000(1). Orthotopic liver transplantation is one of the definitive treatments for such patients. Selection of patients for transplantation is complex and varies among centers. While psychiatric evaluation is nearly universal in the screening process for liver transplantation, it is unclear what role mental health comorbidities and their treatments play in the outcomes of liver transplantation.

Mental health issues are common among patients with advanced liver disease. One study showed that of 73 patients listed for transplant, 17% had symptoms of depression and 33% had symptoms of anxiety (2). The rates are likely higher for patients with substance-related liver disease and those who are not on a transplant list. Of 71 patients in another study, 66% with alcohol-related and 32% with non-alcohol-related liver disease met criteria for affective disorders (3). In addition to affective disorders, patients with liver disease have delirium, encephalopathy, psychiatric side effects of interferon therapy, and substance abuse. All of these factors confound the measurements of psychiatric disease and complicate care.

Despite the high prevalence of mental health issues among patients with liver disease, there is little data assessing how mental health prior to transplant impacts outcomes of transplantation. Most studies have investigated psychosocial outcomes after transplantation. One study found that poor mental health pre-transplantation was predictive of alcohol recidivism (4). There is one study that assessed pre-transplant mental health and medical outcomes among recipients of liver transplantation. This study investigated a variety of psychosocial factors among a combined group of recipients of liver, heart, and lung transplants and found that pre-transplant medication non-adherence predicted late acute rejection, and living without a stable relationship predicted graft loss, but anxiety and depression did not predict poor outcomes (5).

Studies among recipients of heart transplants point to a significant role for pre-transplant psychiatric factors on outcomes. Among recipients of heart transplantation, pre-morbid psychiatric disease was found to predict the need for postoperative psychiatric intervention (6), and pre-transplant suicide attempts, poor adherence to medical recommendations, previous drug or alcohol rehabilitation, and depression significantly predicted attenuated survival times (7). However, another study found no difference in mortality by pre-transplant mental health status if patients were selected for medical adherence (8).

While little is known about pre-transplant mental health and outcomes of liver transplantation, depression post-transplant is associated with unfavorable medical outcomes among other types of transplant. Among recipients of bone marrow transplants, depression after transplantation increased post-transplant mortality (9). Among renal transplant recipients, post-transplant depression was associated with increased graft failure, return to dialysis, and death (10). This study did not assess pre-transplant mental health. Among heart transplant recipients, depression after transplant was associated with coronary artery disease but not mortality (11).

Despite knowledge that mental health disorders are common and the suggestion that psychosocial factors play a role in outcomes, there is limited research addressing how pre-transplant mental health affects post transplant outcomes in those receiving liver transplants. The purpose of this exploratory study was to assess whether the outcomes after liver transplantation differed for patients with a history of depression prior to transplant. We hypothesized that in patients undergoing liver transplantation, those with a prior diagnosis of depression would have less favorable outcomes than those without depression. We also hypothesized that treatment of depression with antidepressant medications would attenuate these less favorable outcomes in the subgroup of patients with depression.

Experimental Procedures

In this pilot study, a retrospective cohort analysis was performed by extracting data from the charts all patients over 18 years of age who underwent primary orthotopic liver transplantation at Massachusetts General Hospital (MGH) from January 2004 to December 2008. Recipients were assessed for mental illness prior to transplant and outcomes of transplantation. Institutional Review Board Committee approval was obtained for this project from MGH. Because there were no identifiers linking patients to the records, and because of the retrospective nature of the analysis and data collection, the need for consent from individual patients was waived.

Patients’ mental health data were collected from the chart, including history of depression and other psychiatric comorbidities. Depression was defined based on psychiatric assessment or diagnosis referenced in notes in the chart by a treating physician. Structured measurements of psychiatric disease were not available for patients at our center. We determined occurrence of depression based on assessments performed prior to transplantation. We did not differentiate among classifications of depression and included patients with minor and major depression, adjustment disorder, and mixed anxiety and depression, both past and present. Antidepressant use was ascertained via the medication list at the time of admission for transplant using the computerized notes from the hospitalization and/or psychiatric records.

Baseline characteristics were assessed for patients with and without depression. Race and ethnicity were self-defined and abstracted from data provided by the patients to hospital registration. Race was collapsed into white and non-white for further analyses. Ethnicity was independent of race and defined as Hispanic vs. non-Hispanic. Education was self-reported and categorized as less than high school, high school graduate, and college or more than college, and was taken from registration data provided by the patients. Age was defined as age at the time of transplant. Marital status was taken from the registration data or from the hepatology notes when not available from hospital registration data. It was classified into single, married, living with a partner, widowed, and divorced and collapsed into married or living with a partner versus unmarried/no partner for modeling purposes. At our center there is a requirement for abstinence from drugs and alcohol for at least 6 months prior to transplant. History of prior drug abuse was collected, as well as tobacco use, which was categorized into none, past, and present, and collapsed into ongoing vs. not ongoing for analysis. Model for End-stage Liver Disease (MELD) score was that collected by the transplant surgery service for reporting to United Network for Organ Sharing (UNOS) at the time of transplant. Donor risk index (DRI), was calculated using data provided by UNOS: donor age, cause of death (anoxia, trauma, or cerebrovascular accident), cold time in hours, organ location (local, regional, or national), and donor race (African-American, white, or other), donor height in centimeters, and donation after cardiac death. This risk score was calculated using the formula described by S. Feng et al. (12). Renal transplant was defined as whether patients had a concomitant transplant. In order to categorize patients by degree of illness at time of transplantation, we assessed whether the patients were in the intensive care unit (ICU) at the time of transplantation. For patients that had more than one liver transplant in the designated time period, data from the first liver transplant were used in the study. Etiology of liver disease was that listed in the primary hepatologist’s assessment of the patient. This was then further classified into alcohol and non-alcohol related liver disease, based on data that those with alcohol-related liver disease have both better outcomes of transplantation and more psychiatric comorbidities than other patients with liver disease. Hepatocellular carcinoma (HCC) was listed as distinct from the other etiology of the liver disease and was based on the diagnosis in hepatology assessments. The mean follow-up time for each group was also assessed.

Outcomes included infection within 1 year of transplant and readmission to the hospital within 1 year of transplant. Other outcomes were collected until death or December 31, 2009. Infection was defined as the time from transplantation to first acute infectious illness necessitating antibiotic treatment or hospitalization. Readmission was defined as the time of first hospitalization after discharge from transplant admission, and was collected by looking at discharge summaries and transplant surgery notes to ascertain outside hospital admissions. The time to first acute rejection episode was, in the majority of cases, defined by a positive biopsy result. Four of the patients were empirically treated for acute rejection without obtaining a biopsy, and these were included in this outcome. None of the patients were noted to have more than one acute rejection episode within the study period. Graft failure was based on the UNOS definition of graft failure and included need for re-transplantation or death. Death was ascertained by chart review and confirmed with data reported to UNOS. Psychiatric morbidity was defined as requirement of increased care with the addition of therapy, medication, or hospitalization and was collected by reading through the electronic chart to assess the time from transplantation to one of the above events.

The data were analyzed using the R statistical package, version 2.10.1(13). The primary objectives of these analyses were to determine if pre-transplant depression worsens clinical outcomes and whether outcomes differ by use of antidepressant medications in those with depression. A secondary analysis was performed to assess for differences in outcomes when those labeled “past depression” were excluded and only actively depressed recipients were assessed.

In the primary analysis, patients with and without depression prior to transplant were compared. The multivariable analysis was conducted using Cox proportional hazard modeling, first with only depression as a covariate, and then adjusting for the variables that changed the coefficient for the outcome by more than 10%. This change-in-estimate cut-off approach is thought to yield better statistical properties when assessing for confounding as compared to more general significance testing methods (13). In addition, the overall clinical focus on depression status and treatment aligns more closely with this approach. The following were the potential confounders that were assessed in the modeling: age and MELD score at transplant, which were kept as continuous variables, gender, non-white race, Hispanic ethnicity, HCC status, ongoing tobacco use at the time of transplant, history of past illicit drug use, concomitant transplant of other organ(s), MICU stay prior to transplant, married status (including those that were married or living with a partner), education less than high school, high school graduate, and beyond high school, DRI, and alcoholic liver disease. In all cases 95% confidence intervals were applied.

Analyses were then completed for patients with depression based on antidepressant use. The baseline characteristics of patients with depression were compared by antidepressant use using t-tests for continuous variables of MELD score, DRI, and age. Chi-square tests were used for categorical variables. Cox Proportional Hazard Models were then used to assess the impact of antidepressant use on the analysis. This was done using a single model with dummy variables for patients with depression on antidepressants and for patients with depression not on antidepressants, so that the differences from the baseline non-depressed group could be assessed. Patients on antidepressant medications without depression (3 total), were included in the not depressed group. Hazard ratios were generated for each of these groups compared to non-depressed. Based on evidence of potential confounding given changes of more than 10% to the point estimate, selected confounders were included in the final models. The number of death events was too small among those not on anti-depressants to create final models for this outcome. Multivariable analysis were then made between depressed subjects on and off medications.

Kaplan-Meier curves were created to visualize and assess the timing of the outcomes in each of the groups of non-medicated depressed, medicated depressed, and not depressed subjects.

Finally, a secondary analysis was performed to assess the associations between depression and antidepressant use in patients with active depression or adjustment disorder at last assessment using chi-square analysis. The group of depressed patients was divided into active and past depression and assessed for outcome rates by antidepressant medication usage or non-usage. “Active depression” included those with active adjustment disorder or depression as listed in the chart or by psychiatry evaluation. “Past depression” included those in the chart listed as “past depression”, “past adjustment disorder,” or having depression on a problem list or in a note in the chart without further elaboration.

Multicollinearity in the final models was assessed using the variable inflation factor calculation.

Results

From January 2004 to December 2008, 179 patients underwent primary orthotopic liver transplantation at MGH, of which 103 (57%) had no mental health problems, 28% had depression, 8% had depression and anxiety, 3% had anxiety alone, 2% had depression and PTSD or another mental health problem, and 2% had other mental health issues (such as anger disorder). Of the 65 patients from the group with mental health disorders classified as depressed (including those with other comorbid disorders), 54% had active depression, 15% had past depression, 17% had an active adjustment disorder, 9% had past adjustment disorder, and 5% had depression not otherwise specified but listed as a diagnosis in the chart. Hereafter, recipients with past or present depression and/or adjustment disorder will be referred to as “depressed”.

Compared to non-depressed transplant recipients, depressed recipients had more alcohol-related liver disease (37% vs. 25%), more illicit drug use in the past (35% vs. 25%), and more frequently underwent simultaneous renal transplantation (17% vs. 10%). There were fewer depressed patients in the MICU preoperatively (9% vs. 14%). None of these differences were statistically significant. While the MELD scores were statistically different between depressed and non-depressed patients (28 vs. 30, p =0.02), the difference was not deemed to be clinically meaningful. The other baseline characteristics were similar between the two groups (Table 1). Table 2 illustrates that the majority of patients in the cohort had alcohol, hepatitis C, or a combination as the etiology of their liver disease, and that depressed patients were more likely to have alcohol plus hepatitis C as an etiology (23% vs. 10% in the non-depressed group). Other etiologies were similarly distributed between the groups.

Table 1.

Baseline statistics for sample by depression status and medication use

	Depression (N=65)	No Depression (N=114)	P value	No Antidepressants (N=25)	Antidepressants (N=40)	P value
Age (in years)	51.0±9**	52.8±11	0.23	51±12*	51±7	0.94
Gender
Male	44 (67%)	84 (74%)	0.50	19 (76%)	25 (62%)	0.39
Female	21 (33%)	30 (26%)		6 (24%)	15 (37%)
Race
Caucasian	60 (92%)	95 (83%)	0.35	23 (92%)	37 (92%)	0.93
African-American	3 (5%)	10 (9%)		1 (4%)	2 (5%)
Asian American	2 (3%)	7 (6%)		1 (4%)	1 (2%)
Other	0 (0%)	2 (2%)
Hispanic Ethnicity	8 (12%)	13 (11%)	0.95	3 (12%)	5 (12%)	0.77
Education†
< High School	7 (13%)	9 (8%)	0.80	3 (12%)	4 (10%)	0.50
High School Graduate	28 (51%)	53 (46%)		13 (52%)	15 (37%)
>High School	20 (36%)	31 (27%)		7 (28%)	13 (32%)
Social Supports
Single	13 (20%)	18 (16%)	0.20	5 (20%)	11 (27%)	0.23
Married	38 (58%)	75 (66%)		17 (68%)	21 (52%)
Partner	0 (0%)	5 (4%)		0 (0%)	0 (0%)
Divorced	10 (15%)	11 (10%)		1 (4%)	7 (17%)
Widowed	4 (6%)	3 (3%)		2 (8%)	1 (2%)
Tobacco Use
None	29 (45%)	63 (55%)	0.41	11 (44%)	19 (47%)	0.62
Past	24 (37%)	34 (30%)		11 (44%)	13 (32%)
Ongoing	11 (17%)	16 (11%)		3 (12%)	8 (20%)
Former Illicit Drug Use	23 (35%)	28 (25%)	0.19	9 (36%)	14 (35%)	0.85
Other Organ Transplant	11 (17%)	11 (10%)	0.28	2 (8%)	9 (22%)	0.24
MICU pre-transplant	6 (9%)	16 (14%)	0.50	2 (8%)	4 (10%)	0.89
MELD	28±6	30±8	0.02	31±7	26±5	0.01
ETOH-related liver disease	24 (37%)	28 (25%)	0.11	10 (40%)	14 (35%)	0.89
HCC	24 (37%)	38 (33%)	0.75	8 (32% )	16 (40%)	0.70
Donor Risk Index	1.373±0.35	1.386±0.34	0.82	1.428 ± 0.37	1.293 ± 0.30	0.13
Length of Follow up	33.3±18.0	33.2±19.7	0.943	38.9 ± 16.9	29.6± 17.9	0.04*

^*the numbers after the “±” represent standard deviations

^†numbers do not sum to total b/c of missing values

Table 2.

Etiology of liver disease by depression status^*

Etiology of Liver Disease	Depression (N=65)	No Depression (N=114)
Hepatitis C only	21 (32%)	38 (33%)
Alcohol only	8 (12%)	16 (14%)
Hepatitis C and Alcohol	15 (23%)	11 (10%)
Acute Liver Failure	2 (3%)	4 (4%)
Hepatitis B	2 (3%)	7 (6%)
NAFLD	1 (1%)	4 (4%)
Alpha-1-antitrypsin deficiency	1 (1%)	4 (4%)
Cryptogenic	3 (5%)	3 (3%)
Primary sclerosing cholangitis	1 (1%)	8 (7%)
Autoimmune hepatitis	1 (1%)	5 (4%)
Other†	10 (15%)	14 (12%)

^*HCC was found in 1/3 of patients in each category, in addition to the above diagnoses

^†“other” included Caroli’s disease, Primary biliary cirrhosis, Hemochromatosis, Sarcoidosis, polycystic liver disease, and combinations of the above

Among depressed patients, those on antidepressants had a slightly lower average MELD score at transplant (26 vs. 31, p=0.01) and a shorter follow up time of 30 vs. 40 months (p=0.04) (Table 1). Those on antidepressants had non-significantly lower Donor Risk Index (1.29 vs. 1.43), were less likely to be married (52% vs. 68%) and more likely to have a concomitant renal transplant (22% vs.8%). However, the other baseline characteristics were similar between depressed patients on and off antidepressants.

Comparing the non-depressed and combined depressed populations, there were no significant differences in rates of infection or readmission at 1 year or in rates of acute rejection, graft failure, or death (Table 4). However, depressed patients had increased post-transplant psychiatric morbidity (37% among depressed, and 18% among non-depressed). The hazard ratio was 2.28 for this comparison (95% CI=1.27, 4.11). Regardless of medication use, those with depression did worse in terms of psychiatric morbidity than the non-depressed group (Table 4).

Table 4.

Cox proportional hazard models for depression vs. no depression and liver transplant outcomes

	Depressed vs. non-depressed unadjusted	Depressed vs. non-depressed adjusted *	No Antidepressants vs. non-depressed**	Antidepressants vs. non-depressed**
Infection 1 year	1.29 (0.82, 2.05) p=0.27	1.29 (0.82, 2.05) p=0.27	1.18 (0.60, 2.29) p=0.63	1.74 (0.96, 3.17) p=0.07
Readmission 1 year	1.19 (0.81, 1.75) p=0.37	1.19 (0.81, 1.75) p=0.37	1.00 (0.58, 1.73) p=1.0	1.34 (0.86, 2.10) p=0.2
Acute Rejection	0.90 (0.48, 1.69) p=0.75	0.90 (0.48, 1.69) p=0.75	1.33 (0.61, 2.90) p=0.47	0.34 (0.10, 1.14) p=0.08
Graft Failure	1.06 (0.59, 1.90) p=0.85	1.37 (0.50, 3.77) p=0.54	0.73 (0.15, 3.50) p=0.70	2.14(0.67, 6.80) p=0.20
Death	1.14 (0.58, 2.26)p=0.69	1.27 (0.28, 5.86) p=0.76	N/A	N/A
Psychiatric Morbidity	2.28 (1.27, 4.11) ‡, p=0.006	2.28 (1.27, 4.11)‡, p=0.006	2.65 (1.22, 5.71) ‡, p=0.013	2.36 (1.27, 4.75) ‡, p=0.024

^*Numbers in parenthesis are 95% confidence intervals. Graft failure and death adjusted for MELD, MICU, DRI, and education.

^**Hazard ratios adjusted as follows: Graft failure for education, MELD, MICU, and DRI, infection and psychiatric morbidity for MELD and DRI, and acute rejection for age, DRI, MICU, and education

^†numbers in parenthesis are 95% confidence intervals

^‡significant to the p<0.05 level

There were no differences in infection 1 year post-transplant between not depressed, unmedicated depressed, and medicated depressed patients (Table 4), but the Kaplan-Meier curves crossed at 8 months, with both groups of depressed patients having more post-transplant infections than the non-depressed group (48% vs. 39% at 1 year). This was not statistically significant.

Among depressed patients, those taking antidepressants had less acute cellular rejection than unmedicated patients (40% vs. 13%). The group on medications had less acute rejection than the non-depressed group (13% vs. 25%), and those with unmedicated depression had more acute rejection than the non-depressed group (40% vs. 25%) (Table 3). While the overall Cox-proportional hazard models did not yield statistically different results for the comparisons of the depressed groups to the non-depressed group (Table 4), the pairwise comparison of the depressed groups on vs. off medications yielded an adjusted hazard ratio of 0.14 (0.03, 0.62) (Table 5).

Table 3.

Outcomes by depression status and antidepressant use

	Not Depressed (N=114)	Total Depressed (N=65)	P value*	No Antidepressants (N=25)	Antidepressants (N=40)	P value*
Infection 1 year	44 (39%)	31 (48%)	0.27	11 (48%)	20 (53%)	0.35
Readmission 1 year	68 (60%)	43 (66%)	0.37	16 (64%)	27 (68%)	0.37
Acute Rejection	28 (25%)	15 (23%)	0.75	10 (40%)	5 (13%)	0.037‡
Graft Failure	29 (25%)	18 (28%)	0.85	4 (16%)	14 (35%)	0.10
Death	21 (14%)	14 (12%)	0.69	2 (8%)	12 (30%)	0.052
Psychiatric Morbidity	21 (18%)	24 (37%)	0.006‡	10 (40%)	14 (33%)	0.90

^*p values are from unadjusted cox proportional hazard modeling

^‡statistically significant to the 0.05 level

Table 5.

Cox proportional hazard models for antidepressant use vs. no antidepressant use among depressed recipients

	Hazard Ratio	Adjusted HR*
Infection at 1 year	1.42 (0.68, 2.97) p=0.35	1.99 (0.80, 4.97) p=0.14
Readmission 1 year	1.32(0.71, 2.47) p=0.37	0.34 (0.75, 2.91) p=0.25
Acute Rejection	0.32 (0.11, 0.93) ‡ p=0.04	0.14 (0.03, 0.62) ‡ p=0.009
Graft Failure	2.54(0.83, 7.71) p=0.10	4.83(0.41, 56.44) p=0.21
Psychiatric Morbidity	0.95(0.42, 2.14) p=0.90	0.91 (0.39, 2.09) p=0.83

^*Adjusted as follows: Infection for MELD, age, race, and DRI, Readmission for education, Rejection for MELD, MICU, Hispanic ethnicity, renal transplant, and education, Psychiatric morbidity for Hispanic ethnicity, and Graft failure for MELD, age, MICU, education, renal transplant, and DRI

^‡significant to 0.05 level

Readmissions over the course of the year did not differ statistically overall for the groups, but the Kaplan Meier curve demonstrates that this is likely due to an initial large amount of readmissions among all patients. At the end of the time period, both medicated and unmedicated depressed patients had more readmissions than the non-depressed group, though this was not statistically significant.

There was no significant difference in graft failure and death between the three groups of patients. However, there were trends toward more deaths (30% vs. 8%) and graft failure (28% vs. 16%, adjusted HR=4.83, p=0.21) in the medicated compared to unmedicated depressed patients (Tables 3 and 4). The Kaplan-Meier curves (Figure 1) demonstrate that the unmedicated depressed patients had less death (8% vs. 14%) and graft failure (16% vs.25%) than the non-depressed group and those on medications had more death (30% vs. 14%) and graft failure (35% vs.25%) than the non-depressed group. Of note, the numbers of deaths were too small to adjust for potential confounders and too small to carry out Cox proportional hazards analyses, but the models for graft failure were non-significant. The patients on antidepressants who died had causes of death listed as follows: 4 from infection, 2 from neoplasm, and 1 each of recurrent liver disease, pulmonary hypertension, cerebrovascular accident, motor vehicle accident, cardiac arrest, and drug overdose on SSRI and pain medicine.

Figure 1.

Kaplan Meier Curves for outcomes by depression status and antidepressant use

Kaplan Meier curves for outcomes of death(A) graft failure (B).

Secondary analyses conducted to assess the differences between patients labeled as having “active depression or adjustment disorder” vs. those with history of depression, adjustment disorder, or unspecified diagnosis of depression in the chart, demonstrated that results did not qualitatively change with inclusion of only actively depressed patients, though the sample was underpowered to detect statistically significant differences in outcomes. However, death did reach statistical significance in the subgroup of actively depressed patients in univariate analysis, with 8 in the group on medications and 1 in the non-medicated group with mortality.

When comparing the groups of patients with “ongoing” and “past” depression, we found that 16 of 21 of those with “past depression” were on antidepressants and that their numbers looked similar to those classified as “ongoing depression” on medications (Table 6). Similarly, the trends among the “past depression” group off medication reflected those of the actively depressed group off medication.

Table 6.

Outcomes by depression status and medication use

	Ongoing depression No meds (N=20)	Ongoing depression+Meds (N=24)	Past depression No meds N=5)	Past depression+Meds (N=16)
Infection 1 year	10 (50%)	11 (46%)	1 (20%)	9 (56%)
Readmission 1 year	13 (65%)	15 (63%)	3 (60%)	12 (75%)
Acute Rejection	7 (35%)	4 (17%)	3 (60%)	1 (6%)
Graft Failure	3 (15%)	9 (38%)	1 (20%)	5 (31%)
Death	1 (5%)	8 (33%)	1 (20%)	4 (25%)
Psychiatric Morbidity	9 (45%)	6 (25%)	1 (20%)	8 (50%)

No evidence of multicollinearity was found within the final models using.

Discussion

We have shown in this retrospective cohort analysis that, among patients with a history of depression pre-transplant, there was an increase in psychiatric morbidity one year after transplantation, requiring stepped-up treatment or hospitalizations. This highlights the importance of screening patients for depression prior to transplant and following them closely. The other medical outcomes were notably similar for patients with and without depression prior to transplantation.

Patients with depression were more likely, though not statistically so, to have a history of alcohol use and illicit drug use in the past, were more likely to have a combined hepatitis C and alcohol etiology of their liver disease, and more likely to have alcohol-related liver disease in general. This was not surprising given that substance abuse and affective disease are clearly linked in patients with liver disease. In one study of 71 patients, 66% with alcohol related and 32% with non-alcohol related liver disease met criteria for affective disorders (3). In past studies, up to 40% of those with alcohol-related liver disease had a history of illicit drug abuse (14). This connection highlights the importance of screening those with hepatitis C and alcohol related liver disease for affective disorders in the transplant setting.

A recent review of psychiatric disease and transplantation recommends that patients with depression not be excluded from transplant but rather be treated in a similar manner to the rest of the population; preferably with medication and psychotherapy, often over long periods of time (15). Transplantation psychiatric intervention is optimal with pre-transplant secondary prevention and postoperative continuity of care. Failure of adequate treatment is not uncommon, be it a result of patient non-adherence or of professional intervention. Clinicians may be uncomfortable about prescribing medications in the setting of liver failure, so communication between the treating practitioner and the transplant team (including appropriately trained mental health professionals) can be of great benefit in arriving at appropriate dosing of antidepressants. Factors for consideration to determine a safe and efficacious dose of antidepressants in patients with severe liver dysfunction include knowledge of drug-drug interactions with antirejection medications and the specific agents that are associated with hepatotoxicity. Specifically, there is a black box warning against use of duloxetine in hepatic failure, and nefazadone, trazodone, and venlafaxine have been associated with hepatotoxicity. Several elements help in safety monitoring such as knowledge of and attention to evidence of toxicity and adverse events with specific agents, close follow up of laboratory markers of toxicity, and use of serum levels of active medication metabolites. Determination of the above variables is routinely availability when the team has access to consultation with a dedicated transplant psychiatrist, pharmacologist, and hepatologist.

We expected that, among recipients with depression, those who were on antidepressants would have improved outcomes. Our finding of reduced acute cellular rejection among depressed patients taking antidepressants could be a function of adherence to transplant-related medications and is consistent with other studies showing a correlation between adherence and acute rejection. Past research has shown that 57% of those who did not attend their appointments regularly experienced acute rejection, as opposed to 2% of those who attended appointments, which supports the notion that medical adherence mediates acute rejection (16). One recent paper found that medication non-adherence predicted increased late acute rejection among patients undergoing heart, liver, and lung transplantation (5). Depression is one among many factors in medical non-adherence, including age, distance from and access to the transplant center, active substance abuse, memory impairment, and some types of personality impairment. Butler et al found that depression was among the leading causes of decreased adherence among patients awaiting renal transplant (17). Taking these findings together, untreated depression may be associated with decreased adherence and thus acute rejection.

The typical post-transplant immunosuppressive regimen in our institution includes one calcineurin inhibitor (e.g. cyclosporine or tacrolimus), an antiproliferative agent (e.g. mycophenolate mofetil or azathioprine), and a steroid for up to 1 year post transplant. While this study was not designed to assess medication adherence, a post-hoc analysis showed no significant difference in tacrolimus levels at the time of rejection based on antidepressant use. This lack of difference may be due to reduction in the goal medication level based on the time since transplantation or variable adherence to or doses of other non-measured anti-rejection medications. Assessment of adherence should be a part of future study on this topic. However, it may be that this relationship is related to other pathophysiologic factors than adherence. There could also be a relationship of untreated depression with modulation of the immune system, as it is postulated that depression is associated with increased activity of the hypothalamic-pituitary-adrenal axis and glucocorticoid resistance (18), making rejection more likely.

While depressed patients did not differ in acute rejection from non-depressed patients, this is possibly due to the decreased amount of acute rejection among patients on antidepressants in this sample, as medicated patients had non-significantly less acute rejection than non-depressed patients. Interestingly, none of the three patients on antidepressant medications for reasons other than depression experienced acute rejection. There may be an unmeasured effect of psychiatric medications on anti-rejection medication metabolism to account for this, or that taking an antidepressant is a marker for adherence to other transplant-related recommendations. While it is possible that the change in outcomes for patients with depression based on antidepressant use is mediated by newly diagnosed psychiatric issues, the lack of significant differences in psychiatric morbidity between those on and off medications suggests that the psychiatric medication use and not post-transplant psychiatric morbidity mediate this difference.

An unexpected finding in this study was a trend toward increased mortality among those on antidepressants compared with those not taking antidepressants. The number of depressed patients who suffered mortality was too small to allow for adjustment for potential confounders, and no conclusions can be drawn from this result. Though this is potentially quite important, and there has been some evidence suggesting increased death among those on antidepressants (19) potentially mediated by increased bleeding or hepatotoxicity (15), in this study the number of outcomes was too small to be considered stable. Given the causes of death of the recipients in this study as listed in the results section, it is difficult to implicate antidepressant medications as the cause of the overall increased rate of death. There was one recipient on antidepressant medications who had drug overdose as the cause of death. While there is literature indicating that suicide is increased among young people on SSRIs, more recent findings indicate this is likely not the case in those over age 25 (20), which was the case for this recipient. However, this relationship warrants further investigation in a prospective trial, as antidepressant use may be a marker of more severe illness at baseline in this observational study.

The increased psychiatric morbidity for patients with a history of depression as well as the decreased rates of acute rejection for those with depression on antidepressants highlights the importance of appropriate peri-operative teaching, psychiatric intervention, and chronic care for transplant recipients. Patients should be educated in the importance of adherence with required psychiatric treatment. Patients at risk may benefit from increased frequency of visits to transplant clinic as well as from screening for major life stress, substance abuse, and altered mental status. Use of standardized self-testing for depression is another consideration. Communication between the transplant team and the treating psychiatrist can be helpful.

While this study is unique in its assessment of antidepressant use and medical outcomes, it is an exploratory study with several limitations. These include the small numbers of patients and outcomes, especially of death and graft failure, limiting the ability to interpret the analysis of these outcomes in terms of antidepressant use. Also, the multiple comparisons made in this study limit the interpretability even in the setting of small p-values.

At our institution patients with a history of psychiatric disease, hepatitis C, HCC, or substance abuse are typically referred for psychiatric assessment. There is no formal protocol for psychiatric follow-up, however patients are only considered for transplant if a psychiatrist deems their conditions stable and has a plan for ongoing care. Future studies could benefit by using standardized DSM-4 instruments to diagnose depression in the entire sample as well as standardized follow-up. In this study, the definition of depression included people with past depression and comorbid anxiety, and anxiety could have played a significant and independent role in outcomes. There was no differentiation between diagnostic categories of depression or timing of depression due to lack of information in the chart in terms of psychiatric symptoms at the time of transplant. Given the high rate of recurrence of depression in the general population, and the vulnerability of patients to relapse at times of stress, it was thought that patients with a history of pre-transplant depression would benefit from treatment with medications. Though we cannot speak to the severity of preoperative depression due to lack of data, we interpreted the data as demonstrating that patients taking anti-depressants were gaining benefit in functional capacity.

Consideration was given to including only patients with diagnoses of active depression in the study, however, the majority of patients that were labeled as past depression were on antidepressant medications (16 of 21 patients), indicating that they were actively being treated for depression and could not be fairly classified as “non-depressed”. Additionally, the timing of the psychiatric analyses were non-uniform, so there was concern for misclassification bias if the patients with “past depression” were included in the non-depressed group. There was insufficient data to determine the severity or timing of mood disorder, so patients were included. Those with non-active psychiatric disease had rates of outcomes that were strikingly similar to their counterparts with active disease, so all were included. Future study would benefit from uniform timing of assessments with standardized instruments both before and after transplantation.

Lack of randomization in this observational study means that the differences based on treatment may be a function of unmeasured confounders. An important unmeasured confounder is psychotherapeutic treatment of depression. This variable was not used because of the difficulty in tracking the frequency of psychotherapy visits retrospectively in the referral center setting. Generalization of study findings is limited because as a result of single center participation in this effort. Individual centers have different ways of evaluating, screening for and treating depression in patients with advanced liver disease being evaluated for liver transplantation(15).

In summary, this study assessed medical outcomes based on pre-transplant psychiatric comorbidities and antidepressant use for patients undergoing orthotopic liver transplantation. Our results suggest that patients with pre-transplant depression may benefit from treatment with antidepressants, and this treatment may reduce rates of acute rejection after liver transplantation. Results of this study suggest that depression as a diagnosis or symptom complex is insufficient to determine adverse effects on post transplant prognosis, though patients with pre-transplant psychiatric illness should be monitored for increased post-transplant psychiatric morbidity. Future studies of mental health and liver transplantation are needed to definitively assess the effects of antidepressant medications on medical outcomes.

Abbreviations

MGH

Massachusetts General Hospital

HCC

hepatocellular carcinoma

MICU

medical intensive care unit

MELD

model for end-stage liver disease

UNOS

United Network for Organ Sharing

OR

odds ratio

ETOH

alcohol

DRI

donor risk index