|Home | About | Journals | Submit | Contact Us | Français|
While depression after liver transplantation (LTX) is associated with decreased survival, the effects of treating depression remain unknown. We assessed a previously-described, prospective cohort of 167 patients transplanted for alcohol-related liver disease from 1998–2003. Depressive symptoms were measured with the Beck Depression Inventory serially throughout the first post-transplant year. Adequacy of antidepressant treatment was measured with the Antidepressant Treatment History Form. Using Cox-proportional Hazards modeling, survival times were assessed for recipients with no depression vs. depression with adequate medications vs. depression with inadequate medications. Seventy-two recipients had depressive symptoms in the first post-transplant year. Of these, 43% (n=31) received adequate pharmacotherapy and 57% (n=41) received inadequate (n=7) or no pharmacotherapy (n=34). After a median follow-up time of 9.5 years, 32% of the inadequately treated depressed group survived vs. 52% of the adequately treated group and 56% of the non-depressed group (p=0.006). Compared to the non-depressed group, those with adequately treated depression had no significant difference in survival. However, recipients with depression and inadequate pharmacotherapy had decreased survival times compared to non-depressed recipients (HR for death=2.44, 95% CI=1.45, 4.11), controlling for other known confounders. The factor most strongly linked to long-term mortality after liver transplantation in this cohort was untreated depression.
Depression is a common mental health disorder among recipients of liver transplantation (LTX), with a prevalence of more than 60% prior to LTX(1) and 30–40%(2) after LTX. Depression is treatable, and early identification and treatment of those suffering with the symptoms of this disorder can greatly improve quality of life and restore more normal functioning. Beyond the importance of quality of life and symptomatic relief, depression and depressive symptoms have also been associated with increased all-cause mortality among the elderly(3) and decreased survival in specific disease processes, including diabetes,(4) heart failure,(5) and post-coronary artery bypass.(6)
Our recent research has similarly demonstrated the importance of depression to long-term post-LTX survival in alcohol-related liver disease.(2) Both persistent and worsening depressive symptoms in the early post-transplant period were associated with decreased long-term survival. In fact, depression scores were more strongly associated with survival than MELD, donor age, hepatitis C virus (HCV), or other transplant-related factors. A similar study found that depression predicted mortality after LTX among recipients with all etiologies of cirrhosis.(7)
While ongoing advances continue to improve outcomes in LTX, the 10-year deceased-donor survival rate remains at only 54%(8) and strategies to identify modifiable factors impacting survival are needed. Given our determination that depression is associated decreased survival times, mental health may be one such potentially modifiable risk factor. To date, there has been descriptive work assessing the treatment of depression after organ transplantation,(9) but the one study assessing post-transplant survival by antidepressant medication treatment status had too few deaths to examine long-term survival in a controlled fashion.(10) Using our previously examined cohort we collected additional data on prescribed antidepressant medication and examined its association with long-term survival. We hypothesized that the association of early post-transplant depression with decreased survival would be attenuated by adequate, early treatment of depressive symptoms.
This cohort is derived from a previously reported study of depressive symptoms in the first post-transplant year.(2) The parent study was a prospective, longitudinal cohort study in which liver transplant (LTX) recipients with alcohol-related liver disease (ALD) were recruited from the University of Pittsburgh Starzl Transplant Institute (STI) between May 1998 and July 2003. The pre-transplant diagnosis of ALD was determined by hepatologist, surgeon, and psychiatrist interviews wherein the definition of alcohol dependence or abuse was based on structured psychiatric clinical interviews using the DSM IV criteria. From a cohort of 177 eligible recipients, 173 completed the depression measures to be included in the study. As the present study examined survival following the first post-LTX year during which depression trajectories were established, those who died within the first year (n=6) were not included in the analyses leaving a final cohort of 167.
Using the Beck Depression Inventory (BDI), depressive symptoms were assessed prospectively and repeatedly across four time points every three months for each patient in the first year after LTX. The BDI is a commonly used self-report instrument assessing 21 symptoms of depression. The measure yields a summary score reflecting the severity of symptoms, which was further characterized as low (0–9.5), moderate (9.51–16.5), and high (≥16.51) using standard thresholds. (11)
Dose and duration of antidepressant medication use in the first post-transplant year were ascertained by retrospective chart review. Patients were followed closely by the clinical team during the first post-LTX year. Patients had weekly appointments for the first 2–3 months after transplant hospitalization discharge and then were seen every 1–2 months thereafter and at one year post-LTX. At each appointment their medication list was reviewed and updated in the electronic medical record, wherein a start and stop date for all medications including dates of dosage changes were recorded. This provided detailed assessment of medication usage in the first post-LTX year. The Antidepressant Treatment History Form (ATHF) was used to rate the adequacy of antidepressant treatment. This instrument has been used to rate the adequacy of treatment trials for antidepressant medications.(12, 13) Adequacy scores range from 1–5, based on medication dose, which is determined for each medication based on the minimal dosage found to be effective in randomized control trials, and duration, with ≥4 weeks considered as adequate. A score of ≥3 defines adequate antidepressant therapy.
Demographic variables including race, marital status, education, and employment, were obtained by self-report. Age, gender, medical variables (donor age and HCV infection) and date and cause of death were obtained by review of the medical records. Patients with allograft failure contributing to their death were considered to have graft failure. Those with graft failure followed by an acute event such as an infection were considered to have infection and graft failure as causes of death. Alcohol use in the first year was ascertained using timeline follow-back (14), a validated calendar method of data collection. These data were included as a continuous variable (standard drinks in the first post-transplant year) and a categorical variable (any alcohol use in the first post-transplant year). Model for End-Stage Liver Disease (MELD) score was calculated at the time of transplant. Employment was categorized by Hollingshead Index of Social Position Occupation Scale (15) with a cutoff of 5 used to define “professional employment”. The Charlson comorbidity index is calculated using 22 comorbid medical conditions and is based on the relative risk for one-year mortality of each.(16) For this analysis we created a score including the comorbidities of recipients through the first year of post-LTX follow-up.
We collected results on all liver enzymes (alanine aminotranferease (ALT), aspartate aminotransferase (AST), and gamma-glutamyltransferase (GGT)) out to the furthest post-LTX time point for each participant. Transplant recipients undergo weekly lab testing for the first 3 months following transplant, then monthly to one year and then every 2–3 months thereafter. Using standard normal values at our chemistry lab (ALT <63 IU/L, AST <41 IU/L, GGTP <41 IU/L and creatinine < 1.5mg/dL) we calculated the percentage of each individual’s lab results that were abnormal. Patients underwent biopsies based on clinical necessity. We similarly calculated the percentage of biopsies with acute rejection or as having steatohepatitis based on assessment by an expert pathologist in the STI Transplant Pathology department. Each biopsy is rated on specific features including the presence of fat, bile duct and portal inflammation, venous endothelial inflammation, and lobar architecture and inflammation. Acute rejection required both a pathologist’s descriptive report of acute rejection and a Rejection Activity Index score ≥ 3.(17) For the purposes of this analysis we considered as a cutoff those with ≥50% of laboratories or ≥20% of biopsies with these findings to be abnormal. Recipients were followed from the time of transplant (1998–2003) until March 30, 2012.
Details of the analysis used to obtain the three depression trajectory groups were previously reported (2) but are briefly summarized here. BDI scores for each recipient at each of the four time points were first categorized by commonly used clinical cutpoints as low (0–9.5), moderate (9.51–16.5) or severe (≥16.51). We then used hierarchical agglomerative cluster analysis with these data to identify groups of individuals who clustered along similar trajectories of symptom severity over the year. The resulting three trajectories were: consistently low depressive symptoms, depressive symptoms that rose from low to moderate/high levels at the end of the first year, and consistently high BDI scores throughout the year.(2) Because in our prior analysis the high and rising depressive symptoms groups represented those at significant risk for increased mortality with similar survival curves, these groups were combined into one group for the purposes of the present analysis. We then determined whether these individuals received adequate antidepressant treatment in the first year based on the ATHF. These treatment groups were compared to the remaining individuals who had consistently low depressive symptoms throughout the first year post-transplant.
Statistical analyses were completed using R version 2.14.0. Baseline characteristics were compared between the three groups (non-depressed, depressed on adequate medications, depressed on inadequate or no medications). ANOVA was used to compare continuous variables and Chi-square and Fisher’s exact tests were used to compare categorical variables. Cox proportional hazards modeling was used to compare recipients with adequately treated depression symptoms and those with inadequately or untreated depression symptoms to those with low depression scores, controlling for variables in the univariate model with p≤0.2 including age, marital status (married vs. not), hepatitis C viral status, and Charlson comorbidity scores. Alcohol use was used in the model first as a categorical variable (any alcohol in the first post-transplant year). Results were checked for robustness by then repeating the modeling substituting number of standard drinks per year. MELD at transplant and donor age were included in the model because they have been shown to be important to survival outcomes in past literature, though they were not significant in univariate testing. Recipients were censored at the time of death or the end of follow-up. A final, parsimonious model was based on the step AIC command, which uses automated backwards and forwards elimination to optimize Akaike Information Criterion scores. A Wald test-derived p-value was reported for the overall variable of depression treatment, with individual level p values corrected with the Holm-Bonferonni correction for multiple comparisons. Models were checked for multicollinearity using a pre-specified variance inflation factor of 5 as a cutoff.
Survival curves were created in Statistical Package for the Social Sciences (SPSS) software using the full model. Causes of death were compared for the three groups, and Chi-square and Fisher’s exact tests were used for comparisons between the three groups based on cell counts. For graft-failure-related death, an exploratory competing-risks regression model was made using depression treatment status and HCV as the independent variables and death from other causes as a competing risk. The numbers of outcomes were too small to control for other variables.
The 167 participants were followed for a median of 9.5 years (interquartile range, 5.5 to 11.1 years). Overall, the mean age was 51.7±7.9, and the group was 95% Caucasian and 84% male. High (n=31) or rising (n=41) BDI scores were found in 72 LTX recipients in the first post-transplant year. Of these, 43% (n=31) received adequate treatment and 57% (n=41) received inadequate (n=7) or no treatment (n=34). In terms of factor analysis of the BDI in this sample, the Cronbach’s alpha in our sample across all time points was 0.86–0.95, demonstrating very good internal consistency between the affective and physical symptoms.
Table 1 illustrates the differences in baseline demographics by the three study groups. Those with adequate treatment of depression were younger (48 vs. 53 and 51) the other groups. Both groups of depressed patients had higher rates of hepatitis C (68% and 51% vs. 38%) compared to non-depressed. The Charlson comorbidity scores were lowest for non-depressed and highest for the inadequately/untreated depressed group. Those with any depression were more likely to have alcohol use in the post-transplant period, though this was not statistically significant. Those with high BDI scores were significantly more likely to be treated adequately than those with rising BDI scores (65% vs. 27%, p=0.003).
By the end of follow-up only 32% of the untreated/inadequately treated depressed group survived compared to 52% of the depressed adequately treated group and 56% of the non-depressed group. There was a significant overall difference in survival rates between the three study groups in the uncontrolled Cox proportional hazards model (Wald test, p=0.006). Pairwise comparisons of each depressed group to the non-depressed group demonstrated no difference in survival times for the adequately treated group (HR=1.27, CI=0.70, 2.29, p=0.43) but significant differences for the inadequately treated group (HR=2.19, CI=1.35,3.53, uncorrected=0.0014, corrected p=0.003). In multivariable modeling, compared to the non-depressed group, those with adequately treated depression had no significant difference in time to death. However, recipients with inadequately treated depression had decreased survival times in the full (HR=2.37, 95% CI=1.40, 4.01) and final models (HR=2.44, 95% CI=1.45, 4.11) when compared to the non-depressed group. Recipient age and Charlson comorbidity score were the other significant covariates in the model. Figure 1 shows the survival curves based on the Cox proportional hazards model. Results did not change significantly if drinks in the first year was included as a continuous variable in the model rather than using alcohol in the first year as a categorical variable.
Both groups of recipients with depression had more deaths related to graft failure than non-depressed recipients (60% and 43% vs. 29%), as shown in Table 3. Those with treated depression had increased deaths due to graft failure compared to those with untreated depression, but once controlling for HCV in the competing-risks regression model, compared to the non-depressed group the hazards ratios for graft-failure-related death in the untreated (HR=2.34, CI=0.97,5.64) and treated depression (HR=1.97, CI=0.78,5.01) trended towards but did not reach statistical significance. HCV in this model had a hazard ratio of 5.30 (CI=1.92,14.62) for graft-failure-related death. However, given the small numbers, other covariates could not be included in the model. Corresponding to this, recipients with depression had increased percentages of abnormal biochemical markers (ALT and GGT), though the percentage of biopsies showing acute rejection was not significantly related to depression treatment status (Table 3).
Table 3 also shows that the percent of deaths due to cancer was 29% in the non-depressed, 0% in the adequately treated and 21% in the inadequately treated depressed groups. However, the group with cancer death also had higher mean Charlson Comorbidity scores (8.2±2.3) than those with other causes of death (7.2±1.2) or no death (6.6±1.0) (ANOVA p=0.006), though the numbers of cancer deaths were too small to perform a competing risk assessment to determine whether depression status determined cancer death controlling for medical comorbidities.
Table 4 shows the initial medications used by the participants in the study. 74% of recipients on adequate medications were taking selective serotonin reuptake inhibitors (SSRIs).
This is the first demonstration that treating depression early in the post-LTX course is associated with improved long-term mortality. We have shown that recipients with untreated and under-treated depressive symptoms in the early post-transplant period had decreased survival, while those with treated depression had survival rates similar to those in the non-depressed group. Treatment of depression was more strongly linked to survival than traditional risk factors in this population, including MELD score, donor age, or HCV status.
Depression is common in patients with liver disease after transplant and particularly among those with alcohol-related liver disease.(18) Our rates of depressive symptomatology in the first post-transplant year are similar to those previously described.(7) Among the recipients with high or rising BDI scores, 29% had a pre-transplant history of depression compared to 17% of those with low depressive symptoms post-transplant. Thus, there are many people with de novo depressive symptoms after transplantation. Given that treating these symptoms may improve survival, these data support continually screening for depressive symptoms as a non-surgical risk factor for survival in the post-transplant period. Conversely, depression is not “cured” by replacing the liver, such that many patients remain with their psychiatric disease and require close follow-up.
While it is logical to expect that antidepressant medications would modify the deleterious effects of depression on survival, controversy exists regarding the associations of antidepressant medications with morbidity and mortality in general. For example, antidepressant use in postmenopausal women has been found to increase the risk of all-cause mortality even when adjusting for depression severity.(19) While there is little work on the effects of antidepressant therapy and outcomes in liver disease, cardiac outcomes have been better investigated. Initial studies found increased myocardial infarction (MI)(20, 21) among patients on antidepressants. However a more recent study, which assessed adequacy of treatment, found that patients with adequate (versus inadequate) antidepressant duration had decreased MI and all-cause mortality.(22) Accordingly, one recent study found that the highest risk of mortality after MI was in patients with insufficiently treated depression compared to those on adequate treatment.(23) Similarly, among people with congestive heart failure, antidepressants were associated with increased death and hospitalization in one study,(24) but in other studies mortality (5, 25) and worsened cardiovascular outcomes (26) were not found to be increased after controlling for severity of depression or randomizing the patients to the medications vs. placebo. Notably, one of these studies still found increased mortality associated with fluoxetine prescriptions.(5) It is likely that much of the discrepancy in terms of the evidence regarding the safety/efficacy of antidepressant medications relates to not accounting for severity of depression or adequacy of medication dosing or response. Though the effects of antidepressant medication use on mortality remain controversial, in our cohort we found improved overall survival with medication doses when accounting for adequacy of medication dosing and duration.
The mechanism by which antidepressant medications could improve survival after liver transplantation is unknown, but several possibilities exist. Treating depression may ameliorate depression-related weight gain and physical inactivity, thus aiding in rehabilitation postoperatively. Treating depression may also improve motivation and adherence to complex anti-rejection medication regimens and post-operative follow-up. Additionally, there may be biological underpinnings to the relationship between the treatment of depression and survival. Activation of the immune system and high levels of proinflammatory cytokines, which are found in untreated depression, are thought to be modulated by antidepressant medications(27) and may allow for decreased rejection. Specifically, IL-1-β, TNF-α, IL-6, c-reactive protein (CRP), and NF-κB are all elevated in depression and are thought to decrease with antidepressant medications, in part related to the medication’s effects on melatonin.(27) NF-κB in its role as a transcription factor for inflammatory cytokines has specifically been related to acute allograft rejection,(28) and high CRP expression in post-perfusion liver biopsies predicted early graft dysfunction.(29) Thus, modulating pro-inflammatory cytokines via antidepressants may directly decrease acute rejection. Additionally, serotonin may act as a growth factor for hepatocytes.(30) Selective serotonin reuptake inhibitors (SSRIs) work through receptor HT-2b,(31) and activation of this receptor in murine models has been shown to improve graft survival in small-for-size models of liver transplantation.(30)
Hepatitis C was more common in the two depressed groups and was a common cause of death, which is consistent with literature on the increased burden of mental illness in HCV. HCV rates were highest in the group with adequate treatment of depression, possibly reflecting the increased screening and more aggressive treatment of these individuals for mental illness given their need for antivirals. Interestingly, however, HCV was not a significant predictor of long-term overall survival, in contrast to treatment of depression. It may be that HCV treatment-related variables are important predictors of overall survival, however, due to the small sample size we were unable to control for various treatment-related variables. This will likely be important for future studies. However, HCV was, as expected, a critical predictor of specifically graft-failure-related death. We have also found in this study a trend towards increased graft-failure-related death with untreated depression, and these factors may be interrelated. Antidepressants are generally thought to help adherence to HCV regimens pre-transplant without change in rates of viral responsiveness.(32) We previously found (in a different population) that LTX recipients on antidepressants at the time of transplant had significantly less acute rejection compared to unmedicated depressed recipients, (10) supporting the notion that antidepressants may help prevent graft failure through several pathways.
There are several limitations of this study. All patients had alcoholic liver disease ± HCV, so the results may not be generalizable to other types of liver disease, particularly given the strong association between alcohol-related liver disease and mood disorders.(18) One recent report showed depressive symptoms among 54% of those with alcohol-related vs. 27% of non-alcohol-related cirrhosis among LTX candidates.(33) Additionally, alcohol and drug recidivism are both difficult co-factors to measure and are key to distinguishing whether recidivism underlies depression and the associated poor survival. However, our validated measures of alcohol use in the first year after transplant did not impact survival in this cohort, while depression and treatment did. For the purposes of this study, we combined all antidepressant medications into one category. Though different medications have different mechanisms of action and different drug-drug interactions, the vast majority used SSRIs (75% of the adequately treated depressed group). We also did not have information regarding medication adherence and were unable to assess other treatment modalities including psychotherapy that may have affected our results. Outside of research protocols, receipt of psychotherapy can be difficult to standardize and quantify but is an important consideration for future study. Beyond the first year, we did not have assessments of depression trajectories or adequacy of treatment of depressive symptoms. However, these results are notable in that depressive symptoms and their treatment in the first year alone can predict long term results. We assessed adequacy of medication treatment rather than response to treatment. There has been controversy as to the ideal doses of the antidepressant medications, and this may be particularly true in the case of patients with hepatic and renal disease whose doses may require reduction due to metabolic alterations. Future studies are needed to determine the ideal manner in which to treat depression in a post-LTX cohort. Given that causality cannot be inferred from observational studies findings will require replication in a randomized, controlled trial of different treatment modalities in order to assess how to optimize the survival benefits of treating depression. While any observational study carries the risk of bias, we adjusted for the Charlson comorbidity score and a validated measure of alcohol recidivism in addition to standard LTX variables and still found a strong and significant negative impact of untreated depression. In order to understand the mechanism behind the observed improvement in survival with antidepressant medications, serial depression inventories over years will be needed to determine whether the survival benefit relates to resolution of depression or the medications themselves.
Depression is associated with decreased long-term post-LTX survival, and we have shown that it is truly those with inadequately treated depression who account for the decreased survival, while those on adequate dosages of antidepressant medications have no change in survival from the non-depressed group. Remarkably, depression treatment status proved to be more strongly associated with survival than MELD, donor age, or HCV status, traditional risk factors for mortality in this population. This suggests that depression is a modifiable non-surgical, non-hepatologic risk factor for poor LTX outcomes and that may be improved by early identification and treatment.
Shari Rogal was supported by NIH-T32 grant number DK063922. The project described was supported by the National Institutes of Health through Grant Numbers UL1 RR024153 and UL1TR000005. This research is funded by grant numbers K23-AA0257 from the National Institute of Alcohol Abuse and Alcoholism and R01 DK066266 from the National Institute of Digestive Disorders and Kidney Diseases Rockville, MD, USA.
The authors have nothing to disclose.