1. Top of page
  2. Abstract

Calcineurin-inhibiting immunosuppressive medications are the mainstay of posttransplant immunosuppression. Although these highly beneficial drugs are critical for posttransplant survival, significant numbers of transplant recipients experience side effects, some requiring a switch to a different immunosuppressive regimen. Neurotoxicity is one of the most debilitating side effects because of its impact on mental status and cognition. As our center uses tacrolimus as the initial immunosuppressant for all liver transplant (LTX) recipients, we were interested in those patients who required a switch because of neurotoxic side effects. Over a 5-year period, 827 adult LTX recipients received their first graft at our center. Ninety-four patients were no longer on tacrolimus by 2 months post-LTX (86 switched because of concerns over neurotoxicity, and 8 switched because of renal function concerns). Of those experiencing neurotoxic side effects, the majority (64%) had altered mental status, and 26% had seizures (first onset post-LTX). On the basis of our prior work, we hypothesized that patients with a pre-LTX history of excessive alcohol use would be at higher risk for neurotoxic effects. We also hypothesized that the elderly and those who had more advanced illness (that is, higher Model for End-Stage Liver Disease scores) at LTX would be at risk as well. We found that patients with a pre-LTX diagnosis of alcoholic liver disease were not more likely to be switched from tacrolimus. Furthermore, we found that in addition to older age and higher Model for End-Stage Liver Disease scores, poorer hepatic functioning was significantly associated with a switch from tacrolimus. We discuss the implications of these findings and the relevance for future clinical care in these high-risk patients. Liver Transpl, 2008. © 2008 AASLD.

Until clinically effective tolerance-induction protocols can allow substantial diminution of or liberation from chronic nonspecific immunosuppression for transplant recipients,1 the toxic side effects of calcineurin-inhibiting immunosuppressive medications (tacrolimus and others) will continue to be a clinical challenge. Neurotoxicity is one of the more disturbing of these side effects because of its potential impact on mental status and cognition. Calcineurin-induced neurotoxicity can be severe and debilitating, requiring a mandatory switch to an alternative immunosuppressive medication. This complication-driven medication switch in the recipient's primary immunosuppressive therapy can lead to subsequent acute cellular rejection episodes and eventual allograft dysfunction. For most transplant recipients, neuropsychiatric symptoms of tacrolimus are mild and include tremulousness, headache, restlessness, insomnia, vivid dreams, hyperesthesias, anxiety, and agitation.2 The moderate to serious neuropsychiatric side effects (that is, cognitive impairment, coma, seizures, focal neurologic deficits, and delirium) occur less often but can reach 21% to 32% in the early postoperative period.3

Our prior investigations at the Starzl Transplant Institute have helped to identify the profile of neuropsychiatric side effects of tacrolimus and have informed our hypotheses about recipients potentially at risk for neurotoxicity. In one study of 294 consecutive transplant recipients on tacrolimus (238 liver, 53 heart, 2 double lung, and 1 heart-lung), those with preexisting central nervous system (CNS) damage (for example, strokes and multiple sclerosis) were at higher risk for developing neurotoxic side effects.4 From this, it was hypothesized that more serious neurotoxic side effects (focal neurologic abnormalities, speech disturbances, hemiplegia, and cortical blindness) may occur from higher CNS levels of tacrolimus in patients who have a disrupted blood-brain barrier.4 In addition, liver transplant (LTX) recipients appeared to be more susceptible to neurotoxic side effects than heart and lung transplant recipients.4

In a randomized controlled trial of tacrolimus versus cyclosporine in LTX recipients, we identified that in the early postoperative period, 35% of the recipients experienced some cognitive impairment (based on neuropsychiatric testing scores), with 29% of those on tacrolimus experiencing moderate to severe cognitive impairment.5 Interestingly, we also found that those in the tacrolimus group with a pretransplant diagnosis of alcoholic cirrhosis appeared to be at higher risk for neuropsychiatric symptoms.6 Nevertheless, a large-scale study investigating multiple predictors of tacrolimus neurotoxicity has not been done.

Therefore, we hypothesized that LTX recipients would be at higher risk for tacrolimus neurotoxicity if they had possible preexisting CNS damage due to heavy alcohol exposure [that is, those with a pre-LTX primary or secondary diagnosis of alcoholic liver disease (ALD)], a finding that we have preliminarily noted.6 Additionally, we considered 2 other variables possibly representing a vulnerable CNS: (1) pre-LTX Model for End-Stage Liver Disease (MELD) scores, given that elevated MELD scores have been associated with post-LTX mental status changes,7 and (2) advanced age, which has been associated with impaired CNS “homeostatic reserve.”8 Because all adult LTX recipients at the Starzl Transplant Institute receive tacrolimus as their primary immunosuppressant (that is, first choice after LTX), we were able to test our hypothesized predictors in a large cohort by identifying those LTX recipients switched off tacrolimus because of neurotoxicity.


  1. Top of page
  2. Abstract

All adult LTX recipients who received their first liver allograft over a 5-year period from June 1, 2001 to May 31, 2006 and survived 2 months post-LTX were included in the study (n = 827). Immunosuppressant medications at 2 months post-LTX were noted, and patients in whom tacrolimus was discontinued because of neurotoxicity were identified. Our primary categorical outcome in this study was simply whether a patient remained on tacrolimus or required a change to another immunosuppressant because of neurotoxicity. In addition, the clinical features of the neurotoxic symptoms were retrospectively analyzed through the medical records as well as documentation of electroencephalogram (EEG) results and brain imaging [computed tomography (CT) and/or magnetic resonance imaging (MRI)] when done. The initial 2-month postoperative period was chosen as the time frame for identifying the primary immunosuppressive medication, as clinical evidence of serious tacrolimus neurotoxicity necessitating a switch would usually develop in the first weeks post-LTX.9, 10 Demographic and medical history variables were also collected from the medical records. In addition, specific laboratory values of interest were collected, including tacrolimus blood levels, sodium, magnesium, total bilirubin, and creatinine. Because patients' blood samples were checked repeatedly, we chose to calculate the median value for each biochemical measure (either up until the switch from tacrolimus or throughout the total 2-month period for those not switched) as the estimate of central tendency. Finally, we collected data on medical outcomes to compare the 2 groups, including primary nonfunction, renal failure requiring dialysis, and the times to acute cellular rejection, retransplantation, or death.

Statistical Analysis

Continuous variables are presented as the mean ± standard deviation, and categorical variables are presented as proportions. Independent t tests were used to test continuous variables, and the chi-square test was used for categorical variables. The time to specific medical outcomes was calculated by Kaplan-Meier survival analyses and compared by the log-rank test. MELD scores were log-transformed prior to analysis because of positive skewness. Untransformed scores are presented in the tables and figures for clarity. We chose median values for the biochemical tests rather than mean values, which could be affected by extreme values and not indicate central tendency.

To test our first hypothesis, we compared the proportion of patients switching immunosuppressants among those with a pre-LTX primary or secondary diagnosis of ALD versus patients with all other liver diagnoses. We then examined the proportions switching medications among patients with other common liver disease diagnoses [that is, hepatitis C virus (HCV)] compared to all others. In addition to age and MELD score at LTX, we also examined gender, race, date of LTX, and biochemical values as potential predictors of our primary outcome in an exploratory univariate analysis. Colinearity was checked between variables, and the final model was tested by logistic regression using variables that had achieved P < 0.05 at the univariate level. Differences between groups on specific medical outcomes were also calculated. All analyses were performed with Statistical Package for Social Sciences version 12 for Windows.


  1. Top of page
  2. Abstract

Cohort Characteristics

All 827 adult LTX recipients received tacrolimus as their primary immunosuppressant. By 2 months post-LTX, 86 recipients (11%) had been removed from tacrolimus because of neurotoxic side effects. Eight patients were switched from tacrolimus because of concerns over renal toxicity and were not included in the analyses. The majority of patients (93%) were switched to cyclosporine as their primary immunosuppressant medication. Six patients received various combinations of sirolimus, mycophenolate, or azathioprine. Twelve patients (14% of those switched) were eventually switched back to tacrolimus anywhere from months to years later. The cohort was predominately Caucasian (90%) and male (61%), with ages ranging from 18 to 78 years of age (mean 54 ± 11). MELD scores at LTX ranged from 6 to 40 (mean 16 ± 7). One hundred twelve recipients received a living liver donation; 707 recipients had deceased donor grafts. Thirty-three percent of the cohort had either a primary or secondary diagnosis of ALD; of these, 42% were additionally coinfected with HCV (Table 1).

Table 1. Total Cohort Demographics (n = 819)
  1. Abbreviations: HCV, hepatitis C virus; MELD, Model for End-Stage Liver Disease; NASH, nonalcoholic steatohepatitis; PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; SD, standard deviation.

Age (mean ± SD)54 ± 11
Gender (% male)61
Race (%) 
 Middle Eastern1.9
Married (%)71
Liver disease (%) 
 Alcohol and HCV14
MELD score (mean ± SD)16 ± 7

The majority of patients (64%) had symptoms of altered mental status (including symptoms of stupor/coma, confusion/disorientation, agitation, and/or psychosis). Twenty-six percent experienced seizures (Table 2). Of those switched from tacrolimus, 60% had an EEG performed and 79% had CT and/or MRI imaging of the brain to investigate the etiology of the mental status symptoms (see Table 3). The most common finding on EEG was generalized, diffuse slowing (65%). Only 23% showed either active or prodromal seizure activity. Twelve patients had characteristic abnormalities on radiological imaging (mostly MRI) of posterior reversible leukoencephalopathy syndrome (low attenuation of white matter on CT scan and/or corresponding hyperintense lesions on T2-weighted MRI images). Interestingly, there were 3 patients who had both MRI and CT scanning where the MRI was read by the radiologist as being consistent with posterior reversible leukoencephalopathy syndrome yet the CT scan was either read as normal or nonspecific. This suggests the better distinguishing value of MRI in identifying such radiographic changes.

Table 2. Neurotoxic Symptoms Among the 86 Patients Who Were Switched off Tacrolimus
  • NOTE: Percentages add to >100% as some patients had more than 1 symptom.

  • *

    This included stupor/coma, confusion/disorientation, agitation, or hallucinations/psychosis.

Altered mental status*6456
Tremors/involuntary movements1917
Table 3. EEG and Radiological Imaging Results Among the 86 Patients Who Were Switched off Tacrolimus
  • Abbreviations: CT, computed tomography; EEG, electroencephalogram; MRI, magnetic resonance imaging.

  • *

    This category included radiographic changes consistent with periventricular white matter changes, small vessel disease, old infarcts, or volume loss.

EEG (n = 52)  
 Generalized slowing6534
 Seizure/seizure tendency2312
 Focal or other63
CT scans (n = 53)  
 Low intensity63
MRI scans (n = 45)  
 High-intensity T2 signal2612

Associations with Tacrolimus Switch

Of our 3 primary hypotheses, a previous diagnosis of ALD was not directly associated with a subsequent switch from tacrolimus by 2 months post-LTX. However, both older age and higher MELD scores were significantly associated with switching from tacrolimus due to neurotoxicity (Table 4). The association between MELD score and tacrolimus switch was linear (see Fig. 1). For age, there was a discreet cut point such that above 60 years of age the risk increased substantially (see Fig. 2). There was a significant correlation between age and MELD score (P = 0.013), although it was small (r = −0.06). When we controlled for MELD scores, the association between age and tacrolimus switch remained, and similarly, when we controlled for age, the association between MELD scores and tacrolimus switch remained.

Table 4. Potential Predictors of Continuing on Tacrolimus Versus Switching off Tacrolimus Because of Neurotoxicity After Liver Transplantation
 Remaining on Tacrolimus (n = 733)Switched off Tacrolimus (n = 86)Test Statistic*P Value
  • Abbreviations: ALD, alcoholic liver disease; LTX, liver transplant; MELD, Model for End-Stage Liver Disease; SD, standard deviation.

  • *

    χ2 (1) was used for dichotomous variables, and a t test was used for independent variables.

  • Prior to switch for switched group.

ALD (%)32391.80.17
MELD score, mean (SD)15.6 (7.3)18.7 (8.0)3.60.000
Age (years), mean (SD)53 (11)57 (11)3.20.002
Era of LTX (days), mean (SD)911 (452)747 (517)3.10.002
Gender (% male)60600.0270.90
Race (% Caucasian)89943.00.69
Creatinine, mean (SD)1.4 (1.0)1.6 (0.9)1.80.07
Tacrolimus level (ng/mL), mean (SD)9.7 (1.6)9.5 (2.4)1.50.14
Total bilirubin, mean (SD)3.7 (4.0)5.5 (4.3)3.70.000
Magnesium, mean (SD)1.5 (0.2)1.7 (0.2)9.20.000
Sodium, mean (SD)137 (2.4)137 (3.5)0.550.58
thumbnail image

Figure 1. Percentage switched off tacrolimus by the Model for End-Stage Liver Disease (MELD) score.

Download figure to PowerPoint

thumbnail image

Figure 2. Percentage switched off tacrolimus by age.

Download figure to PowerPoint

In addition, the date of transplant was significantly associated with being switched from tacrolimus such that within the 5-year cohort, those transplanted earlier (that is, closer to 2001) were more likely to remain on tacrolimus, whereas those transplanted more recently were more likely to be switched off tacrolimus.

To further examine liver disease diagnoses, we analyzed whether those with HCV were more likely to be switched compared to all others. Although HCV was associated with a less likely chance of being switched (χ2 = 6.1, P = 0.014), the association did not remain after controlling for age, MELD score, and time of LTX.

There were also notable differences in specific laboratory values between groups (see Table 4). Because we were investigating laboratory values that would predict a switch from tacrolimus, we report the median values either up until the switch from tacrolimus or throughout the total 2-month period for those not switched. Thus, the laboratory values reflect events before the switch. We chose to investigate hypomagnesemia because of its association with tacrolimus neurotoxicity and hyponatremia because of its association with immunosuppression and central pontine myelinolysis. Although there was no difference between groups in median sodium levels, the group switched off tacrolimus had significantly higher magnesium levels (1.7 versus 1.5, P = 0.000) perhaps reflecting magnesium supplementation. In an attempt to keep patients on tacrolimus, the dosage and blood levels were initially reduced in nearly all patients except when the clinical situation was more urgent, necessitating an immediate switch. Although our immunosuppressive protocol has changed over time, tacrolimus is the mainstay of our therapy, and we strive for a target whole blood level of 8 to 15 ng/mL. Ninety percent had median tacrolimus values of 11.5 ng/mL or less, and only 5 patients had median values greater that 15 ng/mL. No patient had a median tacrolimus value greater than 18 ng/mL. Median tacrolimus levels were not different between groups (see Table 4).

We also considered acute and chronic medical factors that may have contributed to our findings. Both acute renal failure (represented by those who required dialysis within 2 months post-LTX) and chronic renal insufficiency (represented by elevated median creatinine) were examined between groups. Although the creatinine values were slightly but not significantly different between groups, patients switched off tacrolimus were significantly more likely to require dialysis within 2 months post-transplant (34% versus 22%, P = 0.02). In addition, liver function (represented by median total bilirubin levels) was significantly worse for the switched group. Older age was correlated with lower total bilirubin levels, although this association was small (P = 0.01, r = −0.09). Information on intraoperative hypotension was not available; however, we had data on the total number of packed cells used during the transplant operation on 478 patients. The median number of packed cells used was 6 units and was not different between groups.

There were also important differences in medical outcomes between groups. Patients switched off tacrolimus were slightly, although not significantly, more likely to have acute cellular rejection (37% versus 30%) by 1 year (see Fig. 3). However, they were not more likely to have primary nonfunction (1% versus 1%), require retransplantation (12% versus 9% at 5 years), or have received a donation from a living donor (14% versus 14%). Of special concern was the overall significantly poorer survival for those switched off tacrolimus beginning at the point of transplantation and continuing out to 5 years post-LTX (see Fig. 4). Although we considered the contribution of age and pre-LTX MELD score to this outcome in the switched group, in a multivariate Cox regression including the variables of age and MELD score, being switched off tacrolimus still added a significant contribution to poorer post-LTX outcomes.

thumbnail image

Figure 3. Time to acute cellular rejection.

Download figure to PowerPoint

thumbnail image

Figure 4. Survival: tacrolimus versus patients switched off tacrolimus.

Download figure to PowerPoint

We also investigated whether treatment with an anticonvulsant was associated with tacrolimus switch. De novo anticonvulsant therapy was initiated in 37 recipients by 2 months post-LTX (35 on phenytoin, 1 on valproic acid, and 1 on oxcarbazepine). Those on anticonvulsants were significantly more likely to be switched off tacrolimus (P = 0.000) and represented patients that we suspect developed neurotoxic seizures necessitating both anticonvulsant therapy and a tacrolimus switch. The majority (68%) were nearly simultaneously switched off tacrolimus and started on an anticonvulsant; therefore, a drug interaction between tacrolimus and the anticonvulsant was an unlikely cause for the switch. It is of note that 3 patients were already on cyclosporine before developing seizures and requiring anticonvulsant therapy, highlighting the neurotoxic potential of cyclosporine as well.


  1. Top of page
  2. Abstract

Although elevated blood levels of tacrolimus2, 11, 12 and electrolyte and other biochemical abnormalities3, 13–15 have been associated with neurotoxicity, these chemical derangements are correctable and often do not require discontinuing tacrolimus. Thus, our hypotheses were based on the premise that beyond reversible chemical derangements, core characteristics exist that can predict tacrolimus neurotoxicity and hence the need to switch to a different primary immunosuppressant. In addition, our group had previously noted that neurotoxicity is most likely multifactorial.9 By identifying these risk factors, we may be able to predict and possibly prevent this outcome. Furthermore, these results may lead to the development of studies on the potential mechanisms of tacrolimus-induced neurotoxicity.

As we hypothesized, older age and higher MELD scores at LTX did predict those who required a discontinuation of tacrolimus. Increased age is associated with increased sensitivity to a number of neuronally active medications.16 For several reasons, including loss of neurons, altered receptor levels, impaired adaptability, and decreased homeostatic reserve, the older brain has elevated vulnerability to the effects of pharmacologic challenges.8 Additionally, increased susceptibility to blood-brain barrier disruption is evident in the aged brain.17 Small changes associated with cerebrovascular disease, such as small lacunar infarcts, may further impair the blood-brain barrier.18

Neuropsychologic deficits have previously been found to be higher in liver disease patients with elevated pre-LTX MELD scores.19 In a cohort of 80 LTX recipients, MELD scores greater than 15 predicted post-LTX mental status changes occurring within the first postoperative month and lasting more than 3 days.7 The mechanistic pathway resulting in these neural deficits is not clear. However, our results suggest that elevated pre-LTX MELD scores are additionally associated with increased vulnerability to post-LTX tacrolimus neurotoxicity. These findings are especially important given the progressive trend toward a recipient population that is older and has higher MELD scores than in the last decade.20

Our prior work in a small cohort of subjects6 also led us to hypothesize that patients with heavy pre-LTX alcohol exposure would have developed “vulnerable” brains and be at higher risk for tacrolimus neurotoxicity. Although we did not examine neurotoxicity directly, we did not find evidence in support of this hypothesis. However, because alcohol differentially affects different organ systems (that is, the CNS and the liver), using the diagnosis of ALD as a proxy for alcoholic brain damage may have masked a potential association between alcohol-related structural or microvascular brain abnormalities and a potential for tacrolimus neurotoxicity. A much more sophisticated investigation, optimally using structural or functional imaging of the brain, may be required to resolve this question.

Although not considered as a priori hypotheses, we found poorer post-LTX hepatic functioning distinguished those switched from tacrolimus. Those switched had significantly higher bilirubin levels and slightly but not significantly higher creatinine levels and were more likely to be on dialysis within 2 months following LTX. As we had removed from the data set those switched from tacrolimus because of concerns over renal toxicity, the patients in our analyses represented those switched because of neurotoxicity. Thus, in addition to pre-LTX medical debility (that is, pre-LTX MELD scores), ongoing post-LTX hepatic and renal insufficiency may identify those prone to tacrolimus neurotoxicity; this finding was noted in a previous study from our center.9

One additional finding is that our prescribing practices may have changed over time. More recently, tacrolimus discontinuation is more common, and this suggests a current lower clinical threshold to switch. The recent availability of other powerful, potentially less neurotoxic non–calcineurin-inhibiting immunosuppressants (that is, sirolimus) may have provided an adequate alternative to tacrolimus as the primary immunosuppressant and a more recent impetus to switch. Nevertheless, for the majority of our patients, tacrolimus was replaced with cyclosporine, which has a similar neurotoxic profile.21 As we found, even after the switch to cyclosporine, 3 patients developed new onset seizures.

A striking finding was the poorer survival in the group switched off tacrolimus. Although this group was older and had higher pre-LTX MELD scores, this did not fully explain their poorer outcomes. Perhaps some other medical factors that we did not explore resulted in their being switched off tacrolimus and also would explain their poorer survival. Nevertheless, this underscores the overall higher risk in this medically compromised group.

There are several limitations to the present study. The retrospective design prevented us from characterizing the severity of the neurotoxic events necessitating the switch from tacrolimus. Thus, the severity of the events may have varied, and the decision to switch may have differed by clinician. Nevertheless, that 26% of those switched also required the addition of an anticonvulsant suggests that the neurologic events were significant and often severe. In addition, our clinical team has been stable for years, and protocols for clinical care are routine. As this was a retrospective study, we also lacked detailed information on our patients' past medical histories, information that may have further clarified the predictors that we found to be significant. For example, pretransplant hepatic encephalopathy, a prior history of a stroke, or a neurodegenerative disorder would be expected to contribute to differences between groups and may have further explained the association of older age with those requiring a switch. Finally, our choice of median laboratory values to represent the central tendency was an attempt to avoid the influence of laboratory value outliers on mean values. However, we realize that median values may not fully represent the biochemical influences either at the point of immunosuppressant medication switch or throughout the 2-month period. These limitations notwithstanding, we have identified key variables predicating those at risk for being intolerant of tacrolimus.

Even with these considerations, this study is valid as an initial attempt to assess the variables related to the incidence of clinically relevant tacrolimus neurotoxicity in liver allograft recipients. The importance of these parameters is especially relevant to the current clinical arena, where older patients with higher MELD scores are undergoing liver transplantation. The findings of this retrospective study emphasize the need for increasing awareness of this neurotoxicity-prone population, who should be carefully considered as a high-risk group for this undesirable calcineurin-inhibiting immunosuppressant–driven complication. For our future clinical care, these high-risk patients should be carefully monitored for potential neurotoxic side effects, and a lower threshold to switch should be considered, especially when mental status changes persist post-LTX.


  1. Top of page
  2. Abstract