Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Potential conflict of interest: Nothing to report.
This article first published online on January 3, 2011; the abstract has since been updated. The correct version appears in print.
The diagnosis of hepatic encephalopathy (HE) relies on clinical, neurophysiological, psychometric and laboratory variables. The relationships between such tests remain debated. The aim of this study was to determine the laboratory correlates/prognostic value of neurophysiological/psychometric abnormalities in patients with cirrhosis. Seventy-two patients and 14 healthy volunteers underwent EEG and paper-and-pencil psychometry (PHES). Blood was obtained for C reactive protein (CRP), interleukin 6 (IL6), tumor necrosis factor (TNF)α, ammonia and indole/oxindole. Patients were followed prospectively for a median of 22 months in relation to the occurrence of death, transplantation and HE-related hospitalizations. Thirty-three patients had normal PHES and EEG, 6 had abnormal PHES, 18 abnormal EEG and 13 abnormal PHES and EEG. Patients with abnormal PHES had higher CRP (17 ± 22 vs 7 ± 6, P < 0.01), IL6 (32 ± 54 vs 12 ± 13, P < 0.05) and TNFα (17 ± 8 vs 11 ± 7, P < 0.001) levels than those with normal PHES. Patients with abnormal EEG had higher indole (430 ± 270 vs 258 ± 255, P < 0.01) and ammonia (66 ± 35 vs 45 ± 27, P < 0.05) levels than those with normal EEG. Psychometric test scores showed significant correlations with CRP, TNFα and IL6; EEG indices with ammonia and IL6. CRP and TNFα concentrations were independent predictors of abnormal PHES, ammonia and indole of abnormal EEG on multivariate analysis. Seven patients were lost to follow-up; of the remaining 65, 20 died and 14 underwent transplantation; 15 developed HE requiring hospitalization. PHES and EEG performance were independent predictors of HE and death (P < 0.05). Conclusion: PHES and EEG abnormalities in patients with cirrhosis have partially different biochemical correlates and independently predict outcome. (HEPATOLOGY 2011;53:558-566)
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The diagnosis of hepatic encephalopathy (HE) relies on clinical, neurophysiological, and psychometric variables. Patients with severe overt HE almost invariably exhibit both neurophysiological and psychometric abnormalities, whereas more compensated patients can present with isolated psychometric deficits or electroencephalogram (EEG) slowing.1, 2
The pathogenesis of HE is only partially understood, but there is general consensus that it is due to impaired hepatic clearance of gut-derived neurotoxins, because of hepatocellular failure and/or portal-systemic shunting. Several neurotoxins have been implicated, including ammonia,3 the tryptophan derivative indole, and its tissue metabolite oxindole, which is believed to have direct sedative effects.4 More recently, it has been suggested that inflammation may also play an important role.5 Infection has been recognized as a precipitating factor for HE for some time6; lipopolysaccharides have been shown to enhance ammonia-induced changes in cerebral hemodynamics in animal models,7 and markers of a systemic inflammatory response have been related to the presence of neuropsychiatric impairment in patients with both acute and chronic liver failure.5, 8
However, the relationship between the behavioural/neuropsychiatric features of HE and the circulating levels of substances which have been implicated in its pathogenesis has generally been deemed poor and remains debated.
The aims of this study were: (1) to determine the relationship between psychometric/EEG abnormalities and blood levels of ammonia, indole, oxindole, and a set of markers of the activated inflammatory cascade in a group of patients with cirrhosis with no or grade I overt HE; and (2) to determine the prognostic value of psychometric, EEG, and HE-related laboratory abnormalities in relation to the subsequent development of HE-related hospitalizations and death.
CRP, C-reactive protein; EEG, electroencephalogram; HE, hepatic encephalopathy; IL-6, interleukin-6; MDF, mean dominant frequency; MELD, model for end-stage liver disease; PHES, psychometric hepatic encephalopathy score; TNFα, tumor necrosis factor α.
Patients and Methods
The patient population comprised 72 consecutive outpatient attendees (49 men, 23 women; age, 54 ± 9 years [mean ± SD]) with established cirrhosis, who had been followed up regularly and had already experienced at least one episode of hepatic decompensation (advanced disease). The diagnosis of cirrhosis and its etiology had been determined by use of clinical, laboratory, radiological, and, where needed, histological variables. The functional severity of the liver disease was assessed using the Child-Pugh grading system9 and model for end-stage liver disease (MELD) score.10 Patients were excluded if they were <20 years or >80 years of age, could not comply with the study procedures, had misused alcohol in the preceding 6 months, had a history of significant head injury, cardiovascular/cerebrovascular disease or significant neurological/psychiatric comorbidity, were taking neuroactive drugs, had overt grade II or higher HE according to the West Haven criteria11 or symptoms/signs of infection.
The reference population comprised 14 healthy volunteers with a mean age of 51 ± 15 years. None of the volunteers drank alcohol in excess of 20 g/day or were taking prescription medication.
Neuropsychiatric assessment was conducted in one morning session after breakfast. Brief rest breaks were offered between tests.
Each patient's mental status was assessed by an experienced physician (S. Montagnese, A. Biancardi, or P. Amodio) prior to the psychometric/neurophysiological evaluation. The assessment included: (1) a detailed and comprehensive medical history wherein evidence was sought for changes in memory, concentration, attention, and vigilance and in the ability/modality of approaching the activities of daily living; (2) a comprehensive neurological examination, looking particularly for evidence of subtle motor abnormalities, including hypomimia, dysarthria, increased tone, reduced speed, and difficulty in executing rapid alternating movements and tremors, especially asterixis; (3) exclusion of concomitant neurological disorders (e.g., subdural hematoma, Wernicke's encephalopathy) or other metabolic encephalopathies (e.g., those associated with glucose or electrolyte imbalance, thyroid dysfunction, renal failure, and intoxication with drugs or alcohol); and (4) a clinical grading of the neuropsychiatric abnormalities according to the West Haven criteria.11 Patients were finally qualified as having or not having grade I overt HE and were excluded from the study if they had overt HE of grade II or higher.
Psychometric performance was assessed, under standardized conditions, using number connection tests A and B, the digit symbol subtest of the Wechsler adult intelligence scale, and the line tracing and serial dotting tests.12 Individual psychometric test results were scored in relation to age- and education-adjusted Italian norms.13 Psychometric performance was classified as impaired if the sum of the standard deviations for the individual tests, known as the psychometric hepatic encephalopathy score (PHES), was ≤ −4.13
EEGs were recorded for 10 minutes, with eyes closed, in a condition of relaxed wakefulness, using a 21-electrode EEG cap. Electrodes were placed according to the International 10-20 system; the ground electrode was Fpz; the reference electrode was Oz; impedance was kept below 5 kΩ. Each channel had its own analogue-to-digital converter; the resolution was 0.19 μV/bit (Brainquick 3200, Micromed, Italy equipment). One continuous 80-100 second period of artifact-free EEG tracing was selected for subsequent spectral analysis by Fast Fourier Transform. The following spectral parameters were calculated on the P3-P4 derivation: the mean dominant frequency, which is an estimate of the background frequency of the EEG, and the relative power of the spectral bands delta (1-3.5 Hz), theta (4-8 Hz), alpha (8.5-13 Hz) and beta (13.5-25.5 Hz). EEGs were classified as normal/abnormal based on the spectral criteria proposed by Van der Rijt et al.14 and subsequently modified by Amodio et al.15
Patients were qualified as having minimal HE if psychometric and/or neurophysiological abnormalities were present.
Venous blood was obtained for routine full blood count, renal function and electrolytes, protein profile, glucose, aminotransferases, bilirubin, clotting screen and C-reactive protein (CRP); ammonia was measured in the emergency laboratory immediately after blood had been drawn in an iced tube. Serum was frozen at −20°C for subsequent measurement of interleukin-6 (IL-6), TNFα, indole, and oxindole.
IL-6 and TNFα were measured using a solid-phase immunological method with two antibodies: a monoclonal immobilizing murine antibody and a polyclonal enzyme-labeled (bovine alkaline phosphatase) antibody. The system was coupled in a chemoluminescent sequential immunometric assay, and the method was automated in a DPC Immulyte One analyzer (Medical Systems, Genova, Italy).16 The results are not confounded by hemolysis or by bilirubin and lipids in clinical sample concentrations. Interference by heterophilic antibodies was ruled out by standard laboratory procedures. The intra-assay coefficient of variation (CV) was 3.5-4% for IL-6 and 2.6-3.5% for TNFα; the inter-assay CV was 5.1%-5.3% for IL-6 and 5.3%-6.5% for TNFα; the limit of detection was 2 μg/L for IL-6 and 1.7 μg/L for TNFα.
Indole plasma concentrations were measured using a procedure based on high-performance liquid chromatography separation and fluorescence spectrophotometer detection as previously described.17 Briefly, 2 mL of pure methanol was added to 1 mL of plasma. The mixture was centrifuged (18,000g for 20 minutes), and an aliquot of the supernatant was injected into the high-performance liquid chromatography apparatus. The column was an 18 SpheriSorb octadecyl silane 10-μm column (Alltech, Deerfield, IL) and the mobile phase was water/methanol (40%/60%) at a flow rate of 1.2 mL/minute. Detection was obtained at wavelengths of 285-nm excitation and 340-nm emission.
Oxindole plasma levels were evaluated as previously described18: 2 mL HClO4 0.4 N was added to 1 mL plasma to precipitate proteins. After centrifugation (18,000g for 20 minutes), the supernatant was collected. The procedure was repeated to improve plasma extraction. The supernatants were then mixed with 8 mL of chloroform for at least 5 minutes. The organic layer was collected and evaporated under a stream of N2. Residues were dissolved in HClO4 0.4 N (200 μL), and a portion was injected into a high-performance liquid chromatography apparatus equipped with an ultraviolet spectrophotometer detector. A 25-cm reverse-phase 18 SpheriSorb octadecyl silane 10-μm column and a mobile phase of 0.5 M acetic acid/acetonitrile at a ratio of 90%/10 % wt/wt at a flow rate of 1.5 mL/minute were used.
Patients were followed prospectively for a median (interquartile range) of 22 months (range, 11-36 months) from the date of initial assessment. Information was obtained on the occurrence of death/hepatic transplantation and episodes of HE requiring in-hospital admission. Hospital admissions were qualified as HE-related if the reason for hospitalization was HE itself. Thus, inpatient stays during which an episode of HE occurred in an individual who had been admitted for a different reason or a major precipitant (i.e., gastrointestinal bleeding, sepsis) were not included.
Differences between groups were examined using Mann-Whitney U or Kruskal-Wallis tests (post hoc comparisons: Mann-Whitney U test, applying the Bonferroni correction for multiple comparisons). Correlations were tested using the Spearman coefficient. Survival analysis was performed with the Cox proportional hazards model or with the Kaplan-Meier cumulative survival method, as appropriate. Patients who underwent transplantation were qualified as alive and censored on the day of transplantation; the analysis was also conducted excluding transplanted patients. The predictive validity of different variables on the occurrence of HE-related hospitalizations was also assessed using survival analysis methods; patients who were hospitalized because of HE were qualified as complete cases.
The protocol was approved the Hospital of Padua Ethics Committee. All participating subjects provided written, informed consent. The study was conducted according to the Declaration of Helsinki (Hong Kong Amendment) and European Good Clinical Practice guidelines.
Demographics, Neuropsychiatric Performance, and Laboratory Variables.
The etiology of cirrhosis was viral (hepatitis C, B, or B plus D) in 38 (53%) patients, alcohol in 22 (30%) patients, primary biliary cirrhosis in 10 (14%) patient, and cryptogenic in two (3%) patients. Functionally, 14 patients (19%) were classified as Child-Pugh grade A, 38 (53%) as Child-Pugh grade B, and 20 (28%) as Child-Pugh grade C. The average MELD score was 12 ± 7.
On average, patients with cirrhosis had significantly worse neuropsychiatric performance than healthy volunteers (Table 1). Patients with alcohol-related cirrhosis had significantly worse neuropsychiatric performance than their counterparts with non–alcohol-related cirrhosis (Table 2).
Table 1. Summary of Neuropsychiatric and Laboratory Indices in Healthy Volunteers and Patients with Cirrhosis
On the day of study, 38 (53%) patients were classified as neuropsychiatrically unimpaired and 34 (47%) patients were classified as having grade I overt HE according to the West Haven criteria. Thirty-three (46%) patients had normal PHES and EEG performance, six (8%) had abnormal PHES, 18 (25%) had abnormal EEG, and 13 (18%) had both abnormal PHES and EEG. Of the 34 patients who were classified as having grade I overt HE, 11 (32%) had normal PHES and EEG performance, 5 (15%) had abnormal PHES, nine (26%) had abnormal EEG, and nine (26%) had both abnormal PHES and EEG. However, these 34 patients had significantly worse performance than their counterparts classified as clinically normal on most stand-alone psychometric and EEG indices (P < 0.05).
On average, patients with cirrhosis had significantly higher ammonia and tryptophan derivatives concentrations than healthy volunteers, as well as elevated inflammatory markers (Table 1). Patients with alcohol-related cirrhosis had significantly lower sodium and higher CRP and IL-6 concentrations than their counterparts with non–alcohol-related cirrhosis (Table 2).
Thirteen (18%) patients had mild hyponatremia, 47 (65%) had mild-moderate anemia, 37 (54%) had high CRP, 41 (61%) had high IL-6, 48 (72%) had high TNFα, 40 (71%) had hyperammonemia, 58 (86%) had high indole, and 43 (64%) had high oxindole.
Psychometry and Laboratory Variables.
Patients with abnormal PHES had significantly higher CRP (17 ± 22 versus 7 ± 6; P < 0.01), IL-6 (32 ± 54 versus 12 ± 13; P < 0.05), and TNFα (17 ± 8 versus 11 ± 7; P < 0.001) concentrations than their counterparts with normal PHES (Fig. 1, Table 3). Significant, consistent correlations were observed between stand-alone psychometric test results and CRP, IL-6, and TNFα (Table 4).
Table 3. Laboratory Parameters in Patients with Cirrhosis with Normal and Abnormal PHES and EEG
Table 4. Matrix of Correlations Between Laboratory Parameters and Stand-Alone, Crude Psychometric Test Results
Values are correlation coefficients (Spearman's R); bold typeface: P < 0.05.Abbreviations: CRP: C reaction protein; IL6: Interleukin 6; TNFα: tumor necrosis factor α; TMT-A/B: trial making test A/B; DS: digit symbol; SD: serial dotting; LTT/E: line tracing time/error.
CRP and TNFα concentrations were also independent predictors of an abnormal PHES performance (overall model, χ2 = 16; CRP, β [± SE] = 0.10 ± 0.04, P = 0.02; TNFα, β = 0.09 ± 0.04, P = 0.03); a trend (0.05 < P < 0.1) was maintained also when the degree of hepatic failure, either in the form of the Child-Pugh or MELD score, was taken into account.
EEG and Laboratory Variables.
Patients with abnormal EEG had significantly higher indole (430 ± 270 versus 258 ± 255; P < 0.01) and ammonia (66 ± 35 versus 45 ± 27; P < 0.05) concentrations than their counterparts with normal EEG (Fig. 2, Table 3). Significant correlations were observed between spectral EEG indices and a number of laboratory variables; these correlations were more consistent for ammonia and IL-6 (Table 5).
Table 5. Matrix of Correlations Between Laboratory and EEG Spectral Parameters
Values are correlation coefficients (Spearman's R); bold typeface: P < 0.05.
Indole and ammonia concentrations were independent predictors of an abnormal EEG (overall model, χ2 = 15; indole, β = 0.003 ± 0.001, P = 0.008; ammonia, β = 0.02 ± = 0.01, P = 0.03); this also held true for indole (overall model, χ2 = 20; β = 0.004 ± 0.001, P = 0.005) when the degree of hepatic failure, either in the form of the Child-Pugh or MELD score, was taken into account.
Seven patients were lost to follow-up. Of the remaining 65 patients, 20 died (median, 11 months [interquartile range, 6-23 months]) and 14 underwent transplantation (median, 10 months [interquartile range, 3-16 months]). During the follow-up period, 15 (23%) patients developed an episode of HE requiring in-hospital admission (median, 7 months [interquartile range, 3-14 months]).
No differences in the length of survival or the risk of developing HE over the follow-up period were observed in relation to the etiology of cirrhosis (alcohol-related versus non–alcohol-related).
Both the PHES and EEG analysis (categorical [PHES/EEG normal or abnormal] or continuous/semicontinuous variables [total PHES score, EEG mean dominant frequency]) were independent predictors of death (Table 6) and occurrence of HE-related hospitalization (Fig. 3, Table 7). This held true when the degree of hepatic failure (either MELD or Child-Pugh score) was included in the model and when patients who underwent transplantation were excluded from the analysis (data not shown).
Table 6. Relation Between PHES/EEG Performance and Survival
Overall Model Fit (χ2)
Transplanted patients were included and were classified as alive and censored on the day of transplantation. The degree of hepatic failure was not included in the model.
Table 7. Relation Between PHES/EEG Performance and the Subsequent Occurrence of HE Requiring Hospitalization
Overall Model Fit (χ2)
Transplanted patients were included and were classified as alive and censored on the day of transplantation. The degree of hepatic failure was not included in the model.
Laboratory markers in the form of continuous variables or when qualified as normal/abnormal did not predict death. However, hemoglobin, ammonia, and IL-6 levels independently predicted the subsequent occurrence of HE-related hospitalizations (overall model, χ2 = 24; hemoglobin, β = −0.51 ± 0.176, P = 0.003; ammonia, β = 0.04 ± 0.01, P = 0.001; IL-6, β = 0.02 ± 0.01, P = 0.04).
Finally, in a model including PHES, EEG, hemoglobin, ammonia, and IL-6 levels, all variables except for IL-6 maintained independent predictive value.
In this study, PHES and EEG abnormalities in patients with cirrhosis were found to have partially different biochemical correlates, the former being mostly associated with elevated inflammatory markers, the latter with high concentrations of ammonia and indole. In addition, both PHES and EEG performance independently predicted the occurrence of HE-related hospitalization and death.
Both the PHES battery and the EEG have been recommended for the diagnosis of minimal HE and the quantification of overt HE.19-21 However, their degree of agreement and their relationship with laboratory markers of HE, particularly venous ammonia, has generally been deemed poor.22 In addition, both PHES and EEG analysis have limitations in relation to the diagnosis of HE,19-21 and some degree of experience is required for their interpretation. In the present study, strong associations were observed between overall/stand-alone PHES performances and elevated inflammatory markers, whereas EEG abnormalities were associated with high levels of ammonia and the tryptophan metabolite indole. These data point to partially different mechanisms for different features of the HE phenotype: whereas PHES seems to be particularly susceptible to the neurotoxic effect of an activated inflammatory cascade, the EEG seems to better reflect the cerebral metabolism of gut-derived toxins, such as ammonia or tryptophan metabolites that the liver fails to dispose of. No significant differences in sodium, hemoglobin, or glucose concentrations were observed between patients with normal/abnormal psychometric/EEG performance, although a number of significant correlations were observed between sodium/hemoglobin levels and psychometric/EEG stand-alone indices. These data suggest that in a population of relatively well-compensated outpatients with cirrhosis, sodium, hemoglobin or glucose levels have no major confounding effects on standard measures of mental status used for HE assessment. However, screening for all possible metabolic causes of mental derangement is probably appropriate on a single-patient level.
A relationship between an activated inflammatory cascade and impaired cognitive performance has been observed in a number of clinical settings. Very low-dose endotoxemia causes an increase in circulating IL-6 and a parallel worsening in memory function, which are independent of physical stress symptoms or the activation of the hypothalamo-pituitary-adrenal axis.23 Decreased ability to concentrate and sickness behavior have been related to the release of proinflammatory cytokines.24 Recent studies suggest that inflammation, particularly the acute phase response, might also play a role in the pathogenesis of mild cognitive impairment, a transitional stage between normal cognitive aging and Alzheimer dementia.25 The fact that in these conditions limited or no changes are observed in the EEG analysis might suggest that cytokines are neurotoxic at a subcortical rather than cortical level.
The EEG analysis is sensitive to the influence of nutrition and energy-providing metabolic pathways, to electrolyte balance and to the clearance of toxic substances. Therefore, it is not surprising that high blood/cerebral levels of ammonia/indole or their metabolites would affect electrogenesis. Increased cerebral levels of ammonia determine an increase in the conversion of glutamate to glutamine and, in turn, an alteration of the inhibition/excitation neurotransmitter balance toward inhibition.26 Interestingly, some of the EEG changes observed in patients with HE are reminiscent of those observed in the physiological transition between wakefulness and the early stages of sleep,27 supporting the hypothesis of neural inhibition and vigilance reduction. Oxindole is believed to reduce neuronal excitability by modifying the function of voltage-operated sodium channels, and therefore to have direct sedative effects.4 In our study, only indole and not oxindole serum concentrations were related to EEG slowing and to the occurrence of severe HE/death. Although this might seem surprising, it is important to remember that indole is metabolized to oxindole in almost every organ and tissue of the body, including the brain. It is therefore possible that serum indole concentrations might reflect cerebral oxindole concentrations better than serum oxindole itself. These results are in line with the recent observations by Riggio et al.17 Finally, it is also possible that ammonia and tryptophan derivatives might potentiate their respective effects on cerebral electrogenesis. Indeed, the cerebral pathway of the kynurenic metabolites of tryptophan has been shown to be affected by hyperammonemia in an experimental model of hepatic failure.28
Although psychometric and EEG abnormalities were found to have at least partially different biochemical correlates, they both predicted the subsequent onset of severe, overt HE, indicating that both the neuropsychiatric effects of inflammation and those of hyperammonemia/elevated tryptophan metabolites contribute to the medium/long-term clinical outcome. This seems an important finding and might represent a relatively direct clinical correlate of the interaction between ammonia- and inflammation-related neurotoxicity that has been observed both in experimental cirrhosis and in experimental sepsis.7, 29, 30 It also lends support to the hypotheses that ammonia and inflammatory cytokines may act synergistically31 and that they might induce astrocyte swelling/dysfunction as a common pathogenic endpoint.32
The observation that patients with alcohol-related cirrhosis are more likely to exhibit neuropsychiatric abnormalities than their counterparts with non–alcohol-related cirrhosis is not entirely novel and most likely due to the direct damage that alcohol misuse causes to the brain, regardless of the degree of hepatic involvement.1, 33, 34 In addition, the enhanced activation of the inflammatory cascade observed in patients with alcohol-related cirrhosis may also play a role.
The findings of this study have several direct and indirect implications: First, the discrepancies between EEG and psychometric abnormalities often observed in patients with cirrhosis depend, at least to some extent, on the different pathways leading to such abnormalities. Second, PHES and EEG analysis are both useful for an optimal HE evaluation in that they reflect different aspects of the pathogenesis of HE and they independently predict the subsequent occurrence of severe overt HE and death. Third, for the same reasons, and for purposes of differential diagnosis, the results of a comprehensive neuropsychiatric examination should probably be included in the decision process leading to selection for hepatic transplantation. Finally, it is possible to hypothesize that the effects of ammonia/indole-lowering therapeutic strategies are more likely to be measured by neurophysiological rather than psychometric tools. In contrast, the effects of new drugs aimed at modulating the inflammatory cascade are probably best assessed by psychometry.
In conclusion, PHES and EEG abnormalities in patients with cirrhosis have partially different biochemical correlates and independently predict outcome. If confirmed, these results suggest that, despite the demands of routine hepatology practice for simple tools for the evaluation of neuropsychiatric status, meaningful and prognostically useful results can be obtained only with protocols including both psychometry and neurophysiology and, where possible, measurement of venous ammonia/indole and an inflammatory marker. This will be even more important within research and clinical trial settings.
The authors are grateful to Dr. Antonietta Sticca for technical assistance.