Summary of findings
Description of the condition
Hepatic encephalopathy is a complex neuropsychiatric syndrome seen in severe liver failure (Gitlin 1996; Ferenci 2002). Symptoms range from minor neuropsychiatric changes to deep coma (Conn 1979). Hepatic encephalopathy may be clinically overt or may consist of mild neurocognitive impairments, which have been identified in a substantial percentage of patients with liver disease (Randolph 2009). The course of the disease may be episodic, with recurrent symptoms, or chronic, with more stable symptoms (Bajaj 2011). The exact underlying pathophysiology is not known. Experimental studies suggest that symptoms develop as the result of accumulation of toxic agents that have not been metabolised by the liver (Gitlin 1996). Other potential mechanisms include the generation of false neurotransmitters and an abnormal interaction between astrocytes and other cellular elements with cerebral oedema and alterations in glioneural communication (Haussinger 2000; Cordoba 2001).
Description of the intervention
Many patients with hepatic encephalopathy present with extrapyramidal symptoms and have changes in the basal ganglia, as detected by magnetic resonance imaging and proton spectroscopy (Spahr 2000). These symptoms are comparable with those seen in Parkinson's disease and suggest an impairment of dopamine neurotransmission (Blei 1999; Jover 2003). Patients with Parkinson's disease are less likely to experience dyskinesia and dystonia when treated with levodopa (Stowe 2008). Uncontrolled trials suggest that levodopa or bromocriptine could be beneficial in the treatment of patients with hepatic encephalopathy (Parkes 1970; Jorge 1973). The effects of dopamine agents have also been assessed in randomised clinical trials (Uribe 1979; Michel 1980; Morgan 1980), and previous guidelines suggested that the intervention may be considered in patients with chronic hepatic encephalopathy (Blei 1999; Lizardi-Cervera 2003).
Why it is important to do this review
We have previously published a systematic review on dopamine agents for hepatic encephalopathy (Als-Nielsen 2004a). The results of this review were inconclusive. We have been unable to identify any further meta-analyses or systematic reviews on the topic. To determine the strengths and weaknesses of the current evidence, we have updated our previous review (Als-Nielsen 2004a).
To evaluate the beneficial and harmful effects of dopamine agents versus placebo or no intervention for patients with hepatic encephalopathy.
Criteria for considering studies for this review
Types of studies
This review included all randomised trials, regardless of publication status, language, or blinding. Unpublished trials were included if the methodology and the data were available in written form. We planned to include observational studies reporting harms, but we identified no observational studies reporting relevant data.
Types of participants
Patients with hepatic encephalopathy were included, irrespective of the aetiology of the underlying liver disease. The diagnostic criteria could include psychometric tests, clinical scoring systems (such as the West-Haven criteria), electroencephalography (Guerit 2009), or biochemical findings (including ammonia levels). Based on the diagnostic criteria used in the included trials, participants were classified as having overt or minimal hepatic encephalopathy, and the latter was classified further as recurrent or chronic.
Types of interventions
The intervention comparisons assessed were dopamine agents (e.g., levodopa, bromocriptine) versus placebo or no intervention. Studies were included irrespective of the dose or duration of therapy.
Types of outcome measures
- Mortality (all-cause).
- All cause non-fatal serious adverse events.
- Morbidity. This outcome measure was assessed on the basis of the number of participants who showed no improvement in manifestations of hepatic encephalopathy as defined by the authors of included trials.
- All-cause non-serious adverse events (number and type) (ICH-GCP 1997).
- Qualitiy of life.
Search methods for identification of studies
We searched the Cochrane Hepato-Biliary Group Controlled Trials Register, the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, and Science Citation Index-Expanded (Royle 2003). Search strategies with time spans of the searches are given in Appendix 1.
Searching other resources
Reference lists in relevant articles and conference proceedings were scanned for additional trials not identified in the electronic searches. We wrote to authors of identified trials and pharmaceutical companies to enquire about additional trials. Ongoing and completed trials were also identified through searches in the World Health Organization Trial Search Portal (www.who.int/trialsearch/).
Data collection and analysis
Selection of studies
All review authors participated in the selection of trials. AEJ listed the potentially eligible trials. Subsequently, trials that fulfilled all inclusion criteria were identified. Excluded trials were listed along with the reasons for exclusion.
Data extraction and management
Three review authors (AEJ, BA-N, and LLG) extracted data independently. All disagreements were resolved through discussion before analyses.
We extracted data on the design of the trial (country of origin, parallel or cross-over design, and bias control), participant characteristics (aetiology of underlying liver diseases and type of hepatic encephalopathy, mean age, proportion of men), and the intervention regimen assessed (type, dose, and duration of therapy).
Assessment of risk of bias in included studies
We assessed the risk of bias in the trials independently in accordance with the instructions provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) and the Cochrane Hepato-Biliary Group Module (Gluud 2013). Because of the risk of overestimation of intervention effects in randomised trials with high risk of bias (Schulz 1995; Moher 1998; Kjaergard 2001; Wood 2008; Lundh 2012; Savovic 2012, Savovic 2012a), we assessed the influence of risk of bias on trial results using the following domains.
Allocation sequence generation
- Low risk of bias: Sequence generation was achieved by using computer random number generation or a random number table. Drawing lots, tossing a coin, shuffling cards, and throwing dice were adequate if performed by an independent person not otherwise involved in the trial.
- Uncertain risk of bias: The method of sequence generation was not specified.
- High risk of bias: The sequence generation method was not random.
- Low risk of bias: The participant allocations could not have been foreseen in advance of, or during, enrolment. Allocation was controlled by a central and independent randomisation unit. The allocation sequence was unknown to the investigators (e.g., if the allocation sequence was hidden in sequentially numbered, opaque, and sealed envelopes).
- Uncertain risk of bias: The method used to conceal the allocation was not described so that intervention allocations may have been foreseen in advance of, or during, enrolment.
- High risk of bias: The allocation sequence was likely to be known to the investigators who assigned the participants.
Blinding of participants, personnel, and outcome assessors
- Low risk of bias: Blinding was performed adequately, or the assessment of outcomes was not likely to be influenced by lack of blinding.
- Uncertain risk of bias: Information was insufficient to permit assessment of whether blinding was likely to induce bias on the results.
- High risk of bias: No blinding or incomplete blinding was performed, and assessment of outcomes was likely to be influenced by lack of blinding.
Incomplete outcome data
- Low risk of bias: Missing data were unlikely to make treatment effects depart from plausible values. Sufficient methods, such as multiple imputation, had been employed to handle missing data.
- Uncertain risk of bias: Information was insufficient to permit assessment of whether missing data in combination with the method used to handle missing data were likely to induce bias on the results.
- High risk of bias: The results were likely to be biased as the result of missing data.
Selective outcome reporting
- Low risk of bias: All outcomes were predefined and reported, or all clinically relevant and reasonably expected outcomes were reported. The trial was registered on the www.clinicaltrials.gov web site or on a similar register, or the protocol was published.
- Uncertain risk of bias: It was unclear whether all predefined and clinically relevant and reasonably expected outcomes were reported.
- High risk of bias: One or more clinically relevant and reasonably expected outcomes were not reported, and data on these outcomes were likely to have been recorded.
- Low risk of bias: The trial appears to be free of other components that could put it at risk of bias.
- Uncertain risk of bias: The trial may or may not be free of other components that could put it at risk of bias.
- High risk of bias: Other factors in the trial could put it at risk of bias (e.g., for-profit involvement, authors conducting trials on the same topic).
Trials with unclear or high risk of bias methodology in one or more of the above domains were considered trials with high risk of bias. The remaining were considered trials with low risk of bias.
Measures of treatment effect
All outcome measures were dichotomised and were expressed using odds ratios (ORs) with 95% confidence intervals (CIs).
Unit of analysis issues
The primary analyses included data from trials using a parallel-group design and from the first treatment period of cross-over trials. Additional analyses were performed that included paired data from the cross-over trials (Becker 1993; Elbourne 2002).
Dealing with missing data
Data on all participants randomly assigned were sought to allow intention-to-treat analyses that included participants irrespective of compliance or follow-up. For participants with missing data, carry-forward of the last observed response was used. We originally planned to analyse the influence of missing data using imputation (Higgins 2008). We planned to impute missing values as failures, successes, same as control group, same as experimental group, and same as own group (Higgins 2008). We did not perform these analyses because no losses to follow-up were described.
Assessment of heterogeneity
Intertrial heterogeneity was assessed on the basis of I
Assessment of reporting biases
We planned to evaluate the risk of reporting bias by comparing trial protocols and published reports. Furthermore, reporting biases were assessed on the basis of the extent to which clinically relevant outcome measures (hepatic encephalopathy, mortality, and adverse events) were reported.
Analyses were performed in Review Manager 5 (RevMan 2012) and in STATA 12 (STATA 12). Primary meta-analyses were performed by using random-effects models because of anticipated variability between trials regarding participants and interventions.
Subgroup analysis and investigation of heterogeneity
Originally, we planned to perform several subgroup analyses to assess sources of intertrial heterogeneity (bias control, participant characteristics, and intervention regimens). However, because of the limited number of trials in the meta-analyses of the primary outcomes, we were able to perform these subgroup analyses only for the outcome measure of mortality. Likewise, regression analyses (Egger's test) that were planned to estimate the risk of publication bias and other biases (small-study effects) were performed only for the outcome measure of mortality.
Trial sequential analysis
We performed trial sequential analysis (CTU 2011; Thorlund 2011) to control risks of random errors due to sparse data and repetitive testing of cumulative data (Brok 2008; Wetterslev 2008; Brok 2009; Thorlund 2009; Wetterslev 2009; Thorlund 2010). To minimise the risk of random error, we calculated the required information size, defined as the required sample size necessary to detect or reject intervention effects after adjusting for diversity (Brok 2008; Wetterslev 2008; Brok 2009; Thorlund 2009; Wetterslev 2009; Thorlund 2010). The information size was calculated on the basis of a risk ratio (RR) reduction of 20% or the results of included trials with a low risk of bias (Brok 2008; Wetterslev 2008; Brok 2009; Thorlund 2009; Wetterslev 2009; Thorlund 2010). We presented the results of the analysis in a graph. with individual trials added on the basis of their year of publication. If more than one trial was published in a year, trials were added alphabetically according to the first author's family name. The results of the trials were presented as a cumulative Z-curve. The trial sequential monitoring boundaries were constructed and the diversity-adjusted required information size calculated with a type 1 error of 5% and a type 2 error of 20%. The results were displayed as a graph with the cumulative meta-analysis results entered. The trial sequential analysis shows firm evidence of intervention effects (or no intervention effects) if the cumulative Z-curve crosses the monitoring boundaries; it also shows that additional trials may be needed if the boundaries are not crossed.
The robustness of the results was assessed by repeating the meta-analyses using a fixed-effect model. No additional sensitivity analyses were performed because of the limited number of trials identified.
Description of studies
Results of the search
In total, 294 references were identified through the literature searches (Appendix 1). After duplicates and clearly irrelevant references (references to papers that did not describe trials of dopaminergic agents for participants with hepatic encephalopathy) were excluded, 17 references were retrieved for further assessment (Figure 1). Of these, eight references referred to five randomised trials that were eligible for inclusion (Uribe 1979; Vij 1979; Michel 1980; Morgan 1980; Koshy 1982). Through correspondence with the authors of two trials (Uribe 1979; Morgan 1980), additional information was obtained on trial results and methods. For the remaining trials, data were gathered from published reports.
|Figure 1. Figure 1. Study flow diagram.|
All of the included trials were described in at least one full-paper article published from 1979 to 1982. Three trials used a parallel-group design (Vij 1979; Michel 1980; Koshy 1982), and two trials used a cross-over design (Uribe 1979; Morgan 1980).
In total, 144 participants with overt hepatic encephalopathy were included. Three trials (66 participants in the treatment group versus 65 participants in the control group) assessed acute episodes of hepatic encephalopathy (Vij 1979; Michel 1980; Koshy 1982). Two trials (seven participants in the treatment group versus six participants in the control group) assessed chronic hepatic encephalopathy (Uribe 1979; Morgan 1980). Two trials included participants with acute fulminant liver failure due to viral hepatitis (Vij 1979; Koshy 1982). Three trials included participants with cirrhosis (Uribe 1979; Michel 1980; Morgan 1980). The proportion of participants with alcoholic liver disease ranged from 0 to 80%. The proportion of participants with viral hepatitis ranged from 0 to 100%, and mean age ranged from 32 years to 57 years.
Three trials assessed levodopa (Vij 1979; Michel 1980; Koshy 1982), and two trials assessed bromocriptine (Uribe 1979; Morgan 1980). The mean daily dose was 4 grams for levodopa and 15 grams for bromocriptine. The median duration of treatment was 14 days (range seven to 56 days). None of the trials followed participants after the end of treatment. None of the included trials assessed health economics.
Nine references to eight trials were excluded because they turned out not to be randomised or referred to cross-over trials that compared dopamine agents versus interventions for hepatic encephalopathy considered potentially active (Characteristics of included studies).
Risk of bias in included studies
All trials had a high risk of bias in the assessment of one or more than one of the bias risk domains.
|Figure 2. Figure 2. Risk of bias graph: review authors' judgements about all risk of bias items presented as percentages across all included studies.|
Incomplete outcome data
Two trials accounted for all participants with missing outcome data (Uribe 1979; Morgan 1980). In the remaining three trials, no dropouts or withdrawals were described, giving the impression that no losses to follow-up occurred, although this was not specifically stated.
Other potential sources of bias
No sample size calculations were reported. None of the included trials received industry funding.
Effects of interventions
Random-effects meta-analyses found no difference in mortality between participants randomly assigned to dopamine agents versus controls (OR 1.11, 95% CI 0.34 to 3.54; Analysis 1.1). Little intertrial heterogeneity was noted (I
The trial sequential analysis graph showed that the cumulative Z-curve does not cross the monitoring boundary (Figure 3). The analysis showed a diversity-adjusted required information size of 673 participants (the number of participants needed to reach firm evidence of an intervention effect of 20% risk ratio reduction). The number of participants included corresponds to only 21% of the diversity-adjusted required information size. Accordingly, we lack evidence to recommend or refute dopamine agents for hepatic encephalopathy.
The primary random-effects meta-analyses showed no significant effects of dopamine agents on hepatic encephalopathy compared with placebo or no intervention when data from parallel-group trials were analysed (OR 0.33, 95% CI 0.01 to 11.25; Analysis 1.7) or when paired data from the cross-over trial reporting this outcome measure were included (OR 0.68, 95% CI 0.17 to 2.67; Analysis 1.8). The results were confirmed by fixed-effect meta-analyses including data from parallel-group trials (OR 1.08, 95% CI 0.45 to 2.62), but also when paired data from the two cross-over trials reporting this outcome measure were included (OR 1.04, 95% CI 0.75 to 1.43).
We were able to retrieve data on adverse events only from the two cross-over trials (Uribe 1979; Morgan 1980). In total, seven of 13 participants experienced non-serious adverse events during treatment with dopamine agents. No adverse events were reported during control periods. No clear difference was observed between intervention and control groups ( Analysis 1.9). No serious adverse events were registered. Adverse events included hypomania (n = 1), hallucinations and headache (n = 1), constipation (n = 3), and nausea and vomiting (n = 2).
Quality of life
None of the included trials reported data on quality of life.
Summary of main results
Patients with cirrhosis may present with extrapyramidal symptoms similar to those seen in Parkinson's disease (Jover 2003). Further similarities between participants with hepatic encephalopathy and participants with Parkinson's disease include alterations in the basal ganglia (Spahr 2000). In theory, dopamine agents that are effective in Parkinson's disease could alleviate manifestations of hepatic encephalopathy. However, the present systematic review found no evidence to recommend or refute the use of dopamine agents for patients with hepatic encephalopathy. The available evidence includes only a limited number of small trials published before 1983. No clear effects were identified for any of the outcome measures assessed. Additional analyses found no specific subgroups that indicated potential effects when the results of included trials were separated on the basis of the type of hepatic encephalopathy at inclusion, the type of underlying liver disease, or the intervention assessed. The dose and duration of the interventions assessed were similar across trials. Data from participants with Parkinson's disease (Miyasaki 2002) show that the dose of both levodopa and bromocriptine and the duration of the intervention regimens assessed in included trials should be sufficiently high to detect a clinical response. The combined evidence is not promising. However, the statistical power is low, and evidence is insufficient to support or refute beneficial or harmful effects of the interventions assessed.
Overall completeness and applicability of evidence
To ensure completeness of the evidence, we performed extensive literature searches. Our regression analyses showed no clear evidence of publication bias or other small-study effects. Still, the regression analysis was not sensitive because of the limited number of trials.
The main problem with the included trials is the fact that a number of potentially effective interventions for patients with decompensated liver disease have been identified after the trials were completed. These interventions include treatments for hepatic encephalopathy (Bass 2010), bleeding oesophageal varices (Abraldes 2007), and spontaneous bacterial peritonitis (Wiest 2012). Likewise, the diagnostic assessment and nomenclature for hepatic encephalopathy have been updated (Bajaj 2011). Accordingly, extrapolation of results from the present review to current clinical practice is of limited value.
Quality of the evidence
Adequate internal validity depends on the control of bias and random errors. Because three trials had unclear randomisation (Michel 1980; Morgan 1980; Koshy 1982) and consequently an unclear control of selection bias, the internal validity of their results and of the results of our meta-analyses can be questioned. The use of a cross-over design as applied in two of the included trials (Uribe 1979; Morgan 1980) is also debatable. Even chronic hepatic encephalopathy may have a fluctuating course (Basile 1991); therefore, manifestations of hepatic encephalopathy may change during the course of the trial, irrespective of the interventions assessed. The underlying condition and the ability to respond to treatment may not remain stable from the first to the second treatment period. We therefore used only data from the first study period of the cross-over trials in our primary analyses. Unfortunately, these data were available for only one trial (Morgan 1980). The sensitivity analysis on paired data did not change our overall result.
Potential biases in the review process
Identification and selection of trials are essential to the assessment of bias in the review process. To limit bias in the selection process, we included trials irrespective of language or publication status. We also chose to include trials regardless of the dose or duration of the interventions assessed. This led to a relatively heterogeneous group of trials. We did, however, choose to exclude trials with an active comparison group. This choice was made on the basis of lack of evidence supporting several of the interventions assessed for patients with hepatic encephalopathy. The strategy resulted in the exclusion of two small, low-quality, cross-over trials on chronic hepatic encephalopathy (Messner 1982; Uribe 1983). The control groups in these trials received lactulose or neomycin, which could affect the course of hepatic encephalopathy. The total number of participants randomly assigned in these two trials was only 15, and this limits the value of these results.
Agreements and disagreements with other studies or reviews
At present, dopamine agents are not recommended for patients with hepatic encephalopathy. Previous guidelines state that bromocriptine may be considered for patients with chronic hepatic encephalopathy that is unresponsive to other interventions (Blei 1999). In agreement with more recent recommendations (Phongsamran 2010), the present review contradicts these recommendations, suggesting that no evidence is available to support the use of dopamine agents for chronic hepatic encephalopathy.
Implications for practice
This review does not provide evidence to recommend or refute the use of dopamine agents for patients with hepatic encephalopathy.
Implications for research
However, we cannot exclude the possibility that dopamine agents may have beneficial effects that were overlooked because of the limited statistical power of the included trials. On the other hand, other interventions for hepatic encephalopathy (such as non-absorbable disaccharides, branched chain amino acids, and antibiotics) appear potentially more promising than dopamine agents (Als-Nielsen 2004a; Bass 2010; Les 2011). The value of additional trials on dopamine agents is questionable. Should anyone wish to conduct further trials, we recommend that the dopamine agent used should be tested against placebo in parallel-group superiority trials conducted according to the SPIRIT guidelines (SPIRIT 2013; SPIRIT 2013a) and reported according to the CONSORT guidelines (www.consort-statement.org).
We thank Marsha Morgan and Misael Uribe for providing additional information on their trials, and Sarah Klingenberg for performing the electronic literature searches.
Peer reviewers: FG Romeiro, Brazil; Diego Sánchez-Munoz, Spain.
Contact editor: Goran Bjelakovic, Serbia.
Data and analyses
- Top of page
- Summary of findings [Explanations]
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. Search strategies
Last assessed as up-to-date: 13 January 2014.
Contributions of authors
Anders Ellekær Junker (AEJ) and Lise Lotte Gluud (LLG) drafted the revised version of this updated review with methodology updates based on the most recent recommendations in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). AEJ, Bodil Als-Nielsen (BA-N), and LLG participated in the literature searches, identified trials eligible for inclusion, extracted data, and performed the statistical analyses. All authors revised the review and have approved the final version.
Declarations of interest
Sources of support
- Copenhagen Trial Unit, Denmark.
- The 1991 Pharmacy Foundation, Denmark.
- Danish Center for Evaluation and Health Technology Assessment (DACEHTA), Denmark.
Differences between protocol and review
- We have changed the term 'dopaminergic agents' to the MeSH term 'dopamine agents' throughout the review.
- Based on reviewer comments, we have omitted the outcome 'Number of participants with hepatic encephalopathy recovery' because the definition of this outcome is highly variable. The outcome of (lack of) improvement in hepatic encephalopathy includes participants with complete as well as partial recovery from hepatic encephalopathy.
- In our original protocol, we planned to include health economics as an outcome. This outcome was omitted from our previous and present review on the basis of reviewer comments and evidence concerning the best methods for assessing this outcome. We have gathered data on whether health economics were assessed and have included these data in our table of included trials.
- Based on the most recent recommendations regarding the assessment of bias control, we have included bias tables and have assessed the bias control components of allocation (selection bias), blinding (performance bias and detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other potential sources of bias (sample size assessments).
- We have included additional analyses on small-study effects (Egger's test).
Medical Subject Headings (MeSH)
MeSH check words
* Indicates the major publication for the study