Non-absorbable disaccharides for hepatic encephalopathy: systematic review of randomised trialsBMJ 2004; 328 doi: https://doi.org/10.1136/bmj.38048.506134.EE (Published 29 April 2004) Cite this as: BMJ 2004;328:1046
- Bodil Als-Nielsen, research fellow ()1,
- Lise L Gluud, research fellow1,
- Christian Gluud, chief physician1
- 1Cochrane Hepato-Biliary Group, Copenhagen Trial Unit, Centre for Clinical Intervention Research, Copenhagen University Hospital, Department 7102, H:S Rigshospitalet, DK-2100 Copenhagen, Denmark
- Correspondence to: B Als-Nielsen
- Accepted 20 February 2004
Objective To assess the effects of non-absorbable disaccharides (lactulose and lactitol) in patients with hepatic encephalopathy.
Data sources Cochrane Hepato-Biliary Group controlled trials register, Cochrane Library, Medline, and Embase until March 2003; reference lists of relevant articles; authors and pharmaceutical companies.
Review methods Randomised trials that compared non-absorbable disaccharides with placebo, no intervention, or antibiotics for hepatic encephalopathy were included. The primary outcome measures were no improvement of hepatic encephalopathy and all cause mortality.
Results 22 trials were included. Compared with placebo or no intervention, non-absorbable disaccharides seemed to reduce the risk of no improvement in patients with hepatic encephalopathy (relative risk 0.62, 95% confidence interval 0.46 to 0.84, six trials). However, high quality trials found no significant effect (0.92, 0.42 to 2.04, two trials). Compared with placebo or no intervention, non-absorbable disaccharides had no significant effect on mortality (0.41, 0.02 to 8.68, four trials). Non-absorbable disaccharides were inferior to antibiotics in reducing the risk of no improvement (1.24, 1.02 to 1.50, 10 trials) and lowering blood ammonia concentration (weighted mean difference 2.35 μmol/l, 0.06 μmol/l to 13.45 μmol/l, 10 trials). There was no significant difference in mortality (0.90, 0.48 to 1.67, five trials).
Conclusions There is insufficient evidence to support or refute the use of non-absorbable disaccharides for hepatic encephalopathy. Antibiotics were superior to non-absorbable disaccharides in improving hepatic encephalopathy, but it is unclear whether this difference is clinically important. Non-absorbable disaccharides should not serve as comparator in randomised trials on hepatic encephalopathy.
Hepatic encephalopathy is a complex neuropsychiatric syndrome, which may complicate acute or chronic liver failure.1 It is characterised by changes in mental state including a wide range of neuropsychiatric symptoms ranging from minor signs of altered brain function to deep coma.2
Treatment of hepatic encephalopathy aims at reducing the production and absorption of ammonia,which is involved in the pathogenesis.3 4 As colonic bacteria are the primary source of ammonia, treatment initially consisted of poorly absorbed antibiotics, especially neomycin.5 6 This treatment was implemented without appropriate scientific documentation. Lactulose was introduced as a safer alternative.3 On the basis of two small trials,5 6 lactulose was considered to be as effective as neomycin. Subsequent trials and meta-analyses concluded that lactitol and lactulose were equally effective.7–10 Since the 1980s, nonabsorbable disaccharides (lactulose and lactitol) have been considered as the standard treatment for hepatic encephalopathy.11 12 Recent guidelines state that lactulose is the first line pharmacological treatment for hepatic encephalopathy.12 Antibiotics can be considered a therapeutic alternative to non-absorbable disaccharides in acute hepatic encephalopathy but in chronic encephalopathy should be reserved for patients who respond poorly to non-absorbable disaccharides.12
We performed a systematic review to assess the beneficial and harmful effects of non-absorbable disaccharides for hepatic encephalopathy and to compare them with antibiotics.
We searched the Cochrane Hepato-Biliary Group controlled trials register, the Cochrane Library, Medline, and Embase up to March 2003. Included terms were “hepatic encephalopathy or cirrhosis”, and “lactulose, lactitol, or disaccharide”, and “random* or clinical”.13 We screened bibliographies of relevant articles and conference proceedings and wrote to experts and pharmaceutical companies.
Selection—We included all randomised trials that compared non-absorbable disaccharides (lactulose and lactitol) with placebo, no treatment, or antibiotics for hepatic encephalopathy. Inclusion was regardless of publication status, language, or blinding. Included patients had acute, chronic, or minimal hepatic encephalopathy.
Validity assessment—Two reviewers independently assessed trial quality15 16 by examining three components: generation of allocation sequence (classified as adequate if based on computer generated random numbers, tables of random numbers, or similar), concealment of allocation (classified as adequate if based on central randomisation, sealed envelopes, or similar), and blinding (classified as adequate if the trial was described as double blind or had blinded outcome assessment).13 We classified trials with adequate concealment of allocation and adequate blinding as high quality.
Data abstraction—Two reviewers (BA-N and LLG) independently extracted data from each trial. Our primary outcome measures were the numbers of patients without improvement of hepatic encephalopathy and all cause mortality. Improvement was defined as partial or complete resolution of clinical or subclinical symptoms of hepatic encephalopathy. Secondary outcome measures were adverse events, number connection test result, and blood ammonia concentration. In the number connection test, participants are instructed to connect numbers printed on a page consecutively from 1 to 25 as quickly as possible. The test score is the time the patient needs to perform the test, including the time needed to correct any errors. A low score represents a good performance. All outcomes were assessed at the end of treatment and maximum follow up.
Trial characteristics—We extracted the type and cause of the underlying liver disease, type of hepatic encephalopathy (acute, chronic, or minimal); mean age; proportion of men; number of patients randomised to each intervention arm; type, dose, and duration of treatment; mode of administration; trial quality15 16; trial design (parallel or crossover); duration of follow up; and number of dropouts. We sought data on all patients, irrespective of compliance or follow up. Primary investigators were contacted if data were missing.
Quantitative data synthesis—All data were analysed on the basis of intention to treat, including all randomised patients irrespective of compliance or follow up. If patients had missing outcome data, we carried forward the last reported observed response.17 Data from the first period of crossover trials were included. Binary outcomes were expressed as relative risks with 95% confidence intervals. Continuous outcomes wereexpressed as weighted mean difference with 95% confidence intervals. We used a random effects model18 because we anticipated clinical variability between trials. Statistical heterogeneity was explored by the χ2 test with significance setat P < 0.1. Potential sources of heterogeneity were explored through subgroup analyses with regard to the quality of methods and type of hepatic encephalopathy. We used the test of interaction19 to compare the difference between the estimates of subgroup analyses. Analyses were performed in Review Manager version 4.2.2. for Windows and SPSS version 11.0 for Windows.
Figure 1 summarises the literature search. We included 22 trials that assessed lactulose or lactitol versus placebo, no treatment, or antibiotics.5 6 20–39 Two trials were published as abstracts.32 37 The remaining were published as full articles. Eighteen trials used a parallel group design and four a crossover design. All trials were described as randomised, but adequate generation of the allocation sequence was described in only four.22 30 31 39 Treatment allocation was adequately concealed in 10 trials,5 6 20–23 26 30 36 39 double blinding was reported in 15 trials,5 6 20–25 27 32–34 36 38 39 and one trial had blinded outcome assessment.30 We classified nine trials as high quality.5 6 20–23 30 36 39
Lactulose or lactitol v placebo or no intervention
Ten trials with 280 patients (75% men) assessed lactulose or lactitol versus placebo or no intervention (table 1).20–29 All patients had cirrhosis and acute,25 chronic,20 22–24 acute or chronic,21 or minimal hepatic encephalopathy.26–29 Eight trials assessed oral lactulose,20–24 26 28 29 one assessed oral lactitol,27 and one assessed lactitol enemas.25 The daily mean doses of lactulose ranged from 30 g to 84 g (median 50 g). In six trials the dose was adjusted to obtain two to three semisoft stools per day. The median duration of treatment was 15 days (range 5 to 360 days). None of the trials followed up patients after the end of treatment.
Trial results were homogeneous. Compared with placebo or no intervention, lactulose and lactitol seemed to reduce the risk of no improvement of hepatic encephalopathy (relative risk 0.62, 95%confidence interval 0.46 to 0.84, six trials; fig 2). This result was not robust when trials were stratified by quality. High quality trials found no significant effect of lactulose or lactitol on the risk of no improvement (0.92, 0.42 to 2.04, two trials; fig 2), whereas low quality trials found a significant beneficial effect of lactulose or lactitol (0.57, 0.40 to 0.83, four trials; fig 2). Although this difference in treatment response was not significant (P= 0.3 by test of interaction), it is noteworthy that the event rate in the control groups was significantly associated with quality of methods (high quality trials 38%, low quality trials 78%; P = 0.0005 with χ2 test). The event rate in the experimental group was not significantly different in trials with high (35%) and low (43%) quality (P = 0.5 with χ2 test). The treatment responses in acute, chronic, and minimal hepatic encephalopathy did not differ significantly. However, there was no significant effect of lactulose or lactitol on acute (0.27, 0.02 to 3.28, two trials) or chronic hepatic encephalopathy (1.33, 0.41 to 4.33, one trial). Trials in patients with minimal hepatic encephalopathy found that lactulose or lactitol significantly reduced the risk of no improvement assessed by various psychometric tests (0.61, 0.47 to 0.79,three trials). These trials were all of low methodological quality.
Compared with placebo or no intervention, lactulose and lactitol had no significant effect on mortality (0.41, 0.02 to 8.68, four trials) or the number connection test result (weighted mean difference −9.0 seconds, −20.1 to 2.1, one trial) but tended to lower blood ammonia (−8.16 μmol/l, −16.44 μmol/l to 0.18 μmol/l, four trials). Data on adverse events were incompletely reported. Most trials mentioned adverse events associated only with non-absorbable disaccharides. We were therefore unable to perform a reliable meta-analysis of this outcome. None of the reported adverse events were serious, and all originated from the gastrointestinal tract (diarrhoea, flatulence, abdominal pain, or nausea).
Lactulose or lactitol versus antibiotics
Twelve trials with 698 patients (72% men) assessed lactulose or lactitol versus antibiotics (table 2).5 6 30–39 All patients had cirrhosis and acute,6 32 39 chronic,5 31 35 36 38 acute or chronic,30 or presumed chronic hepatic encephalopathy.33 34 37 Nine trials assessed oral lactulose,5 6 30 31 33–37 and three trials assessed oral lactitol.32 38 39 The daily mean dose of lactulose ranged from 30 g to 120 g (median 59 g) and of lactitol from 30 g to 60 g (median 60 g). The antibiotics were neomycin,5 6 30 ribostamycin,31 vancomycin,32 or rifaximin.33–39 The median duration of treatment was 15 days (range 5-90 days). One trial assessed all outcomes 15 days after the end of treatment,38 and onereported mortality 28 days after the end of treatment.39 All other trials followed the patients only to the end of treatment.
Trial results were homogeneous. Compared with antibiotics, patients taking lactulose or lactitol had a significantly higher risk of no improvement of hepatic encephalopathy (1.24, 1.02 to 1.50, 10 trials; fig 3). We found no significant difference in response to treatment between aminoglycosides and rifaximin (P = 0.2 by test of interaction) or when trials were stratified by quality or type of hepatic encephalopathy. We found no significantly different effect on mortality between nonabsorbable disaccharides and antibiotics (0.90, 0.48 to 1.67, five trials) or on adverse events (1.62, 0.57 to 4.58, eight trials). None of the reported adverse events were serious, and all originated from the gastrointestinal tract (diarrhoea, flatulence, abdominal pain, or nausea). Compared with antibiotics, patients on lactulose or lactitoltook on average six more seconds to complete the number connection test (weighted mean difference6.4 seconds, 1.4 seconds to 11.3 seconds, six trials) and had higher blood ammonia concentrations(2.35 μmol/l, 0.06 μmol/l to 4.64 μmol/l, 10 trials).
We did not find sufficient evidence to determine whether lactulose or lactitol have a significant beneficial effect on patients with hepatic encephalopathy. In our overall analysis non-absorbable disaccharides seemed to improve encephalopathy, but this effect was seen in only low qualitytrials.
The beneficial effect in low quality trials was related to significantly worse rates of improvement in the control group. This finding concurs with empirical evidence showing that low qualitytrials exaggerate the beneficial effects of treatment.15 16 40 Accordingly, the overall result may reflect bias because of the low methodological quality of most of the included trials. Our results may also be inflated by publication bias.
We found no significant effect of non-absorbable disaccharides on acute or chronic hepatic encephalopathy. Only low quality trials in patients with minimal hepatic encephalopathy found that lactulose had a beneficial effect, as assessed by various non-validated psychometric tests. The clinical relevance of these tests is uncertain.41
Lactulose has been used as the standard treatment for hepatic encephalopathy, and its efficacy has been considered to be beyond doubt.2 7 24 25 42 However, when it was introduced, the few trials that compared lactulose against placebo found no beneficial effect of lactulose.21 23 It was implemented in clinical practice because two trials found it “equally effective” to neomycin,5 6 which had been the standard treatment for hepatic encephalopathy since 1957.43 There are two major pitfalls in this reasoning. Firstly, the efficacy of neomycin in hepatic encephalopathy has never been shown. We identified only one randomised trial that compared neomycin with placebo44 and one that compared neomycin plus lactulose with placebo,45 both for acute hepatic encephalopathy. Both trials found no significant beneficial effects of neomycin. Secondly, lactulose was considered as equally effective to neomycin because event rates in intervention groups were not significantly different.5 6 However, lack of statistical significance does not imply that treatments have equal effects.46 Both trials were small,5 6 and neither reported sample size calculations based on an equivalencehypothesis or stated a margin of equivalence.46 47 It would require a far larger sample size than these two trials (a total of 78 patients) to establish with confidence that lactulose and neomycin have comparable effects.
Later on, new trials compared other antibiotics to non-absorbable disaccharides for hepatic encephalopathy. None was set up as an equivalence trial. Sample size calculations with statements implying an equivalence hypothesis or a margin of equivalence were not reported in any of the trials. All were underpowered to show equivalence. Nevertheless, all trials concluded equivalence from the lack of statistical significance.30–39 It seems that the research was continuously building up on both insufficient evidence and inadequate methods. Our analyses indicate that antibiotics are statistically superior to non-absorbable disaccharides in improving hepatic encephalopathy and lowering blood ammonia concentrations. However, it is unclear whether the effects are clinically important. Considering this, the lack of effect of antibiotics in placebo controlled trials,44 45 the risk of multiresistance,48 and the potential risk of severe adverse events5 lead us to conclude that there is insufficient evidence to recommend the use of antibiotics for hepatic encephalopathy.
When assessing intervention effects for hepatic encephalopathy, it is important to consider the fluctuating course as well as the impact of treating precipitating factors in acute hepatic encephalopathy. Well conducted placebo controlled trials on the use of ornithine aspartate in patients with minimal or chronic hepatic encephalopathy49 50 and lactulose plus neomycin45 in those with acute hepatic encephalopathy found improvement rates in the placebo group ranging from 40% to 70%. Many clinicians claim to have witnessed beneficial effects of nonabsorbable disaccharides on patients with hepatic encephalopathy. This effect may represent a high rate of spontaneous improvement and successful treatment of precipitating factors.
Non-absorbable disaccharides seem to have been introduced into clinical practice without appropriate documentation. This leads to at least three major problems. Firstly, patients are given a treatment of uncertain efficacy. It might be beneficial; it might be unfavourable. Secondly, there is reluctance towards performing randomised trials to assess lactulose or lactitol versus placebo because it is considered unethical. Thirdly, most randomised trials on new treatments for hepatic encephalopathy use lactulose as comparator. New treatments are considered effective if improvement rates do not differ significantly from the group treated with lactulose, although trials are vastly underpowered to show equivalence. This approach is most problematic. Nonabsorbable disaccharides should not serve as comparator in randomised trials on hepatic encephalopathy until other trials have shown that lactulose or lactitol has any beneficial effect on hepatic encephalopathy.
What is already known on this topic
Non-absorbable disaccharides are considered standard treatment for hepatic encephalopathy
Non-absorbable disaccharides serve as control treatment in most trials of new drugs for hepatic encephalopathy
What this study adds
There is insufficient evidence to determine whether non-absorbable disaccharides are of benefit to patients with hepatic encephalopathy
Antibiotics seem superior to non-absorbable disaccharides in improving hepatic encephalopathy,but it is unclear whether this difference is clinically important
Non-absorbable disaccharides should not be used as the comparator in randomised trials on hepatic encephalopathy
We thank the patients who took part in the reviewed trials; the researchers who provided us with additional information; J⊘rgen Hilden, department of biostatistics, University of Copenhagen, for statistical support; and Peter G⊘tzsche, Nordic Cochrane Centre, for valuable comments on an earlier draft of this review. This review is an abbreviated version of a Cochrane systematic review. The full version will be published in the Cochrane Library 2004, Issue 2. Cochrane Reviews are regularly updated as new evidence emerges and in response to comments and criticisms, and the Cochrane Library should be consulted for the most recent version of the review.
Contributors BA-N drafted the protocol and paper, performed the literature searches, identified trials, extracted data, and performed the statistical analyses. LLG identified trials and extracted data. All reviewers contributed to the writing of the protocol and review and all have approved of the final version. BA-N is guarantor.
Funding anish Centre for Evaluation and Health Technology Assessment (DACEHTA), Danish Medical Research Council, and Copenhagen Hospital Corporation's Medical Research Council
Competing interests None declared
Ethical approval: Not required.