Placebo controlled treatment trials in hepatic encephalopathy are feasible and ethical


  • Potential conflict of interest: Nothing to report.

Gentile S, Guarino G, Romano M, Alagia IA, Fierro M, Annunziata S, Magliano PL, Gravina AG, Torella R. A randomized controlled trial of acarbose in hepatic encephalopathy. Clin Gastroenterol Hepatol 2005;3:184–191. (Reprinted with permission from Elsevier.)


Background & aims

Hepatic encephalopathy in cirrhosis is contributed to by toxic products deriving from the proteolytic bacterial flora–related degradation of dietary nitrogen substances. Acarbose is a novel hypoglycemic agent acting through the inhibition of glucose absorption in the gut and the promotion of intestinal saccharolytic bacterial flora at the expense of proteolytic flora. We assessed whether acarbose exerts a beneficial effect on hepatic encephalopathy and on postprandial hyperglycemia in cirrhotic patients with low-grade hepatic encephalopathy and type 2 diabetes mellitus.


One hundred seven cirrhotic patients with grade 1–2 hepatic encephalopathy and type 2 diabetes mellitus were randomized to acarbose 100 mg 3 times daily or placebo for 8 weeks; after a 2-week washout period, treatments were switched, and patients were followed for 8 more weeks. Ammonia blood levels, Reitan's number connection test, intellectual function, fasting and postprandial glucose levels, glycated hemoglobin values, and C peptide values were determined 2 weeks before and 4, 8, 11, 14, and 18 weeks after treatment.


(1) Acarbose significantly decreased ammonia blood levels and improved Reitan's test score and intellectual function score compared with placebo (P< .01). (2) Acarbose caused a 33% decrease in fasting glucose level and an approximately 50% decrease in postprandial glucose level compared with placebo (P< .01). (3) Acarbose significantly lowered glycated hemoglobin values and postprandial C peptide compared with baseline values, whereas placebo did not. (4) No change in biochemical parameters of liver function was observed after acarbose treatment.


Acarbose is a safe and effective drug in cirrhotic patients with low-grade hepatic encephalopathy and type 2 diabetes mellitus.

This prospective, placebo-controlled, double-blinded, randomized crossover trial is interesting from a number of perspectives. Firstly, it confirms that disaccharidase inhibitors can improve measures of hepatic encephalopathy (HE). This was previously reported by Uribe and colleagues using AO-128, another disaccharidase inhibitor, derived from valeolamine.1 Considering that this trial was also a prospective randomized placebo-controlled trial one is left to wonder why follow up on this mode of HE therapy has not occurred until now. Potentially, the side effect of severe diarrhea at full doses of AO-128 was a factor.

The second intriguing point raised by this current publication is the lack of worsening of HE in the 107 patients on placebo for 8 weeks. Over 50% of the patients had stage II HE at the time of randomization after a 4-week period of no HE therapy. Potentially therefore, if one includes the run in period, 8 weeks of placebo treatment and 2 weeks of washout, patients had no treatment for HE for as long as 14 weeks without any suggestion of worsening mental status. This type of persistent HE has been recognized to exist according to the new consensus terminology of HE.2 However, what is surprising to many is its presence (especially stage II HE) in such a large population of well compensated (Child–Pugh Class A n = 93, Class B n= 14) patients with cirrhosis. The complete absence of precipitating factors for HE in this large group of patients presumably is because of this lack of hepatic decompensation.

Could the investigators in this current trial be guilty of over diagnosing stage I-II HE in patients with well-compensated cirrhosis? There is strong internal validation that this was not the case in this study. The blinded assessment determined improvement only in patients on acarbose and their clinical estimates of mental status were paralleled by the performance times in the number connections tests (NCT). A typographical error in table 4 in the right hand column may have confused readers. The mental status score should have been in the 1.0-1.4 range and not 6.5 as stated in the table after the second placebo cycle. The correct value can be estimated by subtracting the value for NH3and NCT from the Encephalopathy global score.

There are some data omitted in this manuscript which might have a bearing on the very stable, unidirectional HE (i.e., improved on acarbose, no change on placebo) observed in this trial. The authors inform us that they excluded patients with stage III-IV HE from the study, but do not state if a history of this more severe form of HE was an exclusion criteria. Also, we are not given the details on how many patients were previously on active HE therapy. These two points are of significant importance in study design because Human Investigator Review Boards (HIRB) in the United States until now have expressed great reluctance in permitting patients to stop HE therapy. This is despite the fact that the need to be on maintenance HE therapy is often poorly documented.3

However in this study, patients had to have stage I to II HE to be enrolled. This suggests there was no deterioration in mental status during the run in no therapy phase. The authors really should have given specific information on the issue of percent of patients on therapy before the run in period. For instance, if more patients on prior therapy were randomized to the placebo arm, it could have biased the results. Another issue with this study is the potential unblinding of the patient and/or the assessor by the diarrhea induced by acarbose. This might most likely occur in patients previously on the therapy which causes diarrhea as they associate this symptom with efficacy. This is a study design issue with all types of treatments for HE treatments that cause diarrhea.

The major message of this publication to designers of HE treatment trials is there are patients with well-compensated cirrhosis and stage II in the outpatient clinic. If this HE population can be identified then design of placebo-controlled treatment trials is both feasible and ethical. Heretofore it has been assumed that the large clusters of patients with decompensated cirrhosis in liver transplantation centers would be the best pool from which to enroll patients for clinical treatment trials. However, patients with decompensated cirrhosis have frequent precipitating events and widely fluctuating levels of severity of HE.4 Almost by definition, designing clinical treatment trials in these patients is a formidable task and placebo treatment is generally neither advisable nor ethical to consider.

Assuming we can overcome our reluctance to believe that stage II HE can be commonly found in patients with Child–Pugh Class A cirrhosis, it can be predicted that patients similar to the ones in the trial by Gentile and colleagues can be identified. This identification process may require detailed face-to-face discussions between clinicians throughout the world interested in HE to standardize the clinical staging of HE severity. At this point there are few if any assurances that the clinical staging of HE is the same through out the world. Until there is a standardized mental status evaluation or an accepted neuropsychological measuring technique that is available we cannot easily compare studies. Hence meta-analysis or systematic reviews of the evidence in favor of treatments of HE may be inappropriate to perform.

In the meantime, all HE investigators would be well advised to look into these clues for stage II HE in patients with well-compensated cirrhosis. If few or none can be found, it might be worthwhile for all of us to visit Naples, Italy and determine what differences there are between our HE staging system and those of Gentile and his co-investigators. However, we predict these types of patients will be found opening the way for placebo-controlled treatment trials in HE treatment. Trials would then be far easier to conduct than if we continue to enroll patients with advanced cirrhosis.

One final point raised by this study is an issue that has arisen after discussion with the Food and Drug administration (FDA).5 This is whether a drug proven to be effective in mild (stage I, II) HE should be approved to treat more severe HE. Short of delivery problems of the active agents into the gastrointestinal tract there seems to be no rationale to suggest that an agent proven to be effective in mild HE should not also be approved for treatment of severe HE. Clearly circumstances may arise where a parenteral therapy would be desirable (i.e., ileus, bowel obstruction). However, as with any therapy for HE, it may be only feasible to conduct a study in patients with less severe HE simply because severe HE is generally seen in unstable patients with advanced liver disease and multiple concurrent precipitating factors and other confounders. Therefore, we may want to concentrate our efforts on mild HE in more stable patients with cirrhosis once we are more confident in the accuracy of mental status evaluation of less severe HE.