Baclofen Study Results from Berlin 2016 – in detail

The new baclofen for alcoholism treatment studies were unveiled on Saturday 3rd September 2016 in Berlin at the World Congress for Alcohol and Alcoholism. It was a packed session as this was the most interesting news of the Congress.
At the Congress overall, there were a lot of rather obscure studies in animals and humans and studies observing patterns of alcohol use in various groups.
Here are some examples:
1. Investigating the role of one-carbon metabolism pathway in complicated alcohol withdrawal states.
2. SIAH1 mediates ethanol-induced apoptosis in neural crest cells by CBP/p300-mediated acetylation of P53.

It was as if the Congress participants were blind to a simple truth: the only useful way to treat alcohol related disease is to assist people to reduce or stop their excessive alcohol intake. The rest is just decoration. There was a lot of decoration at this Congress.

The baclofen session presented data from four studies:

1. BACLAD – this was published in 2015 and they had no new patient data to present. Their study was rigorously controlled (double blind, placebo controlled) and used individually titrated baclofen up to a maximum of 270mg/day with a three month study phase with success defined as complete abstinence for the whole three months. They had found 68.2% success in baclofen treated patients vs 23.8% in the placebo group, a whopping 44% improvement. The main problem with the BACLAD study was its small size, with only 43 patients in total.
At the Congress, they presented some MRI data done on the patient group but no data on any further patients.

2. Bacloville.-
Basically the results were: 56.8% of patients taking baclofen vs 36.5% of the placebo group achieved WHO criteria safe drinking levels measured in the last month of the 12 month study period.
Bacloville was the best of the studies by far: 60 general practices, 320 patients, all patients taken with no exclusions for psychiatric problems, psychotropic medications, or other types of substance abuse like opiates, THC or amphetamines. In fact they had their 320 patients after initial review of 323 patients proposed for the study: they excluded only three who were simply too close to death to go into a 12 month study.
This is very different from all the other studies. I found it hard to believe that the other studies could recruit ANY patients because they excluded anyone with really any problem (medical, psychiatric, other substance abuse) and this made their patient populations very atypical. But Bacloville steadfastly kept to a real life situation and accepted anyone with significant alcohol dependence.
As a result, Bacloville had a patient population with an elevated alcohol intake – it averaged 13 standard drinks per day (against WHO safe levels >2 for woman and >4 for men) so these are serious punters.
Bacloville was very different from the others in that it was done entirely in a General Practice setting with no detox or rehabilitation (inpatient or outpatient), no requirement to stop drinking or aim for abstinence and no prescribed psychosocial program. The baclofen was individually titrated to each patient’s needs with a maximum dose set at 300mg/day.
The primary outcome measured was simple: the % of patients who achieved WHO defined safe drinking levels (>2/day for woman and >4/day for men) in the last month of the 12 month study period. Result: 56.8% in patients taking baclofen vs 36.5% in the placebo group, an improvement of 20.3%
The Bacloville study only presented this primary outcome at the congress. The reason for this was the data was only unblinded (as to who got baclofen and who got placebo) 3 weeks ago so they did not have much time to pull out results. There are many other secondary outcomes that will be calculated for an upcoming research paper for publication. This will be very interesting as it will break down the data in more detail.
The 20.3% difference does not seem like a lot but you have to understand the details of the study to see why this is ok. The study used rigorous methodology that included strict “Intention to Treat”. What this means is that once a patient was entered into the Bacloville study, their outcome was counted, no exceptions. If the patient signed up for the study at a first visit and never came back, or stopped their study treatment after 3 weeks or died, they would be counted as a treatment failure even if it had nothing to do with baclofen treatment or the study.
This means that the baclofen failure group will include people who were not taking baclofen in the last month of the study when the outcome was measured and even people who had died months before. This is good methodology because it stops data being manipulated to improve results by excluding unfavourable patients but it does reduce the positive results, especially when the study period is long, in this case 12 months.
While initially I was disappointed that the difference was relatively small, on reflection I realized that it was what I should have expected. Of course what I really wanted to see was a big success rate in the baclofen group of 70-90% but this would be completely unrealistic given the “real life” conditions of the study. All baclofen prescribers know that if you take “all comers”, a reasonable proportion will simply drift away from baclofen treatment. It’s not that baclofen is not effective but the pain of being sober is not worthwhile for many patients whose lives have been decimated by alcohol.

There was a disarming simplicity and honesty about the Bacloville study.

Bacloville showed firstly that just having a GP supporting treatment for alcoholism was already a powerful therapeutic tool with a 36.5% success rate in the placebo group and that’s a very important message.
And if you add baclofen to the treatment, nearly 60% of these previously really heavy drinkers are now drinking at safe levels – this could be abstinence or low level drinking. That’s going to change their health and life expectancy a lot.

That’s pretty damn remarkable – alcohol dependence is considered to be largely untreatable. Not anymore.

And it’s not as if these GPs were specialised in this area although they are clearly motivated to use baclofen by agreeing to participate in the study. Each GP was only treating small numbers of patients – an average of 5.5 patients each. There are estimated to be around 10,000 baclofen treating GPs in France with most having only a couple of patients on baclofen.

The Bacloville study was a slap in the face for the many Addiction specialists present because it showed GPs treating a severe group of alcoholics and doing better than the other two new studies (shown below) which were done in specialized Addiction Services, one with less severely dependent patients.

I’m hoping that the secondary analysis of the Bacloville data includes looking at the patients who continued to see their GP took the medication for the whole treatment period of 12 months. I would expect more difference between these baclofen and placebo groups but, of course, only the data can tell us the truth!

So after some reflection, I think that the results of Bacloville are spot on. Not extraordinary because there is so much complexity in these patients and the cohort studied was severely affected by alcohol – multiple patients died from alcohol related illness in both the baclofen and placebo groups during the study. Someone from the audience questioned this as a dangerous effect of baclofen, even though it happened in the placebo group in equal numbers. The presenter, Prof Philippe Jaury, just shrugged his shoulders in a very french way and said “these were very ill people”. That’s right. Alcoholism causes a lot of premature death.

You can hear Prof Jaury’s actual Berlin talk on an MP3 recording (link) and see the slides (link).

The other two baclofen studies were negative in that there was no difference baclofen vs placebo. But the study designs were of such complexity with so many restrictions and such restrictive outcomes that they were always going to have problems.
The contrast between them and the simplicity of Bacloville was striking.

3. The Aldapir study was funded by the French pharmaceutical company Ethypharm which hopes to market higher dose baclofen tablets in France and thinks they can make a profit by charging a premium price for the convenience. At present only 10mg tablets are available in France and this means handfuls of pills are needed for patients taking high doses of baclofen.
Aldapir was done in a couple of French specialist alcohol treatment clinics as outpatients with a total study group of 316 patients. Despite this specialist setting, they had mild alcoholics overall, averaging 9.5 standard drinks/day (vs average 13/day for Bacloville). In the baclofen group, 13.5% had WHO safe levels of drinking as were 15.5% of the placebo group which is complete nonsense! Why would they be in a treatment trial? The presenter had trouble explaining why this was. In addition, 18.1% of the baclofen group and 14.2% of the placebo group were drinking at moderate risk levels. Only 70% of their patients were drinking at high or very high risk levels.
The reason they ended up with this skewed group is that their list of exclusions was monumental including any psychiatric illness or needing any “heavy” psychological or psychiatric support, any psychiatric medications apart from stable antidepressants. Also no epileptics, no people who had ever attempted suicide plus they needed to be detoxed for at least 3 days. I’m surprised they could find any patients who survived all their exclusion criteria! Then they had 40% drop out rate during the study.
They aimed to get all patients to a baclofen dose of 180mg/day over a 4 week titration period, irrespective of if they needed it. They “only” got 66% of the baclofen treated patients to 180mg but 90% of the placebo patients got there.
Interestingly there was no difference in adverse effects or side effects between baclofen and placebo even though they tried to push everyone up to 180mg/day, only stopping if the patient had too many side effects to tolerate more. It was therefore no great surprise that over 95% of the baclofen treated group had side effects/adverse effects but it was very surprising that the placebo group had an identical 95% rate!
Alpadir’s criteria for success was continuous abstinence for 20 weeks (after 2/4 weeks of the titration period) and not surprisingly, they had a low rate of success with ~11% both the baclofen and placebo groups (11.9% in baclofen and 10.5% in placebo).
The Alpadir group seemed to have realized that their group of relatively mild alcoholics was not ideal to get positive results – remember this is a drug company looking to sell high dose baclofen. So they did a series of post hoc analyses to show that there was more effect from baclofen in the 70% of the group who were heavy or very heavy drinkers ie they left the low and medium risk patients out. They also loosened their success criteria from continuous abstinence for 16 weeks down to looking at patients’ alcohol consumption levels. Then they found 10% improvement in the baclofen vs placebo group– not a great result for their cause but some data to use for government approval for marketing high dose baclofen tablets.
The Alpadir study really boxed itself in to a corner with multiple restrictions and the eminent professor who presented it looked pretty sheepish, especially after some very pertinent questions from the audience about the way the study was done.

4. The Dutch baclofen study presented had many of the same problems.
It started as a study of 151 patients in three groups: placebo vs baclofen 30mg/day vs high dose of up to 150mg/day but they rapidly abandoned the 30mg/day group due to a lack of money for the study and this left just 100 patients in the study. They had a 6 week titration phase and 10 week study phase. The primary outcome was time to first relapse into heavy drinking at which time they were considered treatment failures. So essentially, like the Alpadir study, success was continuous abstinence over the 10 week study phase.
Like Alpadir, the Dutch study had a big list of exclusion criteria – any severe psychiatric or medical illness, any other substance abuse, any other current or recent alcoholism treatment and the patients had to be abstinent for at least 4 days (the mean was 12 days) prior to enrolment in the study. The average daily alcohol consumption for their patients was 14 standard drinks/day, sitting above Alpadir at 9.5/day and equivalent to Bacloville at 13/day.
But the weirdest thing was their odd mix of patients. They were recruited from two inpatient and three outpatient treatment centres. In fact 70% were from inpatient centres, one of which kept patients in hospital for 6 weeks and the other kept them in for least 4 weeks. The inpatient group had all the baclofen/placebo titration done as an inpatient coupled with very intensive psychosocial support (at least twice daily!).
This meant that the baclofen treatment dose was fixed in this group while they were an inpatient. In the study protocol, the baclofen dose could not be further adjusted once they were discharged into the outpatient phase of 10-12 weeks. Once in the outpatient phase, they dropped from intensive support to twice weekly psychological support. But the 30% of patients recruited in the outpatient clinics had no inpatient time and only ever had twice weekly psychology sessions.
It’s a very heterogeneous group. I asked the presenter if they had looked at the differences between in- and outpatient centre treatment groups but the reply was that the groups were too small to analyse.
So 70% of patients had the baclofen dose set while in a sheltered rehab environment which it could not be adjusted once outpatients. Not surprisingly their average baclofen dose was quite low at 94mg/day.
The Dutch study measured their patients’ outcome in various ways but the simplest to compare with Alpadir is continuous abstinence in the 10 week high dose phase. The Dutch patients had a very high success rate in both groups – 62.5% continuous abstinence in the high dose baclofen group and 65.9% in the placebo group so no difference. This was despite the Dutch patients having a low average dose of baclofen (94mg/day vs 160mg/day in Alpadir) and high average daily alcohol consumption (14 std drinks/day vs 9.5/day for Alpadir). The placebo group has done so well that it would be hard for baclofen to outperform it.
The Dutch study presenter also looked pretty sheepish as the odd study design became fully apparent once the audience asked some questions. I asked about the inpatient vs outpatient groups as per above and whether the baclofen dose could be increased after patients were discharged into outpatient treatment as I was surprised by the low average dose of baclofen they used compared to the other studies.

So superficially we have two similar studies but with completely contrasting rates of continuous abstinence as their primary outcome – Alpadir 11% over 20 weeks and Dutch has 64% over 10 weeks. It wasn’t that Alpadir’s patients had low drinking levels similar to the Dutch cohort early in the first 10 weeks which rose over the extra 10 weeks in the Alpadir study period – the patients’ average total alcohol intake in both baclofen and placebo groups dropped in the first 4 weeks of the titration phase then stayed the same for the whole 20 week study period.

I’ve spent a long time combing the data on the slides of the Alpadir and Dutch studies and I can’t get a good feel for why they are so different except that the long inpatient stays in the Dutch study is very likely to play a role.
And I’ll throw in a mischievious suggestion which a Swedish researcher, Bo Soderpalm, talked about with me. It’s that alcohol intake via self reported intake is not an objective measurement. It can be pretty obvious when looking at a patient on treatment for alcoholism if they are doing well or badly but that’s not very accurate. What Dr Soderpalm was interested in was getting objective and sensitive measurements of alcohol intake into studies on treatment of alcohol dependence. There is such a test, phosphatidylethanol, a molecule only formed in the body when alcohol/ethanol is in the bloodstream. And it is broken down slowly in the body so it gives a measure of how much alcohol has been consumed over the last week or so. That’s very useful.

Dr Soderpalm was involved a trial of the anti-smoking medication Varenicline for treating alcohol dependence. When they measured the result of the Varenicline vs placebo groups by self reported drinking levels, both groups reported a marked reduction in alcohol intake with no difference between them. Both placebo and varenicline worked equally well. But it was another story when the levels of phosphatidylethanol were looked at. They showed that the placebo group were actually drinking the same amount of alcohol as they had been before they started the study medication but the alcohol intake of the varenicline treated group had dropped significantly. The paper doesn’t try to explain this – did the placebo group actually believe they were drinking less or did they know they were drinking as much but didn’t want to disappoint the researchers?
Interesting. It raises a lot of questions. I hope any future studies of treatments for alcohol dependence include this type of objective measurement so that we can be more confident of the data produced.
Here’s the reference for the paper if you’re interested:Varenicline for Treatment of Alcohol Dependence: A Randomized, Placebo-Controlled Trial: Andrea de Bejczy
et al. Alcoholism: Clinical and Experimental Research: Volume 39, Issue 11 November 2015 pp 2189–2199

A big positive for the Baclofen Study session was that all the studies showed good safety data with no significant differences between adverse events and side effects on baclofen vs placebo. This has long been an argument of baclofen critics: that baclofen has too many side effects and is too dangerous to use in alcoholic patients at the higher doses often required. The combined results of the four studies should help with getting baclofen approved for use in alcohol dependence in countries in the world where there are restriction on its use.

What became apparent in the Baclofen session was that the four studies were very different in severity of alcohol dependence, exclusion criteria, measures of success and duration of the study period.

The Bacloville study really stood apart from the others in terms of describing clinical practice which is widely applicable: no restrictive criteria applied to patients, no detox or rehab, individual titration of dose for each patient, general practice based care, looking for a realistic outcome of reduction to safe drinking levels and giving patients a year of treatment to start with. This is how I manage my patients and what I will be talking about to GPs in Australia. The results of the secondary analysis of Bacloville will be very interesting. I’ll bring them to you as soon as they reach me……