The Thirteenth Step: Addiction in the Age of Brain Science

WHEN I WAS still in medical school, one of my instructors asked me what I wanted to do after I finished my training. I answered that I wanted to develop better treatments for people with addiction. He said, “Oh, you want to be a psychotherapist? Do you really need medical school for that?” I said no, I had seen people attend therapy for years while continuing to take drugs, and ultimately succumbing to them. I wanted to understand the brain mechanisms gone awry in addiction and see if we could do better, by finding medications that could somehow help those mechanisms get back on track. The instructor just shook his head. “Young man, how can you ever hope to treat a chemical addiction with yet another chemical?” This conversation has stuck with me. In endless variations, the statement is one I have since encountered countless times.

The statement was, of course, demonstrably wrong already at the time. Only four years later, in 1988, Vincent Dole of Rockefeller University in New York received the Lasker Medical Research Award for a medication that came to save countless lives and continues to do so. Here is the citation from the Lasker jury:

Although heroin addiction may seem like an extreme, methadone is a good place to start a discussion of medications for addictive disorders, for a couple of reasons. First, it provided the first proof-of-principle that an addiction could indeed be effectively treated with a medication. Second, and perhaps even more important, methadone maintenance illustrates the importance of the addiction mechanisms detailed earlier in this book and shows how those can be targeted with medications. The changes in brain function I earlier labeled allostatic neuroadaptations are lasting shifts in the baseline of the brain’s reward and stress systems. Once these shifts are in place, patients are miserable in the absence of drugs. They crave drugs both to get rid of that negative emotional state and to reexperience the high. Drugs such as heroin mimic the actions of endogenous morphine-like substances, or the endorphins, and activate the mu-opioid receptors through which these act to produce pleasure and suppress pain. But because it is so short acting, heroin potently triggers allostatic neuroadaptations, by first bombarding the mu-opioid receptors with a high intensity, then quickly coming off them. In contrast, methadone is extremely long acting. When adequately dosed, it is able to slowly activate the opioid receptors to the point of normalizing the patient’s state and suppressing craving for a long time. At the same time, it prevents the patient from experiencing the high from heroin should he or she still try it. Both effects work together to prevent relapse.

By now the data are indisputable. Methadone maintenance effectively helps people stay in treatment and reduce their heroin use.² It is consistently better than various control conditions, even when the methadone treatment is bare bones and has few other elements besides the medication. However, the outcomes vary markedly between programs and studies and become progressively better with increased intensity of psychosocial treatment.³ As a rule of thumb, in the absence of methadone maintenance, no more than 10 percent of patients are still in treatment after six months to a year, no matter how slow the detox from heroin, or how intensive the psychosocial treatment offered.⁴ When methadone is provided, even with little or no psychosocial treatment, retention in treatment is typically about 50 percent over the same time frame. In the best programs, those that combine methadone maintenance with intensive psychosocial treatment, the way Gunne’s program has always done in Uppsala, the corresponding number can reach 75 percent or better.⁵ Over the years, when data like these were presented internationally, they were often met with disbelief. Surely outcomes of treating heroin addiction, one of the most deadly addictions, just could not be that good. Maybe the results somehow reflected a selection of less severely addicted patients? Maybe there was something special about breathing the air in Uppsala?

I learned most of what I know about opioid maintenance treatment from Lars Gunne, one of the most generous, caring, and brilliant people I have ever met. Several years after we had first been in touch and he gently started to instill in me his wisdom, my group applied his treatment principles to a population of unselected heroin addicts, literally taken right off the street. As much as I admired and respected Lars, I was still amazed to see that we got results very similar to those he had obtained.⁶ The differences between programs are important to properly understand differences in outcomes beyond simply staying in treatment or reducing heroin use. Staying in treatment is just an indirect marker of success. What we really want to achieve is, first and foremost, that people don’t die. After that, we want to facilitate change so that patients don’t remain trapped in crime and prostitution. When all different methadone studies are put together in a meta-analysis, a reduction in mortality and crime rates is seen only as a trend, one that does not reach statistical significance. However, in studies with the better outcomes, such as those from Gunne’s program, both death and crime rates are robustly reduced.⁷ Perhaps one way of saying this is that methadone alone provides an opportunity to stay off heroin. The addition of psychosocial treatment provides the opportunity to make use of that pause from heroin to change one’s life.

Yet methadone maintenance is not without its challenges. On the physician’s part, it takes a while to teach people how to safely start up and manage the treatment. There are also some side effects that, although typically mild, can be problematic in a treatment that frequently needs to last many years or even decades. Almost all methadone patients become constipated, and quite a few become drowsy, so much so at times that this gets in the way of rehabilitation. The most important issue by far, however, is the risk for accidental overdose and death. Patients in methadone treatment develop tolerance to the medication and end up receiving daily doses—typically 70 to 120 mg—that will easily kill a person without tolerance, by suppressing their breathing. The risk is particularly high if people inject the methadone solution. Because of the overdose risk, great effort has to be devoted in any methadone program to prevent diversion of medication to illicit street use. In the early stages of treatment, patients have to visit a methadone clinic daily to receive their dose because their lives have not yet stabilized to the point that they can take responsibility for bringing a potentially deadly drug with them outside the clinic. Even once patients have been stable for a long time, most programs would hesitate to send along medication for more than perhaps two weeks. After years of successful treatment, by which time most patients are back to work, this creates unwelcome restrictions. And resources devoted to preventing diversion would of course be better used to treat more people.

In the 1990s a unique partnership among government, academia, and industry ⁸ brought to market a treatment for heroin addiction that has many of the advantages of methadone maintenance while addressing several of its downsides. Buprenorphine was first developed, in the 1970s, as a somewhat unusual pain suppressant. Similar to methadone, it is long acting and activates mu-opioid receptors, but it does so with a twist. It is what pharmacologists call a partial agonist. This means that like any drug similar to morphine or endorphins, buprenorphine recognizes the mu-opioid receptors and binds to them. The difference is that once buprenorphine is sitting on the receptor, it only partially activates it. This partial activation is most of the time sufficient to eliminate or at least reduce cravings for heroin, but it is not strong enough to give the patient much of a high or to suppress breathing. It does not matter how much the patient takes. The way partial agonists work, beyond a certain level, irrespective of the dose, there will not be more of an effect. We have reached a plateau. This means that the medication is easy to manage, is very safe to use, and does not pose an overdose risk, at least not as long as it is not mixed with a lot of other drugs. As a bonus, buprenorphine becomes even more tightly bound to the mu-opioid receptor than methadone. This means that it is better at blocking a possible high from heroin.

Just like one would expect from all this, stringent clinical studies showed that buprenorphine was highly effective as a treatment for heroin addiction,⁹ something that has since been confirmed by numerous controlled trials.¹⁰ The high safety and ease with which buprenorphine can be managed may sound mostly like a technicality, of interest only for the prescribing physician. In fact, in the real world, it has translated into tremendous opportunities for expanding access to treatment. Just as an example, when my group introduced buprenorphine in Sweden, only about 10 percent of heroin addicts there were in methadone maintenance treatment, waitlists were long, and people died while waiting to be admitted into treatment. When buprenorphine was introduced, waitlists were quickly shortened or eliminated, many times without the addition of any new treatment resources. By now close to 40 percent of Swedish heroin addicts receive treatment, half of them with buprenorphine.¹¹ In the United States the Food and Drug Administration approved buprenorphine for opioid addiction treatment in 2002, and the development was very similar. For the first time, it was now possible for a physician to take a course and then provide maintenance treatment to patients in his or her office, rather than in a traditional methadone program. The ease and lower cost of delivering treatment under these conditions quickly translated into a major expansion of treatment. A study from Yale University also showed that, just as one would expect, office-based treatment was able to attract people who otherwise had found treatment in a traditional program too stigmatizing to seek it out.¹²

A particularly clever use of simple pharmacological principles further minimized the diversion risk associated with buprenorphine treatment. Buprenorphine is taken under the tongue, from where it is taken up into the bloodstream. This route of absorption is far too slow to result in a high. But if someone really wanted to use buprenorphine to get a high, they could still try to dissolve it and inject the solution. To prevent this, a formulation was created in which buprenorphine was mixed with the opioid antidote naloxone. If naloxone is taken by mouth, it gets broken down in the liver before it has a chance to reach the brain.¹³ So as long as patients are using their buprenorphine-naloxone combination as they are supposed to, the naloxone ingredient simply doesn’t make a difference. But if the medication is injected intravenously, then the liver is bypassed. Both naloxone and buprenorphine now reach the brain. Naloxone blocks the opioid receptors and prevents buprenorphine from causing any receptor activation. Thus this is a package that retains its ability to effectively treat an otherwise deadly addiction but now has a markedly reduced overdose risk compared to methadone.¹⁴

The development of buprenorphine is a major success story of government, academia, and a drug company working together. The clinical benefits have been tremendous. With that as a background, this is perhaps the time to stick out my neck and say one more thing. Over the years, I have carefully guarded my independence from any drug company interests. To be credible, particularly in a field as contentious as this one, one simply cannot at the same time have a vested financial interest in any drug company, and its successes.¹⁵ But many people fail to appreciate that the realities of pharmaceutical research in the field of addiction are very different from those in other therapeutic areas, those that the industry considers profitable.

Marketing departments of pharmaceutical companies typically do not consider addiction medications to be worth pursuing. Because of that, the prevailing stigma, and concerns about risks involved in treating this complicated population, the problem we have in the addiction field is not too much pharma involvement—it is too little. With the exception of smoking cessation treatments, it continues to be very difficult to engage the pharmaceutical industry to develop medications for addictive disorders. That is a major barrier that results in very little translation of scientific advances in the addiction field into new treatments. Academia and government agencies can do many of the steps involved in medication development well. But there are other parts of this process that simply require the involvement of the pharmaceutical industry. In this context, it is important to point out that buprenorphine, in addition to saving countless lives, has also been a commercial success. Contrary to common perceptions, there is a market for effective, safe medications for addictive disorders. The partial agonism properties of buprenorphine are, by the way, probably attractive beyond the treatment of heroin addiction. We will revisit them when we get back to smoking cessation.

If experiencing the high were all there was to heroin addiction, a daily dose of naltrexone would be a great treatment for this condition. It is not. Because naltrexone does not result in any receptor activation, heroin-dependent patients continue to experience intense cravings, until most of them discontinue the medication and relapse to heroin use. Overall, outcomes with naltrexone for heroin addiction are no better than without treatment. As one would expect when compliance is the key issue for treatment success, the medication has shown some positive effects in studies where patients have been forced to follow naltrexone treatment, for example, as a provision of their parole.¹⁶ To improve compliance in other patient populations, injectable depot formulations of naltrexone have also been tried, so that the patient “only has to make one good decision a month.” Short-term results with those formulations have been encouraging.¹⁷ These results have been obtained by some of the best people in the field, whose judgment I have tremendous respect for. These colleagues are convinced that depot naltrexone is a treatment approach with a good potential in heroin addiction. I hope they are right. Yet knowing how difficult heroin cravings are to endure, I am skeptical whether patients will keep coming back year after year, unless they receive a treatment that dampens those cravings. And if they don’t come back for the next injection, the risk of relapse is there again.

Finally, to put medications for opioid addiction in context, it is important to realize that the face of this condition has been changing over time. Once, this therapeutic area was dominated by the need to treat severe cases, typically intravenous heroin addicts, many of whom had been recruited into drug use in the 1970s. Methadone maintenance, despite its challenges and the need for years or a lifetime of treatment, was usually appropriate in those cases, and that remains to be the case. Then there was a wave of new recruitment into opioid use among young people. It was not clear whether a decision to initiate methadone maintenance in an eighteen-year-old after a year or two of irregular heroin smoking could be justified. Yet at the same time, a failure to provide this treatment could mean that the patient would die of an overdose the next day. Buprenorphine offers an attractive treatment for this group. Finally, over the past decade, the United States has experienced a continuous rise in nonmedical use of prescription opioid analgesics. These medications act at the same brain receptors as heroin and are highly addictive. It is currently estimated that close to two million people in the United States meet abuse or dependence criteria for this substance class. Their use accounts for a larger number of overdose deaths than those from heroin and cocaine combined, or about twelve thousand each year.

With this changing panorama, I think that there is a need for a progression of treatments that can be tailored to the different levels of problem severity. Buprenorphine and methadone offer a logical progression that can be systematically followed. Buprenorphine, the safer, more easily managed medication, can be offered first. Even if it is somewhat less effective, it is reasonable to always try it first, just like penicillin should first be tried for pneumonia before using a broad spectrum antibiotic. If the patient does well on buprenorphine as first-line treatment, as about half the patients do, then the challenges of providing methadone treatment can be avoided. If, on the other hand, the first-line treatment is insufficient, then a rapid transfer to methadone can be offered. This approach maximizes the benefits while minimizing the risks and the downsides.¹⁸ Ideally, however, we would also like to have an option for cases where even initiating buprenorphine might seem questionable. In those cases, a medication without an addiction liability of its own would be ideal. Depot naltrexone may to some extent fill that niche. But if my hunch is correct and many patients will not stick with that treatment, then we should work hard to find other options.

In this book I have so far largely stayed away from smoking, other forms of nicotine addiction, and their treatment. As I pointed out at the beginning, there is no question that these disorders cause a tremendous disease burden, one that in aggregate takes a not-so-proud first position on the list of addictions. I have also mentioned that I have a couple of excuses for avoiding these important conditions. First, they are in a way a special case, in that the addictive drug nicotine, as such, is not particularly harmful. The harm is caused when motivation to obtain nicotine serves as an incentive for consumption of harmful products such as cigarettes. That is clearly different from, say, heroin addiction. Second, nicotine, as opposed to other addictive drugs, does not cause an immediate mental impairment. It improves cognitive function and is happily accessed, for instance in the form of nicotine gum, by some of my most responsible colleagues when they write important papers. Third, tobacco products, although regulated and equipped with warnings, remain legal. Combined, these differences have resulted in smoking cessation traditionally being offered outside my world of addiction medicine. And so I have to confess that I don’t have much hands-on experience from treating nicotine addiction. Having said that, it is rather clear that much of what applies to treatment of other addictions is also true of nicotine. This is certainly also true for pharmacological treatment, in ways that will nicely echo the principles just covered for treatment of heroin or other opioid addiction.

In brief, the principle of maintenance treatment to some extent works for smoking as well. The logic of that principle is to use a slow-onset and long-lasting stimulation of the same receptors that the addictive drug itself otherwise activates with rapid on-off dynamics. The brain receptor to target for smoking cessation is the nicotinic acetylcholine receptor, whose activation ultimately turns on classical dopaminergic brain-reward circuitry. Nicotine addicts certainly display an allostatic shift of their affective and hedonic balance after extensive drug use, leaving them miserable when abstinent. To have a vivid illustration of that state, one need only to talk to someone who is trying to quit. In this case, the nicotine patch provides the equivalent of a methadone effect, although the slow dynamics in this case are the result of the slow-release patch delivery replacing smoking. Of course it helps that nicotine as such is reasonably benign. A recent meta-analysis that included studies carried out in some forty thousand people found that the nicotine patch, or “nicotine replacement therapy,” improves chances of quitting by 50–70 percent.¹⁹ This seems to be the case irrespective of what behavioral treatments replacement therapy is combined with. Once again, maintenance with a medication that fully activates the target receptor works.

And so does “maintenance” with a partial nicotinic agonist, a principle similar to buprenorphine for opioid addiction. In 2006 the Food and Drug Administration approved the partial nicotinic receptor agonist varenicline, sold in the United States as Chantix and in the rest of the world, oddly enough, as Champix. It seems that varenicline is able to activate nicotinic receptors in the brain enough to suppress cravings for nicotine. However, in an analogy to buprenorphine, if the patient attempts to smoke, the receptors are already occupied by the medication. This makes smoking or other attempts to obtain nicotine effects largely meaningless. Just as one would expect from these properties, varenicline is a robustly effective smoking-cessation medication.²⁰ It is less clear why, in this case, the partial agonist is more effective than maintenance with the fully active agonist nicotine itself. One possibility is that the cravings for nicotine, pronounced though they may be, perhaps are not as intense as cravings for opioids, making the receptor blockade relatively more important for treatment success.

Finally, you may wonder if there is a smoker’s equivalent to naltrexone, which would block nicotinic receptors in nicotine addicts. Indeed there is. Mecamylamine is the right kind of nicotinic antagonist and has been tested for its ability to improve quit rates in smokers.²¹ It actually did help. Having read about how hard it is to make opioid addicts stick with their antagonist treatment, you will perhaps not be entirely surprised to learn that no pharmaceutical company has so far pursued developing this or any other antagonist, despite the gigantic market for smoking cessation.

Medications for treatment of alcoholism unfortunately do not follow the appealing logic of opioid or nicotine addiction treatments. That was perhaps only to be expected. After all, opioids and nicotine target highly specific neurotransmitter systems, in each case with an established role in the motivational circuitry of the brain. In contrast, alcohol does not directly interact with any specific brain receptor. It was long thought to affect the brain in entirely nonspecific ways, by acting as a detergent of sorts on fat layers from which membranes of all cells are made. It turns out that is not how it works. As I have described, alcohol influences several major neurotransmitter systems of the brain, including opioids, dopamine, glutamate, and GABA.²² It is hard to know which of those effects that should be targeted for treatment. To make matters worse, as mentioned previously, people differ in their responses to alcohol, based, among other things, on family history. This probably reflects genetic differences in how much different neurotransmitter systems contribute to a person’s alcohol response. The contributions probably also vary depending on the person’s sex, age, or stage of addiction. No wonder it has been difficult to develop effective medications for “alcoholism.” If we consider these individual differences, developing medications that are helpful for every patient may simply not be a reasonable expectation. I am convinced that, as new medications get developed, one of the most important tasks will be to identify the optimal intervention for every patient from a menu of possible options.²³

Before discussing neuropharmacological treatments for alcoholism, I need to briefly mention the alcoholism medication that is probably still best known to the public. Disulfiram, discovered in the 1920s, continues to be marketed, under the trade name Antabuse. Disulfiram does not intervene in the motivational mechanisms underlying alcoholism at all. Instead it simply blocks a step along the sequence through which alcohol is broken down and eliminated from the body. As a result of this blockade, the toxic substance acetaldehyde accumulates in the bloodstream. This in turn results in an extreme version of the Asian flush syndrome discussed in the genetics chapter. What that means is that if alcohol is consumed while a patient is on disulfiram, the accumulation of acetaldehyde results in facial flushing, as well as an extremely unpleasant and potentially dangerous reaction that includes a pounding heart and elevated blood pressure. This is intended to deter alcohol use, but of course it does nothing to reduce cravings for alcohol. If those cravings become too difficult to manage, the patient is still able to discontinue the medication and to relapse. As one would expect from this, disulfiram, offered as a prescription without supervised administration, is no more effective as a treatment than is placebo. It does, however, have some degree of efficacy if the medication is given in a supervised fashion, with a nurse watching and making sure that the patient actually takes the tablets every day.²⁴

The first medication approved for the treatment of alcohol addiction that can be considered modern, in that it acts on the brain, is one discussed in the opioid treatment chapter, naltrexone. In 1980 it was reported that when naltrexone was given to eight rhesus monkeys, it suppressed their alcohol drinking at doses that did not significantly affect water consumption.²⁵ Influenced by these data, Chuck O’Brien and his colleagues at the Philadelphia VA Medical Center in 1983 obtained approval from the Food and Drug Administration to use naltrexone as an experimental treatment in alcoholics. In pilot studies, several patients reported a lack of enjoyment from drinking alcohol while taking naltrexone. Based on those observations, Chuck and his trainee Joe Volpicelli carried out the first double-blind, placebo-controlled trial of naltrexone in alcoholism. Male military veterans in an outpatient treatment program all received regular counseling and group therapy using the twelve-step methods of AA. In addition, however, they were also randomized to receive either 50 mg naltrexone daily or placebo. The dose was selected because it had previously been observed to block the high from heroin. Whether through sheer luck, amazing foresight, or both, the outcome that was chosen was not the traditional, which is to stay totally abstinent. Instead the thinking was that if naltrexone blocks the reward from alcohol, then maybe people would still have slips but in the absence of reward would not progress further and have a relapse to heavy drinking.

Despite the intensive behavioral treatment, 54 percent of patients receiving placebo relapsed to heavy drinking within three months. As we have seen, that result is fairly typical. In contrast, relapse occurred in only 23 percent of patients receiving naltrexone. When Stephanie O’Malley and her colleagues at Yale conducted a similar study in both male and female outpatients, they obtained very similar results. Both sets of results were published, back to back, in the same issue of a leading scientific journal in 1992.²⁶ As a result, naltrexone was approved for alcoholism treatment in 1994 under the name Revia.

More than thirty years after these seminal studies, and some fifty randomized controlled trials later, there is no question that naltrexone treatment has positive effects in alcohol addiction.²⁷ But the story of naltrexone is one that invites both a bottle half full and bottle half empty perspective. On one hand, the development of this medication was a groundbreaking discovery. For the first time there was solid evidence that alcohol addiction could be treated with a medication that acted on the brain. Once again, my instructor in medical school and many other skeptics were proven wrong. Importantly, in contrast to methadone, this was not simply about replacing the addictive substance with a medication that acted as a stabilizer while still having an abuse liability of its own. Naltrexone does not have any addictive potential. Because of this, and because its positive effects were obtained in combination with traditional behavioral treatments, one could have hoped that its use would become uncontroversial. On the other hand, once all the naltrexone studies were put together in a meta-analysis, the effect size was not very impressive. In fact, it was on the same order as the better behavioral treatments. And the large American COMBINE study suggested that if a good behavioral treatment was given, naltrexone possibly did not offer any additional benefit.²⁸ With a limited effect size, no major pharmaceutical company committed to marketing it, and unexpected resistance among many treatment providers to medications of any kind, naltrexone just did not catch on. For a long time its use remained a marginal phenomenon, largely confined to academic treatment centers. And even there, less than 5 percent of patients with an alcoholism diagnosis received a naltrexone prescription.

But those of us who did take up prescribing naltrexone noticed something quite interesting. A significant minority of naltrexone-treated patients seemed to improve dramatically. Many of those patients had for years not been able to touch alcohol without having a major relapse. They were, if you remember the quote from the Big Book of AA, “powerless against the first drink.” Now, while on naltrexone, if they slipped, they did not seem to have much of a high and did not relapse to heavy drinking. This was indeed striking. Others, on the other hand, did not seem to show any response to the treatment at all. Sifting through the various naltrexone studies revealed that the characteristics of a naltrexone responder seemed to be that he was male and had a strong family history of alcoholism. Meanwhile, it had been reported that subjects with a family history of alcoholism had a significantly greater endorphin response to alcohol in the laboratory. This was at the time only possible to measure in plasma, but it could be speculated that a similar release occurred in the brain. A working hypothesis emerged from these findings. Perhaps alcohol activated endogenous opioid release in the brain, producing reward via some of the same pathways as heroin ²⁹ and ultimately activating classical dopaminergic reward circuitry? This was appealing as a hypothesis. But for a long time it remained just that—a hypothesis.

So there was a crossroads. If patients seemed to respond differently to naltrexone treatment and the overall effect was modest, one possible response was to write off the treatment and move on. That is essentially what most of the field did. Of course, in the age of the sequenced human genome and emergence of personalized medicine, there was an alternative option. If patients seemed to respond differently to a medication, maybe that was because they were, fundamentally, different. In this case there was particular reason to pay attention to that possibility. In 1998 Mary Jeanne Kreek’s group at the Rockefeller University in New York had reported that people differ in an important manner in the DNA code for the mu-opioid receptor, the target for naltrexone.³⁰ About 15–20 percent of white people had a variant of the receptor that seemed to respond differently to activation. Maybe this difference had something to do with medications targeting this receptor?

To test if these genetic differences could be related to the differences in naltrexone response, Chuck O’Brien and his colleague David Oslin set out to bring in DNA samples from people who had participated in three different naltrexone studies. Half the samples were missing. This was an “after-the-fact” analysis, of which geneticists are inherently suspicious. There were also other limitations. But if one was willing to look beyond those, the results seemed quite striking. It appeared that only a minority of patients—those with the less common receptor variant—had a benefit from naltrexone treatment. For them, on the other hand, the benefit seemed quite substantial. The remaining majority of patients did not seem to have any benefit at all. Once the pronounced effect obtained in the smaller responder group was mixed up with the lack of effect among the nonresponder majority, the average effect appeared small.³¹

These data seemed to potentially explain much of the naltrexone experience, but the results were controversial. Some analyses of other studies seemed to support the proposed pharmacogenetic effect; others did not. Some people welcomed the prospect of personalized treatment; others did not. When the Penn team submitted a grant application to fund a decisive, prospective study to answer the question, one anonymous grant reviewer wrote that “this could not be true; and even if it were, we would not want to know about it, because it would imply that not everyone is equally entitled to a treatment.” And the biological mechanism that could explain how the different receptor variants influenced naltrexone responses continued to elude the field. The controversy continued.

Meanwhile, our lab had carried out a series of studies in rhesus monkeys that could be pointing to the underlying mechanism.³² Following up on those findings, we were able to test Oslin and O’Brien’s hypotheses in people. To figure out the role of the mu-opioid receptor gene variants, Vijay Ramchandani in our program recruited one group of social drinkers in whom both copies of the receptor gene were of the common variant. He also recruited another group, of equal size, in which all participants carried at least one copy of the unusual receptor gene variant.³³ We recruited only men because women do not seem to respond to alcohol with much of a brain reward system activation in the first place.³⁴ We gave our subjects a very precisely controlled alcohol infusion in the PET scanner so that in fifteen minutes they went from not having any alcohol in the blood to 80 mg/dl, the legal limit for driving in most states. As we infused alcohol, we were able to measure the amount of dopamine released in the brain reward circuitry. The results were striking. Only people who carried a copy of the less common gene variant responded to alcohol with a robust dopamine release.³⁵ These were exactly the people who in Oslin and O’Brien’s analysis had seemed to respond to naltrexone.

The chain of events appeared quite clear. Just as Chuck O’Brien had hypothesized, alcohol must be releasing endogenous opioids. These then act on their mu-opioid receptors, ultimately activating the “final common pathway” of dopaminergic brain reward circuitry. This seemed to be happening predominantly, or perhaps exclusively, in people with a particular kind of genetic makeup, namely, those who carry the less common mu-opioid receptor gene variant. Obviously, if naltrexone intervenes into the cascade alcohol → endorphin release → dopamine release → reward, the medication can have an effect only in people in whom alcohol turns on this cascade. In others, there is simply nothing for naltrexone to work on.

Appealing though these results were, the proof that the gene variant actually caused the difference was still not watertight. Even if we just count genetic differences between people that are made up of one DNA letter exchanges, people differ from one another at more than three million places in the genetic code. Because of our origin as a species, and because of the way these differences have developed through our evolutionary history, these differences tend to travel in packs. What that means is that if you have a particular variant at a certain address in the genome, you are also more likely to carry a number of specific variants at nearby locations. So when we recruited people with the less common mu-receptor gene variant, we could, for all we knew, also have been recruiting people with some other variant that we did not even know about. That unknown variant could be causing the differences between people.

Normally this would be the end of the road. There is no way, in a human research subject, to determine if the differences we saw in the PET camera were really caused by the variant we thought caused them or some unknown neighbor. Working with my colleagues Wolfgang Sommer and Annika Thorsell, we therefore turned to genetically modified mice to address that final question. We spent several years making mouse lines in which we cut out the mouse mu-opioid receptor gene and replaced it with the human gene. We inserted the most common human gene version in one of the lines without any additional changes. In the other line, we started out with the same DNA code, but before inserting it into the mouse, we changed just one letter—the one that distinguishes the two different variants of the human receptor gene. When all this was done, we had mice that were genetically identical throughout the three billion letters of their genetic code except in the one place we wanted to probe. Working with Larry Parsons at the Scripps Research Institute in La Jolla, we were then able to directly measure the dopamine release in their reward circuitry when they received alcohol. Just like the people in our PET study, the mice with the less common gene variant released four times more dopamine.

Annika and Wolfgang, who by now run their own labs in Sweden and Germany, respectively, have since shown that these mice also consume more alcohol and have a much better response to naltrexone treatment. The story is remarkably consistent. The one-letter change in the genetic code has turned out to be critical. It seems that Chuck O’Brien had been right all along. By now a meta-analysis of clinical treatment studies confirms that carriers of the less common mu-opioid receptor gene variant are twice as likely to respond to naltrexone as patients without this variant.³⁶ The question is, what are we to do with these data in clinical practice? For another few years, we cannot yet count on every patient carrying his or her genome sequence on a card in a wallet. Neither can we yet expect the individual physician to order a lab test to determine the genetic makeup of a patient. But the days when a genetic profile will be available in clinical practice are not too far off. And until then we should remember that a good history may do a good enough job of identifying naltrexone-responsive patients. If someone is male, has a strong family history of alcoholism, and had a stimulant-like response from alcohol when first beginning to drink, chances are the patient will respond. There is little to lose from trying.

In a somewhat different way, the challenges to finding the right match between patient and medication are also illustrated by the other brain-acting medication that is approved for treatment of alcoholism. The U.S. Food and Drug Administration approved acamprosate, marketed under the name Campral, as an alcoholism medication in 2004. It had been approved in France already in 1989, and in the rest of Europe over the following years. In 1996 two ambitious evaluations of acamprosate were carried out at a large number of European treatment centers and published in leading medical journals.³⁷ Together these studies included more than seven hundred patients. The participants had been treated for a year in one of the studies, and forty-eight weeks in the other. Subjects had then been followed up for an equal amount of time after the study medication was discontinued. Notably both studies focused on a different outcome from that of the original naltrexone studies, which measured the ability of the medication to prevent relapse to heavy drinking. The outcome chosen in the acamprosate studies was instead the traditional one: the ability to increase chances of sustained abstinence. That is clearly a more challenging treatment goal. In both studies from 1996, only a small proportion of placebo-treated patients—about 7 percent in one and 17 percent in the other—were continuously abstinent throughout the duration of the respective study. In both cases this proportion was approximately doubled when acamprosate was given, a set of results that were both highly statistically significant. What was particularly remarkable was that the beneficial effects of acamprosate were sustained during the follow-up period, when patients no longer received any medication.

Today meta-analyses support the efficacy of acamprosate to improve abstinence rates.³⁸ There is, however, an interesting complication to those data. The meta-analyses are dominated by studies carried out in Europe, because that is where most of the acamprosate research was done. You wouldn’t think that would be all that important. After all, how different can a medication effect be, say, in England compared to New England? But something strange happened with acamprosate as it traveled across the Atlantic. A large study, with about six hundred participants, was carried out in twenty-one American treatment centers to support an approval of acamprosate as an alcoholism medication in the United States. When the results of that study were in, the acamprosate- and placebo-treated patients did not differ on the predefined outcome measure, the percentage of days patients were abstinent during the six-month-long trial. In some analyses there was some benefit to receiving acamprosate. For instance, among the two hundred or so people who entered the study with an explicit goal of achieving abstinence, the benefit was very clear. But these kinds of after-the-fact analyses, while often able to generate important new hypotheses, are typically not considered valid to directly support approval of a medication.³⁹ Then, a few years later, acamprosate was evaluated as part of the large COMBINE study mentioned previously, where naltrexone showed a modest but significant benefit.⁴⁰ Once again, no beneficial effect was found for acamprosate.

This seemed strange. The European studies were large, solid, and consistently positive. The American studies were large, solid, and consistently negative. What could possibly be going on? Although I cannot claim to have the definitive answer, I think it has by now become pretty clear that the effects of acamprosate differ from those of naltrexone in a fundamental way. For instance, if you give naltrexone to rats that self-administer alcohol, self-administration rates will consistently decrease, in any rat. That is what you would expect from a drug that in part blocks the rewarding, or reinforcing, effects of alcohol.⁴¹ I still remember our disappointment when we first gave acamprosate to rats under the same conditions in my old lab in Stockholm. Nothing whatsoever happened to self-administration rates. But the results were very different when we first induced a prolonged state of physical alcohol dependence in our experimental animals. After we had done that, we let them come out of withdrawal and then sit on a shelf for three weeks. By that time the animals self-administered twice the amount of alcohol compared to animals without a history of dependence. Under those conditions, acamprosate brought down self-administration rates by half, reducing them to the levels of nondependent animals.⁴² These and similar animal experiments suggested that for acamprosate to have an effect, long-term changes in brain function must be in place as a result of alcohol dependence.

With those data in mind, the European and American results make a lot more sense. The patients in the European studies typically had almost twice the alcoholism severity compared to those in the American trials. This is not because there is a European and an American version of alcoholism! It is simply a function of how studies recruit their participants on the two continents. All European countries have publicly funded health care, including addiction treatment centers. The standard approach to recruiting patients into a treatment study is therefore to go into the clinical population that seeks treatment at these centers and offer them a chance to participate. These participants tend to be more severely dependent, treatment-seeking, “real” alcoholics. In contrast, in the United States private health insurance, managed care, and liability issues make it difficult to access a regular clinical population for recruitment into a research study. So to a large extent investigators get their subjects by advertising in newspapers. That way they end up drawing from a population that tends to be socially more stable and in many cases perhaps would not have sought regular treatment. Those people tend to have less severe alcohol problems.

We still don’t know the exact molecular mechanism with which acamprosate works, but these observations fit in well with the things we do know. As alcohol is taken acutely, it dampens the activity of glutamate, the main excitatory neurotransmitter of the brain. But during withdrawal there is a rebound, and glutamate activity goes into hyperdrive. That is why during withdrawal almost all brain systems appear hyperactive and hyperexcited—as an example, just test someone’s knee-jerk reflex as he or she goes through alcohol withdrawal. Reflexes, of course, become hyperactive already at relatively modest withdrawal intensities. During severe withdrawal, excitability can become so pronounced that people go into delirium tremens or have seizures. Over time the excessive glutamatergic activity of acute alcohol withdrawal subsides and things calm down. But with repeated cycles of intoxication and withdrawal, brain function changes progressively, and there is more and more of a residual overactivity of glutamatergic nerve cells. Through mechanisms that are not yet well understood, this contributes to alcohol cravings and escalation of alcohol intake. Acamprosate can normalize this hyperactive glutamate transmission, as first shown in mice in a beautiful paper by Rainer Spanagel and his laboratory in Germany and then confirmed in patients with alcoholism by our own group.⁴³

So the lessons from the acamprosate experience once again highlight the need for personalized treatment, tailored to characteristics of the individual patient. In the case of acamprosate, however, what seems most important is not the genetic makeup of the patient. Instead it is the history, and specifically alcoholism duration and severity. If patients with low severity of dependence are treated, we are unlikely to get clinically meaningful effects. Acamprosate seems primarily to be useful in patients with a high severity and long history of alcohol dependence. Although that may not sound as a fancy or modern as genetically determined personalized medicine, it is of course just as important.

In addition, several other medications, although not approved for alcoholism treatment, have shown promising effects in alcoholism treatment studies. Some of these medications are already on the market, approved for other diseases. Perhaps best supported among them is the epilepsy medication topiramate, which in two large, independent, and well-designed studies were effective in reducing heavy drinking and also reduced the medical consequences of alcohol use.⁴⁴ Another very promising set of results was obtained with the muscle relaxant baclofen, which in an excellent study was found to be safe and effective for treatment of alcoholism in people with severe liver damage, who cannot be treated with naltrexone.⁴⁵ Neither topiramate nor baclofen is likely to receive approval as an alcoholism treatment, for purely commercial reasons. The costs associated with obtaining FDA approval for a new indication are considerable. Both baclofen and topiramate are old molecules; the baclofen patent expired in 1986, and the last patent for topiramate came to an end 2009. Generics have now long been available at low cost, and investing in approval of a new indication simply isn’t commercially viable. With recent lawsuits, it is also unclear whether “off-label” prescription will be possible, even with the best possible scientific evidence supporting these medications. Ironically, perhaps the only hope that alcohol-dependent patients will benefit from the elegant science is therefore if someone develops molecules that are similar enough to these two to have the same effects, but different enough so that they can be patented and sold at premium price.