THE CALLER, a father of two boys, was a leading Swedish industrialist. People spoke of him as a good man, but even so, I found these referrals somewhat awkward. Most of our patients were disadvantaged; many were homeless. All had neglected medical needs. I had mixed feelings about treating or advising people who had vast resources available to them, and who expected us to give them special attention. We just didn’t do VIP treatment. When the caller asked if I could please see him and his sons in my private practice, I explained I didn’t have one. He insisted he could not be seen at the clinic with his sons. It was a simple statement of fact: they would not come with him. Could I please see them somewhere else? I apologized and said the clinic was the one place where I saw patients. He insisted. This would actually not be about treatment at all, he promised; it was just that his family needed advice. He thought maybe his youngest son in particular would find it easier to receive that from a stranger than from his father. I quietly thought back to my own father and the stormy years until he became my best friend. I guess I could relate. And I did of course have an office in our home on the outskirts of Stockholm, where I tried to write scientific papers at night, after the kids had fallen asleep. It would have to do.
As I was greeting them, I was struck by what a curious trio they were. The father, a commanding man in his fifties, blond with a bit of gray here and there, was still strikingly fit. He dressed with simple, understated elegance, and it was easy to see what a forceful presence he could be. But at the time, he was measured, almost restrained. After he introduced himself and his sons, he sat down and let them do the talking. The older son, on the Swedish national track and field team, had the cliché looks of an elite athlete. In this case that basically meant a younger version of his dad, with more muscle but the same quiet, restrained energy. It was the younger son who drew the most attention. Thinner than the other two men, fidgety and visibly less confident, he clearly found the situation awkward and took every chance to monkey around just to relieve the tension. His older brother looked embarrassed, and the jokes really were juvenile. But it was also clear that this was a sweet guy who adored his big brother and had tremendous respect for his father.
In a way the story was quite simple. Both sons were exceptional track-and-field talents, recognized as such from their youngest years, and fulfilling the promise at every stage. Good genes, I remember thinking. Their father, it turned out, had been a promising college athlete, too, before engineering and business studies, and then running a rapidly expanding company, set him on a different path. Over the course of the conversation, I learned that he still beat both of his sons at tennis and still did the classic Swedish cross-country ski race, Vasaloppet, which runs over a distance of 56 miles every winter. The older son was about to face the decision his father had once had to make—whether to pursue a career as an athlete and try to become competitive at an international level or to keep the good memories and focus on getting his engineering degree. He seemed to slowly be getting ready for the latter alternative but wasn’t there quite yet. The younger son was at a very different point. He was only now approaching the junction when he would reveal his true potential, as the junior team was occasionally invited to join the national team for training camps, and the coaches scouted for talent. Except every time that happened, things went terribly wrong. After the hard workouts, they partied. And every time they partied, this pleasant, polite young man drank everyone else under the table and did crazy things. On one occasion he would get into a fight over something silly; on another, he would walk out on a bridge, get up on the handrail, and do a balancing act 30 feet up in the air that only luck prevented from ending with a fatal outcome.
As I started to piece together the family history, the picture was clear. It had not been talked about much in the family, but the grandfather, an army officer, must have been an alcoholic. The father had decided to stay away from alcohol at a young age, while himself in the military service. Prior to that decision, there had been some episodes that involved crazy use of firearms but fortunately didn’t end with a tragedy. One day the young man who would become the father of the two boys in front of me decided never again to touch alcohol, and he kept that promise ever since. He had not talked with his sons about this experience until the older one tried alcohol around the age of eighteen and came to him, troubled by his own reaction. The two were very close. After a long conversation with his father, the older boy started a determined effort to practice skills that helped him handle drinking, much the way he would have practiced a new high-jump technique. He would still take the occasional glass but learned to stop at that. And for the most part, he just told people that alcohol would spoil some special training regimen he was following. That was hard to argue with. But his younger brother did not seem willing or able to master those kinds of tools. He just wanted to be one of the boys, like his big brother. And now he was about to be kicked off the team because of bad behavior.
I explained what we knew about how and why alcoholism risk runs in families. Thankfully the older two men had the good sense to keep quiet, sometimes nodding to affirm some of the things I said, sometimes just asking a clarifying question. I basically summarized what they had told me and put it in a broader context. The younger son stopped fidgeting and listened. After a while he became visibly upset and actually quite hostile. I shrugged it off. “Look, it’s not your fault and it’s not mine,” I said. “People are born certain ways and can neither take credit nor be blamed for it. You are blond and my hair is dark. I’ll never run 200 meters under 22 seconds, and you will always have a hard time controlling your alcohol reactions. It is not a big deal. But whether you like it or not, you are the one who will have to live with it.” His anger dissipated and turned into a sadness that for a while filled the room. We were quiet and let it sink in. As he started asking questions, it was a very different, subdued, more mature person talking. At some point I could also see curiosity take over. “How does it work?” he asked. I summarized some of the key studies. I also told him all the things we still did not know.
The young man had some unusual resources available to him. Like his father and his older brother, he was very smart. He also had their experience to relate to. He thought things were unfair, but he also realized that Mother Nature did not care much about his opinion in the matter. He got it, he said. But what if he still wasn’t ready to completely give up drinking? he asked. Would I still have some useful advice to offer him?
Everybody in the addiction field knows what to say in response at this point: “Young man, you have not yet fully accepted that you are powerless under your addiction and poor you until you do.” Except this guy was not even an alcoholic, at least not yet. And this was around the time we were beginning to figure out some things that challenged conventional wisdom. We already knew, for instance, from analyses that pooled results from many different controlled clinical studies, that a strong family history of alcoholism was a sign people had a better than average chance of responding well to naltrexone. Of course, the approved use of naltrexone was to take the medication every day. For someone who drank only a few times a year, that just didn’t seem reasonable.
I don’t know what got into me. I said, “You really should follow your father’s example. See how well he has done with his decision. But until you are ready to do that, here is something else we can try, almost as an experiment. I’ll write you a naltrexone prescription. When you absolutely want to go to a party where there is a chance you will use alcohol, take the medication a couple of hours ahead of time. Then come see me three months from now, at the clinic, and tell me how you are doing. And call us if you get in any kind of trouble.” He nodded quietly. I followed him for about two years. He relapsed to heavy drinking once. After more than a year of controlled and seemingly harmless alcohol use, he decided he was probably cured and tried to skip the medication. Next time, he drank himself to a blackout, fought with his best friend without later being able to recall over what, and ended up in a police cell. By the time he came back for an appointment to tell me about it all, he had already returned to the medication, and things were back on track.
Then, after about two more years of follow-up, he said he no longer needed to come back. I worried about him, but as he was leaving my office, he gave me a funny smile and reassured me. Half a year later he sent me a letter. In it he thanked me and explained why he no longer needed my help. He was close to twenty-one now. He had come to realize that when he drank without being on naltrexone, he was walking a tightrope over the abyss and would sooner or later fall. When he drank on naltrexone, it was pretty much pointless, with none of the bright light flooding his mind, and the most prominent result instead being headaches. As he was mulling over his options, he had encountered the old Mark Twain quote: “When I was a boy of fourteen, my father was so ignorant I could hardly stand to have the old man around. But when I got to be twenty-one, I was astonished at how much the old man had learned in seven years.” He talked to his father. After a series of conversations, he decided that for them, drinking was just not an option.
Alcoholism runs in families. Laura Bierut, a leading genetic epidemiologist of addiction at Washington University, St. Louis, likes to remind her audiences that this insight is in many ways unremarkable. Already the ancient Greeks seem to have attributed a significant part of alcoholism risk to genetics. Aristotle wrote that “drunken women bring forth children like themselves,”
1 while Plutarch stated that “one drunkard begets another.” Some two thousand years later, similar observations were noted in Scandinavian church books, one of the best-preserved sources of multigenerational family history information recorded to this time. In the margins of those books, notes can frequently be found from long forgotten ministers describing how drunkenness had run in a family for generations. Interestingly, that insight never seems to have prevented those ministers from preaching that the drunkard could and should change his ways. Clearly, in their mind, acknowledging a genetic influence did not imply a view of genes as destiny.
The fact that alcoholism and other addictions run in families does not, on its own, prove that there is a substantial genetic component to the risk for developing these disorders. It could well be that the risk, for the most part at least, instead comes from growing up in a family affected by addictive disorders. Other shared environmental factors, outside the family, could contribute too. Addiction is after all strongly associated with growing up and living in poverty, being exposed to violence, or having a low level of education. Each or all of these factors could theoretically contribute to addiction risk. Whether most of the risk comes from genes, the shared environmental factors, or a combination of the two is therefore an empirical question. Simply stated, good-quality data about the world are needed to decide it. It cannot be settled on the basis of one view being more appealing than the other. Nor is it helpful to consider which view is more in line with an enlightened political agenda. Yet that is exactly how smart, well-educated, and well-meaning people for a long time argued about these matters.
In short, for decades the argument of progressive intellectuals and their followers was that acknowledging a role of heritable, biological factors for complex behaviors such as drug seeking and excessive drug use leads to “a reductionist, biological determinist explanation of human existence. Its adherents claim, first, that the details of present and past social arrangements are the inevitable manifestations of the specific action of genes.”
2Amazingly, world-class scientists such as the Harvard evolutionary biologist Richard Lewontin helped lend credence to this type of argument, and the quote above is in fact from a book Lewontin coauthored. For a long time those of us who dreamed of improved conditions for poor, disadvantaged, and addicted people were expected to subscribe to the notion that the human mind must start out as Locke’s blank slate,
3 fully malleable by society’s influence.
4 Only that would be compatible with progress. Anything else, and in particular considering an important role for inherited biological factors, rapidly qualified one for a label as a reactionary.
This was very peculiar, even as a hypothesis, for several reasons. First, humans have since the dawn of written history known that physical traits such as size have a strong genetic influence. For instance, we have long known and made use of the fact that domestic animals can be bred for those traits. A chihuahua will not grow into a mastiff no matter how much it is fed. But just like breeding large dogs and small dogs, we have also bred watchdogs that are territorial and aggressive, as well as child-loving black labs that will happily lick burglars all the way to the door when they leave with their loot. So to claim that there could not be a genetic influence on temperament or other complex behavioral traits would require that humans somehow be fundamentally different in their genetic transmission or their brain function from all other mammalian species. That is a claim for which there has never been a basis.
Second, the very reason physicians like myself want to know about genetic risk factors is that we think they may offer our patients means to intervene and change potentially devastating outcomes. Consider if your grandfather, your father, and your brother all got in trouble with alcohol, and if, on first trying alcohol, you experienced that you could drink more than anyone and enjoy it. Would you under those circumstances not want to consider the possibility that there is something about the response to alcohol running in your family that puts you at risk? But if you realized that to be the case, decided alcohol is just too dangerous for someone with your disposition, and stopped using it, you would radically change your fate, would you not? Or maybe, in the age of personalized medicine, you would instead decide to use a medication that prevented that enjoyable reaction to alcohol, allowing you to avoid the risk that way. Now, how is that for “biological determinism”?
There is in fact nothing deterministic about understanding the influence of genes for behavior. Knowledge is power. As every reader of Percy Jackson knows, gnothi seauton is the first advice the Oracle has to offer anyone on a quest to achieve something difficult and important.
Having said that, it is not a trivial task to parse out the contribution to addiction risk from genes, on one hand, and environmental factors, on the other. It is even harder to find the specific genes that contribute to the genetic risk. As we will see in the following, the past thirty years or so have seen some remarkable scientific advances in the former domain. Converging conclusions have been reached using quite different and complementary scientific methods. While each of those may have its weaknesses and potential sources of error, those sources are different for the different methods. The convergence therefore makes it highly probable that the conclusions are correct. I will in a moment explain in more detail what that statement means, and also, importantly, what it does not mean. The fundamental issue is, however, largely settled: addictive disorders are all moderately to highly heritable.
The other set of questions, prompted by that conclusion, has turned out to be much harder. What characteristics are being inherited that put people at risk for addiction? And what are the gene variants that contribute to the heritability of addiction that research has established? How many of those variants does it take for a significant level of risk—a few, a few hundred, a few thousand? We don’t know, but it looks like it might be more than many of us thought just ten years ago. Progress in finding these risk variants has been frustrating, but there are advances in that area as well.
Umeå, home to Sweden’s northernmost medical school, is located about 250 miles south of the polar circle, but you wouldn’t know it from the weather. As far as the climate is concerned, this could just as well be the North Pole. The saying goes, if you survive the first winter, maybe you’ll stay, but most people don’t. By the time I met him some fifteen years ago, Michael Bohman had long retired as chair of child psychiatry in Umeå, had moved back to his native Stockholm, and was mostly pursuing his literary interests.
It was only when I actually met him that I realized Michael had started out as a psychoanalytically trained child psychiatrist. As a young physician, he told me, he had taken for granted the psychoanalytic teaching that early life experiences shape people’s lives in major ways. He got involved in adoption studies because that line of research seemed to offer a promising way of finding out about the influence of early family environment and parsing it out from any genetic factors. In principle, if early family environment was influential the way psychoanalytic teachings had hypothesized, then various behaviors of children who were adopted early in life should be more influenced by their adoptive than their biological parents. Michael realized, as many genetic epidemiologists have since, that Sweden offers unique opportunities for this kind of research. Say what you will about the social democratic welfare state, but Sweden is sure good at keeping track of people, from birth to the grave. After all, how else would we know where to send all those welfare checks to everybody?
Michael collected the data for the Stockholm Adoption Study before landing his faculty position and moving north, but he continued to work on them after the move. The study, he wrote in a 1981 paper,
was suitable for cross-fostering analysis because it involves a large population in which there has been no selection among the adoptees for particular types of parents or children. The population includes all persons born out of wedlock in Stockholm from 1930 through 1949 and placed for adoption. Most of the children were separated from their biological relatives in the first few months of life. The average age at final placement in the adoptive home was 8 months. Sweden is an ideal site for an adoption study. Extensive social and medical records about the adoptees and their parents have been obtained from several public sources by Bohman and his associates. Alcohol is readily available for consumption, but its abuse is under rather intense social sanction. A Temperance Board in each community is legally responsible to maintain sobriety. This board imposes fines for intemperance and supervises the treatment of alcoholics. Other records about diagnoses and treatment are kept by agencies of the National Health Insurance.
In contrast, in the United States, records about biological parents obtained by adoption agencies are often incomplete and limited to the period prior to adoption, which usually occurs when the biological parents are quite young and not through the age of risk for psychiatric illness. The availability of objective and systematic lifetime data in Sweden avoids the potentially serious biases inherent in assessments based on self-report.
5
It was clear that the data collected through the Stockholm Adoption Study were unique, but Bohman and his collaborator Sigvardsson, a psychologist, were initially struggling to make sense of them. The results simply did not seem to fit their hypothesis by any stretch of imagination. In fact, the data suggested something very different. Now, most scientists, when faced with data that don’t fit their hypothesis, will conclude that there is something wrong with the dataset or that the experimental measurements are somehow flawed. Or they will produce a softened variant of the hypothesis, one that does not need to be rejected in the face of the available data after all.
In what we should now begin to recognize as the hallmark of a truly great scientist, Bohman did nothing of the kind. Instead he began to realize that Mother Nature was trying to tell him something he had not anticipated. But he still didn’t know how to analyze the data to best address the research questions. He began reading papers on separation experiments, published by a young American psychiatrist who had trained with early leaders in the field of statistical genetics and who seemed to be developing innovative approaches to parsing out the influence of genetic and cultural inheritance. They met at a conference and discussed the possibility of a collaboration. The American immediately realized the value of the Swedish data. In what became the beginning of a lasting collaboration, he came up with a sophisticated statistical approach that enabled the cross-fostering analysis.
In brief, the analysis, which over the years has become an official citation classic, showed that adopted men ran a risk of developing early onset, severe alcoholism that reflected the risk of their biological, not their adoptive, fathers. This confirmed indications from earlier, more limited adoption studies such as that carried out by Donald Goodwin and Sam Guze of Washington University, St. Louis, using a Danish adoption registry.
6 For good measure, Bohman and Sigvardsson replicated their Stockholm Adoption Study fifteen years later, using the adoption registry from the second largest Swedish city, Göteborg. The results were nearly identical.
7 Similar data followed from other groups.
Of course, if this research really is all that important, readers in the United States may wonder why they have not heard of these groundbreaking Swedes. It is a safe bet that neither Bohman nor Sigvardsson is a familiar name to most Americans, not even to those who study addiction. Instead psychiatrists and addiction scientists all over the world are more likely to be familiar with another name, Bob Cloninger, who is one of the most cited psychiatric researchers of all time. The author of some four hundred scientific research papers and a member of the Institute of Medicine of the U.S. National Academy of Sciences, he was for years chair of the psychiatry department at Washington University. Cloninger is the young American with whom Bohman once initiated a collaboration, and with whom he then continued to publish for many years.
Although clearly a landmark, the adoption studies could still be questioned. In some of them, it was not possible to exclude that exposure to the intrauterine environment was a factor, although that could not be the case in the Stockholm Adoption Study, which for that very reason restricted the analysis to fathers and sons. Even with that caveat taken care of, few adoptions occur just outside the delivery room. One could speculate that even a few months at the beginning of life could have a decisive, lasting influence on how people turn out. But the adoption studies marked the beginning of an era in which much improved insights into the role of heritable and environmental factors for addiction risk were achieved. In the more than three decades that have passed since these seminal studies, a much more complete picture has emerged from a range of different approaches. The basic findings of a substantial genetic contribution to the risk for developing addictive disorders have received firm support over these years. The most important source of this expanding knowledge has been twin studies, originating from large twin registries in Denmark, Sweden, Minnesota, and Virginia, and from the Vietnam era twins.
Before going into those findings, I need to say a few words about twin studies. Despite a fascination with twins that occasionally results in newspaper headlines like “They Got Married on the Same Date Without Even Knowing It,” the scientific opportunities offered by twin studies are in my experience not well understood by the public. It is not unusual to hear research results originating from twins dismissed with comments such as “of course twins will be similar to each other in their risk for addiction, they grow up in the same families. So any similarities don’t have to reflect genetics, at all.” In fact, twin studies do not base their assessment of genetic influences on the degree of similarity between twins in general. What they do instead is make clever use of the fact that twins come in two varieties. Fraternal twins are genetically no more similar than any other pair of brothers or sisters—they share, on average, 50 percent of their genes. Identical twins, on the other hand, are just that—identical. All the three billion or so letters in their DNA code are the same. So it really doesn’t matter whether, for instance, one twin being alcohol dependent is accompanied by the other sharing the same fate in 10, 40, or 70 percent of cases. What matters is whether that degree of similarity, called concordance, is the same or different in identical and fraternal twins.
If concordance is the same in fraternal and identical twin pairs, despite the much higher degree of genetic similarity within the identical pairs, then genetics simply cannot matter. In that case, heritability, the proportion of risk contributed by genes, must be zero, and any differences between twins must be attributed to environmental influences or chance. But if, on the other hand, the concordance within identical twin pairs is significantly higher than that within fraternal pairs, then genes must contribute, at least if some simple assumptions hold. This is not to say that shared family environment could not still play a role. The beauty of this approach is, however, that the role of the environment does not interfere with our ability to measure the genetic risk. This is because environmental factors presumably affect twins to the same extent whether they are fraternal or identical. Your environment—whether, for instance, your parents were alcoholics themselves—presumably does not care much if the fertilized egg from which you originated happened to split moments after conception, resulting in there essentially being two of you.
8 If this assumption holds true, then whatever the contribution of the environment was, it can be simplified out of the equation.
Applying some sophisticated math—called structural equation modeling—to this kind of data allows an equation consisting of three terms to be estimated. For any trait, such as in our case a diagnosis of an addictive disorder, and for any level of total risk for developing that trait, that risk is decomposed into three components that can be expressed as percentages, obviously adding up to 100:
• a heritable component, or “heritability”
• an environmental component that is shared between children in a family, such as growing up with alcoholic parents, or “shared environment”
• a term that bundles all the different life experiences that are unique to the individual—say, having been in a car accident—including sheer good or bad luck, or “individual environment”
In 2012 Ken Kendler of Virginia Commonwealth University received the Mark Keller Award from the National Institute on Alcohol Abuse and Alcoholism for his pioneering work on the genetics of alcoholism. This work, much of it carried out with his former colleague Carol Prescott, now at the University of Southern California, has used twin data to answer many of the key questions of the field. The award lecture, one of the most enjoyable and thought provoking I have heard in many years, showed what a great rabbi the world lost when Ken decided to become a physician scientist, and what a wonderful scientist and academic teacher we gained in the bargain. Pooling available research from many sources, the different twin studies agree to a remarkable degree on the heritability of alcoholism, which they estimate to be somewhere around 55 percent. The same twin data show that the shared environment also contributes to a significant extent, accounting for just under 40 percent of the risk. The heritability of other addictive disorders has been estimated in the same range, with heritability of addiction to opiates and cocaine perhaps being slightly higher than that to alcohol.
9 This places addictions in a range of genetic influence that is higher than that for depression or anxiety disorders but lower than for schizophrenia, bipolar illness, or autism.
The pooling of the various twin studies yields some other important insights. In contrast to what Bohman and his colleagues thought, heritability of alcoholism seems to be almost identical for men and women. And in a case in which it has been possible to examine whether heritability of alcoholism has changed over half a century of changing societal conditions, little difference over time was in fact found. By the way, these heritability estimates from the twin studies are if anything likely to be underestimates. As Kendler points out, establishing a diagnosis of an addictive disorder such as alcoholism is fraught with uncertainty, and that adds noise to the calculations. If the data are corrected for that uncertainty, the true heritability of alcoholism is likely closer to 70 percent.
By now maybe I have convinced you about the proportion of addiction risk that comes from genes. Or maybe you did not need convincing, in which I case I have perhaps reinforced everything you already knew, hopefully in an interesting way. In either case, it is time to make life a bit harder. Having come this far, how about if I told you that the coveted, hard-earned heritability numbers we have painstakingly gone over perhaps are not meaningless but mean something quite different from what one may first think? To illustrate the fundamental issue at hand, let’s carry out a thought experiment. Clearly addictions are in the broad category of complex traits—think, for example, of positive emotionality or mental toughness, both estimated to be about 50 percent heritable
10—to which both genes and environment contribute, and which are important for how people fare in life. Let’s now imagine a dream society in which we have identified every environmental factor that adversely influences these traits, and where we have waved a magic progressive wand to eliminate each of those factors. Likewise, in this society, we have identified every positive environmental influence and made sure that every citizen has equal access to it. These interventions would clearly have a major positive impact overall. But because everyone is now exposed to the same optimal environment, there can no longer be a contribution from the environment to differences in how people turn out. Because everyone is exposed to the same perfect environment, any remaining differences between people, although perhaps smaller than before, must now entirely be caused by genes. With the exception of a small term due to chance, heritability all of a sudden approaches 100 percent.
This extreme example of an ideal society has not been achieved anywhere yet, not even in Sweden. But it illustrates a fundamental aspect of heritability. As a measure of how much influence genes have, it is relative rather than absolute. And any estimate of this measure is valid only under the conditions under which it was obtained. Following this thought, maybe the heritability of some important trait on which people differ is lower in a rough, capitalist society but higher in an affluent, well-educated liberal paradise? If you think I am joking, think again. The sweeping claim is often made that aspects of intelligence are highly heritable. People who don’t like that idea dismiss this claim in an equally sweeping manner. But neither statement may be all that meaningful. A remarkable piece of work assessed twins at different levels of socioeconomic status and found that in impoverished families, environmental factors accounted for 60 percent of individual differences, while the contribution of genes was close to zero. In contrast, in affluent families the result was almost exactly the reverse.
11 Another way of saying this is that if a person’s material resources are very limited, they put a limit on achievement so far below the person’s full, genetically determined potential that those heritable determinants don’t matter much. If, on the other hand, a person doesn’t have any constraint from material resources, then only biological endowment will set the limit.
To properly interpret the results of research on how much genes contribute to addiction risk, we must thus be careful how we frame them. The message so far can perhaps be stated like this: the contribution of genes to the risk for developing alcoholism and other addictive disorders is somewhere between 55 and 70 percent in mostly white populations of industrialized, Western countries in the past half century. Beyond that, it is anybody’s guess. There are no records of good twin studies in the papyrus rolls recovered from ancient Egypt.
Establishing the extent to which addictive disorders are heritable is a major scientific advance and settles issues that are as important as they have been contentious. Yet at the same time, these data do not tell us much that is useful about what characteristics or “traits” people inherit that put them at risk for addiction. The heritability numbers say even less about the specific gene variants that contribute to those characteristics. Answers to those questions are not just of academic interest. Those are the answers that ultimately will be needed if we are to put knowledge about genetics of addiction in the service of patients and their families.
In my experience, physicians, policy makers, and others who are keen to improve public health can talk all they want about the dangers of drinking too much or using drugs. And we do. People nod politely and then go on with their lives. It is only when people are able to relate the data to their own lives—perhaps when they can see parallels to that uncle who didn’t fare so well—that the risk feels real, and that the information is able to plant a seed of change. Because of this psychology, it will only be once we are able to make predictions about individual risk that we will be able to make personalized interventions and more effectively prevent harmful outcomes. Those interventions can for starters be of the simplest possible kind, one for which we have already seen an example. We can simply advise people at high genetic risk to abstain from a drug that might otherwise kill them, even though others may be able to use the same substance without much harm and with enjoyment. In the process, this distinction effectively gets us away from a double standard, a moralizing position of talking about how bad alcohol is while allowing it to be marketed and sold.
And beyond prevention there is of course the promise of personalized treatment. Once individual heritable factors behind addiction become better understood, we will finally be able to use advances of genetics to guide development of new medications. We will see in the following that different genetic factors contribute in different cases, and once we are able to identify those factors, we will be able to choose the right medication for a patient. For all this promise to materialize, we will need to know much more about the specific gene variants that are involved, not only on average in the population but also in the individual case. Some progress has been made to advance this type of knowledge, but in general the specific traits and genes that carry addiction risk have turned out to be a lot harder to identify than many of us had hoped.
An important reason is probably that different people, even though they may qualify for the same addiction diagnosis, are in fact very different. One person may quietly increase his alcohol consumption over many years while holding a job and supporting a family, until at the age of fifty or so he can finally no longer be without alcohol and starts planning his life around the drug. In contrast, someone else may start drinking early in life, tolerate excessive amounts of alcohol right away, and become aggressive when drunk, in ways so disruptive that she never succeeds in getting an education or a job. One person may be anxious and drink just so that he can leave his home for the grocery store. Someone else may be impulsive and not give much thought to consequences of consuming excessive amounts of drug for performing adequately at work the following day. To make things even more complicated, people who have a high genetic risk for addiction may not require much environmental influence to develop a diagnosis. But others, even without much genetic vulnerability, will ultimately develop the same diagnosis if they keep using a drug for a long time, to the point that their brains undergo changes described in prior chapters.
12In the practice of medicine, we tend to focus on what people who seek treatment have in common so that we can establish a diagnosis and use that diagnosis as a basis for treatment. Addiction medicine is no exception. But while people with an addiction diagnosis may all seem quite similar, they are in fact, in many equally important ways, fundamentally different. A way of thinking about this is that there are many different pathways through which a person can get to the point of qualifying for a clinical diagnosis of an addictive disorder. By the time patients seek treatment, they may be similar in the sense that their lives center around seeking out and consuming alcohol or drug in excessive amounts, at the expense of other aspects of life. That does make it critically important to identify the addictive disorder as a clinical problem that requires attention. But that is just a snapshot, here and now; it does not tell us much about the pathway through which the patient got here. It may therefore entirely miss equally critically important information about what biological and psychological mechanisms have become engaged along the way. To me this is much like a diagnosis of heart failure. Most people are familiar with the symptoms of a patient who receives that diagnosis. The swollen legs and the shortness of breath are characteristic. But that is also an end-stage condition. In one patient it perhaps resulted from coronary disease. In someone else it was caused by hypertension. If you had one of these underlying conditions, you would definitely not want to be treated for the other.
What hints do we have, then, about traits that set people on different pathways to addiction? One important indication comes, once again, from twin studies. When thousands of twin pairs were evaluated for ten of the most common psychiatric and addictive disorders, there was a clear pattern. Judging from their genetics, the disorders seemed to fall into two distinct groups, indicating that the conditions within each group belong together genetically. What that means is that people who have them are likely to share many gene variants that result in risk. But the disorders in each of the groups also seemed to share clinical characteristics. Scientists typically classify the psychiatric disorders within one of these groups as internalizing, and the other as externalizing.
13 These concepts are quite intuitive. In the internalizing group we find conditions in which people seem to turn inward, such as in fear or anxious misery. In this group we find the different types of anxiety disorders as well as major depression.
In the externalizing category are conditions where people instead turn their attention and actions outward. Addictive disorders for the most part fall in this category, together with antisocial personality disorder and conduct disorder. Before we proceed, there is an important caveat to interpreting this finding. Antisocial personality is a rather stigmatizing condition, characterized by a habitual pattern of lying, lawbreaking, and lacking in empathy with other people. Finding addictive disorders in the same group as this condition does not imply that the two are the same. What it means is that genes and traits that put individuals at risk for one of the disorders to some extent also contribute to the risk for developing the other. And among the most prominent traits shared by disorders within the externalizing group are various aspects of impulsivity. Having read in a prior chapter how rash actions, steep temporal discounting, and other aspects of impulsive behavior predispose for compulsive drug use in both people and animal experiments, this should hardly come as a surprise. In short, a major pathway through which genes are likely to put people at risk for addiction seems to be through various impulsive traits. It is easy to see how that might happen. Drugs have powerful reinforcing or rewarding effects in the short term; their adverse consequences are more distal in time. If, for genetic reasons, you are unable to properly gauge the value of short-term outcomes against their long-term cost, your risk of excessive drug use would only be expected to be higher.
Other characteristics are perhaps more surprising. Marc Schuckit at the University of California, San Diego,
14 found some of these by applying an ingenious approach. He recruited several hundred people in their twenties who themselves had not yet developed alcohol problems. Some of them were sons of alcoholics; others did not have any alcoholism in the family. Presumably the former group would carry a genetic susceptibility to developing addiction, while the latter would not. The idea was to carefully examine both groups early enough so that any differences between them would have a better chance of reflecting their preexisting susceptibility without being complicated by alcohol problems of their own. The researchers then planned to follow their subjects over years to see what characteristics observed prior to the onset of alcohol problems predicted the development of alcoholism.
In the end the research group engaged and examined over four hundred research participants. Among those who had alcoholism in the family, there was an interesting characteristic: they did not seem to be very sensitive to alcohol. In particular, they did not seem to be much influenced by the ability of alcohol to result in impaired balance, something that can be measured as increased body sway. Not every son of alcoholics had this characteristic, but it was present among almost half of them. The same was occasionally found in people without a family history as well, but in the latter group only between one in ten and one in twenty people showed the trait. The importance of these findings emerged over the following years. People with the “low response” to alcohol went on to develop alcoholism four times more often than those with a normal level of responding in 56 versus 14 percent of cases, respectively. A simple way of interpreting the data is that people normally have a brake on their alcohol intake. Alcohol activates brain reward systems, but it is not very potent, so quite a bit of intake is required for more pronounced effects. But at those levels of intake, most people also feel quite impaired, and so they respond with an aversion or simply become unable to continue their intake. If, because of genetic factors, this built-in brake is absent, then it makes sense that the risk of progressively increasing consumption and ultimately becoming addicted grows.
15There are other traits, too. For instance, people with social anxiety are more likely to develop all kinds of substance use disorders. A clinical diagnosis of social anxiety disorder is only modestly heritable, with genes being responsible for about 30 percent of the risk. But the personality trait central to developing the clinical conditions—fear of being negatively evaluated by others—has a heritability around 50 percent.
16 It is easy to see a pathway that leads to addiction for a person with this trait. For these people, attending every social function is torture; it feeds scary fantasies about what terrible thoughts others may have about them. Even just eating lunch in a diner is a nightmare. In the mind’s eye, everyone is watching, just waiting for you to spill ketchup all over your tie, something that makes your hands shake just enough to spill the ketchup all over your tie. If you repeatedly make the experience of these fears go away after taking alcohol as a pretreatment, then of course very powerful learning occurs. This is a classic case of establishing and escalating drug use for its negatively reinforcing properties. But as we have already learned, repeated drug taking will result in neuroadaptations in the brain pathways that control stress and anxiety reactions. This is therefore a path that leads straight to the dark side of addiction. In the absence of drug, the fears will get even worse, and the incentive to resume drug use will become progressively greater.
What we have described here is a very different pathway to addiction than that which starts as an impulsive, externalizing daredevil, walking the handrails of highway overpasses. The genetic risk factors are unlikely to overlap much. Yet the end result, by the time the patient seeks treatment, will satisfy diagnostic criteria for the same disorder, tempting the thought process of the treating physician to apply the same treatment. There are other traits, too, but you get the idea. We need to, as Danielle Dick at Virginia Commonwealth University likes to say, “deconstruct addiction” and find the different component traits that make up both the risk factors and the diagnosis itself.
17 I expect that understanding those component traits will make it easier to identify specific biological mechanisms that contribute to an addiction in the individual case and to target it with treatment appropriate for that particular individual. Many years from now, once we have understood the component traits and the biological mechanisms that drive them, maybe we will be able to reassemble diagnostic categories that make more sense and provide better help for choosing the right treatment. Perhaps those will be something like “impulsive alcoholism,” “opioid-reward dependent alcoholism,” or “socially anxious alcoholism.”
Although I wish this view were my invention, it is of course nothing of the kind. It is merely an application to addictive disorders of a very influential concept proposed in the early 1970s and then reiterated more recently by the behavioral geneticist Irving Gottesman of the University of Minnesota in the context of his research on schizophrenia. To put things in the language of a professional geneticist, a trait—such as having a diagnosis of hypertension or an addictive disorder—is called a phenotype, from the Greek words
phainein, “to show,” and
typos, “type.” As is often the case, the Greek roots tell us something important about the concept. Because showing is an inherent part of the word, it is clear that the trait should be directly visible or observable. But if things that under the surface are quite different can appear the same way to unaided observation, then we can hope that, beneath the surface, we might be able to find other traits, closer to the biological underpinnings of the disorder, and ultimately the genes that shape the biological risk factors. Gottesman describes these component traits, which he calls endophenotypes, like this:
Endophenotypes, measurable components unseen by the unaided eye along the pathway between disease and distal genotype, have emerged as an important concept in the study of complex neuropsychiatric diseases. An endophenotype may be neurophysiological, biochemical, endocrinological, neuroanatomical, cognitive, or neuropsychological (including configured self-report data) in nature. Endophenotypes represent simpler clues to genetic underpinnings than the disease syndrome itself, promoting the view that psychiatric diagnoses can be decomposed or deconstructed, which can result in more straightforward—and successful—genetic analysis.
18
Next we will see what the concept of endophenotypes has to offer for gene findings, and how it helps us understand some of the challenges that have so far prevented much of the promises of molecular genetics of addiction to materialize.
So what about the genes? Scientists are beginning to learn about some genes that contribute to addiction risk, too, and I am now almost ready to talk about them. But before I do, I still need to discuss how genes vary between people, introduce some terms with which that variation is described, and talk about how the influence of a gene variant for the risk of developing a disease can be determined. If this sounds like a crash course on aspects of modern molecular genetics, that is because it is. It is actually not possible to make sense of newspaper headlines such as “Alcoholism Gene Found”
19 without understanding these basics. Though I will only bring in the facts and terms that really matter, even those will be simplified, so I apologize ahead of time to my geneticist friends.
Remember, a gene is, loosely defined, a stretch of DNA string from which a protein is produced. Proteins are stitched together through the workings of cellular machinery by adding, one by one, individual building blocks made up of amino acids. There are twenty of those in nature. In contrast, the DNA code consists of only four letters—T, C, A, and G—and these are used to code what amino acid to add next to the growing protein string, using “words” of three letters each. As the astute reader will quickly realize, this means there are many more code words than there are amino acids. Indeed, as the 1968 Nobel laureates Nirenberg, Khorana, and Holley found, each amino acid has more than one triplet encoding it. This is important because it means that even when we find a mutation that changes the genetic code, that mutation may leave the meaning of the code unaltered. The sequence may still code for the same amino acid, in which case it is called “synonymous,” and may not matter at all. This is in contrast to “nonsynonymous,” meaning that the protein will actually contain a different amino acid in that position.
Preceding the DNA code that contains the blueprint for which amino acids to put together to make a protein, a gene typically starts with a sequence of code letters that contain “promoter elements.” These are places where components of cellular machinery bind and determine, not what the protein that is about to be produced will look like, but rather how much of it will be made. The DNA code for consecutive amino acids then follows but is contained in pieces, or “exons.” These are segments of code that are interspersed with “introns,” stretches that will not be translated into protein. To send a work order from the DNA blueprint to the protein-building factory of the cell, the DNA code is copied to a messenger RNA (mRNA). The immature form of the work order is a copy of both exons and introns, but before reaching the protein factory, it matures into the final protein-making template by “looping out” and removing the segments of code that correspond to introns, through a process called splicing. The mature mRNA finally has a tail that does not contain any protein making code but can allow various cellular signals to fine-tune how much protein will ultimately be made.
The point of this crash course is that the proteins are the business end of the genome. They fall into several categories. Some become structural elements of cells or what lies between cells. Others are enzymes that turn over fuels to energy or carry out other cellular functions. A third class make up a slew of signaling molecules, such as hormones or neurotransmitters. Although people vary in their DNA sequence, for this variation to have functional consequences, it has to influence the function of proteins, by changing either their amino acid composition or the amount in which they are made. There are less than thirty thousand genes in the human genome, but because of the opportunities for rearrangement of the building blocks that come from individual exons—alternative splicing—there are many more proteins. There are millions of places in the genome where two different people may have a different code, or be “polymorphic.” The simplest case is a replacement of a single DNA code letter for another one, say C → G, called a “single nucleotide polymorphism” (SNP). Other cases may be where a couple of letters have been lost or gained, called insertion-deletion polymorphism. When two people carry gene variants that contain different code letters, they are said to have different alleles of that gene, and the most common way people differ in their genetics is called allelic variation.
20 But remember, allelic variation will not matter much unless it influences the structure or amount of a protein. So what we are after in the end is
functional variation.
How do we go about trying to find gene variants, or alleles, that contribute risk for addictive disorders? Although this is a monumental task, the basic logic is deceptively simple. Let’s say we find an SNP where some people carry a C, while others have a G. Of course we all carry two copies of almost all genes—one from mom, the other from dad—so in reality people can be CC, CG, or GG at this position in the genome. Now let’s find one hundred people with a diagnosis of alcoholism and another hundred without it, determine their genotype, and count how many alleles read C in the respective group. Remember, since every individual carries two copies of the gene, there are two hundred alleles in each group. Let’s say 152/200, and thus 76 percent of these are C among alcoholics, while only half that, 76/200 or 38 percent, are C among the healthy controls. If one applies some relatively simple statistics to these numbers, it turns out that the probability of this happening by chance is less than 1/10,000. So as true scientists, we reject the null hypothesis that this was a random event. Instead we conclude that we have found an association between this particular gene variant and alcoholism. With only slightly more sophisticated math, we can also calculate what percentage of the risk for alcoholism can be attributed to this variant.
Simple, right? Not really. There are more sources of error for this kind of analysis than there is room in this book. Just to take one issue that is common and has plagued this field, there are obviously systematic differences in DNA sequence between people based on their ancestry. Otherwise our babies would not differ in visible ways! But if C happens to be more common among people of African descent than among Caucasians, and there happen to be more of the former than of the latter in our patient group, we would get a higher allele frequency and a statistical association no matter what disease we looked at. That association would have nothing to do with what we wanted to study, and everything to do with how we had selected our subjects. Many of the early gene findings were simply artifacts of this nature. There are increasingly good ways of correcting for that kind of error. But even once we establish a real, valid association, it does not tell us anything about how this variant contributes to the risk. In fact, because the genome contains so many variants, and because these tend to travel together in packs, glued to one another as chromosomes are transmitted from parents to offspring and occasionally rearranged through a process called recombination, we have no idea if the statistically associated variant is the one that contributes to the differences between patients and controls. It could just as well be another variant that does the job, one that is located some ways up- or downstream of the allele that we have studied within the DNA sequence, and that we did not even know about but that statistically travels together with our C, say, 90 percent of the time.
In the early days of the gene-finding field, scientists would literally count frequencies of one or a few markers in genes they suspected might have something to do with addiction risk, perhaps in regions of the genome identified through a different methodology, called linkage analysis. Several of the earliest gene associations were established using this kind of cumbersome approach, for instance, by the long-running, NIAAA-sponsored Collaborative on the Genetics of Alcoholism (COGA). As technology evolved, however, it became possible to simultaneously determine people’s DNA code at a million or even several million SNPs at the same time and carry out what are called whole genome association studies. Once we are able to apply that kind of brute force, surely we must be able to crack which gene variants contribute to addiction risk?
Not really. It is true that there are some findings. The best-established gene associations are with variants of genes from which the body makes enzymes, the little chemical factories that break down alcohol. After it has been ingested and done its job in the brain, alcohol gets broken down, first to the rather nasty chemical acetaldehyde, which normally is not present in the body, and only then to acetate, which the body can use for fuel. Although acetaldehyde makes people feel really sick, this is usually not much of a problem because the rate at which it in turn is converted to acetate is so fast that we don’t accumulate much of the toxic intermediary. But about 80 percent of Asians or their descendants have a gene variant that makes them generate acetaldehyde from alcohol about fifty to a hundred times faster than the rest of us. Because this is much faster than the body can convert aldehyde to acetate, the toxic metabolite will now accumulate. In about half these people, this is made even worse by another genetic variant that means that the next step in the reaction, the breakdown of acetaldehyde, is made unusually slow, combining to a particularly pronounced accumulation of the toxic metabolite. These people will react to alcohol intake with a flushing reaction, also called the Asian flush syndrome or Asian glow. Most of them will for that reason experience alcohol as being quite unpleasant. After all, this is the same effect that results from drinking after having taken Antabuse, the very purpose of which is to deter drinking by making alcohol intake intensely unpleasant. It is perhaps unsurprising, then, that people with the gene variants resulting in the flushing reaction have a consistently decreased risk of developing alcoholism. Note that in this case we don’t only have an “association”—we have a mechanism.
There is more. In 2004, COGA investigators found an association between alcoholism and a variant of the brain receptor for the neurotransmitter GABA. This is the receptor through which both alcohol and Valium act to produce their dampening effects, so it clearly seems to be relevant. Most important, the association is real. Other associations had been reported before but often failed to replicate in independent studies. The GABA-receptor finding, in contrast, has been independently replicated several times by now. Yet to this day, while we have an association, we don’t quite have a mechanism. The SNPs that show the association are located in introns, and it is so far unclear if they are functional.
21 An even more important advance came in 2007, when a couple of genome-wide association studies pointed to an association between nicotine addiction and a gene that codes for one of the proteins from which the nicotinic receptor is assembled. That association was then replicated by the Icelandic DECODE group, which uses the entire population of Iceland as a genetic laboratory.
22 The DECODE paper also elegantly showed that the guilty gene variant, as would be expected if it significantly contributes to people’s smoking, is also associated with lung cancer and cardiovascular disease. In this case there is a bit more of a mechanism. The SNP that causes the association does lead to a receptor that has altered function, and imaging studies have shown that it alters brain function, by modulating the strength of connection between two components of the brain reward circuitry.
23There are other advances, but this should convey the idea. Here is, however, the dilemma. We have previously established that 50–70 percent of addiction risk is caused by genes. Yet if we add up the risk of the most likely hits in whole genome association studies, we can account for, at best, a few percent of the disease risk. But if we are scanning the whole genome, and if genes contribute as much as we know they do, then most of the heritability is missing. The situation is similar for all complex behavioral disorders, as well as complex nonbehavioral conditions such as diabetes. This has been called the missing heritability or, more fancifully, the dark matter of the genome. It has resulted in quite a bit of disappointment after the initial fanfare to celebrate the completion of the Human Genome Project, which raised hopes that we would soon understand the genetic basis of most medical conditions.
Scientists have different takes on the issue. Many hard-core geneticists say that our sample sizes simply are still too small. If we didn’t find enough genes in studies that used 10,000 people, then we should use sample sizes of 100,000 instead. Studies of populations this size have in fact been published, often by pooling many different studies carried out in different countries. I am personally not convinced this approach is going to solve the problem we are facing. It is true that with increasing sample sizes, our chances to find associations between gene variants and disease do increase, but we will be finding associations that have smaller and smaller contributions. So I can’t see how that is going to address the problem of the missing heritability.
There is a complementary view that uses a different approach. The whole-genome association approach rests on the widely held assumption that common gene variants cause common disorders. For us to find an association, the same common gene variant, or allele, that contributes to risk needs to be present in many of the people who have the disorder. But what if many different, less common variants can all have similar consequences? What if any such unusual variant, not only anywhere in, say, the GABA-receptor gene, but anywhere in one of the many genes that are involved in GABA neurotransmission, can increase alcoholism risk? If that is the case, that is not something we will ever find in a whole-genome association study, at least not the way those are done now.
If many unusual, rather than a few common, gene variants contribute to addiction risk, then other approaches may be better suited to identify the pathways that are involved. Or at least this other approach offers a complementary strategy. To explain it, let me use an example provided by a recent paper from the laboratory of David Goldman at the NIAAA.
24 David’s group obtained DNA from one of the most extreme phenotypic groups you can imagine: Finnish men convicted for impulsive murder. The lab sequenced parts of each participant’s genome in search of mutations that are rare but have a dramatic effect: they disrupt a gene altogether, creating the human equivalent of a knockout mouse. The Goldman lab found that one such mutation can occur in a gene that codes for a particular serotonin receptor. When mice had the same gene knocked out, they also showed highly impulsive behavior. The human mutation was present only in Finns, so it can’t cause impulsivity in other populations. But it confirms that serotonergic function is critical for impulsivity, and it suggests that any mutation in this pathway might influence the risk for impulsive traits, and therefore also for addiction.
Although my own laboratory does a fair amount of genetics, all that work is in the context of pharmacology. I am not a real geneticist. I follow the advances described here with equal amounts of fascination and frustration. In many ways the field has made unbelievable progress. On the horizon is the day when each of us will be able to carry our whole genome sequence, obtained once and for all at a cost of less than $1,000, on a card in our wallets, or perhaps on a chip injected once and for all under the skin. Yet the search for individual gene variants that are mechanistically important for addictive disorders, in a way that would help me as a clinician treat patients, has for the most part been less successful than most of us hoped and expected even a decade ago.
For now it seems that a reasonable position is to be fascinated by genes but to let clinical work be guided by the much more mature knowledge of classical genetics. We know useful things, and we should make a habit of putting those to good use. If a person had a father and a grandfather who were alcoholics, if that person himself or herself appears to have low sensitivity to depressant alcohol effects and perhaps also shows impulsive traits, then it important for that individual to be counseled about the high genetic risk of alcoholism.
If you ask me, taking a good family history is probably more important at the moment than sequencing the patient’s genome.