Diversity and disease

Most physicians are aware that some diseases are more prevalent in certain ethnic groups. For example, prostate cancer is more prevalent in African-American men, and diabetes is more common among Native Americans, compared to other groups in the United States. Physicians will, therefore, often use some estimate of ancestry to assess disease risk. These are often crude ancestry estimators, based on things like skin color or last name. This brings up an important and complex issue; given that ancestry can be informative, and that racial categories are not that biologically meaningful, how can physicians most effectively use ancestry in the clinic?

The social element of ethnicity really complicates the issue. Ethnic groups often differ from one another in cultural and socioeconomic status, in addition to differing in where they ancestors originated. Many health differences between ethnic groups can probably be largely explained by social, cultural, and economic factors. In the US, this effect is made obvious  by the many diseases that are more common in minority ethnic groups, which often suffer at the lower end of the socioeconomic spectrum. This trend is likely explained by socioeconomic factors, such as limited access to quality health care, higher stress, and differences in diet. Removing socioeconomic differences between groups would go a long way towards removing differences in health care outcomes. This is, primarily, a social science, and, more problematically, a political issue.

While social factors are probably the larger issue for many conditions, genetic factors may also make an important contribution to differences in health-related traits between ethnic groups. Genetics may be especially important in certain health issues, like how people respond to drugs. Pharmaceutical companies have recognized this, and a drug tailored specifically to African-Americans, called BiDil, was approved by the FDA back in 2005. The drug  was originally intended for use by the general population. When it failed to be approved by the FDA in 1997, the drug company noticed that it worked much better in among self-identified African-Americans in the drug trial. So they indicated that it was for African-Americans, the FDA approved, and it became the first drug intended for a specific ‘race’.

So what’s complicated about this? Why don’t drug companies tailor drugs to specific ethnic groups all the time?

Well, it’s a bit trickier than that. What is defined as ‘race’, based on a people’s physical features, is not really a biologically meaningful way of dividing people up. But it’s a subtle issue, because a very small number of biological traits do correlate quite closely with self-identified ‘race’ (such as skin color), some traits correlate loosely with ‘race’, and a large number are completely uncorrelated.

Geneticists are quick to point out, justifiably I might add, that things like skin color are generally not meaningful predictors of other biological traits. Skin, eye and hair color, along with other physical features associated with ‘race’ only reflect variation in a few genes. Differences in these genes between people in different parts of the world probably reflects adaptation to local environments. For example, it has long been thought that UV exposure levels have driven differences in skin color between populations, where darker skin was favorable in geographic regions with lots of UV radiation (as a protection against skin cancer), while lighter skin was favorable in regions with less intense sunlight (to allow higher levels of vitamin D synthesis).

It is important to note that many other traits, including disease risk, could be completely uncorrelated with these adaptive traits. Even if traits were correlated to physical features in ancestral populations, the random assortment of genetic variants due to inter-marriage between groups has likely reduced this correlation in recently admixed populations, which includes  almost all groups currently living in North and South America.

The eventual goal of biomedical research, in my opinion, is to be able to perfectly explain the origin of a disease in any patient on the molecular level. This will include identifying genetic variants that can predict health outcomes, like how a patient will respond to a drug. Of course, if we knew the genetic variants that predicted BiDil response, there would be no need to use ‘race’ in deciding to whom one should prescribe it. But personalized medicine is in it’s infancy, so we don’t have these genetic tools yet. In the meantime, it may be necessary to use crude genetic proxies, like ethnicity or family history, to better inform treatment strategies. Still, I think the widespread use of ‘race’ as biological marker should be accompanied by an understanding that it is really not a meaningful proxy for the vast majority of biological traits. This will help prevent the medical profession from putting too much stock in an imprecise diagnostic tool. It has already been noted that BiDil is ineffective in some African-Americans, and effective in some European-Americans. Importantly, taking racial differences in medical traits with a grain will also help prevent an unwitting scientific validation of racist ideologies about inherit, and derogatory, differences between groups.

Next time you find that a mouse has invaded your house, perhaps you should take a minute to say thank you before you ‘dispose’ of it. Why, you ask? Well, it turns out that mice may be a vital part of developing the new field of personalized medicine, where genetic tests can help doctors tailor treatment to your unique biological make-up.

A study came out this week in the journal PLoS Genetics reporting a set of common mutations whose presence or absence can help predict an individual’s systolic and diastolic blood pressure. They used a very interesting approach, starting with a mouse model and working back to humans.

They took a large panel of inbred mouse lines, where every mouse in each line is genetically identical.  They then looked for genetic differences between lines that correlated with differences in blood pressure (imagine mice with tiny, little sphygmomanometers wrapped around their upper arms). This analysis revealed a section of mouse chromosome 9 that  showed strong evidence that variation in this region was associated with blood pressure. For technical reasons, the resolution of these sorts of analyses is not great. This is due to ‘linkage’ between multiple, physically clustered, positions in the genome that are inherited together as a chunk of DNA from one parent. This causes mutations close to each other in the genome to be often correlated. A variable position in the genome is due to a mutation of the original sequence. Some individuals have mutations and some individuals have the original DNA sequence. For two linked variable positions in genome, individuals with a mutation at one position are likely to also have the mutated sequence at other. This causes entire chunks of the genome to be correlated with traits, making it difficult to figure out which of the mutations in a region is actually relevant.

Human and mice are very similar in many ways, which allows work in mice to be applicable to human disease. The two species share a common ancestor, but that was almost 100 million years ago, and all genetic variation shared between the two lineages was lost long ago. Consequently, you cannot find genetic variants in mice that are also in humans. What you can find are regions of the genome where mutations in mice affect a trait, and, therefore, where different mutations in humans may also affect the trait. This second goal requires two steps, however, first you have to find regions in the mouse genome, and then you see if there are mutations in humans that also correlate with this trait. The authors did just that. They took the part of the human genome that corresponds to the chromosome 9 region they found in mice, and they searched for human mutations correlated with blood pressure. Using this method, they successfully identified a set of mutations, common in humans, in this region that predicted an increase in systolic pressure (1.3 to 1.5 mmHg) and diastolic pressure  (0.8 mmHg).

So you may ask, why not just start with humans and skip the mouse step? After all, you have to test everything you find in mice in humans as well. The answer is power. Statistical power to be specific. Mouse experiments are much more controlled, so there is much less noise in the measurements compared to human studies. The noise in human studies can obscure the signal you are looking for, making it difficult to know if negative results are because your hypothesis is wrong or just because you had too much noise.

Imagine that finding a mutation in humans is like trying to get to your grandma’s house in suburban Phoenix from your house in Seattle. Using a mouse model is like taking an airplane. You can cover large distances with a tremendous amount of power and speed. On a local scale, however, a jet becomes too large and cumbersome to allow you to travel to specific places within a smaller region. For that you need a car. You could drive all the way down from Seattle, but it’s more efficient from a time perspective to use both methods in combination.

I have become very interested recently in the role hormones play in shaping individuals’ unique medical needs. They are so important for so many different aspects of human heath, and different hormones work together as a complex system to maintain homeostasis. Even more interesting, the dynamics of these systems may differ from person to person, and may explain differences in disease prevalence in different parts of the world.

Recently I’ve begun to think about a lot about testosterone in particular. Everybody knows about testosterone. It’s often thought of as the hormone that gives men their manliness, so to speak, although it also has important functions in women’s health, and it helps regulate a variety of things like libido and muscle growth. Testosterone is also famous as the steroids that baseball players and pro wrestlers take to enhance their athletic abilities.

What you may not know, however, is that the amounts of testosterone produced naturally by the body varies quite a bit from man to man. Well, you may have already assumed that was true, perhaps based on traumatic memories of schoolyard bullies, but it is important to remember there is more to testosterone than aggression and muscles. In fact, recent evidence suggests that  higher levels of testosterone do not necessarily correspond to increased aggression. Instead, there is some evidence that psychological disturbances are associated with extremely low levels of testosterone, for example in men treated for testicular cancer or in men with mutations in proteins that mediate testosterone response.

While variation in testosterone levels may not explain differences in behavior, it may help explain why some men are more likely to develop conditions like type 2 diabetes and cardiovascular disease. Multiple epidemiological studies, such as a recent study in men over 70 from the European Journal of Endocrinology (Yeap et al.), have found that (free) testosterone levels are correlated with risk factors of metabolic syndrome like insulin resistance. Specifically, lower levels of testosterone are associated with higher risk of metabolic syndrome and corresponding conditions like diabetes and heart disease. Variation in testosterone levels have been shown to be heritable, based on twin studies, and some common genetic variants have been shown to predict testosterone levels. Correlation does not mean causation, and the so the exact relationship between genetic variation, testosterone and metabolic disease still needs to be investigated. The necessary sorts of experiments to investigate this relationship may be readily available currently. For example, testosterone acts on the cellular level by turning on a protein that regulates gene expression. Techniques exist that could allow us to look for genetic variants that disrupt this activity, and these variants are likely to be involved in metabolic syndrome. Finding variants that cause increased metabolic syndrome risk could shed light on new drugs, and also allow physicians to identify people at high risk. The current system for identifying high-risk people involves asking about disease incidence in known relatives, which is a very crude estimate of genetic risk compared to what we are capable of with modern human genetics.

Uncovering the genetics of metabolic syndrome risk could also help reveal the history of this condition, and, perhaps, help us understand why it is so common and why the prevalence varies throughout the world. More about this in a later post…