risk

That's the final line in this story from the Globe and Mail about vitamin D and cancer risk.

The new findings that prompted the story haven't yet been published, and the story seems one-sided (no prominent skeptics are quoted), but it's a story that mixes traditional anthropological narratives about adaptation with new medical research, and it's full of punchy quotes. Here's the lede:

In June, U.S. researchers will announce the first direct link between cancer prevention and the sunshine vitamin. Their results are nothing short of astounding.

A four-year clinical trial involving 1,200 women found those taking the vitamin had about a 60-per-cent reduction in cancer incidence, compared with those who didn't take it, a drop so large — twice the impact on cancer attributed to smoking — it almost looks like a typographical error.

The story can potentially pull in several other kinds of information, including the lower cancer rates in developing countries, the different rates in U.S. racial groups, and the increase in cancer rates over time. Still, we've seen other correlations in the past that could pull these together, such as diet, stress, and substance use (smoking, alcohol) -- which are sometimes interrelated -- so we'll have to see how the result accounts for multiple correlations. It sounds like a simple treatment-control experiment, which should lead to a strong conclusion.

If vitamin D at the "normal" levels is actually a deficiency, then the obvious conclusion is that people should get out in the sun more. The article doesn't miss this:

Those studying the vitamin say the hide-from-sunlight advice has amounted to the health equivalent of a foolish poker trade. Anyone practising sun avoidance has traded the benefit of a reduced risk of skin cancer — which is easy to detect and treat and seldom fatal — for an increased risk of the scary, high-body-count cancers, such as breast, prostate and colon, that appear linked to vitamin D shortages.

The sun advice has been misguided information "of just breathtaking proportions," said John Cannell, head of the Vitamin D Council, a non-profit, California-based organization.

"Fifteen hundred Americans die every year from [skin cancers]. Fifteen hundred Americans die every day from the serious cancers."

The article includes lots of quotes from the vitamin D industry guy -- it almost sounds like a press release -- so we'll have to see if the results are really as provocative as suggested here. Last year there were other stories that pointed in this direction, so it seems not too unlikely. Vitamin D isn't just for rickets anymore.

From p. 1 of Risk, by John Adams:

The apprehension, determination and intense concentration that can be observed in the face of a toddler learning to toddle, the wails of frustration or pain if it goes wrong, and the beaming delight when it succeeds -- are all evidence that one is in the presence of a serious risk-management exercise.

Time has a short article describing the work of risk assessment expert John Adams.

The point, stresses Adams, is that drivers who feel safe may actually increase the risk that they pose to other drivers, bicyclists, pedestrians and their own passengers (while an average of 80% of drivers buckle up, only 68% of their rear-seat passengers do). And risk compensation is hardly confined to the act of driving a car. Think of a trapeze artist, suggests Adams, or a rock climber, motorcyclist or college kid on a hot date. Add some safety equipment to the equation -- a net, rope, helmet or a condom respectively -- and the person may try maneuvers that he or she would otherwise consider foolish. In the case of seat belts, instead of a simple, straightforward reduction in deaths, the end result is actually a more complicated redistribution of risk and fatalities. For the sake of argument, offers Adams, imagine how it might affect the behavior of drivers if a sharp stake were mounted in the middle of the steering wheel? Or if the bumper were packed with explosives. Perverse, yes, but it certainly provides a vivid example of how a perception of risk could modify behavior.

Adams makes two points:

(1) People who will tolerate a given level of risk will respond to an objective reduction in risk by doing riskier things. (Elsewhere, Adams calls this idea the "risk thermostat" model.)

(2) Risk is interactive and unequally distributed, so that decreasing the risks for one category of individuals may increase it for others.

Car accidents make pretty good illustrations for both. Some people who drive large SUVs feel a greater safety margin and therefore drive more aggressively. Providing these people with a "safer" driving platform tends to increase the risks for drivers of smaller cars. If these effects were strong enough, putting a "safer" class of vehicle on the road would actually increase the overall number of fatalities.

A web search for Adams' work brings up a technical report written for the Cato Institute, discussing risk perception in more detail. After a review of the effect of seat belt laws, he embarks on a discussion of risks identified through science. These kinds of risks are not directly accessible to the senses, and their magnitude can be appreciated only by studying large populations of things. The long-term risk of smoking is one example.

Before discussing these kinds of risks, Adams considers our attitudes toward risks that are accessible to the senses:

Directly perceptible risks are "managed" instinctively; our ability to cope with them has been built into us by evolution--contemplation of animal behavior suggests that it has evolved in nonhuman species as well. Our method of coping is also intuitive; we do not do a formal probabilistic risk assessment before we cross the street. There is now abundant evidence, particularly with respect to directly perceived risks on the road, that risk compensation accompanies the introduction of safety measures that do not reduce people's propensity to take risks. Statistics for death by accident and violence, perhaps the best available aggregate indicator of the way in which societies cope with directly perceived risk, display a stubborn resistance, over many decades, to the efforts of safety regulators to reduce them (Adams 1999:10).

Adams notes that much of the decrease in premature mortality during the past 150 years has been brought about by better understanding and communicating about invisible risks. The germ theory of disease is probably the most notable example.

But he points out the difficulty of measuring reductions in risk. At least, we can measure overall mortality rates -- if they decline after an intervention, then presumably it was effective. But activity-specific mortality rates won't do:

Moreover, risks can be displaced. If motorcycling were to be banned in Britain it would save about 500 lives a year. Or would it? If it could be assumed that all the banned motorcyclists would sit at home drinking tea, one could simply subtract motorcycle accident fatalities from the total annual road accident death toll. But at least some frustrated motorcyclists would buy old clunkers and try to drive them in a way that pumped as much adrenaline as their motorcycling did, and in a way likely to produce more kinetic energy to be dispersed if they crashed. The alternative risk-taking activities that they might pursue range from skydiving to glue sniffing, and there is no set of statistics that could prove that the country had been made safer, or more dangerous, by the ban (Adams 1999:20).

Now, I'm reading this because I'm evaluating strategies toward risk in human evolution. Mortality reductions are a major trend in the emergence of modern humans. That would tend to indicate an objective decrease in risks of various kinds.

But a decrease in mortality risk may not translate to an increase in fitness. For instance, if more adult males survive to older ages, they may prevent younger males from reproducing until they are older. If this happened, a reduction in mortality would impose a tradeoff of a reduction in fecundity for younger individuals. This tradeoff would not prevent the change, by any means -- in the transient after the appearance of a risk-reducing strategy, males who adopted the strategy would immediately have a fitness benefit. But the tradeoff itself might obscure the reasons for the change, or even suggest wrong hypotheses (for instance, the hypothesis that low fecundity for young males forced them to reduce their mortality risk).

Anyway, if professional statisticians are so bad at evaluating the risk landscape, evolutionary biologists are no better. Many evolutionary hypotheses deal explicitly in risks -- with increases in some risks being explained by declines in others. But if you have ever seen an attempt to quantify those risks in terms of fitness, you probably understand how shaky the foundations of such hypotheses can be.

This is the beginning of a multipart series on evolution and risk. There will be some math involved -- calculus, even! -- but at the end something very important will emerge. Risk is the hinge connecting the evolution of human life histories to the evolution of the human brain.

I'm reading a bit about risk in large animal hunting, and I ran across an article by Dereck Joubert on elephant hunting by lions in Botswana.

Over the 4 years, we observed a total of 74 elephants killed by lions, including eleven elephants in 1993, seventeen in 1994, nineteen in 1995, and 27 in 1996, suggesting an increasing hunting success rate. All the elephants killed, with one exception, were from breeding herds (females and young). The exception was an adult bull, previously wounded by another bull, who remained alive for several days before eventually being killed by the lions. The great majority of the young elephants killed were males, and two-thirds of the kills were of elephants in the age range 4-15 years, with highest hunting success achieved for elephants aged 4-9 years (Table 1). The animals killed were commonly on the periphery of, or straggling behind, the breeding herds, with nearly half killed more than 50 m away from the main herd. Hunts were less commonly attempted on calves which were under the age of 4 years, which remained more closely associated with their mothers. Hunting success for elephants older than 4 years apparently doubled from 33% (n = 9) in 1993 to 62% (n = 61) in 1996. Many attempts to kill adults bulls were made in
1996, when we saw lions attacking elephant bulls almost nightly although only one hunt was successful. All except one of the kills were made at night, and hunts occurred more commonly on dark moon nights than when the moon was bright.

Well, hunting elephants ought to be pretty risky (otherwise, lions would do it all the time, right?). So how many lions got hurt during all these hunts?

There was a close resemblance between the methods that the lions used to hunt elephants and the technique commonly used to hunt buffalo. This tactic included first opportunistically detecting a straggler, or targeting a vulnerable member of the herd, then circling behind the selected prey. The lions then attacked by running in as a group. One or more lions leapt up onto the back or lower flanks and orientated along the spine of the prey. They then bit down on the backbone. The lion positioned highest up the spine would still be behind the ears of the elephant and just far enough back to be out of reach of the extended trunk. The elephant was then pulled down to its knees, not collapsed because of any fatal bite to the spine. Another approach involved a running hunt causing confusion and bunching of the elephant herds. This often resulted in one elephant falling or getting separated. In all cases a rear attack was employed, never a frontal attack. In one notable case, a single male lion ran at nearly full speed into the side of a 6-year-old male calf with sufficient force to collapse the elephant on its side. On only one occasion was a lion injured by an elephant in these hunts. In that case, the elephant collapsed on top of the lion. The resulting injury to the head was therefore recorded as accidental rather than as a result of a counterattack by the elephant.

OK, so the lions mostly limited their hunts to a class of most vulnerable elephants (subadults old enough to be isolated from their mothers, and inattentive to predators -- males amounted to 236 confirmed attempts versus 38 for females!). They adopted a special hunting style that they use for other dangerous large prey animals, attacking from the back by ambush. And during all these hunts (which totaled 74 kills out of 323 attempts) only one lion was confirmed injured. The paper doesn't say how serious the injury was, or if it
was eventually fatal, but elephant-falling-on-lion can't be a good situation.

Now, the relevant measure of risk in this instance is the injury rate (or even better, death rate) per successful kill. Unsuccessful attempts might fail for many reasons, including injury, but none of these unsuccessful attempts satisfy anybody's energetic requirements. So we have one serious injury per 74 kills. There may have been other injuries that weren't major enough to be observed or counted. Limiting to the one that was counted, we have a rate of serious injury of around 1.33 percent per kill; divided among the average number of lions that participated, which isn't specified.

From the elephant perspective, there appears to be a case for strategic indifference of adult males to predation on the younger males:

When these young elephants finished and called out to their families, the lions attacked. There was surprisingly little response from other nearby elephants. Older calves were attacked and killed within 50 m of the drinking bulls. The distress calls of the young elephant and lion growls seldom distracted them from drinking.

Tough to be a young male elephant.

References:

Joubert D. 2006. Hunting behaviour of lions (Panthera leo) on elephants (Loxodonta africana) in the Chobe National Park, Botswana. Afr J Ecol 44:279-281. DOI

The NYTimes has been very helpful for human geneticists lately, at least when it comes to providing good articles for class discussions. The latest is this article by Amy Harmon, titled "Couples cull embryos to halt heritage of cancer."

The article gives a good overview of preimplantation genetic diagnosis of embryos -- a technique that takes a single cell from an early-stage embryo in order to probe its DNA for undesirable alleles. The current twist, and really the hanger for the story, is that people who carry risk alleles for cancers and other late-life conditions are using the technique to avoid bearing children who carry their genetic "curse." PGD has been used for many years now to screen embryos that would inherit conditions like Tay-Sachs or cystic fibrosis. But testing for genes like BRCA1 may be perceived very differently, since the associated risks occur relatively late in life, and they are treatable (although certainly not always so). In other words, it is a different balance of risk versus fear.

Knowing that Mr. Kingsbury had tested positive for the colon cancer mutation, the Kingsburys started with the basic laws of genetics: because children randomly inherit half of each parent's genes, he had a 50 percent chance of passing it on. Since the mutation raises the risk of developing the cancer by about twentyfold, that means any child of theirs conceived the traditional way would have about a one in three chance of getting it, usually around age 45. Those who did develop the cancer would also have a nearly 90 percent chance of surviving it, but only if it was caught early.

The article talks about the procedure as becoming increasingly common, which is certainly true. But as I tell my students, there are pretty clear logical limits. Screening for one Mendelian gene is certainly possible, and has a reasonable chance of obtaining three or four "desirable" embryos if the couple starts with ten or twelve fertilized zygotes. But screen for two Mendelian genes at once, and you square the necessary number of fertilized eggs. The supply of ova is very limited, so picking offspring characteristics with PGD is not going to happen at any large scale.

Which, I would imagine, will increase the pressure for reproductive cloning. A father who carries a dominant cancer-risk gene can avoid passing it on entirely, if the couple decides to clone the mother instead.

Or maybe they will move on to deliberate chimerization. The father's colon cancer gene would be no problem, if all the progenitor cells for the reproductive tract were maternal clones. The rest of the child's body could be the normal fertilized child of both parents. Or a paternal clone.

Thanks to an enterprising student, I have an AP story about the discovery of a risk allele for prostate cancer that has different frequencies in different groups.

Scientists have identified a common genetic marker that signals a 60 percent heightened risk of prostate cancer in men who carry it, and it may help explain why black men are unusually prone to the disease, a new study says.

The DNA variant may play a role in about 8 percent of prostate cancers in men of European extraction and 16 percent of the cancers in blacks, researchers said.

So, what does somebody like me do with this information? I'm interested in cases where populations are different as illustrations of human biological variation and its evolutionary causes. Here, we have medical data and genomic information that appears to show a geographic difference among men in their risk. But we have to look at the study to see what that means. Here's the abstract:

With the increasing incidence of prostate cancer, identifying common genetic variants that confer risk of the disease is important. Here we report such a variant on chromosome 8q24, a region initially identified through a study of Icelandic families. Allele -8 of the microsatellite DG8S737 was associated with prostate cancer in three case-control series of European ancestry from Iceland, Sweden and the US. The estimated odds ratio (OR) of the allele is 1.62 (P = 2.7 x 10^-11). About 19% of affected men and 13% of the general population carry at least one copy, yielding a population attributable risk (PAR) of ˜ 8%. The association was also replicated in an African American case-control group with a similar OR, in which 41% of affected individuals and 30% of the population are carriers. This leads to a greater estimated PAR (16%) that may contribute to higher incidence of prostate cancer in African American men than in men of European ancestry.

This is actually a great case study in the complexity of finding the genetic causes of disease:

As only the microsatellite allele showed significant association in the African American case-control group, and it is contained in a smaller LD block in African Americans than in populations of European ancestry (Supplementary Fig. 2 online), we propose that the region most likely to contain the functional variant can be narrowed down to positions 128.414-128.474 Mb (NCBI build 34). This region contains one spliced EST (AW183883) and three single-exon ESTs (BE144297, CV364590 and AF119310) in addition to a few predicted genes, but no known genes. No microRNAs have been detected within the block. Expression analysis in various cDNA libraries confirmed only the expression of the AW183883 EST (Supplementary Methods). We identified four different splice variants of AW183883 by 5'and 3' RACE that were verified by RT-PCR and RNA blot analysis (Fig. 2a). Using the AW183883 EST as a probe on an RNA blot, we detected a 1.5-kb signal only in testis, consistent with the size of the two longer forms. The two shorter transcripts harboring exons 6-8 were detected only in normal (0.6-kb transcript) and malignant (0.6- and 0.9-kb transcripts) prostate cell lines, not in the other tissues analyzed (Fig. 2b,c). The predicted ORFs for these transcripts did not show significant homology to known proteins.

In other words, although they have narrowed down a region around this marker as a risk factor for prostate cancer, they don't know what specific genetic change causes the risk. It isn't any of the SNPs in the HapMap set. They can tell from the HapMap variation that the linkage block is narrower in Africans, so that the causative variant must be fairly near it. There are no characterized genes and no known microRNAs in the region, but there are some possible genes that haven't been characterized. "cDNA" libraries are complements of mRNA expressed within cell lines, only one of these may correspond to the region in question, and one variant is expressed in testis and prostate cell lines. They don't know what it is or what it does.

All of that analysis is bioinformatics -- taking databases of known gene expression in cell lines, genetic variation among people, and identified genes, SNPs and microsatellite markers and integrating them into a picture of associations. What is left is a lot of biology -- what does the key genetic variant change; what gene is it part of; what is the normal role of the gene within prostate (or elsewhere); why does it become pathological; etc. And most important, how can we fix it?

My interest is at the opposite end --- why did this allele become common; why does it have different frequencies in different groups; how old is it?

There are no answers to these questions -- there rarely are in studies like this. But put several of them together and we may start to uncover much about human prehistory and the conditions of life in ancient populuations. Does this risk allele correspond to different ancient environments? Different population dynamics? Or is it just different by chance?

That is the breakdown on the one hand between bioinformatics and biology, and on the other between functional biology and evolutionary biology.

References:

Amundadottir LT and 45 others. 2006. A common variant associated with prostate cancer in European and African populations. Nature Genet (online early) DOI link

According to Nature News, Brian Knutson and Camelia Kuhnen of Stanford have discovered that the interaction between two brain regions is involved in determining whether people take risks:

As centres for pleasure and anxiety battle it out, a simple brain scan of the two can actually predict what a person will chose to do a few seconds before they do it: when joy beats worry in our brain, a risky decision is made.

Studies of how the mind handles risky behaviour have highlighted a number of neural hotspots. One is a peanut-sized region of the brain called the nucleus accumbens, which is loaded with the molecule dopamine and becomes active in anticipation of pleasure. The nucleus accumbens is known to play a role in the addictive affect of drugs.

Another region, known as the anterior insula, is stimulated in anticipation of a bad sensation. This area lights up in those predicting the onset of physical pain, and in generally anxious individuals.

This doesn't complete the causal chain behind such decision-making. Indeed, the relative activity of these two regions may be better understood as a correlate of decision-making. Whether these dualing motivations are resolved in one way or another depends on other, earlier links in the chain.

It would be interesting to know if normally anxious people had a different threshold for action than normally risk-taking people, or whether instead, one or the other of the brain areas were simply more active in one than in the other.

The question for this week: what will my anterior insula let me get away with writing about...