Epidemiology, biomedical 'causation', and my widow-maker: how to make sense of it, and what terms to use

As I've posted previously, I have been recovering from multiple bypass heart surgery.  I had some angina, vague superficial chest pain that is a symptom of impending heart attack (or, more properly) coronary artery blockage.  Fortunately, I am educated enough to recognize that this chest pain wasn't from my doing a new kind of exercise, I went to the doc and--to make a long story short--was sent right off to the hospital for heart bypass surgery (replacing clogged coronary arteries).

The radiography showed that at least some of my heart arteries were clogged--with whatever radio-opaque goop, presumably including cholesterol, and by whatever clogging mechanism.  These causal facts are, as I understand things, complex and not completely understood, but the upshot was clear: surgery.....or else!

Now, the doctors would say that, given this evidence, I was at high risk of potentially lethal heart disease.  I'm sure had the opportunity been there (and it may be in some future doctor's appointment), I will be chided--or scolded--for my bad diet, too much cholesterol, etc.  It will be assumed that my voluntary lifestyle choices caused my blockage and my need for preventive artery replacement.  Bad boy!  Bad diet! Tsk, tsk, tsk....

But is that right, or might it be the opposite of a more serious truth?

What is bad behavior, health-wise?
I am 77.  This is beyond the usual 76-ish life expectancy for US males (searching the ad-laden web to find data has become mainly a challenge to wade through the relentless commercialism).  So, my lifestyle cannot be viewed as bad behavior in this respect.  Indeed, I have already lived longer than half my birth cohort!  So perhaps my diet and whatever else can, or should be viewed as having been protective.  After all, I was symptom-free until after my expected lifespan.

It is very difficult to understand what 'risk' means in such regards.  If my lifestyle led to my artery becoming clogged up, but it didn't happen until after I'd out-lived my average peer, can I legitimately think of that lifestyle as having been protective rather than risky?  We all have to get some final disorder and some point, so is the absolute cause the relevant fact, or is it the relative?  How can we decide such questions, if indeed they are meaningful ones that can even have meaningful answers?

If my behavior (for whatever reason, including just plain luck) led to my surviving in very good health, except for one weakest-link, then does that link suggest I've behaved badly, or does my overall great state of health suggest the opposite?  More to the point, how can such questions even be answered in a meaningful sense?  They seem meaningful....until you think a bit more carefully about them. . . . .

The philosophical quick-sand doesn't stop there.  If my arterial clog would have led to a relatively quick death--not a 'premature' death at my age!--but saved me from some worse, more prolonged or debilitating fate, can we seriously view that as preventive or protective, with me now facing those dreadful fates?

When we have competing causes and inevitable mortality, we have to view the causes, and what causes them, in a rather different light.  That doesn't mean there are consensus, much less easy, answers.  But it may mean that rules for 'healthy' behavior are not so obvious as they seem to be.

Red meat makes a good, scary cancer story....but is it?

It's off again, on again:  don't eat processed meat, don't eat red meat, or you'll get colon cancer!! Eat fish (well, unless it has mercury) or chicken (unless it has salmonella), or 'the other white meat': pork (remember the billboards?). They're safe!

A few years ago we seemed to have been given some relief when stories suggested that red meat (beef) was OK after all (of course, the lives of the cows were awful, and eating beef meant you doped up on antibiotics, but at least it didn't give you colon cancer).

Recently, a statement (now apparently offline) released by the International Agency for Cancer Research, a part of the World Health Organization, asserts that eating processed meat and red meat, 'causes' cancer.  Actually, the report was a bit more nuanced than the headlines, but journalists have to make a living, no?

Bacon, Stock photo

In response to strong backlash, the WHO quickly was forced to 'clarify' their clarion call to vegetarianism -- here's a link to their Q&A on the subject.  They now acknowledge, or 'clarify' that what they had done was simply add the meats to a list of known nasties, that cause cancer.  Putting meat on a causal list is one thing, but dishing it out to the media is another, and a rather irresponsible way to play for publicity (of course, if the news media made an exception and actually did their job of being skeptics, this wouldn't have unwound as it did).

In any case, the bottom line was basically that even two strips of bacon a day increases your colon cancer rate by 18%.  That sounds like a whopping and terrifying difference!  The WHO put this in the same carcinogenic-substance category as asbestos and tobacco. As they quickly clarified, that is in a sense a warning list, but the 18% figure is what got in the news and may have, at least temporarily, slammed the bacon and hamburger industry, if anybody still listens to the daily Big Warnings. However, let's hold all cynicism for the moment, hard as that is to do, and look a bit more closely at was said.

First, there seems little doubt that processed meats 'cause' cancer.  That doesn't mean an innocent-looking strip of bacon will give you cancer.  Instead, what it means is that various high quality studies have found a dose-response pattern in which higher or longer exposure levels earlier in life are associated with higher cancer incidence later on.  We know that correlation is not the same as causation, and that lifestyle factors are highly correlated.  Thus, for example, those in dire poverty don't eat tons of processed meat, and those who eat less salami also eat more brussels sprouts, take vitamins, don't smoke, lay off the double gin tonics.....and of course, go to the Ashram regularly to get your mind off the bacon you didn't eat at breakfast and the aftertaste of your dinner's brussels sprouts, and say a mantra to stay calm after you've given up everything that's fun.

Now, in the west, the lifetime risk of colon cancer is about 5%.  That means that if you tote up the probability of having cancer at age 40, 45, 50, .... 100, if the 18% figure is credible, it means that risk is about that much higher in those who dose up on pastrami and burgers.  Actually, this was the estimate based on eating 2-strips of bacon or the equivalent every day.  Of course, by far most of these cancers occur in older people (over the age of 60, say).  That means that the risk figures mainly apply to you if you live to old age, and of those who die earlier of other things their actual risk turned out to be zero--they enjoyed their visits to McD's and the deli!  That's why smoking is, in a literal epidemiological sense, a preventive relative to colon cancer (smoking will kill you of something else first).  There's no joking about cancer, but the basic idea is that for those who lived long enough, about 5% get colon cancer at some age.  Actually, while we don't know about meat-eating habits, but risks have been declining in recent years in developing countries (and, I think, increasing in other countries as they westernize).

Eat meat and lower your risk!
At a baseline of 5%, an 18% increase means a lifetime risk of about 6%.  Now if you hog up even more, your risk will go higher, perhaps much higher.  But wait a minute.  How many people actually dish up so heavily on processed meat (including steak and burger)?  Surely some do.  In fact, we don't know exactly where the lifetime risk estimate of 5% comes from; if from a population sample, then it wouldn't have regressed out meat-eating, and the figure would already include meat-eaters. However, let's ignore all these potential confounding or confusing issues and just consider the 18% figure on its own, as a given, as risk differences between abstainers and sausage gluttons.

Now in modern countries with health care systems, one routine health-care procedure is regular colonoscopy in older adults.  There was a recent estimate that regular colonoscopy can prevent about 53% of colon cancers; the reason is that precancerous polyps are found and excised so they can't transform into cancer.  Actually, you can find even more dramatic estimates of the preventive effectiveness if you scan at the web.  Likewise, you'll find many other lifestyle factors widely cited as having protective effects, including exercise, vitamins, eating vegetables, and the like.

Let's just do a bit of back-of-the-envelope numerology to make the point that if you're a bacon hog but have regular screening, get your exercise and all that, and you reduce your meat-elevated risk by 50%, then your net risk is around 3%, about half the 'average' of 5%.  One can surmise that if you stop your bacon fix, but then figure you're fine and don't do the other preventives, many of which are likely to be wanting in the meat-hog's normal lifestyle, then the actual effect of your 'healthy' baconless diet change will be to increase your cancer risk!

This is a lesson in complex causation and oversimplified news stories.  Processed meat may be a risk factor for colon cancer, but throwing irresponsibly simplified figures like raw meat to the news media leads to worse, rather than better information for the public.

So, as Hippocrates said, moderation in all things.  Eat your reuben (OK, yes, along with some broccoli).  But go one better than Hippocrates:  get scoped!

Hot under the (epidemiological) collar

Blogs like this are venues for expressing views on the current scene, in our case, related to genetics, evolution and a few other things we throw in.  If you express a view, unless it's just plain vanilla, you will irritate some readers.  In a sense, if you don't then there's no point in writing the blogpost.  In this case, we heavily criticized the recent NYTimes article reporting that the government has now backed off its claim that dietary cholesterol is a heart disease risk factor.  We try to be responsible, but that doesn't mean we have to expect agreement or to mince words!

We argued that the kind of herky-jerky yes/no results from huge long-term megastudies are so common that this shows the studies are rather useless and we think should be phased out, the results to date archived for anyone who wants to mine them, and the funds put to something that actually generates more trustworthy and stable results (if risks are stable enough to be estimated in these ways).

Well, this generated a very heated message from an old friend, a prominent genetic epidemiologist, who said that if we were listened to, it would lead to throwing the baby out with the bathwater.  He was upset because he said it was not the data but the analysis of these big epidemiological (environmental or genetic) studies that was at fault.  The studies are based essentially on correlation or regression models that assume everyone starts out as an equal blank slate, and whose individual risk is the basically additive total of the various risk factor exposures.  Your sex gives you a 'dose' of risk, which age adds to, then smoking history, diet, and so on.  Once all your risk-factor measures are toted up, your net risk can be estimated.  It is this general approach that regardless of sophisticated details in the statistical method, is not good at finding what we really should be looking for.

Our friend's idea is that, for example, dietary cholesterol may on average not be harmful, but there are likely some subsets of the population for which it is a risk.  Standard models may be convenient to apply, and everyone knows how to do that....but they miss the boat.  The key problem is basically that risk factors interact and searching for complex interaction is rarely done because it is very demanding in terms of sample size, sample structure, and analytic tools.  But there is no reason, for example, to think that males and females respond to a given risk factor equally per exposure dose.  So interactions ('epistasis' in relation to genome elements)  are given very light treatment and basically wished away.

My irate friend basically argued that what is needed is not an end to the data but to the methods. 
One recent paper I was referred to applies application of a method for finding high-risk subsets, and has references to earlier descriptions of the methodology is this: Int J Epidemiol. ("Subgroups at high risk for ischaemic heart disease:identification and validation in 67 000 individuals from the general population", Frikke-Schmidt R et al."), 2015 Feb;44(1):117-28 (but unfortunately it is not freely available).

How effectively this will find really different subgroups is open.  We know, for example, that males and females are not at equal risk and respond differently to other factors, as mentioned above.  We know genomic components interact.  We know that as you get older you get closer to various risks, such as heart disease or cancer, and that the same exposure has different impacts with age, and so on.

The idea both in public health and in medicine (and in evolutionary inference) of identifying causation as effectively as possible, and that includes identifying high risk individuals as early as possible, is of course absolutely the right thing.  There are many instances of genetic risk factors like some variants in the gene responsible for cystic fibrosis, or in the BRCA1 gene related to breast cancer, where the Who Cares? principle applies: the single factor's effect is so predictably strong that one intervenes regardless of the details of how much risk is associated or what other factors might slightly modify the outcome.  We know that the nominal risk factor (e.g., a mutation) doesn't always lead to the same degree of severity, but the variation isn't enough to cause doubt: Who Cares about the details?

But whether in general this sort of method of searching for statistical associations can identify risk earlier enough, where preventive measures might be more helpful, is unclear.  We know about age and sex and smoking and so on, and maybe we don't really gain much from adjusting the exact values.  Or maybe we would.  Likewise for the complex interactions among hundreds of contributing genomic factors.  But there the number of factors and the assumption of independence and so on need to be recognized as being as daunting as they are, relative to statistical risk analysis.

In my view, this is still walking in dreamland.  The major factors will be identified, perhaps with more precision, but we face a huge, open-ended kind of 'multibody' problem.  It's just not possible to analyze all the possible combinations of factors and their interactions to get combination-specific risk estimates.  First, risks are contingent, one factor's effect depending on what else is present, as discussed above.  Second, not all combinations will show up in the data, even in huge samples, so estimating risks if interactions must be accounted for will simply come up short, or perhaps better-put, with unknown or even unknowable precision.

Third, we know very well even from just the recent few decades, that incidence of outcomes changes hugely, yet the genomes basically don't, and while we may estimate the effects of the particular combinations of environments our sampled individuals were exposed to, we simply cannot, even in principle, know what environmental factors current individuals for whom we are being promised precise predictions, will be exposed to.  Yet heritabilities, with all their problems, clearly show that genomes contribute typically far less than half of all risk.  Environmentally and genomically, specific factors or variants come and go, and no two people are identical or even close to it.

A major issue is not just that there is no way, that is no way to know what risks are associated with most risk factors, much less interactions among them, even in principle, but we have no way of knowing the degree of precision of predictions.   This is why, among other things, even if increasing our understanding is a very noble pursuit, promising 'precision' in prediction based on genomes or, really, almost any other risk factors, is irresponsible.

And it's even worse for Big Data.....

Last week we pointed out that the history of technology enabled larger and more extensive and exhaustive enumeration of genomic variation, to apply to understanding the cause of human traits, important diseases or even normal variation like the recent stature paper.

We basically noted that even the earlier, cruder methods such as 'linkage' and 'candidate gene' analysis did not differ logically in their approach as much as advocates of mega-studies often allege, but more importantly, all the approaches have for decades told the same story.  That is, that from a genomic causal point of view, with a few exceptions, complex traits that don't seem to be due to strongly-causal simple genotypes aren't.  Yet the profession has been vigorously insisting on ever bigger and more costly, long-term, too-big-to stop studies to look for some vaguely specified pot of tractably simple causation at the end of the genomic rainbow.

When a genetic effect is strong, we can usually find it, in family members; Wikipedia

But the problem is even worse.  The history we reviewed in last week's post did not include other factors that make the problem even more complex.  First, is somatic mutation, about which we wrote separately last week.  But environment and lifestyle exposures are far more important to the causation of most traits than genomic variation.  The obvious and rather clear evidence is that the heritability is almost always quite modest.  With a few exceptions like stature (which is only an exception if you standardize height for cohort or age, that is, for differing environments), most traits of interest have heritability on the order of 30% or so.

But why are environmental contributions problematic?  If we can Next-Gen our way to enumerating all the (germline) genomic variants, why can't someone invent a Next-Env assessor of environments? There are several reasons, each profound and challenging to say the least.

1. Identifying all the potentially major environmental contributors, which may occur only sporadically, decades before someone is surveyed, unbeknownst to the person, or simply unknown to the investigator, is at least a largely insuperable challenge. 

2.  Defining the exposure is difficult and measurement error probably large and to an unknown or even unknowable extent.  We might refine our list, but how accurately can we really expect to measure subtle things like behavior habits in early childhood, exposures in utero, or even such seemingly obvious things like fat intake, or even smoking exposure. 

3.  Somatic mutations will interact with environmental factors just as germline ones will, so our models of causation are simply inaccurate to an unknown extent. 

4.  Suppose we could get perfect recall, and perfect measurement, of all exposure variables that affected our sampled individuals from in utero to their current ages (and, assuming pre-conception effects on gene expression by such things as epigenetic marking).  That would enable accurate retrospective data fitting, but there is simply no way to predict a current person's future exposures. The agents and exposure levels, and even identity of risk factors is simply unknowable, even in principle.

These facts, and they are relevant facts, show that no matter the DNA sequencing sophistication, we cannot expect to turn complex traits into simple traits, or even enumerable, predictable ones.

More on history
In last Friday's (Oct 17) post on Big Data, we noted that history had shown that even cruder messages were successful at giving us the basic message we sought, even if it wasn't the message we wanted to find.

The history of major, long-term, large-scale studies in general shows another reason we invest false expectations in the desire for Big Data.  Forgetting physics and other sciences, there have been a number of very big, long-term studies of various types.  We have had registries, studies of radiation exposure, follow-ups on diets, drug, hormone and other usage, studies on lipids in places like Framlingham, sometimes multi-generational follow-ups.  They have made some findings, sometimes important findings, but usually two things have happened.  First, they quickly reached the point of diminishing returns, even if the investigators pressed to keep the studies going.  Second, many if not perhaps even most of the major findings have turned out, years or decades later, even in the same study, not to hold up very well.

This isn't the investigators' fault.  It's the fault of the idea itself.  Even things like, say, lipids and other such risk factors are estimated in different ways later in a study than they were earlier, as ideas about what to measure evolves.  Measurement technology evolves so that earlier data are not considered reliable or accurate or cogent enough.  The repeated desire is to re-survey or re-measure, but then all sorts of other things happen to the study population (not least of which is that subjects die and can't be recontacted).  Yet, and we've all heard this excuse for renewal in Study Section meetings, "after all that we've invested, it would be a shame to stop now!"

In a very real scientific investment sense, the shame in many such cases is not to stop now.  That involves some very difficult politics of various kinds, and of course scientists are people, too, and we don't just react to the facts as we clearly know them.

Is an Awakening in progress?
Here and there (not just here on MT!) are signs that not only is the complexity being openly recognized as a problem, but perhaps there's an awakening or a formal recognition of the important lessons of Big Data.

Although the advocates of Big Data Evermore surely remain predominant, a number of Tweets out of this week's American Society of Human Genetics meetings suggest at least a recognition among some that caution is warranted when it comes to the promises of Big Data. The dawning has not yet, we think, reached the serious recognition that Big Data is a wastefully costly and misleading way to go, but at least that genotype-based predictions have been over-promised.  How much more time and intellectual as well as fiscal resources will be dispersed to the winds before a serious re-thinking of the problem or the goals occurs?  Or is the technology first approach too entrenched in current ways of scientific thinking for anything like such a change of direction to happen?

Environmental risk factors - can we ever know when we're done?

Complexity, confounding, heterogeneity: these are just a few of the issues that can and do play havoc with biomedical research where we attempt to infer causation from observational data. A paper just published in Cancer Research ("Circadian and Melatonin Disruption by Exposure to Light at Night Drives Intrinsic Resistance to Tamoxifen Therapy in Breast Cancer," Dauchy et al.) is a good example of some of the issues.  Reported all over the web, including here by the BBC, the study suggests that low levels of light might inhibit the effectiveness of breast cancer drugs.

Many breast tumor cells have receptors for estrogen on their cell membranes; the hormone binds to the receptors and activates cell growth and proliferation, the signal characteristics of cancer cells.  To inhibit that growth, many women with estrogen-positive breast cancers are treated with an estrogen agonist or antiestrogen such as tamoxifen, the metabolites of which bind to the receptor and slow or halt further growth of the tumor.  However, 30 - 50% of tumors become resistant to tamoxifen over time.

Antiestrogens: National Cancer Institute

Dauchy et al. write:

Resistance to endocrine therapy is a major impediment to successful treatment of breast cancer. Preclinical and clinical evidence links resistance to antiestrogen drugs in breast cancer cells with the overexpression and/or activation of various pro-oncogenic tyrosine kinases. Disruption of circadian rhythms by night shift work or disturbed sleep-wake cycles may lead to an increased risk of breast cancer and other diseases. Moreover, light exposure at night (LEN) suppresses the nocturnal production of melatonin that inhibits breast cancer growth.

Many previous studies have shown a relationship between sleep disturbance and cancer risk, and more specifically, between melatonin and cancer risk, so Dauchy et al. focused their research on the role of melatonin as a mediator between tamoxifen and cancer cell growth.  Human breast tumor xenografts (cross species tissue transplants) were implanted into rats and the animals then spent 12 hours in light and 12 hours in dim light or in total darkness every day, and the progression of the tumors followed. The animals exposed to dim light had lower melatonin levels, and their tumors were larger than the rats in total darkness.  Once the tumor reached a specific size, the animals were treated with tamoxifen, and progression or regression of the tumor was then documented.  Further, supplementing the mice in dim light with melatonin lead to smaller tumors and lower tamoxifen-resistance.  The equivalent of the dim light the rats were exposed to would be, according to the authors, the light entering a room under a door.  They suggest that exposure to light from electronic devices before sleep may also lower melatonin levels.

Senior author Steven Hill was quoted in The Telegraph as saying that because resistance to tamoxifen is an increasing problem, this study could be significant.

“Our data, although they were generated in rats, have potential implications for the large number of patients with breast cancer who are being treated with tamoxifen, because they suggest that night-time exposure to light, even dim light, could cause their tumours to become resistant to the drug by suppressing melatonin production.

How melatonin encourages tamoxifen to inhibit breast cancer growth is not understood, although melatonin may play multiple roles in inhibiting cancer cell growth.  The result, if it is replicable, is important however, and because this is not just an association study but an experiment based on a prior hypothesis it is at least that much more persuasive.  But we know from legions of work on inbred mice that mimicking a human effect in one strain doesn't seal the deal, because the results widely differ among strains when investigators are conscientious enough to test multiple strains -- and sometimes some strains show essentially no effect. Presumably this is for genetic reasons, which could mean that the current rat-based study may apply to some, but not to all women.  That would not make it a bad study nor a dismissible result, but it would temper that result.  

The rats in this study were exposed to complete darkness or dim light consistently, 12 hours every day, so it's impossible to know whether variation in light exposure by phase of the moon, say, or other  kinds of varying light sources would make a significant difference in the amount of melatonin a woman makes, nor in how much melatonin is required for tamoxifen to maintain its effectiveness.  
Often, one doesn't know when study results are important or convincing enough to warrant complex changes in drugs, treatment, or behavior.  In this case, it seems rather benign to ask women in treatment to avoid light at bedtime. But even here, should they be compulsive about it?  Shut the curtains in a full-moon?  Nothing to light their way to the bathroom?  And, the authors caution that women shouldn't begin taking melatonin as a result of this study.  

But this work is interesting in its own right, and clearly worth paying attention to.  It also is a reminder of why it's so hard to study environmental risk factors for disease.  As far as we know, when the benefits of tamoxifen were first shown, there was no a priori reason to suspect that it would lose its effectiveness, at least in part, because of exposure to light at night, or interruption in melatonin cycles.  This is a sobering but important reminder that there is always the potential for unidentified and unexpected interactions between unknown and unexpected risk factors (confounders).

We now know at least that there seems to be an interaction between melatonin and tamoxifen in rats, at least in regard to implanted human tumor cells, but we don't know how much of the tamoxifen resistance in human breast cancers this particular interaction might explain (if any, since these were rats, not humans).  But complexity is a hallmark of diseases like cancers, and there's not going to be a single, simple explanation.

Breast cancer and exercise; is this study convincing?

Breast cancer will affect roughly 13% of women in the western world who live to old age, a plague of terror and worry, even when it's eventually cured.  But since cures are only sometimes possible, here's a situation where prevention is worth its weight in....lives.  So any findings could be important -- but it is important to look closely at what might be misleading or false hope.

A recent paper ("Recreational Physical Activity and Leisure-Time Sitting in Relation to Postmenopausal Breast Cancer Risk", Hildebrand et al., in Cancer Epidemiology, Biomarkers and Prevention, Oct, 2013), described on the BBC website, reports that walking an hour a day reduces breast cancer risk in post-menopausal women by 14%.  An association between exercise and lower risk of many diseases, including breast cancer, is often reported, but this paper describes a prospective study by the American Cancer Society of 74,000 women followed for 17 years, a non-trivial study, the results of which we should surely take seriously.

Baseline information on demographic, behavioral, reproductive, medical and environmental factors were collected from women when they enrolled in the study and they answered follow-up questionnaires every 2 years thereafter. This study looked at the association between exercise and risk and found that women who exercised 42 or more MET-hours per week (METs are 'metabolic equivalents', a way to standardize different forms of activity for comparison purposes) were 25% less likely to get breast cancer than the least active women, who reported 0 to 7 MET-hours per week.  Forty seven percent of women reported that walking was their only exercise and among this group, those reporting what the investigators call an intermediate level of exercise, ≥7 hours per week, were at 14% lower risk than women who walked 3 or fewer hours a week.

Nordic walkers; Wikipedia

Other breast cancer risk factors, reported by many earlier studies, include body mass index, weight gain in adulthood, and use of postmenopausal hormones.  Among women in the sample with breast cancer, this study tested whether risk associated with these factors may vary by estrogen receptor (ER) status of tumors.  And, perhaps it's not exercise per se that's important, but time spent sitting, another variable collected for this study, so they looked at that too but of these factors only exercise was found to be associated with risk.

Get out and walk?
What does this all mean?  Should all postmenopausal women now walk 1 hour a day, or more?  Well, as with all studies reporting the effects of purported risk factors as probabilities, the answer isn't clear.  Risk of breast cancer by age 90 is around 13%, as noted earlier.  That is, 1 out of 8 women will develop breast cancer by the age of 90.  So, this means that all women have a 13% risk of breast cancer?  No, that's on average, and cumulative risk is lower at younger ages, increasing with age.  And it isn't even clear what 13% risk means because some women are at no risk, and some at very high risk (women with a high risk BRCA1 or 2 variant, for example, but even these women aren't all at 100% risk, and -- or because -- environmental factors affect risk, though which factors isn't entirely clear).  And, if you are at risk of dying of some other cause before age 90, you need to know the risk by that age, and it will be lower than 13%.

But statistically, a 14% drop in risk, from 13%, would, on average, reduce risk to around 11.2%.  Is that significant?  Should postmenopausal women all start walking an hour a day?  It's up to each woman to decide.  But since no one knows their actual personal risk, and because, based on many findings by numerous other studies over decades, replicated risk factors such as age at menarche, age at first birth, number of births, age at menopause, genetic risk, and various environmental factors such as alcohol use or smoking, and so on, may be more significant than physical activity anyway.

So, given all the other possible risk factors, known and unknown, how much a daily walk would affect a given woman's risk is impossible to predict.  Indeed, one study found that for women with BMI lower than 22 -- thin women -- exercise lowered their risk by 27%, but for women whose BMI was over 30, their risk was lowered by 1%.  The Hildebrand study, however, didn't find any difference in affect of exercise on risk by weight. One can -- should?  must? ask why that is so, and what it means about how epidemiology goes about its business.

Hildebrand et al. also report that 

Physically active women tended to be leaner, more likely to maintain or lose weight during adulthood, more likely to drink alcohol, and less likely to currently smoke. They were also more likely to use PMH and to have had a mammogram in the past year.

These are confounders, variables that may have an effect on risk of breast cancer, but which weren't included in the analysis.  So, it may be that it's because women who walk one hour a day are thinner or don't smoke that their risk is lower than that of less active women, but this study doesn't tease that out.  And how many potential confounders weren't identified specifically, but end up being built into the analysis as if they were due directly to exercise?  This is not an easy question,  nor the fault of any specific study by any means.  But one needs to ask.

The authors state that their study found that "[w]alking on average at least 1 hour/day was modestly associated with lower risk, even in the absence of other recreational physical activities."  Modestly.  But they finally conclude,

Given that breast cancer is the most common cancer affecting women, and that walking is a common activity among postmenopausal women, the finding of a possible lower risk with an average one or more hours/day of walking is of considerable public health interest.

"Considerable"?  In our view, this study doesn't in fact earn that word.  We must ask our usual skeptical question as to whether the investigators may be looking for more funding by promoting this particular finding (and not other risks that might be more surprising but that they didn't find)?  If you do a  huge study, you're naturally compelled to make as much of it as you can....whether or not there is that much to make.

Another side of this is that we already know of the many salubrious effects on health and longevity of exercise.  In that light, this study at best confirms what we already know.  That's good....though if the authors didn't see the connection with body weight, one must wonder about that, and wonder whether such a huge study to find small effects related to something already very well established, was worth doing.  Would smaller, more extensive or focused studies, that could detect specific risks that were really substantial, be a better way? You'll have to answer that for yourself.

It's stressful, worrying about dementia

Remember the big news just a few months ago about the declining incidence of dementia?  There was a story in the New York Times, and many other sites, about a couple of papers in The Lancet (papers here and here.)  We blogged about it at the time.  Gina Kolata in the NYT said this:

A new study has found that dementia rates among people 65 and older in England and Wales have plummeted by 25 percent over the past two decades, to 6.2 percent from 8.3 percent, a trend that researchers say is probably occurring across developed countries and that could have major social and economic implications for families and societies. 

Another recent study, conducted in Denmark, found that people in their 90s who were given a standard test of mental ability in 2010 scored substantially better than people who had reached their 90s a decade earlier. Nearly one-quarter of those assessed in 2010 scored at the highest level, a rate twice that of those tested in 1998. The percentage of subjects severely impaired fell to 17 percent from 22 percent.

Er, but now a study published in BMJ Open this week reports that the more stress women experience in middle age, the higher their risk of dementia as they get older.  This was a prospective study of 800 Swedish women born in 1914, 1918, 1922 and 1930, who underwent a psychiatric examination in 1968, and who were re-examined in 1974, 1980, 1992, 2000 and 2005. They were asked whether they had undergone any of 18 major stressors, including divorce, widowhood, work-related stress and illness of a relative.

Anonymous German picture puzzle, 19th C; source

In the 37 years of follow-up, 19.1% or 153 women developed dementia (425 of the subjects had died over the course of the study), and number of stressors was found to be associated with risk.  And, importantly, risk and number of psychosocial stressors was "independent of long-standing perceived distress." 

Our study shows that common psychosocial stressors may have severe and long-standing physiological and psychological consequences. However, more studies are needed to confirm these results and investigate whether more interventions such as stress management and behavioural therapy should be initiated in individuals who have experienced psychosocial stressors.

What's the mechanism?  The idea is that long-standing exposure to stress hormones may cause "dysregulation in neuroendocrine systems".   But the researchers don't find a one-to-one correlation between stress in midlife and dementia -- that is, not everyone who reported stress became demented and not everyone with dementia had reported major stresses.  The investigators suggest that that is because individuals respond differently to stress.

Well, that's a bit hand-wavy.  Of course, it's always difficult, or even impossible, to apply associations found at the population level to individuals.  Not everyone with high cholesterol levels has a heart attack, and not everyone with high stress will become demented, even if cholesterol and stress are real risk factors.  It means that there are other, unmeasured risk factors involved, that the effect of stress or cholesterol depends on or interacts with unknown variables. So, of course more research is needed (she said snidely).  Or... a different approach to understanding causation. 

But, let's go back to the Lancet papers of July, that reported that incidence of dementia is going down.  Yes, that's on a population level, and it's possible that everyone with dementia in these studies (samples in Denmark, England and Wales) experienced more stress in midlife than those without.   But, if stress is a strong risk factor for dementia, and we were to accept that rates of dementia are really going down, then this would mean that levels of stress in midlife are declining as well.  That is, by the measures in the BMJ Open study, less work-related stress, less divorce, less family illness and fewer parents dying.  Not likely.

Now, it's possible that the younger cohorts in the Swedish study will or did experience less dementia than those born earlier, as in the Danish and British studies, but that question wasn't asked of the data, and anyway the sample size is too small to show a reliable effect if it's stratified by birth cohort.  So, we don't know if incidence of dementia is falling in Sweden as it seems to be elsewhere.  But if it is, that means, to us at least, that something is overriding the effect of stress as a risk factor. 

Do we know more now than we did last year about predicting who'll get dementia in old age?  Another way to put this is: will we know more in 6 months than we do now?  And yet another way is: at what point should we start believing any of these stories?  More generally, is there a better way to understand causation?  At present, for whatever reason, we seem to be doing little more than groping for a black cat in the dark. 

Magical science: now you see it, now you don't. Part II: How real is 'risk'?

Why is it that after countless studies, we don't know whether to believe the latest hot-off-the-press pronouncement of risk factors, genetic or environmental, for disease, or of assertions about the fitness history of a given genotype?  Or in social and behavioral science....almost anything!  Why are scientific studies, if they really are science, so often not replicated when the core tenet of science is that causes determine outcomes?  Why should we have to have so many studies of the same thing, even decade after decade?  Why do we still fund more studies of the same thing?  Is there ever a time when we say Enough!?

That time hasn't come yet, and partly that's because professors have to have new studies to keep our grants and our jobs, and we do what we know how to do.  But there are deeper reasons, without obvious answers, and they're important to you if you care about what science is, or what it should be--or what you should be paying for.

Last Thursday, we discussed some aspects of the problem when a set of causes that we suspect work only by affecting the probability of an outcome we're interested in.  The cause may truly be deterministic, but we just don't understand it well enough, so must view its effect in probability terms.  That means we have to study a sample, of repeated individuals exposed to the risk factor we're interested in, in the same way you have to flip a coin many times to see if it's really fair--if its probability of coming up Heads is really 50%.  You can't just look at the coin or flip it once.

Nowadays, reports are often of meta-analysis, in which, because it is believed that no single study is definitive (i.e., reliable), we pool them and analyze the lot, that is, the net result of many studies, to achieve adequate sample sizes to see what risk really is associated with the risk factor. It should be a warning in itself that the samples of many studies (funded because they claimed and reviewers expected them to be adequate to the task) are now viewed as hopelessly inadequate.  Maybe it's a warning that the supposed causes are weak to begin with--too weak for this kind of approach to be very meaningful?

Why, among countless examples, after having done many studies don't we know if HDL cholesterol does or doesn't protect from heart disease, or antioxidants from cancer, or coffee is a risk factor, or obesity is, or how to teach language or math, or avoid misbehavior of students, or whether criminality is genetic (or is a 'disease'), and so on--so many countless examples from the daily news, and you are paying for this, study after study without conclusive results, every day!

There are several reasons.  These are serious issues, worthy of the attention of anyone who actually cares about understanding truth and the world we live in, and its evolution.  The results are important to our society as well as to our basic understanding of the world.

So, then, why are so many results not replicable?
Here are at least some reasons to consider:

1.  If no one study is trustworthy, why on earth would pooling them be?
2.  We are not defining the trait of interest accurately
3.  We are always changing the definition of the trait or how we determine its presence or absence
4.  We are not measuring the trait accurately
5.  We have not identified the relevant causal risk factors
6.  We have not measured the relevant risk factors accurately
7.  The definition of the risk factors is changing or vague
8.  The individual studies are each accurate, and our understanding of risk is in error
9.  Some of the studies being pooled are inaccurate
10.  The first study or two that indicated risk were biased (see our post on replication), and should be removed from meta-analysis....and if that were done the supposed risk factor would have little or no risk.
11.  The risk factor's effects depend on its context:  it is not a risk all by itself
12.  The risk factor just doesn't have an inherent causal effect:  our model or ideas are simply wrong
13.  The context is always changing, so the idea of a stable risk is simply wrong
14.  We have not really collected samples that are adequate for assessing risk (they may not be representative of the population at-risk)
15.  We have not collected large enough samples to see the risk through the fog of measurement error and multiple contributing factors
16. Our statistical models of probability and sampling are not adequate or are inappropriate for the task at hand (usually, the models are far too simplified, so that at best they can be expected only to generate an approximate assessment of things)
17.  Our statistical criteria ('significance level') are subjective but we are trying to understand an objective world
18.  Some causes that are really operating are beyond what we know or are able to measure or observe (e.g., past natural selection events)
19. Negative results are rarely published, and so meta-analyses cannot include them, so a true measure of risk is unattainable
20. The outcome has numerous possible causes; each study picks up a unique, real one (familial genetic diseases, say), but it won't be replicable in another population (or family) with a different cause that is just as real
21. Population-based studies can never in fact be replicated because you can never study the same population--same people, same age, same environmental exposures--at the same time, again
22. The effect of risk factors can be so small--but real--that it is swamped by confounding, unmeasured variables. 

Overall, when this is the situation, the risk factor is simply not a major one!
This situation--and our list is surely not exhaustive--is typical and pervasive in observational rather than experimental science.  (In the same kinds of problems, lists just as long exist to explain why some areas even of experimental science don't do much better!)

A recent Times commentary and post of ours discussed these issues.  The commentary says that we need to make social science more like experimental physical science with better replications and study designs and the like.  But that may be wrong advice.  It may simply lead us down an endless, expensive path that simply fails to recognize the problem.  Social sciences already consider themselves to be real science.  And presenting peer-reviewed work that way, they've got their fingers as deeply entrenched into the funding pot as, say genetics does.

Whether coffee is a risk factor for disease, or certain behaviors or diseases are genetically determined, or why some trait has evolved in our ancestry...these are all legitimate questions whose non-answers show that there may be something deeply wrong without current methods and ideas about science.  We regularly comment on the problem.  But there seems to be no real sense that there's an issue being recognized, in opposition to the forces that pressure scientists to continue business as usual---which means that we continue to do more and more and more-expensive studies of the same things.

One highly defensible solution would be to cut support for such non-productive science until people figure out a better way to view the world, and/or that we require scientists to be accountable for their results.  No more, "I write the significance section of my grants with my fingers crossed behind my back" because I know that I'm not telling the truth (and the reviewers, who do the same themselves, know that you are doing that).

As it is, resources go to more and more and more studies of the same that yield basically little, students flock to large university departments that teach them how to do it, too, journals and funders make their careers reporting their research results, and policy makers follow the advice.  Every day on almost any topic you will see in the news "studies show that....."

This is no secret: we all know the areas in which the advice goes little if anywhere.  But politically, we haven't got the nerve to make such cuts and in a sense we would be lost if we had nobody assessing these issues.  What to do is not an easy call, even if there were the societal will to act.

Magical science: now you see it, now you don't. Part I: What we mean by 'risk'

The  life, social, and evolutionary sciences have a problem.  We posted about the issue of their non-replicability last Friday but that is only part of the problem.  They also have non-predictability (see a recent Times commentary), but both replicability and predictability are key elements of science as we know it. 

It is difficult to make rigorous assertions that have the kind of predictive power we have come (rightly or wrongly) to expect of science, on the typical if often unstated assumption that our world is law-like, the way it seems that the physical and chemical universe are.

A clear manifestation of the problem is the way that findings in epidemiology and genetic risk say yes factor X is risky, then a few years later no, then yes again, and so on.   Can't we ever know if factor X is a cause of some outcome Y?  In particular, is X a risk factor?  X could be a reason for natural selection, a component of some disease, and so on.

Often, we think of this as a probability.  As we've posted before, that means that exposure to that factor yields some probability that the outcome will be observed.  If you have two copies of a gene, there is usually a 50% risk or probability that a given one of your children (or parents, or sibs) will also have it.

So, in a strange turn of phrase, coffee or a given HDL cholesterol level is said to be a risk factor for heart disease, or a given genetic variant is a risk factor for cancer.  And we try to estimate the level of that risk--the probability that if that factor is present, you will manifest the outcome.  Among various possible risk factors, the modern concept of science has it that you are at some net or overall risk of the outcome, like having a heart attack, depending on your exposure to those risk factors.

In evolutionary terms, having a particular genetic variant can have some probability  (or some similar measure) of reproducing, or surviving to a particular age.  Among various possible genetic risk factors, what you have puts you at some net risk of such outcome, which is your evolutionary fitness in the face of natural selection, for example. 

If we assume that enumerated causes of this sort, and that they really are causes, are responsible for a trait then your exposure level can be specified.  The causes might truly be deterministic, in the way gravity determines the rate an object will fall--here or anywhere in the universe--but that our incomplete level of knowledge is such that we can only express its effect in terms of probability.

Still, we assume that probabilistic causation is real.  When things are the result of probabilities, we can know the causes but can't predict the specific outcome of any given instance.  This is the sense in which we know a fair coin will come up Heads 50% of the time, but can't predict the result of a given flip.  Actually, and we've posted about this before but the issue of probability is so central to much of science that we keep repeating it, the coin may be perfectly deterministic but we just don't know enough, so that for all practical purposes the result is probabilistic.

In such cases, which are clearly at the foundation of evolutionary inference and of genetic and other biomedical problems, we must estimate the risk associated with a given cause by choosing a sample from all those at risk, and seeing what happened to them.  Then, we assume we know the causal structure and can then do what we must be able to do, if this is actual science: predict the outcome.  This must be so if the world is causal, even if our predictions are expressed in terms of probabilities: given your genotype you have xx probability of getting yy disease.

So, with our huge and munificently funded science establishment, why is it that day after day the media tout the latest Dramatic Finding....that is just as noisily touted the next day when the previous assertion is overturned?

Why is it that we don't know if coffee is a risk factor for disease?  Or isn't?  Or isn't for the moment until some new study comes along?  Or maybe until some environmental factor changes, like the type of filter paper McDonald's uses in its coffee maker, say--but how would we ever know whether that explains the flip-flopping findings?

Why indeed do we have to continue doing studies of the same purported risk factor to see if they are really, truly risks?  These are fundamental questions not about the individual studies, but about the current practice of science itself.

If we look at the reasons, which is tomorrow's post, we'll see how shaky our knowledge really is in these areas, and we can ask whether it is even 'science'.

Does salt really cause cancer?

A new report on cancer incidence and mortality in the UK, described on the Guardian website, suggests that 40% of cancers in women and 45% in men are preventable, due to lifestyle choices. This is great news for people pushing healthy diets and exercise (perhaps not such great news for those pushing genetic causation).

Dr Rachel Thompson, deputy head of science for the World Cancer Research Fund, said: "This adds to the now overwhelmingly strong evidence that our cancer risk is affected by our lifestyles.

"We hope this study helps to raise awareness of the fact that cancer is not simply a question of fate and that people can make changes today that can reduce their risk of developing cancer in the future. 

So, what is this overwhelmingly strong evidence? The authors chose 14 different risk factors, as listed in the table below (taken from the paper).

Table 1. Exposures considered, and theoretical optimum exposure level

            Exposure                             Optimum exposure

Tobacco smoke                                                 Nil

Alcohol consumption                                       Nil

Diet

1.      Deficit in intake of fruit and veg        ≥5 servings (400 g) per day

2.      Red and preserved meat                     Nil

3.      Deficit in intake of dietary fiber        ≥23 g per day

4.      Excess intake of salt                             ≤6 g per day

Overweight and obesity                                BMI ≤25 kb m^-2

Physical exercise                                            ≥30 min 5 times per week

Exogenous hormones                                    Nil

Infections                                                         Nil

Radiation – ionizing                                      Nil

Radiation – solar (UV)                                 As in 1903 birth cohort

Occupational exposures                               Nil

Reproduction: breast feeding                      Min of 6 months

The study calculated the "population attributable fraction" of each risk factor, that is, how much excess cancer was due to exposure to the risk factor. They compared cancer incidence in those exposed with incidence in those not exposed and assumed any excess (or, in theory, deficit) was due to the risk factor.

They chose risk factors based on the following criteria:
1. There was sufficient evidence on the presence and magnitude of likely causal associations with cancer risk from high-quality epidemiological studies.
2. Data on risk factor exposure were available from nationally representative surveys.
3. There were achievable alternative exposure levels that would modify the risk.

They calculated the relative risk per unit of exposure for cancers with probable or convincing causal associations with each risk factor, based on observational epidemiological studies. These would be the same studies you see reported in the news every day, telling you that you should or shouldn't eat butter, should or shouldn't go out into the sun, should or shouldn't eat sugar.

But there is something rotten in the state of Denmark, because despite billions of dollars, hundreds of thousands of study subjects, countless studies over decades of time by the most prominent (or, that is, highly placed) epidemiologists, the actual truth about the data is very different, surprising as that may seem. Indeed, as far as we know, the only truly convincing risk factors in this list are tobacco, papilloma virus and radiation, and even there it isn't really clear how much radiation exposure is too much (and many would say no exposure is the only completely safe exposure, though some exposures may detect treatable dangerous conditions and be good in the net).

The other behavioral risk factors have been shown in some studies to account for a small fraction of risk, though the results aren't always replicable. Indeed we can assert what we've just said because, by chance we just heard a talk by Gary Taubes, a science journalist for the New York Times and Science, among other outlets, who has systematically been debunking the idea that low fat diets have been shown definitively to prevent heart disease and cancer. He says the data just aren't there and never have been, but that it's been a belief so entrenched that it can't be denied because of all the vested interest that would challenge.

Hey, we like a good heretic as much as the next guy. And Taubes has written some of the best stuff out there on why observational epidemiology can't answer basic questions about cause and effect (here and here, for example). His work on dietary fat is very convincing, and his more general point that observational epidemiology can't be the basis for dietary recommendations is equally convincing. (So it's confusing that he's now a strong advocate for the idea that processed sugar is toxic, and responsible for the obesity and diabetes epidemics all over the globe -- conclusions largely based on the same kinds of observational studies he debunks when it comes to risk factors he doesn't like.)

But we digress. We are more than willing to accept that environmental risk factors can lead to disease. If not, only genetic variation would cause disease, and that clearly isn't so! We write about this all the time on MT. We just aren't nearly as ready to accept that we know definitively what those risk factors and their associated risks are. Nor that everyone is equally at risk from every factor.

Some of the optimal exposure levels in the list probably come under the category of 'wouldn't hurt', but public health measures are, by design, meant to be population-based, and the economic costs of encouraging lifestyle changes on a population level are not trivial. Nor is the cost of lost credibility when the risk factors turn out to be less important than we've been told after all.

Worse, risk is always and necessarily estimated retrospectively by relating outcomes quantitatively to exposure histories. But what we want to know is the future risk, and we know very well that we cannot predict the mix or amount of exposures to who-knows-what risk factors in the future. This is a deeply troubling problem, since major changes in risk for many or even most complex disease have occurred, often because of unclear behaviors or exposures, just in the past 50 years or so.

So, here's the safest conclusion to date -- do (most) everything in moderation, and don't worry about it. Something will get you in the end, so try to have the best time you can before that.

If genetic causation is complex, why should risk factors be any less so?

Every day, it seems, the forces of biological simplism -- the hunger for, and vested interest in simple answers to complex questions -- suffer a setback. Today, it's large-study results that show that screening for a simple marker for early prostate cancer detection seems to be ineffective ( New York Times prostate cancer article ). It may be harmful in the sense of leading to the detection of benign cases, and then some intervention with its associated risk of morbidity. Earlier this year somewhat similar results appeared for mammographic screening for breast cancer. The point is not to coldly denigrate attempts at early detection, but to show the importance of recognizing nature's complexity. Those who suffer from cancer--and we all know such people, and many of us will be such people--deserve all the care and concern that can be mustered. But can we think of better ways to approach this genetically complex problem? Is standard reductionism, trying to identify individual risk factors, or even single risk factors, the way to go? Or will some smart young researcher give us the benefit of conceptually innovative ideas?

For most risk factors, genetic or otherwise, the situation is similar: cholesterol, blood pressure, even obesity have complex and poorly understood associations with subsequent disease outcomes, and with prior genetic risk factors. It is already known, however, that the most effective way to head off chronic diseases is not to smoke, get exercise, and eat a moderate, balanced diet (including even to have a drink now and then!).

But, that conceptually innovative idea is not going to come anytime in the next month or so -- biology is on holiday. This is not like France, where everyone goes to the seaside in August. No, it's because of our 'stimulus' package's ad hoc grants program. Like lemmings to the sea, or hogs to the trough, every scientist and his relatives (living or deceased) is charging headlong for the new money. Whether this is a moral way to spend these funds is an open question. But everyone's now too busy putting together their hoped-for bonanza grants to do any actual scientific work. Ironically, the stimulus package's 'challenge grants' may turn out to be a NON-work initiative for science!

Presumably, the crush will end and we'll all get back to work. One can predict that, due to the gold rush the funding percentages won't be any better, and they may be worse for this 'easy money'. Time will tell.