Processed meat? Really? How to process epidemiological news

So this week's Big Story in health is that processed meat is a risk for breast cancer.  A study has been published that finds it so.....so it must be true, right?  After all, it's on CNN and in some research report.  Well, read even CNN's headliner story and you'll see the caveats, the admissions, softened of course, that the excess risk isn't that great, but, at least, that the past studies have been 'inconsistent'.

Yummy poison!!  source: from the web, at Static.zoonar.com

Of course, with this sort of 'research' the weak associations with some named risk factors can easily be correlated with who knows how many other behavioral or other factors, and even if researchers tried to winnow them out, it is obvious that it's a guessing game.  Too many aspects of our lives are unreported, unknown, or correlated.  This is why week after week, it seems, do-this or don't-do-that stories hit the headlines.  If you believe them, well, I guess you should stop eating bacon.....until next week when some story will say that bacon prevents some disease or other.

Why breast cancer, by the way?  Why not intestinal or many other cancers?  Why, if even the current story refers to past results as being 'inconsistent' do we assume this one's right and they, or some of them, were wrong?  Could it be that this is because investigators want attention, journalists need news stories, and so on?

Why, by the way, is it always things that are actually pleasurable to eat that end up in these stories?  Why is it never cauliflower, or rhubarb, or squash?  Why coffee and not hibiscus tea?  Could western notions of sin have anything to do with the design of the studies themselves?

But what about, say, protective effects?
Of course, the headlines are always about the  nasty diseases to which anything fun, like a juicy bacon sandwich, not to mention alcohol, coffee, cookies, and so on seems to condemn us.  This makes for 'news', even if the past studies have been 'inconsistent' and therefore (it seems) we can believe this new one.

However, maybe eating bacon sandwiches has beneficial effects that don't make the headlines.  Maybe they protect us from hives, antisocial or even criminal behavior, raise our IQ, or get fewer toothaches.  Who could look for all those things, when they're busy trying to find bad things that bacon sandwiches cause?  Have investigators of this sort of behavioral exposure asked whether bacon and, say, beer raise job performance, add to longevity, or (heavens!) improve one's sex life?  Are these studies, essentially, about bad outcomes from things we enjoy?  Is that, in fact, a subtle, indirect effect of the Protestant ethic or something like that?  Of the urge to find bad things in these studies because they're paid for by NIH and done by people in medical schools?

The serious question
There are the pragmatic, self-interested aspects to these stories, and indeed even to the publication of the papers in proper journals.  If they disagree with previous work on the purportedly same subject, they get new headlines, when they should perhaps not be published without explicitly addressing the disagreement in real detail, as the main point of the work--rather than the subtle implication that now, finally, these new authors have got it right.  Or at least, they should not headline their findings.  Or something!

Instead, news sells, and thus we build a legacy of yes/yes/no/maybe/no/yes! studies.  These may generally be ignored by our baconophilic society, or they could make lots of people switch to spinach sandwiches, or many other kinds of effects.  This latter is somewhat akin to the quantum mechanical notion that measurement gives only incomplete information but affects what's being measured.

Epidemiological studies of this sort have been funded, at large expense, for decades now, and if there is anything consistent about them, it's that they are not consistent.  There must be a reason!  Is it really that the previous studies weren't well done?  Is it that if you fish for enough items, you'll catch something--big questionnaire studies looking at too many things?  Is it changing behaviors in ways not being identified by the studies?

Or, perchance, is it that these investigators need projects to get funded?  This sort of yo-yo result is very, very common.  There must be some explanation, and that inconsistency itself is likely as fundamental and important as any given study's findings.  Maybe bacon-burgers only are bad for you in some cultural environments, and these change in unmeasured ways, and that varying results are not 'inconsistent' at all--maybe it's the expectation that there's one relevant truth, so that inconsistency suggests problems in study design.  Maybe the problem is in simplistic thinking about risks.

Where do cynical possibilities meet serious epistemological ones, and how do we tell?

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.

Breast cancer and other unknowns -- between a (probalistic) rock and a (probabilistic) hard place

The New York Times Magazine cover story on Sunday was disturbing.  With "Our Feel-Good War on Breast Cancer," Peggy Orenstein, herself a woman with a history of breast cancer, confronts the frustrating lack of progress in understanding what causes this disease that has or will touch so many of us, the best ways to detect it, how to treat it, which messages to convey to at-risk women (which is, of course, essentially all of them) and how best to do it.

It has been four decades since the former first lady Betty Ford went public with her breast-cancer diagnosis, shattering the stigma of the disease. It has been three decades since the founding of Komen [the best-funded breast cancer foundation in the US]. Two decades since the introduction of the pink ribbon. Yet all that well-meaning awareness has ultimately made women less conscious of the facts: obscuring the limits of screening, conflating risk with disease, compromising our decisions about health care, celebrating “cancer survivors” who may have never required treating. And ultimately, it has come at the expense of those whose lives are most at risk.

A central ongoing debate is over mammography and whether it saves lives. It's clear that it doesn't detect all tumors, and that it does detect tumors that would disappear with no intervention, and tumors that will never metastasize, and may itself be a risk factor, and yet it's the primary tool for health education and prevention that the breast cancer advocacy groups have, and which they fiercely defend. Despite statistical evidence that it may not save lives and could be causing a lot of harm.

What is 'risk' and how do we know?
People who pay attention are told that this procedure, or that treatment, has certain 'risk' associated with it, or some 'probability' of success. These are difficult words for ordinary people and even for professionals. And the problem is by no means confined to complicated biomedical situations. There is the emotionally subjective aspect of risk, as in "oooh, that's risky", in which emotions override considerations of numbers, often very misleadingly. And there is the scientifically subjective aspect of risk, as in "significant at the 0.05 level." (0.05 is just an arbitrary cutoff for 'signficance' -- another loaded and widely misperceived word).

But even if we overlook these aspects of subjectivity, there is another that is at least as important. A risk is a value between 0 (can't happen) and 1 (certain to happen). How is that determined? It's estimated from some sample of data, in which some number, say m, of observations were made, and some number, n, of them experienced the outcome for which the risk is estimated as n/m. But that is retrospective, meaning it only relates to what happened to those we studied, and there are all sorts of problems in making the data 'representative' of what we want to estimate. Did we measure all the proper variables that might affect the risk. Did we measure the exposures and outcomes accurately enough? Did we sample enough people, and were they 'random' relative to the causes and effects we want to understand?

Even if all went well, there is the inescapable fact that what we want for making decisions, such as clinical decisions on treatment, are clear-cut answers, and we can't get them because they don't exist. Every patient is different, there are multiple kinds of breast cancers, as most other cancers, and multiple ways to get the disease, so predicting an individual's risk of disease is impossible to do with very much precision. Even women with BRCA1 or 2 mutations that are strongly associated with risk aren't at 100% risk, and at some times in history, have been at much less risk than others.

And risks cutting in all directions
It is not just the risk of cancer we have to consider. It is the risk of cancer that might be due to having a mammogram. It is the risk (or probability) that the test will accurately detect a cancer. It's the risk that the cancer being detected would otherwise have progressed to become a health danger (rather than regressing and just going away unnoticed). There is the emotional risk of having a diagnosis.

Then there is the risk of any sort of treatment: biopsies, surgeries of this or that type, radiation treatment, chemotherapy. And risks associated with aspects of life that might yet induce a cancer if you don't already have one, or induce another if you do. And risks are often in these cases not just some simple probability, but the probability of living 1, 5, or 10 years, or of no recurrence.

Every one of these is a probability of sorts that has to be estimated. Since risk factors, such as lifestyles that might be associated with disease, are always changing, and we don't even know what most of those factors are, and since any treatment must be followed up on a large enough number of people to know how well (in probability terms!) it works, women are caught between many probabilistic rocks and probabilistic hard places.

It's exquisitely painful to have to deal with so many competing uncertainties in so very personal an area, when everyone is doing his or her best to evaluate options and knowledge for the shared goal of avoiding or treating disease. Our society is very poorly trained to deal with probabilities in any serious quantitative way; we're more used to high and low values that we can think of as won't-happen or will-happen. And the subtleties play on the scientists and physicians, too. No matter how sincere you are, you want your patients to do what you think is best, as unemotionally as possible, based on your personal assessment of this host of competing probabilities. And you are also embedded in this subjectivity, because you want your view to be correct, and you are always vulnerable to shading the evidence to suit your preferences.

There is plenty of advocacy, dissembling, careerism, vanities, and all that in those dealing with these issues, both patients and perhaps especially researchers and clinicians. New findings that challenge the currently accepted verdict of probabilities are emotionally adopted or resisted.

One might say that much of this includes improper or uninformed behavior. But the problem is that our knowledge, and probabilistic 'knowledge' in particular, is shaky and difficult to handle and would be if we were all saints with IQs of 200.

Trying to balance incomplete information, probabilistic estimates imperfectly arrived at, emotional reactions, and the poor understanding of probabilities plagues even the most honorable and objective of us. And it's generally even worse.  Probabilities and risks in cases like this are averages based on analysis of groups -- patients treated this way or that, screening populations, and so on.  But the risks you need to know about are for you, though is far from clear that everyone in one of these risk groups has the same risk.

It is not even clear that there is advice one can give: since the many relevant probabilities are all imperfectly known, to imperfectly known extents, the most we can do perhaps is acknowledge the realities, that the kind of evidence we must weigh is just the way science works these days.  When, whether, or how we might conceive of the problems differently, to get closer to actual individuals rather than groups, when everybody is different, is an open question--but one of widespread relevance.

Unfortunately, at this time in history of course, even in so painfully urgent and important areas as diseases like cancer, life is a roll of the dice -- and we don't even know what kind of dice we're rolling.

The consequences of casual concepts of causation

Exposure to X-rays can cause cancer, so their use as a routine and repetitive screen for tuberculosis (and, for readers old enough to remember, to look for the fit of shoes at a shoe-store), or for routine CT scans, and so on, is questionable.  Dental x-rays seem to be so safe that their risk, which is probably not zero, is nearly unmeasurable and presumably worth the dental problems they can find (hopefully, though not certainly, they are not taken too often as a source of profit).

Developing breast tissue in young girls is vulnerable, so they were not routinely given chest x-rays.  But breast cancer is common and serious, so mammography was seen for a long time as clearly a valuable life-saver, if used on peri- or post-menopausal women.  But recent studies have raised questions.  Interestingly, this is not because of new cancers that may be caused by the screening (though the number may not be zero).  It was because they could detect small anomalies that were then followed up.   Some would turn out to be cancer, even if in an early stage.  But the follow-up is psychologically traumatic and has its own unexpected consequences.  And even more, studies have shown that some of these cancers would regress on their own.  So mammography, despite so many being convinced of its value, is now under scrutiny:  when and how often and on who is this false positive risk too great to justify routine mammographic screening?

PSA testing has become routine for finding prostate cancer in older men, because prostate cells that are too active churn out PSA (prostate specific antigen), so high PSA levels, just as with suggestive mammograms, have been considered indicative of the need for follow-up.  Again, that has its own morbidity--including, gulp, impotence!--so it, too has come under scrutiny.  Indeed, as with mammography, studies have shown that the intervention's risk and the fact that many of the tumors would regress, or would stay silent until some other cause took the poor guy away, suggests that routine PSA testing be stopped.

Now, a new report suggests the same thing for ovarian cancer screening.  The reasons are the same, and surely there will be as much controversy.   What is this all about, and what is one to do, and why do we see this?  Surely and hopefully, it cannot be all, or even primarily, due to the profitability of screening services and follow-up.

More likely this reflects the belief in technology, fed by and into the hunger for early diagnosis and treatment of very nasty diseases that threaten the quality of life or even life itself.  Is it that early ideas about what might be early risk factors, based on some first rounds of studies, lead researchers anxious for important findings, and clinicians anxious for effective detection, to believe what are not very sound results?  This must be the case, unless the early studies were seriously flawed in their methodology.

Probably more importantly, this reflects a profound modern-day problem in science: the way that complex, multi-factor causation is studied by statistical studies, and the difficulty of getting good enough samples, well-enough understood, to generate reliable results.  Plus, many factors are lifestyle-related, and they change over time.

But could these findings be a reflection of a point we often wonder about, and that applies to evolutionary reconstructions as well--namely, that the assumptions underlying statistical studies of these types make and test assumptions?  Is it that the methods assume a type or level of regularity that simply does not reflect how things really are?

If that is the case then we have to await the next brilliant insight that will transform  how we think.  Meanwhile, we are apparently stuck not knowing why results and opinion change so often, or whether we can trust the latest study more than the previous studies it overturns.....or whether we have to believe that the latest study, too, will shortly be overturned.

Either the situation is straightforward but we just haven't done the right studies, or we're in a deeper epistemological hole than most people would like to think, calling for the kinds of creative thinking that no grant or research 'system' can order up, but simply depend on the lucky arrival of the required genius.

Genetic methods and technologies with potentially important payoff

Last week a group of cancer researchers, largely in the UK, announced in Nature that after characterizing the genome and transcriptome of 2000 breast tumors, they have found that breast cancers can be divided into 10 subgroups, or even 'different diseases', each with characteristic aggressiveness and response to therapy.  The 'transcriptome'  is the set of the 20-25,000 human genes that the tumor cells are actually using (other genes are presumably quiescent and not needed by those cells).

To us, this demonstrates a welcome appropriate use of genetics in an important problem -- even if  not so welcome is the typical hype that's come with the announcement -- this will 'revolutionize' treatment, and this is 'breakthrough research', and so on.  It's a problem when every story is accompanied by such hype, a boy-crying-wolf kind of problem.  This study may well be very important work, but let's let it prove itself before shouting from the rooftops.   The human genome sequence, after all, was going to allow us all to live forever by 2020 if not earlier, leaders proclaimed in the '90s.

From the Nature paper:

We present an integrated analysis of copy number and gene expression in a discovery and validation set of 997 and 995 primary breast tumours, respectively, with long-term clinical follow-up. Inherited variants (copy number variants and single nucleotide polymorphisms) and acquired somatic copy number aberrations (CNAs) were associated with expression in ~40% of genes, with the landscape dominated by cis- and trans-acting CNAs. By delineating expression outlier genes driven in cis by CNAs, we identified putative cancer genes, including deletions in PPP2R2A, MTAP and MAP2K4. Unsupervised analysis of paired DNA–RNA profiles revealed novel subgroups with distinct clinical outcomes, which reproduced in the validation cohort.

('Unsupervised analysis' refers to a particular statistical method used.)

Both germline (transmitted in sperm or egg) and somatic (body cell) variation was found to contribute to tumor occurrence and architecture.  The terms are technical but essentially all sorts of variation was found to be involved:  CNA's (copy number aberrations), CNV's (copy number variants) and SNPs (single nucleotide polymorphisms), either on the same chromosome as the contributing gene (cis) or on a different chromosome (trans):  all these contributed to variation in expression of genes associated with tumors.

a, Venn diagrams depict the relative contribution of SNPs, CNVs and CNAs to genome-wide, cis and trans tumour expression variation for significant expression associations (Šidák adjusted P-value ≤0.0001). b, Histograms illustrate the proportion of variance explained by the most significantly associated predictor for each predictor type, where several of the top associations are indicated. [Figure and caption from the paper.]

If, as the paper suggests, integrating genetic information about germline as well as somatic aberrations, as well as tumor type helps to clarify decisions about treatment of breast cancer, this will prove to be a valuable application of genetic technologies and information.

This is just the the first step, however.  As the lead author, Carlos Caldas said, interviewed on the BBC Radio 4 program, Material World, on April 19, they've identified what he equated with continents, but the rivers, mountains, plains and other aspects of the landscape are yet to be determined. 

Indeed, this is a study of 'primary' tumors, and it is not clear what the story is if the tumor has spread to other parts of the  body ('metastasized').  Other recent studies have shown what cruder methods had previously shown, that new genetic changes occur that allow those secondary tumors to spread and grow.  Likewise treatment itself selects for cells that are by their good luck, and the patient's bad luck, resistant.  This study appears to show, assuming no post-study tumor recurrences, that the predictive methods can lead treatment to stay a step ahead of the tumor's evolution.

It's been clear for many years that somatic genetic changes may be important in disease, and cancer, which is a cascade of cells descendant from a founding aberrant misbehaving cell is the classic archetype.   In such instances, it makes sense to search for variation among cells within the individual.  Whether or not the pattern is too complex to be very useful, only time will tell.

If such work can reveal useful information as this paper claims, and treatment can be focused on patient-specific traits, there may indeed be something to shout from the rooftops.  Whether or not complexity again rises to bite, this study shows an appropriate use of high-throughput genetic technology.

Mammography and breast cancer mortality: correlation is not causation

Breast cancer mortality has been declining in the industrialized world over the last several decades.  At the same time, mammographic screening has become the norm in most of Europe, Australia and North America.  Does early detection explain the reduced mortality? A paper in the Aug 1 British Medical Journal concludes that it does not.

The study looked at three country pairs (Northern Ireland vs the Republic of Ireland, the Netherlands vs Belgium and Flanders, and Sweden vs Norway), in which breast cancer screening was introduced at different time periods.  The paper points out that it would not be unprecedented for screening to lead to lower death rates because that's exactly what happened with cervical cancer mortality, which quickly decreased in the countries in which it was first introduced, with a lag time as others adopted the practice. That seems clearly because early detection led to curative treatment.

Thus, the researchers decided to test the idea that mortality from breast cancer fell first in countries that first had high screening rates (for the purposes of this study, over 70%).  So, they looked at breast cancer mortality in these neighboring pairs of countries, which happened to have implemented screening many years apart, and they controlled for confounding factors that might have influenced mortality risk.

All data were population data: breast cancer death rates for each country were drawn from a World Health Organization mortality database, and confounding variables (obesity, age at first birth and total fertility, all of which have been shown to affect risk of breast cancer) from nation-wide statistics on these, by age.  Data on breast cancer management was also population-based; they looked at expenditures on anticancer drugs, the 'uptake of new anticancer drugs in general...after their introduction in the country, and the uptake of trastuzumab [a recently introduced cancer medication that is effective against a subset of breast tumors]...after its introduction in the country.'

National organized screening was first introduced in Sweden in1986, and by 1990 90% of Swedish women had been offered a screening.  Screening rates there are among the highest in any country.  In Norway, in contrast, organized screening wasn't introduced until 1996 as a pilot project, and was gradually expanded until in 2005 the program was nationwide.  Breast cancer mortality fell by 16% in Sweden and by 24% in Norway between 1989 and 2006. 

Mammography screening and mortality, Sweden and Norway

Trends are similar in the Netherlands and Belgium, with organized screening beginning in 1989 in the former, and not until 2001 in the latter, with only 59% of Belgian women undergoing screening by 2005.  Breast cancer mortality fell by 25% in the Netherlands from 1989 to 2006, and by 19.9% in Belgium, but  by 24.6% in Flanders.

Finally, screening began in the 1990s in Northern Ireland, but was first introduced in the Republic of Ireland in 2000, and it wasn't until 2008 that more than 70% of women over 50 were screened.  Breast cancer mortality decreased by 29.6% in Northern Ireland and by 26.7% in the Republic of Ireland between 1989 and 2006.

The authors report that, at the population level, obesity levels and reproductive variables didn't differ significantly between countries, and cancer treatment, measured by drug expenditures, didn't differ significantly between countries, though uptake of recent drugs seemed to be slower in Norther Ireland than other countries.

This is a population-based study, which means that confounding variables can't be linked to individuals.  Thus, for example, it's possible that obesity rates differ significantly among women who undergo mammography and those who don't.  Given this limitation, the study concludes

The contrast between the timing of breast cancer screening being implemented and the similarity in mortality reduction between the country pairs do not suggest that a large proportion of the mortality reduction after 1990 can be attributed to mammography screening. Improvements in treatment and in the efficiency of healthcare systems may be more plausible explanations. Our study adds further population data to the evidence of studies that have used various designs and found that mammography screening by itself has little detectable impact on mortality due to breast cancer.

Given that there is a small, perhaps even undetectable, increase in risk of breast cancer due to exposure to the screening process itself, this study is another that gives pause to the idea that all women over either 40 or 50, depending on who is doing the recommending, would benefit from routine annual, or biannual, mammography.  X-ray screening for breast cancer is, however, a practice by now considered by many to be state-of-the-art medical care for women in many countries, and previous attempts to recommend that its use be curtailed, based on similar findings in 2009, met with loud and widespread disapproval.

There is the other fact that mammography leads to early diagnosis and intervention, but that means (1) higher rates of reported cases and (2) morbidity and trauma (including psychological) due to the treatment of the detected tumors.  But evidence we discussed in earlier posts (here and here) showed that many of those tumors would have resolved on their own, needing no treatment. So even if lifestyles are responsible for the overall reduction in cases, screening has the potential negative effect of over-diagnosis.

The fact that mammography is big business can't be ignored here.  Just as DNA sequencing is big business, and it's in the interest of sequencing manufacturers for science to find more uses for their machines, the makers of mammographic equipment, as well as mammography clinics, can't take kindly to the idea that women might need less screening.

But even if we don't throw the complicating factor of vested interests into the mix, as with other health issues that are studied on the population level, it's difficult to know how to apply these results to individuals.  Population-level data lead to population-level recommendations.  That is the job of public health, of course, so the problem is not simple.  In 2009, much of the negative reaction to the recommendations for less screening came from women who testified that mammography saved their life.  Surely there are such instances, but given the vagaries of tumor type, stage at which it's found, speed of growth, response to chemotherapy and so forth, which they are is hard, if not impossible to determine.  And certainly impossible to predict.

Why? Risk and decisions about risk

Breast cancer isn't even that rare, unfortunately, but there is apparently still a lot that isn't clear about best practice when it comes to diagnosis and treatment.  We have earlier posted about stories related to whether mammograms are worth the risk--that the radiation will induce too many cancers relative to what they detect, or detect tumors that would mainly regress on their own.  A positive mammogram leads to some sort of follow up, and this has its own risks and morbidities.

A story about a new study reports conclusions that more extensive biopsies ('open biopsies') are being done far too often, rather than less costly and less traumatic 'needle' biopsies.  If a tumor is detected, usually surgery is required, but in the former case this means two surgeries, and the story says this is considerably more difficult than a needle biopsy and one surgery.

There was a recent related story saying that lymph node biopsies or removal (in the armpit area through which breast cancer often metastasizes, when it spreads) were not worth doing, as judged by subsequent course of the disease.  And another story claims, at least, the discovery of another breast-cancer related gene--another type of test which, depending on risk estimates, will then lead to further decisions about further tests or treatment.

We know that when an absolute risk is very rare, and must be assessed by aggregate results of very large numbers of instances, it is difficult to make much less evaluate policy.  In the case of radiation, we can estimate the per dose-rate effect of high doses, but must extrapolate the dose-response curve to make a guess at what the low-dose risk, if any, might be.  This is the case with mammograms and even more so with exposures to radiation workers, dental x-rays, CT scans, and the like.

The same issues arise in GWAS or efforts to detect natural selection at the gene level.  Very small effects are difficult to detect, evaluate, or prove.  We usually do so with statistical significance criteria, but often even large samples are not adequate because too many sources of variation impair the ability to convincingly detect the effect.  Things that are real but small can go undetected and tests for them are vulnerable to interpreting fluke positives true positives.

These are challenging issues for science, because we're very good at picking up strong signals, that behave well relative to statistical evaluation criteria (like significance testing.  That ability itself may lure us to try--and expect--to be successful with very weak effects if can but collect huge enough samples.  At present, it's not working very well: at least, reaching consensus is not easy.

And the problems apply even to breast cancer which, in this context, is not even that rare.

Repeat after us: Correlation is not causation, correlation is not causation, correlation is not....

Confounding is probably the single most important explanation for irreproducible and even nonsensical results in epidemiology--and probably in genetics and evolutionary reconstructions as well. In the mid 1900's, for example, before everyone had telephones, researchers found that having a phone was a risk factor for breast cancer. How could that be? As it turned out, having a phone was associated with middle or upper class status, and money was associated with a diet that included increasingly more fat--or with increased age at first birth, or fewer children, and so on--risk factors for breast cancer subsequently identified by many studies that noted increased risk as socioeconomic status rose. Confounding is notoriously difficult to control, largely because many associations can't be anticipated in advance of the design of a study. (Whether the same-sounding argument applies to the idea that use of cell phones 'causes' brain cancer is not known.)

The BBC is reporting that meat eating causes early menarche. Or rather, eating a lot of meat. This according to a paper in Public Health Nutrition (though this is only the latest of the papers reporting this correlation). The age at menarche--first menstrual period--dropped throughout the 20th century and many people have wondered why. It had been thought that this was due to increased nutrition in general, but arguing against this idea is the observation that, as obesity rates increased, age at menarche didn't further decrease. That is, it's apparently not a simple matter of body size or nutritional intake.

The question of what has caused early periods prompted a group of researchers in Britain to look prospectively at a cohort of 3000 girls, including nutritional intake at age 3, 7 and 10. They identified a group of girls at birth in 1991 or 1992, and followed them up for about 13 years both by questionnaire and clinically. Early menarche was considered 12 years 8 months or younger, experienced by about half of the sample.

Girls with 'high' meat intake were eating 8 or more portions of meat per week at age 3 and 12 or more portions at age 7. Since early menarche has been associated with increased risk of breast cancer (the odds ratio is 1.5 - 2 times higher for women who started having periods at or before age 12 vs. women who started at 15 or older), the authors of the paper note that early menarche should be of concern. (But remember that 1.5 - 2 times a fairly low risk is not that high, and that, anyway, the comparison group, women who reached menarche at 15 or older has been a very small fraction of women for decades, at least in the developed world, and they perhaps were late for reasons that also protect against breast cancer, or at least may have different hormonal profiles. So whether this increased odds ratio is meaningful is up to you to decide.)In this large group of contemporary girls we have found a number of associations between dietary intakes throughout childhood and the occurrence of menarche. We have confirmed previous findings of higher energy intakes among girls reaching menarche earlier, reflecting their larger body size. We have also found evidence that intakes of meat and total and animal protein, and also possibly PUFA [polyunsaturated fatty acids] in early to mid-childhood may increase the chances of menarche by 12 years 8 months. However, we found no evidence that the chances of reaching menarche increased with higher total fat intakes, or reduced with higher intakes of fruit, vegetables or NSP. Unexpectedly, higher vegetable intakes at 3 years were associated with an increased chance of reaching menarche, although this may have reflected the positive association between meat and vegetable intakes at 3 years.

But could the meat/menarche association be due to confounders, variables that are associated with meat consumption and are the true explanation for the correlation? Things like ethnicity, socioeconomic status, mother's behavior during pregnancy, and so on, which would influence diet? The researchers controlled for some of these, but ethnicity was classified crudely as white/non-white, for example, and in the UK, as elsewhere, non-white can cover a lot of different diets, so that something else that rides along with meat could be the explanation instead. Kind of cooking oil, for example--we have absolutely no evidence that this is the case, but the point is that it's possible. Not to mention that high meat consumption is often associated with increased consumption of processed foods in general, which was not controlled for.  That is, high meat consumption could indicate a different diet, with some unidentified causative variable, compared with the diet of kids who eat less meat.

Another possible confounding variable that has long been discussed in the literature--though as far as we can tell, the association is still only suggestive--is exogenous hormone levels in meat (a brief time out for a bow to the web, where you can find just about anything: here, for example, is a link to the "Museum of Menstruation and Women's Health", which we stumbled upon in searching for papers on meat hormones and age at menarche). Animals, of course, are given a panorama of hormones to increase the speed at which they grow. Whether these hormones are found in excess in meat, or are active in our bodies when we eat that meat isn't clear, at least to us. But it's certainly a possible alternative explanation for the association between meat consumption and age at menarche.

So, should we stop feeding meat to little girls? Early age at menarche, whatever its cause, may be a risk factor for breast cancer (any increased risk is not to be taken lightly, though see above), but it has also been found to protect against osteoporosis--both presumed to be due to the increased length of exposure to estrogen. And, of course age at menarche is by no means the only factor associated with risk of breast cancer or osteoporosis. And the protein and other ingredients in meat are likely good for other aspects of growth and health.

Of course epidemiologists (and journal editors) have to eat, and to do that they have to publish lots of studies!  Does this study tell us anything of significance? You decide.

Mammography: Grim tales of real life.

The use of x-rays to detect breast cancer, known as mammography, started around 1960. The idea was that x-rays could give an in-depth picture of the breast that would be superior to palpation for detecting small tumors that had not yet become obvious or symptomatic. It seemed like a very good idea that could save many lives, and became not just widespread but formally recommended as part of preventive care.

This was based on the belief, or perhaps even dogma, that tumors are 'transformed' cells that are out of control and will continue dividing without the normal context-specific restraint. The tumors induced vascularization that nourished its cells, and eventually cells flake off into the blood or lymph systems, to be carried along to other sites where they would eventually lodge, spreading the tumor (this is called metastasis). If anything, treatment or just competition would lead this distributing clone of transformed cells to gain an increasing evolutionary advantage over the woman's (and, in much rarer instances, men's) normal tissue: tumor cells would continue to accumulate mutations at the regular or even an accelerated rate, that would give them even further growth advantage.

Sometimes tumors seemed to regress, but this was difficult to explain and often it was thought that perhaps the initial diagnosis was wrong. If the tumor had escaped immune destruction when it was only a single or few cells large, what could then later make it regress?

Thus the general dogma in cancer biology that the earlier it was caught, the less likely it would spread. That also meant the earlier in life one was screened, the better. Local surgery could then cure the disease.

But there was a problem: the same x-radiation used to detect different cell densities between tumor and normal tissue, is also a very well-known mutagen and cause of cancer!

Worse, the more actively dividing cells were, the more liable to mutation and thus transmission to increased numbers of a descendant line of daughter cells in the tissue. Since breast tissue grows every menstrual cycle, pre-menopausal women would be particularly vulnerable to iatric carcinogenesis. Yet the idea was that earlier screening was better!

Even further, early onset cases are more likely to be or to become bilateral (both breasts) or multiclonal (more independent tumors), and it was suspected and is now known that some of this, at least, is due to inherited susceptibility mutations (in BRCA1 and BRCA1 and a few other genes). These mutations put a woman at very high risk, so earlier and more frequent screening--but higher total radiation doses!--could be important.

Especially after the atomic bombing of Japan in World War II, and the subsequent fallout from nuclear reactors and bomb tests, and the proliferation of diagnostic x-rays, many extensive studies were done to document the risk, and for example chest x-rays used in routine tuberculosis screening were shown to be a risk for cancers including breast cancer.

So, to screen or not to screen? The obvious answer to this Hobson's choice was a grim cost-benefit analysis: how many cancers are detected and cured vs those that are caused by mammographic screening? Even grimmer, this could be evaluated by age, so that recommendations could be made based on a judgment as to how favorable the age-specific balance between cause and cure was. And there's more: radiation-induced carcinomas take years to develop before they would appear as clinically detectable tumors, so evaluating and attributing risk was (and is) not easy.

Breast cancer is unfortunately quite common, but the differences being considered, among many additional variables known and unknown, are small. That means very large, long-term studies needed even to come to a tentative rational ('evidence-based') conclusion. The result was recommendations of occasional mammograms for women in their 40's, with more frequent screens in 50's and beyond.

This made sense....until a few studies recently began to appear with curious results. Several studies showed that the number of cancers in women not screened was lower than those in women who had been screened. How can this be? The answer appears to be that screening leads to detection, reporting, and treatment of tumors that would eventually disappear on their own. So screening led to interventions of various types, some rather grim in themselves, in a substantial fraction of cases that would go away without any treatment with its associated cost and trauma.

The same has been found recently in PSA testing of men for prostate cancer, so it's not a fluke of the study design. Scars of remitted tumors have been found, showing clearly that they regressed without diagnosis or treatment.

So now a panel of experts has recommended backing off, and doing screening less often (except in those who, in a grim kind of good luck, know they carry a high-risk mutation and hence need to be checked carefully, and often, where early detection can more clearly be effective).

Now if that isn't 'evidence' what is? Yet this is controversial, because it goes against accepted practice. In the Wednesday NY Times it's reported that some physicians don't plan to change their recommendations (what will insurance companies, our most noble citizens, and the entities that will actually drive this decision, do?). The NIH Secretary also backed away from this new report. This is curious to say the least and relevant, of course, to the notion of 'evidence based' medicine that we discussed in a recent post, and why we think the notion of evidence is actually rather slippery.

This strikes close to home for many of us, who have very close relatives who have died of breast cancer. For us, research on this subject could hardly be more important. If you're a young woman you face these grim or even terrifying choices. But in real life, rather than fairy stories, there's no easy answer.