Friday, September 19, 2014

Faith in science? Industrialized agriculture and antibiotic resistance

Someone asked me the other day on Twitter whether I thought that the words "science" and "belief" were compatible.  I said yes, though I know that a lot of scientists think (...believe...) that faith has nothing to do with science.  Science is facts, faith is religion, based on sacred texts and the like, which are basically hearsay without empirically acceptable evidence.  But, the history of science indicates that this distinction is far from being so simple -- there was a time when people believed that the moon was made of cheese, diseases were caused by bad air, Newton was right about physics, the continents didn't move.  And these beliefs were based on empirical evidence, observation -- dare I say 'facts'? -- not mere guesswork.

In that light, two recent pieces about the role of agriculture in the rise of antibiotic resistance are interesting.  The New York Times described a new study in the Journal of Occupational and Environmental Medicine ("Persistence of livestock-associated antibiotic-resistant Staphylococcus aureus among industrial hog operation workers in North Carolina over 14 days," Nadimpalli et al.)
that reports that workers at industrial hog farms can carry antibiotic-resistant bacteria, Staphylococcus aureus, in their nostrils for up to four days.
Twenty-two workers provided 327 samples. S. aureus carriage end points did not change with time away from work (mean 49 h; range greater than 0-96 h). Ten workers were persistent and six were intermittent carriers of livestock-associated S. aureus. Six workers were persistent and three intermittent carriers of livestock-associated multidrug-resistant S. aureus. One worker persistently carried livestock-associated methicillin-resistant S. aureus. Six workers were non-carriers of livestock-associated S. aureus. Eighty-two per cent of livestock-associated S. aureus demonstrated resistance to tetracycline. A majority of livestock-associated S. aureus isolates (n=169) were CC398 (68%) while 31% were CC9. No CC398 and one CC9 isolate was detected among scn-positive isolates.
As the NYT piece notes, eight-six percent of this sample of hog farm workers carried bacteria for at least 24 hours, compared with about one-third of the non-farm worker population.

This is a problem because the resistant variety of S. aureus, MRSA, has made its way into hospitals and is responsible for thousands of deaths.  Further, many people believe that industrial farming is the cause of much of the antibiotic resistance that is now becoming such a problem, because animals are fed antibiotics to speed their growth, and many of those antibiotics are used to treat human diseases.  Indeed, the majority of the antibiotics used in the industrialized world are given to animals.  When bacteria on the farm become resistant to antibiotics, as this study shows, they don't necessarily stay on the farm.  How they spread has been difficult to document, but might include consumption of contaminated meat, and Nadimpalli et al. report another pathway.

Hog farm; Wikipedia
Responding to the increase in antibiotic resistance that many believe industrial farming to be responsible for, the US Food and Drug Administration this year put a voluntary ban on the use of antibiotics for growth promotion. Critics saw this as a weak response to a very large problem, but pharmaceutical companies and some farmers say it will do what it is meant to do; reduce the use of antibiotics for non-medical purposes, and thus reduce the possible evolution of resistant bacteria that are harmful to humans. Of course one always has to ask the political question of who wields the power and influence over any sort of decision that may affect a particular industry.

But much of this is controversial. Is agricultural use of antibiotics in fact to blame for the problem, or is it overuse of antibiotics by the medical system?  Indeed, there's less of a problem in, say, Scandinavian countries where for decades physicians have prescribed antibiotics at a much lower rate than they have done in the US. Do resistant bacteria really spread in considerable numbers from farm to city?  This may be less controversial with the publication of the Nadimpalli et al. paper, but critics will say that the sample size was small and anyway, documenting a mechanism doesn't mean this is what has happened.

We all tend to pick and choose facts to support our convictions.  Indeed, if you look at how scientists, in any field, cling to their explanations, 'convictions' is perhaps a muted term for what is being clung to.  How we think about these questions may well reflect what we believe more generally about the food system, how or even whether animals should be farmed for meat, whether we patronize farmers'  markets rather than industrially produced food, and so forth rather than what we, or anyone, actually know about the causes of antibiotic resistance.  That is, our personal sociopolitical positions seem clearly be correlated with, if not strongly influencing, our scientific position.

Yesterday, an opinion piece by Iowa veterinarian and pig farmer Howard Hill appeared in our local paper, and in papers around the country.  Hill believes that farmers are being unfairly blamed for antibiotic resistance in humans.
...the claim that "70 to 80 percent of all antibiotics sold in the United States each year are used in livestock" is a straw man. More than a third of those drugs aren't used in human medicine, another third are not considered highly important to human medicine, and most of them aren't used for growth promotion. Critics also ignore the fact that there are a lot more cows, pigs and chickens than people. In 2011, for example, 30 million pounds of antibiotics were sold for use in more than 3 billion livestock and poultry, compared with 7 million pounds for 311 million people, meaning each person used nearly five times more antibiotics than were used in each food animal.
Is he making selective use of the data?  Yes, but isn't everyone who talks about this issue?  And does that make our assertions wrong?  Doesn't prior belief influence our understanding of what the data show?

While Rome burns
President Obama yesterday issued an executive order aimed at combating antibiotic resistance.  The order accepts that industrial agriculture may have a role in increasing resistance, but it adds little to the FDA order of several months ago:
The Food and Drug Administration (FDA) in HHS, in coordination with the Department of Agriculture (USDA), shall continue taking steps to eliminate the use of medically important classes of antibiotics for growth promotion purposes in food-producing animals.
Not many teeth here.  Years ago Europe took much the same approach, requiring that the use of antibiotics for growth promotion be reduced, but a lot of reclassification of antibiotic use for medical purposes followed, as many expected in the US following the FDA announcement last December, which we blogged about here,  and again with this Executive Order.

Again in Scandinavia, the use of antibiotics for growth promotion has been banned, beginning in Sweden in 1986, but farmers have not suffered.  According to a piece in the BCMJ in 2011:
In 1986, Sweden became the first country to regulate the withdrawal of antibiotics used in food animal production. By 2009, Swedish sales of antibiotics for use in agriculture were reduced from an average of 45 tons of active substance to 15 tons. Sweden was followed by Denmark, the United Kingdom, and the Netherlands. 
Danish swine and poultry production continued to flourish with gradual reductions of antibiotic use beginning in 1992 and continuing to 2008 (latest data). During this time, Danish farmers increased swine production by 47% while reducing antimicrobial use by 51%. As well, poultry production increased slightly while reducing antimicrobial use by 90%. Denmark remains one of the largest pork ex­porters in the world.
So, whether or not growth promoting antibiotic use in animals is a major cause of resistance is not really an issue, and we needn't even continue to have the discussion.  If there is any chance it is, why not ban it entirely?  Experience in Scandinavia suggests there won't be dire economic consequences -- unless you're a pharmaceutical company making antibiotics for animals.

Faith in science
We have often written here about the economic interests that drive the course of Big Science.  Can we have faith in science if there is considerable faith in science?  People are, after all, only human, and people of all faiths, including science, defend their faiths.  Further, it's often impossible to disentangle belief from vested interest.   If you've got a hammer, or a hammer to sell, everything looks like a nail.


Alex Stoddard said...

I posit we are all implicitly Bayesians in any reasoning when we admit (on going) observations to the discussion. Very loosely "faith" may be taken as the strength of our belief in a prior probability. It is also entirely reasonable to critique an adopted prior.

Many a revolution in science could be described in terms of one prior replacing another as that commonly believed by scientists.

There may be a lot of benefit to being more explicitly Bayesian in a lot of our scientific endeavors, perhaps especially when seeking scientific support and guidance for policy. At least it would force us to get the priors out on the table where they can be reviewed by all.

Ken Weiss said...

That makes sense, and if we had to make our priors explicit a lot of them might seem too generic (or, one might say 'faith-based' to be of much use at least relative to the claims of knowledge that are made. But it would perhaps show people more about their thinking.

Of course whether 'probability' is even the right way to think of one's predispositions (or, often, even of one's results) is to me a serious but deeper question.

Anne Buchanan said...

I like your description of revolutions in science, one prior replacing another. But if all priors are equal, in effect, this doesn't answer the question of how we know if we have found a 'truth', does it?

Ken Weiss said...

Reply to discussion

I think I'll draft a post, maybe for Monday, about scientific revolutions. On thinking about it, I don't think I agree after all that 'revolutions' are changes in priors; at least to me that that isn't their major feature. I think that most real 'revolutions' (or 'paradigm shifts') are something else, and I'll try to say what I'm thinking when there's more space--to see if it's cogent or not.

Even without toss-up priors, I agree that this doesn't seem to get us closer to knowing what 'truth' is, without the very strong prior, that this whole enterprise of science gets us there (and that there is one, objective truth). But that now gets us into philosophy....

Anonymous said...

"Someone asked me the other day on Twitter whether I thought that the words "science" and "belief" were compatible."

Let's see. In 1960s, scientists believed that all eukaryotic organelles evolved through slow modifications, when Lynn Margulis hypothesized that mitochondria and chloroplast were symbiotic bacteria. A decade later, DNA evidence showed that mitochondria and chloroplast have own chromosomes. Margulis also hypothesized that eukaryotic cilia came from outside.

"mitochondria originated inside cell through slow evolutionary process" - belief

"mitochondria came from outside" - strong hypothesis in 1970, based on microscopic data

"presence of genome show mitochondria were independent bacteria" - strong hypothesis, starting a new belief system

"cilia was another bacteria - (spirochete)" - belief/weak hypothesis

"earth is alive as in gaia" - belief/weak hypothesis

I think scientists can believe in anything, but that belief part should not be treated as science. Here is the real problem. When science turns into 'scientific consensus' in a democratic system, the beliefs of many naive people turn into scientific fact.


Ken Weiss said...

I think it is a mix. Religions often have a belief in what is taken to be an unchanging truth (a 'received wisdom', or sacred text). Science has theories in which sometimes a comparably strong belief rests, but it can be titrated by facts. There are possibilities of change (of course, religions also produce sects), but consensus is one of many criteria for what is accepted. The sociocultural aspect has to do to a great extent with funding, jobs, and so on. So there are differences, I think, but there are also similarities.

And we have no rigorous criteria for inference; instead we have a set of things we use in ways a consensus accepts; these can change.

Anne Buchanan said...


But, you're looking at the history of belief about mitochondria with the benefit of hindsight. We can dismiss earlier ideas as naive, and wrong, because we know better now. But what we don't know, and can't know, is whether what we now consider to be true will be overturned at a later date. It's possible that future scientists will consider us naive to have believed what we now believe. Knowledge about epigenetics has overturned some of what we thought we knew about gene expression, discovering miRNA expanded our knowledge of gene expression, etc. Increasing knowledge _has_ to mean that some of what we think is true now won't be true in the future. For me, that makes it difficult to disentangle 'belief' from 'knowledge.'

Alex Stoddard said...

In reply to Anne Buchanan,

I content all priors absolutely are not equal. That is why it is to the benefit of scientific reasoning to force there declaration and review.

It is true that I may arbitrarily change the conclusion arrived at from the identical set of data by an arbitrary choice of prior (either its "shape" or its "strength"). However when I have done so my prior is stated and open to evaluation and criticism.

What the prior is doing is encoding my initial belief along with how strong the evidence needs to be for me to change my belief.

Often in practice one may choose a very non-committal and weak prior but it ain't necessarily so. See

Ken Weiss said...

I think Anne was speaking figuratively about 'equal' in that one has no basis for preference of one over another (of course she may answer separately about this).

It's true that the Bayesian approach is designed to 'disequify' hypotheses. Here the question (I think) is what the basis is for establishing the alternatives as priors for which one can establish a credible value, or the extent to which new data can adjudicate among very different hypotheses. In well-behaved situations, like coin-flipping, it's relatively clear. And this has been true in many areas of science, certainly, and whole theories of causality.

We have a couple of posts upcoming on the relevance of this to scientific revolutions, but that's a somewhat different subject.

This post was not about the Bayesian debate, but about more elusive and harder to evaluate explanations, I think.

Anne Buchanan said...


I'm not schooled in Bayesian statistics, so here's a simple-minded question: why do we need statistics at all if we load the dice, so to speak, in favor of the answer we believe we should get?

Alex Stoddard said...


The point, as I see it, of Baysian statistics is that by adopting an explicit prior we gain a lot of utility for reasoning about how much evidence we need to change our beliefs. And we can treat the process as iterative.

A nefarious or utterly pig-headed bayesian can load the dice at the point of prior selection but that is explicit and open to critique. Anyone reviewing the conclusions can choose different (much more reasonable) priors and re-run an analysis to see in quantitative terms, what the data should lead them to conclude (or put another way - what the data should lead them to change about their beliefs.)

I have enjoyed thinking about this stuff in light of the work (and frank advocacy) of Bayesian reasoning by John Kruschke.


and the article

Ken Weiss said...

Reply to Alex
I'm no expert but in the course of a career in genetics and evolution of course have come across or collaborated with people using Bayesian approaches. I agree with your comment and tomorrow will venture (hopefully not too far off base) my own view of how this applies to the idea of 'revolutions'.

Earlier in my career, the big debate was between frequentists and likelihood advocates, and more recently Bayesian methods have gained fashion. Likelihood advocates, at the time at least, stressed that a likelihood ratio was just a comparison of two notions, not a test of truth.

What I know of Bayesian approaches comes (as I cite tomorrow) from Jaynes' book. There is also Pearl's book Causality, but to me what I say tomorrow is relevant to that point of view.

Basically, and I think we'd agree, if you have a well-posed question and its numerically evaluable and involves sampling etc., Bayes gives a good way to go--and no illusions about 'truth'.

Anne Buchanan said...

Thanks, Alex. I like the idea that the point is to make decisions about changing one's view. Or at least re-evaluating. And I like the idea of crowd sourcing truths.