Monthly Archives: January 2011

Unhealthful News 21 – would you like some cereal with that milk?

On the occasion of the U.S. Surgeon General announcing a “call to action” to encourage breastfeeding, which I am sure will be every bit the stunning public health victory that was her call to action against smoking, I thought I would follow up on the reporting about a new recommendation about breastfeeding that created quite the uproar.  I mentioned it in Unhealthful News 14 to make one specific point, that people who studied just enough epidemiology to be dangerous have a habit of overstating the value of artificial experiments (randomized trials) compared to other evidence.

Notwithstanding that epistemic error (and others noted below), the authors of the study actually wrote a fairly modest paper and almost kept their conclusions appropriately tentative.  They cannot be blamed for the news frenzy that followed the publication (presumably BMJ, the journal it was published in, or the authors’ university’s publicists deserve that blame, though it is possible that the authors become less accurate when talking to reporters – a lot of them do).  Basically the authors observed that, contrary to the official WHO advice, it does not appear that there is any strong argument for women in wealthy countries to exclusively breastfeed (no other food) their babies for 6 months rather than just 4.

The justification for the extra two months is mainly infection risk, which the authors of the new paper found appears to be a bit higher in rich countries, but not much.  This contrasts with poor countries where prolonged exclusive breastfeeding is the best defense against foodborne and waterbourne disease, or starvation for that matter.  WHO is notorious for making recommendations as if the world were a lot more homogeneous than it really is, and local health officials commit the even less forgivable error of not bothering to figure out when WHO is offering them bad advice.  So, the main bottom line was that British women who wanted not to exclusively breastfeed for the extra two months should not feel bad about that, despite Britain adopting the WHO recommendations.

That is about it.  If only the authors had simply said “There is no compelling evidence favoring 6 months over 4.” everyone would be a lot happier and better informed.

Indeed, the authors acknowledge that they are working with evidence that is inadequate for saying much more:

Yet infants exclusively breast fed for 6 months represent, globally, a small, potentially biased subgroup (for example, under 1% of UK infants in the 2005 UK Infant Feeding Survey [reference note]), that presumably excludes those perceived by their parents as signalling [sic] hunger and so requiring weaning foods earlier. Generalisation from this subgroup must therefore be questioned.

So they recognize that we do not really have much data on 6 versus 4 – enough to detect a big difference but not a small one.  Well, maybe they recognize that.  They do not say it, and what they actually say is wrong:  If many babies apparently need supplemental feeding before 6 months, or the mothers otherwise bail on exclusive breastfeeding despite the current recommendations, then this advice does not apply to them (other than to perhaps say “do not sweat it so much”).  The question is what to tell those who think that going the full 6 months – is it worth the extra effort, or is it even a good idea.

This is another case of researchers knowing just enough epidemiology to be dangerous.  They understood the concept that it is possible to have data from a subgroup that is unrepresentative of the whole population.  But apparently they only remember being told that this is a bad thing.  The missed the detail that this is a bad thing only if you want to extrapolate to the entire population.  If you actually want to study that subgroup, then it works out quite nicely.  In this case, it is obviously that subgroup that is of interest.  Mothers who were not going to exclusively breastfeed for 6 months anyway do not need to be advised to consider not doing so.

It never fails to amaze me how often epidemiologic analyses draw conclusions based on rules of thumb that clearly do not apply to the particular case.

As for affirmative reasons to introduce other foods before 6 months, the authors went on to suggest that there are a couple of reasons.  A single study that they found suggested that the extra two months have a huge effect on whether babies suffer from serious iron deficiency.  This result seemed rather incredible, an increase in the risk from about 2% to about 10%, particularly because there was no suggestion that this result had ever been replicated despite the fact that it is (a) huge, (b) important for people to know, assuming it is really true, and (c) relatively easy to check in any number of ways.  This strongly suggests that this is a classic case of “our study is the first to show…” calling for the conclusion “…and therefore it is almost certainly wrong.”  That study was based on mining a general health survey, which in itself is not necessarily a problem, but given how many dedicated breastfeeding studies there are, we would not expect something real to show up there that was missed everywhere else.  I cannot report much more about it, though, because I could not get a copy (who knew that something published in the journal “Breastfeeding Medicine” would be so difficult to get a copy of; after all, there must be dozens of libraries that subscribe to it – a perfect case of “publication” as score-keeping rather than actual communication of information).  Thus, it seems like a big mistake to have taken this claim seriously.

The second argument against longer exclusive breastfeeding was rather more convincing, being based on several studies.  Some research has found that failing to introduce gluten (grain protein) before 6 months increases the risk of celiac disease (an inability to eat gluten due to an autoimmune reaction), which is quite a problem given our grain-centric diets.  Assuming we take this caution seriously, the advice from the new study really should have been not just, “There is no compelling evidence favoring 6 months” but also “and it is probably a good idea to introduce some grain, if nothing else, before 6 months”.  A careful reading of the paper (if you skip over the faith in the one study that claims a huge risk of iron deficiency) shows that the authors basically report this.

So what went wrong?  Why did this cause headlines, fearful reports on the television news, and substantial consternation?  Because it was reported as if it were some radical change in advice.  That is, the naive WHO advice had been touted by naive British officials as if it were definitive and precise, and represented a huge variation in health risk, when actually it was tentative and as far as anyone could tell, whichever way was best was only slightly better.  And the vague suggestions from the new review were presented the same way.  (And why did American reporters make a big deal about it?  Apparently because they did not even stop to notice that the “new” advice conforms to what is typically advised in the US, notwithstanding the random assertions of the Surgeon General today.)  While it is gratifying that reporters and their readers have so much faith in epidemiology, it might be useful to realize it is much fuzzier around the edges than typically portrayed.

At least there is some good news.  Since we cannot be sure whether it is better for someone to believe the “old standard” or the “new advice”, and any difference is small, then whatever people took away from the news frenzy makes little difference.  The bad news is that while some reports touted this as imperative advice, a few recognized the fuzzy edges and did not go overboard with their conclusions, and most were in between, I do not recall seeing a single report that mentioned their one interesting conclusion: that introducing gluten in the fourth or fifth month dramatically lowers celiac disease risk. 

That, I think, was the fault of the researchers and their publicists.  Reporters often bury the lead, but if they had been told that this was the main important piece of advice, I suspect they would have run with it.  It is a perfect story:  “And coming up after the break, the simple food that could save your baby from a lifetime of misery. Stay tuned.”  It is sad to see the hype masters let us down so badly.

On the other hand, I am not sure how good the evidence is about celiac disease, but I intended to find out in the next 2-3 months and will probably find it blog worthy, so stay tuned.

Unhealthful News 20 – Just because someone tells you something does not mean it is true.

A recent report from of a Washington lobbyist/pundit company about taxes on tobacco products (see more about why I picked this topic and about that report below) was picked up in this news story last week.  The story began “A national study of state tobacco taxes, released Thursday, found that a lower rate on smokeless products could encourage users to switch from cigarettes.”  In the middle of the article, the reporter wrote, “Analysts say that most tobacco taxes are geared toward recouping the cost to society of health-care expenses related to its use. However, some state-level politicians want to fill budget shortfalls with a larger tax on those products.”  The trouble with both of these is that though someone was claiming them, they are wrong.

The second of these assertions is a common claim, but a bit of checking shows that it is not so.  Cigarette taxes have often been justified based on medical costs, though that traces back to a time when there was not even good evidence that the medical costs were positive (i.e., the increased costs of treating smoking caused diseases is offset by the decreased costs of not treating people who have died prematurely, and it took a while to get good measurements for both sides of that).  But the taxes (technical note: particularly if we include the “Master Settlement Agreement” payments which are widely portrayed as a liability award paid by cigarette companies but are actually a sales tax on smokers) are quite a bit higher than the net increase in medical costs.  Moreover, if we do not try to isolate medical care consumption as if it were special, but consider total effects on smokers’ lifetime consumption (i.e., we consider that people who die earlier consume nothing, healthcare or otherwise), smoking saves resource costs (or pension / Social Security payments if  you prefer to think about money rather than resources).  Obviously this does not imply that people dying from smoking is good, but merely that it is disingenuous to claim that taxes on cigarettes can be justified because smokers impose higher costs on the rest of the community.

Moreover, taxes on smokeless tobacco, the subject of the story, could not possibly be justified based on this because smokeless tobacco causes risks that are so low that they cannot even be measured.  And anti-tobacco activists who demand even higher taxes have largely stopped making the claim about them being some kind of fair user fee, instead focusing almost exclusively on how the taxes support their political goal of trying to punish people who choose not to quit using tobacco, whether it is risky or not.

(Oh, and sorry, but saying “Analysts say…” does not excuse a reporter from following with something that just ain’t so.  If you start with “Congressman Smith said…” then you are off the hook, but “Analysts say…” is just as much a way of saying “you, dear reader, should believe the following without question” as is just asserting the statement.)

As for the lead, the “study” that the story reports on did not, in fact, find that lower taxes on smokeless tobacco could reduce smoking by encouraging product switching, and therefore benefit public health.  The author of the study asserted that conclusion, and it is undoubtedly right (hundreds of others of us have said the same thing).  But it turns out nothing in the study actually supported the claim – it was pretty much disconnected from the analysis.  I wrote a few more details about that here if you are interested.  Indeed, the reason I wrote a brief post and about this topic today is that I used up all my blogging time for the day writing a long analysis of how tobacco harm reduction advocacy seems to be tending toward embracing bad analysis – anything that sounds like science and supports the cause – which I see as a serious threat to the cause (along with a couple of other forces I mention).

If you found your way to my writing because of my work on THR, you really should read that other piece.  If you are not interested, I will just leave you with today’s message (reporters should not just believe assertions about what analyses tell us; the rest of us should not assume that the reporters have taken that advice) and an assurance that, thanks to regression toward the mean, tomorrow’s post is likely to be more interesting.

Unhealthful News 19 – cost-benefit analyses that ignore the benefits

Cost-benefit analysis (CBA), if you are not familiar, describes a concept and set of methods used in policy analysis to compare – as you might guess – the benefits of a policy (or standard, or recommendation) and the costs.  It is a sensible way to look at things, for obvious reasons, though it has its problems (e.g., sometimes some of the costs or benefits are difficult to put in the same currency as the others).  But it really does not work well at all if you only look at the costs.

The most obvious example right now is the debate over health care financing in the U.S.  There is an argument between those who look at the actual science and conclude that the costs are actually negative (that is, it saves money, so the “cost” side is really a benefit) and the right-wingers who traffic in analytic nonsense to claim that this is not true.  But what is even more astonishing than what the latter group gets away with claiming about the costs is that they get away with ignoring the benefits.  That is, the punditocracy seems to think if they can show that there are net dollar costs then it must be a bad policy.  But, of course, the main benefit of the policy is that far more people get good affordable health care coverage (not me, but lots of people).  Obviously, even if it costs something to bring this about, it might be a good idea.

Another example is the habit of writing about only the costs of nicotine use, which leads to all kinds of absurd conclusions and analyses, as well as a bad outcome in terms of social welfare.  I have written about this many times, and Elaine Keller recently contributed an essay on the topic to our Tobacco Harm Reduction blog.

Lacking an exciting health news story this morning, I fall back on this story from last week, which I had a bit to say about, but was overshadowed by more exciting stuff.  The story involved a move by U.S. regulators to reduce the amount of acetaminophen (Tylenol) in some medicines that include it along with other drugs, such as Vicodin.  The problem is that acetaminophen has an unfortunately low ratio between a seriously harmful dose (in particular, that damages your liver) and the therapeutic dose that you want to take.  Among popular drugs (medicinal or recreational), nothing else has such a low ratio.  That means that accidentally taking more acetaminophen than you think you are taking can be quite dangerous, like when it is the second, less well publicized drug in a drug cocktail medicine and you do not realize you are even taking it, and so take more in some other form also (with “you” being especially older people or others that are being medicated for many conditions).

To try to do something about this problem, regulators have lowered the amount of acetaminophen in cocktail drugs.  Some commentators praised this, as did some news stories (most of which, in a bout of brilliant reporting, just asked the guys who were changing the regulations whether it was a good idea and – surprise! – they praised it).  A very few commentators questioned it.  Not one of them ever looked at the benefit side.  That is, not one commentator – regulator, researcher, or news reporter – ever addressed what the benefit of having acetaminophen-blended drug cocktails is. 

One possible assessment is that there is no benefit:  Acetaminophen is extremely cheap to acquire by itself, so there is no apparent reason for blending it with other drugs.  Instead, a physician who wanted to prescribe something like Vicodin could just say “take one of these (the new formulation of the old medicine, minus the acetaminophen) and at the same time take a 500 mg. acetaminophen pill”.  This would allow the additional instruction, “but if you are taking multiple such combinations that include acetaminophen, do not add up the acetaminophen doses! just take it once.”  This seems like an excellent solution to the problem that the new regulations are supposed to address.  It solves the problem much better than the halfway solution that is being employed, which is to reduce the doses but still leave the possibility that someone might make the same mistake they are currently making.

The current policy also has the potential problem that the new formulations will contain less than the optimal amount of acetaminophen, so if someone is taking just one of them, they will be underdosing acetaminophen.  Of course, this assumes that the current doses are reasonably close to optimal for the particular drug cocktails.  We might know, had anyone bothered to analyze the benefit side of having acetaminophen blended into these medicines.  If there is some good reason for the cocktails (perhaps there is evidence that people will not take their acetaminophen supplement if instructed as in the above example), and the current doses are right, then presumably the new policy represents some compromise between the best dose for someone who is taking just one cocktail and a limit on the dose for people who are combining multiple cocktails.  But we have no idea about that.  No one had any business praising this as a good policy without looking at the benefit side, and no reporter should have allowed the commentators to dictate the news stories without asking a few simple questions.

If this is a good way to regulate pharmaceuticals, then a good way to regulate cars would be to consider only on the goals of safety and fuel use.  It leads to a pretty silly analysis:  I saw a really cute little car at the toy store that uses no fuel and, barring a Jackass-style stunt involving a steep hill, cannot manage injury-producing speeds.  We would get a rather different recommendation if the regulators worried about such matters as transport (or pain management).

Though in the case of acetaminophen cocktails, I am inclined to believe that actually considering the benefits side would call for greater regulation (which is kind of odd, because usually a failure to consider the benefits side leads to too much regulation rather than too little).  Perhaps I am overlooking something, but I really cannot see any substantial benefit for anyone (other than pharma companies – at the expense of the rest of us – and lazy physicians) of continuing to allow acetaminophen cocktails instead of just telling people to take two pills.  In other words, what was portrayed as a wonderful step forward in solving a health problem was actually an obviously inadequate intervention that was the worst of all worlds, making it more difficult to get the right dosage and still leaving people vulnerable to taking too much acetaminophen without knowing it.  If only the regulators had thought to consider the “B” part of “CBA” we might have had a better policy.  And if any reporter or health policy commentator who wrote about it had thought the situation through and asked about the benefit side, he might have written an insightful story instead of misleading stenography.

[P.S. Helpful note to anyone ever writing a paper that uses a long and impossible to type word like “acetaminophen” quite this many times:  Program a macro or at least put the word in your cut-and-paste clipboard.  I think my brain would have exploded if I had not done that.]