Monday, March 26, 2012

When a Risk isn't Really Risky

Let's forget kids and talk red-meat and risk.  Look at these findings I weaned from Carl Bialik, the Numbers Guy at WSJ.

A recent study out of Harvard's Public Health found eating one extra serving of processed red meat a day increases a person's risk of death each year by 20%.

An extra helping of processed red meat a day means 2 more deaths per 1,000 people per year - or less than a .2% increased risk of death.  

Thankfully Carl translated it even further:
The results translate into a typical 40-year-old man in the study living to 79 instead of 80 if he chooses to add a hamburger a day to his diet. That extra year may not seem worth the change in diet to some red-meat fans. When Risk Is A Red-Meat Issue, Wall Street Journal, March 23.

Aha.  That's what the meat-lovers (or sanctimonious vegetarians, hand raised) wanna know.

How does it effect me!

If only all health writers would translate the results into what they mean for the average cog in the industrial-food complex.  

Unfortunately for most of us the media favors the first, more dramatic statements up top that reflect relative risk.

They tell us how one group compares to another, in this case, those who eat a smaller amount of red-meat versus an extra serving (not sure how much exactly, didn't read the study).  Researchers and public health wonks particularly enjoy relative risk because it gives a measure of the success (or failure) of one treatment, behavior, new drug, etc and if you're interested in the welfare of many people (i.e. entire populations) than you'd better consider this risk. The FDA clearly should care whether a new pharmaceutical has an effect and the only way to do this is to compare a group who didn't get the new drug with one that did. 

Of course relative risks appeal to the media as well given their more sensational nature. Twice as ill.  150% more cases of autism. Five times as likely to die.

Relative risk doesn't give us a sense of the absolute risk or what we commonly think of as risk. We normal folk generally think about overall risk:
  • The one in whatever risk of being killed on the way to school by the mother driving the enormous SUV while smoking, chatting on the smartphone and passing out goldfish.
  • How many children out of the total number of skiiers ever to visit the slopes plunge to their deaths riding the chair lift. 
If we're thinking beef jerky et al. the average wife might wonder how much sooner she'll be picking out a coffin for her spouse ordering the double pink-slime burger.  For those troubled by the beefy results, my husband reminds me the study didn't take into account other factors that might explain the hastened death of red-meat enthusiasts.  

Speaking of missing controls, suspect links and heightened drama... let's turn to the parenting realm. 

Take breastfeeding.  We often hear about relative risks in the media.  For instance, breastfed kids suffer 30% fewer ear infections, 50% fewer colds, half the number of behavioral problems.  Formula babies face twice the illnesses, three times the emotional distress, blah blah blah.  Less often do we get the absolute risk - how many ear infections or colds a kid may get if he isn't breast-fed vs breast-fed. 

I'm making up the stats but check the next claim in the news.  The exception seems to be studies finding links between breast-feeding and intelligence that rarely fail to describe how many IQ points a baby looses from slurping formula. 

Here's an article in the New York Times last year that reported on a study finding formula-fed babies given solid food before 4 months face an increased risk of obesity (see my post on the study).  Tara Parker-Pope, who writes the Well column and thus reports on plenty of health studies doesn't fail to deliver the relative risk:
...formula-fed babies who had been introduced to solid foods before 4 months of age were six times more likely to be obese by age 3. Timing of Baby Food Tied to Obesity Risk, New York Times, Well, February 2011.
Six times.  Six many toddlers are obese anyhow?
...about two-thirds of the mothers breast-fed their babies for at least four months, while the remaining mothers gave their babies formula. By the time the children were 3 years old, 75 of them, or 9 percent, were obese.

I appreciate the attempt at absolute risk but it's not really all the baseline info I need to make a truly informed feeding choice.  Looks like 9% of all the kids in the study were too heavy and that's helpful because it shows the overall risk of obesity is pretty low for all kids but breastfed kids might be less "obese" than formula-fed kids in general.  The risk for formula babies of eating real food before 4 months might only increase the risk of getting fat by a couple percentage points.  Aside from the near ridiculousness of calling babies and toddlers clinically obese (remember the obese baby study?) I take issue with the confusion.  It would be great to know how many breast-fed kids are fat and how many formula kids (both those who eat solids before and those after 4 months, respectively). 

Even a good attempt at clarity fails to totally clarify the risks.

I'd like to see every single study result translated into what it means for the average child.  So what's the average child's risk of autism/adhd/asthma/bullying and also the increased risk in plain language (IQ points, pounds, bouts of gastrointestinal upset, tantrums) given the (bad parenting/bad genes/processed meat). I suspect journalists don't report this practical information because it usually requires scouring the actual study or bugging the authors who don't always publish the info and sometimes might not even know it.   

Still, I believe we deserve it even if we're slogging through a glut of information aka the new media environment, The Age of Big Data, Web 2.0 or whatever some techy type with a large audience platform calls it.


Awesome Mom said...

With the formula study did they also have breastfed kids that got solids before 4 months of age? What was their obesity rate? To me the real thing that could possibly causing issues is giving a baby solids before they are even ready for them not if they are getting formula or not. It frustrates me how they try and make the studies more dramatic and scary than they really are.

Polly Palumbo, Ph.D. said...

Hey Awesome, sorry trouble with comments here, finally resolved! I think my main complaint about this study was that the researchers never asked what the kids were eating. I'll have to go back and check.

Sarah V. said...

Do you know, I've been trying to do precisely this? But it is a *mammoth* task, potentially requiring slogging through dozens and dozens of studies to translate figures into absolute risk and find the range of different absolute risks in different studies. So don't expect a post any time soon. But, yeah... working on it, and you've just given me further incentive to do so!

Polly Palumbo, Ph.D. said...

Wow, Sarah, that is a mounumental task! How have you narrowed down the field of studies? Are you enlisting the help of anyone, say, a tireless grad student or someone with a very good eyes?

Glad I could be of some help in the endeavor...carry on!

Dr Sarah said...

I figured I'd use the AHRQ report as my starting point, since their remit is to pick out the research that is 1. high quality with good adjustment for confounders and 2. based in the developed world. So, for the commoner things like otitis media or gastroenteritis, I look through the studies they reference as being of decent quality to find the cohort studies, then dig the actual risk factors out of there and do the subtraction to get to attributable risk. Oh, with some extra arithmetic along the way in my clumsy attempts to allow for the fact that most studies only give unadjusted figures for the actual risks.

For the rarer outcomes - diabetes, childhood leukaemia, SIDS - I just take the consensus on the odds ratio and then find the most recent figures I can on incidence to calculate the attributable risk. For example, I did already work out the leukaemia risk, which comes out at around 1 in 100 000. Diabetes and SIDS are going to be more complicated because of the other risk factors - with diabetes, I'll have to figure out how to work out separate figures for babies with a family history of DM and those without, as the risks will be rather different, and with SIDS I'll have to find some way of allowing for modifiable behaviours such as smoking and how/where baby is put down to sleep (which may be impossible to allow for because the data on what proportion of SIDS cases they account for is probably just not available).

Oh, and on top of this I'm trying to work out figures on how mixed feeding (with different proportions of breastfeeding) or breastfeeding for shorter or longer durations might affect risk, rather than the all-or-nothing breastfeed-or-formula-feed attitude that seems to be so prevalent in discussions of the subject.

Phew. I don't even know if all that made sense as typed, but I do have it moderately clear in my head. But you can see why I'm not going to be accomplishing that project any time soon.

Polly Palumbo, Ph.D. said...

Where are you publishing it? This sounds like a journal article somewhere.

So for leukemia it's 1 in 100,000 if you don't breastfeed at all? What if you do? Yeah SIDS will be hard. In fact didn't several gov agencies say they didn't think there was good evidence of a link?

Well, keep me posted!