UAB is wonderful place to be a postdoc, or at least it is in my experience. I am searching for a home for post-postdoc life and one of my questions to potential institutions (yes! I have had 2 interviews… fingers crossed for that all important second interview) is always a variation on: “UAB is wonderfully supportive to new investigators… while they may not be an Ivy League University they go out of their way to give the resources, and the intellectual stimulation to early career scientists to match the opportunities at those places. What efforts do you make to do the same?”. Along the lines of the intellectual stimulation, UAB (or at least the Departments of Biostatistics and Epidemiology and the Nutrition Obesity Research Center) go out of their way to get diverse high-quality speakers to give seminars to the students. It is a fantastic way to meet other researchers, hear about new research and diversify your thinking. Below are my thoughts on one such talk; I often share thoughts on NORC weekly seminars, but this one was from the Epidemiology Department. The NORC-related researchers also get emailed a round up of new research that “stands out” (yes, an entirely subjective description). And I have permission to share my view on the subjectively subjective “what stands out from that which stood out”.
I think all the sugar from licking my fingers during pie baking has addled my brain.
So – this weeks’ research roundup.
Tiffin & Arnoult’s data showed that their stimulation of a tax on food containing saturated fat combined with a subsidy on fruit and vegetables increased fruit and vegetable intake to be in line with dietary recommendations, but did not sufficiently reduce saturated fat intake. My thoughts? Yes, I have a number of problems with this. (1) This is a simulation study and we have no idea how this would translate to real-world intakes and (2) what was going on in the rest of their diet? to name a few. But my real problem, and this perhaps reflects my current bête noir, is that they write in the results section “Once the changes in diet are converted into changes in the risks of disease, the impacts of the policy are negligible. A substantial part of the population continues to consume an unhealthy diet.” So far, so good. But they conclude “Fiscally based interventions should be considered amongst a suite of policy interventions”. As far as I can see – this is not substantiated by their data! Their data suggests that fiscally based interventions, if we wish to reduce diet-mediated disease, should be abandoned for the development of more effective strategies. It is starting to annoy me: people research popular dogmas (sports at school reduces obesity / access to sugar sweetened beverages increases obesity) find that along with a host of other data these dogmas are instantiated and refuted, but are too scared to conclude that and say “well… we should still have more sports in school, and no sugar sweetened drinks, but we need to look at other things too”. Why not be bold and say “if we want to reduce obesity increasing sports and reducing access to sugary drinks doesn’t work (according to our data) and so we need to examine other avenues.” You’ll see later that this was my problem with the Epidemiology talk.
Thank goodness for Taber et al. They compared consumption of sugary drinks between states that (1) banned soda at school (2) banned sugary drinks at school and (3) had no state policy. They found actual purchasing overall (so including time outside school) was hardly different between the groups, and consumption not at all. No jiggery-pokery, they say “State policies that ban all SSBs [sugar sweetened beverages] in middle schools appear to reduce in-school access and purchasing of SSBs but do not reduce overall consumption”. Of course, this was a snap shot, and I wonder what effect a wider ban or societal view about sugary drinks would have. I don’t drink SSBs (although I grew up on “full-fat” Coca-Cola, and sugary squash) and tend to think children should see that SSBs should be treated with caution. However, this study suggests that banning them in schools has no effect – longer term, and household bans? That we don’t know…
We know socioeconomic status (SES) is related to BMI, but is it causally related? Fontaine et al argue, in part, the effects of shared genes that are confounded with SES (i.e. you share your SES and genes with your parents) and direct environmental transmission equally contribute to the association, using a particularly strong methodology (adoption study). So SES may be partially causally related to BMI, but not fully as genes play a role (according to the authors). I got very lost in the methods here to really make any useful comment, but I am tentatively not sure I agree that their conclusions are supported by the data. For example the first paragraph of the results writes that adoptee BMI was equally correlated with biological as adoptive parents’ SES and this
” This supports the notion of a possible causal effect of the SES of the rearing environment”. because ” the correlation between the genetic influence on BMI and its influence on SES each accounted for roughly half the correlation between rearing environment SES and BMI”
So Fontaine et al find (1)a correlation between biological parent SES and adopted child BMI and (2) a correlation between adopted parent SES and adopted child BMI, that is 50% mediated by parent BMI.
Now – this took a break from blogging (hence the delay in post), a 15K run and a post Gossip-Girl Eureka moment to ‘get’. The theory goes as follows: if biological parental SES and child’s BMI are correlated, then SES must cause BMI. The obvious criticism is ‘what if there is some 3rd factor, C, which causes both SES and BMI – the two would be related but not causally’. The most common example is genes – what if genes cause the association between them? So Fontaine et al (partially) control for genes (by controlling for parental BMI) and show that the association still exists (although to a lesser degree).
But… we still have not accounted for non genetic factors in the association. Perhaps there is an environmental cause of both BMI and SES… yet then we would expect a correlation between the adoptive parent’s BMI and child’s BMI (as they are subject to the same environmental forces)- which there is not. So.. it is a pretty neat design. However, there are, of course, confounders:
e.g. SES could be shared between the parents, and may be primarily driven by one parent’s characteristics [genetics] (e.g. the father’s) and Fontaine et al control for for only one parent’s BMI. It is conceivable they only control for the other’s (i.e. the mother’s) meaning that the genetic component is not controlled for in the control for BMI. Now there is evidence of assortative mating for BMI indicating that it could be partially controlled for, but perhaps that is why we see an attenuation of the correlation, not a disappearance.
I don’t know. I am still wrapping my head around the intricacies of the underlying methodology. Until then, I consider it ‘supportive evidence’ but no randomized controlled trial (which of course is totally possible with human children. Ahem).
So – an interesting week for research. Not such an interesting week for talks. I was (ashamedly) harassed and distracted with this week’s NORC talk (love that rhyme) but here is a roundup of
“The Social environment and childhood obesity”
This week’s Epi seminar
The speaker introduced plausible social determinants of childhood obesity; how exclusion, discrimination and low self esteem: could be possible psychosocial influences on childhood health. OK, all in agreement, right? And so she said she was going to discuss the home environment and family structure, school policies on sugar sweetened beverages and friendship.
Wait. Did I go to sleep here? Social circles? Maybe – friends less likely. But family structure and sugar sweetened beverages – as the follow-up to exclusion, discrimination and low self esteem? Am I missing something? Ever been bullied for drinking sweetened juice? Ever felt really low about yourself because your school offered you Cola not milk? It’s not that the speaker directly linked the two, but she did seamlessly segue from one to another, with that tacit, implied conclusion that might register at the back of your mind if you weren’t reading over notes to write a blog post. It is what Baroness Susan Greenfield did, casually mentioning the twin rises of facebook and autism – never directly give a causal link (because you can’t) but bring it all up in a tut tut way and move quickly on.
Shoddy at best (as I suspect was the speaker’s case), unethical at worst (Greenfield).
So, our speaker introduced her datasets (which were great): 2 large longitudinal national data sets :
1. the Early childhood longitudinal study; kindergarten cohort (ECLS-K) -and-
2. the National longitudinal study of adolescent health (add health)
She boldly stated that early childhood family environment was important to study, because [and this is a quote] “our preferences [the affect our adult BMI] are learned early on”. I didn’t get a chance to ask why she thought this against the wealth of contrary evidence. But… moving on.
The speaker examined: “What family activities matter for children’s BMI and BMI change?” noting, observationally, that family sports is associated with a lower BMI. Note: not, ‘children who participate in their family’s sporting activites have a lower BMI that those who do not’, but [my phrasing]: families who tend to be active have children with lower BMIs and smaller weight gain. So – no control for the methodological loophole I discussed above: no control for the third, latent, causal variable. Like, I don’t know: genetics? Or even, SES 😉
She discussed the role of family structure in a child’s BMI, noting that the presence of a grandma was associated with obesity. She didn’t mention if the presence of a grandma was associated with any other variables that might be affected BMI. Oh, I sound like a broken down record, I know, but the time has come to stop accepting nice right-sounding conclusions, and to start collecting hard evidence. But, do you really think, if I transplanted an older lady into families, everyone’s BMI would shoot up? Really?
She reported associations with family breakfast, tv watching and free playing and BMI but not if these “risk factors” have the same associations in different family structures; i.e. if there was any control for that which she was studying.
But, sigh, if you are interested: children living with both parents most likely to be obese and the presence of other adults is associated with weight and weight gain. So, if you are worried about your child’s BMI get divorced, live on your own, but make sure you don’t lower your SES in the process as research suggests that is linked to a higher BMI. Sigh.
Moving on to school nutrition policies. She acknowledged that sugar sweetened beverages are considered unhealthy and consumption is on the increase, alongside obesity, yet the evidence for an association between sugar sweetened beverages and BMI is at best very weak and, more accurately: inconsistent
So, she compared school that banned sugar sweetened beverages with those that did not, and found a very low relationship relationship cross sectionally with self-reported sugar sweetened beverage consumption (presumably they are being brought in / purchased without the school knowing, drunk outside of school) , and a null longitudinal relationship with sugar sweetened beverage access and BMI.
Good stuff. Slightly weak, but OK data. Her conclusion: “There is no effect of sugar sweetened beverage access to consumption of BMI, but access to them should still be limited.” The mind boggles.
We moved on to whether our friends affect our weight? An whether there are influences from
– friends’ weight status
– friends’ weight related behavior
– friends’ communication about weight and weight related behaviors
She found only weak association, which was interesting as often there is a stronger association found between dining companions BMI and their food behaviours (although it is a complex relationship). Differences between two friends’ BMI increased across time as well, suggesting even less of an effect. The speaker said up front that she was not going to “bore” us with her methodology, which was slightly odd if she wanted us to accept her conclusions as the scientists that we are. So, it was not clear if she tracking the relationship of BMI between friend A and friend B across time, or friend A and friend A’s new friends (B,C and D) and finding that the relationship got weaker. She alluded to ‘making your own family’ but I am not sure that this helped me understand.
However, if we do want to predict this relationship, it would be the self-reported characteristics of respondent, and the friend-reported characteristics of the friend, not the quality of the friendship. Or something like that.
What I learned from this, is the vital need to move away from conjecture, dogma and poor inference. What obesity research really needs is strong methodologies, the conclusions of which from several designs can be used to address the the other designs methodological weaknesses, a lack of going in to ‘prove’ a hunch, and a brave soul to stand up and make new conclusions. We are not solving the obesity epidemic going down our current roads: someone needs to forge new ones.