Thursday, November 29, 2007

Follow-up: Active Parents Raise Active Children

Alas, I was right in my condemnation of Mattocks et. al.'s study into the causes of children's degree of physical activity: genetics was not controlled for at all and, in fact, no attempt was made to do so. Shorty after writing my blog entry on the study, I emailed the lead author, Culum Mattocks, pointing out that genetics was a possible confound and asking him whether I somehow missed how they controlled for it. He replied:
Thanks for your interest in our article. We were unable to control for genetic factors in this study. We do intend to look at genetic influences on physical activity in the future in our study but were not able to do so at this time as that will need analysis of DNA samples.
This, frankly, is simply not good enough. As I pointed out to Mattocks in reply, the point of science is to understand the causes of observed phenomena - and we come to such an understanding by subjecting our causal hypotheses about what's going on to empirical tests that can distinguish between alternative theses. In its current form, their study tells us, basically, "active parents raise active children OR active parents have active children OR some combination of the two OR some other factors are at play". It tells us, in short, nothing we didn't know before Mattocks and his co-authors spent their large grant on fancy accelerometers and other paraphernalia. This is absolutely criminal in my opinion: Mattocks et. al. squandered valuable scientific resources, took up the time of the Avon cohort, missed an opportunity to find out something of value about an important topic, misled the public and, worse of all, engaged in bad science and sloppy thinking. At least BMJ published my "rapid response" to the paper on their website (basically a precis of my blog entry) , it can be found here: "A possible confound: genetics - Michael Meadon".

A couple of other brief observations. Firstly, one most certainly does not "need analysis of DNA samples" to control for genetics, as Mattocks suggests in his email. (Besides, that seems to be already available). My colleague (and supervisor) David Spurrett summed up the reason rather aptly: "you don't need DNA samples to tell how related children are to their parents." What you would need to control for genetics in such a study is some sort of intervention (asking parents to be more active than they normally would be around their young children) or a twin study. Secondly, and finally, one may wonder how Mattocks et. al.'s paper got past peer review and onto the pages of the BMJ in the first place. Perhaps the fact that BMJ published this study sheds some light on the matter.


  1. And, as you well know, there's quite a bit of evidence of activity levels being heritable in specific ways in other species, including mice and flies. So Baik and colleagues produced D2R (Dopamine receptor) knockout mice that moved dramatically less than unmodified individuals. Kelly and colleagues produced different D2R knockout mice, that initiated dramatically fewer bouts of movement in unfamiliar environments, a difference accounting for most of their reduced overall movement.

    Baik J H, Picetti R, Saiardi A, Thiriet G, Dierich A, Depaulis A, Le Meur M, Borrelli E. Nature. 1995; 377: 424–428.

    Kelly M A, Rubinstein M, Asa S L, Zhang G, Saez C, Bunzow J R, Allen R G, Hnasko R, Ben-Jonathan N, Grandy D K, Low M J. Neuron. 1997; 19: 103–113.

  2. Thanks for that David. The animal models - and indeed everything else we know about behavioral genetics - makes it likely that activeness is heritable. Of course, the mere fact that it's plausible that activeness is heritable requires scientists to take genes into account as a possible confound.

  3. You must be somewhat naive if you think that "asking parents to be more active than they normally would be around their young children" is a practicable intervention, or you don't recognise an article from the BMJ Christmas issue.

  4. Sorry if I was unclear. I didn't mean to suggest that "asking parents to be more active than they normally would be around their young children" is a practicable experimental modality - I was simply thinking through the methodological logic to come up with ways in which genetics could, logically, be controlled for. Clearly, somehow incentivizing parents to be more active than they normally would be fits the methodological bill, althoug it's certainly impractical, as you point out.

    Thanks for correcting me on the Christmas issue; I didn't spot that at all. I'll strike that bit of my entry.