Tuesday, December 11, 2007

Active parents raise active children: the authors reply

Recently, I wrote a rather scathing blog entry (see also my follow-up post) and a more measured "rapid response" to a BMJ (British Medical Journal) article that argued parents socialize their children into being physically either more active or more sedentary. Now the authors have replied to my criticism with a rapid response of their own. (Actually, the response was published almost a week ago - oh the dangers of going on holiday!). Suffice it to say I don't think Mattocks et. al. have neither met my criticism or understood it fully, so I'm going to respond to them both here and at the BMJ website. (Submissions to BMJ's rapid response system aren't peer-reviewed but are approved by an editor, so there's no guarantee my response will in fact be published. I'll link to my response if it does get published).

A quick recap of the study: the authors (Mattocks et. al., 2007) wanted to determine which early life variables (before age 5) affected the physical activity of children at ages 11-12 (which was objectively measured with accelerometers). They found that (among other things) "children are slightly more active if their parents are active early in the child's life and suggest "helping parents to increase their physical activity... may promote children's activity." It's clear from the quotations (and the rest of the study) that the authors were thinking solely in terms of socialization. As I pointed out in my critical blog entry (which I recommend you read if you haven't done so), there's an obvious possible confound here: genetics. Since (1) all behavioral traits are heritable (Turkheimer, 2000) and (2) children inherit 50% of their genes from each parent, "genetic factors are always possible confounds when relating parenting style (or other parental behavior) to outcomes in children". Consequently, because genetic factors were not controlled for, Mattock's et. al.'s study does not distinguish between the relevant possible causal hypotheses and therefore fails to add very much to our knowledge.

The authors' reply to my criticism, I think, amounts to the following: 'controlling for genetic factors is really hard' (first paragraph), and (2) 'not all the hypotheses we tested are confounded by genetic factors' (most of the second paragraph). Let's take these replies in turn. The literature about the determinants of physical activity is certainly not one of my specialities so I don't have an opinion about whether or not it's possible to control for genetics at the present time. Let's grant, for argument's sake, that Mattocks et. al. are correct: that it's not possible to control for genetic factors. What follows? Does assuming this proposition at all support the truth of their finding concerning the link between active parents and active children? Clearly not - the fact that we cannot control for a confound manifestly (and unfortunately) does not make it causally inert. Because genetics is a possible confound (something Mattocks et. al. do not dispute in their reply), we simply can't draw a conclusion one way or another because, to repeat, their data fails to distinguish between the relevant alternative causal hypotheses. I note furthermore that Mattocks et. al. seem to have been intellectually careless - they don't seem to have considered genetic factors as possible confounds when designing the study at all. Firstly, electronic text-searches confirm my impression from reading their paper: they fail even to mention genetics. Secondly and damningly, their paper does contain a section entitled "Possible confounders" but it does not list genetics as a possible instance.

Mattocks et. al. are correct in saying that not all the variables they considered are possibly confounded by genetics. I focused on a subset of their variables and findings because I found the study through this ScienceDaily article (title: "Active Parents Raise Active Children") and because the authors themselves emphasize the correlation (what they regard as causation) between maternal physical activity during pregnancy and early life and children's physical activity later in life (see the conclusion of the abstract). Furthermore, it seems a majority of the study's positive findings are at least possibly confounded by genetics, even if in some cases a particularly plausible causal mechanism is absent. (The sum total of the positive findings were that activity at 11-12 was 'modestly associated with': "mother's body mass index before pregnancy, parents' smoking status during pregnancy, mother's age at birth of the child, mother's physical activity, parity, and season of birth.")

In short, genetics is a possible confound (a fact that remains unaltered whether or not it's possible to control for it), but, despite this, the authors didn't even try to control for it and happily drew causal conclusions in the absence of controls. My criticisms stand.

Bibliography

Mattocks, C., Ness, A., Deere, K., Tilling, K., Leary, S., Blair, S.N., Riddoch, C. (2007). Early life determinants of physical activity in 11 to 12 year olds: cohort study. BMJ (British Medical Journal). DOI: 10.1136/bmj.39385.443565.BE

Turkheimer, E. (2000) "Three Laws of Behavior Genetics and What They Mean," Current Directions in Psychological Science, 9(5): 160-164.

No comments:

Post a Comment