Monday, March 31, 2014

Early childhood interventions and health in later life -- or not?

On the heels of our Friday post on the causes of obesity comes a paper in Science ("Early Childhood Investments Substantially Boost Adult Health," Campbell et al.), reported by the NYTimes, ScienceNOWand other venues, about the effects of early childhood intervention in poor children on the children's subsequent educational achievement and, relevant to our discussion of obesity and cardiovascular disease, their health in later life.

The study
In 1972 researchers in North Carolina began to follow two groups of children from poor families. The children were enrolled as infants (so that experience they already had would not confound the study) and were randomly assigned to one of two groups. One group was given full-time day care and most of their meals up to age 5, and then intensive assistance with what those children were learning at school until age 8, while the other group of children were given only baby formula until they were 18 months old.

Preschool class; Wikimedia

The question the investigators asked was whether early intervention would lead to improved learning abilities later in life, but in fact the effects were fairly immediate, with the groups diverging even by age 3. The groups have been followed up periodically since 1972.  Among other long-term effects are that those in the treatment group were 4 times more likely to have graduated from college.  The authors' bottom-line interpretation of this study is that the earlier the intervention the better, with pre-school effects being much more long-lasting than the effects of school-age intervention.

Now, it turns out that apparently the effects weren't only on cognitive abilities, but were more extensive in important ways.  As the Times reports,
Men in the treatment group, now mostly in their mid-30s, were less likely to develop hypertension than those in the control group. They also had significantly higher levels of so-called good cholesterol, and none had developed metabolic syndrome, the medical term for a group of risk factors that together substantially raise the chances for heart disease, diabetes and stroke. In contrast, a quarter of the men in the control group had the syndrome. 
As for women, those in the treated group were less likely to develop pre-hypertension or abdominal obesity, which tends to be a risk factor for heart problems. They also had healthier habits. They were significantly less likely to have started drinking before age 17, and more likely to be physically active and eat nutritious food, than the women in the control group.
In addition, treated males were more likely to have health insurance at age 30, and to receive medical care when they were ill, although this wasn't statistically associated with health outcomes.  As Campbell et al. write, the intervention group also included a nutritional and health care component.  Children were given healthy meals and a snack every day, and they also were given both well- and ill-child pediatric care, thus, the question of longterm health effects is not an unreasonable one.

Current notions
There has been a lot written in the last decade or so about the effects of uterine environment, early childhood nutritional status and so forth, on health in later life, although cause/effect relationships haven't been solidly established.  Perhaps it's epigenetics, the environmentally-triggered modification of DNA and thus gene expression that is hypothesized to be responsible for just about everything these days, or perhaps exposure to some environmental variable at a given early age triggers a cascade of responses, or perhaps we have no clue.  Studies of these kinds of relationships are in contrast with, in these days of DNA rapture, the much more common and fashionable argument that inherited genotypes are responsible for everything in life.  

Campbell et al. write that the "precise mechanisms" by which early childhood interventions in her study affected adult health "remain to be determined" but that much of the effect seems to be, in the authors' words, "mediated by" (that is, at least were correlated with) low body mass index when children, most significantly males, were 1 year of age.

If so, this would suggest that we know much of what we need to know about how to lower risk of heart disease, stroke, hypertension; feed an infant well, or by the time the child is a year old it's too late.  Forget genes, diet in middle age, cholesterol levels at age 60, sugar consumption, antibiotics, etc.  If obesity or metabolic syndrome or stroke are triggered by what happens in infancy, and similarly, can be prevented with early childhood interventions, looking for causation decades later is fruitless.

This study may well serve to encourage researchers to change the focus of their search for causal risk factors for cardiovascular disease.  At least, these might be the interpretations, if we could take these results seriously.

But can we?
There's a huge caveat.  The original study was of 111 children, 57 in the treatment group and 54 controls.  That's already a small sample.  And then by the time subjects were age 35, attrition had reduced the sample size considerably, until there were only 9 men in the control group who completed both the physical exam and the laboratory component of follow-up, 19 men in the treatment group, 22 women in the control group and 18 in the treatment group.  These are numbers that took a while to find, I might add: I had to dig them out of the supplementary material, though in fairness the main paper did mention that they had had to correct for small sample sizes, which is why I went hunting for the numbers. Indeed, the sample sizes are so small that it makes one wonder why this paper was accepted for publication in Science.  And was picked up by any news outlet at all, never mind many.

ScienceNOW writes:
It’s a grim fact of life in the United States: Children born into poor families are sicker and die earlier than their well-off counterparts, particularly from obesity-related diseases such as heart attack and stroke. Now, new data from a famous North Carolina study of early childhood education suggest that such disparities are not carved in stone. Children who grew up poor but participated in an intensive, 5-year day care program are significantly healthier in their mid-30s than similarly impoverished children who did not receive the same care, researchers report. The study provides rare experimental evidence that such programs can give poor children a better shot at living longer, healthier lives.
Given this kind of write-up, I was ready to really like this paper.  The reporting suggested that longterm, significant health effects could be had with relatively little investment up front.  Regular readers of MT know that that's the kind of health expenditure we would readily support.

Indeed, Campbell et al. concluded,
Whatever the channel, our evidence supports the importance of intervening in the first years of life and suggests that early childhood programs can make a substantial contribution to improving the health of adult Americans and reducing the burden of health care costs. An intervention that lasted 5 years and cost $67,000 [in 2002 dollars] produced sustained and substantial health benefits. Early childhood interventions are an unexplored and promising new avenue of health policy.
Is this science?
This paper should not have made the kind of splash it did. Indeed, its struggle to find convincing samples, measures, and statistically meaningful analyses of same, mainly raise questions about how such research should be done, if indeed it even can be done.

The results as reported are provocative, although the paper itself is less convincing, even if, as the authors write,
We use exact permutation tests to account for small sample sizes and conduct a parallel bootstrap confidence interval analysis to confirm the permutation analysis. We adjust inference to account for the multiple hypotheses tested and for nonrandom attrition. 
Even so, these sample sizes seem 'under powered' to be convincing for this sort of study, to say the least.  Adjusting for attrition and multiple testing as the authors did, even if adequately representing all the data exploration done, with this small and irregular sample, does not convincingly mean that the results are very strong.  The authors went further, to make numerical relative risk estimates on these very small samples, that have to be regarded as very crude, at best, and without much heft.

It has been said with good reason that if you can't see an effect in a sample of 30 or so, the effect isn't big enough to matter.  This argument would apply to simple enough situations where a single variable could effectively be isolated, such as a particular drug vs placebo. Or a gene with strong effect. Unfortunately, in a complex causal web as this paper is reporting, sampling issues are more difficult.  Every sample may be too different to compare, or measures so numerous, interacting, incomplete, imprecise, and context-specific as to require unachievable samples for very definitively specific results to be reached--especially if causation is by complex 'variables' such as diet and so on.

Our personal predilection is that inherent genotypic effects are being heavily oversold these days, for reasons that have as much to do with scientists and their enterprise as they do with the actual facts on the ground about health and societal position.   We would be very happy to be able to laud Campbell et al.'s results as support for our views.

But another view we try to express consistently is that studies should not claim more than they actually find, and that the media should not trumpet them (or in many cases, should refuse even to report them), and should be far more critical of the studies they do decide to report.

In this case, unless we somehow are badly misinterpreting the study, it should be reported at the very most as finding that early effects can be very long-lasting, and that there may be many environmental reasons for this, that could in principle be important to identify.

But as it stands, it just seems highly premature to draw the authors' conclusions, much as we'd like to believe them.

No comments: