Monday, 12 September 2011

How to become a celebrity scientific expert

Maybe you’re tired of grotting away at the lab bench. Or finding it hard to get a tenured job. Perhaps your last paper was rejected and you haven’t the spirit to fight back. Do not despair. There is an alternative. The media are always on the look-out for a scientist who will fearlessly speak out and generate newsworthy stories. You can gain kudos as an expert, even if if you haven't got much of a track record in the subject, by following a few simple rules.

Rule #1. Establish your credentials. You need to have lots of letters after your name. It doesn’t really matter what they mean, so long as they sound impressive. It’s also good to be a fellow of some kind of Royal Society. Some of these are rather snooty and appoint fellows by an exclusive election process, but it’s a little known fact that others require little more than a minimal indication of academic standing and will admit you to the fellowship provided you fill in a form and agree to pay an annual subscription. So sign up as a Fellow of the Royal Society of Medicine, and keep good company with a range of naturopaths, homeopaths and chiropracters who have discovered this easy route to eminence. The really nice thing is that even academics can be hoodwinked by this one.

Rule #2. Find a controversial topic. This is key. You have to be willing to take a definite position on something that people have strong views about. A scare story is good - we’ve all been doing X for years but it could damage us. Finding someone to blame is also good - people who do Y are feckless. And the buzz word of the decade is neuroscience, so if you can work that in, success is guaranteed. If you're short of ideas, the list of the right might help inspire you. A recent article in the Biologist hits the spot with “The biological effects of day care”, managing to get us worried about an everyday activity, blame working mothers, and get in a neuro message all at once. It's even spiced up with a bit of conspiracy theory: experts know that day care is bad for children’s brains but nobody is allowed to speak out because it is too politically sensitive. This presses so many buttons that few journalists could resist the story.

Rule #3. Specify a causal chain. As we shall see when it comes to assembling evidence, it is particularly useful to have a causal chain with several steps. For instance:

The point here is that if you can usually find at least some studies that provide evidence for bits of the causal chain. Although it may be inconvenient if, as in this case, studies looking for a link between A and D fail to come up with clear evidence (Lucas-Thompson et al, 2010), you can rely on two things: first, few readers will be familiar with the research literature, so they will only know as much as you tell them. And second, step C, brain abnormality, is highly salient and once you start talking about that, it will distract attention from the other levels of description.

Rule #4. Avoid rigorous peer review. You don’t want to have your views critiqued by someone who knows the literature, or checks your sources. Writing books is a safe bet for avoiding pre-publication scientific critique. As far as journals go, the Biologist is ideal. This publication for the members of the Society of Biology claims to be peer-reviewed, but, as we shall see, the review process is far from rigorous.

Rule #5. Assemble supportive evidence. Note, it is important not to present all relevant studies, just those with findings that can be fitted into the causal chain.
The author of the paper in question, Dr Aric Sigman, presents us with so much positive evidence that he manages to give the impression that the whole casual chain has been validated. He starts by mentioning studies that investigated the link between A and B and find that salivary cortisol is increased in the afternoons in children attending day care. This is a product of the hypothalamo-pituitary-adrenal (HPA) axis, which is elevated in response to stress. This result is well-established, both from studies comparing groups of children who do and don’t attend day care, and from comparing the same children on days when they stay at home or go to day care. This is a potentially concerning finding, if it can be shown that there are consequences for children’s learning and behaviour. As Sigman notes “Of central concern is that the routine stress experienced at day care could cause permanent changes in the child’s neuroendocrine networks, with long-term consequences for their mental and physical health as adults.” (p. 30). The article that he cites does discuss this issue, but notes the complexity of causal relationships and cautions against assuming that the cortisol elevation is harmful. In fact the authors draw attention to an animal study suggesting a very different conclusion:
As in the work on cortisol responses to fullday child care, these separations in squirrel monkey infants produced marked and repeated activations of the HPA axis. However, followed into the late juvenile and early adult age, animals exposed to this form of early life stress were found to be less fearful, to produce lower rather than higher cortisol responses to stressors, and to show more optimal development of prefrontal regulatory brain circuits; consistent with these findings, they also performed better on tests of executive functioning. Thus, at least for this animal model, repeated separation stress early in life fostered a form of resilience.” Gunnar et al (2010) (my emphasis)
Sigman avoids mentioning any of this and turns instead to look at the links between cortisol levels and ill health, starting with studies that show a link between cortisol and cardiovascular disease. He does not explain that these were done on people aged over 65 years, but rather implants in people’s minds the notion that this is relevant for his arguments about daycare in toddlers.  Next comes the serious stuff: research linking elevated cortisol to the brain. We are told that “Cortisol is considered neurotoxic and has a global impact on cerebral size (e.g. McEwen 2007; Sheline 2003).” (p. 29). Again, we are left with the distinct impression that children who attend day care will have small brains, but to find out what actually is meant here we need to read the cited articles. When we do, we find that McEwen (2007) is a thorough research review of the physiology and neurobiology of stress and adaptation that nowhere mentions the terms ‘neurotoxic’, ‘global’ or ‘cerebral size’. Rather, in this article McEwen develops a complex theory that considers both positive and negative impact of stress. It actually has a section subtitled “Protection and damage: the two sides of the response to stressors” which discusses animal studies demonstrating how elevated cortisol can either improve or interfere with brain function, depending on context. Neurotoxicity does feature in the other article, by Sheline (2003), but this is concerned with mood disorders in adults, and discusses effects of hypercortisolemia, a condition where there is chronic elevation of cortisol, rather than a temporary increase at specific times of day. Subsequent citations are to papers that considered the role of cortisol in psychiatric disorders such as anxiety and depression: note that now the evidence is focused on a link between cortisol and adult psychiatric disorders in the opposite direction (from D to B), yet it is presented in the context of discussing consequences of high cortisol in children who attended day care. Sigman further states: “a higher cortisol awakening curve may be a biological marker for an underlying disposition towards developing depressive and anxiety disorders” (p. 30), even though the studies of toddlers attending day care show a cortisol response that develops through the day, rather than a chronically raised level: see Vermeer and van IJzendoorn (2006) for a well-balanced discussion of such evidence.
It would be tedious to wade through every cited article, but it’s worth considering just one more example. We are told “In human grey matter, the quality of a mother’s care in early childhood is thought to alter the size of the hippocampus (Buss et al, 2007)” (p. 30). Erm no. Buss et al clearly stated there were no differences in left or right hippocampal volume between those categorised as having high or low maternal care. What they did find was a complicated interaction between birth weight, gender and maternal care, such that birth weight predicted hippocampal volume only in female subjects reporting low maternal care. But even this limited result doesn’t support Sigman’s case. We are talking about ‘low maternal care’, not ‘attendance at daycare’. And guess what? When we look at ‘low maternal care’ we find it measured from a self-report questionnaire, where low care is partly identified in terms of maternal overprotection. Those categorised this way were more likely to have endorsed items describing their mother in such terms as:
  • Tended to baby me
  • Tried to make me feel dependent on her
  • Felt I could not look after myself unless she was around
  • Was overprotective of me
I’m not an expert in the neurobiology of stress. I can track down articles that appear to have been cited by Sigman (the reference list is behind a paywall) and see where he's given a misleading account, but what I don’t know is how much relevant literature has been omitted. This, of course, is what the celebrity scientist can rely on: there's only a handful of people who both have the expertise in the area, and are obsessive enough to trawl through your writing (if they can access it) and challenge any misleading statements.
Rule #6. Anticipate criticism but don't let it worry you. People who actually do research in the area you are reviewing may get irritated, but most scientists in the field wouldn’t bother on the grounds that they don't know who you are, and aren’t interested in pursuing academic debates outside the domain of mainstream journals. The worst you may get is a few nerdy bloggers such as Gimpy or Mind Hacks criticising you for lack of scholarship, sensationalism and cherrypicking of evidence. Or, if you're really unlucky, you might be up against Ben Goldacre on Newsnight. But meanwhile, your purpose as celebrity scientist has been achieved: your views are all over the media.

Buss, C., Lord, C., Wadiwalla, M., Hellhammer, D. H., Lupien, S. J., Meaney, M. J., et al. (2007). Maternal care modulates the relationship between prenatal risk and hippocampal volume in women but not in men. Journal of Neuroscience, 27(10), 2592-2595.

Gunnar MR, Kryzer E, Van Ryzin MJ, & Phillips DA (2010). The rise in cortisol in family day care: associations with aspects of care quality, child behavior, and child sex. Child Development, 81 (3), 851-69. PMID: 20573109

Lucas-Thompson, R. G., Goldberg, W. A., & Prause, J. A. (2010). Maternal work early in the lives of children and its distal associations with achievement and behavior problems: A meta-analysis. Psychological Bulletin, 136(6), 915-942. DOI: 10.1037/a0020875
McEwen, B. S. (2007). Physiology and neurobiology of stress and adaptation: Central role of the brain. Physiological Reviews, 87(3), 873-904.doi: 10.​1152/​physrev.​00041.​2006 (Open Access)

Sigman, A. (2011). Mother superior? The biological effects of day care. The Biologist, 5 (3), 29-32.

Vermeer, H. J. & van IJzendoorn, M. H. (2006). Children's elevated cortisol levels at daycare: A review and meta-analysis. Early Childhood Research Quarterly, 21(3), 390-401.doi 10.1016/j.ecresq.2006.07.004

Thursday, 1 September 2011

Early intervention: What's not to like?

If a child has language problems, when would be the best age to intervene? At 18 months of age, when they’re just at the outset of learning language, or at five years, when they’re in school? Most people would say this is a no-brainer, with early intervention being preferred on two counts:
  • There are all kinds of secondary consequences of language difficulties: effects on self-esteem, educational outcomes and social interactions. Potentially, early intervention can avoid these.
  • It is easier to influence the course of development while the brain is still plastic. An analogy can be made with vision, where it is well-recognised that amblyopia (or "lazy eye") needs to be corrected early in life, because otherwise visual pathways in the brain do not develop normally, and the potential for good vision in the lazy eye is lost.
Currently, interest by policy-makers in early intervention has focussed mainly on children’s social and emotional outcomes, with a report by MP Graham Allen emphasising the benefits, not just for children’s outcomes, but also in economic terms. The argument is that by preventing problems from developing, we have the potential to save millions of pounds that would otherwise be spent in dealing with problems that manifest later in childhood.
The Allen report does not say much about children’s language development, but similar arguments are often made, and in some areas of the country, speech and language therapy services put most of their resources into intervention with preschoolers.
There is, however, a problem with early intervention that is easily overlooked, but which is well-documented in the case of children’s language problems. This is the phenomenon of the "late bloomer". Quite simply, the earlier you identify children’s language difficulties, the higher the proportion of cases will prove to be "false positives" who spontaneously move into the normal range without any intervention. We’ve known about this phenomenon for many years: For instance, a study conducted by Fischel et al in 1989 followed 26 two-year-olds recruited because their parents reported that they understood complete sentences but could say only a few words. Five months after initial assessment, one third still had problems, one third had made some improvement, and one third were in the normal range. Another study by Thal et al in 1991 followed ten children who scored in the bottom 10% for expressive vocabulary at the age of 18 to 29 months. One year after initial assessment, six had caught up, whereas the remaining four still had delayed language. These early small-scale studies have since been confirmed by much larger population-based studies in the Netherlands and Australia.
The late-bloomer phenomenon was neatly demonstrated in a study just published in the British Medical Journal by an Australian team headed by paediatrician Prof Melissa Wake and speech pathologist Prof Sheena Reilly. They recruited children from a large population-based study, where parents were asked to complete a Sure Start vocabulary screening measure when their child was 18 months of age, as well as a child behaviour checklist. Around 20 per cent of children were reported as having no or very limited spoken words. 301 of these children were randomly allocated to intervention or control groups. The intervention, "Let's Learn Language", was based on a widely-used approach where parents are trained to adopt strategies to enhance communicative interactions with their child. The children were then given a detailed assessment at two years of age, and again at three years. Results were striking: there were no hints of any difference between children in the intervention group and control group on any language or behavioural measures, either at 2 years or at 3 years.
The study authors noted various strengths and weaknesses of their study. Among these they discussed the possibility that the intensity of the intervention (six weekly sessions, each lasting 2 hours) may not have been enough. But they went on to note that “the normal mean language and vocabulary scores achieved by both intervention and control children by age 3 years suggest that natural resolution, rather than our intervention’s intensity being too low, explains the null findings.”
They then point out the sobering conclusion to be drawn: quite simply, if you intervene with children who are likely to improve spontaneously, there will considerable waste of government’s and families’ resources.
Does this mean we should give up on early intervention? No. But it does mean that we need to target such intervention much more carefully. At present, one of the big questions for those of us investigating late talkers is to find characteristics that will allow us to identify those children who won’t make spontaneous progress. This has proved to be surprisingly difficult.
Another important message applies to intervention studies more generally. If you provide an intervention for a condition that spontaneously improves, it is easy to become convinced that you’ve been effective. Parents were very positive about the intervention program. There was remarkably good attendance, and when asked to rate specific features of the program and its effects, around three quarters of the parents gave positive responses. This may explain why both parents and professionals find it hard to believe such interventions have no impact: they do see improvement. Only if you do a properly controlled trial will the lack of effect become apparent, not because treated children don’t improve, but rather because the control group gets better as well.

Reference: (Open Access) :-)
Wake M, Tobin S, Girolametto L, Ukoumunne OC, Gold L, Levickis P, Sheehan J, Goldfeld S, & Reilly S (2011). Outcomes of population based language promotion for slow to talk toddlers at ages 2 and 3 years: Let's Learn Language cluster randomised controlled trial. BMJ (Clinical research ed.), 343 PMID: 21852344