This week's journal reviews on Doctors.net.uk
Journal Watch is a service provided to summarise some of the most popular medical journals.
Doctors.net.uk has a panel of specialist advisers responsible for reviewing a range of journals of general medical interest and some more specialised publications.
General Journal Watch is written by Dr Druin Burch, Consultant in Internal Medicine
This week's journals include....
A randomised controlled trial of Weight Watchers
A fascinating paper for many reasons. Through a grant to the Medical Research Council, the commercial operation known as Weight Watchers has submitted itself to trial by science. It's hard not to believe that them doing so was brave and admirable, and that many other policies or commercial businesses claiming to offer health benefits should copy their example. It's also wonderful to see effort being invested in looking at the effects of simple weight-loss interventions, rather than diet and exercise being ignored in favour of vast sums being spent on dubious and dangerous drugs. The primary outcome was positive too: Weight Watchers works. The other messages are probably even more important but they're also a damn sight bleaker. Firstly, these sort of trials are difficult - the drop-out rates were huge. Secondly, the benefits are small. Those randomised to Weight Watchers were a couple of kilos lighter at the end of a full year. Combating obesity is fiendishly difficult; this is a decent but very tiny step in the right direction.
"Estimating treatment effects for individual patients…"
Critics of randomised controlled trials make excellent points. These trials are stupid ways of trying to work out what we should do for individuals. They tell us something about average responses and precious little about what will happen to a particular person. They have more offences than we have thoughts to put them in, imagination to give them shape or time to tell them in. This paper attempts to do better by predicting what will happen for particular people. The only problem with this wonderful idea is that it can't be done. The reason RCTs are of value is not that they're particularly good, just that they're not as bad as the alternatives. For treatments that have moderate effects - most of medicine - figuring out average impacts via RCTs is the best we can do. It's sad, but not as sad as the mess you get yourself into if you start pretending it isn't the case.
Evidence-based social engineering
A quite incredible study. Miraculously, the American Department of Housing and Urban Development agreed to a multi-site randomised trial whose intervention was the re-siting of poor people. Women with children in areas of terrific poverty were either let be, were offered housing vouchers to use as they wished, or were offered housing vouchers that could only be redeemed if they moved to areas without a massive degree of poverty. The trial's primary outcomes, not discussed here, were those of employment, income and education - this study looks at BMI and glycated haemoglobin. After more than a decade of follow up, with 4,500 families enrolled, rates of extreme obesity and diabetes were decreased in the group who'd been pushed away from areas of concentrated poverty. The implications - for health, for epistemology, for freedom - are staggering. What this trial really tells us, however, is not whether such things are right or wrong, but that it's possible to work out what the actual impacts are of a profound piece of social engineering. Discussions of whether we should engage in it should be based on this sort of data.
Low levels of BNP cause diabetes
Mendelian randomisation is a form of natural experiment. Take a genetic variant that's inherited randomly. If it matches up with a disease then you're looking at a relationship that's causal rather than brought on by a confounder. There are qualifications and limitations, but it's broadly true. Some time back these reviews of mine looked at studies of this sort done on C-reative protein (CRP). CRP was known to be linked to heart disease but what was unknown was whether it was associated through confounding or causally involved. By looking at people who'd inherited a single mutation that meant they didn't produce CRP, the answer became clear: CRP isn't causally linked to heart disease, for all that it serves as a marker of it. This paper starts by noting an association between type 2 diabetes and low levels of brain natriuretic peptide. It finds that those with a mutation causing them to produce lower levels of BNP are at higher risk of developing diabetes. The relationship, in other words, is causal. As ever with PLoS Medicine
, the editor's summary does an impressively good job of explaining the paper, pulling off with aplomb the rare arts of putting it in context and making clear what it doesn't say.
Go to the Journal Watch page
to see all monthly and weekly reviews.