NEJM 2 Aug 2007 Vol 357

You may remember that in April JAMA published a paper by the indefatigable Harlan Krumholz and his team, who looked for 85 reported genetic risk markers for coronary disease in 811 patients with acute coronary events without confirming a single significant association. Clearly the way these polymorphisms have been sought for up to now is pretty useless. The haystack – 3 billion base pairs in the human genome – is huge, and the sought-after needles are few. Rummaging is not going to succeed; but what if you could chop the haystack up, go over each section with a giant magnet and watch the needles fly into your grasp? The genomic equivalent is a very expensive tool called the SNP chip (SNP stands for single-nucleotide polymorphism). Currently available SNPs can perform more than 500,000 comparisons, but the down side is that the statistical threshold for “genomic significance” is therefore 0.05 divided by half a million, i.e. 10-7. So your magnet may pick up the needles, but there may still be a big fuzzy bale of hay around them. The way to get rid of this is to compare two large genome-wide case-control studies: here they come from the Wellcome Trust Case Control Consortium and the German Myocardial Infarction Family Study. And now at last we can see that nine loci are probably related to coronary risk, and the most definite lies at the start of chromosome 9. Naturally, we still need many more studies in different populations, with bigger, better SNPs. You can wake up now.

Let’s race from the genomics lab to the labour ward, where a mother with twins has gone into premature labour. It’s too late to stop her now, but should we have been using 17-alpha-hydroxyprogesterone injections from week 16 onwards to try and prevent this? Apparently the stuff works in singleton pregnancies; but this placebo-controlled study in twin pregnancies failed to show any effect.

Another group of women who are at risk of premature delivery are those with a short cervix – as measured by ultrasound. Here a multinational study looked at the effect of giving progesterone 200mg vaginally every night from 24 to 34 weeks’ gestation. This time the progestagen worked, with a definite reduction in spontaneous delivery before 34 weeks and a probable reduction in neonatal morbidity.

I was taught medicine before the era of modern imaging, so both cardiology and neurology depended greatly on clinical acumen – i.e. confident guesswork and pompous assertion. In those days, rheumatic heart disease lay at the very heart of clinical teaching because it could conjure up such a wide range of real or imaginary clinical signs. The only problem was that it had ceased to exist in Oxford 30 years earlier, due to a new drug produced there called penicillin. So the only case I have ever seen was in Iran, 33 years ago; and yet childhood rheumatic heart disease is still common in places like Cambodia and Mozambique, where these investigators looked for it in children by echocardiographic screening. The pick up rate was ten times that of good old fashioned clinical acumen.

The review this week deals, fittingly, with the prevention of preterm delivery, an area where there is some blurring of the boundary between prevention and treatment. Good evidence is hard to come by, especially since it is tricky to define the population at highest risk. Calcium-channel blockers seem to emerge favourite, and before 32 weeks the authors use indometacin and other cyclo-oxygenase inhibitors without apparently causing premature closure of the ductus arteriosus.