An exercise in critical reading in epidemiology

This is an article from New Scientist based on a paper in Epidemiology. On the right are questions and comments that might occur to an enquiring reader.

Electrical appliances linked to miscarriage

10:54 10 January 02

Anil Ananthaswamy, San Francisco

The strong magnetic fields produced by some electric appliances and vehicles increase the risk of miscarriage, claim researchers in California. Their findings also suggest that most previous investigations into the health effects of electromagnetic fields (EMFs) have been measuring the wrong thing.

"The studies really represent state-of-the-art research (1) into the causes of pregnancy loss," says epidemiologist David Savitz of the University of North Carolina in Chapel Hill. Nevertheless, he says the researchers' interpretation of their findings may be wrong. California is already initiating public hearings to discuss the findings.

One study was led by De-Kun Li, a reproductive epidemiologist at the Kaiser Foundation Research Institute in Oakland, California. His team asked 1063 (2) women around San Francisco who were in the first 10 weeks of pregnancy to spend a day (3) wearing a meter around their waists that measured magnetic field levels every 10 seconds. Overall, they found that women exposed to peak levels of 1.6 microteslas (4) or greater were nearly twice (5) as likely to miscarry as women not exposed to such strong fields.

More significantly, says Li, among the 622 women who said the measuring period had been a typical day (6), those who experienced high peak fields were three times (7) as likely to have a miscarriage. "That's another confirmation that the effect is due to EMF," says Li. (8)

Shavers and hairdryers

Other factors can have a more dramatic effect, however. The risk of a miscarriage increases tenfold as women age, for example, from 5 per cent for women under 30 years old to 50 per cent for those in their mid-40s. (9)

Li's team didn't look at what was producing the fields, but appliances such as shavers, hairdryers and vacuum cleaners can produce strong alternating magnetic fields, as can electric vehicles such as trams and trains (10). The key is proximity to the source, as fields drop off rapidly with distance.

Alternating magnetic fields also have associated electric fields (11). The few previous studies of the effect of low-frequency EMFs on miscarriages, such as one involving 727 women done in 1991 by Raymond Neutra's group at the California Department of Health Services in Oakland, have been inconclusive.

But Li thinks this is because Neutra looked at people's average exposure to electromagnetic fields over time, not peak values. "People have never looked at peak EMFs before (12)," Li says. "My study opens a new chapter for these EMF effects. Not just for miscarriages, but for other health effects."

EMF spikes

When Neutra reanalysed the data from his earlier study, which has only now been published (13), he discovered the results were similar to Li's. Women exposed to peak EMF levels greater than 1.4 microteslas were nearly twice as likely to miscarry (14).

But Savitz sees things differently. "Both studies found a reassuring lack of association for the most well-established measures of magnetic field exposure, that is average magnetic fields." The correlation between exposures to EMF spikes and an increased miscarriage rate may be due to other factors, he says. For instance, women who have a healthy pregnancy are more likely to suffer from nausea. This may make them more likely to stay at home and do less, and thus also reduce their exposure to magnetic fields. (15)

After Savitz's comments, Li's team looked at their data again. They also checked for other possible risk factors such as drug use, carrying heavy loads and previous induced abortions (16). "We adjusted for 30 different kinds of risk factors. Nothing changed," says Li. (17)

Localised heating

But Michael Bracken, an epidemiologist at Yale University, is unconvinced. "There are numerous ways of measuring these fields, and one worries that if you do it enough times, then you are going to find positive associations," he says. "There's a real risk in these things getting over-interpreted and scaring the dickens out of people." (18)

In the past, EMFs have been blamed for various other ill effects, especially leukaemia in children. But no one can explain how relatively weak fields might cause the DNA mutations that lead to cancer, and most studies have failed to find evidence of a link.

The peak values measured by Li are way below the recommended exposure limit of 1600 microteslas. Above this level, EMFs can induce electric currents in the body, which leads to localised heating. Li speculates that EMF spikes could cause miscarriages by subtly disrupting cell-to-cell communication (19). "But as epidemiologists, we should not feel weaker because we don't understand the mechanisms." (20)

Journal reference: Epidemiology (vol 13, p 1, 9 and 21)


10:54 10 January 02



1. What is state-of –the –art about this research? Weak field magnetometry is old hat, so are EMF scares.

2. 1063 is the Trojan number, the total in the survey. Those who actually contribute to the statistics are smaller in number and get fewer as we go on.

3. One day is used to represent the whole of whatever is the putative critical period for exposure. It is rather unlikely (say one chance in 40) that the true peak was caught.

4. 1.6 microteslas is a very low flux density. There are no known biological effects at this level. It is the level you would experience one metre away from a straight wire carrying 8 amps. Anyway, what is so special about this threshold? Did it happen to give the best results?
If you wonder why magnetometers are so interesting to fringe science, it is because they are simple and sensitive. You see on television programs people claiming to be ghost hunters, who wander round houses thrilled to bits because the numbers on their instruments go up and down. Nothing to do with the matter in hand, but very impressive to naďve punters. It so happens that I published papers in the early eighties describing one of the first integrated intelligent microsensors. Magnetic sensing was chosen because of its electrical simplicity, allowing us to concentrate on the aspects of intelligence and communication.

5. "Nearly twice" means a risk ratio of less than two, which is not acceptable in real science.

6. What is the scientific definition of a typical day? Is it the same for a dangerous sports addict as it is for a couch potato? It has provided a convenient way of dredging for a data subset showing apparently greater significance. Those who had an atypical day must therefore show less significance. Anyway, it is a purely anecdotal characteristic.

7. Three times is apparently an acceptable risk ratio, but one obtained by dubious data manipulation. Assume from the figures below an overall miscarriage rate of about 10%. This means that about 15 were below the threshold and 45 above – small numbers compared with the initial Trojan number and room for many confounding effects.

8. It is no confirmation. Correlation is not causation!

9. Age is clearly a dominant variable. Age dependent habits would therefore be major confounding factors.

10. The "researchers" did not look at the very thing that most real scientists would look at. An apparent correlation would only be the starting point in real science. There are many other potential sources of exposure; for example, some of the highest electric currents to which you are likely to be proximate occur when you start your car. Driving is then associated with all sorts of other potential stresses.

11. A complete red herring. Electric fields depend on two things, the magnetic flux magnitude and its rate of change. For the cognoscenti: . At normal mains frequency the rate of change is very small. Switching surges might be another matter, but they are not mentioned.

12. There are all sorts of properties of waveforms you can dredge for. They have not yet started on spectral analysis.

13. Why has this only now been published? This apparently innocuous statement is the most damning of all. It is suspected that one of the greatest influences in such "research" is publication bias (negative results are simply not published). How many other unpublished negative results are there?

14. Why has the threshold changed? Was it to optimise the results? Nearly twice is a risk ratio of less than two.

15. The possibilities for confounding factors are almost infinite. Did the women carry on working, possibly at mechanically and mentally stressful jobs, and did they receive their exposure at work? Etc.

16. More red herrings, such risks could not possibly be confounded with magnetic field exposure.

17. They "adjusted" for thirty different factors, which is about the same number as the difference between the numbers above and below the selected threshold. What assumptions were made? Did they understand how their statistical package dealt with them? When epidemiologists inadvertently reveal the details you find that they use, for example, post codes as a proxy for deprivation, when there are multimillionaires and paupers who share the same postcode. Also, however honest the researchers, there is enormous scope for the human subconscious to intrude in order to optimise the results. In Sorry, wrong number! it is shown that even in purely physical measurements, the subconscious can act to "optimise" results.

18. Couldn’t have put it better. Interesting that the sane comments come from Yale, where the influence of Feinstein is apparent. He is one of the few epidemiologists who seem to adhere to the scientific method.

19. Oh yeah?

20. A masterly quotation that sums up the genre. This was the attitude of the alchemists and astrologers. It was only when people broke away from this form of mental indolence that real science began. It highlights the fact that modern epidemiology is a throwback.






moncler outlet usa Moncler outlet hermes outlet prada outlet gucci outlet dior outlet lv outlet chloe outlet