Thursday, August 09, 2007

On Breast Implants and Suicide

The correlation between the two has been in the news recently:

Women who get cosmetic breast implants are nearly three times as likely to commit suicide as other women, U.S. researchers reported on Wednesday.

The study, published in the Annals of Plastic Surgery, reinforces several others that have shown women who have breast enlargements have higher suicide risks.

Loren Lipworth of the Vanderbilt University Medical Center in Tennessee and colleagues followed up on 3,527 Swedish women who had cosmetic breast implant surgery between 1965 and 1993. They looked at death certificates to analyze causes of death among women with breast implants.

Only 24 of the women had committed suicide after an average of 19 years, but this worked out to triple the risk compared to the average population, they reported. Doctors who perform cosmetic breast surgery may want to monitor patients closely or screen them for suicide risk, Lipworth said.

Correlation is a neat little statistical measure, which reflects the extent to which two phenomena appear to move in some connected ways. We call the correlation positive if the phenomena move in the same direction. For example, when one goes up, so does the other, on average. Negative correlation means that when one phenomenon goes up the other one tends to go down.

Now, correlation doesn't necessarily mean that one phenomenon causes the other. This could be the case, but many other explanations are possible. This study is a good example of the kind of study where the correlation is unlikely to be the same as causation. Unless we find something in the implant materials that makes people more suicidal, the best guess for this correlation is that there are some women who elect to have breast surgery for psychological reasons and that these same reasons may also later contribute to suicides. Depression and low self-esteem are examples that crop to mind as such underlying psychological reasons.

The article I link to makes this point clearly, and that is good. But it's always useful to remind readers that correlation doesn't always mean causality.