TCS Daily

Life or Death Decision?

By Iain Murray - September 16, 2002 12:00 AM

Abortion is possibly the most contentious issue in American domestic politics. The battle lines are sharply drawn, and both sides are always looking for more ammunition to expend in their cause. Scientific studies are the heavy artillery in this battle, providing withering fire unless their results can be debunked or somehow deflected. In recent weeks, the pro-life side has wielded a new form of ordnance in the shape of a study that claimed to find a link between abortion and early death. If true, this could have been the equivalent of a daisycutter bomb. Instead, it turned out to be a dud.

The study, by researchers at the Elliot Institute, a non-profit corporation founded in 1988 "to perform original research and education on the impact of abortion on women, men, siblings and society," attempted to build on a 1997 Finnish study that had found significantly increased risks of death among women in the first year after they had had an abortion. The researchers' results, published in the Southern Medical Journal, suggested that Californian women on Medicaid who had had an abortion were 62 percent more likely to die in the eight years after the event than women who had successfully delivered a child. The rates for death from suicide, accidents and AIDS were 154 percent, 82 percent and 118 percent greater respectively.

These figures look terrible at first glance, but one of the cardinal rules when considering studies such as this one is that correlation does not imply causation. Is there, for instance, something else driving both abortion and early death, such as psychological problems? The researchers, to their credit, controlled for this possibility by accounting for any psychiatric problems the women might have had in the year before their death, but the chances of increased death overall hardly changed. Women who had aborted were still 61 percent more likely to die than women who had successfully delivered. Only death by accident ceased to be statistically significant.

Yet the researchers stopped there. They did not or could not refine their findings by controlling their model for further potentially confounding factors. For instance, they found a 93 percent greater chance of death by homicide among women who had aborted. Yet, as the latest National Criminal Victimization Survey shows, persons who are married are much less likely to be victims of violence than persons who have either never married or are divorced or separated. This is because married people are much less likely to place themselves in risky or dangerous situations. Marital status might therefore account for some of the difference in outcomes after pregnancy, assuming that married women are less likely to have abortions. Unfortunately, because the data the researchers used were collected from government records, no information was supplied as to race or marital status. These are perhaps the two most important variables in determining health and susceptibility to crime, so without accounting for them, any study is bound to be flawed.

That might not be the case if the study had found more emphatic results, but even a 100 percent difference in outcome is thought by most epidemiologists to be poor evidence of a causal effect; a difference of 200-300 percent, in a well-designed study that controls for the most obvious confounding factors, is felt to be the minimum necessary to suggest causation. This study did not really approach that standard.

This is a familiar tale in the abortion debate. There has been much discussion, for instance, of whether abortion contributes to breast cancer. A 1996 review of 23 independent studies on the subject found, overall, a 30 percent increased risk of breast cancer following abortion. This is far too small a difference for anyone to say that abortion causes breast cancer. However, there does appear to be evidence that something happens. Women with a family history of breast cancer who undergo an early abortion, for instance, do appear to have a significantly increased chance of contracting the disease. That is why research on the subject is ongoing and reputable journals like the British Medical Journal are willing to publish such studies.

The most recent studies have done little to gain the pro-life side any ground in the abortion debate, having merited barely a mention in the press. If there is a connection between abortion and early death, the public does need to know. But if the studies are not well-designed enough, they really need to return to the proving grounds before they are deployed.



TCS Daily Archives