Statistics: Proceed with Caution [Scott]

Jill Stanek's excellent blogsite has a discussion over an article in the journal Contraception. The article covers contraception policy in Spain and the abstract for the study concludes as follows:

STUDY DESIGN: Since 1997, representative samples of Spanish women of childbearing potential (15-49 years) have been surveyed by the Daphne Team every 2 years to gather data of contraceptive methods used.

RESULTS: During the study period, 1997 to 2007, the overall use of contraceptive methods increased from 49.1% to 79.9%. The most commonly used method was the condom (an increase from 21% to 38.8%), followed by the pill (an increase from 14.2% to 20.3%). Female sterilization and IUDs decreased slightly and were used by less than 5% of women in 2007. The elective abortion rate increased from 5.52 to 11.49 per 1000 women.

CONCLUSIONS: The factors responsible for the increased rate of elective abortion need further investigation.

The ensuing discussion about the study (in the comments section of Jill's blog) contains some fiesty exchanges about contraception in general, with some pro-lifers insisting this study proves increased contraceptive use results in more abortions. At least one post claims the pill itself is abortifacient.

What follows is what I said on Jill's blog:

Admittedly, I’m no statistican. However, I have two concerns: First, statistics are slippery things and it’s tempting to quote them when they seem to support our view and downplay them when they don’t. With any study involving the extensive use of stats, it’s best to withhold a declaration of victory until 1) the internal numbers are fully analyzed for strengths and weaknesses, and 2) the methodology of the study is subject to further peer-review. To cite one example, in 1994 (some) pro-lifers went crazy citing a statistic from the Daling study which allegedly indicated that women undergoing first trimester abortions were 800% more likely to develop breast cancer by age 35. Not long after, the alleged statistic was shown to be a fluke, a random statistical error that sometimes creeps up in large samples. Applied to the study in question, I’m not suggesting that pushing contraceptives on unmarried couples has no impact on abortion rates. I’m only appealing for caution while we await further commentary.

Second, regarding some comments above on BC pills functioning as abortifacients, my own organization takes a cautious view while we await further evidence. That is, while we don’t think there is sufficient evidence to say for certain that the pill functions as an abortifacient in the event of breakthrough ovulation, we do think there’s sufficient evidence to indicate it may function that way. Thus, given human life is at stake, we should err on the side of caution while we await further evidence, meaning we do not endorse its use. But we are careful not to overstate our case and claim certitude when the evidence is still open to debate. (We have dealt with that debate on our blog, so I won’t go into it here.)

In short, it’s important that pro-lifers function with intellectual integrity, meaning we go only so far as the evidence allows. Perhaps the Spain study is in the main accurate; if so, it’s useful for confronting the lies of PP. But while we await further review, I think we should hedge our conclusions just a bit.

Kevin DeYoung has helpful posts on being cautious with statistics here and here.

You may also want to read this.