Tap water tempest

You know the news isn’t going to be good when the report of a new study cites sources in City Hall rather than the scientific literature. That’s exactly how the story broke in February, reporting a link between tap water and miscarriages in a California study. On February 10, the Los Angeles Times warned its readers in a front page report that “drinking five or more glasses of cold tap water per day could increase the risk of miscarriage...” The story noted that the report, to be published in the 18 February issue of Epidemiology, was all the buzz in City Hall. “Local officials, noting that they had not yet read the study, expressed concern.” The study, it turns out, followed 5,144 women from three California counties. None of the news reports remarked on its odd design: researchers first collected the data about women’s water consumption and pregnancy outcomes, then matched those to the estimated levels of total trihalomethanes (THMs) in the women’s tap water. Instead of starting with an explicit hypothesis to test, the researchers sifted through their data looking for a pattern. They found that if they focused on tap water with more than 75 parts per million THMs, and singled out women who drank at least five glasses of this water a day, they could achieve a statistically significant result: 15.7% of the women in this now very small group (2% of the study population) had miscarriages, compared with 9.5% among the remainder. Epidemiologists call this kind of analysis a legitimate way to generate a hypothesis — but not to test it. Reporters apparently skipped the Methods section of the paper (as they often do) and sounded the alarm. “In a study that may spark changes in regulation of the nation’s drinking water supplies,” reported the Sacramento Bee, “California researchers said Monday that they have found a correlation between high miscarriage rates and a common byproduct of chlorine found in tap water.” The Bee report did note that the findings contradict a previous study in North Carolina, which found no such correlation. To help women sort out this confusing situation, the report quotes the state’s top health official, proffering this advice: “Where the strength of the science is, we still don’t know one way or the other. It’s up to the woman to decide. We think it’s worth discussing with an obstetrician/gynecologist.” Why the family doctor would know more about the as-yet unpublished study than the state health department is left for the reader to decide.