Assorted Links

Thanks to Steve Hansen.

6 Replies to “Assorted Links”

  1. Seth,

    That magnesium study has precisely the same problems as the Willett studies that we had been discussing. Look at Table 1. Across the board, almost every indicator of health (except physical activity, but more on this below) increased monotonically across the magnesium quartiles. The folks in the highest quartile were inherently healthier than those in the lower quartiles. This inherent (i.e., unmeasured) difference cannot be adjusted for in statistical models. Incomplete adjustment for “potential confounders” (or even “known confounders”) can just as easily increase the net confounding bias as reduce it.

    Regarding the physical activity measure, I’m skeptical. They measured it in hours per day, and the averages ranged between 4.5 and 4.9 (hours per day!!) across the quartiles. I’d wouldn’t put much weight on that one!

    I wouldn’t use this study as a reason to start taking magnesium!

    1. The real problem isn’t the research that is done (which is easy to ignore), it is the research that isn’t done. E.g., the narrowness of research at medical schools. Ioannidis has never shown any sign of understanding that.

    1. Ben Goldacre’s misunderstanding, as revealed in that article, is epic. First, he doesn’t explain how he knows statins do cause muscle pains. I’m sure they do — but how does Goldacre, the evidence snob, know? Second, his explanation for why his study got the wrong answer makes no sense. “Incomplete data”? All studies have incomplete data.

Comments are closed.