|
medpundit |
||
 |
 |
|
Tuesday, April 22, 2003Unfortunately, I couldn’t find the raw data from the older Nurse’s Health Study this morning due to time constraints, but I remember that the actual differences were very small, and that there was initially some debate about whether or not the findings really meant that hormone replacement therapy protected against heart disease. That debate was quickly brushed aside, however, as media reports repeated again and again the researcher’s spin that heart disease was “decreased by 30%” in hormone users. I do, however, have the data from last year’s Women's Health Initiative study which claims that women who take estrogen have “40% more heart attacks." (Unfortunately, the data isn’t linkable anymore. JAMA revamped its website and the link no longer works.) This is what that 40% increase represents in reality: 0.37% of estrogen users had heart attacks compared to 0.30% of placebo users. In the real world, that isn’t any difference at all. The Times article quotes one expert as saying that when the Women's Health Initiative came out, he realized that “public health has a real problem.” He meant that doctors have been giving the wrong advice to women about estrogen replacement therapy, but the real problem is that our profession has completely forgotten the difference between statistical significance and clinical significance. And even more to our shame, we allow researchers to use statistics to exaggerate their findings without any criticism. Like sheep, we take their press releases at face value. That appears to be true whether we’re in private practice, editing medical journals, or doing research. Shame on us. UPDATE: DB thinks we act on marginal results because our tendency is to want to do the best for patients. That's probably true, but it isn't necessarily for the greater good. It's one thing to mention marginal benefits when counseling a patient about preventive measures. It's quite another thing to base standards of care on marginal results. posted by Sydney on 4/22/2003 08:02:00 AM 0 comments 0 Comments: |
|
