Time for my second cup of coffee

Being a serious coffee drinker, I love studies that show drinking coffee is good for you.  This one is particularly gratifying—drinking three to five cups a day seems to be linked with a markedly decreased likelihood of developing dementia!These are the kinds of studies that get a lot of press.  I guess all the other coffee-addicts out there also want to hear their habit is good for them (note its Number One status in most-emailed articles in the New York Times, Jan. 26, 2009).  It wasn’t too long ago, though, that caffeine was the devil-incarnate of hot beverages and anyone who cared about her health felt pressured to quit her coffee addiction.  So what’s going on?The researchers in this study are careful to point out that their findings only hint at a link between coffee and decreased risk of dementia.  No conclusions can be drawn; no recommendations can be made.But they did feel that the study was unusual in the kind and amount of data available to them.  Of the original 2000 subjects who were selected 21 years ago, 70% were still available for examination.  Because the subjects had reported their coffee consumptions at the beginning of the study, there was less risk that people were inaccurately recalling their consumption.It’s surely a rare thing, to have a good longitudinal group of subjects, but ultimately, it still means that this finding comes from a group of 2000 people x 70% = 1400 people.  And as the researchers pointed out, any self-reported data is subject to inaccuracies.  So multiple inaccuracies in a sample of 1400 people—hmm, maybe I can’t congratulate myself on my coffee consumption after all.When I talk to people about the research potential of online data collection through the Common Data Project, they’ll almost always say to me, “But can any conclusions from online data collection be accurate?”  But for me, the question should be, “Could any conclusions from online data collection be more accurate than what’s available now?"  How sure are we that the conclusions we're drawing now are accurate?  You can imagine that longitudinal studies in particular, that rely on self-reporting anyway, could greatly benefit from online data collecting tools that would reduce the costs of collecting, monitoring, and updating information on thousands, maybe even tens of thousands of people.I’m looking forward to seeing what future longitudinal studies will say about the health benefits of my coffee addiction.