Roy Gordon wrote:
> The study in question didn't compare caffeine intake of persons while they
> had AD to those without AD, but over a 20 year period prior to being
> diagnosed with probable AD, according to the abstract.
>> Yes, incipient AD may have been a factor, but for 20 years previous to
> diagnosis? (Also, caffeine is a central nervous system stimulant which
> supposedly reading the financial pages isn't, at least in any obviously
> similar sense.)
>> Again the study is very small and we don't have the full text.
>> Another question is the following, quoting from an unrelated article in the
> NY Times:
>> --------------------------------------
>> When so many variables [as it looks to be in this study] are in question
> ... it is quite likely that chance alone will cause a few of them to appear
> significant. Scientists usually employ statistical formulas to ensure that
> none of the associations they find can be attributed to chance...
>> --------------------------------------
>> So, did the study in question take care here?
>> Also, per the newspaper article, prospective studies are needed.
>> -- Roy
>>>>A similar observation that could have been made is that the Alzheimers
>>patients read the financial pages of the local newspaper significantly
>>less often than those without AD and then to infer that if only you
>>could force those AD patients to read the financial section they might
>>not have AD.
>
I was about to remark that what makes this study striking, despite the
small size, is the very substantial difference in the caffeine intakes
of the two groups, but then I noticed the enormous range of error
associated with the numbers. Perhaps a statitician can comment more
intelligently on this, on how much confidence this should give us on the
size of the difference.
It somewhat bothers me that they have framed it in terms of caffeine
intake. I assume that they didn't actually have real measurements of
caffeine intake, but estimated these from the subjects' reported daily
intake of coffee, tea, etc (the reason why the margin of error is so
large?). To me that's putting an assumption on the data from the outset,
that caffeine is the (only) relevant component of what they are actually
measuring.
For example, if coffee were the primary source of the caffeine intake
difference in the groups, might there not be another component of coffee
that could responsible for the protective benefit alone or in
combination with others (including caffeine)? In any case coffee, tea,
etc are the real ways most people ingest caffeine, so knowing the
statistical relationship between actual coffee consumption, tea
consumption, etc. is more valuable to most of us anyway, regardless of
the ability to determine what compounds are responsible.
Eric
