[Parts combined.]
 
Yo, all!
 
I was thinking about the various debates and it occured to me that some
of us might be using different meanings of the word "prove".  WARNING:
this message is super-long.
 
For the purposes of the "drugs vs herbs (and now homeopathy)" discussion,
I'm using the word prove in a scientific sense -- or at least what I
/think/ is scientific since I am not a sciency type.  Bob C, Sukie, et
al, please PLEASE tell me if I'm off the mark here.  "Prove" and "proof",
to me, refer to theories which have a lot of controlled experiment and
study to back them up.  Other folks might interpret the word "prove" in
other, much less technical and much more subjective ways (for example,
accepting testimonial evidence as proof).  It's important to define what
we mean by "proof".
 
Scientifically speaking, a statement or idea (e.g., "ferrets whose diets
are low in carbohydrates suffer fewer incidences of adrenal disease than
ferrets whose diets are high in carbohydrates") is a hypothesis until
controlled tests are done and show the statement to be true, and they are
reproduced by independent researchers doing /the exact same test/.  Then
it becomes a theory.  In science, even the most tried and true "facts"
are simply theories -- there's always the possibility that they'll be
proven (again, through controlled tests) wrong.
 
So what's the scientific process here?  First you have to make the
statement.  The simpler the statement, the easier it is to test.  This
is important.  If you add too many factors, it becomes easier and easier
to get confused results.
 
Next, you might have to add some definitions (e.g., explaining just what
percentage of carbohydrates in a diet defines the diet as "low" or
"high").
 
Now: design the test.  The test needs to be designed in such a way that
there are as few variables involved as possible -- the test subjects need
to be subjected to /exactly the same test except for the single thing
you're testing for/.  It's really important to make sure that every other
possible factor (lighting, playtime, etc...) is the same so that if
/they/ have an effect, that effect is averaged out in favor of effects
from whatever it is you're testing.
 
Say I want to try to prove the statement about carbs.  I'm going to
define the "high", "low" and "average" carb levels and a length of time
for the test.  How long is long enough?  According to "Ferrets for
Dummies", most ferts develop the disease after 3 1/2 years.  So let's
say four years for the test.  Wow, good thing I'm a millionaire.
 
Now I'm going to have to obtain a kits from the same source.  I'm trying
to test for diet here, so I want to make sure they started off having the
same diet and it's nice to know that they come from roughly the same
breeding pool.  That way, any predisposition towards or against adrenal
disease is (hopefully) evenly spread throughout my population.  And I
find out exactly what that diet was and at what times of the day the kits
were typically fed (because, who knows, time of feeding might be a
factor).
 
Let's keep the sample size low: 60 kits total, randomly male and female,
all descented and neutered and all apparently healthy.  The fact that the
sample isn't large adds a huge caveat to the results, because individual
"quirky responses" will have a greater impact.  But still, I need to keep
my sanity here.  Sixty it is.
 
I would also say that, since color seems to be linked to some problems,
preferably all of the same color and pattern.  Let's say, oh, your
standard sable.
 
Next, I design three diets: one high in carbs, one low, and one "average"
(about so-so, carb-wise).  I make recipies as specific as I can -- I'm
trying to make sure that everyone eats the same amount.  It's almost
impossible for me to guarantee that each ferret gets exactly the same
meal (unless I whip them up totally artificial meals from raw chemicals,
since that's the only way to completely control dosage), but I'm going to
do my best and hope that things average out in the long run.  Every
ferret will receive some "constants" in addition to the diet, such as the
same kind of hard kibble to clean their teeth, the same vitamin/fatty
acid suppliments, etc...
 
I set up 30 cages so that each woozle has a cage mate.  I enlist a hoard
of volunteers and separate them into "handling", "recipe", and "feeding"
volunteers.
 
Now I'm ready to bring in the ferts.  I have special "handling
volunteers" to do this.  Each ferret is sequentially assigned a number by
a handling volunteer as they come in.  Because they might inadvertedly be
"sorted" somehow by my volunteers, I use a computer to randomly reassign
numbers.  The numbers that are spitted out are done so in an encoded
fashion and printed out in barcode, because some people have prejudices
about numbers and that might inadvertedly mess up the results ("seven is
my lucky number!  I'm going to play especially well with number seven!").
 
I'll have my volunteers relabel each ferret with a collar bearing its
final encoded barcode (and name).  Plus there's a tag hanging from the
cage with a color photo and the tag number/name in case there's some
confusion.  I include instructions that if there is any question as to
who is who, it must be noted.  There's a notebook for each carpet shark
by the cages for this reason.
 
Next, I randomly assign them cage mates.  For about, oh, a month or so,
I let them get used to one another (making sure I feed them all the same
stuff at the same time the source fed them).  If there are any problems,
my handling volunteers can move them around until they're comfortable.  I
can also get replacements if a kit ends up being obviously sick from the
get-go.
 
I make sure that each cage has the same kind of lighting and ventilation
(especially since lighting may also be linked to cancer in ferrets).  And
I take careful notes, lots of 'em, especially for minor factors I can't
control (e.g., the kits' personalities, their weight at the start of the
experiment, etc..).
 
Once they're all settled and happy, I randomly assign each kit to either
the low carb group (20 kits total), the high carb group (20 kits total),
or the "average" group (20 kits total).  I don't know which kit goes with
which number (I don't even ever see the kits), so I can't inadvertedly
mess up the results by assigning some fert to a particular group
("Charlie is my favorite, so I'm putting him in the low-carb group").
 
Now, the average group is going to function as my control group;
alternately, I could just choose to feed them something approximating a
"standard" ferret diet, but the reason why I choose "average" is because
from what I've read, there is no "standard" ferret diet.  Anyway, on to
the test.
 
Now I have some "recipe volunteers" whip up the recipes.  I make periodic
inspections to make sure that instructions are being followed and that
volunteers aren't, say, using the choicest meat and the beginning of the
preparation and the crappiest towards the end.  And anyway, I have a
computer that randomly spits out a label (ferret barcode from within the
carb group being prepared) so that the volunteers aren't tempted to, say,
package the meals in sequential order.  And the recipe-making volunteers
never ever handle the ferrets or know who they are.  This way, they can't
inadvertedly mess up the results ("Oh, this batch is for Charlie, who's
my favorite, so I'm going to slip in a few extra raisins").  Preferably,
these are people who know nothing about ferrets whatsoever, since I
don't want any ingrained prejudices regarding diet to cause them to
inadvertedly mess up the results ("this can't be serious -- this recipe
needs WAY more molasses!").
 
Each mealtime, the handling volunteers come back in (they're now
"playtime volunteers") in to take one fert from each cage ("gang one")
into a play area and play with them (the ones who remain are "gang two").
I make sure that they alternate cagemates between gang one and gang two
every feeding time.  This is because I want to average out the effect of
feeding/playing time if those are indeed a factor -- it's the /diet/ I'm
testing for, see?
 
During playtime for group one, the "feeding volunteers" feed the ferts
still in their cages and note the time that they were fed.  When playtime
for gang one is nearing its end, the feeding volunteers remove the
feeding dishes and clean up any spills.  At the end of playtime for gang
one, the playtime volunteers swap ferts -- making notes in the
appropriate notebooks as to playtime and session -- and play with gang
two.  The "feeding volunteers" now feed gang one, etc..
 
I make sure that the play area is out of sight of the housing area,
because I don't want the playtime volunteers to inadvertedly mess up the
results ("oh, I can see/smell that Charlie has the 'low-carb' diet and I
believe in the low-carb diet, so I'm going to play a bit more thoroughly
with Charlie today" -- if activity is a factor, I want to make sure
everyone plays the same so that activity is averaged out -- it's the
/diet/ I'm testing for).
 
I can't keep my volunteers from choosing favorites due to fert
personality, of course.  I think I'll just have to suck up this variable
and hope it doesn't mess me up too badly.  Wait -- I just realized that
I want the playtime volunteers to be constant, because having human
playmates constantly switched up on the ferrets might end up having a
factor on their stress level, which also might be linked to adrenal
disease.  This means that I'm going to have to /hire/ the playtime people
and pay them well to minimize the turnover.  It's a /really/ good thing
I'm a millionaire.
 
Now I do this for four years, training and retraining my volunteers when
necessary, and inspecting things (or having a trusted helper inspect
things, if it would involve violating the double-blind process) to make
sure that everything's being done properly.  I'm having my playtime
employees check in from time to time to make notes on the "down-time"
activity levels of the ferts (does this pair tussle more per day than
this other pair?  It might be a factor later).
 
During this whole time, I'm having vets come in and check the beasties
regularly.  The vets /especially/ will have no clue as to what the
animals are being fed.  I track all of these results, and by the end
of the four years I should have some numbers and some results: either
diet has an effect (in which case one of the groups will show a marked
difference from the other two groups) or it does not (in which case
adrenal cases will occur with about the same frequency in all three
groups).
 
Obviously, this is a LOT of work.  It's probably way more controlled than
it needs to be.  Or maybe it's not controlled enough.  But I wanted to
give you an idea of what it means to create a test.  In order for a
statement to have a true scientific basis, it has to go through this
process, and it has to be reproduceable by other people in a completely
different area by simply following the protocol (e.g., people in Europe
will have the same /kinds/ of results, though the actual levels or
occurances of adrenal might be higher or lower depending on local
factors, such as the region's gene pool for ferrets).
 
If you ask "why aren't there more studies to test for x,y,z", this is
why -- tests, even the simple ones, are large, involved and expensive.
Imagine what it must be for /humans/.  And this is what I mean by
"proof" -- it means that the statement in question has undergone this
process, aka "the scientific process", and that other scientists have
review the work, found it to be sound, and replicated it with the same
or similar results.
 
Okay, I'm going back to lurking now.  :)
 
 -- Kat
[Posted in FML issue 3935]