Inkwell: Authors and Artists

    
You wouldn't be a good placebo group, anyway, because, being
self-selected, the odds were good that you were different, possibly in
an important way, from those who accepted treatment.  Random assignment
is crucial for testing placebos versus treatments.

I'm very glad you are doing well, and that you've lived so long since
the.
  

    
Maria,
Thanks for your input. 

"does having a hodge-podge of treatments at once, with no way to even
estimate what effects any of them may be having, expand the horizons of
knowledge? "

We're trying to save our own lives and help each others. If we can
expand the horizons of knowledge, great, but it's pretty low on our
list of others priorities. 


"but without any scientific studies to support the claims that he
makes (quite boldly) for the harms lectins supposedly do to those of
certain blood types."

I'm not sure what you would consider a scientific study, Maria.  If
you mean a large-scale clinical trial, you're right. No such
large-scale study exists to support (or , for that matter, refute)
D'Adamo's work.  YET.  I hope you're not saying that the absence of
large studies proves his work is invalid, though, are you? 

"And if the person is going to make a lot of money off of claims that
any particular treatment is going to help, then I REALLY, REALLY want
that person to provide some evidence.

Thanks for the chuckle. 

As one who goes on and off the D'Adamo Type O wagon, I can tell you
that I feel much better on it.   That's been the experience of everyone
I know who has tried his diet for their blood type.  

I think his diet is of particular value for Type A's.  And if you are
a Type A and you should get breast cancer, I hope you'll remember to
read what he says about the helix pomatia snail lectin, despite the
lack of a large study.  
  

    
Maria,

"I'm very glad you are doing well, and that you've lived so long since
the(n)." 

Thanks.

"You wouldn't be a good placebo group, anyway, because, being
self-selected, the odds were good that you were different, possibly in
an important way, from those who accepted treatment.  Random
assignment is crucial for testing placebos versus treatments."

Ah. I'm glad you pointed  that out.  I wish I knew what kinds of
differences you might mean; I mean, our decisions to avoid chemo are so
different -- 

-Some of us have multiple chemical sensitivities.
-A few did chemo once before, then recurred, and came to us.
-One has MS
-I avoid drugs in general because I have such peculiar negative
reactions to them.  Even my oncodoc doubted I could get through any
more than one round of chemo.
-Others saw a friend or loved one suffer with chemo and die anyway,
and decided they'd never allow that to be done to them.  
-Some come from long traditions of avoiding 'doctoring,' and so it
never occurred to them to deal with their bc by any EXCEPT natural
means.  
-Some would have done it if their physicians had been willing to do
vitamin c or other infusions as well as chemo  
-Some were not told they needed chemo in the first place, as their
cancers were small or well-differentiated, or mild in some other ways. 
-We have different blood types, eat different foods, come from
different social, economic, and religious backgrounds... 
-The only area of homogeneity I see is our general level of education:
 most are college graduates, many have MA's, one's an attorney, one's
a nurse, several of us have taught at the college level ..
--We had different cancers with different characteristics, 
--We were diagnosed early and late, young and old.   
- Our youngest member's in her late forties.  Our oldest has an 88th
birthday coming up.  

In short, self-selected though we are, *I*'d call us pretty random. 

What would you look at if you were determining whether we'd serve as a
random group or not?
  

    
I want to thank everyone, and expecially Carole, for a
thought-provocking discussion. However, today Inkwell moves on to a new
discussion. As always, this topic will remain open indefinitely for
further discussion.
  

    
 #120  
Debunix

"I am concerned that iodine might be getting extra play . . . So
anything connected to the thyroid (and iodine is certainly associated
with it) may be hot stuff right now.  
"

Fair enough.  I can see why you might be concerned:  I have been
talking about iodine and the thyroid, there is a connection between
thyroid and breast problems, and Lugol's solution (potassium iodide
plus iodine) is useful for both.  

Please rest assured, though: There is no thyroid bandwagon up my
sleeve.  
  

    
Thanks Julie, and everyone else who participated.  It was a pleasure
to meet and talk with you all. And I do mean ALL.

cb
 
  

    
Thanks, Carole.

Here's a really quick summary of experimental design principles, which
may help explain why random assignment is so crucial in experiments:
<http://www.socialresearchmethods.net/kb/desexper.php>.  

As for the blood type diet, you talk to those who say it works for
them.  That is great, for them as individuals, and for you.  But it
ignores the thousands of people who tried the diet for whom it didn't
work. (To get a handful of these--and there must be many more, since
this was a bestselling book--google "blood type diet didn't work" and
read the anecdotes on that side.)  This is a reason that studies are so
useful (and Dr. D'Amo has had a lot of time to do a study if he really
cared about helping people change their lives--if he were right, it
would be a medical breakthrough of massive proportions).  We all tend
to look for information that confirms our beliefs--it is hard to force
ourselves to look for disconfirming evidence and to take it seriously. 
Studies require at least some discipline--they force researchers to
set out ahead of time what will count as confirming or disconfirming,
and to be as careful as possible about taking possibly confounding
factors into account.

Even the best designed study is bound to have some flaws, but it is
better than the alternatives, if you are talking about giving people
advice on a large scale.  (Saying something worked for you is one
thing; saying that all people who have blood type A would benefit is
where you need a study supporting that, rather than assuming that you
and your type A friends who have tried the blood type diet are
representative of type As in general.)

(The Blood Type diet also has some arguments about the evolution of
blood types that are incorrect--for example, he says that Type O is the
oldest blood type, when actually it is the most recently evolved:
<http://www.npr.org/templates/story/story.php?storyId=18504473>.  This
matters a lot for his theory, since the argument that Type Os need a
high protein diet is based on the argument that they evolved first,
back when humans were mainly hunters.)
  

    
Diet books!   All kinds of incomplete information!   

(On tiny example, the South Beach Diet books say that beer is full of
maltose, which is a highly digestible sugar.  Except that's what the
yeast uses to make alcohol.  So drinking the unfermented beer wort is
going to give a blast of simple sugar to your system, but drinking a
finished beer replaces that particular sugar with alcohol. Some shoddy
nutrition researcher looked up the proportion of sugars in malted
barley and neglected to learn about fermentation. Those little bizarre
credibility errors are extensive in the world of diet books.)
  

    
 #132, and #125. Placebo and random assignment. 

 Though the topic is moving on, yes, let there be no doubt whatsoever,
it is unethical to deny treatment to people. 
Even if those people volunteer. And even if they volunteered, they
would not be a randomized control group.
 
 As to the random assignment here's the reason (others may state it
more eloquently), 
At the point, when medical science is at -equipoise-, the point at
which there is no certainty toward one treatment or another, then it is
time to randomize.

Here's a link to equipoise: 
http://en.wikipedia.org/wiki/Clinical_equipoise

  And randomize does the following.

  Say patients are randomized to two treatment groups (this extends to
any number of groups), then randomization, 
-guarantees- that the only differences between the two groups, other
than treat, are due simply to chance alone. There are no systematic
factors involved.

  And randomization, was, an invention, of R.A. Fischer, though his 
work was in agriculture.

  Randomization in clinical trials, 
 -must-
 be accompanied by an informed consent, where the patient knows and
understands in advance the known risks.

 The principles of informed consent, always refer to the Declaration
of Helsinki.

http://en.wikipedia.org/wiki/Declaration_of_Helsinki

http://www.wma.net/en/30publications/10policies/b3/index.html

  Past bad experience with experiments on humans were in World War II
at the concentration camps.
  

    
Unfortunately, similar lapses in ethics occurred after WWII right here
in the US of A.

<http://www.cbsnews.com/8301-503543_162-20018272-503543.html>

Clinical research involving humans is governed by Title 45 CFR (Code
of Federal Regulations) Part 46.

<http://www.hhs.gov/ohrp/documents/OHRPRegulations.pdf>

Developed partly in response to the unethical trials referred to
above.

Chris, that is one of the best and concise explanations of equipoise
I've ever seen. 
  

    
 thx. The randomized trial has a couple additional aspects.
 First the randomization is a coin toss (or the electronic version of
the coin toss).

 And the best randomized trials are double blind.

  That is, neither the patient nor the doctor know what
 treatment the patient gets.
   Sometimes, its impossible to blind, say, if the treatment is 
 surgical removal of a section of the colon.

 The blinding eliminates the possibility of bias, that is, the doctor
somehow changes how they treat, based on what they know about the
treatment and the disease.

  The first big randomized trials after the war, were the Salk Polio
vaccine trials.
  

    
I agree with you Chris, with the caveat that I think effect size* is
one of the most important factors to take into account when
interpreting the outcome of a clinical trial, in addition to the trial
design itself.

The biomedical research literature is peppered with clinical trials
and epidemiologic reports which when deconstructed show absolutely
trivial and therefore in a sense meaningless effect sizes.

The rejection of good observational data in favor of evidence gathered
via a well-designed trial with good power, an issue brought up in this
topic, has its downside.

A humorous treatment of the subject is found in this paper: Smith GCS,
Pell JP. (2003). Parachute use to prevent death and major trauma
related to gravitational challenge: systematic review of randomised
controlled trials. BMJ, 327(7429), 1459-1461. DOI:
10.1136/bmj.327.7429.1459

Some excerpts:

"Advocates of evidence-based medicine have criticised the adoption of
interventions evaluated by using only observational data."

"We think that everyone might benefit if the most radical protagonists
of evidence-based medicine organised and participated in a
double-blind, randomised, placebo-controlled, crossover trial of the
parachute."

The Monty Python flourish at the end:

"GCSS had the original idea. JPP tried to talk him out of it. JPP did
the first literature search but GCSS lost it. GCSS drafted the
manuscript but JPP deleted all the best jokes. GCSS is the guarantor,
and JPP says it serves him right."

Staph aureus was exquisitely sensitive to penicillin in the 40's, huge
effect size - so that the first trial of the drug involved an n of 3
without a placebo arm. It would have been ridiculous, and in some ways,
unethical, to wait for for a good clinical trial before proceeding
with its use.

* <http://en.wikipedia.org/wiki/Effect_size>
  

    
And then there's the problem of  hiding behind the placebo control
study, which is what happens with "me-too drugs." Some people would
like to pit drugs for which there is already proof of concept, like
serotonin-enhancing antidepressants, against each other, so that Paxil,
say, would have to show benefit over Zoloft, rather than only over
placebo, in order to gain approval by FDA. 

On the other hand, makers of drugs with lousy signal (high placebo to
drug response ratio) would love to drop the placebo controls. MAkes
them look bad.

As for Fischer, the British agronomist who invented modern statistics,
he was nearly blind, and did most ofhis math in his head. Try that at
home.
  

    
Yes indeed - the effect size for many antidepressants is relatively
small. Yet there seems to be an emerging consensus that stratifying
population response to antidepressants by allelic polymorphisms in the
function and/or expression of proteins involved in certain
neurotransmitter pathways may greatly improve the signal to noise; as
well as giving some important guidelines for personalizing
antidepressant therapy.   

Fisher was a giant in the field. He formulated the theory of fiducial
inference, which is one method of statistical inference. But assigning
him the invention of modern statistics is stretching it a bit, I think.
  

    
Right. That was sloppy. What Fisher did was to figure out, along with
some other researchers, how to design experiments that could use the
power of statistics to discern probabilities and distinguish outcomes
from random chance. 

There's a really cool paper by Ian Hacking abou Fisher. I'll try to
dig up the reference.

As for using proteomics to strengthen the antidepressant signal, my
money is on this method, but with one huge predicted caveat. The people
who have those expressions will turn out to be a tiny subset of the
people who meet the clinical criteria for depression, which means that
if depression is defined by biomarkers, and antidepressant use is
pegged to the results, the market is goiung to shrink tremendously. 
  

    
I agree with your assessment of the impact of personalized medicine /
genomics on antidepressant therapy. I would quibble about the number of
people who are loaded with "risk" alleles based on antidepressant x
allelic variation response studies.  It's probably higher than a tiny
subset.

I almost got into the genetic testing business for allelic variants
that might influence behavior and antidepressant response.  Back when I
was doing bench science at NIH my lab found that combinations of
allelic variations influenced the risk for binge drinking.* Similar
variations are associated with risk of depression and antidepressant
response to the various classes of antidepressants.

I decided not to get into the business because the CLIA oversight for
genetic testing labs is non-existent, and the competition does not
perform adequate quality control - ergo it is not possible to make a
profit if one's goal is to do the proper quality assurance.

The other reason I jettisoned the idea is that pharmaceutical
manufacturers, for the reasons you state, do not want to restrict their
market based on allelic variations to antidepressant response, hence
they are not supportive of such testing in Phase II trials.

*
<http://alcalc.oxfordjournals.org/content/38/5/446.long>

<http://onlinelibrary.wiley.com/doi/10.1002/ajmg.b.30135/abstract;jsessionid=CD
D118305D26BEEE4CA4D29BC65BB995.d03t01>