hi arnie,

my apologies for the delay, as well. a

    > First, it was not my idea to compare the incidence of PG (pathological gambling)

    Arnie> True, but neither draws the conclusion which Mackenzie does, that there is "zero evidence" for any relationship between DAs and PG. Rather, they think it is a real effect and are trying to put the rate of occurrence in context.

it is either ok to derive meaning, or it is not. if it is ok for them to, then it is ok for me to, particularly when the datasets from which the meaning is derived are exactly the same, and when the authors have given the comparison of the two groups,, uncontrolled though they be, their blessing by comparing them in the first place.

to say that the incidences are not appropriate for statistical comparison because they are derived from groups that are not controlled in any way, while at the same time sanctioning the authors’ direct comparison of those very same incidences, from the very same uncontrolled groups, is illogical.

therefore, on the basis of control or lack thereof for demographics, etc., either both methods of comparison are appropriate or neither of them is.

And Dr. Stacy unquestionably makes straight comparisons among the various incidences directly and actually acknowledges not only that the difference is vanishingly small:

“Because the reported incidence of pathologic gambling in the general population ranges from 0.3% to 1.3%, the 1.5% incidence found in the pramipexole therapy group may not be unexpected. Certainly, the overall incidence of excessive gambling in PD, including the 752 subjects prescribed pergolide and ropinirole, in this retrospective, uncontrolled analysis, does not exceed the accepted general population expectations.�

But he then goes on to articulate a huge risk factor that could easily account for such a small difference:

“In addition, the availability of casinos in this retirement and vacation setting may allow subjects to discover this behavior after developing PD.�

If one is going to derive significance from the variable age, for example, as you do, i.e., the PD gamblers are older than your average novice pathological gambler and therefore there is likely something other than natural inclination at the root of the onset of the behavior then one *must also* derive signifiicance from other variables, including proximity of casinos, history of depression, the fact that every single one of these people just underwent an adjustment to their meds (from which it would be logical to extrapolate that they have experienced a deterioration sufficient to require more meds - which is depressing!)

To say that this variable has significance and another does not strikes me as illogical.

it also strikes me as illogical to take such a tiny difference as this, directly state a risk factor that could easily acccount f or such a miniscule difference, and STILL go on to imply that there is an association - but Dr. Stacy does just that:

“However, these subjects illustrate that excessive gambling may occur in advancing PD�

Please note the non-committal nature of the wording - he does not *say* he found an association - he just says “gambling could happen.� well, *anything* could happen.

Nonetheless, he goes on to recommend that doctors warn their patients of the risk:

“it may be appropriate to inform subjects of a potential risk of this behavior. “

i’m sorry but i don’t see the logic - but given the following statement, made by Dr. Stacy in his study, perhaps that is not so surprising:

“Given the complex world of the patient with PD— dopaminergic therapy, chronic illness, and casino availability—�

casino?? what casino? my PD didn’t come with  a casino - did yours? i demand to speak with the management!

So to sum up, if you are going to take issue with the uncontrolled nature of the groups being compared, that issue applies whether the comparison is direct or statistical. Since you clearly do not take issue with straight comparison, then the disqualifying of statistical significance must be made on a basis other than the uncontrolled nature of the groups.

So that leaves the significance and validity of statistical significance itself, and given the non-existent nature of a statistically significant difference (i.e., low likelihood that the difference (or lack thereof) is due to chance - significance), along with the fact that  in medical research, incidences are routinely compared to the general population (validity), and, perhaps most telling, in his study, Dr. Stacy does not ever say that he found an association, i think it would be difficult to argue that the absence of a statistically significant difference could be masking a significant difference.

Arnie Kuzmack <[log in to unmask]> wrote: Sorry for the delay in responding.  Doing my taxes took me
down to the wire this year.  Let me address each of
Mackenzie�s points.


> First, it was not my idea to compare the incidence of PG
> (pathological gambling) among those on DAs to that of the
> general population - Both Drs. Stacy and Voon (the April
> 5th study) held up the indcidence and prevalence,
> respectively, in the general population for comparison.
> Stacy does not calculate signficance, but he does compare
> the two numbers - and derives meaning from the one being
> marginally higher than the other. Voon does the same.


Arnie> True, but neither draws the conclusion which
Mackenzie does, that there is "zero evidence" for any
relationship between DAs and PG.  Rather, they think it is a
real effect and are trying to put the rate of occurrence in
context.


> Second, perhaps i misunderstand you, but it seems to me
> that you are saying two things.
>
> first, it seems you are saying that statistical
> significance is insignificant in that it tells us nothing
> more than the degree to which certain results may or may
> not have been influenced by random fluctuations, and that
> tells us absolutely nothing of worth about the sources
> from which the raw data was drawn.
>
> If i am understanding you correctly, then i have to ask
> why the calculation is undertaken as a matter of course in
> countless studies, and even more basic than that, why
> bother inventing (?) a calculation that merely give us a
> number - say, 42 -  from which we may derive no meaning
> whatsoever?


Arnie> Yes, all that a test of statistical significance does
is determine whether a result might be due to chance.  Check
any statistics textbook, or see one of these websites or
dozens of others you can find by googling "statistical
significance":

http://en.wikipedia.org/wiki/Statistical_significance
http://www.statpac.com/surveys/statistical-significance.htm

I am not saying that tests of statistical significance are
useless in general � they would take away my degree if I
said that � but that they need to be used appropriately.

I suspect I did not explain this very well, so I will try
again.  Ruling out random fluctuations as the cause of an
empirical result is essential to being able to rely on it.
But the comparison needs to be appropriate, and there are
lots of ways one can go astray.  For example, children who
grow up in older housing which has lead-based paint do more
poorly in school than the population as a whole.  But there
are many other differences besides lead-based paint: rheir
parents are less well educated, they are poorer, fewer of
them have intact families, etc.  If you don�t take account
of all these other factors, you will not be able to
correctly determine the impact of lead-based paint (which
is, in fact, very important, by the way).

If there does not appear to be an association between two
variables, there are several possible reasons: there really
is no relationship; there is a relationship, but it is
smaller than you thought, so you need a larger sample size;
and, there is a relationship, but it is being masked by
other factors that you are not considering but which impact
the factors you are looking at.  An example of the third
type (this is made up, so don�t go quoting it):  The total
number of deaths from cancer in the population remains about
the same, despite progress in treating it.  Other factor:
people are living longer, so more of them are reaching the
ages where cancer is more common.

The general principle here is to be careful when comparing
two different populations or two different variables, since
they may differ in ways you did not anticipate.  This is
where the problem is in the "lack of statistical
significance" in the Parkinson�s pathological gambling
studies, as I will discuss below.


> second, it seems to me you may be saying that beyond the
> uselessness of calculating significance itself, the two
> groups are not appropriate for comparison because, and
> please correct me if i have this wrong, PG among PWP is
> associated with the starting and stopping of DAs, and PG
> among the general population is not. could you clarify
> that for me?


Arnie> Exactly.  I�m glad *something* I wrote was clear.


> because such comparisons are standard procedure  - take
> the question of suicide risk among cancer patients -
> epidemiological studies have indicated that the risk of
> suicide in cancer patients is higher than that of the
> general population.  suicide is to the onset of cancer
> as gambling is to the onset of dopaminergic medication (in
> theory, at least)- and both suicide and gambling occur in
> the general population as well.  in what way is the
> comparison stacy and voon make different from cancer
> comparison?


Arnie> I am not familiar with the literature on suicide risk
among cancer patients, but I would assume that the studies
follow standard practice by using a control group or
comparison group that is matched or adjusted for variables
known to affect suicide risk, such as age, gender,socio-
economic status, etc.  Thus, the comparison is between
cancer patients and a group of non-patients who are as
similar to them as the investigators can arrange.

In contrast, the comparisons quoted by Mackenzie are between
the occurrence rates among the subpopulation who are PWPs
getting a particular medication compared to the whole
population.  The statements were made to indicate the
general magnitude of the effects reported and did not
pretend to be a complete, detailed analysis.  Thus, the
distribution of demographic variables is quite different
between the PWP group and the comparison group.
Pathological gambling has been studied, and, as I wrote
earlier, it usually starts in adolescence or early adulthood
and builds up over many years until it reaches the point of
being "out of control".  There typically does not seem to be
any particular event that sets off the behavior, and it
continues until treatment is sought and is successful.
Treatment is usually similar to that for alcohol addiction.
Overcoming it is a very difficult, lifelong struggle.

By using these comparisons to discount the reality of the
experiences of PWPs in these studies, Mackenzie is basically
arguing that they are not real *because* a similar
proportion of young people without PD experience a
long-term addiction to gambling.  Obviously, this is not the
intention, but it�s what the argument from lack of
statistical significance means.  According to this argument,
a few percent of PWPs taking DAs experience pathological
gambling, mostly  in their 50s or higher, with sudden onset
and rapid resolution, while a few percent of young people
experience a long-term addiction to gambling.  Since the
percentages are similar, there is argued to be "no
statistically signficant difference" and hence "zero
evidence" of an effect.

If someone were to make a comparison of suicide rates among
cancer patients with suicide in the total population *and*
use it to discount the seriousness of the problem, that
reasoning would be equally questionable.

I hope I have made it clear that pathological gambling as a
result of treatment with DAs is very different from PG in
the general population.  They start at different ages, have
a different pattern of development, and are totally
different in their patterns of treatment and the efficacy of
treatment in bringing them under control.  Therefore, it
does not matter that the proportions of the very different
populations that are affected by each happen to be similar.


> and as far as the difficulty determining whether anyone
> has the behavior because of the tendency to lie about it -
> people lie about suicidal thoughts, about binge eating,
> about starving themselves, about pulling their hair out -
> people lie about lots of things and still studies are
> done, such obstacles and variables are taken into account
> and conclusions are drawn. this situation is no different.


Arnie> True, but they are taken into account by recognizing
that the study is likely to underestimate the rate of
occurrence of the behavior in question and by reflecting
this in assessing the meaning of the low rates documented
in the study.


> regarding the statement that no study has shown an
> association btw gambling and levodopa monotherapy, i will
> actually put it in all caps - THE MOLINA STUDY found an
> incidence of 4-4.5% of PG among those on levodopa - no
> other drug is mentioned - just levodopa - and for some
> reason it is like that study is invisible.
>
> anyone wants to read that study, i have it, happy to send
> it.


Arnie> I would like to see that study.  I have not gotten
ahold of the full report, but I have seen the abstract and
the descriptions by Dr. Dodd, et al.  As is noted, no other
drug is mentioned, but it does not say that no other drug
was used.  In fact, Dr. Dodd states that "adjunctive drugs
were employed but not listed".  Of course, if it turns out
that Molina�s patients were receiving levodopa monotherapy,
I would have no problem dropping that part of my argument.
And, of course, this may be overtaken by events if the study
by Voon is replicated and a clear pattern emerges of which
drugs are most likely to have which effects.


> however, i totally agree with your assertion that whether
> or not it is a *real* effect is of tantamount importance,
> along with the mechanism of causation (is it mirapex, or
> DAs, or levodopa, or some combination thereof?) I also
> agree that the frequency of use of a particular drug must
> be taken into account when calculating the incidence of a
> certain behavior occuring on that drug vs other drugs of
> its kind.
>
> but, i disagree that there is substantial evidence to
> support the existence of the association we are
> discussing - the Voon abstract *seems* to address some of
> the as yet unaddressed issues, but i am going to wait to
> read the full study before coming to any conclusions about
> that - and until then, the stacy, dodd and szarfman
> studies are all we have and they remain just as flimsy as
> they have been - in my opinion


Arnie> It is generally a good idea to read the full study
and not rely on abstracts and press reports.

While I certainly remain open to new data, I think the
existing studies (even before Voon) show a consistent
pattern of behavior that includes onset of pathological
gambling after treatment with some PD drug or drugs, at ages
usually in the 50s or later, sometimes accompanied by
hypersexuality and pathological shopping, which resolves
after stopping the medication or reducing the dose.  None of
these are observed in typical cases of PG in the general
population.

I would not change the conclusion of my previous
post:  "The evidence is less convincing than a randomized
double-blind trial but goes well beyond isolated anecdotal
reports."

> anyway - if your takeaway were the final word and the goal
> of all of this sound and fury were to alert folks to this
> "danger," i think it is VERY safe to say "mission
> accomplished."
>
> would it be nice if that were the final word? but it's
> not - not by a long shot - this is going to spawn dozens
> of studies (time will tell if i am right) and syphon money
> away from research that could really help people - not to
> mention the danger of people being frightened away from
> DAs and into the arms of that most excellent drug,
> levodopa.

Arnie> People have lost their life savings and destroyed
their marriages as a result of this behavior.  If these
results can be prevented, that certainly seems to me to be
"really helping people".  I agree that it would be
unfortunate if people are frightened away from DAs, but I do
not see any evidence that this is actually happening.

----------------------------------------------------------------------
To sign-off Parkinsn send a message to: mailto:[log in to unmask]
In the body of the message put: signoff parkinsn



---------------------------------

----------------------------------------------------------------------
To sign-off Parkinsn send a message to: mailto:[log in to unmask]
In the body of the message put: signoff parkinsn
Yahoo! Messenger with Voice. Make PC-to-Phone Calls to the US (and 30+ countries) for 2¢/min or less.