I have prepared an article on the misuse of statistics in medical
research. If the subject is of interest to you or if you work with
cancer, you might be interested to read it. Your opinion would be
appreciated.

To read the article, click here:
http://www.geocities.com/benny_patrick/stats.html

Fabrikant

Report this post to a moderator

[  Post Follow-up to this message ]

"Fabrikant" <benny_patrick7@hotmail.com> wrote in message
news:f855b5d5.0503011429.5e40fa05@posting.google.com..
>I have prepared an article on the misuse of statistics in medical
> research. If the subject is of interest to you or if you work with
> cancer, you might be interested to read it. Your opinion would be
> appreciated.
>
> To read the article, click here:
> http://www.geocities.com/benny_patrick/stats.html
>
> Fabrikant

Some of your complaints, such as the misuse of relative risk and the quite
small benefits from certain medical interventions are valid.  But this does
not look right to me ---

"Now let us see, how the study results deal with the pitfalls mentioned in
the previous section.   We do not have to deal here with stability of the
basic data, because on the page 784 of the publication [3] it is mention
ed
that at three years of study, 81% of patients in perindopril group were
actually taking the medication.  This means, that 19% of the perindopril
group did not take the medication and de-facto belonged to the placebo
group.  The number of patients in perindopril group has to be re-calculated
as n2=6110*0.81=4950 and the difference of 1160 patients added to the
placebo group, so that n1=6108+1150=7268.  The percentages are computed as
p1=603/7268=8.3% and  p2=488/4950=9.9%.  Now the placebo group is safer than
the perindopril group and their difference p1-p2= -1.6%, according to
formula (1), is statistically significant with the confidence interval
greater than 99.7%.  Does this prove that perindopril is dangerous for the
health?  Of course, not, but it does prove, just how shaky is the basis, on
which the authors of the study made their conclusions. "

Those patients not taking the drug at three years would presumably have
taken the drug for a while, and it destroys the whole point of randomisation
to later transfer them to the placebo group.   It CANNOT be done.   It is
proper to include them in the "treated" arm, even if doing so actually has
the effect of diluting any benefits from taking the drug.

Dilutes the benefits?   Yes.  You are deceptively including them in the
placebo group without knowing, or allowing for,  what their primary outcomes
were.   If these patients showed the same morbidity as the rest of the
placebo arm then the figure of 603 in the above calculation is quite wrong -
it should be increased by (9.9% of  1160) =  116), making a mockery of your
calculations.

Peter Moran. .

Report this post to a moderator

[  Post Follow-up to this message ]

"Peter Moran" <moringa@gil.com.au> wrote in message
news:42250f2c$0$246$61c65585@uq-127creek-reader-03.brisbane.pipenetworks.com.au..
>
> "Fabrikant" <benny_patrick7@hotmail.com> wrote in message
> news:f855b5d5.0503011429.5e40fa05@posting.google.com.. 
>
> Some of your complaints, such as the misuse of relative risk and the quite
> small benefits from certain medical interventions are valid.  But this
> does not look right to me ---
>
> "Now let us see, how the study results deal with the pitfalls mentioned in
> the previous section.   We do not have to deal here with stability of the
> basic data, because on the page 784 of the publication [3] it is menti
oned
> that at three years of study, 81% of patients in perindopril group were
> actually taking the medication.  This means, that 19% of the perindopril
> group did not take the medication and de-facto belonged to the placebo
> group.  The number of patients in perindopril group has to be
> re-calculated as n2=6110*0.81=4950 and the difference of 1160 patients
> added to the placebo group, so that n1=6108+1150=7268.  The percentages
> are computed as p1=603/7268=8.3% and  p2=488/4950=9.9%.  Now the placebo
> group is safer than the perindopril group and their difference
> p1-p2= -1.6%, according to formula (1), is statistically significant with
> the confidence interval greater than 99.7%.  Does this prove that
> perindopril is dangerous for the health?  Of course, not, but it does
> prove, just how shaky is the basis, on which the authors of the study made
> their conclusions. "
>
> Those patients not taking the drug at three years would presumably have
> taken the drug for a while, and it destroys the whole point of
> randomisation to later transfer them to the placebo group.   It CANNOT be
> done.   It is proper to include them in the "treated" arm, even if doing
> so actually has the effect of diluting any benefits from taking the drug.
>
> Dilutes the benefits?   Yes.  You are deceptively including them in the
> placebo group without knowing, or allowing for,  what their primary
> outcomes were.   If these patients showed the same morbidity as the rest
> of the placebo arm then the figure of 603 in the above calculation is
> quite wrong - it should be increased by (9.9% of  1160) =  116), making a
> mockery of your calculations.
>
> Peter Moran. .

BTW You are also wrong about breast cancer.   In every advanced country the
mortality rate is distinctly decreasing even as the incidence continues to
rise.   You can confirm with the national statistical agencies of England
America, and Australia.

Peter Moran

>
>
>
>
>
>

Report this post to a moderator

[  Post Follow-up to this message ]

"Peter Moran" <moringa@gil.com.au> wrote in message
news:422511b9$0$263$61c65585@uq-127creek-reader-03.brisbane.pipenetworks.com.au..
>
> "Peter Moran" <moringa@gil.com.au> wrote in message
> news:42250f2c$0$246$61c65585@uq-127creek-reader-03.brisbane.pipenetworks.c
om.au.. 
>
> BTW You are also wrong about breast cancer.   In every advanced country
> the mortality rate is distinctly decreasing even as the incidence
> continues to rise.   You can confirm with the national statistical
> agencies of England America, and Australia.
>
> Peter Moran


Although this is largely a stage migration effect, Peter...

Report this post to a moderator

[  Post Follow-up to this message ]

"Steph" <steph@vancouver.island> wrote in message
news:MaaVd.547410$8l.112466@pd7tw1no..
>
> "Peter Moran" <moringa@gil.com.au> wrote in message
> news:422511b9$0$263$61c65585@uq-127creek-reader-03.brisbane.pipenetworks.c
om.au.. 
>
>
> Although this is largely a stage migration effect, Peter...

No it's not.   These are absolute mortality figures.  These are the number
of people per hundred thousand of population who end up with "breast cancer"
as the cause of death  on their death certificates.

You are thinking of survival rates, which will be affected by overall
earlier diagnosis and other factors.

Peter Moran

>
>

Report this post to a moderator

[  Post Follow-up to this message ]

"Peter Moran" <moringa@gil.com.au> wrote in message
news:4225610c$0$258$61c65585@uq-127creek-reader-03.brisbane.pipenetworks.com.au..
>
> "Steph" <steph@vancouver.island> wrote in message
> news:MaaVd.547410$8l.112466@pd7tw1no.. 
>
> No it's not.   These are absolute mortality figures.  These are the number
> of people per hundred thousand of population who end up with "breast
> cancer" as the cause of death  on their death certificates.
>
> You are thinking of survival rates, which will be affected by overall
> earlier diagnosis and other factors.
>
> Peter Moran
> 
>
>

No, if the proportion of early stage cancers diagnosed goes up (as it has
due to mammography and other screening), absolute mortality goes down.
Stage for stage, the 5, 10 and 15 year survival for breast cancer is only
marginally better than it was 20 years ago, if at all

Report this post to a moderator

[  Post Follow-up to this message ]

"Steph" <steph@vancouver.island> wrote in message
news:AleVd.552854$6l.437552@pd7tw2no..
>
> "Peter Moran" <moringa@gil.com.au> wrote in message
 
[vbcol=seagreen]
>
> No, if the proportion of early stage cancers diagnosed goes up (as it has
> due to mammography and other screening), absolute mortality goes down.

Yes.  That is usually called earlier diagnosis, and it is one probable
factor in the declining mortality .   Stage migration is usually used to
refer to spurious stage-specific benefits, when there is no true reduction
in either overall or individual mortality.

> Stage for stage, the 5, 10 and 15 year survival for breast cancer is only
> marginally better than it was 20 years ago, if at all

Then where is the declining mortality coming from?   There must be fairly
significant improvements in survival rates in early stage breast cancer to
produce the figures shown in  the graph at
http://www.statistics.gov.uk/cci/nugget.asp?id=575.

Stated there:  "Death rates gradually increased up to the mid 1980s and then
began to fall around the time that screening started. By 1998 mortality was
around 20 per cent lower than it would have been (based on predictions of
pre-screening rates in various age groups)."

The widening disparity between incidence and mortality of breast cancer
could have many contributing factors:  earlier diagnosis, better treatment,
changes in breast cancer biology, and even a small contribution from
overdiagnosis in marginal pathology, but there is no disputing the declining
death rates.

Peter Moran


>
>

Report this post to a moderator

[  Post Follow-up to this message ]

"Peter Moran" <moringa@gil.com.au> wrote in message
news:42261feb$0$678$61c65585@uq-127creek-reader-01.brisbane.pipenetworks.com.au..
>
> "Steph" <steph@vancouver.island> wrote in message
> news:AleVd.552854$6l.437552@pd7tw2no.. 
> 
> 
>
> Yes.  That is usually called earlier diagnosis, and it is one probable
> factor in the declining mortality .   Stage migration is usually used to
> refer to spurious stage-specific benefits, when there is no true reduction
> in either overall or individual mortality.
> 
>
> Then where is the declining mortality coming from?   There must be fairly
> significant improvements in survival rates in early stage breast cancer to
> produce the figures shown in  the graph at
> http://www.statistics.gov.uk/cci/nugget.asp?id=575.
>
> Stated there:  "Death rates gradually increased up to the mid 1980s and
> then began to fall around the time that screening started. By 1998
> mortality was around 20 per cent lower than it would have been (based on
> predictions of pre-screening rates in various age groups)."
>
> The widening disparity between incidence and mortality of breast cancer
> could have many contributing factors:  earlier diagnosis, better
> treatment, changes in breast cancer biology, and even a small contribution
> from overdiagnosis in marginal pathology, but there is no disputing the
> declining death rates.
>
> Peter Moran
>
>

But the biggest factor is that the proportion of early (and VERY early)
cancers compared to the total is now much higher than 20 years ago.
More early cancers in the group = better overall mortality in the group.
That's a stage-migration effect in this colony, at least.....

Report this post to a moderator

[  Post Follow-up to this message ]

"Steph" <steph@vancouver.island> wrote in message news:<8%uVd.558599$Xk.4455
91@pd7tw3no>..

>
> But the biggest factor is that the proportion of early (and VERY early)
> cancers compared to the total is now much higher than 20 years ago.
> More early cancers in the group = better overall mortality in the group.
> That's a stage-migration effect in this colony, at least.....

Your dispute with Moran would benefit, if both of you operate with
facts and numbers, rather than having a discussion of the kind "is
not!" - "is so!" I have looked at the ACS data from 1930 to 1990.  The
mortality practically did not change during 60 years, variating
between 25 to 28 per 100,000 of population, and this variation looks
quite random, clearly independent from the level of mammography and/or
any medication.

Seekers of success though found a reason to celebrate: they claim that
the increase of number of diagnosed cases, which did not result in
increased mortality is a 20% success.  How about a simple explanation
that the increased diagnostics is due to increase in false positive?
Nobody knows, how many perfectly healthy women lost their breasts for
nothing and live under illusion that they survived breast cancer?
What is the probability that the early diagnosis and treatment would
save almost exactly the number of lives, which would leave the total
mortality almost unchanged?

Fabrikant

Report this post to a moderator

[  Post Follow-up to this message ]

>
> Then where is the declining mortality coming from?   There must be fairly
> significant improvements in survival rates in early stage breast cancer to
> produce the figures shown in  the graph at
> http://www.statistics.gov.uk/cci/nugget.asp?id=575.
>
> Stated there:  "Death rates gradually increased up to the mid 1980s and th
en
> began to fall around the time that screening started. By 1998 mortality wa
s
> around 20 per cent lower than it would have been (based on predictions of
> pre-screening rates in various age groups)."
>
> The widening disparity between incidence and mortality of breast cancer
> could have many contributing factors:  earlier diagnosis, better treatment
,
> changes in breast cancer biology, and even a small contribution from
> overdiagnosis in marginal pathology, but there is no disputing the declini
ng
> death rates.
>
> Peter Moran
>

Please, read my article attentively: I do discuss the argument that
the transfer of some patients from medicated group to placebo group
involves possible transfer of the number of outcomes.  My point is not
uniquely based on this transfer, but rather on instability of data,
where a +/-5% of change in the initial data changes the conclusion.  I
also show that 2% difference between groups can be easily explained by
reasons, other than the study medication.  So, I suggest that you take
the article as a whole and address all the argumentation.

About success being diluted.  I remind you the main proposition of
null-hypothesis: you have to presume the medication to be USELESS,
unless proven otherwise.  So, you can not presume, that those, who do
not take medication are in worse situation that those, who do.

Fabrikant

Report this post to a moderator

[  Post Follow-up to this message ]