Diabetes drugs and combinations of drugs have no effect on all causemortality

x-no-archive: yes


Antihyperglycemic Agents Have No Effect on All-Cause Mortality
No single agent or combination of oral antihyperglycemic agents has any
significant effect on all-cause mortality, according to a new study.

Those findings come from a study of diabetics in the Veterans Health
Administration (VHA), led by Dr. Kristijan H. Kahler of the VA New
Jersey Health Care System in East Orange, and Novartis Pharmaceuticals
in East Hanover, New Jersey, and are published in the July issue of
Diabetes Care.

The team has been studying 39,721 diabetics divided into groups
according to oral antihyperglycemic therapy: sulfonylurea monotherapy,
metformin monotherapy, metformin plus sulfonylurea, thiazolidinedione
(TZD) monotherapy or in combination with other oral agents (TZD users),
and a group on no drug therapy.

The primary outcome measure was all-cause mortality.

The VHA study has been ongoing since 1996. During its first decade, the
adjusted odds ratios for all-cause mortality were 0.87 for metformin
monotherapy users, 0.92 for metformin plus sulfonylurea users, and 1.04
for TZD users, relative to sulfonylurea monotherapy users.

"We did not find any significant drug effect on all-cause mortality for
any oral treatment cohorts relative to sulfonylurea oral monotherapy,"
Dr. Kahler and associates report.

"Future work should assess whether long-term exposure to oral
antihyperglycemic medications reduces all-cause or cause-specific
mortality," the investigators note.
Diabetes Care 2077;30:1689-1693.

===============================

"Susan" <nevermind@nomail.com> wrote in message
news:5g7oiuF3fm309U1@mid.individual.net...
> x-no-archive: yes
>
>
> Antihyperglycemic Agents Have No Effect on All-Cause Mortality
> No single agent or combination of oral antihyperglycemic agents has any
> significant effect on all-cause mortality, according to a new study.
>
> Those findings come from a study of diabetics in the Veterans Health
> Administration (VHA), led by Dr. Kristijan H. Kahler of the VA New Jersey
> Health Care System in East Orange, and Novartis Pharmaceuticals in East
> Hanover, New Jersey, and are published in the July issue of Diabetes Care.
>
> The team has been studying 39,721 diabetics divided into groups according
> to oral antihyperglycemic therapy: sulfonylurea monotherapy, metformin
> monotherapy, metformin plus sulfonylurea, thiazolidinedione (TZD)
> monotherapy or in combination with other oral agents (TZD users), and a
> group on no drug therapy.
>
> The primary outcome measure was all-cause mortality.
>
> The VHA study has been ongoing since 1996. During its first decade, the
> adjusted odds ratios for all-cause mortality were 0.87 for metformin
> monotherapy users, 0.92 for metformin plus sulfonylurea users, and 1.04
> for TZD users, relative to sulfonylurea monotherapy users.
>
> "We did not find any significant drug effect on all-cause mortality for
> any oral treatment cohorts relative to sulfonylurea oral monotherapy," Dr.
> Kahler and associates report.
>
> "Future work should assess whether long-term exposure to oral
> antihyperglycemic medications reduces all-cause or cause-specific
> mortality," the investigators note.
> Diabetes Care 2077;30:1689-1693.
>
> ===============================

Um... Dur... What exactly does this mean?

On Thu, 19 Jul 2007 01:47:31 GMT, "Julie Bove" <juliebove@verizon.net>
wrote:

>Future work should assess whether long-term exposure to oral
>> antihyperglycemic medications reduces all-cause or cause-specific
>> mortality," the investigators note.
>> Diabetes Care 2077;30:1689-1693.

This is along the lines Jim was asking about yesterday....

Will, T2

"Will, T2" <wmmckee@cox.net> wrote in part:

>On Thu, 19 Jul 2007 01:47:31 GMT, "Julie Bove" <juliebove@verizon.net>
>wrote:
>
>>Future work should assess whether long-term exposure to oral
>>> antihyperglycemic medications reduces all-cause or cause-specific
>>> mortality," the investigators note.
>>> Diabetes Care 2077;30:1689-1693.
>
>This is along the lines Jim was asking about yesterday....
>
>Will, T2

Yes. But I wouldn't say the subject chosen ("Diabetes drugs and combinations
of drugs have no effect on all cause mortality") describes the result.

First, all patients were treated and the comparison is with one of the
treatments, not with a placebo.

Second, metformin produced a 13% reduction in ALL-CAUSE mortality in just
TEN YEARS. The result apparently is not statistically significant (may have
occurred by chance), but it is most definitely clinically significant (of
benefit to the patient) if it is not a chance result.

The study may not be sufficiently powered (not enough patients in each group
or not enough duration). If the study is continuing, it may well conclude
after the second decade that metformin was superior to sulfonylurea
monotherapy.
--
Jim Chinnis Warrenton, Virginia, USA

On Thu, 19 Jul 2007 03:01:24 GMT, Jim Chinnis
<jchinnis@SPAMalum.mit.edu> wrote:

>"Will, T2" <wmmckee@cox.net> wrote in part:
>
>>On Thu, 19 Jul 2007 01:47:31 GMT, "Julie Bove" <juliebove@verizon.net>
>>wrote:
>>
>>>Future work should assess whether long-term exposure to oral
>>>> antihyperglycemic medications reduces all-cause or cause-specific
>>>> mortality," the investigators note.
>>>> Diabetes Care 2077;30:1689-1693.
>>
>>This is along the lines Jim was asking about yesterday....
>>
>>Will, T2
>
>Yes. But I wouldn't say the subject chosen ("Diabetes drugs and combinations
>of drugs have no effect on all cause mortality") describes the result.
>
>First, all patients were treated and the comparison is with one of the
>treatments, not with a placebo.
>
>Second, metformin produced a 13% reduction in ALL-CAUSE mortality in just
>TEN YEARS. The result apparently is not statistically significant (may have
>occurred by chance), but it is most definitely clinically significant (of
>benefit to the patient) if it is not a chance result.
>
>The study may not be sufficiently powered (not enough patients in each group
>or not enough duration). If the study is continuing, it may well conclude
>after the second decade that metformin was superior to sulfonylurea
>monotherapy.


Thanks, Jim

Another thing I appreciate about you is that you think. You really
think. And, it should make some of the rest of us slow down and think,
as well...

Will, T2

In article <tkkt93tbtul6ub0kje9pn5ovvqghf52qj7@4ax.com>,
Jim Chinnis <jchinnis@SPAMalum.mit.edu> wrote:

> "Will, T2" <wmmckee@cox.net> wrote in part:
>
> >On Thu, 19 Jul 2007 01:47:31 GMT, "Julie Bove" <juliebove@verizon.net>
> >wrote:
> >
> >>Future work should assess whether long-term exposure to oral
> >>> antihyperglycemic medications reduces all-cause or cause-specific
> >>> mortality," the investigators note.
> >>> Diabetes Care 2077;30:1689-1693.
> >
> >This is along the lines Jim was asking about yesterday....
> >
> >Will, T2
>
> Yes. But I wouldn't say the subject chosen ("Diabetes drugs and combinations
> of drugs have no effect on all cause mortality") describes the result.
>
> First, all patients were treated and the comparison is with one of the
> treatments, not with a placebo.
>
> Second, metformin produced a 13% reduction in ALL-CAUSE mortality in just
> TEN YEARS. The result apparently is not statistically significant (may have
> occurred by chance), but it is most definitely clinically significant (of
> benefit to the patient) if it is not a chance result.

But, if the result isn't statistically significant, on the basis of this
result you absolutely cannot conclude that the observed difference isn't
due to chance.
>
> The study may not be sufficiently powered (not enough patients in each group
> or not enough duration). If the study is continuing, it may well conclude
> after the second decade that metformin was superior to sulfonylurea
> monotherapy.

That kind of outcome shopping is generally frowned upon,
methodologically, as it comes perilously close to saying you'll continue
the study until you get numbers you like.

What I'd really like to see is a finer-grained analysis. While there may
be no difference between groups in all cause mortality, there may well
be different breakdowns (simplistically, one group may have more cancer
deaths, while the other has more stroke deaths)

--
AF
"Non Sequitur U has a really, really lousy debate team."
--artyw raises the bar on rec.sport.baseball

Alice Faber <afaber@panix.com> wrote in part:

>In article <tkkt93tbtul6ub0kje9pn5ovvqghf52qj7@4ax.com>,
> Jim Chinnis <jchinnis@SPAMalum.mit.edu> wrote:
>
>> "Will, T2" <wmmckee@cox.net> wrote in part:
>>
>> >On Thu, 19 Jul 2007 01:47:31 GMT, "Julie Bove" <juliebove@verizon.net>
>> >wrote:
>> >
>> >>Future work should assess whether long-term exposure to oral
>> >>> antihyperglycemic medications reduces all-cause or cause-specific
>> >>> mortality," the investigators note.
>> >>> Diabetes Care 2077;30:1689-1693.
>> >
>> >This is along the lines Jim was asking about yesterday....
>> >
>> >Will, T2
>>
>> Yes. But I wouldn't say the subject chosen ("Diabetes drugs and combinations
>> of drugs have no effect on all cause mortality") describes the result.
>>
>> First, all patients were treated and the comparison is with one of the
>> treatments, not with a placebo.
>>
>> Second, metformin produced a 13% reduction in ALL-CAUSE mortality in just
>> TEN YEARS. The result apparently is not statistically significant (may have
>> occurred by chance), but it is most definitely clinically significant (of
>> benefit to the patient) if it is not a chance result.
>
>But, if the result isn't statistically significant, on the basis of this
>result you absolutely cannot conclude that the observed difference isn't
>due to chance.

You never can. It is a sliding scale, and the level chosen for statistical
significance is arbitrary. Usually, one says a result is statistically
significant if the probability is less than 0.05 that it would have
occurred by chance if there were no difference due to treatment...

>> The study may not be sufficiently powered (not enough patients in each group
>> or not enough duration). If the study is continuing, it may well conclude
>> after the second decade that metformin was superior to sulfonylurea
>> monotherapy.
>
>That kind of outcome shopping is generally frowned upon,
>methodologically, as it comes perilously close to saying you'll continue
>the study until you get numbers you like.

Well, usually you will never get the numbers you like if they disagree with
reality. But you are right that the duration of the trial should be
specified in advance, as I hope and assume it was.

>What I'd really like to see is a finer-grained analysis. While there may
>be no difference between groups in all cause mortality, there may well
>be different breakdowns (simplistically, one group may have more cancer
>deaths, while the other has more stroke deaths)

Statistically, that isn't kosher. If the full group showed a statistically
significant difference between metformin and sulfonylurea, then it would
make sense to see if subgroups differed. But when no overall effect is
significant, one cannot fish for a significant subgroup.
--
Jim Chinnis Warrenton, Virginia, USA

In article <r6pt93t70b8f1j5m0p9q43mn5lvlvvf82b@4ax.com>,
Jim Chinnis <jchinnis@SPAMalum.mit.edu> wrote:

> Alice Faber <afaber@panix.com> wrote in part:
>
> >In article <tkkt93tbtul6ub0kje9pn5ovvqghf52qj7@4ax.com>,
> > Jim Chinnis <jchinnis@SPAMalum.mit.edu> wrote:
> >
> >> "Will, T2" <wmmckee@cox.net> wrote in part:
> >>
> >> >On Thu, 19 Jul 2007 01:47:31 GMT, "Julie Bove" <juliebove@verizon.net>
> >> >wrote:
> >> >
> >> >>Future work should assess whether long-term exposure to oral
> >> >>> antihyperglycemic medications reduces all-cause or cause-specific
> >> >>> mortality," the investigators note.
> >> >>> Diabetes Care 2077;30:1689-1693.
> >> >
> >> >This is along the lines Jim was asking about yesterday....
> >> >
> >> >Will, T2
> >>
> >> Yes. But I wouldn't say the subject chosen ("Diabetes drugs and
> >> combinations
> >> of drugs have no effect on all cause mortality") describes the result.
> >>
> >> First, all patients were treated and the comparison is with one of the
> >> treatments, not with a placebo.
> >>
> >> Second, metformin produced a 13% reduction in ALL-CAUSE mortality in just
> >> TEN YEARS. The result apparently is not statistically significant (may
> >> have
> >> occurred by chance), but it is most definitely clinically significant (of
> >> benefit to the patient) if it is not a chance result.
> >
> >But, if the result isn't statistically significant, on the basis of this
> >result you absolutely cannot conclude that the observed difference isn't
> >due to chance.
>
> You never can. It is a sliding scale, and the level chosen for statistical
> significance is arbitrary. Usually, one says a result is statistically
> significant if the probability is less than 0.05 that it would have
> occurred by chance if there were no difference due to treatment...

It's arbitrary, sure, but different fields have different conventions
for what level is reportable.
>
> >> The study may not be sufficiently powered (not enough patients in each
> >> group
> >> or not enough duration). If the study is continuing, it may well conclude
> >> after the second decade that metformin was superior to sulfonylurea
> >> monotherapy.
> >
> >That kind of outcome shopping is generally frowned upon,
> >methodologically, as it comes perilously close to saying you'll continue
> >the study until you get numbers you like.
>
> Well, usually you will never get the numbers you like if they disagree with
> reality. But you are right that the duration of the trial should be
> specified in advance, as I hope and assume it was.
>
> >What I'd really like to see is a finer-grained analysis. While there may
> >be no difference between groups in all cause mortality, there may well
> >be different breakdowns (simplistically, one group may have more cancer
> >deaths, while the other has more stroke deaths)
>
> Statistically, that isn't kosher. If the full group showed a statistically
> significant difference between metformin and sulfonylurea, then it would
> make sense to see if subgroups differed. But when no overall effect is
> significant, one cannot fish for a significant subgroup.

It's not fishing here, necessarily (though, of course, it could be).
There are two ways you could end up with no difference in all cause
mortality. One is if, for every cause of death sufficiently represented,
there is no difference. The other is off-setting differences. A
reasonable study design would want to distinguish these two
possibilities. And knowing the answer, while it wouldn't answer the
insurance company question of which treatment, metformin or sulfs, is
best from a cost-benefit point of view, would have the potential for
enhancing understanding of both the progression of diabetes and the
mechanisms by which these two medications work (as well as to the
development of newer therapies that might work even better).

--
AF
"Non Sequitur U has a really, really lousy debate team."
--artyw raises the bar on rec.sport.baseball

Alice Faber <afaber@panix.com> wrote in part:

>In article <r6pt93t70b8f1j5m0p9q43mn5lvlvvf82b@4ax.com>,
> Jim Chinnis <jchinnis@SPAMalum.mit.edu> wrote:
>
>> Alice Faber <afaber@panix.com> wrote in part:
>>
>> >In article <tkkt93tbtul6ub0kje9pn5ovvqghf52qj7@4ax.com>,
>> > Jim Chinnis <jchinnis@SPAMalum.mit.edu> wrote:
>> >
>> >> "Will, T2" <wmmckee@cox.net> wrote in part:
>> >>
>> >> >On Thu, 19 Jul 2007 01:47:31 GMT, "Julie Bove" <juliebove@verizon.net>
>> >> >wrote:
>> >> >
>> >> >>Future work should assess whether long-term exposure to oral
>> >> >>> antihyperglycemic medications reduces all-cause or cause-specific
>> >> >>> mortality," the investigators note.
>> >> >>> Diabetes Care 2077;30:1689-1693.
>> >> >
>> >> >This is along the lines Jim was asking about yesterday....
>> >> >
>> >> >Will, T2
>> >>
>> >> Yes. But I wouldn't say the subject chosen ("Diabetes drugs and
>> >> combinations
>> >> of drugs have no effect on all cause mortality") describes the result.
>> >>
>> >> First, all patients were treated and the comparison is with one of the
>> >> treatments, not with a placebo.
>> >>
>> >> Second, metformin produced a 13% reduction in ALL-CAUSE mortality in just
>> >> TEN YEARS. The result apparently is not statistically significant (may
>> >> have
>> >> occurred by chance), but it is most definitely clinically significant (of
>> >> benefit to the patient) if it is not a chance result.
>> >
>> >But, if the result isn't statistically significant, on the basis of this
>> >result you absolutely cannot conclude that the observed difference isn't
>> >due to chance.
>>
>> You never can. It is a sliding scale, and the level chosen for statistical
>> significance is arbitrary. Usually, one says a result is statistically
>> significant if the probability is less than 0.05 that it would have
>> occurred by chance if there were no difference due to treatment...
>
>It's arbitrary, sure, but different fields have different conventions
>for what level is reportable.

The size of the effect is reported (a 13% reduction in all-cause mortality
in this case) and the significance level or whether the significance level
exceeds the prespecified criterion.

The issue of what the probabability is that metformin reduced all-cause
mortality by 13% or more, or 5% or more, or whatever, is an issue in
Bayesian inference. It's the number(s) we want to know, but can't get
objectively.

But someone who is in the position of deciding which drug to take should, in
my opinion, take into account the nonsignificant but sizable reduction in
mortality this study reported. If there's a 10% chance that it is the result
of luck in a small study, take that into account, too.

I would choose metformin, all else being equal.

>> >> The study may not be sufficiently powered (not enough patients in each
>> >> group
>> >> or not enough duration). If the study is continuing, it may well conclude
>> >> after the second decade that metformin was superior to sulfonylurea
>> >> monotherapy.
>> >
>> >That kind of outcome shopping is generally frowned upon,
>> >methodologically, as it comes perilously close to saying you'll continue
>> >the study until you get numbers you like.
>>
>> Well, usually you will never get the numbers you like if they disagree with
>> reality. But you are right that the duration of the trial should be
>> specified in advance, as I hope and assume it was.
>>
>> >What I'd really like to see is a finer-grained analysis. While there may
>> >be no difference between groups in all cause mortality, there may well
>> >be different breakdowns (simplistically, one group may have more cancer
>> >deaths, while the other has more stroke deaths)
>>
>> Statistically, that isn't kosher. If the full group showed a statistically
>> significant difference between metformin and sulfonylurea, then it would
>> make sense to see if subgroups differed. But when no overall effect is
>> significant, one cannot fish for a significant subgroup.
>
>It's not fishing here, necessarily (though, of course, it could be).

I'd say it is a post-hoc test with a non-significant group effect. That's
fishing. It may be useful to fish that way, since you can see things and
some of them might turn out to survive a rigorous test in a new study.

>There are two ways you could end up with no difference in all cause
>mortality.

There was a big difference in all-cause mortality here. It wasn't zero!

> One is if, for every cause of death sufficiently represented,
>there is no difference. The other is off-setting differences. A
>reasonable study design would want to distinguish these two
>possibilities. And knowing the answer, while it wouldn't answer the
>insurance company question of which treatment, metformin or sulfs, is
>best from a cost-benefit point of view, would have the potential for
>enhancing understanding of both the progression of diabetes and the
>mechanisms by which these two medications work (as well as to the
>development of newer therapies that might work even better).

I agree that the data should be worked over in great detail--in order to
gain possible insights that might result in ideas to be tested in future
studies. It is conceivable that there are massive off-setting differences
that would be so dramatic that they pass the interocular traumatic test... I
imagine the researchers looked for such things.
--
Jim Chinnis Warrenton, Virginia, USA

Jim Chinnis <jchinnis@spamalum.mit.edu> wrote:
: "Will, T2" <wmmckee@cox.net> wrote in part:

: >On Thu, 19 Jul 2007 01:47:31 GMT, "Julie Bove" <juliebove@verizon.net>
: >wrote:
: >
: >>Future work should assess whether long-term exposure to oral
: >>> antihyperglycemic medications reduces all-cause or cause-specific
: >>> mortality," the investigators note.
: >>> Diabetes Care 2077;30:1689-1693.
: >
: >This is along the lines Jim was asking about yesterday....
: >
: >Will, T2

: Yes. But I wouldn't say the subject chosen ("Diabetes drugs and combinations
: of drugs have no effect on all cause mortality") describes the result.

: First, all patients were treated and the comparison is with one of the
: treatments, not with a placebo.

: Second, metformin produced a 13% reduction in ALL-CAUSE mortality in just
: TEN YEARS. The result apparently is not statistically significant (may have
: occurred by chance), but it is most definitely clinically significant (of
: benefit to the patient) if it is not a chance result.

: The study may not be sufficiently powered (not enough patients in each group
: or not enough duration). If the study is continuing, it may well conclude
: after the second decade that metformin was superior to sulfonylurea
: monotherapy.
: --
: Jim Chinnis Warrenton, Virginia, USA

Do we know if all the patients were under treatment my the VA? We have,
certainly, been hearing stories about the difficulties of gaining good
contrl under their treatment regimins and standards. that, it seems to
me, is enough to skew the results. Bad treatment with several drugs
doesn't help. Well, duh!

Wendy

Jim Chinnis wrote:

> "Will, T2" <wmmckee@cox.net> wrote in part:
>
>
>>On Thu, 19 Jul 2007 01:47:31 GMT, "Julie Bove" <juliebove@verizon.net>
>>wrote:
>>
>>
>>>Future work should assess whether long-term exposure to oral
>>>
>>>>antihyperglycemic medications reduces all-cause or cause-specific
>>>>mortality," the investigators note.
>>>>Diabetes Care 2077;30:1689-1693.
>>
>>This is along the lines Jim was asking about yesterday....
>>
>>Will, T2
>
>
> Yes. But I wouldn't say the subject chosen ("Diabetes drugs and combinations
> of drugs have no effect on all cause mortality") describes the result.
>
> First, all patients were treated and the comparison is with one of the
> treatments, not with a placebo.
>
> Second, metformin produced a 13% reduction in ALL-CAUSE mortality in just
> TEN YEARS. The result apparently is not statistically significant (may have
> occurred by chance), but it is most definitely clinically significant (of
> benefit to the patient) if it is not a chance result.
>
> The study may not be sufficiently powered (not enough patients in each group
> or not enough duration). If the study is continuing, it may well conclude
> after the second decade that metformin was superior to sulfonylurea
> monotherapy.

The management of type 2 diabetes is more than a choice of oral
hypoglycemics. Less than 8% of T2s have good control over blood
glucose, blood pressure, and lipids. The following graphic shows the
intersects between fasting, postprandial glucose, and the relative risk
of death. It is a nice 3 dimensional graphic but it leaves out the
blood pressure and lipid factors that are also significant in type 2
diabetes management as well as the metabolic syndrome.

Slide 30. DECODE Trial: Relative Risk of Death Shown by Blood Glucose
Level - The Pathophysiology of Type 2 Diabetes: New Perspectives and
Their Clinical Impact - http://www.medscape.com/viewarticle/558472
(Note: This graphic was part of the report on DECODE.)

The pillar on the left, lower position on the graph would seem to confer
little risks.

One of the points that Steven Nissen was making in regard to oral
hypoglycemics in the article he co-authored on Avandia (rosiglitazone)
was that there wasn't enough focus on other aspects of diabetic meds
besides blood glucose control.

Framingham, SCORE, and DECODE Risk Equations Do Not Provide Reliable
Cardiovascular Risk Estimates in Type 2 Diabetes -
http://www.medscape.com/viewarticle/559075

Frank Roy
Jefferson, Md.

On Thu, 19 Jul 2007 14:48:08 GMT, Jim Chinnis
<jchinnis@SPAMalum.mit.edu> wrote:

>There was a big difference in all-cause mortality here. It wasn't zero!

Sure made a difference to those in the 13%.


Cheers, Alan, T2, Australia.
d&e, metformin 1500mg, ezetrol 10mg
Everything in Moderation - Except Laughter.
--
http://loraltraveloz.blogspot.com/
latest: Mossman Gorge in the Daintree Rainforest
http://loraldiabetes.blogspot.com/
latest: Self-Testing and Type 2 Management