overmatching in case-control studies

[Raoul]

I have a simple question:

Can overmatching in a case-control study lead to bias? I have
consulted various epidemiologists and get different
answers all the time. Some say it can, but others say it can't and it
only reduces the efficiency of the study (i.e., power). I would be
interested what people think. Thanks.

Raoul

[Steve Simon, P.Mean Consulting]

Raoul wrote:

Both answers can be right.

As far as I know, overmatching can't lead to a liberal bias, a bias that
makes a treatment effect look stronger than it really is. It can only
produce a conservative bias, a bias towards the null hypothesis. If you
think that liberal biases, ones that produce false positive findings,
are the only ones worth worrying about, then you would call the tendency
to conservative biases (producing too many false negatives) as something
like loss of power rather than bias.

Here's a very nice reference that gives a good practical example of
overmatching.

Removal of radiation dose response effects: an example of over-matching.
J. L. Marsh, J. L. Hutton, K. Binks. Bmj 2002: 325(7359); 327-30.
http://www.bmj.com/cgi/content/full/325/7359/327
--
Steve Simon, Standard Disclaimer
Free statistics webinar, Wed, Oct 14, 10am CDT.
"P-values, confidence intervals, and the Bayesian alternative"
Details at www.pmean.com/webinars

[ציפי שוחט‎]

I think overmatching may lead to confounding between the risk factor of interest and the matched factors if those who possess the risk factor have very different profiles in the matched factors than those who don't. Say in an extreme case all men smoke and all women don't. Matching on gender would make it immposible to distinguish between the risk factor - smoking and the matching factor - gender.

 

Tzippy

 

2009/10/16, Steve Simon, P.Mean Consulting <n...@pmean.com>:

[John Sorkin]

Tzippy
The problem you mention has nothing to do with over-matching, at least
as far as I define over-matiching, viz. something other than 1:1
matching. In any matched study, no matter what the matching ratio is,
1:1, 1:2, 1:n, etc. it is not possible to make inferences about the
factor used in the matching. In your specific case, not matter what
matching ratio is used, it would not be possible to distinguish the
effect of smoking vs. the effect of sex.
Off the top of my head, I am hard pressed to see why over matching could
causes bias. It might make a biased study achieve significance that
would not be found in a 1:1 matched study by increasing study power, but
in and of itself, I don't see how it could lead to bias.

I await a counter example.

John



John Sorkin M.D., Ph.D.
Chief, Biostatistics and Informatics
Baltimore VA Medical Center GRECC,
University of Maryland School of Medicine Claude D. Pepper OAIC,
University of Maryland Clinical Nutrition Research Unit, and
Baltimore VA Center Stroke of Excellence

University of Maryland School of Medicine
Division of Gerontology
Baltimore VA Medical Center
10 North Greene Street
GRECC (BT/18/GR)
Baltimore, MD 21201-1524

(Phone) 410-605-7119
(Fax) 410-605-7913 (Please call phone number above prior to faxing)
jso...@grecc.umaryland.edu
>>> ציפי שוחט <tz.sh...@gmail.com> 10/17/09 5:11 AM >>>
I think overmatching may lead to confoundoing between the risk factor of

Confidentiality Statement:
This email message, including any attachments, is for the sole use of
the intended recipient(s) and may contain confidential and privileged
information. Any unauthorized use, disclosure or distribution is
prohibited. If you are not the intended recipient, please contact the
sender by reply email and destroy all copies of the original message.

[John Whittington]

At 10:45 17/10/2009 -0400, John Sorkin wrote:
>The problem you mention has nothing to do with over-matching, at least
>as far as I define over-matiching, viz. something other than 1:1
>matching. In any matched study, no matter what the matching ratio is,
>1:1, 1:2, 1:n, etc. it is not possible to make inferences about the
>factor used in the matching. In your specific case, not matter what
>matching ratio is used, it would not be possible to distinguish the
>effect of smoking vs. the effect of sex.
>Off the top of my head, I am hard pressed to see why over matching could
>causes bias. It might make a biased study achieve significance that
>would not be found in a 1:1 matched study by increasing study power, but
>in and of itself, I don't see how it could lead to bias.

John, I may be wrong, but I don't think the 'overmatching' being talked
about here has got anything to do with numbers (i.e. the case:control
ratio). Rather, I think it refers to imprudent choice of the factors by
which one matches/stratifies, in particular circumstances. An oft-cited
paper which discusses this can be found at:

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1123834/

Kind Regards,

John

----------------------------------------------------------------
Dr John Whittington, Voice: +44 (0) 1296 730225
Mediscience Services Fax: +44 (0) 1296 738893
Twyford Manor, Twyford, E-mail: Joh...@mediscience.co.uk
Buckingham MK18 4EL, UK
----------------------------------------------------------------

[John Sorkin]

John
From one John to another,
Thank you,

John

John Sorkin M.D., Ph.D.
Chief, Biostatistics and Informatics
Baltimore VA Medical Center GRECC,
University of Maryland School of Medicine Claude D. Pepper OAIC,
University of Maryland Clinical Nutrition Research Unit, and
Baltimore VA Center Stroke of Excellence

University of Maryland School of Medicine
Division of Gerontology
Baltimore VA Medical Center
10 North Greene Street
GRECC (BT/18/GR)
Baltimore, MD 21201-1524

(Phone) 410-605-7119
(Fax) 410-605-7913 (Please call phone number above prior to faxing)
jso...@grecc.umaryland.edu

>>> John Whittington <Joh...@mediscience.co.uk> 10/17/09 11:07 AM >>>

At 10:45 17/10/2009 -0400, John Sorkin wrote:
>The problem you mention has nothing to do with over-matching, at least
>as far as I define over-matiching, viz. something other than 1:1
>matching. In any matched study, no matter what the matching ratio is,
>1:1, 1:2, 1:n, etc. it is not possible to make inferences about the
>factor used in the matching. In your specific case, not matter what

>matching ratio is used, it would Frobe possible to distinguish the

>effect of smoking vs. the effect of sex.
>Off the top of my head, I am hard pressed to see why over matching could
>causes bias. It might make a biased study achieve significance that
>would not be found in a 1:1 matched study by increasing study power, but
>in and of itself, I don't see how it could lead to bias.

John, I may be wrong, but I don't think the 'overmatching' being talked
about here has got anything to do with numbers (i.e. the case:control
ratio). Rather, I think it refers to imprudent choice of the factors by
which one matches/stratifies, in particular circumstances. An oft-cited
paper which discusses this can be found at:

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1123834/

Kind Regards,


John

----------------------------------------------------------------
Dr John Whittington, Voice: +44 (0) 1296 730225
Mediscience Services Fax: +44 (0) 1296 738893
Twyford Manor, Twyford, E-mail: Joh...@mediscience.co.uk
Buckingham MK18 4EL, UK
----------------------------------------------------------------