Percent change and power analysis
SR Millis
A physician is consulting with me on a clinical trial to compare the effectiveness of 2 medications to treat osteoporosis in patients with spinal cord injury (SCI). Bone loss after SCI leads to increased fracture risk in lower limbs of patients with paraplegia.
Patients will be randomized to one of 2 groups, bone mineral density (BMD) will be assessed at baseline (pre-treatment) and post-treatment.
We're currently working on the power analysis. This physician is fixated on characterizing the effect size as a percent change in BMD from baseline. I have always worked with the "raw" units of measure of the efficacy endpoint in my sample size calculations---rather than percent change, which has a highly skewed distribution, among other things.
Suggestions?
Thanks,
Scott Millis
John Sorkin
Please forward replies you receive to me; I have often been in the same situation as you and would like to know what our colleagues have today.
John
John David Sorkin M.D., Ph.D.
Chief, Biostatistics and Informatics
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)
>>> SR Millis <srmi...@yahoo.com> 4/7/2009 6:14 PM >>>
Suggestions?
Thanks,
Scott Millis
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.
BXC (Bendix Carstensen)
- freely available from www.bmj.com
Log-transform of the orginal data will take you to the relative scale.
This should give you most of the information you need to set up a simulation machinery to do the power (and precison, hopefully) calculations.
Best regards,
Bendix
_______________________________________________
Bendix Carstensen
Senior Statistician
Steno Diabetes Center
Niels Steensens Vej 2-4
DK-2820 Gentofte
Denmark
+45 44 43 87 38 (direct)
+45 30 75 87 38 (mobile)
b...@steno.dk http://www.biostat.ku.dk/~bxc
www.steno.dk
SR Millis
Bendix,
I don't have any original data on which to base the simulation.
Scott Millis
Neil Shephard
clinicians percentage to the change in raw units?
Obviously you'll need to know the baseline BMD for the sample, but he
should be able to provide reasonable estimates of what to expect.
Neil
--
"The combination of some data and an aching desire for an answer does
not ensure that a reasonable answer can be extracted from a given body
of data." ~ John Tukey (1986), "Sunset salvo". The American
Statistician 40(1).
Email - nshe...@gmail.com
Website - http://slack.ser.man.ac.uk/
Photos - http://www.flickr.com/photos/slackline/
Peter Flom
>
>
>A physician is consulting with me on a clinical trial to compare the effectiveness of 2 medications to treat osteoporosis in patients with spinal cord injury (SCI). Bone loss after SCI leads to increased fracture risk in lower limbs of patients with paraplegia.
>
>Patients will be randomized to one of 2 groups, bone mineral density (BMD) will be assessed at baseline (pre-treatment) and post-treatment.
>
>We're currently working on the power analysis. This physician is fixated on characterizing the effect size as a percent change in BMD from baseline. I have always worked with the "raw" units of measure of the efficacy endpoint in my sample size calculations---rather than percent change, which has a highly skewed distribution, among other things.
>
>Suggestions?
>
Just as a sort of start of an idea .... could you use e.g. the arcsine transformation to start points and end points?
I'm not sure how that would work .... but it's a useful transformation in some cases.
Peter
Peter L. Flom, PhD
Statistical Consultant
www DOT peterflomconsulting DOT com
SR Millis
Neil,
Yes, that is the approach that I'm trying to take---along with getting an estimate of variance in raw units. The physician was also characterizing the variance in terms of percent change!
Scott Millis
--- On Tue, 4/7/09, Neil Shephard <nshe...@gmail.com> wrote:
BXC (Bendix Carstensen)
Bendix
> -----Original Message-----
> From: MedS...@googlegroups.com
> [mailto:MedS...@googlegroups.com] On Behalf Of SR Millis
> Sent: 8. april 2009 00:32
> To: MedS...@googlegroups.com
> Subject: {MEDSTATS} Re: Percent change and power analysis
>
>
>
Doug Altman
Vickers AJ. The use of percentage change from baseline as an outcome in a controlled trial is statistically inefficient: a simulation study. BMC Med Res Methodol. 2001;1:6.
http://www.biomedcentral.com/1471-2288/1/6
Doug
_____________________________________________________
Doug Altman
Professor of Statistics in Medicine
Centre for Statistics in Medicine
University of Oxford
Wolfson College Annexe
Linton Road
Oxford OX2 6UD
email: doug....@csm.ox.ac.uk
Tel: 01865 284400 (direct line
01865 284401)
Fax: 01865 284424
www:
http://www.csm-oxford.org.uk/
EQUATOR Network - resources for reporting research
www:
http://www.equator-network.org/
Frank
It is great to know about this paper. Andrew Vickers has written a
few small articles for medscape.com which are fantastic (see
especially his ones on data torture and the sample size samba).
Analysis of percent change is one of the great mistakes of medical
statistics. Most practitioners do not even know that one cannot
compute any summary statistics on percent change other than min and
max (without converting to log ratios). Very few clinicians seem to
know that the average of +100% and -50% change is 0%. On a deeper
level, many clinicians seem to feel that 'change' is what a parallel
group randomized trial is intended to analyze, but that violates the
spirit of the parallel group design. The only change that a parallel
group design is intended to analyze is the change from one treatment
group to another. I find that spending time on change, with all of
its problems such as regression to the mean (and computing change
without making the appropriate transformation on the raw measurements
first), gets in the way of thinking. Most clinicians don't even
bother to make a Bland-Altman plot to verify that their favorite
change score is a valid one.
We have devoted a section to this on our "Checklist for Authors" at
http://biostat.mc.vanderbilt.edu/ManuscriptChecklist
Frank Harrell
On Apr 8, 3:20 am, Doug Altman <doug.alt...@csm.ox.ac.uk> wrote:
> See also
>
> Vickers AJ. The use of percentage change from baseline as an outcome
> in a controlled trial is statistically inefficient: a simulation
> Fax: 01865 284424
> www: http://www.csm-oxford.org.uk/
>
> EQUATOR Network - resources for reporting research
John Sorkin
Do you have a further explication of the problems associated with using %change or fractional change? I could use a written document to give to my boss and other folks who insist I analyze data using % or fractional change?
Thanks,
John
John David Sorkin M.D., Ph.D.
Chief, Biostatistics and Informatics
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)
>>> Frank <f.ha...@vanderbilt.edu> 4/8/2009 8:46 AM >>>
Doug,
It is great to know about this paper. Andrew Vickers has written a
few small articles for medscape.com which are fantastic (see
especially his ones on data torture and the sample size samba).
Analysis of percent change is one of the great mistakes of medical
statistics. Most practitioners do not even know that one cannot
compute any summary statistics on percent change other than min and
max (without converting to log ratios). Very few clinicians seem to
know that the average of +100% and -50% change is 0%. On a deeper
level, many clinicians seem to feel that 'change' is what a parallel
group randomized trial is intended to analyze, but that violates the
spirit of the parallel group design. The only change that a parallel
group design is intended to analyze is the change from one treatment
group to another. I find that spending time on change, with all of%
Frank Harrell
Confidentiality Statement:
Frank
In addition to the loss of power documented in the paper provided by
Doug, the problems are many and severe. This is one of the worst of
statistical practices. We have tried to document the problems in the
web site listed below. -Frank
On Apr 13, 10:12 pm, "John Sorkin" <jsor...@grecc.umaryland.edu>
wrote:
> Do you have a further explication of the problems associated with using %change or fractional change? I could use a written document to give to my boss and other folks who insist I analyze data using % or fractional change?
> Thanks,
> John
>
> John David Sorkin M.D., Ph.D.
> Chief, Biostatistics and Informatics
> 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)
>
> Doug,
>
> It is great to know about this paper. Andrew Vickers has written a
> few small articles for medscape.com which are fantastic (see
> especially his ones on data torture and the sample size samba).
> Analysis of percent change is one of the great mistakes of medical
> statistics. Most practitioners do not even know that one cannot
> compute any summary statistics on percent change other than min and
> max (without converting to log ratios). Very few clinicians seem to
> know that the average of +100% and -50% change is 0%. On a deeper
> level, many clinicians seem to feel that 'change' is what a parallel
> group randomized trial is intended to analyze, but that violates the
> spirit of the parallel group design. The only change that a parallel
> group design is intended to analyze is the change from one treatment
> group to another. I find that spending time on change, with all of%
> its problems such as regression to the mean (and computing change
> without making the appropriate transformation on the raw measurements
> first), gets in the way of thinking. Most clinicians don't even
> bother to make a Bland-Altman plot to verify that their favorite
> change score is a valid one.
>
Frank
In addition to the loss of power documented in the paper provided by
Doug, the problems are many and severe. This is one of the worst of
statistical practices. We have tried to document the problems in the
web site listed below. -Frank
On Apr 13, 10:12 pm, "John Sorkin" <jsor...@grecc.umaryland.edu>
wrote:
> Do you have a further explication of the problems associated with using %change or fractional change? I could use a written document to give to my boss and other folks who insist I analyze data using % or fractional change?
> Thanks,
> John
>
> John David Sorkin M.D., Ph.D.
> Chief, Biostatistics and Informatics
> 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)
>
> Doug,
>
> It is great to know about this paper. Andrew Vickers has written a
> few small articles for medscape.com which are fantastic (see
> especially his ones on data torture and the sample size samba).
> Analysis of percent change is one of the great mistakes of medical
> statistics. Most practitioners do not even know that one cannot
> compute any summary statistics on percent change other than min and
> max (without converting to log ratios). Very few clinicians seem to
> know that the average of +100% and -50% change is 0%. On a deeper
> level, many clinicians seem to feel that 'change' is what a parallel
> group randomized trial is intended to analyze, but that violates the
> spirit of the parallel group design. The only change that a parallel
> group design is intended to analyze is the change from one treatment
> group to another. I find that spending time on change, with all of%
> its problems such as regression to the mean (and computing change
> without making the appropriate transformation on the raw measurements
> first), gets in the way of thinking. Most clinicians don't even
> bother to make a Bland-Altman plot to verify that their favorite
> change score is a valid one.
>