Re: Blinded studies

On Mar 18, 11:11 am, Marc Schwartz <marc_schwa...@xxxxxxxxxxx> wrote:
Richard Ulrich wrote:
On 17 Mar 2007 15:31:33 -0700, "franco" <fra...@xxxxxxxx> wrote:

Hi, I am learning statistics.
I have a question: how can one monitor safety data in an ongoing
double-blinded clinical trial? Suppose you have a series of subjects
from a trial (with an X total number of subjects) who present with
adverse events. All you have is the total number of subjects in the
trial (and the number of subjects for each treatment group; eg 1000
total subjects, 500 in each group and 35 present with an adverse event
E. I thought one way to monitor safety in a blinded way is to do a

'...monitor safety in a blinded way...'

I think that safety monitoring of side effects is done by
an independent committee that is not blind. This is very
easy if the side effect is such that the subject is being
dropped because of the side effect.

It depends on what your 'safety' problem is, but in many
instances, complete diagnosis and treatment of the side-effect
will depend on knowing the actual intervention. For instance,
you do *not* have a toxic reaction to the treatment medication
if the assigned medication was a placebo.

[snip]

Rich,

Typically the committee, interchangeably called the DMC (Data Monitoring
Committee) or the DSMB (Data Safety Monitoring Board) will also be
blinded. The blind will usually only be broken (in an executive session
of the committee) when there is a safety concern involving material
differences between the two (or more) arms such that it becomes
imperative to know whether the control/placebo arm or the study
treatment arm is problematic and there is a sufficient basis to consider
stopping the study due to safety concerns.

Otherwise, inter-arm comparative analyses will be presented to the
committee in a blinded fashion.

Taking a step back, increasingly, there are a priori stopping rules that
are defined as part of the study design expressed in the protocol to
enable more rigorous and unbiased decisions to be made. These rules can
be defined based upon:

1. Futility - Interim study results of the primary efficacy endpoints,
based upon actual data to date and the projected results for the
remaining subjects, suggest that there is little likelihood of a
significant difference between the arms if the study continued. Thus,
there is no point to continue to expose the participants to the risks of
the trial and for the sponsor to continue to fund the trial.

2. Safety - Interim results indicate that one or more arms have
experienced an a priori defined number of specific events in a specific
number of subjects. These "events" may actually be an aggregate of
multiple events that individually occur at lower rates. Again, based
upon the existing data and the projections of the experience in the
remaining subjects, there is little likelihood that the comparative
safety experience would change. Thus, the study would be stopped to
cease exposing the subjects to the elevated safety risks of the study.

3. Efficacy - Interim results of the primary efficacy endpoints suggest
that there is a significant difference between the arms and that the
projected experience in the remaining subjects would not alter that
conclusion. Thus the study would be stopped to enable the results to be
collated and reported to the FDA (or other regulatory entities) earlier
than projected. This also reduces the exposure of the subjects to the
safety risks of the study, the sponsor saves money in funding the trial
and can get the product to market sooner, increasing revenue potential.

Such rules are increasingly becoming part of 'adaptive study designs' to
enable a more fluid approach to the conduct of the study and to
facilitate mid-stream changes without compromising the scientific
integrity of the study. Again, these are a priori rules that are
explicitly defined in the study protocol.

Even in the absence of such rules, there is a priori knowledge (defined
in the study protocol based upon prior published data) of the expected
incidences of key endpoints and adverse events. In general, as long as
the observed rates are within the range of the expected and there is not
a significant difference between the arms of the studies, the study
would likely be allowed to continue.

The above are generalities, with more specificity defined in the study
protocols and related documents.

More information on the monitoring of clinical trials is available in:

Statistical Monitoring of Clinical Trials: A Unified Approach
Michael A. Proschan, K.K. Gordon Lan, Janet Turk Wittes
Springer (2006)

Data Monitoring in Clinical Trials: A Case Studies Approach
David L. DeMets, Curt D. Furberg, Lawrence M. Friedman
Springer (2005)

Data Monitoring Committees in Clinical Trials: A Practical Perspective
Susan S. Ellenberg, Thomas R. Fleming, David L. DeMets
Wiley (2002)

HTH,

Marc Schwartz

Somewhere on the internet several years ago, in the context of a
discussion of research ethics, I read an account of someone shutting
down the study he was conducting because he happened to speak briefly
in an airport boarding area with someone, and in the course of the
conversation discovered that the person was on the committee that was
monitoring his study. That's my recollection, but I haven't been able
to find what I think I read. Do you (or anyone else) remember anything
like that?

.

Relevant Pages

  • Re: Blinded studies
    ... how can one monitor safety data in an ongoing ... interchangeably called the DMC (Data Monitoring Committee) or the DSMB will also be blinded. ... The blind will usually only be broken when there is a safety concern involving material differences between the two arms such that it becomes imperative to know whether the control/placebo arm or the study treatment arm is problematic and there is a sufficient basis to consider stopping the study due to safety concerns. ... Statistical Monitoring of Clinical Trials: ...
    (sci.stat.edu)
  • Re: Jindal vs The Volcano
    ... monitoring such a thing as Yellowstone would achieve? ... because we have been monitoring volcanoes we have learned a lot ... "They that can give up essential liberty to obtain a little temporary ... deserve neither liberty nor safety." ...
    (talk.origins)
  • Re: Jindal vs The Volcano
    ... monitoring such a thing as Yellowstone would achieve? ... Alerting people of the danger ... "They that can give up essential liberty to obtain a little temporary ... deserve neither liberty nor safety." ...
    (talk.origins)
  • Re: Jindal vs The Volcano
    ... Alerting people of the danger will ... because we have been monitoring volcanoes we have learned a lot ... "They that can give up essential liberty to obtain a little temporary ... deserve neither liberty nor safety." ...
    (talk.origins)
Loading