Date:         Tue, 20 Sep 2011 07:02:16 -0400
Reply-To:     William Shakespeare <shakespeare_1040@HOTMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         William Shakespeare <shakespeare_1040@HOTMAIL.COM>
Subject:      Study design
Content-Type: text/plain; charset=ISO-8859-1

Suppose there’s a situation where Treatment A and never Treatment B is
always administered to a cohort, there then occurs a change in practice,
and Treatment B and never Treatment A is administered to a different
cohort.  An investigator wishes to compare the incidence of a disease
across these cohorts.  What does this prove other than the incidences are
similar/different?  There are so many threats to internal validity there’s
no way anyone can make any causal attribution for the differences.  This
is certainly the unspoken goal of such a study.  Do any of you face this
situation?  What do you say?  What is the standard design for such a
situation?  There is not really the possibility of a case-control study
after the change of practice since there really is not a relevant control
group because everyone gets the same treatment.  Even if you could find
appropriate controls a case-control study would be a poor choice because
those designs are geared more toward elucidating the factors associated
with disease rather than the effectiveness of treatment.  Has anyone ever
seen case-control studies used to evaluate treatments in a retrospective
sense?