meta analysis is appropriate...the question you ask if if you can pool controlled with uncontrolled studies and the answer is no...
also, it is best to have several RCTs to pool but 2 RCTs are better than 1; once the PICO is in order as to clinical
heterogeneity (lack of), then I say pool the 2 RCTs for an improved summary estimate, even pool the observationals but separate...not together and look at the differences in estimates and tell the reader the story that way too; tell the reader you did not pool the observationals as you should not do this given they differ methodologically and would be plagued by confounding biases in the observationals...tell the reader everything you decided and it will be ok...just lay out your approach BUT do not pool them....
if you do pool the 2 RCTs, then also
discuss the impact of differential risk of bias in explaining uncovered stats heterogeneity (ensure you apply the Cochrane risk of bias tool for the 2 RCTs)...and yes, provide as much narrative information (tables) as possible. people like ot see many tables and data..so give them that. Best, Paul E. Alexander
--- On Fri, 5/10/13, samia Alhabib <[log in to unmask]> wrote: From: samia Alhabib <[log in to unmask]> Subject: To: [log in to unmask] Received: Friday, May 10, 2013, 12:55 AM
thank you all for your feedback. we got only 2 RCT ans and4 prospective cohort and one retrospective study. I guess one cannot combine 2 RCT , we need at least to have more than 3 studies to combine them? The studies are similar in terms of outcomes, stage of disease, and measurements, but differ of course in some aspects such as race, gender... So, I
understand from your comments that metanalysis is inappropriate and narrative one is the option? Best
Regards, Samia
Dr. Samia Alhabib, MD, PhD
-------- Original message -------- From: "Dr. Carlos Cuello" < [log in to unmask]> Date: To: [log in to unmask] Subject: Hi SamiaI agree with Tom that both designs should not be combined. This great question drives us to another question
If you have a body of evidence from RCTs, why would you need the observational studies?
I would say that depends on the outcome, and you need to decide a priori which outcomes will you be evaluating; For example, if you need to evaluate whether your intervention provoques a rare adverse event (let's say, pancytopenia), then perhaps the cohort studies (or other observational designs) will be more useful than the RCTs.
RCTs by default are considered high quality evidence that you can downgrade depending on certain criteria (see GRADE methods). Furthermore, observational studies are considered low quality evidence that can also be upgraded depending on the GRADE criteria (very similar to the Hills criteria, actually). My suggestion is that you could add GRADE SoF tables to your work.
best
|