In message <[log in to unmask]>, Julia Valderrama <[log in to unmask]> writes >Hi, >Thanks a lot for all your comments on intention to treat analysis. > >I am looking at a RCT by Dutta D et als, 1996 (Efficacy of norfloxacin >and doxycycline for the treatment of V cholerae O139 infection) >where 160 patients were enrolled in the study but just 111 analysed. >The difference 49 subjects were not included in the analysis because >they showed culture (-) to Vcholerae O139. There's a table that >shows comparable characteristics of the patients on admission, but >for the 111 not for the 160. Is this the way it should be? OR the >right way would be showing the table with all the subjects entered and >randomized for the study (i.e 160) and as we have analyzed a 69% of >that (i.e. 111) because of 49 not being O139 we can conclude that >realibility of the study is compromised? > >Would this be an example of not having followed the intention to treat >analyis, i.e. once randomised always analysed? > It depends on the inclusion criteria and at what stage subjects were randomised. If one inclusion criterion was +ve culture for V cholera 0139 then only 111 patients should have been randomised, and all should have been analysed. Another way of putting it was that there was an intention to treat only those 111 +ve patients. If, on the other hand, all 160 were randomised and that 49 were excluded post hoc because of -ve culture, then that is no longer an intention to treat analysis. Sometimes otherwise well designed and executed studies exclude subjects from the analysis after randomisation for a variety of reasons - lost to follow up, failed to complete the treatment and so on. This should not happen, and is becoming rare in the EBM era but I guess might have been done in a study published in 1996 (seems like so long ago!). You can often do the intention to treat analysis yourself from the data given (often it will show a smaller absolute risk reduction and hence larger NNT), and a sensitivity analysis on the "worst case" scenario - assume all the patients lost in the treatment arm had a bad outcome and all those lost in the placebo arm had a good one. This may well give you enough information upon which to base a clinical decision. cheers Toby -- Toby Lipman General practitioner, Newcastle upon Tyne Northern and Yorkshire research training fellow Tel 0191-2811060 (home), 0191-2437000 (surgery) Northern and Yorkshire Evidence-Based Practice Workshops http://www.eb-practice.fsnet.co.uk/ %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%