General

Clinical Research : Intention-to-treat Vs Per-Protocol Analysis of Result

Created by:

 
7 Aug 2016 - General
 

In a randomised controlled trial, the resuls for both new intervention and placebo (or standard treatment) group will be compared, at the end of study period, to determine whether the new intervention will have significant better effect comparing to placebo or current standard treatment. Usually, there are 2 ways of analysis of the result, which are intention-to-treat analysis (ITT) and per protocol analysis.  For intention-to-treat, the analysis will compare the outcomes between the study groups (intervention Vs. placebo) with every participant analyzed according to the assigned randomized group, regardless whether the participant adhered to the assigned intervention. Per-protocol analysis will only analyse or include the participants who adhered strictly to the study protocol. For example, in study A, there are 100 participants assigned to intervention group, and 100 participants asssigned to placebo group, and there are 20 participants drop out from intervention group, while there are 5 drop out from placebo group. For intention-to treat analysis, all the results of both 100 participants in interventon and placebo groups will be analysed, But, for per-protocol analysis, only 80 in intervention arm and 95 in placebo group will be anaylsed.  Whic result (ITT Vs. per protocol) will be more appropriate ? Which result will you rely on ?? 

Mark Edmon Tan Nowadays, most studies will have both Intention-To-Treat analysis and Per-Protocol analysis. As you have pointed out, intention-to-treat analysis may underestimate the full effect of the treatment. This is because for ITT analysis, we do included the number of patients who were selected for treatment group but have not complete the treatment as per protocol for various reason, with the drop out. Bu...
 (Total 139 words)
You Yi Hong: I am no expert in statistical analysis. I think if we are considering a clinician's standpoint, it would be prudent to use per-protocol analysis because it only analyzes data that has followed the study design. It would be safer to determine possible effects and adverse effects with this statistical analysis because full course of intervention has been. However, this is not applicable in all dise...
 (Total 135 words)