Apr 02, 2015
Terminology 101: Intention-to-treat analysis in RCTs
Intention to treat: The principle of comparing RCT participants according to the treatment to which they were originally randomized, regardless of which treatment they actually received
Source: Gordis, L. (2014). Epidemiology. (5th ed.). Philadelphia: Elsevier Saunders
Randomized controlled trials (RCTs) often include some participants who do not adhere to their randomized group assignment and others who leave the study altogether (i.e., lost to followup). Traditionally, researchers deleted participants who left the study from the analysis and reassigned those who switched groups to their self-assigned group. In this approach, known as per protocol analysis, the data are analyzed according to the intervention that participants actually receive during the study, irrespective of their random assignment. This approach negates the whole purpose of randomization and presents a serious threat to the validity of the study findings. This is because participants who leave the study or do not adhere to their group assignment are likely to do so because of characteristics that may influence our understanding of the outcome. For instance, adherers tend to be more attentive to their health than non-adherers and are thus more likely to experience better outcomes, irrespective of the true effect of the intervention.
Intention to treat (ITT) is an alternative approach to per protocol analysis in which study participants are analyzed according to their randomized assignment even if they were lost to followup or failed to adhere to the protocol. The ITT approach preserves the randomization of baseline characteristics between the study groups and mirrors real-life situations wherein not every individual fully adheres to his or her prescribed treatment. Thus, ITT analysis tends to give a more conservative, yet more realistic, estimate of the outcome than per protocol analysis.
To illustrate, let’s say that an RCT was conducted to compare surgical wound infections in 100 postoperative patients who were randomly assigned to receive on-site dressing changes (group 1) and another 100 patients who were randomly assigned to receive at-home dressing changes (group 2). Let us assume that 10 patients from group 1 left the study and that five didn’t show up for their on-site dressing changes and instead had them done at home. Let us also assume that five patients from group 2 were lost to followup while 10 other patients showed up for on-site dressing changes because they were not comfortable doing them at home. With the per protocol approach, the investigators would delete the 15 patients who were lost to followup from both groups and would reassign the patients who switched groups. This would yield a final sample of 95 patients in group 1 (10 of whom were originally assigned to group 2) and 90 patients in group 2 (five of whom were originally assigned to group 1). With the ITT approach, however, data from all 200 patients would be analyzed as per their randomized assignment, irrespective of whether any had left the study or changed groups.
Failure to implement ITT in RCTs presents a major threat to the validity of study conclusions. Thus, readers of RCT reports should check whether the researchers have analyzed all cases according to their original group assignment, including those who were lost to followup.
NurseONE.ca resources on this topic
- Bennett, P. (2011). Abnormal and Clinical Psychology: An Introductory Textbook.
- Di Mario, C., Dangas, G., & Barlis, P. (Eds.). (2011). Interventional Cardiology: Principles and Practice.
- Holly, C., Salmond, S. W., & Saimbert, M. K. (2011). Comprehensive Systematic Review for Advanced Nursing Practice