In a randomized trial, consider whether people’s outcomes were counted in the treatment group to which they were assigned.
In a randomized trial, researchers randomly assign participants to treatment (health action) comparison groups. Sometimes participants do not receive or take the treatment to which they were assigned. If their outcomes are not counted in the group to which they were assigned, the results of the trial may be misleading.
Explanation
In a randomized trial, researchers decide by chance (randomly) who is assigned or allocated to which treatment comparison group. This helps to ensure that the groups are similar before they take or receive the treatment or health action to which they were allocated. However, sometimes people do not take the treatment to which they were randomly allocated. Those people may be different from those who do take the treatment to which they were allocated. So, if their outcomes are not counted in the group to which they were allocated, the treatment comparison groups will no longer be similar. This may lead to an underestimate or an overestimate of the effects of treatments.
For example, in a comparison of surgery and drug treatments, people who die while waiting for the surgery should be counted in the surgery group, even though they did not receive surgery. This may seem counter intuitive. But if they are excluded and people who die during the same time in the drug group are not excluded, it will not be a fair comparison. This would make surgery appear better than it actually is.
Counting outcomes in the group to which the participants were allocated is called an intention-to-treat analysis. Randomized trials that do not report an intention-to-treat analysis may be misleading.
Example
A large randomized trial of screening for breast cancer in New York provides a clear illustration of how people who take an allocated health action (in this case, a type of X-ray used for screening for breast cancer called screening mammography) may be different from those who do not. The study found similar numbers of deaths after five years among women offered breast cancer screening and those who were not offered screening. Some women who were offered screening chose not to be screened. If those women are excluded from the comparison (not counted in the group to which they were allocated), it appears that there were fewer deaths in the screened group compared to the women who were not offered screening (22 versus 30 per 1,000 women). However, that comparison is misleading because there were important differences between the women offered screening who chose to be screened and those who chose not to be screened. In fact, there were almost twice as many deaths among women who chose not to be screened compared to women who chose to be screened (40 versus 22 per 1,000 women).
Remember: Be careful about relying on the results of randomized trials if outcomes have not been counted in the treatment group to which the participants were assigned.
- Video: Intention to Treat. This is a Sketchy EBM video lasting 5:51 minutes.
- Video: Intention-to-treat analysis: What is it and why is it important? This is a Terry Shaneyfelt video lasting 4:43 minutes.
- Video: Intention-to-treat analysis (ITT). This is a Cochrane Austria video lasting 6:50 minutes.
- Video: What is intention-to-treat analysis? This is an Allen Shaughnessy video lasting 1:11 minutes.
- Blog: People’s outcomes should be analysed in their original groups. Students 4 Best Evidence.