Intention to treat issues
Intention-to-treat (ITT) analyses are widely recommended as the preferred approach to the analysis of most clinical trials. Systematic reviewers often wish to practice this recommendation and plan to conduct meta-analyses according to the ITT principle. But what does this mean, and how might it be achieved?
The ITT principle
The basic intention-to-treat principle is that participants in trials should be analysed in the groups to which they were randomized, regardless of whether they received or adhered to the allocated intervention. Two issues are involved here. The first issue is that participants who strayed from the protocol (for example by not adhering to the prescribed intervention, or by being withdrawn from active treatment) should still be kept in the analysis. An extreme variation of this is participants who receive the treatment from the group they were not allocated to, who should be kept in their original group for the analysis. This issue causes no problems provided that, as a systematic reviewer, you can extract the appropriate data from trial reports.
The rationale for this approach is that, in the first instance, we want to estimate the effects of allocating an intervention in practice, not the effects in the subgroup of participants who adhere to it.
The second issue in ITT analyses is the problem of loss to follow-up. People are lost from clinical trials for many reasons. They may die, or move away; they may withdraw themselves or be withdrawn by their clinician, perhaps due to adverse effects of the intervention being studied.
If participants are lost to follow-up then the outcome may not be measured on them. But the strict ITT principle suggests that they should still be included in the analysis. There is an obvious problem - we often do not have the data that we need for these participants. In order to include such participants in an analysis, we must either find out whether outcome data are available for them by contacting the trialists, or we must 'impute' (i.e. make up) their outcomes. This involves making assumptions about outcomes in the 'lost' participants.
There are many 'formal' approaches to imputing missing outcomes in clinical trials. A review of these is beyond the scope of this course. We shall look at one particular situation that arises commonly and consider some alternative approaches that might be compared in a sensitivity analysis.
Consider the following trial of a Larium-Qinghaosu combination versus Qinghaosu alone for treating malaria that was included in a Cochrane review. Although 20 were randomized to the former and 18 to the latter, results were available only for the 34 people that did not drop out. These were the findings regarding the presence of parasitic infection after four weeks: