Intention-to-treat analysis
The intention-to-treat analysis is a method of medical statistics that describes a certain data evaluation principle (in German-language medical journals often referred to as intention-to-treat for short ).
It is a principle of analysis of data collected in controlled, randomized clinical trials . As a rule, these studies compare the effectiveness of a particular therapy between the active ingredient ( verum ) and placebo, or between intervention and sham treatment. All patients who were included in the verum group or the placebo group at the beginning of the study are also included in the result evaluation at the end. The data of patients who drop out during the ongoing study - for example at the request of the patient or because another therapy was necessary - and who therefore do not participate in the study treatment until the planned end of the study or who have changed the treatment group for any reason must also be included in the evaluation can be recorded and evaluated in the group that was assigned to them.
This evaluation principle often seems a bit illogical at first, but most closely corresponds to the conditions in daily practice. Often, for example, prescribed medication is inadvertently or intentionally left out.
If, on the other hand, patients were excluded from the data who did not take a drug, for example because their expectations were not met or they are no longer able to do so, the study result would be falsified. In each individual case, the reasons that led to the patient dropping out of the study must always be checked. In the case of death in particular, the question arises whether the death might not be related to the therapy administered. Cases of doubt are to be assessed as failure of therapy.
Or explained using another example: a clinical study wants to compare two different diets (diet A and diet B). Both diets are aimed at weight loss. For this purpose, the study participants are divided into two groups ("study arm A" and "study arm B"), each receiving diet A or B. Diet A is effective, while Diet B mostly does not have the desired effect, neither of which is known at the start of the study. In the course of the study, it is predominantly those study participants who will leave the study who do not have the desired effect (weight loss). As a result, the results in both study arms appear better than they actually are, because it is mainly the study participants who remain who the diet worked for. This bias applies to a much greater extent to study arm B because a particularly large number of study participants leave the study there. The difference in effectiveness between Diet A and Diet B will therefore appear smaller than it actually is.
Intention-to-treat therefore means that the data of all patients who were previously intended ( intention ) to treat ( to treat ) must also be evaluated afterwards . This happens regardless of whether the treatment was actually carried out in the planned form ( per protocol ).
The opposite of this method is the per-protocol analysis , in which only the data of those patients are evaluated who have behaved in accordance with the test plan (in the narrower sense, they were actually treated with the intervention or medication provided in the test protocol). As-treated analysis is a subtype of per-protocol analysis . Here, the data of all the test persons who received the intervention or medication to be tested in the course of the study are evaluated, regardless of which study group they were originally assigned to.