Intention to treat analysis
Editor-In-Chief: C. Michael Gibson, M.S., M.D. [1]
Overview
In epidemiology, an intention to treat (ITT) analysis is an analysis based on the initial treatment intent, not on the treatment eventually administered. The ITT analysis includes all patients randomized to a therapy irrespective of protocol deviations, discontinuation of study drug, drug administration errors, cross-over to another strategy, or withdrawal from the study by the subject. For example, in a trial comparing medical therapy to angioplasty for coronary artery disease, those patients who were initially randomized to medical therapy but who crossed-over to receive an angioplasty instead due to a failure of medical therapy are analyzed as part of the medical therapy cohort. This form of analysis differs significantly from a per protocol (PP) analysis in which only those subjects are included who actually received study drug as specified in the study protocol.
Philosophy of an ITT Analysis
In a way, an ITT analysis tests a strategy of administering the therapy, rather than the therapy itself. It is based on the assumption that, as in real life, sometimes patients cannot all tolerate the initial treatment, they may stop the initial treatment, there may be errors in administering the initial treatment or they may switch to another treatment despite the initial treatment. For the purposes of analysis, the reasons why the patient did not receive the treatment are ignored.
Rationale
Randomization of subjects at the beginning of a trial assures that any imbalances in risk factors, demographics and confounders will be minimized. An ITT analysis preserves that initial randomization, and avoids the effects of crossover and drop-out, which may break the randomization to the treatment groups in a study. Patients who discontinue study drug therapy may do so due to side effects or a perceived lack of effectiveness. As a result, the remaining population is enriched with those patients who could tolerate the drug or perhaps had their symptoms improved on the treatment. Simply put, this remaining "per protocol" cohort harvests out those subjects who were "success stories", and eliminates those patients who failed. An ITT analysis is a more conservative form of analysis because it incorporates patients who failed therapy, could not take therapy and who may have crossed over to other therapies. A per protocol analysis of those patients who successfully followed the protocol may overestimate the efficacy of a treatment.
Modified Intent to Treat Analyses
The Intent to Treat Analysis can be modified in some situations.
Enrollment of Patients Who Were Actually Inelligible for the Protocol
In some trials, patients may be enrolled prior to the ascertainment of lab values or diagnostic studies that would exclude them from participation. An example would be a trial of an agent to treat adult respiratory distress syndrome whereby patients were enrolled before a chest x ray was performed. In this scenario, it would be reasonable to remove those patients from the strict ITT analysis and perform a modified ITT analysis (mITT) of efficacy. The decision to perform an mITT analysis and the criteria by which an mITT analysis is to be undertaken should be pre-specified in the protocol and statistical analysis plan. While subjects who did not meet entry criteria may be excluded in an mITT analysis of efficacy, they have been exposed to study drug, and they are traditionally included in an analysis of safety.
Limitations of an ITT Analysis
If there are a large number of patients who withdraw from treatment, this will dilute the benefit of the therapy being evaluated. This dilution can be overcome by an increase in sample size.
Appropriate Application of the ITT Analysis
An ITT analysis is appropriate if the crossover, withdrawal, and protocol non-compliance rates are low. Full application of intention to treat can only be performed where there is complete outcome data for all randomised subjects. Complete ascertainment of the primary endpoint in all randomized patients in large clinical trials is extremely rare. Although intention to treat is widely cited in published trials, it is often incorrectly described and its application may be flawed.
Patients who discontinue study drug are included in an ITT analysis. The duration of follow-up varies. In a strict ITT analysis, event data after therapy discontinuation is captured and analyzed until the trial is completed. In a modified ITT, event data after therapy discontinuation is captured and analyzed for a pre-specified period of time, say 30 days after study drug discontinuation. If a patient withdraws consent, and no longer wishes to be contacted, then efficacy and safety data are censored at time of consent withdrawal if no further clinical outcome data are available.
There are eithical issues regarding the follow-up of patients who have withdrawn consent. It is a matter of ongoing debate as to whether a patient who has withdrawn consent can be contacted at the end of the trial to assess their vital status (whether they are dead or alive). The majority view is that this is not appropriate to contact the patient, unless the patient is approached utilizing an IRB approved document. It is also a matter of debate as to whether publicly accessible death registries can be used to ascertain the vital status of the patient if they have withdrawn consent.
Because withdrawal of consent, loss of subjects to follow-up study and drug discontinuation have such a profound effect on an ITT analysis, all reasonable efforts should be made to reduce the rates of these occurrences. A run in period at the beginning of a trial may allow the identification of patients who tolerate the therapy. Patients should be educated regarding the potential side effects and the frequency of follow-up visits so that they can fully comprehend the likelihood of their ongoing participation.
Per Protocol Analysis
In contrast, a per-protocol or efficacy subset analysis selects the subset of the patients who received the treatment of interest--regardless of initial randomization--and who have not dropped out for any reason.
References
See also
External links
- Intention to Treat - Bandolier's definition
- Intention to Treat - Tufts.edu
- What is meant by intention to treat analysis? Survey of published randomised controlled trials - BMJ