Understanding Intention to Treat Analysis
and Per Protocol Analysis
Vikas Roshan1* and Sadamoto Zenda2
1Department of Radiation Oncology, Shri Mata Vaishno Devi Narayana Superspeciality Hospital, India
2Department of Radiation Oncology, National Cancer Center Hospital East, Japanq
Submission: May 23, 2018; Published: June 25, 2018
*Corresponding author: Dr. Vikas Roshan, Consultant Radiation Oncology, Department of Radiation Oncology, Shri Mata Vaishno Devi Narayana Superspeciality Hospital, Kakryal, Katra, Jammu and Kashmir, 182320, India, Email:[email protected]
How to cite this article: Vikas R, Sadamoto Z. Understanding Intention to Treat Analysis and Per Protocol Analysis. J Tumor Med Prev. 2018; 3(2):
555608. DOI: 10.19080/JTMP.2018.03.555608
Randomized control trials (RCT) are usually done to see treatment efficacy and safety profiles. It has two critical flaws, i.e., patient’s non-compliance and loss of data concerning measuring outcomes and the solution to this problem is Intention to Treat analysis .
While conducting clinical trials we come to know about the complexities of analyzing results. There is a lot of protocol violations seen, and as a result, we require some statistical principles to evaluate data and then comes the role of intention to treat analysis (ITT). In randomized controlled trials, ITT use is highly recommended [1-3].
It includes all the patients who are randomized in statistical analysis and usually these patients should be analyzed as per their allocated treatment group even if the patient has refused or discontinued their intervention.
In head and neck cancer study author compared three weekly concurrent chemo radiation versus weekly concurrent Cisplatin. In this study total, 300 patients are randomized into arms of 150 each, but patients who completed treatment are 133 in three weekly arms and 141 in the weekly, but patients included in this trial for ITT are 150 in each division [3-5] (Figure 1).
There is significant confusion in the understanding rationale
that why we are including the patients in the analysis even if these
patients have not received treatment regimen.
These patients are included in the analysis to maintain the rules
of randomization because if self-selection excludes patients, then
the benefits of randomization are lost as randomization is done to
balance the factors in each arm that can introduce bias later on.
The second point of discussion is if we exclude non-adherent
patients like in head and neck cancer study, 34 patients excluded
from weekly Cisplatin and 18 from three weekly Cisplatin. This
point will introduce the bias, and it can be concluded that patients
who receive treatment have better outcome ignoring whether
treatment is useful or not.
ITT has widespread use in clinical practice as it simulates the
real clinical environment because in actual practice also patient
doesn’t often stick to the treatment depending on the various
ITT seems very attractive approach to deal with analysis, but
there are also few issues associated with it. Suppose if inclusion
criteria are not framed strictly and consequently at time of review
there is a problem of excluding lot of patients and which in turn
leads to violation of randomization.
The problem that we face with ITT is that it overestimates
the effect of acceptance of non-compliance, dropping patients
who do not adhere to treatment and protocol deviations. Patients
who completed treatment as usual responded well and show the
substantial effect of treatment. ITT is today standard for analysis in
the clinical trial.
ITT preserves the sample size and hence protect the statistical
power. It also minimizes type I error, and it is possible to generalize
the results of RCT to the general population [5,6].
In contrary to above discussion Per Protocol analysis strictly
adhered to the patients who stick to the treatment, so only those
patients are analyzed who completed treatment. It gives the
reliable estimation of effect between treatment and result. PPA is
complicated to apply in clinical practice, and as per evidence, it is
weak as compared to ITT. CONSORT (Consolidated Standards of
reporting trials) guidelines strongly recommend providing the
details of both estimates in clinical trials because when ITT and
PPA come to the same conclusion, the confidence level in study
results gets increased. There is modified ITT also available that will
strictly deal with attrition and allow us to drop patient even after
randomization like those patients who never started treatment
after randomization. There are limitations to it that it is purely
subjective and may allow users to manipulate data. There are no
strict guidelines for application [7-10].