Scribd
Upload
17 views

Making Sense of Biostatistics: As-Treated and Intention-to-Treat Analysis

The document discusses two methods for analyzing data from randomized clinical trials: as-treated analysis and intention-to-treat analysis. As-treated analysis only includes subjects who completed the study and adhered to the protocol, while intention-to-treat analysis includes all subjects as randomized regardless of adherence. Intention-to-treat analysis better preserves randomization but may underestimate the treatment effect if subjects do not adhere. The document recommends intention-to-treat analysis as the primary analysis method.

Uploaded by

أيهم غزال
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
17 views

Making Sense of Biostatistics: As-Treated and Intention-to-Treat Analysis

The document discusses two methods for analyzing data from randomized clinical trials: as-treated analysis and intention-to-treat analysis. As-treated analysis only includes subjects who completed the study and adhered to the protocol, while intention-to-treat analysis includes all subjects as randomized regardless of adherence. Intention-to-treat analysis better preserves randomization but may underestimate the treatment effect if subjects do not adhere. The document recommends intention-to-treat analysis as the primary analysis method.

Uploaded by

أيهم غزال
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Vol. 7, No. 3, March 2011 “Can You Handle the Truth?

Making Sense of Biostatistics:


As-Treated and Intention-to-Treat Analysis
By William Irish

Randomized clinical trials are so powerful because they randomly assign subjects to the
treatment and control groups. At the end of the study, the biostatisticians can compare the
safety and effectiveness of the treatment versus the control using statistical techniques that
are valid only if the subjects are randomized.
So far, so good, but what do we do with data from subjects who drop out of the study, do
not adhere to treatment regimens, or otherwise deviate from the protocol? These deviations
might not be random. For example, subjects in the treatment group might drop out because
they experience unpleasant side effects, or subjects in the placebo group might drop out
because they see no improvement in their health. These deviations might occur more
frequently in different groups, e.g., depending on concomitant medications or the initial
severity of the condition.
Prior to starting the study, the researchers must decide what to do with the data from
subjects who deviate from the protocol. There are two primary options:
 As-treated. Analyze study data only from subjects who complete the study and
adhere to protocol requirements.
 Intention-to-treat. Analyze study data from all subjects, regardless of whether
they complete the study and adhere to protocol requirements.
As-treated analysis has the advantage of examining only those subjects with complete data.
However, because of the reasons cited above, as-treated analysis is prone to numerous,
potentially important biases. The results have suggestive value but simply cannot be
trusted.
Intention-to-treat (ITT) analysis preserves randomization but has the significant
disadvantage of assuming what simply is not so — that the subjects adhere to the protocol
and take all their doses on schedule. It therefore understates the treatment effect. We
cannot demonstrate the advantages of a new drug if the subjects do not take it. With ITT
analysis, it is thus imperative to design and conduct studies for maximum subject retention
and adherence.
The Coronary Drug Project (1966-1975) compared the five-year mortality rate between
subjects treated with clofibrate versus placebo following an initial heart attack. For subjects
randomly assigned to clofibrate, the five-year mortality rate was 18%, versus 19% in
subjects randomly assigned to placebo. However, the investigators discovered that many of
the subjects assigned to clofibrate did not comply with their treatment (i.e., took less than
80% of their tablets). In fact, the five-year mortality rate was substantially different
between clofibrate compliers (15%) versus noncompliers (25%). This was an exciting
finding because it appeared to show that patients who take the drug as prescribed can
expect a 40% reduction in mortality. However, on closer inspection, it turned out that
subjects in the placebo group who complied with the protocol saw the same reduction in
mortality. In other words, the compliers had some other unknown characteristic that
improved their life expectancy versus non-compliers. Perhaps the health of the subjects
drove compliance/noncompliance, rather than the other way around.
ITT analysis has a second advantage: It estimates the treatment effect in real-world clinical
practice, where patients do not always adhere to treatment regimens. ITT analysis is thus

Subscribe free at www.firstclinical.com


© 2011 First Clinical Research and the Author(s)
equivalent to comparing treatment policies rather than comparing treatments. In other
words, if patients do not take their pills, e.g., because they cause nausea, it does not really
matter how effective the treatment is in the controlled environment of clinical trials.
For these reasons, primary analysis of the data should use the ITT method. The as-treated
method can be used for secondary analyses, e.g., of the likelihood and potential effects of
nonadherence across subgroups. In performing as-treated analyses, we lose the
comparability ensured by randomization. The results are thus more speculative and best
suited for generating new hypotheses for future study.

Author
William Irish, PhD, is Vice President of Outcomes Research and Biostatistics at CTI Clinical
Trial and Consulting Services. Contact him at 513.618.4057 or [email protected].

Subscribe free at www.firstclinical.com 2


© 2011 First Clinical Research and the Author(s)

You might also like