Making sure your study is successful: Formulating hypotheses correctly in a confirmatory trial
Virtually all late-stage trials—and certainly pivotal trials—are confirmatory in the sense that hypotheses are formulated in the protocol and tested using the data subsequently obtained in the trial. When designing a confirmatory clinical trial, hypotheses may often seem relatively obvious and, as a result, little effort is invested in them. However, a well-designed hypothesis can sometimes mean the difference between the success and failure of the study. It is, therefore, essential that when formulating trial hypotheses their implication for the study’s success be assessed.
Here are some common misconceptions about hypotheses:
- For a given medical product, the hypothesis can remain the same from study to study – it’s as simple as copying and pasting
- We can write a draft hypothesis in the study protocol and change it later on
- Hypotheses can be designed with minimal consideration of other features of the study design – hypotheses are only indirectly related to features such as:
- Sample size
- Number of study arms
- Number of analyses (e.g. whether or not interim analyses will be done)
- Methods designed for imputing missing data in the trial
- The population analyzed in the trial (e.g. intent to treat (ITT), per protocol (PP), etc.)
In reality the hypothesis needs much foresight. So, what makes a well-designed hypothesis? A hypothesis that is…
- Individually crafted by the trial statistician, in consultation with clinicians and other study personnel, for the particular study at hand;
- Well thought out so that it relates directly to study goals, thus ensuring that later changes will be unlikely; and
- Developed to take into account all relevant aspects of the study design, including sample size, number of study arms and computation of the endpoint.
The chance of success of a clinical trial is as good as the design of its hypothesis, so it is important to ensure that your study has a strong one.