Welcome to Quantics’ Clinical blog. This blog will explore some of the statistical aspects of clinical trial design including study design, sample size, data management and analysis planning. The aims of this blog are not only to provide thought provoking insight into some of the latest statistical developments for clinical trials but also to create a knowledge base to help those who are new to the area develop a foundation to feel more confident moving forward.
The statistical methodologies used in clinical trial design are becoming increasingly complex and Quantics has a rich history of translating statistics into a language that everybody can digest. In these first posts we will provide an introduction to the part that statistical thought plays in the design of clinical trials from Phase 1 through to Phase 3 drug development.
In this first blog we will focus on sample sizes. What your statistician needs to know and why.
A clinical trial evaluates the safety and effectiveness of a new treatment intervention or compares the effectiveness of a new treatment to that of current best practice or other control group. Clinical trials are required for every type of drug to treat every type of disease.
Statistical thought is vital to any clinical trial. The statistician ensures the design of the study is appropriate for meeting the study objectives. They also ensure a sufficient number of patients is included in the clinical trial to answer its primary objective, but not so many that patients might be subjected unnecessarily to an ineffective or unsafe treatment or that the study would be conducted over an unnecessarily long period of time. They describe in advance the statistical methods to be used to analyze the data, and, when the study is completed, the statistician will analyze the data and write a report about the results.
Our earliest involvement in a clinical trial is usually at the stage of determining how many patients are needed. What information does the statistician need the investigator to provide us with?
1) The primary endpoint for each patient
E.g. Haemoglobin level after 3 months of treatment
2) The variability of the primary endpoint in the study population
E.g. Standard deviation of Haemoglobin level in children with anaemia
3) The impact on the primary endpoint that would be clinically relevant
E.g. Haemoglobin level after 3 months of treatment up by at least 2 g/dl
All this information, combined with the statistical method to be used to compare the primary endpoint between the groups, can be combined to produce the required sample size to achieve the required power, which is defined as:
the chance of a statistically significant result when the true difference between the treatments is at least the clinically relevant level:
o this chance is called the power of the study
o it is usually set at 80% or 90%.
Often the variability if the endpoint is not known very accurately. In that case it’s possible to set a preliminary sample size and then review the variability as soon as enough data has accumulated in the trial. The sample size can then be adapted if necessary. Look out for our blog on adaptive clinical trials where we will discuss this approach.
In more complex trial designs such as those with repeated measurements or those with time-to-event endpoints, it can be really useful to use simulation to find the best sample size. Look out for our blog on simulation techniques in clinical trial design for this.