How many people do you need? Designing the foundations of a clinical trial

How many people do you need? Designing the foundations of a clinical trial

From the outset, designing a trial must take your end goal into consideration. If the trial design, number of subjects, or data collected is not compatible with the type of planned statistical analysis, you may not be able to demonstrate efficacy and safety of your product. Therefore, it is important to understand what your ultimate goal is, and work backwards from it, as the planned statistical analysis of your data will inform how you set your trial up.

However, this is not an easy process as all clinical trials are different - smaller or early phase studies have diverse needs; larger or late phase studies need to mitigate risk while still being cost effective; and safety studies may not need any statistical analyses at all. When planning a study, one must read up on guidelines for your specific study design and indication, such as those provided by the FDA. Once you are aware of what is required, you then need an understanding of several interrelated aspects of the study: its aims, budget, duration, and overall resources available to you.

Study goals and objectives

The first consideration is your primary study goals. These will determine the final study design. It is important to focus on your main goals, such that the overall study results are not diluted. Firstly, consider how this data will be used. Perhaps you are using it as primary data for an FDA submission, supporting data for marketing claims, or is your goal to guide your development pipeline and/or assess proof of concept for a particular therapy. You will need to consider what specific measurements will be assessed.

Your budget, from a resourcing standpoint, may be a limiting factor in your study design, and is why the planning phase is so important. The timelines associated with clinical trials are often underestimated. For example, poor recruitment may delay results and thus increase expected study duration and costs. In addition, delays can have follow-on effects, as tasks have dependencies before being started. Finally, take stock of the resources potentially available to you to carry out the study.  By carefully considering these aspects, you will be able to design a study that maximises both its impact while keeping it cost effective and achievable. 

Considerations for your statistical analysis

When designing your study, another key aspect will be how many subjects are to be enrolled. Your choice has significant implications - advertising, budget, number of sites, study duration, and of course the power of the eventual results all depend on this number. Keep in mind that multiple trial sites will require more time and resources to set up and manage, and the volume of data gathered will need proportionally more data cleaning. If you select too low an estimate, however, you may not have the power available to detect the effects of your therapy, wasting both time and money.

When you are comparing two treatments, you often do not know (and do not want to assume a priori) whether one group's outcome will be higher or lower than the other. A 2-sided test is the standard for clinical trials as it considers both a potential increase and decrease in one outcome compared to the other. This provides a greater certainty that a significant result was not simply due to chance. A 1-sided test does have applications, however, they must be statistically and clinically justified or regulatory bodies may also raise queries as to why a 2-sided test was not used instead.

Characteristics of your primary outcome

Your anticipated standard deviation is a measure of how variable you think the data from your study will be. This can be determined from previous research publications in the same therapeutic area. A low estimated standard deviation means that you will likely need less subjects in your study, but be careful - if you have underestimated the variance of the study and in your actual study it is higher, you could fail to detect an effect. On the other hand, high variability in an outcome means you will need to design your study carefully to minimise this. In addition, there are also certain statistical techniques that can mitigate high variability of results in studies but must be accounted for in the study design ahead of time.

Your anticipated effect size is how large you believe the difference is between your placebo and treatment groups. For example, you may expect your new pain relief product will reduce pain by 40% compared with placebo which may reduce pain by 25% - an expected difference of 15%. The bigger the difference, especially in proportion to the standard deviation, the easier it will be to detect an effect. Low effect sizes could mean probing the same effect in a multitude of ways to build up a body of evidence, or increasing the sensitivity of your measurements. Large effect sizes, while beneficial if true, must not be overestimated. In cases where the effect observed is not as large as thought, a non-significant result may be found.

Overall sample size

The factors discussed above will feed into the calculation of the overall sample size estimate for a study. A study that requires a small sample size will cost less, may only require one study site and it may be quicker to recruit suitable subjects.

A study requiring a large sample size may require several sites and in order to recruit sufficient subjects you may need to extend the reach of your clinical trial within Australia and internationally. Significant costs will be required to establish multiple sites adding to the cost of data collection. Advertising may be required to support patient recruitment over a reasonable timeframe. Speaking generally, a larger study has more at stake due to the increased time and money involved throughout all stages of the trial.

When it comes to clinical trials, there is no "best" way to get the results that you are looking for. It is always a balance between many factors, and an experienced team is required to design an effective study. As such, it can be challenging to find someone who will provide their expertise while respecting your vision and goals. Datapharm Australia understands this, and acknowledges that designing a study is a complex and involved task. Therefore, if you would like to continue the discussion with our experienced statistical team, please contact us today!

Disclaimer: The taxation and legal advice given here are general in nature only. Before acting on this information, seek professional advice specific to your current situation.

Recent Posts