Sample size has always been one of the first design choices a sponsor has to make for a clinical study. But the calculation of sample size is not the most straightforward process and often leaves clinical teams wondering which method to use. As an Electronic Data Capture vendor for medical device companies, we are experiencing greater interest in EDC for PMCF studies and surveys. But we’re also seeing a major increase in requests for feedback on various study design choices - like for example:
How many subjects do I need for my study/survey?
This blog post will provide a sneak peek into the inner workings of The Medical Device Sample Size Cookbook, access to which can be found at the end of this blog. In the e-book, we go in-depth into the process of sample size calculation for medical device studies.
In most cases, it’s impossible to conduct a clinical study that includes the total population of interest (i.e. target population), especially if the population size is indefinite. Instead, we conduct studies with a sample from the population, in the hope that the information gathered about the sample will enable us to make inferences about the total population.
So sample size calculation is actually the process of determining what is the minimum number of samples that would make a studies outcome/s statistically significant. (make your results more reliable in statistical terms)
To calculate the sample size for a clinical study, we use statistical equations that employ inputs that mirror the population(s), study objective, and design. But the problem with the calculation is that it’s based on assumptions on these inputs, and not necessarily the ‘best’ or ‘correct’ values. Therefore, the calculation is only as good as the assumptions you make. If your assumptions are too far off you might end up with a too small or too large sample size, which may affect the study outcomes and conclusions. Thus, to minimize the likelihood of your calculation returning a too small or too large sample size, your assumptions must be evaluated thoroughly.
The EU MDR describes the requirements for both a Clinical Investigation Plan (CIP) and a PMCF plan. In both cases, medical device study sponsors must document the various design choices for a clinical study, such as the choice of sample size, and provide a rationale for the appropriateness of the procedures and the methods used. Even though you are not conducting a clinical study as a part of your PMCF plan, but a survey or a cohort of some kind, you still need to provide rationale or justification for your design choices - including the sample size. Many ethical committees also require justification for the choice of sample size before for ap- proving a clinical study. And depending on the study type, and the market approval pathway in the United States, the FDA requires justification of the sample size calculations as a part of a statistical analysis plan as well.
Many fail to realise that it’s the combination of both the statistical and clinical assumptions that determines the sample size applicability. By leaving out either of the two, you are only providing partial justification on your sample size.
To calculate the appropriate sample size for your study, you need to have knowledge of the expected study results. This is used to determine how the study should be conducted, what data should be collected, and how its results should be analysed (and statistically tested). When it comes to the sample size calculation, you apply statistical equations that are derived from the hypothesis you want to test. But before the equations come into play you need to keep in mind that a few other thing have an impact the method of sample size calculation like:
When it comes to conducting studies whose purpose is to confirm a theory or claim for clinical performance or safety, such as pivotal trials, clinical investigations for market access, or PMCF, you will need to provide justification for sample size calculation (at least according to the EU MDR).
Every study has an objective, i.e. whether you want to test your device's superiority, non-inferiority, equivalence, or another effect/difference. The study objective will have a direct impact on the choice of sample size calculation (the equation) and your statistical hypothesis test.
Depending on the nature of the endpoints in your study, sample size calculation equations have to be adjusted, especially if the endpoint involves multiple comparisons. In some cases, you can use different types of endpoints to evaluate the same clinical outcome. But depending on the endpoint you choose, the sample size can change drastically.
Whatever design you choose, you need to know that it will impact the sample size calculation. For example, a crossover study and a parallel two-group study require two different equations to calculate the sample size, the crossover study only requires a part of what a two-group parallel study needs in terms of sample size.
There are numerous different statistical tests that can be used to analyse and test the results of your study. Usually, this must be de-ned as a part of the statistical analysis plan and is highly related tothe endpoints you are looking to collect. Depending on the statistical tests you choose for your study, you will need to use the appropriate equation, and thus extract the sample size from the equation to cal- culate the sample size for the study.
For most sample size calculations you need five basic inputs. All five inputs are values (or numbers) that you must define or make assumptions on which depend on the study objective, endpoints, and design. Sample size calculation is very sensitive to the inputs used, which invites a large amount of imprecision or error. Due to the nature of some of these inputs, it’s rare to have an ‘accurate value’. But to minimise the imprecision, you need to carefully evaluate each input from both a clinical and statistical standpoint, using existing know-how of the domain of interest, the clinical pathway/setting and your device.
Apart from the ve basic inputs, there are additionally three other factors that should be taken into account for sample size. Depending on the study objective and design, these factors can impact the final number of subjects needed for your study.
For a more detailed explanation of the above mentioned inputs download The Medical Device Sample Size Cookbook
In conclusion, sample size calculation for medical device studies should not be treated lightly and there’s no ‘all-in-one’ solution to determine a clinical study sample size.
To pass the scrutiny of regulators and Notified Bodies, companies will need to provide statistical and clinical justification of the sample size calculation. It’s not enough to solely document the statistical assumptions made, but it’s the combination of both statistical and clinical (scientific) assumptions that determines the sample size calculation applicability. Sample size justification is required for both pre-market clinical investigations and post-market activities.
Not all medical device companies have the privilege of employing a biostatistician or a data scientist in-house to assist with sample size calculation. This forces clinical teams to look for advice outside of their organization (which is also recommended).
Knowing this, we felt it was important to provide clinical teams with insights into the area of sample size calculation, in the hope that this could simplify the process, and better prepare for discussions with biostatisticians.
Which leads us to the the purpose of the e-book which is to explain the process of sample size calculation, in simple language, and exemplify how statistical methods can be used to calculate sample size for medical device studies
If you are interested in how we help clinical teams with their clinical data management contact us