Sample size estimation

Hello,

I want to design a registry study. However, the sample size will need multiple country registries as any one country will not have enough sample size. How do I determine the sample size for each country and when to stop so that the meta-analysis later will have enough power? If only study level data can be used for the meta-analysis, will that cause a problem? How do we convince the agency?
Thank you so much!

Kind regards,
Huifang

Hi,

Generally, sample sizes for registries are predicated not upon a pre-specified null hypothesis, as you would have in a clinical trial, but on the tension or equilibrium between budgets, enrollment timelines, total running time for the registry, and the desire to achieve a reasonable level of precision (not power) in the estimation of key parameters. The latter, perhaps, being the easiest part of the process.

The general term of art, is that you will have a “convenience sample” of some pre-defined, target size.

In a multi-site, much less a multi-country setting, there is the question of whether or not the data within individual countries are representative of that country’s experience, and then, how is the overall aggregate dataset biased by the contributions of the subsets. Do you cap the participation of any site or country, so as to not materially bias the aggregate towards the subset, and if so, at what level, or do you allow each to enroll as many patients as they can, within your enrollment timeline, to increase the likelihood of hitting your enrollment target?

The key issue in the above, is whether or not you will be able to generalize your findings to populations outside of what you have in the registry, or are those findings idiosyncratic of the sample in your registry.

Also, bear in mind that enrollment is generally the biggest challenge for any study.

Part of the pre-study planning process is to get some reasonable estimates of the prevalence of the underlying disease in the populations of interest, get a sense for per site volumes in each geographic area, and then make some presumptions of how many patients can be enrolled at each site, per unit of time, within your overall enrollment time frame.

If you already have a finite budget that you have to work within, you need to model how all of those assumptions fit within the budget, and adjust accordingly. If you are first looking at creating a budget, then you need to model those assumption to conform to your research goals, and estimate a budget based upon those factors. You also need to consider the varied regulatory issues in each country, and what that means operationally and financially for your registry.

It will typically take a multi-disciplinary, team approach, to move towards defining all of these parameters so that you end up with a reasonable design, target sample size, and a protocol that reflects these various factors.

5 Likes

Wow I can’t think of a better person to answer than question than you. Thanks for being here Marc!

1 Like

Frank, that is very kind of you to say. Thank you!

1 Like

Dr. Schwartz,

Thank you so much for providing your insights. I really appreciate it. If there is an alternative to do a registry in one large country (assuming that country would have have enough sample size), that could be a better choice. Do you agree?

Thank you again.

Hi,

That really comes down to your research goals, and to an extent, your budget.

If, for example, a key goal is to be able to assess variations in patient characteristics, the practice of medicine and outcomes across countries or perhaps large geographic regions (e.g. Asia, Europe, North America, etc.), then you need an international registry.

On the other hand, if your belief (perhaps based upon publications, etc.) is that these attributes are largely similar across countries, and you can satisfy a reasonable enrollment target within a single country, especially if you are budget constrained, that would be a reasonable and focused approach.

Even within a single country, there may be some level of variation in these characteristics across multiple sites or types of sites, for example, academic medical centers versus community medical centers, if that makes sense in your domain, and you can get a sufficiently diverse set of sites to participate. That may not be an a priori goal, but may be an exploratory finding that you observe, and may be of interest.

Another strategy to consider, is to start a phase 1 approach with a single country, demonstrate success, and then, presuming you have options for additional financial and logistical support in the future, expand to other countries later. I have seen that done successfully as well.

The likelihood is that you, or members of your team, already have some idea of the sites that you would target for participation, based upon some knowledge of their relevant patient volumes and existing relationships with physicians who would serve as the PIs at each site. If, for some reason, you do not, then an initial survey of these sites, to assess these characteristics, would be warranted.

That would help to guide an informed decision as to what implementation model to consider, and what the budgetary implications may be, so that you can maximize the likelihood of success and ultimately, achieve your research goals.

3 Likes

Dr. Schwartz,

Thank you again for sharing your insights! Much appreciated.

Best wishes,
Huifang

1 Like