Designing a PMCF Study: The Question of Sample Size with a Fictional Example

Post-market clinical follow-up (PMCF) studies are strategic instruments for medical device manufacturers in order to meet the regulatory requirements of MDR (EU) 2017/745, to ensure patient safety and device performance. In a previous article, Efor experts provides an overview of the regulatory framework and the main considerations for designing PMCF studies (read it here). The current article proposes an application of these considerations, focusing on sample size determination through a practical, fictional example.

1. Fictional Example: Context

Consider a fictional manufacturer of implantable devices designed to treat urinary incontinence in women. This medical condition is characterized by involuntary urine loss and significantly impacts quality of life. Its etiology is multifactorial, encompassing anatomical, neurological, hormonal, and behavioral factors. Urinary incontinence can be classified into several types, with stress urinary incontinence (SUI), urgency incontinence, and mixed incontinence.

Stress urinary incontinence, related to pelvic floor weakness or urethral sphincter deficiency, commonly arises due to physical exertion or coughing. Urgency incontinence results from an overactive bladder and causes a sudden, compelling need to urinate. Mixed incontinence combines features of both.

When conservative treatments are ineffective, surgical options may be considered, such as retropubic colposuspension, midurethral slings, which are minimally invasive procedures, injections of bulking agents around the urethra to improve the closure mechanism, and sacral neuromodulation, which can be used for cases of refractory overactive bladder. The success rates of surgical treatments vary depending on the type of procedures and patient population. In the European Union, midurethral sling placement is the preferred surgical approach, but is subject to close regulatory scrutiny due to evolving requirements and implementation protocols.

In this example, the manufacturer markets midurethral slings in the European Union for the treatment of SUI in women. In its PMCF plan, the manufacturer has identified the need to set up a PMCF clinical investigation that will better document re-intervention rates and assess patient quality of life. While performance data such as cure rates are relatively well documented, data on re-intervention rates often lack depth due to limited follow-up periods or irregular occurrence patterns.

2. Defining the Primary Objective Frames the Sample Size Calculation

Clearly defining the primary objective is imperative when determining sample size for any PMCF study. In the fictional example defined above, for instance, the primary objective might be to evaluate the rate of surgical re-interventions in women with pure or predominant SUI within a prespecified postoperative timeframe.

Literature data (Imamura et al. 2019 [1]) report re-intervention rates ranging from 2.2% to 3.6% at 1 year and from 1.5% to 9.4% at 5 years depending on whether the midurethral slings were placed via the retropubic or transobturator route. With limited internal data beyond one year, a follow-up duration of at least 5 years is envisaged. Depending on the expected conclusion (i.e. what will be claimed based on the study’s results), the calculation of the sample size will differ, as illustrated in the following scenarios.

• Case 1: A descriptive and non-comparative approach will aim to estimate the proportion of patients who have had at least one re-intervention at 5 years. According to the literature, a proportion of about 5% would be expected. However, the precision of the estimate matters: estimating a proportion of 5% +/- 1% or one of 5% +/- 5% is not the same thing.  Estimating a proportion of 5% ± 1% offers more confidence that the observed rate falls within the anticipated range (1.5%–9.4%), whereas   a proportion of 5% ± 5% has lower confidence, with a greater chance of the actual rate being closer to 10%. The uncertainty around the estimated proportion is known as the confidence interval (CI, typically 95%), and its range depends directly on the number of subjects in the study population. The larger the number of subjects, the smaller the uncertainty around a given estimate. With a population size of 179 subjects followed for 5 years, it is possible to estimate a proportion of 5.00% with a 95% confidence interval (Clopper-Pearson method) ranging from 2.30 to 9.30%, i.e., at the upper limit of the literature data.

• Case 2: A comparative approach will aim to determine if the estimated re-intervention proportion in the study population is significantly lower than 9.4%. Anticipating a proportion of 5.00%, a sample size of 295 subjects would allow for the detection of a significant difference of -4.40% compared to a reference of 9.40% using a two-sided exact test with an alpha risk of 0.05 and a power of 80%.

Depending on whether the goal is to estimate a proportion with desired precision or to formally test for a statistically significant difference compared to a given reference, the number of subjects will be different, and the formulation of the study’s conclusion will also differ. In the first case, the study would only provide an uncertainty around the observed proportion in the study population and discuss the result in light of the literature data. In the second case, the study might conclude that the rate is significantly lower than a reference that may come from the literature, providing a more robust conclusion.

The statistical approaches described above are relatively simple because they do not consider issues related to lost-to-follow-up subjects, which are common in longitudinal studies. A time-to-event approach, such as Kaplan-Meier survival analysis or Cox’s proportional-hazards models could also be considered. Furthermore, the management of missing data such as imputation techniques is not addressed. Finally, there may be a question of formally differentiating the route of sling placement in the analysis and planning distinct cohorts for which a sample size would be determined a priori.

Once the primary endpoint, its analysis, and the sample size calculation are determined, the data collection methods must be meticulously outlined. In the above examples, data collection would include:

• Safety data: rates of complications or adverse device effects like sling erosion.
• Performance endpoint: subjective and objective cure rates.
• Quality-of-life data: administered through the Incontinence Quality of Life questionnaire

Data would be collected according to a pre- and post-operative visit schedule for up to 5 years using an electronic data capture system such as an eCRF (electronic case report form) to ensure secure documentation while facilitating central monitoring.

3. Conclusion

Designing a PMCF clinical investigation (like any clinical study) requires technical, medical, clinical, methodological, and regulatory expertise. It should align with an anticipated and structured PMCF, considering the state of the art and existing clinical data. Thorough preparation is essential, including clearly defining objectives, sample size calculations, and data collection methods, while leveraging the right expertise.

Need help?

Need help building a PMCF study? The dedicated experts within our clinical CRO, Soladis Clinical Studies by Efor, can support you with:

• Literature review
• Objective definition, study design and methodology planning
• Drafting clinical investigation and core study documents
• End-to-end implementation and conduct of PMCF clinical investigation
• Expert consultation via our hotline for specific questions requiring a rapid response

Contact us today at the following address: TechnicalDivision@efor-group.com to ensure your PMCF clinical investigation reflects the appropriate scientific and regulatory standards.

[1] Imamura M, Hudson J, Wallace SA, MacLennan G, Shimonovich M, Omar MI et al. Surgical interventions for women with stress urinary incontinence: systematic review and network meta-analysis of randomised controlled trials. BMJ 2019; 365