Post Marketing Clinical Follow up (PMCF) study is a continuous process that updates the clinical evaluation which is the assessment and analysis of clinical data pertaining to a medical device to verify the clinical safety and performance of the device when used as intended by the manufacturer. Usually the device studies are used to acknowledge the information to the manufacturer mainly on the below mentioned points.
- To approve the safety and performance of the device throughout its expected lifetime
- To find out the previously unknown side-effects and to track the identified side-effects
- To identify and analyze emergent risks on the basis of authentic evidence
- To ensure the continuous acceptability of the benefit/ risk ratio
- To identify possible systematic misuse or off-label use of the device, with a view to verifying that the intended purpose is correct
- To evaluate the safety and effectiveness of the device in the population expected to be indicated
Finally, a study would be considered as successful if both the safety and effectiveness endpoints are met throughout the expected lifetime of the medical device1.
In order to conduct a PMCF study on medical devices, the manufacturer should have clear idea about the sample size needed for the study. There are certain important factors involved in the sample size calculation. This paper attempts to describe the factors to be considered for sample size calculation at the time of planning the study.
Since there are different formulae available for sample size calculation, the manufacturer should select the appropriate formula depending upon the situation. At this stage, an experienced statistician’s help may be sought for the selection of the most appropriate formula for sample size calculation. Depends up on the study objective and the primary endpoint, the statistician will select the most appropriate formula for sample size calculation. However, the statistician will have a lot of questions in his mind about various inputs for sample size calculation. A dialogue should take place between the statistician and the manufacturer. Statistician would like to know the expected outcome (e.g. number of failure of devices over a period of time, say, for next 5 years) from the proposed study. Usually, the answer from the manufacturer would be that a small proportion (say, 5%) of devices would be failed or malfunctioned during the study period.
The next question from the statistician would be about the expected precision of the estimate. The manufacturer would say that the expected precision would be, for example, about ±2%. This figure could be from their past experience or any other published study already done by others.
Also statistician would like to get the expected drop-out rate for the proposed study. Manufacturer would provide this figure from their past experience. Usually the PMCF studies require long duration follow up of the study subjects, naturally the drop-out rate would be more, say about 25%.
Along with the above inputs, statistician would further assume a 95% confidence interval also for sample size calculation. Let us assume that the objective of the study is to estimate the proportion of failure of devices from the study. Then the below formula2 can be used for estimating the sample size.
n= [Z2 * P (1-P)] / d2
where P is the expected proportion of failure rate; d is expected precision and if we assume 95% confidence interval, then Z = 1.96.
Further, if we assume that a drop-out rate of x%, then the sample size adjusted for drop-out rate would be
n´ = [n/ (100-x)]*100
As an illustration, if P = 5%, d=1% and 95% confidence interval; then the sample size would be
n = [(1.96)2 * 0.05 (1-0.05)] / (0.01)2
= [3.8416 * 0.05 *0.95]/ 0.0001 = 1824.76
Now by adjusting for the drop-out rate of, say 25%, the adjusted sample size would be
n´ = [1824.76 / (100-25)]*100
= 2433.01 = 2433
As per the above calculation, we have to recruit about 2433 subjects into the study. However, the next question is about the availability of subjects for the study. Once again, the statistician should initiate the dialogue with the manufacturer about the availability of subjects and study duration. For example, in order to recruit 2433 subjects into the study, manufacturer should select so many sites and also it would be a costly affair. Hence the statistician can prepare the sample size with different inputs in a tabular form so that the suitable sample size can be decided by both statistician and the manufacturer.
Thus the sample size calculation is not a mere statistical exercise, but a lot of discussion should take place to arrive a suitable number of subjects to be recruited for the study.
