[Clinical Trial] Factors Affecting the Estimation of the Sample Size of Clinical Trials of Medical Devices

The sample size estimation is an extremely important part in the clinical trials of medical devices, which is directly related to whether the research conclusion can reach the expectations.

Although mathematical statistics indicate that the larger the sample size, the closer the sample statistics are to the corresponding overall parameters. However, an excessively large sample size reduces the feasibility of the study and increases the difficulty of quality control, while an excessively small sample size makes it difficult to ensure estimation accuracy and inspection efficiency.

Therefore, sample size estimation is a trade-off process between cost-benefit and inspection efficiency.

At present, according to the current "Quality Management Standards for Medical Device Clinical Trials" issued in 2016, no specific sample size reference value is given, only the instructions in some medical device clinical trial guidelines. However, in the "Guidelines for Technical Guidelines for Clinical Tests of In Vitro Diagnostic Reagents", reference values for sample sizes are given for IVD.

The sample size of the medical device clinical trial is related to the following factors:

1 Determine the design of the clinical trial

1.1 Common types of hypothesis trials for clinical trials of medical devices are: superiority trial, equivalence trial and non-inferiority trail;

1.2 Common design types: parallel control design, single-arm objective performance criteria, paired design and cross-over design..

2 Main evaluation indicators

Effectiveness indicators or safety indicators can be used as the main evaluation indicators, which can be qualitative or quantitative. Qualitative data needs to know the frequency, and quantitative data needs to know the mean and standard deviation.

3 Effect size

It is an important parameter required for sample size estimation. Ways to determine this parameter include: small sample pre-test, previous research of this project and related literature review.

4 Statistical characteristics

4.1 Hypothesis test type I error (ie: the size of the test level α);

4.2 The test II error of the hypothesis test (ie: the size of the test level β) or the size of the test performance (1-β);

4.3 One-sided and two-sided tests. When other conditions are certain, the sample size of the one-sided test is significantly lower than that of the two-sided test.;

4.4 Balanced or unbalanced design

Table1 Relationship between influencing factors and sample size

Influence factor	Value level	Sample size
Test level α	Smaller ↓	Bigger ↑
Test level (1-β)	Bigger ↑	Bigger ↑
Effect size	Smaller ↓	Bigger ↑
One-sided / two-sided test: under the same conditions, the one-sided test is significantly smaller than the two-sided test.
Balanced / unbalanced design: Under the same conditions, balanced design has high test efficiency, that is, the minimum sample size is required.

Note: When estimating the sample size, it is necessary to combine a variety of estimation methods and choose the most conservative method; set a more reasonable rate of loss to follow-up (<20%) according to the test plan; the determination of the effect size also needs to be close to the actual value.