Clinical trials of medical devices refer to the process of
confirming the safety and effectiveness of medical devices to be
registered under normal use conditions in clinical trial institutions
with corresponding conditions. Clinical trials take the subjects
(samples) as observation objects to observe the effect of trial device
on human body or its evaluation ability on the disease and health
status of human body in normal use conditions, so as to infer the
effect of trail device on the population (overall) with intended use.
Due to the inherent characteristics of medical devices, their trial
design has its own characteristics.
This guideline applies to the medical devices with established
product composition, design and performance, including therapeutic
products and diagnostic products, while excluding in vitro diagnostic
reagents.
This guideline is a technical guidance document for applicants
and reviewers, not involving the registration approval and other
administrative matters, nor as law enforcement. If any other methods
can meet the regulations, they also can be used, but detailed research
data and validation shall be provided. This guideline shall be used in
—— 2 —
accordance with the relevant regulations.
I. Purpose of medical device clinical trial
Clinical trials need to set the defined and specific trial purpose.
The applicant can comprehensively analyze the characteristics of
trial device, non-clinical research situation, and the clinical data of
the same kind of product which has been marketed in China
(hereinafter referred to as marketed product) and other factors, in
order to set the purpose of clinical trials. The purpose of clinical
trials determines the design elements of clinical trials, including
main evaluation indicators, test design types, comparison types of
control trials and etc., thus affecting the sample size of clinical trials.
The examples for the purpose of clinical trials under different
conditions are shown as follows:
(i) When the safety and effectiveness of trial device are
confirmed through clinical trials under its intended use, if more
attention is paid to whether the curative effect of trial device can
meet the needs of clinical use, the purpose of clinical trial can be set
to confirm whether the effectiveness of trial device is superior/
equivalent/not inferior to similar marketed products, and the safety
of trial device. Meanwhile, the main evaluation index of clinical
trials is the effectiveness index.
CFDA – No 6 – 2018-01-04 – Medical Device Clinical Trial Design Guideline
— 3 ——
(ii) When the safety and effectiveness of trial device are
confirmed through clinical trials under its intended use, if more
attention is paid to whether the safety of trial device can meet the
need of clinical use, the purpose of clinical trials can be set to
confirm the safety of trial device is superior/ equivalent/not inferior
to similar marketed products, and the effectiveness of trial device.
Meanwhile, the main evaluation index of clinical trials is safety
index. Taking breast implants as example, clinical trials usually
choose complication incidence rate (e.g., capsular contracture rate,
implant rupture rate) as the main evaluation index.
(iii) In the case of adding indications for marketed products,
the purpose of clinical trials may be set to confirm the safety and
effectiveness of trial device for new indications. For example,
hemostatic products add the indications for use in ophthalmology,
neurosurgery and urology on the basis of approved applications
(such as general surgery, obstetrics and gynecology).
(iv) When the applicable population of marketed product
changes, the purpose of clinical trials can be set to confirm the safety
and effectiveness of trial device for the new applicable population.
For example, the membrane oxygenator product increases the
applicable population with the weight ≤ 10kg on the basis of the
—— 4 —
original approved scope of application; therapeutic ventilators add
the application scope for children on the basis of approved
applicable adults.
(v) In case of major design changes in marketed device, the
trial purpose may be set according to the scope of the changes
involved. For example, when the pattern design of coronary
drug-eluting stent platform changes, the purpose of clinical trials can
be set to confirm the influence of the changes on product safety and
effectiveness.
(vi) When the use environment or use method of marketed
device has changed significantly, the trial purpose may be set to
confirm the safety and effectiveness of the product in a specific use
environment and use method. For example, the marketed
implantable cardiac pacemaker usually cannot be compatible with
MRI. If the compatible MRI is applied, the purpose of clinical trial
can be set to confirm the safety and effectiveness related to any
compatible MRI.
II. Basic types and characteristics of clinical trial design
(i) Parallel control design
Randomized, double-blind and parallel-control clinical trial
design can enable the influencing factors of clinical trials distributed
CFDA – No 6 – 2018-01-04 – Medical Device Clinical Trial Design Guideline
— 5 ——
towards equilibrium between trial group and control group, ensure
that investigators, evaluators and subjects have not known grouping
information, and avoid the selection bias and evaluation bias, which
is considered able to provide high-level scientific evidence and can
be generally preferred. For some medical devices, the feasibility of
this design is challenged by the inherent characteristics of the
device.
- Randomization
Randomization is a basic principle of clinical trials for parallel
control, pair design and cross design, referring to each subject in the
clinical trials assigned to trial group or control group with equal
opportunity (e.g., the case number of trial group and control group is
1:1) or other prescribed probability (e.g., the case number of trial
group and control group is n:1), not affected by the investigators
and/or subject subjective intention. Randomization is designed to
ensure the comparability between subjects in trial group and control
group on a variety of known and unknown baseline variables that may
affect the trial result.
Non-random design may result in uneven distribution of
various influencing factors among groups and reduce the credibility
of trial results. On the one hand, covariate analysis may be difficult
—— 6 —
to completely correct the effects of known factors on the results. On
the other hand, the influence of unknown factors on trial results is
also difficult to evaluate. Therefore, non-random design is not
usually recommended. If the applicant has good reason to believe
that a non-random design is necessary, it shall be to elaborate the
reasons why the design must be adopted and specific measures to
control selection bias. - Blind method
If the grouping information is known, the investigators may pay
selective attention to trial group in the use process of device, the
evaluators may have tendency during curative effect and safety
evaluation, and the participants may be influenced by subjective
factors. Blind method is one of the important measures to control
bias caused by “knowing grouping information” in clinical trials.
The purpose is to achieve all personnel unknown the grouping
information in clinical trials. According to the blinding degree, the
blind method can be divided into complete blinding, incomplete
blinding and non-blinding. In the complete blinding clinical trial,
subjects, investigators and evaluators are blinded to the grouping
information.
CFDA – No 6 – 2018-01-04 – Medical Device Clinical Trial Design Guideline
— 7 ——
In many cases, complete blinding is not feasible, based on the
inherent characteristics of the device and the corresponding
treatment modalities. When the trial device and control device are
obviously not the same, blinding is difficult to investigators. For
example, the appearances of trial product and control product for
knee prosthesis could be significantly different, and the
manufacturer laser tag is visible on implant; for the intravascular
metal stent, the trial product and control product are different in
specific structure and pattern. At this moment, blinding is suggested
to subjects to the utmost, i.e., subjects have not known assigned to
trial group or control group, and the third-party blinding evaluation
(such as center radiology, center laboratory and evaluation
committee) and blind-state data audit are used. When trial device
and control device are obviously different in the form and the main
evaluation indexes are sourced from imaging data, blinding is
difficult to investigators and evaluators. Taking biological
absorbable stent as example, when the control product is metal stent,
due to the degradation of biological absorbable stent platform,
blinding is difficult to evaluators when evaluating the late lumen
loss index (this index is evaluated by imaging method). At this
moment, blinding is recommended to subjects to the utmost, and
—— 8 —
blinding data audit is used. The aforesaid situations, in which
blinding is not set to investigators, or to investigators and evaluators
due to the inherent characteristics of device, are incomplete blinding
clinical trial designs.
When the therapeutic method (including device) of trial group
is obviously different from control group, blinding is difficult to the
subjects, investigators and evaluators, so non-blind trial design can
only be adopted, such as the case of comparing interventional
therapy with surgery as well as device therapy with drug therapy. In
order to minimize bias, the following methods can be considered. (1)
Before the screening and enrollment of subjects, neither the subjects
nor the investigators are aware of grouping information (i.e.,
allocation hiding); (2) the subject is unable to know the grouping
information before the completion of treatment under the premise of
ethical permission; (3) blind data audit is used.
The applicant needs to discuss the reason for adopting the
incomplete blinding or non-blind trial design, and elaborate the
concrete measures of bias control (e.g., the available
objectively-judged indexes can be used to avoid bias, and the
standard operating practices are adopted to minimize
implementation bias, etc.).
