Contribution of Methodologies Adapted to Clinical Trials Focusing

on High Risk Medical Devices

C. Vidal

, R. Beuscart

and T. Chevallier

INSERM-CIC 1431, Besançon University Hospital, Besançon, France

INSERM-CIC 1403, Lille University Hospital, Lille, France

IDIL, Nîmes University Hospital, Nîmes, France

Keywords: High Risk Medical Device, Evaluation, Clinical Trial, Adaptive Trial, Bayesian Statistics, Methodology.

Abstract: High risk medical devices clinical trials are complicated, expensive, time-consuming and need an improved

clinical evaluation with better scientific evidence throughout the European Union. The purpose of this study

is to identify methodologies whose use could facilitate the evaluation of the medical device. Adaptive methods

and Bayesian approaches are expert tools that can accelerate access to innovation providing more flexibility

but they are insufficiently used because of a lack of expertise and training in the trial community (clinicians,

statisticians and regulation authorities). Involving stakeholders (regulation authorities, industrial, clinicians,

biostatisticians, end-users) early in the conceptualization of the adaptive design improve adoption,

implementation, feasibility and overall quality of that trial.

1 INTRODUCTION

The clinical evaluation of a new medical device is an

essential stage in the industrialists’ pathway towards

market access. The new European regulation (MDR

2017/745) will be fully in force in May 2020 and

requires clinical investigation particularly for high-

risk medical devices (HRMDs).

Randomized controlled trials (RCTs) have long

been recognized as the gold standard for evaluating

the effectiveness of drugs. Conducting an RCT takes

a great deal of time and financial resources, and great

rigour in trying to isolate the specific effect of the

intervention under study. Compared with drugs,

HRMDs have specific features such as long-term use

and unknown interactions with the human body, the

means of explanting and replacing implantable

devices, the user's skills, the human-machine

interface, the management of data-flow generated,

etc. These specificities require specific evaluation

methods to generate better clinical evidence.

Adaptive methodologies have been developed as an

alternative to the traditional RCT design.

Even though the legislation, particularly

American legislation with the Food and Drug

Administration (FDA), qualifies adaptive

methodologies as “modern” and “new” methods, a

large number of these concepts are old and have

remained unused for many years faced with the

hegemony of RCTs.

The use of adaptive methods in designing clinical

studies has become a major challenge to the

evaluation of the safety and efficacy of a medical

device, faced with the specificities of the field and the

significant financial and temporal restrictions of this

industry composed mainly of start-ups and SMEs. To

do this it is necessary to find methods that take the

specificities of the medical device into account.

Several types of clinical studies may be carried out

according to the different phases of the device’s

development.

The clinical phase is generally split into two

stages, a first stage of collecting information about

safety and performances of the device. This

information is collected during feasibility studies or

clarifications (implantation technique, patient

characteristics, judgement criteria) and a second stage

to evaluate the device’s clinical efficacy in pivotal

evaluation studies to demonstrate the risk-benefit

ratio.

This work consists of reviewing methods that may

be used in the clinical evaluation of high risk medical

devices.

Vidal, C., Beuscart, R. and Chevallier, T.

Contribution of Methodologies Adapted to Clinical Trials Focusing on High Risk Medical Devices.

DOI: 10.5220/0009374503370343

In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2020) - Volume 1: BIODEVICES, pages 337-343

ISBN: 978-989-758-398-8; ISSN: 2184-4305

337

2 CLINICAL INVESTIGATION

Rather than “clinical trial”, the term “clinical

investigation” is generally used in Europe in

reference to research on medical devices. The

expression “clinical investigation” is thus defined in

the ISO 14155 norm, “Clinical investigation on

medical devices for human subjects”, as being “... any

study systematically designed and planned for use on

human subjects, undertaken to check the safety and /

or performance of a specific device.” The term

“clinical investigation” is defined in a slightly broader

way in the American regulations (42 USCS § 1320a-

7h (e)) as being “any experiment involving one or

several human subjects, or products arising from the

human body, and in which a drug or medical device

is administered, dispensed or used.”

Clinical investigations are subject to scientific and

ethical examination. The protocol for clinical

investigation includes justification, objectives,

design, methodologies, control, how to conduct the

clinical investigation and the documentation relative

to results and the analysis method concerning it. The

level of evidence of a study is characterized by its

capacity to answer the question being asked. The

randomised controlled trial is the experimental plan

that offers the highest level of evidence to

demonstrate the efficacy of a device relative to a gold-

standard therapy. However, certain specificities of

medical devices make this type of trial difficult to

perform.

The main limits of resorting to a randomised

controlled trial for medical devices are the

impossibility to randomise patients, the device’s short

life-cycle, the small size of the target population, the

difficulty of double-blinding, the low acceptability of

patients and practitioners, the choice of comparator

and the operator-dependent nature of the medical

device.

Besides, classical trials are often long, which is

incompatible with the evaluation of the medical

device whose life-cycle is short and this can hinder

access to innovation. When the trial is non-

conclusive, this leads to the inclusion of lots of

patients in a pointless trial with inefficient treatment.

When the trial is conclusive with a very effective

device being tested, this poses the problem of patients

in the comparator group not having the chance of

access to progress and delayed access to progress.

For medical devices designed to compensate for

handicap, there is a potential loss of quality of life for

the patients who might be able to benefit from them.

Clinical trials may be expensive and this deters

certain small and medium-sized medical device

companies which, in turn, delays or prevents access

to new technologies and medical progress for patients

and users. It is therefore essential to find new methods

of clinical investigation centred around all these

issues.

3 ADAPTIVE METHODS

3.1 Guides

The first part of this work consisted of gathering all

the available guides in the field of clinical evaluation

of medical devices, and publications on that theme.

The following works were used:

 Methodological choices for the clinical

development of medical devices; HAS evaluation

report dated October 4

, 2013.

 Methodological specificities of the clinical

evaluation of a connected medical device (CMD);

HAS report on the elaboration of the guide on the

specificities of clinical evaluation, in view of its

access to reimbursement dated January 29

, 2019.

 Bernard A, Vaneau M, Fournel I, Galmiche H,

Nony P, Dubernard JM. Methodological choices

for the clinical development of medical devices.

Med Devices (Auckl). 2014 Sep 23;7:325-34.

 Guideline on clinical Trials in small populations;

Committee for medicinal products for human use

on 27 july 2006.

 Guidance for the use of Bayesian Statistics in

Medical Device Clinical Trials; Guidance for

industry and FDA Staff on February 5, 2010.

 Adaptive Designs for Medical Device Clinical

Studies, Guidance for Industry and Food and Drug

Administration Staff, Document issued on July

27, 2016.

3.2 Improve Acceptability by Doctors

and Take into Account the

Operator-dependent Nature

When one arm in the study is less attractive than the

other, studies may be carried out according to a Zelen

plan or according to a complete cohort pattern. These

types of trials introduce flexibility in the attribution

of treatments and allow better acceptability of the

randomisation by the patients and also give us the

possibility of adjusting the results to the

randomisation.

Zelen Plan (Zelen et al., 1983, Zelen et al. 1990):

The patient’s consent is only requested for the new

treatment and not for the gold-standard treatment

ClinMed 2020 - Special Session on Designing Future Health Innovations as Needed

338

(simple consent). It is also possible to ask the patient

randomised to the experimental group what treatment

he/she wants to receive and to give him/her that

treatment, or even in each arm of the randomization,

ask which treatment the patient would like and to give

the patient the treatment he/she wants to have (double

consent).

The patients are analysed in the groups to which

they were initially randomised and not in the arms of

the treatment being received. This plan is only valid

if there is not too great an imbalance between groups;

that is to say, few patients leaving the study (if these

are not related to the treatment) and if the changes of

are not very frequent (fewer than 10% of patients

changing arms).

This type of pattern might be useful in the high

risk medical device area particularly when the target

population is small and you think the recruitment is

going to be very difficult as in the case of studies

focusing on an implantable device (implanted for a

more or less long duration, possible withdrawal /

difficult withdrawal / very difficult withdrawal /

impossible withdrawal) or an invasive surgical

technique with a less invasive or less restrictive

reference arm (with a drug alternative for example).

Comprehensive Cohort Study (Kearney et al.,

2011, Torgerson et al., 1998): the pattern consists of

randomising all patients eligible for research and, at a

second stage, given the patients who refuse

randomisation the treatment they refer. In

methodological terms the main pitfall concerns the

absence of group comparability. However, it is

possible to adjust the results on the randomisation.

3.3 Improve Acceptability of Doctors

and Take into Account the

Operator-dependent Nature

When certain centres only use one of the two

techniques under study and do not know the other

technique or only master the one technique and the

result is operator-dependent, it is possible to use a

trial based on expertise or a cluster trial (or a Stepped

Wedge Cluster trial) to increase the participation of

doctors and the reliability of the evaluation.

Trial based on Expertise (Devereaux et al.,

2005

): in this case the patients are randomised to the

doctor or team that masters the intervention or

technique (for example, prosthetic hip implant

surgery). The doctor only performs the procedure he

fully masters. In this case, the doctor is device user

and he is directly involved in evaluating this. For each

study arm, the doctors master the technique that they

are going to use and have reached the technical

plateau, which avoids any imbalance between the two

groups of the trial during the evaluation and is also

more ethical. This type of trial is still little used. It is

very pertinent when the techniques are different and

complex.

Cluster Trials and SWCs (stepped wedge

cluster) trials (Barker et al., 2016): With this type of

trial, groups of individuals are randomised (hospitals,

services, care units, doctors) and not individuals.

SWC trials are suitable when you want to gradually

implement a new strategy or a new technique without

going back to the previous one.

Centres start with the gold-standard technique and

the time when each centre switches over to the new

technique is randomised. The group experimenting

the new technique can be compared both to itself

based on the initial measures performed on that group

and with the measures from the other patients who are

using the gold-standard technique (independent,

homogeneous control group).

This type of design may be useful when

evaluating a new device, a new technology which is

to be gradually introduced (for example a new device

which is too expensive to use over several centres in

the same area) but the number of clusters must be

sufficient to ensure sufficient statistical power and the

participation of centres/ services/ doctors must be

good especially as these trials may be long

(monitoring of inclusions and motivational strategy to

be established on the scale of the cluster).

3.4 Compensate for a Small Target

Population

When the target population is small, it is important to

optimise and maximise the information collected on

the patients in the study. In some cases, it is possible

to test several strategies on the same patients.

Cross-over Trial (Fuehner et al., 2016,

Haddad et

al., 2010): In this type of trial, each patient receives

two study treatments (or more according to a factorial

design). A weaning period is provided for after the

patient has been given the first treatment. It is the

order of administering the sequences of treatment that

is randomised. This type of design is suitable for

stable pathologies and when judgement criteria can be

read independently over the two periods. The interest

of this type of design is divide by at least two the

numbers planned for the trial and therefore reduce the

duration of the trial. This type of trial may be

proposed in cases of evaluating high risk devices

whose installation and use are not operator-dependent

or if the technical plateau has been reached for all the

investigators before setting up the trial.

Contribution of Methodologies Adapted to Clinical Trials Focusing on High Risk Medical Devices

339

SnSMART Trial (Small n Sequential Multiple

Assignment Randomized Trial) (Tamura et al., 2016,

Wei et al., 2018, Meurer et al., 2017): This type of

trial can be used when a patient is likely to receive

several therapeutic sequences until he/he achieves the

treatment aim (complete recovery, remission, etc).

The sequences are predetermined beforehand and at

the end of each sequence the randomisation is adapted

to orient patients either to pursue their ongoing

treatment if the response is favourable or to use one

of the alternatives being tested in the event of non-

response. The number of arms being tested may be

adapted, if an arm turns out to be ineffective, it can be

removed. These trials potentiate the numbers and may

be used in cases of pathologies focusing on small

target populations (SnSMART). This type of trial is

interesting because it uses the information from the

different sequences to compare therapeutic strategies

and leads to the inclusion of fewer patients.

3.5 Introduce Flexibility to Take

Technological Evolution into

Account and Accelerate and

Optimise Clinical Development

In order to take technological evolution into account

and accelerate clinical development and product

launching whilst allowing early terminations

(futility/efficacy) or protocol adjustments

(evolution/suppression of an arm), it is possible to use

tracker design trials, sequential trials, MAMS trials

and adaptive trials (detailed further on).

These trials rely on planned intermediate analyses

which allow the investigator to glean information

which is useful for adapting the development

strategy. They are particularly interesting in the

context of clinical evaluating medical devices.

Tracker Trial Design (Lilford, et al., 2000): This

type of trial was proposed to evaluate new

technologies. The principle consists of following the

evolution of the technology in the trials based on

flexible protocols without a duration of numbers fixed

beforehand and based on information obtained during

intermediate analyses. It is therefore possible to

interrupt a trial early on if the technique is efficient,

detect poor performances and guide new

developments.

Sequential Trials (Hamilton et al., 2012): The

principle of sequential trials consists of planning

intermediate analyses in order to be able to conclude

early on. The conclusion focuses either on the very

high efficacy/tolerance of the experimental arm

compared with the control arm if the results observed

on the first patients are very promising, that is to say,

beyond what was initially expected, or on its

inefficacity (futility) if the results observed are below

what was initially expected. With this type of trial, it

is possible to quickly conclude on the main criterion.

Multi-Arm Multi-Stage trials (MAMS) (Simon

et al., 1985): MAMS trials are used in the context of

a medical device’s accelerated development plan. In

fact in this type of trial, sequential trials are gathered

into one single protocol (Redman et al., 2015) (e.g.:

several competitive devices with one control arm).

The control group is not obligatory but it is

recommended. The attribution of patients to each arm

is randomised. The arms which do not fulfil the

conditions for minimum efficacy (futility) during the

intermediate analyses are removed and only the most

efficient arms are retained. The first phase is not

directly comparative, the second phase gives us the

probability of selecting the best treatment compared

with the others and the control arm is used to

“estimate” the size of the effect.

4 ADAPTIVE TRIALS

4.1 Principle

With adaptive trials it is possible to modify items in

the protocol during the study, based on data collected

during the planned intermediate analyses without

compromising the integrity and the validity of the

study.

With adaptive methods it is also possible to

strengthen the clinical evaluation of medical devices

by authorising the analysis of lots of evaluation

criteria, carrying out several intermediate analyses,

early terminations in the event of inefficacity,

allocating patients to the most promising arms, re-

evaluating the sample-size and, more especially,

redefining the target population.

These methods also make it possible to combine

the early exploratory phases with the demonstrative

phases which may make it possible to accelerate and

optimise the development and implementation of

innovative devices.

It is also possible with these methods to optimise

the feasibility phases and confirmatory phases by,

proposing much broader, adaptive feasibility trials

leading to better-sized pivot trials or by proposing

adaptive confirmatory trials testing several

hypotheses as required, which would reduce the

number of feasibility studies throughout the course of

the product’s development.

Group sequential design and adaptive sample-size

adjustment were used frequently to make study

ClinMed 2020 - Special Session on Designing Future Health Innovations as Needed

340

durations shorter and include a smaller number of

subjects.

These methods may therefore make it possible to

reduce the requirements in terms of resources, time

necessary to finish the studies and increase the

chances of the study’s success.

There are several possible types of adaptation.

4.2 Response-adaptive Randomization

Trials

The aim of this type of pattern is to treat a maximum

number of patients with the best treatment under trial

and to minimize the number of subjects necessary in

the trial by introducing the possibility of an

anticipated stop.

It may also be used in trials with several arms. It

involves first randomising the patients with a

balanced ratio then, gradually and throughout the

trial, based on information gathered during the

intermediate analyses it is possible to modify the

affectation ratio in order to orient more patients

towards the most effective treatment (Jiang, F, et al.,

2013).

This type of design is an alternative to the multi-

stage multi-arm (MAMS) trials seen above (Wason et

al., 2014, Wathen et al., 2017).

4.3 Sample Size Reassessment Trial

At the time of the intermediate analyses it is possible

to re-evaluate the number subjects necessary for the

rest of the trial if the effect observed seems less than

what was expected at the beginning of the trial

(Magirr et al., 2016).

A misspecification of the expected treatment

effect may result in an underpowered or overpowered

trial. In the flexible framework, the remainder of a

design can be modified at an interim analysis.

In an adaptive trial it is therefore possible to

recalculate the number of participants and to increase

the power of the trial based on new hypotheses

without compromising the validity of the study.

4.4 Seamless Trials

These are trials for which the feasibility and pivot

phases follow on from each other in the same trial

(Thall, 2008). The two phases are based on

complementary criteria (for example: survival

without progression and overall survival).

Certain arms can be removed due to inefficacity

and only the most powerful arms are pursued in the

pivot study.

One control group may be included at the

beginning of the pivot phase or before it.

4.5 Adaptive Enrichment

These are trials for which we observe, during an

intermediate analysis, a better response to treatment

in one of the sub-groups of patients (Simon et al.,

2013, Lai TL et al., 2019).

The underlying idea is therefore to study the effect

of the treatment in the sub-group whose size is not

suitable for analysis beforehand. The eligibility

criteria for the trial are modified and the sample-size

is recalculated so that the size of the sub-group is

sufficient in each arm.

5 BAYESIAN METHODS

5.1 Principle

Bayesian approaches may be used to implement and

analyse clinical trials. They are used because they

give the possibility of combining information

obtained before the trial “prior information”

(previous studies, expert opinion, literature…) and

information obtained during the trial “current

information” to formulate or reformulate a rule for

decision-making.

In a Bayesian clinical trial, any uncertainty about

a parameter is described according to probabilities,

which are then updated during data-collection for the

trial. The probabilities are set beforehand based on

previous data and the probabilities are estimated a

posteriori from the data obtained during the trial.

There are no statistical tests but the probability of the

treatment under experimentation being effective has

a 95% credibility interval. However, it is very

important that the a priori information used does not

influence the final result too much (sensitivity

analysis required). The quality of information

supplied a priori is therefore a key element in the

credibility of results.

5.2 Bayesian Medical Device Trials

Bayesian methods has been supported by the US

Food and Drug Administration’s (FDA) Center for

Devices and Radiological Health for medical

device clinical trials and are used in trials on

medical devices (Pennello et al., 2008, Campbell et

al., 2011, Campbell et al., 2016).

Pennello et al. 2008, explain how these analyses

are particularly suitable in this case: “Device trials

Contribution of Methodologies Adapted to Clinical Trials Focusing on High Risk Medical Devices

341

can be particularly suitable for Bayesian analysis. For

example, if a therapeutic device has evolved in

relatively small increments from previous

generations of the same type of device, then prior

information from the trials of the previous devices

can be predictive of the safety and effectiveness

profile of the new device (Allocco et al., 2010). The

reason the previous trials can be predictive is that the

mechanism of action of a therapeutic device is often

physical, implying a local effect that is often

predictable. In contrast, the mechanism of action of

pharmaceuticals is pharmacokinetic, implying

systemic effects from similar but not identical

formulations, which are often unpredictable. Other

potentially reliable sources of prior information for

device trials include clinical trials of the device

conducted overseas, patient registries, pilot studies,

studies of the device on similar patient populations,

and perhaps nonclinical studies. Historical controls

can also represent prior information for the control

arm of a randomized controlled trial”.

The information collected beforehand is generally

based on previous studies on the same device or on a

similar device ideally under similar conditions of use

(same technique used, training of similar doctors with

the same experience), on the same target population

with the same type of management; it comes

especially from designers (engineers), users

(clinicians, patients) and the academic world

(experts).

They are a more flexible alternative to classical

methods (frequentist approach). They are used to

adapt the randomisation according to the responses

observed (see Bayesian adaptive randomisation).

These methods also make it possible to compare

several sub-groups of patients, several criteria,

several time sequences because multiplicity is

managed better Bayesian statistics. It is also possible

to take missing data into account and to predict an

event depending on what has been observed in other

patients throughout the trial. The underlying

hypothesis is that the patients of a same centre, a same

trial or a same group of trials focused on the same

device or on a similar device are interchangeable.

Meta-analyses also use Bayesian methods to take into

account the heterogeneity between trials and between

groups of trials (for example several versions of the

same device).

6 DISCUSSION

In this review, we noted that there have been many

adaptive methods for decades, but their use is recent

and mainly in the pharmacological area. Adaptive

methodologies have most often been used in

oncology.

Adaptive methods may respond to the

specificities of clinical investigations on high risk

medical devices. Nevertheless, so far they have been

very little used in that area (Ribouleau et al., 2011)

even though a few published examples can be found

in the literature. This observation may also come from

a more general situation concerning medical devices

which most of the time are released without having

undergone a proper clinical investigation. And even

though since 1993 the European ruling has mentioned

the obligation for each new medical device, whatever

its risk category, to undergo a clinical evaluation to

obtain CE marking, few clinical studies are indexed

before they obtain the CE mark in Europe.

These “new” methods have encountered many

suspicions, and the regulatory authorities in charge

evoke methodological failings or data-collection

problems specific to adaptive designs, which delay

the process of product approval. The FDA and the

EMA have had mixed experiences with adaptive

designs (Collignon et al. 2018, Elsäßer et al., 2014).

Experiences have shown that applicants need to

meet early and often with regulators. Adaptive design

and Bayesian clinical trials need to be prospectively

designed and require extensive pre-planning and

model-building from the prior information to

mathematical modeling.

Involving regulation authorities early in the

conceptualization of the adaptive design improve

adoption and implementation of that trial.

Adaptive design and Bayesian clinical trials

require highly technically trained statisticians and

programmers. A particular pedagogical attention

should therefore be paid to accustom all the

stakeholders, and particularly the scientists in charge

of regulation before and during these trials, to these

new uses of new methodologies.

7 CONCLUSION

Overcoming methodological difficulties in

conducting clinical trials is a major challenge.

Barriers encountered in the field of medical devices

lead stakeholders to use new methodologies.

Adaptive methods could be used and has been the

subject of several recent reviews (Bothwell et al.,

2018).

Besides, various studies explored specific aspects

of adaptive trials (Guetterman et al., 2017), including

attitudes and opinions regarding confirmatory

ClinMed 2020 - Special Session on Designing Future Health Innovations as Needed

342

adaptive clinical trials and obstacles to using them

(Meurer et al., 2016, Guetterman et al., 2015).

ACKNOWLEDGMENTS

We wish to thank Teresa Sawyers, Medical Writer at

the B.E.S.P.I.M. for translating this work into

English.

REFERENCES

Allocco,DJ, et al., 2010. A prospective evaluation of the

safety and efficacy of the TAXUS Element paclitaxel-

eluting coronary stent system for the treatment of de

novo coronary artery lesions: Design and statistical

methods of the PERSEUS clinical program, Trials 11:1.

Barker, D, et al., 2016. Stepped wedge cluster randomised

trials: a review of the statistical methodology used and

available. BMC Med Res Methodol 6 ;16:6..

Bothwell, LE, et al., 2018. Adaptive design clinical trials: a

review of the literature and ClinicalTrials.gov. BMJ

Open. 10 ; 8(2).

Campbell, G, 2011. Bayesian statistics in medical devices:

innovation sparked by the FDA. J Biopharm Stat.

21:871-87.

Campbell, G, et al., 2016. Statistical innovations in the

medical device world sparked by the FDA. J Biopharm

Stat. 26 (1):3-16.

Collignon O, et al. 2018. Adaptive designs in clinical trials:

from scientific advice to marketing authorisation to the

European Medicine Agency. Trials 19 (1):642.

Devereaux, PJ, et al., 2005. Need for expertise based

randomised controlled trials. BMJ 330 (7482): 88.

Elsäßer A, et al., 2014. Adaptive clinical trial designs for

European marketing authorization: a survey of

scientific advice letters from the European Medicines

Agency. Trials ; 15:383.

Fuehner, T, et al., 2016. Topical nasal anesthesia in flexible

bronchoscopy- A cross-over comparison between two

devices. PLoS One. 15;11(3).

Guetterman, TC, et al., 2015. Reflections on the Adaptive

Designs Accelerating Promising Trials Into Treatments

(ADAPT-IT) Process-Findings from a Qualitative

Study. Clin Res Regul Aff. 32(4):121-130.

Guetterman, TC, et al. 2017. The life cycles of six multi-

center adaptive clinical trials focused on neurological

emergencies developed for the Advancing Regulatory

Science initiative of the National Institutes of Health

and US Food and Drug Administration: Case studies

from the Adaptive Designs Accelerating Promising

Treatments Into Trials Project. SAGE Open Med..

Haddad, M, et al., 2010. Sacral neuromodulation in children

with urinary and fecal incontinence: a multicenter, open

label, randomized, crossover study. J Urol 184 (2): 696-

701.

Hamilton, C,et al., 2012. Sequential design for clinical trials

evaluating a prosthetic heart valve. Ann Thorac Surg,

93(4): 1162-1166.

Jiang, F, et al., 2013. A Bayesian decision-theoretic

sequential response-adaptive randomization design.

Stat Med, 32 (12):1975-94.

Kearney, RS, et al., 2011. The comprehensive cohort model

in a pilot trial in orthopaedic trauma. BMC Med Res

Methodol ;11:39.

Lai, TL, et al., 2019. Adaptive enrichment designs for

confirmatory trials. Stat Med 38 (4) :613-624.

Lilford, R, et al., 2000. Trials and fast changing

technologies: the case for tracker studies. BMJ, 320

(7226): 43–46.

Magirr, D, et al., 2016. Sample Size Reassessment and

Hypothesis Testing in Adaptive Survival Trials. PLoS

One. 11(2):e0146465.

Meurer, WJ, et al., 2016. Attitudes and opinions regarding

confirmatory adaptive clinical trials: a mixed methods

analysis from the Adaptive Designs Accelerating

Promising Trials into Treatments (ADAPT-IT) project.

Trials. 29;17:373.

Meurer, WJ, et al., 2017. Multiple assignment randomized

Trials: An opportunity for improved design of stroke

reperfusion trials. J Stroke Cerebrovasc Dis. 26

(4):717-724.

Pennello, G, et al., 2008. Experience with reviewing

Bayesian medical device trials. J Biopharm Stat 18 :81-

115.

Redman, MW,.et al., 2015. The Master Protocol Concept.

Semin Oncol 42 (5) :724-30.

Ribouleau, L, et al., 2011. Bayesian statistical method was

underused despite its advantages in the assessment of

implantable medical devices. J Clin Epidemiol.

64:270-9.

Simon, R,. et al., 1985. Randomized phase II clinical trials.

Cancer Treat Rep 69 (12 ):1375-81.

Simon, N, et al., 2013. Adaptive enrichment designs for

clinical trials. Biostatistics 14(4):613-25.

Tamura, RN, et al., 2016. A small n sequential multiple

assignment randomized trial design for use in rare

disease research. Contemp Clin Trials. ; 46:48-51.

Thall, PF, 2008. A review of phase 2-3 clinical trial designs,

Lifetime Data Anal 14(1):37-53.

Torgerson, DJ, et al., 1998. Understanding controlled trials.

What is a patient preference trial? BMJ; 316:360.

Wason JM, et al., 2014. A comparison of Bayesian adaptive

randomization and multi-stage designs for multi-arm

clinical trials. Stat Med. 33 (13):2206-21.

Wathen JK et al., 2017. A simulation study of outcome

adaptive randomization in multi-arm clinical trials. Clin

Trials 14 (5):432-440.

Wei, B, et al., 2018. A Bayesian analysis of small n

sequential multiple assignment randomized trials

(snSMARTs). Stat Med 20 ; 37 (26) :3723-3732.

Zelen, M, et al., 1983. Innovations in the design of clinical

trials in breast cancer. Breast Cancer Res Treat 3 (2):

137-42.

Zelen, M, et al., 1990. Randomized consent designs for

clinical trials: an update. Stat Med 9 (6): 645-56.

Contribution of Methodologies Adapted to Clinical Trials Focusing on High Risk Medical Devices

343