Clinical Science for Medical Devices:
A Guide for Entrepreneurs
20 October 2017
Jim Gustafson
Former Senior Vice President
Clinical Science and Regulatory Affairs
MEDRAD Interventional / Possis
Outline for Today
1. Purpose and Planning
2. Elements of Trial Design
3. Add in the Science!
4. Patient Rights
5. Trial Execution
6. Finishing Well
7. Clinical Science Wisdom
8. Q&A
35+ Years Doing Medical Device Clinical Science
• Class III cardiovascular and ophthalmic implants and therapies
• Always on the Manufacturer’s Side.
• Career path: Clinical Monitor → Sr. Clinical Scientist → Manager →
Director → VP → Sr VP, Clinical and Regulatory Affairs.
• ~ 10 IDE trials, each with 100’s of patients enrolled at dozens of sites
• 1 original PMA approval, based on 6 protocols enrolling 1,100 pts.
• 4 510(k)s involving IDE clinical trials (each > 100 pts and > 10 sites).
• ~ 10 PMS studies (each > 200 pts and > 10 sites).
• ~ 7 large marketing clinical studies.
• Many negotiations with FDA on study design, outcomes, and proofs.
• 3 live presentations to FDA Advisory Panels.
• Watched a bad clinical trial almost kill my company, then led the
successful effort to revive it with new and better clinical trials.
What Won’t Happen Today:
• Tell you if your medical device needs
a clinical study.
• Design a clinical study for you, or
give you a template for doing so.
• Explain everything about clinical
studies for medical devices.
• Make you an expert in clinical
studies for medical devices.
What I Hope Will Happen:
• Appreciate the importance and value
of clinical studies.
• Respect their complexity, yet know
they can be successfully managed.
• Learn the basic elements of clinical
science for medical devices.
• Know what questions to ask to be a
successful medical device
entrepreneur when clinical science is
involved.
1. PURPOSE AND PLANNING
Why Do Medical Device
Clinical Science?
To prove that the device
• Is safe.
• Is effective in a clinically meaningful way.
• Has desirable health economic characteristics.
• Has other desirable features: convenient, easy to use,
compatible with other equipment, devices, and treatments in
the use environment, etc.
What Do We Do with This Proof
• Get regulatory approval.
• Obtain favorable 3rd-party reimbursement.
• Show advantages compared to other treatment options.
• Change health care decisions.
• Other?
Types of Medical Device Clinical Science
• First-in-Human (FIH) to confirm the
device works in humans.
• Feasibility study to confirm target
patients and endpoints.
• Pivotal trial to prove safety and
effectiveness.
• Post-marketing surveillance study for
long-term effects and rare events.
• Marketing study: preference, pricing,
use characteristics, etc.
2. ELEMENTS OF CLINICAL TRIAL DESIGN
Clinical Trials are Always About 3 Things
Patients
Treatment
Outcomes
1. Patients
• Which patients will most
likely benefit?
• How can you describe
them objectively?
• Which patients should be
excluded?
• How can you find the
patients you want?
2. Treatment
• What treatment will be
applied?
• Can it be applied
consistently applied to all
enrolled patients?
• Can it be made the only
variable to drive
outcomes?
3. Outcomes
• Effectiveness
• Safety
• Clinical Utility (value)
• Health Economics
• Other?
Audience Participation
What three things
are clinical studies
always about?
3. BUT TO THESE
THREE, ADD IN THE
Essential Scientific Elements
1. Define Endpoints
2. Minimize Bias
3. Commit to an A Priori
Analysis Plan
Study Endpoints
Endpoint: any clinical effect potentially due to the
treatment under study. Good study endpoints are:
• Objective and quantifiable.
• Meaningful to the patient and the doctor.
Primary Endpoint : the single most important
outcome measured after treatment. Used to
• Formulate the formal hypothesis.
• Calculate the sample size.
• Interpret the outcomes.
Clinical Endpoint
Directly measures how a patient feels,
functions, or survives. Examples:
• Heart-attack free survival at 1 year.
• No repeat surgery to 6 months.
• In-hospital mortality.
• Limb loss.
Surrogate Endpoint
Measures something that robustly
predicts a clinical outcome, even if the
patient does not feel it. Examples:
• PSA for prostate cancer.
• Reduced blood cholesterol.
• A1C for diabetes.
• Imaging results, e.g. CT, MRI, angio
Bias in Clinical Science Comes in Threes
• Selection bias: selecting the wrong kind of patient, or
only a narrow subset of the right kind. Minimize by:
• Defined, objective enrollment criteria.
• Non-enrollment logs.
• Treatment randomization or other active controls.
• Treatment bias: giving different treatments to different
groups of study patients. Minimize by:
• Protocol-defined treatment that follows standard care.
• Site training and monitoring.
• Observation bias: scoring patient outcomes differently
based on treatment received. Minimize by:
• Objectively defined endpoints.
• Blinding the observer.
Types of Study Control Groups
Active Controls:
• Randomized cohort.
• Cross-over control.
• Non-randomized concurrent prospective
reference group.
Passive Controls:
• Matched case samples.
• Site history.
• Literature controls.
Parables of an A Priori
Analysis Plan
• It’s not fair to shoot your arrow, and then
paint your target around where it sticks.
• It’s also not fair to move the target while
the arrow is in flight.
• “3 out of 5?” and “4 out of 7?” aren't
acceptable in clinical science.
Statistics
Wisdom for Clinical Study Design
• Prospective beats retrospective.
• Controlled beats uncontrolled.
• Randomized is the best kind of control.
• Superiority beats non-inferiority.
• Clinical endpoints beat surrogate endpoints.
• Death is the best clinical endpoint:
• Once per patient.
• Cheap and easy to measure.
• No arguments about whether the patient is really dead.
• Bonus: In-hospital death is the hardest endpoint: if you
die in a hospital, you have to really work at it.
4. PATIENT RIGHTS
Benches have no rights.
Animals have some.
Human patients have many.
Your clinical study design must be
reasonably certain to produce
scientifically valid outcomes about your
device, but never at the expense of
patient rights, welfare, or safety.
Informed Consent
• Requirements found in 21 CFR 50.
• Equivalent to the Declaration of Helsinki (1975).
• Before study enrollment, signed by patient or
patient’s representative.
• There is a difference between consent to be treated
and consent to participate in a clinical study!
Elements for Study Informed Consent
1. The consent is for
research.
2. List foreseeable risks
and discomforts.
3. Possible benefit to
patient or others.
4. Degree of confidentiality.
5. Compensation or
treatment for injury.
6. Who to contact with
questions.
7. Participation is voluntary;
patient may quit.
8. Alternative tx available.
9. Additional costs, if any, to
patient.
10.Investigational consent is
separate from treatment
consent.
Institutional Review Board
• Requirements found in 21 CFR 58.
• Represents local standards of hospital
and community.
• Review and approve protocol, patient
consent, and any changes to them over
time.
• Increasingly, IRBs also review grants,
CRFs, etc.
• Must receive minimum yearly reports
from investigator and sponsor.
• EU equivalent is Ethics Committee.
Data Safety Monitoring Board
• Medical and statistical experts
independent of the trial who review
blinded outcomes at pre-specified
intervals.
• Look for evidence trial continuation is
not safe or is unnecessary.
• Has authority to discontinue the trial,
or change it to improve safety or
scientific soundness.
5. TRIAL EXECUTION
Executing Well
Best-laid plans are not enough.
Execution covers many areas that are not
required by regulation, where FDA does not
provide oversight. But inadequate execution
can wreck a clinical study!
• Protocol and CRF development.
• Site and investigator selection and training.
• Field monitoring.
• Standards for clinical data quality and
reporting.
• Data processing and analysis.
Data Integrity
When the study is over, all you have
to show for it is the data. Dataset
integrity is essential to a successful study!
• Source document audits.
• Completion and correction tracking forms.
• Computerized data checks.
• Double data entry.
• Database validations.
Bonus Round: Why is
Pt Enrollment So Slow?
• Study selection criteria are too narrow.
• Investigator is involved in competing trials.
• Investigator hates the control (or the device!).
• Grant is too small.
• Study coordinator has other priorities.
The surest way to cure a
disease is to start a study of it!
Myths of Study Time and Cost
Time: often patient enrollment is not the biggest calendar
time cost for a study. Consider instead:
• 6 months for site selection, approval, training, etc.
• 12 months for follow-up on the last patient enrolled.
• 6 months for dataset prep and report writing.
Cost: grants for patients enrolled is often not most of a
study’s budget. Instead, most of the costs are study-wide,
not patient dependent: database design, protocol
development, site monitoring, etc.
6. FINISHING WELL
Interpreting a Clinical Study is Still
Always About the Same 3 Things
Patients:
• What patients were actually enrolled?
• Do the results apply to a more general
population?
Treatment:
• Was the treatment received the only
significant variable?
• Did the treatment drive the outcomes?
Outcomes:
• Were the endpoints clinically meaningful?
• Are the outcomes compelling?
Interpretation
The right study design and
interpretation lets you say something
like “a multi-centered observer-blinded
randomized trial of early-presenting
STEMI patients with large-thrombus
lesions showed reduced MACE at 30
days in the treatment arm as compared
to the control arm.”
And what could be
more fun than that?
Some Key Interpretive Questions
• Is the enrolled patient sample a fair
representation of the target population?
• Are the treatment and control cohorts well-
matched at baseline?
• Are the endpoints really susceptible to the
treatment received?
• Are the treatment group outcomes really
different?
• What else besides treatment received could
affect outcome?
• Do the primary and secondary endpoint
outcomes conflict?
Have a Plan for Going Public!
Presentations:
• Medical congresses
• Late-breaking results
• State-of-the-arts
• Sponsored sessions
Publications:
• Academic or professional or business?
• Peer-reviewed or other?
• High-level, mid-level, or low-level?
• Sponsored supplements?
Actionable Interpretation
“I recommend the meatloaf.
It’s just received FDA approval.”
7. CLINICAL SCIENCE WISDOM (?)
1. Clinical studies are done by teams, not individuals:
clinical science, statistics, regulatory, medical,
device. etc.
2. Your clinical science project is just the last, longest,
most expensive, and most fraught step in your
design validation process.
3. Unlike bench testing, clinical science is done not by
sponsor employees, but by contractors (doctors
and hospital staff). They have many other priorities
higher than your study, and you can’t change that.
4. Just as you can’t inspect quality into a product, you
can’t wash the dataset from a compromised clinical
study to make it better. Clean trash is still trash.
5. More medical devices have been shown to work
than have been proven to work.
6. The surest way to cure a disease is to start a
study of a new treatment for it – suddenly all the
patients disappear!
7. Medical device clinical studies are designed by
optimists, and interpreted by pessimists.
Clinical Science for Medical Devices: A Guide for Entrepreneurs | Jim Gustafson | Lunch & Learn

Clinical Science for Medical Devices: A Guide for Entrepreneurs | Jim Gustafson | Lunch & Learn

  • 1.
    Clinical Science forMedical Devices: A Guide for Entrepreneurs 20 October 2017 Jim Gustafson Former Senior Vice President Clinical Science and Regulatory Affairs MEDRAD Interventional / Possis
  • 2.
    Outline for Today 1.Purpose and Planning 2. Elements of Trial Design 3. Add in the Science! 4. Patient Rights 5. Trial Execution 6. Finishing Well 7. Clinical Science Wisdom 8. Q&A
  • 4.
    35+ Years DoingMedical Device Clinical Science • Class III cardiovascular and ophthalmic implants and therapies • Always on the Manufacturer’s Side. • Career path: Clinical Monitor → Sr. Clinical Scientist → Manager → Director → VP → Sr VP, Clinical and Regulatory Affairs. • ~ 10 IDE trials, each with 100’s of patients enrolled at dozens of sites • 1 original PMA approval, based on 6 protocols enrolling 1,100 pts. • 4 510(k)s involving IDE clinical trials (each > 100 pts and > 10 sites). • ~ 10 PMS studies (each > 200 pts and > 10 sites). • ~ 7 large marketing clinical studies. • Many negotiations with FDA on study design, outcomes, and proofs. • 3 live presentations to FDA Advisory Panels. • Watched a bad clinical trial almost kill my company, then led the successful effort to revive it with new and better clinical trials.
  • 5.
    What Won’t HappenToday: • Tell you if your medical device needs a clinical study. • Design a clinical study for you, or give you a template for doing so. • Explain everything about clinical studies for medical devices. • Make you an expert in clinical studies for medical devices. What I Hope Will Happen: • Appreciate the importance and value of clinical studies. • Respect their complexity, yet know they can be successfully managed. • Learn the basic elements of clinical science for medical devices. • Know what questions to ask to be a successful medical device entrepreneur when clinical science is involved.
  • 6.
  • 7.
    Why Do MedicalDevice Clinical Science? To prove that the device • Is safe. • Is effective in a clinically meaningful way. • Has desirable health economic characteristics. • Has other desirable features: convenient, easy to use, compatible with other equipment, devices, and treatments in the use environment, etc.
  • 8.
    What Do WeDo with This Proof • Get regulatory approval. • Obtain favorable 3rd-party reimbursement. • Show advantages compared to other treatment options. • Change health care decisions. • Other?
  • 9.
    Types of MedicalDevice Clinical Science • First-in-Human (FIH) to confirm the device works in humans. • Feasibility study to confirm target patients and endpoints. • Pivotal trial to prove safety and effectiveness. • Post-marketing surveillance study for long-term effects and rare events. • Marketing study: preference, pricing, use characteristics, etc.
  • 10.
    2. ELEMENTS OFCLINICAL TRIAL DESIGN
  • 11.
    Clinical Trials areAlways About 3 Things Patients Treatment Outcomes
  • 12.
    1. Patients • Whichpatients will most likely benefit? • How can you describe them objectively? • Which patients should be excluded? • How can you find the patients you want?
  • 13.
    2. Treatment • Whattreatment will be applied? • Can it be applied consistently applied to all enrolled patients? • Can it be made the only variable to drive outcomes?
  • 14.
    3. Outcomes • Effectiveness •Safety • Clinical Utility (value) • Health Economics • Other?
  • 15.
    Audience Participation What threethings are clinical studies always about?
  • 16.
    3. BUT TOTHESE THREE, ADD IN THE
  • 17.
    Essential Scientific Elements 1.Define Endpoints 2. Minimize Bias 3. Commit to an A Priori Analysis Plan
  • 18.
    Study Endpoints Endpoint: anyclinical effect potentially due to the treatment under study. Good study endpoints are: • Objective and quantifiable. • Meaningful to the patient and the doctor. Primary Endpoint : the single most important outcome measured after treatment. Used to • Formulate the formal hypothesis. • Calculate the sample size. • Interpret the outcomes.
  • 19.
    Clinical Endpoint Directly measureshow a patient feels, functions, or survives. Examples: • Heart-attack free survival at 1 year. • No repeat surgery to 6 months. • In-hospital mortality. • Limb loss. Surrogate Endpoint Measures something that robustly predicts a clinical outcome, even if the patient does not feel it. Examples: • PSA for prostate cancer. • Reduced blood cholesterol. • A1C for diabetes. • Imaging results, e.g. CT, MRI, angio
  • 20.
    Bias in ClinicalScience Comes in Threes • Selection bias: selecting the wrong kind of patient, or only a narrow subset of the right kind. Minimize by: • Defined, objective enrollment criteria. • Non-enrollment logs. • Treatment randomization or other active controls. • Treatment bias: giving different treatments to different groups of study patients. Minimize by: • Protocol-defined treatment that follows standard care. • Site training and monitoring. • Observation bias: scoring patient outcomes differently based on treatment received. Minimize by: • Objectively defined endpoints. • Blinding the observer.
  • 21.
    Types of StudyControl Groups Active Controls: • Randomized cohort. • Cross-over control. • Non-randomized concurrent prospective reference group. Passive Controls: • Matched case samples. • Site history. • Literature controls.
  • 23.
    Parables of anA Priori Analysis Plan • It’s not fair to shoot your arrow, and then paint your target around where it sticks. • It’s also not fair to move the target while the arrow is in flight. • “3 out of 5?” and “4 out of 7?” aren't acceptable in clinical science.
  • 24.
  • 25.
    Wisdom for ClinicalStudy Design • Prospective beats retrospective. • Controlled beats uncontrolled. • Randomized is the best kind of control. • Superiority beats non-inferiority. • Clinical endpoints beat surrogate endpoints. • Death is the best clinical endpoint: • Once per patient. • Cheap and easy to measure. • No arguments about whether the patient is really dead. • Bonus: In-hospital death is the hardest endpoint: if you die in a hospital, you have to really work at it.
  • 26.
  • 27.
    Benches have norights. Animals have some. Human patients have many. Your clinical study design must be reasonably certain to produce scientifically valid outcomes about your device, but never at the expense of patient rights, welfare, or safety.
  • 28.
    Informed Consent • Requirementsfound in 21 CFR 50. • Equivalent to the Declaration of Helsinki (1975). • Before study enrollment, signed by patient or patient’s representative. • There is a difference between consent to be treated and consent to participate in a clinical study!
  • 29.
    Elements for StudyInformed Consent 1. The consent is for research. 2. List foreseeable risks and discomforts. 3. Possible benefit to patient or others. 4. Degree of confidentiality. 5. Compensation or treatment for injury. 6. Who to contact with questions. 7. Participation is voluntary; patient may quit. 8. Alternative tx available. 9. Additional costs, if any, to patient. 10.Investigational consent is separate from treatment consent.
  • 30.
    Institutional Review Board •Requirements found in 21 CFR 58. • Represents local standards of hospital and community. • Review and approve protocol, patient consent, and any changes to them over time. • Increasingly, IRBs also review grants, CRFs, etc. • Must receive minimum yearly reports from investigator and sponsor. • EU equivalent is Ethics Committee.
  • 31.
    Data Safety MonitoringBoard • Medical and statistical experts independent of the trial who review blinded outcomes at pre-specified intervals. • Look for evidence trial continuation is not safe or is unnecessary. • Has authority to discontinue the trial, or change it to improve safety or scientific soundness.
  • 32.
  • 33.
    Executing Well Best-laid plansare not enough. Execution covers many areas that are not required by regulation, where FDA does not provide oversight. But inadequate execution can wreck a clinical study! • Protocol and CRF development. • Site and investigator selection and training. • Field monitoring. • Standards for clinical data quality and reporting. • Data processing and analysis.
  • 34.
    Data Integrity When thestudy is over, all you have to show for it is the data. Dataset integrity is essential to a successful study! • Source document audits. • Completion and correction tracking forms. • Computerized data checks. • Double data entry. • Database validations.
  • 35.
    Bonus Round: Whyis Pt Enrollment So Slow? • Study selection criteria are too narrow. • Investigator is involved in competing trials. • Investigator hates the control (or the device!). • Grant is too small. • Study coordinator has other priorities. The surest way to cure a disease is to start a study of it!
  • 36.
    Myths of StudyTime and Cost Time: often patient enrollment is not the biggest calendar time cost for a study. Consider instead: • 6 months for site selection, approval, training, etc. • 12 months for follow-up on the last patient enrolled. • 6 months for dataset prep and report writing. Cost: grants for patients enrolled is often not most of a study’s budget. Instead, most of the costs are study-wide, not patient dependent: database design, protocol development, site monitoring, etc.
  • 37.
  • 38.
    Interpreting a ClinicalStudy is Still Always About the Same 3 Things Patients: • What patients were actually enrolled? • Do the results apply to a more general population? Treatment: • Was the treatment received the only significant variable? • Did the treatment drive the outcomes? Outcomes: • Were the endpoints clinically meaningful? • Are the outcomes compelling?
  • 39.
    Interpretation The right studydesign and interpretation lets you say something like “a multi-centered observer-blinded randomized trial of early-presenting STEMI patients with large-thrombus lesions showed reduced MACE at 30 days in the treatment arm as compared to the control arm.” And what could be more fun than that?
  • 40.
    Some Key InterpretiveQuestions • Is the enrolled patient sample a fair representation of the target population? • Are the treatment and control cohorts well- matched at baseline? • Are the endpoints really susceptible to the treatment received? • Are the treatment group outcomes really different? • What else besides treatment received could affect outcome? • Do the primary and secondary endpoint outcomes conflict?
  • 41.
    Have a Planfor Going Public! Presentations: • Medical congresses • Late-breaking results • State-of-the-arts • Sponsored sessions Publications: • Academic or professional or business? • Peer-reviewed or other? • High-level, mid-level, or low-level? • Sponsored supplements?
  • 42.
    Actionable Interpretation “I recommendthe meatloaf. It’s just received FDA approval.”
  • 43.
  • 44.
    1. Clinical studiesare done by teams, not individuals: clinical science, statistics, regulatory, medical, device. etc. 2. Your clinical science project is just the last, longest, most expensive, and most fraught step in your design validation process. 3. Unlike bench testing, clinical science is done not by sponsor employees, but by contractors (doctors and hospital staff). They have many other priorities higher than your study, and you can’t change that. 4. Just as you can’t inspect quality into a product, you can’t wash the dataset from a compromised clinical study to make it better. Clean trash is still trash.
  • 45.
    5. More medicaldevices have been shown to work than have been proven to work. 6. The surest way to cure a disease is to start a study of a new treatment for it – suddenly all the patients disappear! 7. Medical device clinical studies are designed by optimists, and interpreted by pessimists.