Microsoft PowerPoint - Med Device PKG Sample Size webinar

Medical Device Packaging

Sample Size and Statistical Rationale

Medical Device Packaging

Sample Size and Statistical Rationale

Live Webinar: April 28, 2016

Karen Greene, Life Packaging Technology LLC and Life Pack Labs,

Vista, CA

[email protected]

760 835 2260

Life Packaging Technology and Life Pack LabsLife Packaging Technology and Life Pack Labs

Life Packaging Technology

Provides customized packaging engineering services

on a global basis: design, prototyping, technical and

quality systems consultation and low volume

manufacturing for packaging feasibility.

www.lifepackagingtechnology.com

Life Pack Labs

Provides expert packaging testing services for

healthcare/ pharmaceutical and medical packaging

www.lifepacklabs.com

AgendaAgenda

• Statistical Rationales---the importance and application of

developing an appropriate sample size for testing

• Compliance and guidance for:

– medical device sterile barrier systems

– Food

• Determining sample size for sterile barrier systems (packaging

systems)

– A process for developing an appropriate and statistically

valid test population

• Summary of key takeaways of non-statistically significant survey

of medical device companies on statistical rationales

• Wrap Up

• Question and answer

Statistical Sample Size DevelopmentStatistical Sample Size Development

Why Have a Statistically

Significant Sample Size?

Why have a statistically significant sample size?Why have a statistically significant sample size?

• For sterile medical devices– it is a requirement

– ISO 11607:2006, part 1—section 4.3- “The sampling plans used

for selection and testing of packaging systems shall be

applicable to the packaging systems being evaluated. Sampling

plans shall be based upon statistically valid rationale”

– TIR ISO 16775 section 5.3: “ Sampling plans should be

acceptable to packaging systems, reflective of risk tolerance,

and be based on statistically valid rationale”.

– TIR ISO 16775 section 12.2.3, Packaging system design

validation: “Sampling plan to be used: Sample sizes must be

large enough to provide for statistically significant analysis to

provide a high degree of reliability, and will be dependent on

corporate risk policy, economics and regulatory requirements”.

Why have a statistically significant sample size?Why have a statistically significant sample size?

• For Food– Food Safety and Modernization Act--2011

PREVENTION

– Written food safety plan– Includes:

• Evaluates hazards that are reasonably likely to occur

(risk analysis)

• Steps put in place to minimize or prevent hazards—your

mitigation

• Verification activities– might include validation that the

preventive controls are adequate for their purpose and

are effective in controlling the hazard

• Generally, cGMP provisions would still apply to facilities

that would be exempt from the hazard analysis and risk-

based preventive control requirements

Statistical Sample Size DevelopmentStatistical Sample Size Development

A Process for Development of a

Statistical Rationale

Statistical Sample Size DevelopmentStatistical Sample Size Development

Risk Analysis

Sample Size Development ProcessSample Size Development Process

Statistical Rationale Development for Validation

Process of product/package risk assessment

• Medical Device—corporate application of ISO 14971:2007

“Application of Risk Management to Medical Devices”

• Formal risk analysis of SBS. Check out Annex TIR ISO 16775

• Most Risk Analyses and Design FMEA’s designate Loss

of Sterile Barrier Integrity as a Critical Defect.

• Review and alignment of risk with corporate assignment of

criticality of risk and statistical rigor related to testing

methodologies, based on defect criticality

Statistical Sample Size DevelopmentStatistical Sample Size Development

Design FMEA

Failure modes and effects

analysis of the packaging/device

system

Sample Size Development ProcessSample Size Development Process

Risk Analyses

Device/Project: Performed by:

•Device part/model number(s):

•Type: (Component/ Subsystem/System)

Date:

Component

/Function/

Procedure

Potenti

Hazard

(or

Failure

Mode)

Effects of

Hazard

(or Failure)

Severity

Potential

Cause(s)/Mech

anism(s) of

Failure

Occurrence

Current

Risk or

Design

Control

Measures

Detection

RPN

Recommended

Action(s)

Responsibility

& Target

Completion

Date

Action

Taken

Severity

Occurrence

Detection

RPN

Sterile

Barrier

System

loss of

sterile

barrier

integrity

Nosocomial

infection—

human

patient

device damaging

primary sterile

barrier packaging

components

material

specifications

Packaging Design

Validation

Engineering

216

Sample Size Development ProcessSample Size Development Process

Statistical Rationale Development

So now that I have determined my defect criticality, now what?

Definitions

• Response type for your specific packaging test (s)

• Attribute data– pass/fail, leak/no leak—binary in nature, discrete value

• Variable data– numerical value– for example seal strength– XX lbs.

force/inch width

• Confidence and reliability

• Confidence or risk level—Central Limit Theorem

• Average value of the attribute obtained by your test samples is

equal to the true population value

• 95% confidence level---95/100 samples will have the true

population value within your range of precision specified

• Reliability—percent defective in the population-based on zero failures

for your testing

Sample Size Development ProcessSample Size Development Process

Attribute Data

Sample Size Development ProcessSample Size Development Process

Statistical Rationale Development

•Response type for your specific packaging test (s)

• Attribute data– SAMPLE SIZE DETERMINATION

•The exact binomial distribution

Sample Size Development ProcessSample Size Development Process

Statistical Rationale Development

• Attribute data– SAMPLE SIZE DETERMINATION

•The exact binomial distribution:

Statistical analysis statement with zero failures-- Based on a simple

passing or failing of the test criteria (package leak). Using the exact

binomial distribution with “X quantity” sample size and 0 failures, at

least “Y%” of the population would meet the validation criteria.

Specifically—sample size of 60, at least 95% of the population would

meet validation criteria with 95% confidence.

Sample Size Development ProcessSample Size Development Process

Statistical Rationale Development

• —95% confidence, 95% reliability--This means that one can state

with 95% confidence that no more than 5% of the population will

exhibit the “defect” or unacceptable condition when tested as

specified within the test protocol.

• Is 95%/95% okay? Depends….

• Criticality of defect—risk assessment

• Share survey data for sterile barrier breach

• FDA has not published requirements, audience will

have direct experience

• Company history with—this type of package design,

packaging process, device/product configuration,

validation history

Sample Size Development ProcessSample Size Development Process

Variable Data

Sample Size Development ProcessSample Size Development Process

Statistical Rationale Development

•Variable data– discrete value

• for example—seal strength –lb. force/inch width.

• Typically a specification or specification limit is

established for variable data, for example a minimum seal

strength value.

– Concept of Tolerance Limits for Variable Data, e.g. Seal

Strength

• Practical boundaries of process variability for seal

strength

• Based on your seal strength distribution, the 95%

tolerance limit will be greater than your specified minimum

seal strength spec.

Sample Size Development ProcessSample Size Development Process

Statistical Rationale Development

•Variable data– discrete value

– Frequently, a lower tolerance limit is calculated at a

specified confidence and reliability—for example,

95%/99%

• The lower tolerance limit (LTL) calculation must be equal to or

greater than the specified minimum seal strength value.

• A lower tolerance limit is calculated as follows---- mean of the test

population - (k value) X standard deviation.

Mean – k*s–

• The “k” factor can be obtained from Juran’s Quality Handbook, fifth

edition, by Joseph M. Juran, A. Blanton Godfrey, 1998, page AAII.36.

Sample Size Development ProcessSample Size Development Process

Additional Sampling Plan

Pointers

Sample Size Development ProcessSample Size Development Process

Statistical Rationale Development

•Some additional pointers:

• AQL Sampling Plans are for use in manufacturing sampling

plans—NOT validation sampling plans

• WHY?

• The Burden of proof of “good” shifts

• In manufacturing, the lot is assumed Good, until

proven bad– biased towards the producer’s risk

• Validation—it is assumed that the requirement has not

been met unless testing demonstrates it is so

Sample Size Development ProcessSample Size Development Process

Some Alternative Sample Size

Determination Methods

Sample Size Development ProcessSample Size Development Process

Statistical Rationale Development

•Alternate Sample Size Determination Method:

–

Finite Population Correction for Proportions

• Finite Population Correction for Proportions:

• n = Validation test sample size—what we are solving for..

• N = Validation lot Size from which you will sample, example- -90

packages

• e = level of precision, reliability, 95%

• Power or confidence level = 95%--assumed

• 74 = 90/1 + 90(.05)

• Test-- n = 74 packages for 95%/95%

Statistical Reference: Determining Sample Size by Glenn D. Israel, University of Florida IFAS

Extension, PE0D6, published 1992 and reviewed 2009 and Cliffs Quick Review Statistics by

David H. Voelker, Peter Orton and Scott V. Adams, Wiley Publishing, Inc. 2001, page 79.

Sample Size Development ProcessSample Size Development Process

Medical Device Companies

Sample Size

Survey

Medical Device Companies--SurveyMedical Device Companies--Survey

Sample Size Development ProcessSample Size Development Process

Wrap Up

Sample Size Development ProcessSample Size Development Process

Statistical Rationale Development

Wrap Up

• Understand your industry compliance requirements for a statistical

rationale for validation

• Execute a risk assessment utilizing tools such as FMEA

• Once risk is assessed, work with your organization to establish the

criticality of the risk or failure.

– Major Defect

– Critical Defect

• Develop your statistical rationale based on the criticality of the defect

• Review the “cost of quality” vs. the risk assessment

– Dummy devices can be used in a packaging validation for a

sterile barrier system

– Packaging systems must be manufactured through a validated

process using process extremes

– Validated sterilization process

– Part of the risk assessment should include the leveraging of

previous validations

• Similar packaging design validations?

• Brand new packaging design and/or materials?

Thank You

Questions??

ReferencesReferences

Works Cited

• ISO 11607, Packaging for Terminally Sterilized Medical Devices, Parts 1

and 2. 2006th ed. AAMI, 2006.

• "Medical Devices." U S Food and Drug Administration Home Page. 07

June 2009 <http://www.fda.gov/MedicalDevices/default.htm>.

• ISO/TS 16775:2014, Guidance on the application of ANSI/AAMI/ISO 11607,

Packaging for Terminally Sterilized Medical Devices, Part 1 and Part

2:2006.

• Statistical Reference: Determining Sample Size by Glenn D. Israel,

University of Florida IFAS Extension, PE0D6, published 1992 and

reviewed 2009 and Cliffs Quick Review Statistics by David H. Voelker,

Peter Orton and Scott V. Adams, Wiley Publishing, Inc. 2001, page 79.

• Juran’s Quality Handbook, fifth edition, by Joseph M. Juran, A. Blanton

Godfrey, 1998, page AAII.36.