overcoming the hurdles for pat and implementation of …€¦ · –rtrt/pat then can enable...

Overcoming the Hurdles for PAT

and Implementation of Real-Time

Release Testing - RTRT

Dr. Gert Thurau

MSD Manufacturing Division

Property of Merck Manufacturing Division 2012

Merck & Co., Inc, Whitehouse Station, NJ,. All rights reserved

2

Definition of PAT

Process Analytical Technology:

– In-line, on-line, at-line measurements (but not off-line

= laboratory)

– Real-time or in time-scale of processing step

– Analyzer based and model-based real-time

approaches

Many definitions (FDA, compendia, literature..)

All are useful…

Definition used in this presentation:



3

ICH Q8 (R) Definition of RTRT

• Real-time release testing (RTRT)

– The ability to evaluate and ensure the acceptable

quality of in process and/or final product based on

process data, which typically includes a valid

combination of measured material attributes and

process controls.

• Definition does not specify:

– How fast is “real time”?

– Is Process Analytical Technology used?

– Is there release testing?

– Can refer to drug substance, however most examples discussed are for drug

product



4

Types of RTRT Approaches

A. Analyzer/”test-replacement” type

approaches

– Analytical testing moves from end product

testing to further upstream

– Direct or indirect replacement of conventional

release tests – use of PAT

B. Process-model based approaches

– Comprehensive process models

– To verify process is in Design Space

– To replace most testing* and predict

product quality based on raw material and

processing data

*Current examples mostly to replace some end-product tests

that cannot be put online (i.e. dissolution)



5

A. Analyzer based approaches

• Shift of analytical control from an off‐line, post‐ manufacturing approach to an approach where data is

generated during manufacturing

– Full or partial RTRT - all or selection of tests done during

manufacturing

– Typically RTRT refers to no post-production release testing

• RTRT does not mean less testing - in fact it often means

more analytical data is generated

– Process Analytical Technology (PAT) allows generation of more

measurements without significant additional resources



6

A. Analyzer based approaches,

cont.

• RTRT often requires a combination of favorable product properties, and good process understanding – For example product is stable with no degradates => no need for

degradate testing

– Dissolution testing replaced by disintegration test

• For those analytical tests done “on-line” equivalency to finished product testing is generally needed – PAT methods compared to reference/conventional analytical

methods

– Technical limitations in microbiology and stability tests

A big initial advantage of “test-replacement” approaches is that at any point in time a reference method can be used on a sample.



7

Historical timeline of Analyzer-

Based RTRT Approaches

• 2004-2005 Sanofi-Aventis Tablet Product

– Initial FDA PAT initiative “PAT comparability protocol” including many NIR

methods

– Filed but not implemented

• 2006 MSD/Merck Tablet Product

– Filed and implemented for new product

– Globally accepted in 80+ markets

• Since then

– Multiple implementations, typically by main pharma companies

– New and already marketed products

– Increasing complexity (lower drug loadings, more complex formulations)



8

Analyzer-based RTRT Examples

Pfizer1

MSD/Merck

Sanofi-Aventis2

GSK1

1) IQ industry user meeting/IFPAC 2012

2) IFPAC meeting 2004



9

B. Process Model based approaches Multivariate Data Analysis (MVDA)

• Use of process models (mostly multivariate) to control

process or predict – vs. measure – in-process material

and finished product attributes

Inputs into models:

–Raw material properties

–In-process

measurements

–Equipment parameter

Outputs of models:

–Process within Design

Space?

–Fault detection

–Finished product

properties prediction

Process

Model

• Today multivariate process models are often used in combination with

analyzer based RTRT

• Model predicts finished product property that cannot be analyzed via “real-

time test” (i.e. dissolution)



10

Examples for MVDA Individual unit operations => complete process model

• Fluid Bed Granulation: Input parameters

– Atomizing air pressure and flow

– Solution flow rate & totalizer

– Inlet air temperature, product temperature,

outlet air temperature

– Filter bag differential pressure

– …

• Pan Coating: Input parameters

– Air flow

– Inlet air temperature

– Air dehumidifier

– Pan load

– …

0

2

4

6

8

10

12

14

16

18

20

-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56

T2R

ange

PS

[1 -

Las

t com

p.]

Time in Phase (mins) (shifted)

Janumet_G2_WP69_01.M2:GRANULATING, PS-testset

Predicted Hotelling's T2Range[1 - Last comp.]

T2Crit(95%) = 6.02899 T2Crit(99%) = 9.29155

T2Range[1 - 2] (Aligned) (Avg) T2RangePS[1 - 2] (Batch 2158074_4)

T2Crit(95%)

T2Crit(99%)

SIMCA-P+ 12.0.1 - 2012-05-08 15:32:09 (UTC-5)

Multivariate Trend

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

0 10 20 30 40 50 60 70 80 90 100 110 120t[1]

$Time (shifted)

Janumet_P500_WP69_01.M2:COATING

Scores [Comp. 1]

SIMCA-P+ 12.0.1 - 2012-05-22 10:56:06 (UTC-5)

Model

Model

…in combination with raw material data can formulate predictive models

for finished product properties

• Dissolution behavior



11

Comments on MVDA based

approaches to RTRT • Challenge to fully explore “outside design-space” in full-scale

production – Sensitivity of model to relevant failure modes can be difficult to prove

• So far limited examples for full RTRT with elimination of end-product testing – Typically models run for monitoring, in parallel to conventional testing

– Hesitation to “switch over” to full model control • Industry

• Regulatory agencies



12

Real-Time Release Testing &

Quality by Design

• According to ICH QbD IWG they do not have to

come together

– Can do RTRT without QbD

– Can definitely do QbD without having RTRT!

…however in practice…

–Quality by Design leads to increased process understanding

–Process understanding is foundation of Real-time Release Testing

–RTRT/PAT then can enable further ongoing process understanding (= extension of QbD learning into life cycle of product)



13

Advantages of RTRT (In combination with QbD)

1. Typically more integrated control strategy – Relationships between raw materials, in-process tests and

finished product properties better understood

– Earlier control points than end-product testing

2. Better data for use in process understanding – Often more data to use per batch

– Supports stage III/ongoing process verification

– PAT analyzer can detect process drifts easier

3. Better Quality Assurance than conventional end product testing – Source: EMA guidance on real-time release testing



14

Advantages of RTRT, cont. (In combination with QbD)

4. Involvement of manufacturing operations in Quality Control – More ownership of Quality

– Up-scale of job functions

5. Potential savings through elimination of end-product testing – Capital assets, expenses, staffing

6. The ultimate lean process – Product is tested when it is manufactured

– => no “supply chain issues” between manufacturing and laboratory



15

Challenges with RTRT

1. Technical/scientific limitations – not every product/process is currently amenable to RTRT

2. Investment – Capital investment in PAT and IT systems

– New skill sets required in manufacturing sites

3. Management and execution of innovative approaches to – Analytical testing

– Setting of specification – large number of samples tested, expl. Uniformity of dosage units

– Roles and responsibilities in manufacturing and Quality control



16

Challenges with RTRT, cont.

4. Limited experience with long-term performance of PAT methods (for example NIR) in the pharmaceutical industry

5. Today still more regulatory attention than conventional control strategies

– More intense interaction during review

– More involved pre-approval inspections/subsequent GMP inspections

– Global expectations very different

Case Study – Benefits and

Challenges

MSD/Merck Real-Time Release

Testing Product



18

Product and process introduction

• Indication diabetes (DPP4 Inhibitor)

• Solid oral dosage form – Tablet

– 3 strengths which are weight-multiples

– ~ 30% active by weight

• Direct compression process with taste-masking overcoat

• Initial approval 2006 – FDA, EMEA

– Subsequent approval in > 110 markets world-wide

• Developed into high volume product ( > 850 batches annually)



19

Product attributes that supported

RTRT for this product

• Simple direct compression process with good development history of uniformity – Relatively high active content (~30%)

• Very good stability*

• No degradates*

• BCS class I with well understood disintegration/dissolution behavior

Disclaimer: These attributes are not mean as universal guidance, there are

scenarios/attributes for other products which can support RTRT approaches.

* for approved stability window, under approved storage conditions and packaging



20

Merck Fully Supports QbD

• Merck has embraced QbD as a strategic initiative on

how we develop and manufacture products

– QbD provides a consistent framework for developing high quality

products that provide benefits to our patients and meet our

customer’s needs

– QbD promotes systematic, scientific and risk-based approaches

to product and process development

• Merck is executing a company-wide QbD strategy and

playbook

• All of Merck’s development programs now follow the

QbD approach

– Work processes are established to realize Merck’s QbD strategy



21

RTRT in the context of QbD for this product

• This product was Merck’s first venture into QbD principles in 2005-2006 – Participation in FDA QbD pilot program

• “Quality by Design (QbD)” approaches used in – Development

• “Design Space”

• Systematic assessment of control strategy elements – Analytical/testing

– Process equipment

– Procedures

– Manufacturing • Implementation of control strategy

• Process monitoring plan

• Ongoing improvement/learning

• QbD elements covered in this presentation – Control strategy (RTRT approach)



22

Control Strategy for Product – Real-Time

Release Testing for Product Release

API Blend

Lubricant Blend

Compress

Film coat

PAT Testing Approach

Manufacturing Floor

Disintegration

NIR ID

Appearance

Manufacturing Floor

NIR Composite Assay

On-line Dosage Uniformity by

Weight

Traditional Release

Approach

Laboratory

Lab Tests

HPLC

Content Uniformity

Composite Assay

Degradates

ID

Dissolution

Appearance

NIR technology for tablet

analysis



24

Challenges of this NIR method

implementation

Challenge Solution

In 2005 limited external applicable

guidance and very limited precedence in

industry

•Apply widely accepted standards where

possible (i.e. ASTM)

• Capture more details of strategy and

execution in submission, engage in

discussion with reviewers pre-/post

submission

Existing quality system not in all aspects

conducive to support NIR method for this

purpose

Adopt changes to existing quality system

(vs. creation of new quality system)

Method validation and update expertise

in manufacturing site evolving

Training, knowledge transfer and close

collaboration with central expert group for

life cycle of product



25

Reference

values

PAT CALIBRATION &

DEPLOYMENT

At-line NIR

Assay

Calibration

tablets

y

EMPIRICAL

Multivariate

Calibration

(PLS)

S

Calibration

Spectra

PLS Model

of rank = K

M

variables

N

samples

X

Production

tablet

Prediction

spectrum

Concentration

value by NIR

Outlier metrics:

REAL-TIME

results

confidence!

Deployment

software

x



26

Concentration Data - History

Average Concentration by batch (mg/g)

1,399 Batches plotted (representing nearly five years of production)

Original average and control limits shown – these were set by first 30 batches produced.



27

Within Batch Concentration

Trending

Early Batch Recent Batch

With over 1,000 batches between these two batches and nearly 5 years of production, the consistency of concentration values is excellent.



28

NIR Quality Assurance/

Model Life Cycle Management • NIR does not use daily reference standard or ‘calibrator tablet’

– Instead previously developed calibration is used

– No routine comparison to reference method

• Quality assurance of measurement through model verification via:

– Reference method analysis (initial launch, change control, investigation)

– Spectral outlier diagnostics (routine operation)

• Model life cycle management:

– NIR is able to detect subtle differences in material properties

– Model (but not method) will change over time (example new source of API)

– Well accepted concept -Foundation is strong analytical science and solid change control Quality System

So, how robust are these initially developed NIR calibrations?



29

Multivariate Outlier Metrics

K-dimensional PLS

Model subspace

(plane)

In-space outlier

(high M-distance)

N NIR spectra

(M variables

each) used to

build model

• Calibration spectra

define “normality”

• Any new spectrum

can be abnormal in

2 ways:

1. in-space, and

2. out-of-space

7500 8000 8500 9000 9500 10000 10500-2

-1

0

1

2

3

wavenumber (cm-1)

inte

nsity

Out-of-space outlier

(high F-value)

7500 8000 8500 9000 9500 10000 10500-2

-1

0

1

2

3

7500 8000 8500 9000 9500 10000 10500-2

-1

0

1

2

3

4

• Monitoring of metrics is imperative: REAL-TIME NIR results confidence !

• Two metrics reflect different NIR failure modes both must be monitored!



30

PAT Method Events

1/1/2006 1/1/2007 1/1/2008 12/31/2008 12/31/2009 12/31/2010

25mg

50mg

100mg

MAR, JUL 06: 3 NIR

models put into

service

10.3.2010 C 132 MD alarms over

8 batches! No sign differences

Investigation of outlier metric

limits (too “tight”???)

18.3.2011

(all): new

MD limits set

using 95%

CL

30.6.2011 C:

model

update 29.7.2009

B: 6 MD

alarms

over 2

weeks, no

sign

difference

•Cmodel generated useful metrics right away; A and B models did so after a model

update

• Since 2006, only 6 confirmed tablets with sign. differences (out of >120k!)

• all of which were flagged by the outlier detector!

• none of which were confirmed as product quality issues!

• However, 132 false alarms for C model in 2010

A and B model

updates

B model update effect

on F value

A strength B strength C strength

12.2.2007

C: 2 F

alarms,

sign.

difference

2 4 6 8 10

x 104

0

500

1000

1500

Sample

F v

alu

e

F value

F above limit

F below limit

27.1.2009

C: 1 F

alarm,

sign.

difference

B: 3

separate

MD

alarms,

sign.

difference



31

Conclusions for NIR Method

Robustness

• Spectral outlier use is appropriate to assure

method performance

– Limits are set to detect method performance issues

– With large process history, opportunity to adjust limits

to mitigate nuisance investigations

• NIR method extremely robust over 5 years of

production

– Includes transfer to other instruments, other

manufacturing sites



32

Regulatory Reviews/Inspections

• RTRT/PAT approaches are usually the center of any review or inspection

– Unusually large number of detailed questions

• >80 CMC questions in initial filing

• Many aspects of PAT method scrutinized in recent inspections

– Diversity of background and familiarity with RTRT in inspectors

– Can be intimidating scenario for manufacturing sites

• However:

– Focus of questions typically on learning and understanding, not

challenging

– Strong integration into routine Quality System in sites is assuring (no

“special PAT Quality rules”)

• Routine testing and release flow

• Investigations

• Transfers (within site, between sites)



33

Cultural Change in

Manufacturing Site

Education in chemometrics and

modelling for system administrators

(On a need to basis)

Design space control and update

Periodic method verification to control

and verify model = laboratory

resources

Model management and update

Job enrichment !!!

NIR administration

• RTR process is less labor intensive, but

new activities are necessary



34

Real-Time Release Testing

Business Assessment

• Additional business benefits in

– Inventory Management

Opportunities

– Reduction in Waste

- 97.3 %

Quality Cycle Time per batch

7

1

1

1

0

1

2

3

4

5

6

7

8

9

Conventional

Process

RTR Januvia

Cal.

Days

Lead Time /batch (QA)

Lead Time /batch (Lab)- 75%

RTR Economics - FTE's

0.0

1.0

2.0

3.0

4.0

5.0

6.0

2007 2008 2009 2010 2011 2012 PP

Year

FT

E FTE Savings

Total FTE for RTR



35

Conclusions

• Real-Time Release Testing is advancing within

the pharmaceutical industry

– Number of examples increasing

– Sophistication of approaches increasing

• Challenges exist both on scientific, business

and regulatory aspects

• Experiences within currently running

implementations are clear indications of benefits



36

Acknowledgements

• Eric Ahuja, Niya Bowers, Manoharan

Ramasamy, John Higgins, Nathan Pixley,

Charles Miller, Steve Short

• Beppe Mazzochi, Gianmaria Ghisoni, Paola

Carrera, Jennifer Blatteau

Backup slides



38

Sampling Plan for CU and Assay

• Core tablets are obtained every 15 minutes from the tablet press via automated sampling system – The number of tablets tested varies by dose and batch factors

• Samples for NIR analysis – 3 tablet per “time location” per side (station) of the tablet press

– Typical number of tablets for 100mg dose is 72 tablets tested by NIR

• 36 tablets for side 1, 36 tablets for side 2

• This corresponds to 12 locations (with 3 replicates)

• 12 locations x 15 minutes = 3 hours of compression

• Tablet weight data – Between 240-690 tablets analyzed (distributed by location and

side-of-press)



39

Release calculations

• Composite Assay

– Each NIR result is in concentration units (mg/g of active)

– NIR results are averaged

– Composite Assay (%LC) is calculated using average

NIR mg/g and average tablet weight

– Specification 95-105%

• Content Uniformity

– Use weight uniformity instead of content uniformity

(relatively high drug load)

– Specifications based on large n counting test

overcoming the hurdles for pat and implementation of …€¦ · –rtrt/pat then can enable...

Documents