2ND INTERNATIONAL SYMPOSIUM ON CONTINUOUS MANUFACTURING OF PHARMACEUTICALS, CAMBRIDGE, MA, SEPT 2016

Stephen Conway, Samantha Hurley, Catherine MacConnell, Robert Meyer, Christine Moore, Cindy Starbuck, Manoharan Ramasamy, Brendon Ricart, Frank Witulski, Fan Zhang-Plasket, et al.

Proof of Operations for Continuous Manufacturing and Real Time Release Testing of Film Coated Tablets

Outline

Merck’s Approach to Continuous Manufacturing

PAT Options: RTD Process Model and Blend NIR

Process Model for Material Tracking and Rejection

Sampling for Control and Release

Benefits to Patients and Company

Moving Towards Worldwide Acceptance of CM

Why Continuous Manufacturing (CM)?

High assurance of product quality

Lower cost

Flexibility to meet changing demand

How we’ll start

Leverage extensive experience with an existing product

Build upon existing RTRT, raw materials monitoring and QbD approaches

Conduct short duration runs with efficient, frequent changeovers between strengths

Use batch size flexibility to match changing market demand

What we ask of regulators

During global industry shift from batch CM, alternate production methods are required in parallel

This ensures uninterrupted product delivery to patients

What comes next

Conversion of other marketed products currently produced by batch processing

Addition of other continuous technologies (e.g. granulation)

Efficient development and launch of new products

Merck’s Approach to CM for Oral Solid Doses

How We’ll Use CM to Improve Supply Chains

One batch, every cycle, every strength

• Flexible batch size matched to customer demand

• Fast changeovers between products

High demand products

• Higher quality and efficiency due to elimination of start/stop

Low demand products

• Increases shelf life by eliminating overproduction and inventory hold

Volatile demand products

• Avoid shortages by reacting quickly to changes in demand

Monthly Weekly

Jan Feb Mar Jan Feb Mar

Inve

nto

ry

Time Time

Deliveries

Cycle

Stock

Safety

Stock

Monthly vs Weekly Inventory Levels

Existing Product & Process Description

TABLETING

Co

ati

ng

So

luti

on

Blend

Tablet Compression

Film Coat

• Available >10yr in most markets

• Multiple strengths

• High overall volume

• Volume dependent on strength

Commercial Characteristics

• Drug load > 25%

• BCS I

• Non-functional film coat

• Low segregation & physicochemical stability risks

Product Characteristics

• Direct compression + film coat

• Batch size >500kg

• Assay using transmission NIR

• Tablet weight in lieu of CU

• Hardness / disint. in lieu of disso

• No degradate testing

Manufacturing Description

Merck’s Continuous Manufacturing Process GEA CDC-50, Consigma Coater, Bruker TANDEM

Mapping Different Approaches to CM

-5

-4

-3

-2

-1

0

1

2

3

4

5

-5 -4 -3 -2 -1 0 1 2 3 4 5

System ComplexityFixed / Early

Fixed / Late Flexed / Late

Flexed / Early

Janssen

Vertex

Pfizer

BMS

Merck

Roche

Bayer

AZ

GSK

Lilly

Novartis

Sanofi

Equipment Flexibility / Portability

Po

rtfo

lio D

eve

lopm

ent S

tage

(Pro

duct L

ife

Cycle

)

Comprehensive Control Strategy

8

Enterprise Resource Planning Site Manufacturing Execution System

Supervisory Control and Data Acquisition PAT Management System

Automatic Controls PAT Instrumentation

Diversion (potentially non-conforming)

Process

Mate

rials

Managem

ent

Mate

rials

Managem

ent

Batch defined by number of accepted tablets

Solids Handling Unit

Blender 1 Blender 2 Tablet Press Coater 2

Coater 1

Distribution Arm

Loss in Weight Feeders (x6)

Diverted Tablets Metal detected Disposition Evaluation

On-Line Tablet Test

(W/T/H, Assay by NIR)

Finished

Product

Release

Raw Materials

Monitoring

RTD Process Model (potency)

Blend NIR

(development)

Manual controls and procedures

Appearance Check OOS Check

Refill

Speed Speed Level

Suspension/Air Flow

Temp, RH

Process Validation/Evaluation Statistical Process Control,

model checks, event checks

Mass Flow

Rates,

Background on RTD and NIR

Measure

Spike or step

change API

Blender 1

Blender 2

Entire

System

E(t

) (1

/min

)

Time (min)

E(t

) (1

/min

)

1. Perform Experiment

3. Check Model Predictions

2. Fit Model

Experiments Using Redundant PAT Demonstrate Low Risk If Single Method Used

CC BY-SA 3.0

PAT Summary: Examining Operations & Robustness of RTD Model and Blend NIR

Factor RTD Process Model Blend NIR

Method Basis

Scale of Scrutiny

Prediction Location

Model Robustness

Blind Periods

Signal to Noise Ratio

Fouling / Equipment

Model Maintenance

Requirements

Universal Applicability

First principles calculation Empirical multivariate calibration

Average prediction for >10 tablets Sample size ≈ 1/4 tablet

Sensitive to material properties and process parameters that affect flow

or blending

Sensitive to physical properties and process parameters that affect sample

presentation

Predicts blend and tablet API concentration

Measures API concentration in blend at feed frame entrance

Flow rate assumption during ~3 s feeder refill

No measurement during ~2min probe cleaning

High Medium

Low risk to feeder load cells Dependent on adhesion properties of

product

Updates required based on bulk density, flowability, or process

parameter changes

Updates required based on probe, spectrometer, material property or

process parameter changes

Can predict concentration of any component as needed

Applicable for components with characteristic NIR bands with sufficient

specificity

Blend NIR

Feeder Output (% API)

RTD Model Prediction

HPLC

RTD vs. NIR Deep Dive

Sample Fraction per Second

RTD vs. NIR

= measured tablet

= unmeasured tablet

= sample size

Signal to Noise Ratio Compared with HPLC Results

Assumptions:

50 kg/hr throughput

400mg tablet

1 Hz RTD prediction

NIR: Seven 5mm

rectangular windows

RTD vs. NIR Deep Dive

Model Accuracy vs. HPLC

Model Maintenance

30

31

32

33

34

0 5 10 15 20

NIR

Pre

dic

ted

AP

I C

on

cen

trat

ion

(%)

Time (min)

Jan-15 Nov-15 Jun-16

n = 30 St. Dev.

API Feeder 0.81

RTD Model 0.08

Blend NIR 1.98

Tablet NIR 1.55

Tablet HPLC 0.76

0.0

0.5

1.0

-4 -3 -2 -1 0 1 2 3 4Concentration (mean centered)

API Feeder

RTD Model

Blend NIR

Tablet NIR

Tablet HPLCMethod error ≈ 0.46

Work in Progress: Model Robustness to API Source Change

RTD

Blend

NIR

• Source A

• Source B

• Source C

• Source A

• Source B

• Source C

RTD Model for Raw Material (RM) Tracking Model provides exquisite ability to track RM from start to end

Raw Materials

Feeders Blenders Tablet Press

Tablet Coater

Bulk Tablets

TANDEM

Diverted Tablets

Diverted Tablets

NIR Model Tablet Diversion System:

Material Tracking System:

time

(min)

0 RM Lot 2 RM Lot 1 Drum A

Drum Size = 30kg

RM Lot 1 2 RM Lot 2 Drum A

Mass Flow Rate = 60 kg/hr

10 RM Lot 2 Drum B

15 RM Lot 2 Drum B

RTD Process Model

16

17 Copyright © 2015 Merck & Co., Inc., Whitehouse Station, New Jersey, USA, All Rights Reserved

Implementation of the RTD Process Model

for Material Rejection

17

Co

nce

ntra

tio

n

Time

API Concentration from Feeders

RTD Model Prediction at Tablets

Tablet Action Limit

Tablet Rejection Limit for ModelPrediction

a b c d

API spike

2%

2%

100%

(a) beginning of tablet rejection

(b) first predicted average tablet potency outside acceptable limits

(c) predicted average tablet potency returns to acceptable range

(d) tablet rejection ends

• Yellow shaded regions indicate conservative diversion of tablets

• Red shaded region indicates diversion of tablets predicted to be

outside of acceptable limits

Time

Preliminary Thoughts on Sampling During CM

•Most monitoring & control loops operate at f ≥ 1Hz

•TANDEM used for control and release: weight / hardness / composite assay

•Sampling must be statistically representative

•If process is capable (Cpk>1.3) and in state of control:

–Strict criteria on sampling interval not needed

–Risk between samples is to the business, not the patient

•Monte Carlo modeling will inform final decision

0 500 1000 1500 2000 2500

0

500

1000

1500

2000

2500

3000

0

100

200

300

400

500

600

700

800

900

0 10 20 30 40

Total Mass (kg)

To

tal W

eig

ht

Sam

ple

s f

or

DU

To

tal N

IR S

am

ple

s f

or

CA

Time (hr)

TANDEM NIR Max Rate

Equal Batch Sample %

Weight Control Min Rate

Fast then slow

If RM lot change

Batch Process

Transition

Min batch size

Min samples

Adaptive sample

rate based on

risk

-fast at start &

after change

Moving Towards Acceptance of Continuous Manufacturing Technology

Failure to gain approval of any of these components in any regulatory region sinks the entire ship

What Industry Can’t Do

– We cannot run different control strategies for different regulatory regions

– We cannot maintain different RTRT models for same test for different markets

– We cannot develop a single technology platform without assurance of global acceptance

We embrace O’Connor, Yu and Lee’s proposal (Int. J. Pharm. 509 (2016) p. 492)

– “international harmonization of approaches for expediting the global adoption of emerging technologies.”

An Ethical Dilemma?

– We want the highest assurance of quality of drugs for patients

– We want the most economical manufacturing to benefit our shareholders

– We want to be able to supply all markets

What agreement is needed from global regulators to move CM forward efficiently?

To best serve our patients, we want the flexibility to deliver

our medicines to any patient worldwide

Adjusting

formulation can

improve existing

products

Bioequivalency

studies are

complex & costly

Formulation

Flexibility

Concurrence that

batch and

continuous

processes can

coexist, sometimes

at different

locations, with

slightly different

formulations

Parallel

Production

Processes

Level of

redundancy (eg

RTD model vs

RTD + blend NIR)

based on risk

Sampling

requirements

RTRT vs. end

product testing

Consistent Control

and Release

Strategies Across

Borders

Production

duration and

rate

Future ability

to expand

range

Flexible

Batch Size

Conclusions Continuous manufacturing offers benefits to manufacturers and to patients through quality, agility and flexibility

Equipment Manufacturers

Drug Companies

Health Authorities

Academia

Continuous direct

compression, film coating,

and RTRT for an existing

product provides a risk-

prudent way to

– Demonstrate proof

of operations

– Achieve regulatory

acceptability

Eventually enabling future

applications of continuous

manufacturing technologies

for new products

Collaboration is needed to overcome obstacles and

allow patients to reap the benefits of CM

East

West

North

South

Acknowledgements

Christine

Moore Cindy

Starbuck

Cat

MacConnell

Frank

Witulski

Brendon

Ricart

Sammi

Hurley

Mano

Ramasamy Steve

Conway

Fan Zhang-

Plasket