1

Pharmaceutical supply chains: key issues and strategies for optimisation

Nilay ShahCentre for Process Systems Engineering

Imperial College London

2

Outline

• Some conclusions• Background• The pharmaceutical industry supply and value chains• Supply chain issues• Primary manufacturing:

– Risk management– Process development

• Challenges for the future

3

Some conclusions

• The pharmaceutical supply chain is very complex, with many interacting facets– Difficult to generate radical improvements quickly– Piecemeal approaches (e.g. improved logistics) will generate

incremental benefits• Current process technology is one of the main supply chain

bottlenecks– Many “built-in” inefficiencies that constrain performance– Not a very responsive system

• Current models in the research community are too “company-centric”– Future models need to consider a holistic view of an extended

supply chain of specialist agents• IP generators• Testing specialists• Contract manufacturers• Logistics providers• Healthcare providers/consumers

4

Trends in the pharmaceutical and related industries

• Time to market is the key metric• R&D productivity (numbers of new chemical entities registered

per unit amount of investment) is declining• effective patent lives are shortening• even while active, patents provide lower barriers to entry• many cheaper product substitutes in many therapeutic areas

– alternative compounds (“me-too drugs”)– off-patent generics

• payers of healthcare exerting strong price pressure and influencing prescribing practices

• for approval, new drugs must:– address new therapeutic areas; or– have very significant cost or health benefits over existing

treatments.

5

Value growth

119.1

10.1

50.6

17.5

86.3

14.7

0

20

40

60

80

100

120

140

1981 1999

$ bnUSJapanEurope

6

Product pipeline

DiscoveryPre-Clinical

ClinicalTrials

Launch Commercial

7

Compound success rates by stages

8

Decreasing R&D efficiency

Total Spending on R&D $bn

New Drug Applications to FDA

9

The value chain

• Discovery generates candidate molecules• Variety of trials evaluates efficacy and safety• Complex regulation process• Manufacturing:

– Primary manufacture of active ingredient (usually 1-2 sites)– Secondary manufacture – production of actual doses (up to 20

sites)– Often geographically separate for taxation, political etc reasons– Complex logistics

• Distribution– Supply chains often global– Many third parties become involved

• Distributors, health authorities etc.

• Retail– Pharmacies, doctors and hospitals are main outlets for ethical

drugs

The

“sup

ply

chai

n”

10

The value chain

Research & development 15%

Primary manufacturing 5 - 10%

Secondary mfg/packaging 15 - 20%

Marketing/distribution 30 - 35%

General administration 5%

Profit 20%

Total 100%

11

The supply chain

• Difficult to get value from early stage discovery/trials processes (cf. decreasing R&D efficiency, recent mergers)– Can Process Systems Engineering techniques help generate

more focussed searches/libraries?• Companies view supply chain differently:

– was a means of getting product to where it was needed– now a means of delivering additional value

• (At least) three interesting problems:– Process development and design– Planning of trials/testing and capacity under uncertainty– “Classical” Supply chain planning and management

12

Process development and design

• Problems:– Process chemistry, solvent and catalyst choices result in

• Low material efficiencies (of order 1%)

– Inefficient, very traditional batch manufacturing processes result in• Low velocity ratio or value-added time (of order 1%)

– Sub-optimal design of drug delivery systems results in• Low bio-availability where required (of order 1% for traditional

formulations e.g. pills)

• 1mg delivered to target area:– may require 10kg of materials overall!– ties up a considerable amount of capital!

13

One stage: typical overall mass balance

Effluent 42%

Manufacture

Landfill 9%

Incineration 6%

Material in 100%

Product 10%

Recovery/recycle 29%Solvent ‘loss’ 2%

By-product sold 2%

14

Typical batch process

Measure

Working capital for 4 days

How long does the conversion process really take?

Low value-added time!

Store

2 days

React

12 hrSeparate

12 hr

Store

18 hr

Purify

12 hr Product

Separate

12 hr

15

Longer time: higher costs and lower quality

Cost

Time

Quality

16

Process development and design

• Many research groups now working in relevant fields• More of a design than operations issue

– Significant improvements in material efficiencies required• Involvement of process (systems) engineering at early stage

– Model-based design– True catalysis rather than stoichiometric reagents– Optimise overall material efficiency rather than reaction yields– Large reductions in solvents

» Need for better heat transfer technology– Improve manufacturing performance

• Run processes as close to intrinsic rates as possible• Use small-scale continuous processing where possible• Avoid stage-to-stage isolation where possible• Cleaning and changeovers! (see later)

– Improve drug delivery to be more targeted (new field)

17

The supply chain

• Difficult to get value from early stage discovery/trials processes (cf. decreasing R&D efficiency, recent mergers)– Can Process Systems Engineering techniques help generate

more focussed searches/libraries?• Companies view supply chain differently:

– was a means of getting product to where it was needed– now a means of delivering additional value

• (At least) three interesting problems:– Process development and design– Planning of trials/testing and capacity under uncertainty– “Classical” supply chain planning and management

18

Testing and capacity planning

• “Traditional” sequential approach:

Tests foractivity

Tests fortoxicity

Tests forefficacy

fail

fail

fail

pass

pass

Tests characterised by:Duration*

Cost (in-house or outsourced)*

Resource requirements*

Hard precedence constraintsSoft/conditional precedence constraintsProbability of success

*may be distributions rather than known

Sequential approach conserves resources, butmay increase time to market

19

Optimised planning of tests(Grossmann & co-workers, Pekny, Reklaitis and co-workers)

• Rather than follow sequential approach, approach from a resource-constrained scheduling perspective

• Some tasks have conditional dependence• Degrees of freedom on task precedence• Optimisation balances:

– Risk of unnecessary expenditure– Potential rewards of coming to market earlier– Resource constraints and outsourcing costs

• Resource-constrained stochastic optimisation problem– Conservative approaches (always feasible)– Hybrid simulation-optimisation approaches

20

Capacity and portfolio planning under uncertainty

• What technology/capacity, where, when (plant fabrication lead times!), whether to outsource

• Products to prioritise in the R&D pipeline to structure the future portfolio optimally

• Most severe for pharmaceuticals:– capacity requirements very dependent on outcome of

clinical trials, registration etc.

• Extreme cases– pessimistic: no investment and many successful

products: severe capacity limitations– optimistic: investment → plenty of capacity but no

new products → patent expiry issues

• Need for systematic way to balance risks

21

Product Pipeline and Capacity Plans

promisingCT results

current productsmaterials

entering CT

time

dem

and

successfulproduct life-cycle

How to:• allocate capacity between products ?• plan capacity investment ?

22

Clinical trialsMore than two outcomes!

clin

ical

tria

ls

Phase II (~2 year)Stage 1

Phase III 3-5yearsStage 2 & 3

Registration(1 year)

Success

Failure

0.9

0.1

High

Target

Minimum

Failure

0.10

0.10

0.30

0.50

0.40

0.60

Success

Failure

0.95

0.05

Not Considered yet in this work

High

Low

Deterministic stage Launch

23

Scenario Tree & Clinical Trial Outcomes

Stage 316

1-4

29-32

61-64

+1 StochasticProduct C3

H T L 7

Stage 4

11

2

3

4

+1 StochasticProduct C2

H F

H T

H H TH H

H L

H F F

+1 StochasticProduct C4

1 DeterministicProduct C1

H

Clinical Trial OutcomesH – High Success T – TargetL – LowF – Failure

Stage 1Periods

1

Scn 28H T L F

30 40101 20

Stage 2

Scn 63H F F L

Scn 2H H H T

24

ResultsNPV profile

25

Strategy matrix: risk analysis

Worst Case ( Exposure ) NPVs(-100) Scenarios Legend

Aa AbBa BbCD

FaGH

EbFb

Bc

Ea(INCREASINGRISK)

(-130)

(-160)

Probability of Losses15%30% 0%

26

Combined testing and capacity planning(Grossmann and co-workers)

• Interesting way forward– Holistic analysis of value chain

• Better synchronisation leads to material being ready if tests are successful

– Avoid shortage of material for clinical trials

• Basis for future work; extensions:– Management of risk is a key feature

• Real options techniques should be relevant

– Needs to take account of global trading structures– Needs to include creative possibilities in model

• e.g. placing of low-commitment options with subcontractors

– Needs to take more extended view of supply chain

27

The supply chain

• Difficult to get value from early stage discovery/trials processes (cf. decreasing R&D efficiency, recent mergers)– Can Process Systems Engineering techniques help generate

more focussed searches/libraries?• Companies view supply chain differently:

– was a means of getting product to where it was needed– now a means of delivering additional value

• (At least) three interesting problems:– Process development and design– Planning of trials/testing and capacity under uncertainty– “Classical” Supply chain planning and management

28

“Classical” supply chain planning and management

• Some performance measures– Pipeline stocks may be 30-110% of annual demand– Finished good stocks 10-50% (4-26 weeks) of annual demand– Supply chain cycle times of order 1000s of hours– Value added times 0.3-5% of cycle times– Supply chain costs overtaking R&D costs

• What can better operations deliver?– 30% stock reduction– 30% increase in value-added time– 7% reduction in supply chain costs

• Benefits of improved operation for one large drug:– $30m one-off– $8-16m p.a.

29

Issues/Structure

• Primary production processes usually “slow” and “unresponsive”– lowish yield– labour- and time-intensive – can take 30-200 days from end to end– many QA steps along the way– long changeovers (one to four weeks) force campaign operation

• Secondary processing often geographically separate from primary– transportation lags, but sensible use of API storage can mitigate

• Secondary processing sometimes serves market directly, but more commonly:– regional storage locations/wholesalers and other agents

• Long supply chain cycle times (60-300 days)– Many delays in process– poor responsiveness to changes in demands– High service levels required → high stocks

30

Dynamic supply chain analysis

• Dynamic analysis of existing supply chains generates considerable insight– Understand relationship between policies/parameters and

performance measures

• Use a generic modelling approach which captures physical and business processes

• Library of supply chain objects based on generic node:

Inboundmaterial

management

Handling/processing/conversion

Outboundmaterial

management

31

One SKU: forward look

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

4 5 6 7 8

Weeks Safety Stock

Expected number of stock out (per period)

0.00E+00

5.00E+05

1.00E+06

1.50E+06

2.00E+06

2.50E+06

3.00E+06

3.50E+06

Jan-0

0

Feb-0

0

Mar-

00

Apr-

00

May-0

0

Jun-0

0

Jul-00

Aug-0

0

Sep-0

0

Oct-

00

Nov-0

0

Dec-0

0

Jan-0

1

Feb-0

1

Mar-

01

Apr-

01

May-0

1

Jun-0

1

Jul-01

Aug-0

1

Sep-0

1

Oct-

01

Nov-0

1

Dec-0

1

Average InventorySafety StockHigh ConfidenceLow Confidence

99.4

99.6

99.8

100

4 5 6 7 8

Weeks Safety Stock

Customer service level

0.00E+00

5.00E+05

1.00E+06

1.50E+06

2.00E+06

2.50E+06

4 5 6 7 8

Weeks Safety Stock

Average inventory kept

32

A particular market

• End-user demand fairly steady, but• Internal processes create additional dynamics• Need for high safety stocks to buffer• Aim for high service levels – high perceived

opportunity cost• What if demand was more in line with end-use ?

33

Demand profiles

0

100000

200000

300000

400000

500000

600000

700000

Jan-

97

Mar

-97

May

-97

Jul-9

7

Sep

-97

Nov

-97

Jan-

98

Mar

-98

May

-98

Jul-9

8

Sep

-98

Nov

-98

Jan-

99

Mar

-99

May

-99

Jul-9

9

Sep

-99

Nov

-99

Jan-

00

Mar

-00

May

-00

Jul-0

0

Sep

-00

Pa xil2 01 00(Ac tu a l)

(Fitting)Prediction

includes our forecasting algorithms for future forecasts – these pick up key dynamics

34

Current policy: performance

Service Level

90

92

94

96

98

100

2 3 4 5 6 7 8

weeks cover

%

Average stock

200000

300000

400000

500000

600000

700000

800000

2 3 4 5 6 7 8

weeks cover

pack

s

35

Smoothed performance

Service Level

90

92

94

96

98

100

102

2 3 4 5 6 7 8

weeks cover

%

Average stock

200000

300000

400000

500000

600000

700000

800000

2 3 4 5 6 7 8

weeks cover

pack

s

36

Comparison of two supply chain responses

• Pharmaceutical process– primary production has five synthesis stages– two secondary manufacturing sites

• Two different process principles– Case A: QC at the end of each synthesis stage and the final

products– Case B: QC for the product of the primary process (AI), and

the final product

37

Inventory variation for one SKU

Results for Pack C with QC (Monte Carlo 400 simulations)

0

50000

100000

150000

200000

250000

300000

0 20 40 60 80 100

Time / weeks

Pack

C s

tock

leve

l / p

acks

mean quantitysafety stock95% upper confidence95% lower confidence

Base Case for Pack C

0

50000

100000

150000

200000

250000

300000

0 20 40 60 80 100

Time / weeks

Pack

C s

tock

leve

l / p

acks

mean quantitysafety stock95% upper confidence95% lower confidence

Case BCase A

38

SCM more generally…

• Re-emphasise need for better process technology for 10

manufacture• Currently push-based at back-end of supply chain:

– hard to be responsive– complicated dynamics– can’t exploit short-term opportunities (e.g. tenders)

• aim for much faster processes• avoid too many quality control interventions and isolations• aim for easy to clean/reconfigure (disposable?) plants

– Move towards short-term scheduling• Interim:

– optimise campaign planning (cf. literature of 1980s!)– optimise changeovers using SMED concepts

39

SCM more generally (cont’d) …

• Current SCM methodologies involve a degree of decentralisation– Regional demand management– Primary or secondary planning– Subcontractors– Robust, but built-in inefficiencies

• Need for more integrated, seamless planning– Very large-scale multi-site problems with large geographical span– Looser alliances may arise (e.g. semiconductors, computers)

• Increased co-ordination problems

– Tailored optimisation algorithms will be required– Need to build in robustness (cf. work of Maranas and co-workers,

Sahinidis and co-workers)

40

Some conclusions

• The pharmaceutical supply chain is very complex, with many interacting facets– Difficult to generate radical improvements quickly– Piecemeal approaches (e.g. improved logistics) will generate

incremental benefits• Current process technology is one of the main supply chain

bottlenecks– Many “built-in” inefficiencies that constrain performance– Not a very responsive system

• Current models in the research community are too “company-centric”– Future models need to consider a holistic view of an extended

supply chain of specialist agents• IP generators• Testing specialists• Contract manufacturers• Logistics providers• Healthcare providers/consumers

41

Future perspectives

• Industry– at crossroads?

• “Big is beautiful” v. alliances of specialists• Latter will need next generation of supply chain tools

• Products– More complex, more synthesis stages, more chiral, more active,

smaller lot-sizes– Better drug delivery mechanisms

• Smaller dosages– Likely to become more specialised

• Local solutions to local problems• Genetic research leading to target sub-populations• Current manufacturing and supply chain poorly suited to this

– Economies of scale 1-2 orders of magnitude out– Rapid response vaccines (civilian and military)– More crop-derived products – new supply chains

42

Acknowledgements

• Members of Britest project (www.britest.co.uk)• UK Engineering and Physical Sciences Research

Council• Fellow researchers: Gabriel Gatica, Jonatan

Gjerdrum, Wing Yan Hung, Rodney Johnson, Lazaros Papageorgiou, Nouri Samsatli, Mona Sharif, Trevor Tsang