Effects of Regulation on Drug Launch and Pricing

Patricia M. DanzonThe Wharton School

University of Pennsylvaniadanzon@wharton.upenn.edu

Andrew J. EpsteinYale University

andrew.epstein@yale.edu

Characteristics of On-patent Medicines

• High R&D cost patents are essential

• Most medicines are potentially global

• ∑ ∑(Pjt – MCjt) > F t j• Price-MC margin, across countries and patent life, must pay for R&D

• Ramsey pricing optimal cross-national price differentials

• Prices should vary inversely with “true” price elasticity • (Danzon and Towse, 2004)• GDP per capita may be a rough proxy

• Insurance undermines consumer price sensitivity increased prices and volumes

Third party payers act as cost-conscious surrogates

Regulation of Pharmaceuticals

• 1. Registration and Market Access• Proof of safety, efficacy and quality

• 1990s EMEA and US FDA harmonized requirements and accelerated procedures

• Japan has unique requirements

• 2. Price and Reimbursement Regulation• Price approval a condition of reimbursement in most countries

• Country-specific rules are heterogeneous and complex

Price Regulation in Practice: Two Prototypes

• Internal referencing to other products within class/country• Mark-ups for superior efficacy; pharmacoeconomic data may be required

• Japan; France; Sweden; Canada etc. Cost-effectiveness a condition for reimbursement (>1999 UK, Canada)

• Therapeutic reference price reimbursement: RP = max. reimbursement for

all drugs in a therapeutic group• Netherlands (1991); Germany (2005)

• Generic RP within off-patent molecules, widely used but unlikely to affect launch

• External referencing to same product in other countries• P j,n = min/ mean/ median {Pk,n……. Ps,n}

• External benchmark less discretion for regulators

• Netherlands, Japan, Italy, Portugal, Greece, Canada, France• Informally used in many other countries e.g. UK, US

Price Regulation Affects a Firm’s Incentives for Launch of New Drugs

• Direct Effects • Lower prices lower NPV of sales vs. fixed launch cost

• Non-launch most likely in countries/classes with small volume

• Strategic delay

• Firms may hold out for higher price; converse for regulators

• Indirect Effects• Market segmentation and price discrimination constrained

• Optimal strategy: delay or not launch in low price countries that may undermine potentially higher prices in connected markets

• Parallel trade has same effects as external referencing• Limited to traded share of sales

Welfare Effects Differ by Type of Regulation

• Internal referencing • Regulators should weigh benefit of lower prices against reduced

access to new drugs

• Cost and benefits are internalized to the regulating country

• External referencing • Referencing by high-price country to low-price country launch

delay and/or higher prices in low-price country

• Spillover costs in low-price country are not internalized to high-price country

• Undermines Ramsey-optimal price differentials

Study Objectives: Determinants of Launch Delay and Launch Prices

• Effects of regulation• Direct effect from lower prices

• Spillover effects from external referencing and parallel trade

• Strategic incentives to delay launch

• Dynamic (between subclass) competition vs. Static (within subclass) competition

• Local corporation effects• Uniform across countries?

• Launch and/or price?

Launch and Price Conditions

Expected Net RevenueSpillover

CostsFixedCosts

Launch Condition Firm’s Reservation Price (PAsk)

Launch and Price at Launch

0 Asksj

Offersj PP

)(,,,,;),(),({1

HFdteIRPXNNQQCPYNPRPPE jtrt

T

t jkktktsjtjktgjtbjtjtsjtsjtsktjgjtgjtbjtsjt

Reduced Form Equations

Launch

Price at Launch

HPYINNXQPPhh sKtjjtgjtbjtjtjtgjtbjtsjt ;;,,,,,,,

HPYINjXQPPfP sKtjjjjgjbjsj ;;,,,,,,

Data

• IMS quarterly data on drug sales Q1 1992 – Q4 2003• Revenue at manufacturer prices• Volume in IMS standard units (doses)

• 15 Countries: • 4 High-price EU (Germany, UK, Netherlands, Sweden)• 5 Low-price EU (France, Italy, Spain, Portugal, Greece)• 4 High-price non-EU (US, Japan, Canada, Switzerland)• 2 Middle income (Brazil, Mexico)

• 12 Therapeutic classes, all with new “superior” and old “inferior” subclasses• Anti-ulcerants: H2 antagonists vs. PPIs• Anti-depressants: TCAs vs. SSRI/SNRIs

• “Superior”: 116 molecules; 885 launches in our period• “Inferior” : 259 molecules; 390 launches in our period

Estimation (1): Launch Hazard (Quarterly)

ML discrete-time proportional hazard model, based on complementary log-log regression

Accommodates: -Right censoring and late entry into risk set-Time-varying covariates-Robust s.e.’s for intra-molecule clustering

sjtsjt th expexp1

Alternative clog-log estimators:-Split population model: lim(hazard) as time infinity ≠ 1 -Random (molecule) effects: (Normal distribution)

- Accounts for unobserved molecule heterogeneity

Estimation (2): Price at Launch

• Dependent variable: log price per standard unit • Local currencies adjusted to 2003 using local PPI• Converted to US dollars at 2003 exchange rates

• Estimation• OLS with robust, clustered SEs or Normal random effects

• Estimates conditional on launch

• Clog-log 2-step Heckman estimator with consistent SEs• Unconditional estimates, adjusted for launch selection

• In practice, identification mainly off functional form

Focus on OLS results

Mean Launch Delay by Number of Launches: Superior Molecules

GermanyUKNetherlands Sweden

France Greece

Italy

Portugal

Spain

Canada

Japan

Switzerland

USA

Brazil

Mexico

15

20

25

30

35

40

45

40 50 60 70 80

Number of Superior Molecule Launches

Me

an

La

un

ch

De

lay

(M

os

)

Mean 2003 Price by Number of Superior Molecule Launches

Germany

UK

Netherlands

Sweden

France

Greece

ItalyPortugal

Spain

Canada

Japan

Switzerland

USA

Mexico

Brazil0

10

20

30

40

50

60

40 50 60 70 80

Number of Superior Molecule Launches

Me

an

La

un

ch

Pri

ce

$0

$10

$20

$30

$40

$50

$60

GER UKNET

SWE

FRAGRE

ITA

PORSPA

CANJA

PSW

IUSA

BRAM

EX

Country

Pri

ce

Hi-Price EU

Low-Price EU

Hi-Price non-EU

Low-Price non-EU

Mean Price per Molecule in 2003, by Country:Superior Molecules

Superior Subclasses Inferior Subclasses

  CountryOrig/ Licns Generic Par Imp Oth Brand

Orig/ Licns Generic Par Imp Oth Brand

High-Price EU

Germany 30.31 0.49 9.23 0.94 2.56 0.32 0.33 0.75

UK 13.26 0.64 1.36 0.61 6.18 0.93 0.46 1.27

Netherlands 35.53 0.05 2.22 7.75 10.79 0.45 0.61 0.72

Sweden 24.41 0.34 1.79 0.45 13.08 0.26 0.42 0.98

Low-Price EU

France 6.52 0.41   0.89 0.28 0.39   0.75

Greece 12.84 1.11 0.56 6.26 0.45 0.80

Italy 2.24 0.41 0.77 0.30 0.31 0.61

Portugal 1.83 0.44 0.78 0.33 0.20 0.34

Spain 6.13 0.33   1.06 0.28 0.63   0.69

High-Price non-EU

Canada 49.30 0.57   0.60 1.03 0.93   0.57

Japan 12.27 0.44 0.73 2.62 1.26 0.59

Switzerland 34.64 0.81 0.76 5.81 0.37 1.43

USA 52.70 0.90   14.22 11.05 1.43   12.72

Low-Price non-EU

Brazil 2.34 0.39 0.95 4.21 0.21 0.35

Mexico 18.34 0.69   1.65 5.84 0.26   0.53

Mean Price per Molecule, in 2003, by License Status and Subclass

Superior Subclasses Inferior Subclasses

  CountryTot

Mols

Loc Corp Mols

# Oth Lnchs

Tot Mols

Loc Corp Mols

# Oth Lnchs

High-Price EU

Germany 88 15 10.4 131 32 7.7

UK 80 19 13.3 116 21 10.6

Netherlands 73 0 0.0 112 2 12.5

Sweden 77 4 11.3 82 0 0.0

Low-Price EU

France 69 20 11.0 105 26 7.9

Greece 72 0 0.0 107 1 4.0

Italy 76 21 9.9 127 19 5.0

Portugal 62 4 13.5 113 0 0.0

Spain 76 16 11.4 112 8 5.1

High-Price non-EU

Canada 73 2 7.0 98 3 7.7

Japan 53 29 7.5 158 62 3.0

Switzerland 78 16 10.7 111 27 10.7

USA 86 49 10.6 97 40 10.0

Low-Price non-EU

Brazil 71 2 13.5 92 6 6.5

Mexico 72 0 0.0 105 3 4.0

Number of Molecules Launched by a Local Corporation:Originators and Licensees

Launch Model: Competitor Prices and Volume, By Subclass

Marginal Effects from Clog-log with Robust Clustered SEs

VariablesSuperior

SubclassesInferior

Subclasses

Log Avg Price of Superior Brands (Lag 1Q) 0.0047* -0.0001

Log Avg Price of Inferior Brands (Lag 1Q) 0.0018 0.0001

Log Total Volume of All Drugs in Class (Lag 1Q) -0.0033 0.0001

Num Generic Manufs per Molc in Superior Subclass (Lag 1Q) -0.0001 0.0000

Num Generic Manufs per Molc in Inferior Subclass (Lag 1Q) -0.0001 0.0000

Number of Observations 23,400 96,041

Number of Molecules 111 239

Mean Launch Hazard per Quarter 0.0378 0.0041

Marginal Effects from Clog-log with Robust Clustered SEs

VariablesSuperior

SubclassesInferior

Subclasses

Num Already Launched (UK, Ger) 0.0270*** 0.0006

Num Already Launched (Swe, Net) 0.0231*** 0.0006*

Num Already Launched (Ita, Fra) 0.0120** 0.0001

Num Already Launched (Spa, Por, Gre) 0.0022 0.0002**

Num Already Launched (Can, Jap, Swi, USA) 0.0086*** 0.0000

Mean Launch Hazard per Quarter 0.0378 0.0041

Launch Model: Indirect Regulatory Effects--Cross-National Spillovers

Marginal Effects of Prior Foreign Launch in Low Price EU Countries

Clog-log with Robust Clustered SEs

Prior Launch in a High-Price EU Country 0.0018***

Prior Launch in a Low-Price EU Country 0.0005***

Difference -0.0013***

Prior Launch in a High-Price EU Country 0.0018***

Prior Launch in a High-Price non-EU Country 0.0005***

Difference -0.0012***

Spillover Effects of Prior Foreign Launch on Launch Hazard in Low-Price EU Countries:

Superior Subclasses

Marginal Effects from Clog-log with Robust Clustered SEs

VariablesSuperior

SubclassesInferior

Subclasses

Time Since Global Launch (Yrs) -0.0273*** -0.0003**

Time Since Global Launch Squared (Yrs) 0.0010*** 0.0000**

Any PI Share in Subclass D.V. 0.0011 -0.0003

Launch by Local Corporation D.V. 0.0443*** 0.0043*

Mean Launch Hazard per Quarter 0.0378 0.0041

Launch Model: Effects of Delay, PI Share and Local Corporation

Superior Subclasses Inferior Subclasses

  Country Marg Eff Qtr Haz Marg Eff Qtr Haz

High-Price EU

Germany 0.0070 0.0630 0.0033 0.0034

UK 0.0340 0.0438 0.0034* 0.0039

Sweden -0.0106 0.0436 N/A 0.0025

Netherlands N/A 0.0370 0.0157 0.0034

Low-Price EU

France 0.0145** 0.0315 0.0010 0.0029

Greece N/A 0.0344 0.0032 0.0055

Italy 0.0148** 0.0446 0.0052* 0.0047

Spain 0.0143** 0.0433 0.0113 0.0037

Portugal 0.0001 0.0237 N/A 0.0041

High-Price non-EU

Canada 0.0059 0.0394 0.0008 0.0032

Japan 0.0108*** 0.0184 0.0085* 0.0096

Switzerland 0.0217** 0.0444 0.0004 0.0033

USA 0.0118 0.0542 0.0010* 0.0033

Low-Price non-EU

Brazil 0.0045 0.0373 0.0045 0.0052

Mexico N/A 0.0353 0.0013 0.0041

Marginal Effects of Launch by a Local Corporation on Launch Hazard

Launch Model: Country Fixed Effects for Superior Molecules

(Inferiors Not Significant)

GER is the referent

UK

NET

SWE

FRA GRE

ITA

POR

SPA

CAN

JAP

SWI

USA

BRAMEX

-0.045

-0.040

-0.035

-0.030

-0.025

-0.020

-0.015

-0.010

-0.005

0.000 U K N E T S W E F R A G R E IT A P O R S P A C A N JA P S W I U S A B R A M E X

Country

Clo

g-l

og

Co

un

try

Co

eff

icie

nt

Hi-Price EU

Low-Price EU Hi-Price non-EU

Low-Price non-EU

Normal Random Effects

VariablesSuperior

SubclassesInferior

Subclasses

Log Avg Price of Superior Brands (Lag 1Q) 0.1220*** -0.0064

Log Avg Price of Inferior Brands (Lag 1Q) 0.0800*** 0.1733***

Log Avg Price of Generics in Class (Lag 1Q) 0.0155 -0.0852**

Log GDP per Capita 1.8249** 3.3683***

USD to (ECU or Euro) Exchange Rate -0.0282 1.8420**

Country-Specific Quarterly Producer Price Index -0.0083* -0.0123*

Observations 950 423

Number of Molecules 109 123

R-squared 0.87 0.76

Mean of Dependent Variable (Log of Price) 0.74 -0.39

Price Model: Competitor Prices, Income and Inflation Effects

Normal Random Effects

VariablesSuperior

SubclassesInferior

Subclasses

Log Min Own Price in Hi-Price EU (Lag 1Q) 0.1008*** 0.2740***

Log Min Own Price in Low-Price EU (Lag 1Q) -0.0229 -0.1007

Log Min Own Price in Hi-Price non-EU (Lag 1Q) 0.1366*** -0.0270

Any PI Share in Subclass D.V. 0.0162 -0.4746***

Launch by Local Corporation D.V. 0.0000 -0.0977

Mean of Dependent Variable (Log of Price) 0.74 -0.39

Price Model: Cross-national Spillovers, PI Share and Local Corporation Effects

Marginal Effects of Log Min Own Price on Launch Price in Low-Price EU Countries

 OLS w/ Robust Clustered SEs

Normal Random Effects

Own Price in High-Price EU Countries 0.4274*** 0.2496***

Own Price in Low-Price EU Countries 0.0346 -0.0129

Difference 0.3928*** 0.2625**

Own Price in High-Price EU Countries 0.4274*** 0.2496***

Own Price in High-Price non-EU Countries 0.0526 -0.0034

Difference 0.3748** 0.2530*

Spillover of Foreign Own Price on Launch Price in Low-Price EU Countries:

Superior Subclasses

Normal Random Effects

VariablesSuperior

SubclassesInferior

Subclasses

Time Since Global Launch (Yrs) -0.0455* -0.0032

Time Since Global Launch Squared (Yrs) 0.0008 -0.0013**

First Brand Launch in Ctry-Subclass D.V. 0.1970 1.1693***

Second Brand Launch in Ctry-Subclass D.V. 0.2734*** 0.6104***

Third or Fourth Brand Launch in Ctry-Subclass D.V. 0.1832*** 0.2095

Mean of Dependent Variable (Log of Price) 0.74 -0.39

Price Model: Launch Timing and Order of Entry

Normal Random Effects

VariablesSuperior

SubclassesInferior

Subclasses

Avg Pack Size (Up to 100) -0.0092*** -0.0094***

Pack Size > 100 D.V. -0.9742*** -1.4205***

Avg Pill Strength (g) 0.0927 0.1731

Form: Oral Solid Delayed D.V. 0.0990 0.0950

Form: Injectable D.V. 1.7522*** 1.9598***

Form: Other 0.0987 0.0906

Mean of Dependent Variable 0.74 -0.39

Price Model: Product Characteristics

Price Model: Country Fixed Effects -- Superior Molecules

GER is the referent

-1

-0.5

0

0.5

1

1.5

2

2.5

3

3.5

Country

Ran

do

m E

ffec

ts C

ou

ntr

y C

oef

fici

ent

With GDP per Capita Without GDP per Capita

Hi-Price EU

Low-Price EU

Hi-Price non-EU

Low-Price non-EU

Conclusions: 1. Regulation

• Direct effect of price regulation is confirmed• Launch delay and non-launch more likely where expected

launch prices are low

• Indirect, spillover effects of regulation are significant• Spillover effects are greatest from high-price EU to low-price EU

countries, in both launch and price

• Referencing from high-price to low-price countries in the EU contributes to launch lags in these low price countries

• Launch prices in low-price countries are also influenced by higher foreign prices

• Parallel importation had smaller effects• Plausible, but power may be weak• Our data do not permit identification of PI source

Conclusions: 2. Strategic Bargaining Delay?

• No evidence that firms can obtain a higher price by delaying launch

• Launch hazards decline with time from global launch

• Launch prices (real) decline with time from global launch

• But being first is not necessarily best

• Second and third entrants in superior subclasses get higher real prices on average than first entrant

Conclusions: 3. Dynamic vs. Static Competition

• Launch lags and launch prices are related to prices of competitor products in the same subclass

• Smaller cross-subclass effects

• Other evidence supports dynamic competition

• Late entrants in older subclasses diffuse less broadly

• At lower prices, linked to prices in older subclasses

• Generic prices not significant on average• Country-specific differences not tested

Conclusions: 4. Local Corporation Effects

• Launch by a local corporation increases launch hazard only in certain countries

• Italy, Spain, France, Japan, Switzerland for superior drugs

• Local corporation does not significantly affect price

• Suggests local corporation benefit is in registration, not pricing

• Future work to distinguish originator vs. licensee effects

Conclusion: 5. Country Effects

• Country effects are significant, after controlling for expected prices• Reflect all unmeasured country-specific effects

• Within EU, potential parallel export countries have slower launch• Portugal, Italy, France, Spain, and Greece• But not consistently more negative than some other countries

• Price effects are highly sensitive to control for GDP per capita• Controlling for GDP, low-price countries have relatively high prices • Controlling for GDP, high-price EU countries have relatively low prices

• External referencing from high- to low-price countries contributes to launch lags and higher prices in low-price countries

• If US adopts external referencing or parallel trade, spillover effects to other countries could be larger than within-EU effects here

Normal Random Effects

VariablesSuperior

SubclassesInferior

Subclasses

UK D.V. -0.2741*** -0.1424

Netherlands D.V. -0.0572 -0.0285

Sweden D.V. -0.3271** -0.3940*

France D.V. -0.1218 -0.6705***

Greece D.V. 1.1963** 1.7931**

Italy D.V. 0.0869 -0.0369

Portugal D.V. 1.2180** 2.0567**

Spain D.V. 0.7716** 1.0792*

Canada D.V. 0.0386 -0.2199

Japan D.V. -0.3499 -1.1783*

Switzerland D.V. -0.5989* -1.6100***

United States D.V. -0.3386 -1.0889**

Brazil D.V. 3.1688** 5.7499***

Mexico D.V. 2.3304** 3.7571**

Price Model: Country Fixed Effects

Superior Subclasses Inferior Subclasses

  Country Mols Lnchs

Mean

Delay

Medn

Delay Mols Lnchs

Mean

Delay

Medn

Delay

High-Price EU

Germany 88 72 18.5 9.5 131 18 30.4 17.5

UK 80 58 18.7 6.5 116 24 69.2 41.5

Netherlands 73 56 18.1 10 112 21 43.7 15

Sweden 77 62 17.4 7 82 19 49.9 18

Low-Price EU

France 69 53 30.9 29 105 19 87.6 59

Greece 72 55 30.1 22 107 37 116.8 54

Italy 76 61 24.8 21 127 26 74.7 48.5

Portugal 62 48 37.0 33.5 113 26 85.8 67

Spain 76 62 28.1 21 112 23 43.9 31

High-Price non-EU

Canada 73 62 25.6 16 98 22 91.9 66.5

Japan 53 42 41.0 40 158 43 63.3 28

Switzerland 78 63 23.9 18 111 21 55.5 47

USA 86 72 17.9 8 97 23 86.4 62

Low-Price non-EU

Brazil 71 60 31.2 20.5 92 40 107.1 90.5

Mexico 72 59 28.6 17 105 28 84.0 55

Launch and Molecule Count, and Mean and Median Launch Delay by Country and Subclass

Marginal Effects from Clog-log with Robust Clustered SEs

Variables Superior Subclasses Inferior Subclasses

UK D.V. -0.0069 0.0002

Netherlands D.V. -0.0253*** 0.0006

Sweden D.V. -0.0200** -0.0003

France D.V. -0.0320*** -0.0006

Greece D.V. -0.0319*** 0.0008

Italy D.V. -0.0283*** -0.0001

Portugal D.V. -0.0382*** -0.0002

Spain D.V. -0.0252*** -0.0001

Canada D.V. -0.0283*** -0.0005

Japan D.V. -0.0408*** -0.0002

Switzerland D.V. -0.0287*** -0.0004

USA D.V. -0.0211** -0.0007

Brazil D.V. -0.0304*** 0.0001

Mexico D.V. -0.0321*** -0.0001

Launch Model: Country Fixed Effects