`
`‘5
`
`Medical Care Output
`and Productivity
`
`Edited by
`
`David M. Cutler and
`Ernst R. Berndt
`
`n-ch
`
`md Wealth
`
`'fiie University of Chicago Press
`
`Chicago and London
`
`EARUCH couaGE LIBRARY
`
`IMMUNOGEN 2281, pg. 1
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 1
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`M
`
`The University of Chicago Press, Chicago 60637
`The University of Chicago Press. Ltd. London.
`© 2001 by the National Bureau of Economic Research
`All rights reserved. Published 2001
`Printed in the United States of America
`
`10090807060504030201
`
`12345
`
`ISBN: 0-226s13226-9 (cloth)
`
`Copyright is not claimed for the “Comment” on chap. 4 by Brent R.
`Moulton; chap. 5 by Ina Kay Ford and Daniel H. Ginsburg; chap. 6
`by Dennis Fixler and Mitchell Ginsburg; chap. 9 by Paul Heidenreich
`and Mark McClellan; and chap. 10 by Irving Shapiro, Matthew D.
`Shapiro, and David W. Wilcox
`
`DAVID M. CUTLER is professor of economics at Harvard University
`and a research associate of the National Bureau of Economic
`Research. ERNST R. BERNDT is professor of applied economics at
`the Sloan School of Management of the Massachusetts Institute of
`Technology and director of the NBER’s Program on Technological
`Progress and Productivity Measurement.
`
`'
`
`Library of Congress Cataloging-in-Publication Data
`
`Medical care output and productivity / edited by David M. Cutler
`and Ernst R. Bemdt.
`
`p. cm.—(Studies in income and wealth ; v. 62)
`Includes bibliographical references and index.
`ISBN 0—226-13226-9 (cloth : alk. paper)
`1. Medical care—~Cost effectiveness—Econometric models——
`Congresses.
`2. Medical care, Cost oil—«Congresses.
`I. Cutler.
`David M.
`II. Berndt, Ernst R.
`111. Series.
`
`RA410.5 .M425 2001
`338,4’33621—dc21
`
`00-067235
`
`The paper used in this publication meets the minimum
`requirements of the American National Standard for Information
`Sciences—Permanence of Paper for Printed Library Materials,
`ANSI Z39.48-l992.
`
`IMMUNOGEN 2281, pg. 2
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 2
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`
`
`ii"
`if,
`
`3:5
`
`Iain M. Cockburn and Aslam H. Anis
`
`Hedonic Analysis of
`Arthritis Drugs
`
`W 1
`
`Introduction
`
`1.1
`
`This study examines the market for a group of drugs used to treat rheu-
`matoid arthritis (RA) during the period 1980—92. Rheumatoid arthritis is
`a painful, debilitating, and progressive disease which affects millions of
`people worldwide, with very substantial effects on health and the economy.
`Regrettably, in contrast to some other major health problems such as heart
`disease, depression, ulcers, and bacterial infections, this is an area where
`therapeutic innovations have thus far had comparatively little impact on
`physicians’ ability to reverse the disease. RA currently has no “cure” and
`the effectiveness of available treatments is limited. Compared to other
`drug classes the rate of new product introductions has been slow, and, at
`the time of writing, there have been no breakthroughs of the same order
`of significance as the discovery and development of SSRIs for treatment of
`depression, H2 antagonists for ulcers, or ACE inhibitors for hypertension.
`Nonetheless, the market for RA drugs is far from static. There have
`been significant changes over the past fifteen years in the market shares of
`competing products. Interestingly, relative prices have changed relatively
`
`Iain M. Cockburn is professor of finance and economics at Boston University and a re-
`search associate of the National Bureau of Economic Research. Aslam H. Anis is associate
`professor of health economics in the Department of Health Care and Epidemiology, Univer-
`sity of British Columbia, and team leader of health economics at the Center for Health
`Evaluation and Outcome Sciences, St. Paul’s Hospital. Vancouver.
`The authors thank Ernst Berndt, Zvi Griliches, John Esdaile, and NBER seminar partici-
`pants for helpful comments, and Jennifer Anderson and David Eelwn for access to their
`databases on safety/efficacy profiles. The authors are g.ateful to BC Yharmacare for access
`to claims data; BEA, NBER, and Eli Lilly for financial support; and Merck for access to
`library records. Sophia Wang provided invaluable and very competent research assistance,
`The authors take full responsibility for any remaining errors.
`
`
`
`
`
`IMMUNOGEN 2281, pg. 3
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 3
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`440
`
`Iain M. Cockburn and Aslam H. Anis
`
`little, and these market dynamics appear to be driven primarily by other
`factors. Here we focus on the role played by publication of clinical research
`findings. In contrast to traditional hedonic analysis where product charac-
`teristics are fixed but new products incorporating different quality levels
`appear over time, here the set of products is fixed while their measured
`quality changes over time. New information about the relative efficacy and
`toxicity of existing drugs accumulates through the publication of clinical
`trial results, and this information appears to have had a significant impact
`on the pattern of drug use.
`A number of clinical aspects of rheumatoid arthritis are important
`structural features of the market for drugs used to treat the disease. We
`therefore begin with a brief review of the nature of RA and its treatment.
`We then discuss issues related to the measurement of the relative efficacy
`and toxicity of drug treatments for RA. Next, we present economic data
`on the market for a specific set of drugs used in the treatment of severe RA
`and consider them in the context of models of demand for differentiated
`
`products. We then report the results of estimating price and market share
`equations. In the concluding section, we suggest alternative approaches
`that may provide some additional insight, in particular analysis of the role
`of advertising and promotional expenditures.
`
`11.2 Rheumatoid Arthritis
`
`_.,,.
`
`RA is one of the most prevalent diseases affecting joints and connective
`tissue. RA is an autoimmune disease: For reasons that are still poorly un-
`derstood,
`the body’s immune system begins to malfunction, attacking
`healthy tissue. Like related conditions such as lupus erythematosus, psori-
`atic arthritis, and scleroderma, the disease is systemic and chronic. Tissues
`are affected throughout the body, and although some patients experience
`prolonged periods of remission, most are aflected for a lifetime.‘
`RA is characterized by inflammation of the synovium (a membrane
`which lines the joints) resulting in stiffness, pain, warmth, and swelling in
`joints. As the disease progresses, inflamed cells release an enzyme which
`erodes surrounding bone and cartilage, resulting in increased pain, loss of
`movement, and eventually destruction of the joint.2 Patients experience
`greater and greater pain and loss of mobility. Fatigue often accompanies
`the “classical” joint symptoms. In late stages of the disease, skin and vas-
`cular problems (such as leg ulcers) may develop, along with damage to
`eyes and nerves and inflammation of lymph nodes, heart, and lungs,‘
`i. ;.
`
`1. Brewerton (1994) gives a comprehensive and readable overview of arthritis and its treat-
`ment. See also Cash and Klippel (1994), Wolfe (1990) and Steinman (1993).
`2. Establishing a conclusive diagnosis of RA can be difficult, especially in its early stages,
`since it shares many symptoms with other autoimmune diseases. Note that RA should not
`be confused with osteoarthritis, an even more prevalent disease, which has a distinct clinical
`. . ,. W e . a”- A...
`1"“(31 ~ mi
`firm-{~19 firm)”:
`
`IMMUNOGEN 2281, pg. 4
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 4
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`
`
`m
`
`Hedonic Analysis of Arthritis Drugs
`
`441
`
`1 primarily by other
`‘1 ofclinical research
`[ere product charac-
`Terent quality levels
`hile their measured
`relative efficacy and
`blication of clinical
`a significant impact
`
`ritis are important
`eat the disease. We
`\ and its treatment.
`the relative efficacy
`sent economic data
`tment of severe RA
`(:1 for differentiated
`e and market share
`rnative approaches
`analysis of the role
`
`ms and connective
`
`are still poorly un-
`unction. attacking
`'thematosus, psori-
`nd chronic. Tissues
`)atients experience
`lifetime.I
`
`ium (a membrane
`th, and swelling in
`an enzyme which
`eased pain, loss of
`atients experience
`)ften accompanies
`:ase, skin and vas—
`g with damage to
`t, and lungs.
`
`'arthritis and its treat-
`1993).
`ally in its early stages.
`c that RA should not
`has a distinct clinical
`
`
`US. economy approx1mately $65 blllion per year in direct expenses and
`lost output.
`
`11.2.1 Treatment Options for Rheumatoid Arthritis
`Over the course of the disease, medical treatment of RA patients cone.
`
`
`
`IMMUNOGEN 2281, pg. 5
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 5
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

` 442 Iain M. Cockburn and Aslam H. Anis
`
`HM
`
`
`
`‘
`.
`4‘
`
`
`
`g
`5
`
`chotherapeutic intervention may also play an important role in helping
`patients cope with the impact of the disease, and many patients also turn
`to “alternative” medicine. Physical intervention takes the form ofphysical
`therapy directed toward preservation ofjoint function and surgical procev
`dures to address severe pathologies of specific joints (cg, hip replace-
`ment). Drug treatment, the focus of this study, is given to almost all pa-
`tients who consult a physician: Of the approximately 5.1 million patient
`visits per year in the United States where RA is a primary diagnosis, more
`than 90 percent involved one or more drugs’ being prescribed.
`Two principal classes of drugs are used to treat RA: nonsteroidal anti—
`inflammatory drugs (NSAIDS) and disease-modifying antirheumatic
`drugs (DMARDs). These two classes account for more than 65 percent of
`all prescriptions to RA patients, with corticosteroids accounting for a fur-
`ther 19 percent. It is important to note that drug therapy for RA normally
`follows a treatment hierarchy: Drug treatment begins with NSAIDs and
`moves on to DMARDs as the disease progresses.
`NSAIDs are the most frequently prescribed drugs for RA. Large num-
`bers of drugs fall into the NSAID class; among the most commonly used
`are aspirin, ibuprofen (Motrin), naproxen (Naprosyn), diclofenac (Vol-
`taren), and piroxicam (Feldene). NSAIDs reduce inflammation and have
`an analgesic effect but do not affect progression of the disease. NSAIDs
`act quickly and are well tolerated by many patients but can cause a number
`of dangerous side effects, particularly when used in the high dosages indi-
`cated for RA. Gastrointestinal bleeding is the most frequently encountered
`severe side eflect.‘ While NSAIDs are the first line of defense, they offer
`only palliative treatment of symptoms, and as the disease progresses pa-
`tients will typically be given one of the DMARDs. This does not usually
`imply discontinuation of NSAID therapy, and in fact between 80 and 90
`percent of patients are prescribed drugs from both classes.
`The second-line DMARDs can suppress symptoms and slow the prog-
`ress of the disease, though they cannot halt it. DMARDs are slow acting,
`taking weeks or months before any significant improvement is noticed by
`the patient, and are often poorly tolerated. Different drugs are used with
`varying degrees of success in different patients. Furthermore, many pa-
`tients are forced to discontinue the drug because of serious side effects.
`Minor, though uncomfortable, side effects such as dermatitis, nausea, and
`mouth ulcers are quite frequently experienced. The incidence of serious
`side efiects such as retinal damage, renal failure, liver damage, and reduc-
`tion in blood cell counts, while uncommon, nonetheless requires close
`medical supervision and frequent diagnostic testing.
`
`4. COX-2 inhibitors. 21 new class of NSAIDS with a more selective mechanism of action
`and mm er .zic;denee ot Side eiiects, have recently been introduced into the 0.5. market. A we
`drugs include celecoxib (Celebrex) and rofecoxib (Vioxx).
`
`.
`
`-
`
`l
`
`
`
`IMMUNOGEN 2281, pg. 6
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 6
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`a
`
`nt role in helping
`patients also turn
`e form of physical
`nd surgical proce-
`(e.g., hip replace—
`to almost all pa-
`.l million patient
`'y diagnosis, more
`:ribed.
`
`nonsteroidal anti~
`ig antirheumatic
`han 65 percent of
`ounting for a fur-
`for RA normally
`‘ith NSAIDs and
`
`RA. Large num-
`: commonly used
`diclofenac (Vol-
`mation and have
`iisease. NSAIDs
`1 cause a number
`
`igh dosages indi-
`ntly encountered
`:fense, they offer
`e progresses pa-
`does not usually
`:ween 80 and 90
`s.
`
`:1 slow the prog-
`are slow acting,
`:nt is noticed by
`gs are used with
`
`nore, many pa-
`ous side effects.
`
`.tis, nausea, and
`lence of serious
`age, and reduc-
`; requires close
`
`echanism of action
`US. market. These
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Hedonic Analysis of Arthritis Drugs
`
`443
`
`Table 11.1
`
`DMARD Drugs
`
`Drug
`Auranofin
`Azathioprine
`Gold sodium
`thiomalate
`Aurothioglucose
`Hydroxychloroquine
`methotrexate
`D-penicillamine
`Sulfasalazine
`
`Brand
`Name(s)
`Ridaura
`Imuran
`Myochrysine
`Solganal
`Plaquenil
`Rheumatrex
`Cuprimine
`Azulfidine
`
`US. Market
`Intro
`1985
`1968
`< 1980
`1989?
`1956
`1955
`1963
`1952
`
`Other Indications
`
`Immune suppression
`for transplants
`
`Malaria
`Leukemia, psoriasis
`Chelation
`Ulcerative colitis,
`Crohn’s disease
`
`Manufacturer
`SKB
`Glaxo
`Wellcome
`Merck
`Schering
`Winthrop
`Lederle,
`generics
`Merck,
`Wallace
`Kabi,
`generics
`
`The DMARDs approved for treatm
`ent of RA during the period of this
`study are listed in table 11.1. One
`point to note from this table is that many
`of these drugs are quite old, having been first introduced to the market
`many years ago. Auranofin (Ridaura) was the only strictly new molecule
`approved for RA in the period covered by this study. Other products such
`as methotrexate are new to the market in the sense that they have recently
`gained regulatory approval for treatment of RA, though they may have
`been approved for other indications for many years or may have been used
`informally or in research settings for treatment of RA. (Lederle introduced
`Rheumatrex, a formulation of methotrexate specifically targeted at the RA
`
`'
`
`'
`
`off-patent, but generic production is
`is also important to note that the ori
`of the drugs was not RA. With the 6
`xception of the gold compounds, the
`activity of the DMARDs against R
`A was discovered subsequent to their
`first introduction to the market. M
`ethotrexate was an early treatment for
`cancer, while hydroxychloroquine
`was developed as an antimalarial. The
`precise mechanism of action of mo
`st of these drugs is not well understood,
`though most have their therapeuti
`c efiect through suppressing the immune
`activity of gold compounds appears to
`itic conditions, while the immunosuppressant activity .1
`of azathioprine.and methotrexate is much more general.
`in addition to DMARDs, physicians may also prescribe corticosteroids.
`This occurs in about 20 percent of patient visits in the United States. While
`these drugs can often produce dramatic short-term improvement in symp-
`toms, their long-term use is limited by serious side efl'ects, principally os-
`teoporosis and increased susceptibility to infections. As a last resort,'pa-
`
`-a..._.-.-_.(a..-.i..
`
`
`
`
`
`
`
`
`
`
`
`
`IMMUNOGEN 2281, pg. 7
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 7
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`.4
`
`5
`
`Iain M. Cockburn and Aslam H. Anis
`444
`“Kb
`
`tients may also be prescribed highly toxic third-line immunosuppressant
`drugs such as cyclophosphamide, cyclosporine, or chlorambucil. Without
`a new therapy which induces a lasting remission, physicians face difficult
`decisions and trade-‘ofl‘s in drug therapy for RA.5
`The timing of moving a patient from well-tolerated NSAIDs to the more
`toxic DMARDs is controversial, with some physicians arguing for early
`and aggressive second-line therapy to preempt irreversible joint damage,
`despite serious side effects. Even within the DMARD class it is far from
`clear which drug to prescribe. Only a fraction of patients obtain significant
`benefit from any one agent and even then the elfect is often short-lived,
`typically lasting for only a few months or years. Over the twenty- to thirty-
`year course of the disease, a patient will typically cycle through a series of
`therapeutic alternatives as their physician attempts to arrest, or often
`merely to minimize, the cumulative destruction wrought by the disease.
`Furthermore, professional opinion has changed over time regarding which
`drugs to use, and when. The information base on the relative efficacy and
`toxicity of these agents continues to evolve as new scientific evidence from
`clinical trials is published and physicians individually and collectively ac—
`cumulate more experience. The efficacy/toxicity trade-off lies at the heart
`of the prescribing decision, and changing perceptions of where drugs are
`located in this space drives our analysis of demand for these drugs.
`
`11.3 Measuring the “Quality” of Drug Treatments for RA
`We attempt to measure the characteristics of different DMARD drugs
`in two general dimensions: efficacy and toxicity Unlike some previous
`work on hedonics of pharmaceutical products we pay little attention to
`differences in the dosage regimen. Though characteristics such as the num-
`ber of times a day the patient must take the drug appear to be an impor-
`tant determinant of the relative value of different ulcer drugs and anti—
`depressants (see Suslow 1992, 1996 and Berndt, Cockburn, and Griliches
`1997), we believe them to be much less important here. The very close
`involvement of the physician and the severe nature of the disease suggest
`to us that the impact of dosing regimens on patient compliance is unlikely
`to be an important factor.6
`
`menu—.nwyu-”er
`
`,nerv'erh.
`
`brel), a genetically engineered protein, was approved by the FDA for treatment of RA in
`late 1998, but of most of these “large molecule" drugs are still in the early stages of testing.
`See Wall Street Journal. l7 July 1997, Bl.
`6. As a practical matter, dosage regimens for these drugs vary widely, are difficult to com—
`pare directly, and often involve complicated “ramp-up" schedules paced over man).r weeks.
`For example, the maintenance dose of methotrexate is 7.5 mg spread over a week, while
`
`is
`
`IMMUNOGEN 2281, pg. 8
`Phigenix v. Immunogen
`|PR2014-00676
`
`
`
`1
`
`IMMUNOGEN 2281, pg. 8
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`
`
`fix
`
`me immunosuppressant
`chlorambucil. Without
`)hysicians face difficult
`
`d NSAIDs to the more
`:ians arguing for early
`versible joint damage,
`{D class it is far from
`ents obtain significant
`:t is often short-lived,
`' the twenty- to thirty-
`:le through a series of
`‘s to arrest, or often
`Jught by the disease.
`time regarding which
`: relative efficacy and
`.entific evidence from
`y and collectively ac—
`e-off lies at the heart
`5 of where drugs are
`)r these drugs.
`
`pr RA
`
`ent DMARD drugs
`ilike some previous
`1y little attention to
`ics such as the num—
`ear to be an impor-
`:er drugs and anti-
`burn, and Griliches
`are. The very close
`the disease suggest
`npliance is unlikely
`
`is a new DMARD with
`. number of experimen—
`promise for significant
`orders. Infliximab (En-
`or treatment of RA in
`early stages of testing.
`
`iy, are difficult to com—
`ced over many weeks
`id over a week, while
`
`Random assignment to treatment groups
`Blinded trial (at a minimum single-blinded)
`Appropriate minimum dosage levels
`At least eight weeks’ duration
`
`Imposing these criteria res
`trials being excluded.
`
`ulted in all but 66 of the original set of published
`
`11.3‘1 Efficacy Measures
`
`
`
`IMMUNOGEN 2281, pg. 9
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 9
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`
`
`
`
`Iain M. Cockburn and Aslam H. Anis
`446
`ME
`
`age improvement over baseline or the mean improvement standardized
`by baseline standard deviation
`
`Apart from these measurements, efficacy can also be measured by the re-
`ported rate at which patients dropped out of each trial due to “lack of
`efficacy.”
`
`11.3.2 Toxicity Measures
`
`Toxicity is much harder to measure consistently. We have not been able
`to assemble consistent data on the actual incidence of side effects in each
`trial. Following previous work we have experimented with variables con-
`structed by counting the number of side effects listed under categories such
`as “severe” or “frequent” in standard reference sources, or constructing
`dummy variables reflecting the locus of specific side effects (kidney dam-
`age, central nervous system, retina, etc.) but these perform poorly in exper-
`imental regressions.7 Our preferred measure of toxicity is the reported rate
`at which patients dropped out of clinical trials due to “toxicity" Summary
`statistics for these variables are given in table 11.2.
`
`11.3.3 Changes in Quality over Time
`Because new trials are conducted periodically, information accumulates
`steadily over time, and variables constructed from reported trial results
`form a longitudinal data set. We combine data from different trials in a
`variety of ways intended to capture the evolution over time of the scientific
`information available to prescribing physicians.
`One possibility is to simply assign a value to each variable in each year
`based on the most recently published study. Thus we “ratchet” the level of
`each variable up or down in each year that a new trial came out, and carry
`forward the previous value otherwise. (In tables below we refer to these
`measures as “latest.”)
`A second approach is to do a “rolling” cumulative meta-analysis which
`pools treatment groups over time and across drugs. As new trials are pub-
`lished results for each group of patients are added to the previous total,
`resulting in a continuously expanding sample. Mean treatment effects are
`the weighted sum of treatment effects in all trials to date.
`Third, we modify the cumulative meta-analysis by imposing various
`schemes of declining weights over time to capture “depreciation” of
`knowledge. We expect the results of trials conducted many years in the
`past to weigh less heavily upon current prescribing practice than more
`recent evidence. The simplest such weighting scheme is a three- or five-
`year moving average. Alternatives such as a perpetual inventory deprecia-
`
`7. Clinicians may be most strongly influenced by the relative incidence of severe adverse
`reactions. We have not yet compiled data on these effects. But note that since these events
`are very rare, their probability of occurrence is difficult to measure precisely.
`
`IMMUNOGEN 2281, pg. 10
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 10
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`
`
`
`
`M S
`
` Hedonic Analysis of Arthritis Drugs
`447
`—————————__._______—______
`
`
`Table 11.2
`
`Summary Statistics and Characteristics of DMARDs
`
`
`ample MeansM.
`
`
` Efficacy
`
`
`
`
`
`
`
`
`
`Toxicity
`
`TJC GS ESR DropoutDrug Name and Daily Dose Price“W
`
`
`
`
`
`t standardized
`
`Ired by the re-
`ie to “lack of
`
`not been able
`effects in each
`variables con-
`
`ategories such
`.' constructing
`(kidney dam-
`>orly in exper-
`r reported rate
`ity." Summary
`
`'1 accumulates
`d trial results
`ent trials in a
`fthe scientific
`
`e in each year
`t" the level of
`
`3m. and carry
`refer to these
`
`.nalysis which
`:rials are pub-
`nrevious total.
`:nt effects are
`
`osing various
`ireciation" of
`
`I years in the
`:e than more
`:hree- or five-
`
`:ory deprecia—
`
`31" severe adverse
`ince these events
`:y.
`
`
`
`
`
`
`
`garmaasuaaa.
`
`1.91
`0.16
`10.79
`26.98
`8.44
`auranofin, 6mg
`1.67
`0.27
`13.73
`33.11
`9.78
`Azathioprine, 100mg
`1.07
`0.40
`1079
`38.20
`9.15
`Gold salts, 7mg
`1.42
`0.04
`I 1.41
`39.89
`9.21
`Aitimalarials, 400mg
`1.31
`0.16
`13.49
`33.11
`13.23
`Methotrexate, 125mg
`1.70
`0.33
`22.65
`37.26
`878
`D-penicillamine, 600mg
`0.84
`0.37
`20.64
`28.53
`12.28
`Sulfasalazine. 2.5g
`
`
`
`
`
`4.80 9.74 1.26 0.07Placebo. n.a. n.a.W
`Note: n.a. = not available.
`“1992 US. dollars per daily maintenance dose.
`
`tion scheme or fixed declining weights do not yield materially different re-
`sults.
`
`11.4 Model
`
`The theoretical literature provides little guidance on the appropriate
`functional form for estimating quality—adjusted prices. Following many
`previous hedonic pricing studies (for pharmaceutical products, see Sus-
`low's analysis of ulcer drugs [1996] or Berndt, Cockburn, and Griliches’s
`work on antidepressants [1997]) we use a semilog reduced form:
`
`l 11
`
`(1)
`
`ln(pfl) = (it/[[3 + Zy {- 8],.
`
`where x], represents the measured quality (i.e., toxicity and efficacy) char-
`acteristics of drug j;j = 1,
`.
`.
`.
`, J at time t; Z, is a set of time dummies;
`and pj, denotes the time series of prices for drug j.
`For the market share equation, we follow Berry (1994) and Berry, Levin-
`sohn, and Pakes (1995) in specifying a logit type discrete choice model of
`demand for differentiated products to analyze the DMARD market. See
`King (1996) for a successful application of a modification of this approach
`to the anti-ulcer market. Following Berry we postulate that the utility of
`consumer i for product j is given by the function U(xj, g, pl, @4: v), where
`xj, g], p],
`(9d are observed product characteristics, unobserved product
`characteristics, and price and demand parameters, respectively. The term
`v, is unobserved by the econometrician and represents a consumer-specific
`component of utility. To implement the model, one has to make specific
`parametric assumptions about the consumer-specific variables, analogous
`to the choice of functional form for a_ homogenous good demand equation.
`The utility derived by consumer i for product j can be written as
`
`
`
`IMMUNOGEN 2281, pg. 11
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 11
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`
`
`448 Iain M. Cockburn and Aslam H. Anis
`
`(2)
`
`“1/ = x013, — up]. +§1j + 8‘7.
`
`Averaging over consumers (we assume that the physicians who exercise
`control over the drug consumption decision act as perfect agents for their
`patients) and introducing time subscripts to reflect the fact that the per-
`ceived safety and efficacy characteristics of drugs change over time, we
`obtain a mean consumer utility level from choosing drug j at time t as
`
`(3)
`
`8]: = ijB _ up] + gjr’
`
`that yield invariant oz and B across individuals.
`
`where .5}, may be interpreted as the mean of consumers’lphysicians’ valua-
`tions of an unobserved product characteristic that is not captirred by )9t
`and we use the assumption that E[eij = O.8
`, J, we also assume
`.
`.
`In addition to the competing DMARDs, j = l, .
`the existence of an outside goodj0 with price 120. In this contest, consump»
`tion of the outside good can be thought of as the quantities of NSAIDs
`and all other non-DMARDS consumed by RA patients. (Empirically, al-
`most all RA patients’ visits to doctors result in their being prescribed ei-
`ther an' NSAID or a DMARD or both. Only a tiny number of patients
`receive no drug therapy.) Letting q]. and qj and q0 denote the quantities of
`drug j and the outside. good, respectively, market shares for drug j are just
`sjr = qu/(q1't+ q — j! + qm)
`In this model it is assumed that all aspects of market demand are com-
`pletely determined by the mean utility level 81., and, without going into
`the specifics of supply side dynamics and alternative characterizations of
`market equilibrium, we adopt the special case of the logit model to solve
`for mean utility levels as a function of observed market shares Given the
`utility functionin equation (2), if B. = B for all consumers i, and a is an
`iid variable which follows the type I extreme value distribution, then mar-
`ket share of drug 1 is given by the logit formula
`
`(4)
`
`Sj1(8jr) : exp(8jr)/exp(80r + 28jr)‘
`
`By substitution and by normalizing the mean utility of the outside good
`to equal zero, we get the following linear model for market shares:
`
`(5)
`
`In(sjr) .— 111(501) = 81':
`
`:T’xjefi — “P,- + £17,
`
`where sf, and p}, are the quantity share and price of thejth DMARD at time
`t The unobserved characteristics of the drug, 5,, becomes the error term.
`In our implementation of this basic estimating equation we deflate
`prices by the BLS producer price index for pharmaceuticals to remove
`the general trend of inflation. (This is equivalent to a slightly different
`
`8. See Berry (1994) for more on possible ways of decomposing B. and on the assumptions
`
`IMMUNOGEN 2281, pg. 12
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 12
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`
`
`cians who exercise
`fect agents for their
`e fact that the per-
`inge over time, We
`ugj at time t as
`
`'/physicians’ valua—
`
`iot captured by xfl
`
`J, we also assume
`
`context, consump-
`ntities of NSAIDs
`
`s. (Empirically, al-
`eing prescribed ei-
`tumber of patients
`e the quantities of
`: for drugj are just
`
`demand are com-
`rithout going into
`iaracterizations of
`git model to solve
`shares. Given the
`
`ters i, and 8,]. is an
`.bution, then mar-
`
`the outside good
`set shares:
`
`E,,,
`
`DMARD at time
`es the error term.
`ration we deflate
`iticals to remove
`
`slightly difierent
`
`d on the assumptions
`
`
`
`Hedonic Analysis of Arthritis Drugs
`
`449
`
`specification of equation [2] with a normalization of the utility level of the
`outside good which leaves the price of the outside good in the estimating
`equation.) As argued below, we believe prices to be largely exogenous to
`this market and are therefore unconcerned about endogeneity of this vari-
`able. Given the panel structure of the data, we can address the issue of
`potential correlation between g, and the other explanatory variables by
`including a fixed drug elTect, so that g], = H + 11,-, with n}, assumed to have
`the usual desirable properties.
`
`11.5 Price and Quantity Data
`
`Econometric analysis of the market for DMARDs requires basic data
`on prices and quantities of these drugs sold, and careful attention to the
`definition of the RA market. Our primary data on prices and quantities
`for the DMARDs are drawn from reports of wholesale transactions in the
`United States published by IMS America Inc., a market research com-
`pany. IMS collects information on revenues and quantities of individual
`drug products by wholesale distributors at a very fine level of detail; for
`example, lOO-mg tablets, lOO-count bottle. (We have also collected data on
`retail transactions in British Columbia which were reimbursed under the
`province’s Pharmacare program. The Pharmacare program is universal and
`covers all residents with varying levels of coverage depending on socio—
`demographic status. Trends in these data match the US. wholesale market
`very closely.)
`A major difficulty with these kinds of data, however, is that they are
`collected by drug product, not by disease indication. As pointed out above,
`many of the DMARDs have multiple uses, and in fact their primary use
`may be for quite different medical problems. Analyzing demand for these
`drugs for treatment of RA requires that we distinguish between these uses.
`This may not be important for measuring prices: Lacking some means to
`discriminate among consumers through packaging or reformulation it is
`not unreasonable to assume that one price holds for all sales of a particular
`formulation of a drug regardless of the intended use. This is likely to be
`particularly true for the wholesale market. By contrast, in measuring quan-
`tities it is vitally important to distinguish between markets in the sense of
`different medical conditions. Large (and varying) amounts of these drugs
`are used for treatment of other diseases.
`.
`,
`Figure 11.] presents series on US. wholesale pricés for DMARDs for
`the period 1980—92. Prices are measured in dollars per daily dose unit.
`Daily doses are the “typical maintenance dose” taken from a number of
`standard reference publications such as the Physicians Desk Reference. It
`should be noted that the dosage given to any particular patient may vary
`substantially from the amounts we use here: Treatment of most patients
`may involve considerable experimentation with dosages. Some of the drugs
`33. 1';
`'
`'
`
`
`
`
`
`
`
`
`
`IMMUNOGEN 2281, pg. 13
`Phigenix v. Immunogen
`|PR2014-00676
`
`IMMUNOGEN 2281, pg. 13
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`
`
`Iain M. Cockburn and Aslam H. Anis
`450
`
`
`
`lililllllliltill
`in
`g;
`\3
`i ///
`
`
` ::
`
`
`
`Z
`
`
`
`
`
`
`
`
`
`
`El cortisone ta DMARDs El NSAle
`
`other
`
`Fig. 11.1 Drug class shares of NDTI mentions for RA
`
`also have a fairly complicated “ramp-up” dosage regime lasting many
`weeks before the maintenance dose is treated. Relative prices based on the
`cost of initiating drug therapy and maintaining it for a total of three
`months are very similar to the daily dose prices presented here.
`Perhaps the most striking feature of figure 11.1 is that prices are so
`similar across the major products and move so closely together. Over time
`prices rise st