The Journal of Mental Health Policy and Economics
`J Ment Health Policy Econ 5, 3-19 (2002)
`
`An Analysis of the Diffusion
`of New Antidepressants:
`Variety, Quality, and Marketing Efforts
`
`Ernst R. Berndt,1* Ashoke Bhattacharjya,2 David N. Mishol,3 Almudena Arcelus3 and Thomas Lasky2
`
`1Massachusetts Institute of Technology, Cambridge, MA, USA
`2Janssen Pharmaceutica, Titusville, NJ, USA
`3Analysis Group/Economics, Boston, MA, USA
`
`Abstract
`
`Background: We are not aware of any published research that
`quantifies and compares the importance of effectiveness and side
`effects for pharmaceutical sales, and that simultaneously
`incorporates the impacts of marketing efforts on the diffusion of new
`pharmaceutical agents in the U.S. The overall level and market share
`success of the various selective serotonin reuptake inhibitors
`(“SSRIs”) relative to a representative older generation tricyclic (such
`as Amitriptyline) provides a useful focus for studying such issues.
`Aims of Study: To model jointly the marketing and sales
`relationships of the SSRIs in the U.S., to quantify the extent to which
`marketing efforts are responsive to the availability of new scientific
`information accompanying changes in quality and increases in
`product variety, and in turn to assess how the new FDA indication
`approvals and the enhanced marketing initiatives involving product
`quality and variety affect sales of the SSRI and other novel
`antidepressants.
`Methods: Quarterly US sales, price, quantity and marketing data
`1988Q1-1997Q4 are taken from IMS Health for the eight new
`antidepressants introduced into the US during this time period.
`Measures of physician-perceived quality attributes of the
`antidepressants are drawn from Market Measures, Inc., a medical
`survey research firm. These data are used to construct measures of
`product quality (effectiveness and side effect profile), and attribute
`variety across all antidepressants. Multivariate regression methods
`are used in estimating parameters of a marketing efforts model, a
`sales demand model encompassing the aggregate of the newer
`antidepressants, and a product share model. Simulation methods are
`employed to quantify elasticities.
`Results: Since 1988, and relative to amitriptyline, there has been
`only a rather modest increase in the perceived average effectiveness
`of the SSRIs and related products, but the side effect profiles have
`improved substantially. Variety measures for effectiveness show
`greater increases over time than do those for side effects. Marketing
`efforts respond to science-based events, such as new FDA indication
`approvals, and to effectiveness and side-effect quality improvements.
`
`*Correspondence to: Ernst R. Berndt, Alfred P. Sloan School of Management,
`Massachusetts Institute of Technology, 50 Memorial Drive, Cambridge,
`MA 02142, USA.
`Tel.: + 1-617-253 2665
`Fax: + 1-617-258 6855
`E-mail: eberndt@mit.edu
`Source of Funding: Janssen Research Foundation to Analysis Group/
`Economics.
`
`Copyright © 2002 ICMPE
`
`Total antidepressant sales are positively and significantly related to
`price reductions, increased marketing efforts, and the level and
`variety of side effect profiles involving antidepressants. The level
`and variety of effectiveness does not significantly affect total
`antidepressant sales. Order of entry effects are important in affecting
`product market shares, while marketing efforts and relative quality
`attributes (particularly a more favorable side effect profile) have
`positive and significant impacts on relative market shares.
`Implications for Health Care Provision and Use: Since patient
`response to SSRIs and related products is idiosyncratic, greater
`product variety facilitates better matching of antidepressant with
`patient. Much of the growth of the SSRIs and related antidepressants
`since 1988 can be attributed to increased product attribute variety, to
`improved changes in side effect quality relative to that of the tricyclics,
`and to the marketing of those improvements.
`Implications for Health Policies: Marketing efforts play an
`important role in diffusing product information. Marketing efforts
`increase considerably following FDA approval for indications other
`than depression, and also increase with the average effectiveness and
`the average side effect rating of the products.
`Implications for Further Research: Whether the relatively minor
`role that perceived effectiveness has in affecting sales relative to
`perceived side effect profile is unique to antidepressants, or
`generalizes to other therapeutic classes, merits further examination.
`
`Received 10 December 2001; accepted 11 June 2002
`
`Background
`
`Economic theory suggests that, ceteris paribus, consumers
`benefit from increased product variety.1,2 In the context of
`monopolistic competition, there exists a theoretical literature
`on factors affecting the optimal amount of variety.3 Empirical
`assessments of the effects of variety on overall sales of related
`products are relatively rare, although the empirical literature
`on modeling sales of differentiated products is growing.4-8
`One set of products for which variety could be particularly
`important involves medications to treat illnesses and disorders.
`On a priori grounds, one would expect that since patient
`response to many medications is idiosyncratic and uncertain,
`increases in the variety of medications for treating a particular
`disorder are likely to be valued by society, for as variety
`increases more patients are more likely to be matched with
`
`3
`
`IMMUNOGEN 2279, pg. 1
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`overall market for antidepressants, as well as on sales of
`individual molecules.
`This research focus is important for a number of reasons.
`First, although effectiveness and side effect profiles of
`pharmaceuticals are known to affect product success in the
`marketplace, we are aware of no published research that
`quantifies and ranks the importance of such attributes in
`affecting sales, or provides estimates of the extent to which
`there are trade-offs among them. Here we provide preliminary
`empirical evidence on the relative importance of these various
`attributes in affecting sales. Second, controversy exists
`concerning the role of marketing efforts, and the extent to which
`marketing provides information and/or seeks to influence
`physician prescribing behavior.17, 18, 27, 28 Here we jointly model
`marketing and sales relationships, and quantify the extent to
`which marketing efforts are responsive to the availability of
`new scientific information (e.g., FDA approval of new
`indications) accompanying increases in product variety, and
`in turn how these new indications and the enhanced marketing
`initiatives involving product variety affect sales
`
`Depression and its Treatment: an Overview
`
`Acute depression or major depressive disorder (MDD) is a
`common illness. Estimates indicate that adult lifetime
`prevalence is somewhere between ten to twenty percent.29-31
`Moreover, MDD is often a chronic illness characterized by
`high probabilities of relapse and recurrence.29, 32-37 There is
`considerable evidence that in spite of the availability of a
`number of safe and effective treatments, MDD is
`underdiagnosed and often is inappropriately treated.38-42
`Most forms of depression are treatable, although response
`tends to be somewhat idiosyncratic and unpredictable. Results
`from clinical trials indicate response rates from those
`completing first-line pharmacotherapy for acute-phase
`depression in the range of 50-60 percent, but given the
`increasing variety of antidepressants now available, non-
`responders to first-line therapy often respond to other
`antidepressants.43-45 It is estimated that with the current
`range of available therapies, treatment success rates following
`multiple-line therapy are about 65-80 percent, implying that
`about 20-35 percent of patients may still be resistant to
`antidepressant pharmacotherapy.44-46
`Currently the vast majority of antidepressants block reuptake
`of the neurotransmitters norepinephrine and/or serotonin, and
`fall into four principal classes. The first generations of
`antidepressants were the monoamine oxidase inhibitors
`(MAOIs), which were followed in the 1950s and 1960s by
`tricyclics and tetracyclics (TCAs). The selective serotonin
`reuptake inhibitors (SSRIs) were introduced into the US in
`1988, and in recent years they have become by far the most
`widely prescribed class of antidepressants.47, 48 Recently a
`number of other novel antidepressants have been introduced,
`including serotonin and norepinephrine reuptake inhibitors
`(SNRIs) and other agents.
`Although the clinical and primary care trial evidence to date
`suggests that generally there is no statistically significant
`difference in average treatment response rates among the TCAs,
`
`E. R. BERNDT ET AL.
`
`J Ment Health Policy Econ 5, 3-19 (2002)
`
`effective medicines.9 Medications are one example of what
`Philip Nelson has christened “experience” goods - goods whose
`quality and effectiveness cannot be assessed definitively prior
`to consumption, but can only be determined from consumers’
`own experiences.10,11 By contrast, for “search” goods, quality
`and effectiveness can be largely determined by inspection prior
`to consumption.
`There are at least two important implications that follow
`from the fact that medications are experience goods. First, as
`has been argued by Nelson, in general one should expect
`marketing/sales intensity ratios to be higher for experience than
`search goods (particularly for non-durable experience goods).
`This follows in large part since advertising and marketing are
`envisaged as conveying information about the existence and/
`or quality of the good.12 Thus one should not be surprised that
`marketing/sales ratios are relatively high for medications, both
`prescription and over-the-counter. Moreover, since
`advertising provides greater benefits for higher quality
`experience products in establishing reputation and
`stimulating repeat purchasing, advertising/sales ratios should
`be greater for higher quality experience goods.13-17 An
`implication of this is that once new qualities of an experience
`good are discovered or established (e.g., the Food and Drug
`Administration grants approval to a manufacturer to market
`an existing medication for a new illness or condition), one
`should expect an increase in marketing efforts, ceteris
`paribus.18
`Second, as emphasized by Schmalensee,7 for experience
`goods, order of entry effects are important, and while these
`effects inherently have nothing to do with marketing, in
`practice they may interact. In Schmalensee’s framework, when
`initially skeptical consumers become convinced that the first
`brand in any product class performs satisfactorily, that brand
`becomes the standard against which subsequent entrants are
`rationally judged, and it therefore becomes more difficult for
`later entrants to persuade consumers to invest in learning about
`their qualities than it was for the first brand. To induce
`consumers to make a trial with their brand product, later
`entrants may therefore need to advertise more intensively and/
`or lower the price of their products.19-26
`
`Aims of the Study
`
`In this paper we examine empirically the impacts of product
`attributes, variety in these attributes, marketing efforts, order
`of entry and pricing on the diffusion of a new class of
`pharmaceuticals. The therapeutic class we examine is that for
`the treatment of major depressive disorder. The time frame we
`assess is 1989-97, the decade following introduction of
`Fluoxetine*, the first of a new generation of selective
`serotonin reuptake inhibitors. As measures of quality attributes,
`we utilize data from a medical survey research firm on
`physicians’ changing perceptions of the effectiveness, side
`effects and other quality attributes of antidepressants. Our goal
`is to quantify the impacts of these various factors on the
`
`* The brand name of Fluoxetine is Prozac.
`
`4 C
`
`opyright © 2002 ICMPE
`
`IMMUNOGEN 2279, pg. 2
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`Q388
`
`Q189
`
`Q389
`
`Q190
`
`Q390
`
`Q191
`
`Q391
`
`Q192
`
`Q392
`
`Q193
`
`Q393
`
`Q194
`
`Q394
`
`Q195
`
`Q395
`
`Q196
`
`Q396
`
`Q197
`
`Q397
`
`800.000
`
`700.000
`
`600.000
`
`500.000
`
`400.000
`
`300.000
`
`200.000
`
`100.000
`
`0
`
`Q188
`
`Quarterly Patient-Days of Therapy (000s)
`
`Figure 1. Industry Patients-Days of Therapy for SSRIs and Relative Products Q1 1988 - Q4 1997
`
`SSRIs and SNRIs, there is considerable diversity among them
`in terms of side effect profiles and adverse interactions with
`other drugs.47, 49-51 The SSRIs typically require less titration
`than the TCAs and SNRIs, and thus offer simplicity in dosing,
`a feature that is particularly important to non-psychiatrist
`physicians.50 Since patient tolerability and compliance impact
`medical outcomes, the variability in side effect and adverse
`interaction profiles among the antidepressants has
`considerable clinical significance.
`In particular, because no antidepressant is treatment
`effective in all patients, and because side effects and adverse
`interactions are diverse and to some extent unpredictable, there
`are significant societal benefits to innovations that increase
`the variety of antidepressant treatments available in the
`marketplace. As variety increases, more patients are likely to
`be matched with effective antidepressant pharmacotherapy.
`Within the last decade, growth in sales of the SSRI and
`related antidepressants in the US has been dramatic and
`remarkable. This growth trend is displayed in Figure 1. From
`1988Q1 through 1997Q4, quarterly SSRI and related
`antidepressant sales (measured in patient days of therapy) grew
`from about 5 million in 1988Q1 to 460 million in 1997Q4,
`with particularly high growth since 1993Q3.
`
`THE DIFFUSION OF NEW ANTIDEPRESSANTS
`
`Copyright © 2002 ICMPE
`
`Methods
`
`Theoretical Considerations and Proposed
`Hypotheses
`
`We hypothesize that increases in product variety can facilitate
`the match between a particular patient and a specific
`antidepressant medication, and thus are likely to increase the
`size of the overall antidepressant market.1, 2 *
`Consider the depressed patient searching for appropriate
`antidepressant therapy, aided by a physician. After
`considering the medical history of the patient and his/her
`family as well as the constellation of conditions currently
`being experienced by the patient, and perhaps several other
`factors (e.g., price, the physician’s experiences), the physician
`suggests a particular antidepressant, and informs the patient
`of possible side effects. Perhaps the patient indicates that
`
`* Nonetheless, very little empirical literature is currently available regarding
`optimal treatment choices following failure on an initial antidepressant.
`Further, a related literature dealing with the positive - contagion-mitigating -
`and negative - increased resistance - externalities associated with antibiotic
`prescriptions ascribes a different beneficial role to product variety.52, 53
`
`J Ment Health Policy Econ 5, 3-19 (2002)
`
`5
`
`IMMUNOGEN 2279, pg. 3
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`This reswitching option could significantly lower inertia
`associated with early entrants, and is formalized in a model of
`experience goods studied by Bhattacharjya.55
`Furthermore, even if the new products have the same
`average efficacy in clinical and primary care trials as do
`existing antidepressants, it could be the case that the drug works
`particularly well on one subset of patients (e.g., women), but
`is not as effective in another subset (e.g., men). In such a case,
`while average effectiveness of a new drug may be no better,
`the match between patient and medication may be facilitated
`by the availability of the new variety, for search costs are
`reduced. To the extent marketing efforts reliably
`communicate side effect and effectiveness attributes of new
`products to physicians and patients, both physicians and
`patients will value the information from such marketing
`efforts highly, reducing their search costs.
`The economic reasoning underlying the above arguments is
`drawn in large part from the search literature in labor
`economics.56-60 Suppose an individual with a particular set of
`attributes is looking for employment, and that simultaneously
`there are many employers searching to find employees
`embodying certain characteristics. Both workers and
`employers are heterogeneous. Information about specific wage
`offers is acquired only by search, as is information about
`potential employees, and search takes time and money.
`Employers make offers to selected individuals, and
`individuals then decide whether to accept the offer. Since
`obtaining information on employers is a costly process for job
`searching individuals, and since reliable information on
`potential employee attributes is also costly to obtain
`for the employer, the labor market is one in which there
`is considerable ongoing search behavior. Moreover,
`information can become stale, as conditions change over time.
`As a result, at any point in time, both unemployment and
`help-wanted ads coexist, and wages do not equilibrate supply
`and demand. The resulting unemployment is often called
`“frictional.”
`In the labor market framework, the cost of obtaining
`information by search is a primary determinant of the extent
`of unemployment, for as search costs go to zero, other things
`equal, so too does the number of unemployed at any given
`point in time. Technological and institutional developments
`that reduce search costs by making the acquisition of
`information less costly (e.g., employment services that collect
`information on both workers and employers, low-cost internet
`postings of job offers and job searchers) therefore reduce the
`number of unemployed and increase the number employed,
`other things equal.
`While insights from the matching analogy in labor markets
`are useful, the construction of a formal model of a matching
`process for physicians/patients and antidepressant medications
`is far beyond the scope of this paper. Numerous complexities
`such as the length of search process, formulary restrictions,
`patient compliance and tolerability, step protocols, and
`placebo response are important but difficult to incorporate in
`a formal and rigorous manner. Nonetheless, this framework is
`suggestive of a number of hypotheses that might be assessed
`empirically.
`
`E. R. BERNDT ET AL.
`
`J Ment Health Policy Econ 5, 3-19 (2002)
`
`certain side effects are not acceptable, and so the physician
`suggests an alternative medication. The office visit ends with
`the patient and physician agreeing on a trial treatment.
`The information about the effects of this antidepressant
`treatment trial on a particular patient is costly to acquire. For
`example, it may take six or more weeks for the patient and
`physician to determine whether the patient is responsive to
`this antidepressant treatment. While side effects may manifest
`themselves more quickly, it could still take time to determine
`whether they would subside on their own, or be less intense
`with a lower dosage.
`If the antidepressant is effective without major side
`effects, the patient is likely to remain on treatment. If the
`antidepressant is not effective or if important side effects
`persist, then the physician may prescribe a different
`antidepressant, often called a “second-line” therapy. Some
`patients may have to cycle through a number of different
`antidepressant treatments, taking as long as several years,
`before a suitable match is found between the drug and the
`patient. The available data suggest that for about 20-35
`percent of depressed patients, currently no antidepressant
`offers effective relief of symptoms.
`There are at least two important implications of this costly
`information and search framework. First, the matching model
`helps explain why patient/physician demands for
`antidepressants are likely to be rather price inelastic. A
`patient who has finally found an antidepressant that works is
`likely to develop considerable allegiance to it, and if at all risk
`averse, is likely to resist changing to a different antidepressant
`that has just come on to the market, or because of a reduction
`in the price of another antidepressant. Moreover, physicians
`who find that their patients are responding quite well to a
`particular antidepressant are also likely to continue
`prescribing that drug, at least as a first-line treatment for
`similar patients. Hence antidepressant medications are a good
`example of the order-of-entry phenomena for experience goods
`discussed by Schmalensee.54 That brand loyalty continues even
`after the originator drug loses patent expiration and generic
`drugs enter is well documented in the literature.22, 47
`Second, as new drugs come onto the market embodying
`differing side effect and effectiveness profiles, and as
`information concerning these attributes diffuses, patient/
`physician search costs can be reduced, and the number of
`patients receiving effective antidepressant therapy could
`increase. Product variety, and information concerning that
`variety, can improve the search and matching process.
`Another aspect of variety and experience-based
`information gathering may facilitate evaluation of alternatives.
`Since product quality is revealed to the patient once a
`treatment or a product is tried, the cost of re-switching to a
`certain product after experimenting with alternative treatments
`that prove to be less satisfactory compared to the original
`product in question is negligible, or relatively low. *
`
`* However, there could be a danger that patients who, for whatever
`reason, discontinued an effective antidepressant may not receive the same
`benefit upon resuming use of it.
`
`6 C
`
`opyright © 2002 ICMPE
`
`IMMUNOGEN 2279, pg. 4
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`We hypothesize that marketing efforts will respond
`positively over time to improvements in the side effect and
`effectiveness profiles offered in the antidepressant marketplace,
`both within a product’s life cycle and across products.
`Moreover, we hypothesize that, ceteris paribus, increases in
`product variety and overall product quality will have a
`positive direct impact on total antidepressant sales, in
`addition to the indirect positive impact induced by increased
`marketing efforts. We also hypothesize that order of entry
`effects will be significant factors affecting both marketing
`efforts and sales.
`
`Measurement Issues and Definitions
`
`A very large number of possible attributes are associated with
`a particular antidepressant medication. Side effects could be
`manifested in many different bodily systems and functions
`- agitation, sleep disturbances, gastrointestinal discomfort,
`diarrhea, dryness of mouth, interactions with other drugs, for
`example. Rather than dealing with many distinct product
`attributes (which in some cases are very highly intercorrelated,
`e.g., “incidence of daytime sedation” vs. “effect on quality of
`sleep”), here we develop composite quality measures in two
`dimensions - effectiveness and side effects. Within each
`composite measure, we select several individual attributes for
`inclusion. Each of the attribute measures is based on survey
`research from a physician panel undertaken annually by
`Market Measures, Inc., a New Jersey-based medical
`marketing information firm that collects a variety of survey
`data across a wide range of therapeutic classes and disease
`states (www.mmi-research.com). The physician survey panel
`is recruited in an ongoing basis from a random sample of each
`medical professional universe. For the class of antidepressant
`drugs, and as only one portion of their annual study, MMI
`received completed self-administered questionnaires from a
`panel of approximately 300 physicians (about 100 each of
`psychiatrists, internists and general/family practitioners), in
`which physicians provided rating scores of 1 to 5 to the
`various attributes of a particular drug, with higher scores
`representing better quality. The measures of product quality
`attributes are based on physicians’ changing perceptions of
`how antidepressants perform in actual clinical practice, rather
`than how the manufacturers report them based on information
`from randomized clinical trials. Physicians are surveyed not
`only in terms of their perceptions of various product attributes,
`but also in terms of how important the particular attribute is to
`them. Specifically, physicians are asked to rate each attribute
`on a 1.0 (least important) to 5.0 (most important) scale.
`Physicians’ scores are weighted by their relative
`antidepressant prescribing volume, measured by physicians’
`average number of patients prescribed an antidepressant per
`specialty, as reported by physicians to MMI. The MMI
`quality measures are annual; in the quarterly regressions
`reported below, quantity measures are set to their annual level
`in all four quarters.
`As discussed in further detail below, to construct an
`aggregate measure of effectiveness for each medication, we
`compute a weighted average of physicians’ mean evaluations
`
`THE DIFFUSION OF NEW ANTIDEPRESSANTS
`
`Copyright © 2002 ICMPE
`
`on the effectiveness of a particular medication in treating
`(i) mild to moderate depression, and (ii) moderate to severe
`depression, where the weights are based on physicians’ 1996
`responses to questions asking the relative importance of each
`attribute in prescribing drug therapy to treat depression. For
`side effects, we construct for each product a weighted average
`of responses to six specific side effects queries: daytime
`sedation, anticholinergic side effects, toxicity from overdose,
`incidence of sexual dysfunction, agitation, and suitability for
`long-term therapy.
`We now outline construction of quality measures, for the
`“industry” (the SSRI and related products therapeutic class)
`as a whole, and for individual antidepressant medications.
`
`Product-Specific and “Industry” Measures of Quality
`
`Let a jit represent the rating for attribute j of product i at time t,
`and let w jt be the attribute-specific “importance weight” taken
`from physician survey data. Since these specific weights were
`only explicitly provided for one year (1996) in our 1988-97
`MMI sample time frame, we remove the t subscript on wjt and
`only employ wj as the jth attribute weight. For product i, the
`average quality is constructed as
`= J
`
`
`
`a
`
`it
`
`=
`1
`
`j
`
`a
`
`jit
`
`w
`
`j
`
` (1)
`
`These product-specific quality measures are computed for both
`effectiveness and side effects.
`At the “industry” or therapeutic class level of aggregation,
`average quality measures are constructed as
`
`
`
`A
`t
`
`= J
`
`=
`1
`
`j
`
`wm
`jt
`
`j
`
` (2)
`
`where mjt is the arithmetic mean of attribute j over all SSRIs
`and related products at time t, and wj is the attribute
`´
`importance weight defined above. Note that t, the average
`industry quality index, can vary as new products enter the
`market, and as physicians’ perceptions change.
`It will be useful to develop a relative notion of average
`industry quality, since one research objective is to assess the
`impact of changing average industry quality on the demand
`for the aggregate therapeutic class of SSRIs and related
`products.
`The SSRIs and related products have frequently been
`compared to an earlier generation of antidepressants known
`as tricylics or tetracyclics (TCAs). Perhaps the best known of
`the TCAs is Amitriptyline. We choose Amitriptyline to
`represent the quality of all antidepressants prior to the market
`entry of Fluoxetine, the first SSRI, because aspects of the side
`effect and effectiveness profiles of Amitriptyline are similar
`to those of other TCAs.47 With Amitriptyline representing
`pre-SSRI and related product quality attributes, we then
`construct the industry or therapeutic class average quality
`“frontier” measure Ft as the proportional difference between
`the average quality of the SSRIs and related products, ´
`1t,
`7
`
`J Ment Health Policy Econ 5, 3-19 (2002)
`
`IMMUNOGEN 2279, pg. 5
`Phigenix v. Immunogen
`IPR2014-00676
`
`(cid:229)
`(cid:215)
`(cid:229)
`(cid:215)
`

`

`More generally, we compute variety Vt as
`
`(cid:247)(cid:247)(cid:247)(cid:247)(cid:247)ł(cid:246)(cid:231)(cid:231)(cid:231)(cid:231)(cid:231)Ł(cid:230) --
`(cid:229) (cid:229)
`
`N
`
`=
`
`i
`
`2
`
`a
`
`it
`
`i
`
`1
`
`1
`
`j
`
`=
`i
`
`a
`
`jt
`
`1
`
`(5)
`
`
`
`
`
`(3)
`
`
`
`=
`
`V
`
`t
`
`
`
`AA
`
`,1=
`
`,0
`
`tt
`
`
`
`(cid:247)(cid:247)ł(cid:246)(cid:231)(cid:231)Ł(cid:230) (cid:215) (cid:247)(cid:247)ł(cid:246)(cid:231)(cid:231)Ł(cid:230) (cid:215)
`(cid:229)(cid:229)
`
`m
`
`jt
`
`1,
`
`w
`
`j
`
`J
`
`= J
`
`j
`
`1
`
`and that of the traditional TCA pharmacotherapies, ´
`0t.
`Specifically, the SSRI and related products average frontier at
`time t, Ft, is computed as
`
`=
`
`F
`t
`
`
`
`m
`
`jt
`
`0,
`
`w
`
`j
`
`=
`1
`
`j
`
`where N is the number of new antidepressant products on the
`market at time t. This measure of product variety is
`mathematically equivalent to the measure of product distance
`in the differentiated product space model implemented by
`Stavins.8 It should be pointed out, however, that this measure
`is invariant to changes in the composition of “variety,” insofar
`as it does not allow one to capture the potentially
`idiosyncratic responses of patients to products that may be
`equally ‘varied’ on average, but whose constituent attributes
`may differ in opposite directions. This is a constraint imposed
`on the analysis for reasons of simplification and tractability.
`
`Quantities of Antidepressant Medications
`
`To quantify the diffusion of antidepressant medications, a
`measure is needed that is comparable across different
`products. IMS Health provides data on revenues, units sold by
`product and what they call extended units (essentially number
`of tablets or capsules). Quarterly sales data to retail outlets
`(projected to national levels based on data from 28,000 retail
`pharmacies) were made available to us covering the 1988Q1
`through 1997Q4 time period. The products included in our
`analysis are Fluoxetine, Buproprion HCL, Sertraline HCL,
`Paroxetine HCL, Venlafaxine HCL, Nefazodone HCL,
`Fluvoxamine Maleate and Mirtazapine.* Since typical daily
`dosing is likely to vary across drugs and perhaps over time,
`the extended units measure is standardized by dividing
`extended units by the average number of tablets administered
`per day, using Retail Provider Perspective data from IMS
`Health. This provides a quantity measure of total patient-days
`of antidepressant pharmacotherapy that is consistent across
`products and over time. Price per day of therapy is then
`computed as revenue divided by the patient day quantity
`measure. To adjust for overall inflation, this nominal price
`measure is divided by the overall US Consumer Price Index
`(1982-84=1.00).
`For each time period beginning 1988Q1, total patient days
`of therapy is computed for the benchmark TCA,
`Amitriptyline, and is denoted as Q 0t. Quantity measures for
`each product in the new classes of antidepressant medications
`are noted as q1i, where the subscript i refers to product i in the
`new classes of antidepressant medications (defined in turn by
`subscript 1). Total patient-days of therapy for the class of
`SSRIs and related products is the sum of the individual
`
`* The brand names of the products included in the analysis are: Fluoxetine
`(Prozac), Buproprion HCL (Wellbutrin), Sertraline HCL (Zoloft), Paroxetine
`HCL (Paxil), Venlafaxine HCL (Effexor), Nefazodone HCL (Serzone),
`Fluvoxamine Maleate (Luvox) and Mirtazapine (Remeron).
`
`E. R. BERNDT ET AL.
`
`J Ment Health Policy Econ 5, 3-19 (2002)
`
`where m jt,1 is the mean for attribute j over all SSRIs and
`related products on the market at time t, m jt,0 is the value of
`attribute j for Amitriptyline (a weighted average over the
`number of physicians in the panel), and w j is the perceived
`“importance” weight assigned to attribute j by the physician
`panel.
`Finally, it will also be useful to have product-specific
`relative measures of quality. We focus on quality competition
`within the new class of antidepressant medications by
`calculating the relative distance in product space between
`product-specific measures of average quality and the industry
`average. Specifically, the normalized quality distance for
`product i relative to the industry average is computed as
`
`
`
`=
`
`r
`it
`
`--
`
`it
`
`a
`a
`
`max
`
`A
`,
`t
`a
`
`min
`
`i
`
` (4)
`
`where ´
`t,-i is defined as the industry average quality
`excluding product i and amax and amin are the largest and
`smallest possible quality ratings, respectively. The value r it is
`therefore bounded between –1 (poorest quality) and 1
`(highest quality). During the time period when Fluoxetine was
`the only SSRI competitor in the market, the value of r it is
`defined to equal zero. For each of these industry and
`product-specific relative quality variables, separate measures
`are computed for ef