PHARMACEUTICAL PRICING in a REGULATED MARKET Mats Ekelund and Bjo¨Rn Persson*

PHARMACEUTICAL PRICING IN A REGULATED MARKET Mats Ekelund and Bjo¨rn Persson*

Abstract—We compare how new pharmaceuticals are priced in the price- environment influence the pricing patterns and price com- regulated Swedish market with how they are priced in the U.S. market, as studied by Lu and Comanor (1998). We collect a data set consisting of all petition, compared with the unregulated market? new chemical entities (NCEs) launched in Sweden between 1987 and We do this by examining how the pricing strategies of 1997, and test the same models as Lu and Comanor. In line with their pharmaceuticals in a price-regulated market relate to the results, we find that introductory prices depend on the degree of therapeutic innovation. Contrary to the results from the U.S. market, Swedish pharmaceutical pricing in an unregulated market, using real prices for NCEs fall substantially over time for all classes of identical explanatory variables. More specifically, we con- therapeutic innovation. Also contrary to the findings of Lu and Comanor, sider how the price regulation regime in Sweden rewards we find no effect of the presence of branded substitutes on either introduction prices or price dynamics. Our results indicate that the price therapeutically advanced products in terms of relative regulation discourages price competition between brandname drugs. prices, compared with the unregulated U.S. market. Further- more, we consider how the presence of substitutes affects the nature of price competition in the two different markets. Downloaded from http://direct.mit.edu/rest/article-pdf/85/2/298/1613500/003465303765299828.pdf by guest on 30 September 2021 I. Introduction As a rule, the presence of substitutes has a depressing effect he rapid increase in pharmaceutical expenditures in on prices, but when prices are set by a regulating agency the Tmany countries (see for example Abbott, 1995, and extent of this influence is not clear. We therefore also Besley & Gouveia, 1994) has generated much interest in the consider the empirical connection between pricing patterns pricing of pharmaceuticals. It is sometimes argued that the and the market structure. presence of patent protection and third-party financing There have been a number of empirical studies on the could lead to excessive prices, and therefore many nations pricing of new patented drugs in the U.S. market. Reekie have some form of expenditure regulation. One example is (1978), in a seminal study, examined the introductory prices various forms of price-cap regulations, and many nations of new chemical entities (NCEs) launched in the United worldwide are currently applying these in their attempts to States between 1958 and 1975. He found that the introduc- limit the cost of the pharmaceutical bills. However, this use tory price (relative to existing substitutes) depended on the of price-cap regulation is the subject of some controversy. degree of therapeutic advance, that is, the price premium Opponents argue that the price regulation may adversely was larger for drugs offering important therapeutic gains affect incentives to develop new and better products, since than for drugs offering minor therapeutic gains. Reekie producers are not adequately reimbursed for the massive (1978) also found that prices tended to increase faster for investments needed to bring new drugs to the market. This drugs introduced at lower prices than for drugs with higher standpoint is taken by many in the United States, where no launch prices. price regulation exists (see for example Danzon, 1997, for a The pricing pattern observed by Reekie is consistent with discussion of the adverse effects of price regulations). the observation made by Dean (1969), who distinguished Moreover, it has also been maintained that price caps are not between two pricing strategies for new products: skimming even effective in curbing pharmaceutical outlays (see for and penetration. A skimming strategy involves setting a example Abbott, 1995, and Danzon & Chao, 2000a, for a relatively high introductory price to skim off the highest recent empirical study). willingness to pay, and then lowering the price. Penetration In order to evaluate the effects of price regulations and to is the opposite strategy: lower introductory prices followed resolve the controversies surrounding pharmaceutical pric- by increased prices as demand picks up. Dean (1969) argued ing, we believe that further empirical work is desirable. In that skimming strategies are more often used for products this paper, we take a step in this direction by performing a offering major advantages over existing products, and that basic comparison of pricing patterns in a market with penetration strategies are used for products that offer only regulated prices with those in a market where prices are not marginal improvements over existing products. regulated. In particular, we address two questions: (1) are Recently, Lu and Comanor (1998) (henceforth LC) per- the rewards to therapeutic advance under price-cap regula- formed a study of 144 NCEs introduced in the United States tion similar to those in an unregulated market, and (2) how between 1978 and 1987.1 The principal motive of their does the restriction put on pricing strategies in the regulated paper was to explore and quantify the demand-side determinants for NCE prices. In their view, therapeutic value and Received for publication August 13, 1999. Revision accepted for pub- market structure are the main explanatory variables for NCE lication March 11, 2002. * Stockholm School of Economics. pricing. This is in contrast to the more established supply- We thank Magnus Johannesson for his advice and comments. We have side arguments often used in the debate over pharmaceutical also benefited from comments and discussion with Tore Ellingsen, Ulf prices. Gerdtham, Bengt Jo¨nsson, and two anonymous referees. Financial support from the National Association of Swedish Pharmacies is gratefully ac- knowledged. 1 In their study, the term “new molecular entities” (NME) is used.

The Review of Economics and Statistics, May 2003, 85(2): 298–306 © 2003 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology PHARMACEUTICAL PRICING IN A REGULATED MARKET 299

LC classified the new pharmaceuticals into three different tion exceeding some fixed amount per annum.2 If pharma- classes of therapeutic gain, depending on whether the drugs ceutical producers wish to have their products reimbursed represented a major, a minor, or no improvement over through this system, prices must be set by the appropriate existing drugs. Introductory prices for drugs representing government agency. Prior to 1993, prices were determined important therapeutic gains were on average about three by the National Corporation of Swedish Pharmacies times higher than those for existing substitutes, whereas (NCSP) after negotiations with the producers. The NCSP is new drugs that offered little therapeutic gain were launched government-owned and monopolizes the retailing of all at about the same price as existing substitutes. The real pharmaceuticals in Sweden. When the NCSP was in charge prices for drugs representing important therapeutic gains of the price negotiations, firms were not allowed to sell their were relatively stable over time. However, for drugs repre- products until the prices were agreed upon. If the NCSP and senting small improvements the real prices increased sub- a firm were unable to reach an agreement about the price, stantially. These results are thus consistent with Dean’s the firm could apply to engage the Medical Products Agency 3 (1969) observation of skimming and penetration strategies. (MPA) as a mediator. In 1993, the National Social Insur- Downloaded from http://direct.mit.edu/rest/article-pdf/85/2/298/1613500/003465303765299828.pdf by guest on 30 September 2021 LC also found that the presence of branded substitutes had ance Board (NSIB) replaced the NCSP as the government a clear negative effect on both introductory prices and price agency responsible for administering the drug benefit changes over time. scheme. LC concluded that competitive forces play an important Since that time, participation in the public insurance role in the U.S. pharmaceutical market, and hypothesized program has been optional, so producers are allowed to that an introduction of price regulations may discourage market their drugs outside the program and set prices freely. penetration strategies. However, since consumption outside the benefit program The purpose of the present study is to compare the must be financed completely out of pocket, producers have outcome in a (price-)regulated market with the results in the rarely exercised this option.4 With that exception, the gen- U.S. market by performing the same tests as in LC. To do eral rules and guidelines under which both regulating agen- so, we examine a data set consisting of all NCEs launched cies operate remained unaltered. The general objective of in Sweden between 1987 and 1997, and estimate the same the NCSP/NSIB is to determine pharmaceutical prices in models as in LC. In line with their results, we find that order to introductory prices reflect the degree of therapeutic innovation. Contrary to the results in the U.S. market, the real ● ensure a consistently high level of public health; prices for NCEs fall substantially on average over time for ● ensure that a sufficient range of pharmaceutical prod- all classes of therapeutic gain, so the price-cap regulation ucts are available at reasonable prices; rules out the use of penetration strategies. Perhaps more ● support efficient production of pharmaceuticals; interestingly, we find that there is no significant effect from ● encourage research and development of new pharma- branded substitutes on either launch prices on price dynam- ceuticals.5 ics. It therefore appears that price competition between pharmaceutical brands is indeed less pronounced under Furthermore, when setting the launch price for a new price-cap regulation. In addition, we note that the explana- product, the regulating agency should consider the medical tory power of the models considered is low. This can be merits and the health economic value of the product, the interpreted as evidence in favor of the view that price price in comparable countries,6 and the price (and reim- regulation offsets some of the disciplinary (price) effects bursement cost) for related treatments (pharmaceutical or that are present in the unregulated market. other). The health economic value represents the societal The remainder of this paper is structured as follows. We benefit the new drug brings in terms of savings for the total first give a brief description of the pharmaceutical market in reimbursement bill (for example, through effects on other Sweden. Then we discuss some of the hypotheses that the products comprised in the benefit schedule), savings on regulatory constraint provides and put these in relation to social insurance due to reduced sick leave, healthcare sav- the findings in LC. We continue with a description of the ings in the form of a reduced number of physician visits, and estimated models and the data, then with a presentation of the results. The final section offers a concluding discussion. 2 Presently, this amount is SEK 1800 (approx. USD 180). Although the copayment schedule has changed somewhat over the time for our sample (1987–1997), the state-financed part has always been large. 3 The MPA is roughly the Swedish equivalent of the FDA in the United II. The Pharmaceutical Market in Sweden States. 4 In a random test performed in 1999 for some fifty pharmacies, only In Sweden, a substantial fraction of the pharmaceutical about 0.3% of the prescriptions were for drugs outside the reimbursement consumption is subsidized by the government. For pharma- system (SOU 2000.86: Den nya la¨kemedelsfo¨rmånen). 5 These points are in accordance with the 1990 Transparency Directive ceutical products that are included in the public insurance of the European Union. program, the state will fully reimburse individual consump- 6 Most notably other northern European countries. 300 THE REVIEW OF ECONOMICS AND STATISTICS also savings due to a reduced amount of care outside the pricing is being used for chronic drugs. This is expected, for direct healthcare sector (for example, home care). chronic drugs are to a large extent repeat purchase items, The manufacturer is responsible for providing the regu- and the appropriate strategy for pricing such goods is by a lating agency with the relevant information about the drug.7 penetration scheme: low initial prices in order to attract a This information should also contain, apart from the above, consumer base whose loyalty can be exploited when the forecasts of sales in Sweden in the first two years after producer increases prices later. launch, and predicted sales in a steady state. Furthermore, LC find substantial evidence that the Once an introductory price has been set, it cannot be presence of brandname substitutes has a negative effect raised within the two years after launch. Thereafter, appli- on relative introduction prices. This is also expected: cations for price increases for the whole product line, but even though brandname products are all patented, they not individual products, are considered once per year. As a may still have similar therapeutic properties and func- general rule, the producer may increase prices within some tions. Therefore, prescribing physicians have more prod- specified margin and is granted some leeway for individual ucts to choose from in treating a given condition, which Downloaded from http://direct.mit.edu/rest/article-pdf/85/2/298/1613500/003465303765299828.pdf by guest on 30 September 2021 products in the basket. For instance, price increases on naturally increases the competition between similar phar- certain products may be allowed if the producer lowers maceuticals. prices on other products in the basket. Finally, LC find that on average the presence of generic In 1993, the reference pricing system was introduced in Sweden. This system applies when the patents for brand- competition among the brandname substitutes has a pos- name pharmaceuticals have expired and there are generic itive effect on relative launch prices for NCEs. LC report substances on the market. If generic alternatives are avail- conflicting evidence on the effect of generic entry on able, the reimbursement scheme allows full coverage for NCE prices from earlier studies: Caves, Whinston, and products that cost no more than 1.1 times the lowest-priced Hurwitz (1991), Frank and Salkever (1997), and generic pharmaceutical (belonging to the same therapeutic Grabowski and Vernon (1992). One explanation for the subgroup). If consumers want more expensive pharmaceu- positive price effects, offered both by Frank and Salkever ticals, they have to pay the difference out of pocket. The (1997) and by LC, is that it may be more profitable for a reference price system was intended to encourage price brandname producer to focus on satisfying the inelastic competition so that average prices for off-patent brand segment of demand when generic drugs enter the market. pharmaceuticals within the system would decrease. Except In this instance, other brandname producers may follow for the introduction of this system, little has been done to suit and increase their prices too. In short, LC conclude promote price competition between brandname and generic that the demand factor variables they consider explain pharmaceuticals. For instance, generic substitution has not both launch prices and price dynamics in the American been practiced during the period under study here. We note pharmaceutical market. that brandname pharmaceuticals strongly dominate the We expect some of the effects that LC describe to be Swedish market. In 1991, they made up 70% of total sales present also in the Swedish data. However, given the broad volume and 88% of total sales value in Sweden (Nilsson et regulatory policy objectives stated above, it is not obvious al., 1998). how the two markets compare, given the same explanatory variables. For example, the price regulation in Sweden makes it difficult for firms to raise the price of one of their III. Preliminary Discussion and Hypotheses products without simultaneously lowering prices on other LC argue that demand factors explain well the determi- products. Consequently, we do not expect to observe pene- nants for NCE prices in the United States, and their findings tration strategies. Instead, we are more likely to observe are in line with what standard economic theory predicts. constant or falling real prices over the whole range of NCEs Moreover, their results conform with the hypothesis ad- in the sample, although it is conceivable that the prices of vanced by Dean (1969) and subsequently also supported less therapeutically advanced drugs will fall less than those empirically by Reekie (1978). In particular, they find that of more innovative drugs. therapeutic advantage has a positive effect on relative As far as launch prices are concerned, it is reasonable to launch prices because willingness to pay is increasing in expect that therapeutically advanced drugs will be rewarded quality. Also, pharmaceuticals that are mainly intended for with premia by the regulator as compensation for high R&D chronic conditions command on average lower relative outlays. The regulator could argue that, given that these launch prices than drugs primarily intended for acute con- costs are sunk, it would be rational to offer the producer ditions. Moreover, the former display a pattern of increasing marginal cost coverage only. Such behavior, however, is not prices over time, so there is evidence that penetration in line with the regulator’s objective to “encourage the research and development of new pharmaceuticals.” Also, 7 The regulating agency has no means of conducting investigations of its own, but must rely on the information supplied by the applicant or outside producers would be hesitant to supply the Swedish market sources. with their products in case they were not sufficiently PHARMACEUTICAL PRICING IN A REGULATED MARKET 301

8 compensated. Hence, we expect similar patterns to those in LWRIS ϭ ␣0 ϩ ␣1A ϩ ␣2B ϩ ␣3 ACUTE LC in this respect. However, we do not expect any major differences between drugs intended for acute and chronic ϩ ␣ ϩ ␣ ͭDGͮ ϩ ⑀ 4 LNS 5 (1) conditions, as penetration pricing is not really an option for LG the producers. and LC argue that the market structure, or, more speciﬁcally, the nature of competition, is a strong candidate in explaining LRATIO ϭ ␤0 ϩ ␤1A ϩ ␤2B ϩ ␤3 ACUTE both launch prices and price dynamics. In particular, the ϩ ␤ ϩ ␤ ϩ ␤ existence of more branded substitutes has depressing effects 4 DALP 5 DBLP 6 DCLP on NCE launch prices. In the regulated environment, how- ϩ ␤ ϩ ␤ ͭ BG ͮ ϩ ␦ ever, the market structure may not reﬂect price setting to the 7 LNS 8 LGS , (2) same extent, so the relationship between prices and the number of substitutes is less clear-cut. On the one hand, the

where Downloaded from http://direct.mit.edu/rest/article-pdf/85/2/298/1613500/003465303765299828.pdf by guest on 30 September 2021 presence of branded substitutes provides the regulator with more bargaining power to keep prices low by exercising ● LWRIS denotes the logarithm of the ratio of the launch some form of yardstick regulation. If this effect is dominant, price of an NCE to the weighted average price of we would expect a negative correlation between launch existing brandname substitutes; prices and the number of substitutes. On the other hand, in ● LRATIO is the logarithm of the ratio of the real price the interest of the public health, the regulating agency has a (CPI-deflated) of a new drug four years after launch to commitment to ensure the availability of a variety of hori- its real introductory price; zontally differentiated products. For any given level of ● A and B indicate the degree of therapeutic advance demand (as measured for example by the total number of (defined in the next subsection) and take the value 1 if patients with a certain indication), each product then risks an NCE is classified as A or B, respectively, and 0 obtaining a smaller market share, so the regulator compen- otherwise; sates producers by allowing higher prices. This effect could ● ACUTE indicates whether an NCE is used primarily then counteract the negative influence that the number of for an acute condition; substitutes has on prices. Consequently, the strong connec- ● LNS denotes the logarithm of 1 plus the number of tion between NCE prices and brand substitutes observed in existing brandname substitutes at the time of launch; the U.S. data may be less pronounced in the Swedish data. ● DG indicates whether an NCE had generic competi- As mentioned above, there is mixed empirical evidence tion among its brandname substitutes at the time of of the relation between prices on new pharmaceuticals and launch; the presence of generic competition among the branded ● LG denotes the logarithm of 1 plus the percentage of substitutes. Here, we have no reason to believe that the substitutes that had generic rivals at the time of launch; presence of generics among the brandname substitutes ● BG indicates whether or not an NCE has a generic should affect the NCE prices in any systematic way. The rival four years after launch, when there was none at same ambiguous effects we expect from branded competi- launch; tion are expected to be present also in the case of generic ● LGS is the logarithm of the ratio of the percentage of drugs. Furthermore, brandname substitutes with generic brandname substitutes that had a generic rival after competitors tend to be older drugs located further away four years to the same percentage at the time of NCE from the NCE in product space than newer brands without launch (expressed as whole numbers plus 1); generic competition. Therefore we expect a weak link be- ● DALP, DBLP, and DCLP are interactive variables tween NCE prices and generic competition in our sample. equaling the product of LWRIS and the variables indicating whether the NCE is classified as A, B,orC, respectively. IV. Empirical Analysis These models are in effect identical to the ones estimated A. Equations and Variables by LC, and therefore provide a framework for comparing the results in the two different markets. We now proceed with the estimation of the LC model on the Swedish data. We consider the following two models: B. The Data

8 It can be said that the Swedish market is very small from an interna- The data we use consist of all NCEs approved by the tional perspective, so, given the sunk costs, any price above marginal cost MPA in Sweden between 1987 and 1997. In total the MPA should be accepted by the producers. However, there are additional costs approved 335 NCEs in this time period. Of these NCEs, we associated with the marketing of a new drug: training of pharmaceutical consultants, promotion campaigns, and so on. Therefore, the option of not excluded 89 from the present analysis either because they supplying the drug may be seriously weighed. were never marketed in Sweden, or because no price had yet 302 THE REVIEW OF ECONOMICS AND STATISTICS been set on them at the time of this analysis. Our data set chemical drug class, the substitute was defined as the drug thus includes a total of 246 NCEs. that was most commonly prescribed for that indication in No official rating of therapeutic advance is available for the year before the introduction of the NCE. pharmaceuticals in Sweden. As a part of this research For class A drugs it is often difficult to define substitute project, a rating was carried out by pharmacologists con- drugs, since many of these compounds are intended for nected to the MPA (see Roseń and Beermann, 1999). Roseń indications not previously treated by any marketed drug (see and Beermann had access to all relevant NCE information the above definition of class A drugs). In the ranking of from the MPA and collaborated with the agency to deter- therapeutic advance according to the FDA classification mine the rating. The purpose was to provide a rating of the scheme, Roseń and Beermann (1999) therefore divided the therapeutic advance for all NCEs approved between 1987 class A drugs into: and 1997, according to how the NCE was judged by the MPA at the time of approval. They classified the NCEs (i) drugs that lacked a substitute drug at the time of introduction (drugs for indications not previously

according to the FDA rating system that was used also in the Downloaded from http://direct.mit.edu/rest/article-pdf/85/2/298/1613500/003465303765299828.pdf by guest on 30 September 2021 study by LC. These classes are: treated by drugs), and (ii) drugs with substitute drugs at the time of introduc- ● A—important therapeutic gain: Drug may provide ef- tion (drugs providing markedly improved efficacy fective therapy (by virtue of greatly increased efficacy or safety over existing drugs). or safety) for a disease not adequately treated or diagnosed by any marketed drug, or provide markedly Out of the total 246 NCEs in this study, there were a total improved treatment of a disease through improved of 8 class A drugs with substitutes and 28 without substi- efficacy or safety (including decreased abuse poten- tutes. Naturally, the drugs without substitutes could not be tial). included in the analysis of relative introductory prices, as ● B—modest therapeutic gain: Drug has a modest but their relative prices could not be defined. real advantage over other available marketed drugs; In order to estimate relative prices between NCEs and for example, somewhat greater effectiveness, de- substitutes, comparable dosages need to be defined. For this creased adverse reactions, or more convenient route of purpose we used the defined daily doses (DDD) system administration. recommended by the WHO (1997) for studies of drug use ● C—little or no therapeutic gain: Essentially duplicates (ATC Index with DDDs). A DDD is defined as the average in medical importance and therapy for one or more daily dose of a drug used by an adult for treatment of the existing drugs. main medical indication of the drug. Official DDDs were available for about two-thirds of the drugs in our sample and were used in those cases. For the remaining drugs daily In addition to rating the new NCEs according to thera- dosages were based on the recommended average daily peutic advance, Roseń and Beermann (1999) also classified dosages in Pharmaceutical Specialities in Sweden (FASS), the drugs according to type, that is, whether a drug was a widely used medical reference book that contains instruc- indicated primarily for treatment of acute illnesses or tions and recommendations on the appropriate use of all chronic ones. A drug is defined as acute if it is intended for drugs in Sweden. When recommendations regarding dos- conditions lasting no more than 3 months, and this defini- ages were not sufficiently specific or not available in FASS, tion is also used by LC. we consulted clinical pharmacologists (Roseń and Beer- Roseń and Beermann used the WHO Anatomic Thera- mann) for the appropriate daily dosages. When doses were peutic Classification (ATC) system to define substitute given in amount per square meter (as for creams and drugs [ATC Index with DDDs defined daily dosages lotions) or amount per kilogram of body weight, the daily (DDDs), 1997). The ATC system is divided into 14 anatom- doses were based on 1 m2 and 70 kg, respectively. ical groups (for example, heart and circulation). Each ana- Some pharmaceuticals are available in more than one tomical group is divided into main therapeutic groups, and form of administration. For example, the same drug can be then into therapeutic subgroups, which are further divided taken as a pill or a fluid. As a rule we calculated the price for into chemical-therapeutic subgroups (for example, HMG- tablets when these were available. If there was no tablet, we Co A reductase inhibitors). For example, one such chemical chose the administration form of the package containing the substance is the cholesterol-lowering drug Simvastatin, largest number of daily dosages. which belongs to the HMG-CoA reductase inhibitors We based all prices on the official retail prices of drugs in chemical-therapeutic subgroup. We defined close substitutes Sweden.9 These prices are set once a year and are reported as drugs in the same chemical-therapeutic subgroup that in FASS. We estimated a price per daily dose for all NCEs share the same indication. and all substitute drugs. We estimated the average price for This is similar to the definition used by LC, and for most of the new NCEs substitutes were defined in this way. In the 9 In Sweden the price is independent of the buyer (which is not the case event that an NCE did not have any substitute in the same in, for example, the United States). PHARMACEUTICAL PRICING IN A REGULATED MARKET 303

TABLE 1.—RELATIVE INTRODUCTORY PRICES Sweden United States* Class Type N Med. Mean S.d. Range N Med. Mean S.d. Range A Acute 5 1.43 1.59 0.56 1.10–2.55 3 2.97 3.12 0.88 2.32–4.07 Chronic 3 10.97 9.40 4.88 3.93–13.30 10 2.07 3.11 2.76 0.20–8.08 Combined 8 2.04 4.52 4.83 1.10–13.30 13 2.47 3.11 2.47 0.20–8.08

B Acute 30 1.64 3.75 4.96 0.17–19.28 19 1.72 3.09 3.65 0.40–13.60 Chronic 42 2.55 3.92 4.77 0.09–20.32 29 1.19 1.63 1.14 0.48–5.97 Combined 72 2.25 3.86 4.85 0.09–20.32 48 1.35 2.21 2.53 0.40–13.60

C Acute 63 1.26 2.16 3.09 0.04–16.70 16 1.23 1.37 0.68 0.55–3.13 Chronic 75 1.13 2.18 3.54 0.12–26.17 53 0.94 1.07 0.80 0.37–6.00 Combined 138 1.16 2.17 3.33 0.04–26.17 69 0.97 1.15 0.78 0.37–6.00 * Data from Lu and Comanor (1998). Downloaded from http://direct.mit.edu/rest/article-pdf/85/2/298/1613500/003465303765299828.pdf by guest on 30 September 2021 existing substitutes as the weighted average of the price of regulated market than in the U.S. market. The median all substitutes, with the market shares in the year before the values show smaller differences between the two markets introduction of the NCE as weights. We collected data on than the means, with the exception of the class B drugs. substitute quantities from Swedish Drug Market (SDM) The number of branded substitutes at introduction de- (1987–1997), a database containing the quantities of all creases with increasing degree of therapeutic innovation. drugs sold on the Swedish market, graciously provided by The average number of branded substitutes at introduc- the Swedish Association of the Pharmaceutical Industry. We tion in the Swedish market is 1 for A drugs, 3.9 for B then adjusted all prices to 1997 prices, using the official CPI drugs, and 4.5 for C drugs. These figures are similar to produced by the Statistical Yearbook of Sweden (1999). those in the United States. Furthermore, in both markets the share of branded substitutes that face competition V. Results from generic substitutes at the time of introduction de- A. Summary Statistics creases with increasing therapeutic value. Generic compounds constitute a slightly larger share of the market in Table 1 summarizes some statistics on relative introduc- the United States than in Sweden. This is hardly surpris- tory prices for 218 NCEs introduced between 1987 and ing, given that the copayment rate is substantially larger 1997, excluding the 28 class A drugs without substitutes. To in the U.S. than in the Swedish system, where pharma- facilitate the comparison between the two markets, we also ceutical consumption is reimbursed. reproduce the results from LC in each table. In table 2 we show some summary statistics about the On average, relative launch prices increase with the inflation-adjusted change in the price of NCEs four years degree of innovation. In the regulated market, the average after introduction. This comparison includes all the NCEs in relative launch price is 2.2 for C drugs, 3.9 for B drugs, and our database with a follow-up time of four years or more 4.5 for A drugs. The median relative launch prices are (N ϭ 149). substantially lower than the mean, indicating that prices are The average real price decreases over time for all highly skewed with a few relatively high-priced NCEs. The classes of drugs: for A drugs the decline is 22%, for B median relative launch prices reveal patterns similar to the drugs it is 14%, and for C drugs it is 15%. There is thus average prices, however. The only exception is that the median launch price is somewhat lower for A drugs than for a tendency toward a larger price decline for A drugs than B drugs. This result should be interpreted with some care, for B and C drugs. There is also a tendency toward a given that there are only eight A drugs (with substitutes) in greater decline in prices for chronic drugs than for acute the sample. For drugs in classes B and C, the average drugs. The range of price changes for the individual relative launch prices are about the same for acute and drugs is wide, with some drugs more than doubling in chronic drugs. For A drugs the relative launch prices are price and some decreasing to a third of the introductory lower for acute than for chronic drugs, but again the infor- price. Here we note large differences between the two mative value in this comparison is highly limited due to the markets. In the U.S. data, there is a slight price decrease small number of A drugs. in the most advanced drugs, but the prices for B and C The markets are similar in that the average premia are drugs increase over the four-year period. The price in- increasing in therapeutic value in both markets, and that crease for the least-advanced class C drugs is 23%. The the variability in premia is rather large for all drug range in price changes displays a similar pattern in the classes. The regulated market, however, displays substan- two markets, although it is slightly wider in the Swedish tially more variation in relative launch prices. The aver- data. There is some evidence, however, that prices con- age relative introductory prices are higher in the price- verge across the therapeutic classes in both markets. 304 THE REVIEW OF ECONOMICS AND STATISTICS

TABLE 2.—RATIOS OF INFLATION-ADJUSTED PRICES FOUR YEARS AFTER INTRODUCTION Sweden United States* Class Type N Med. Mean S.d. Range N Med. Mean S.d. Range A Acute 16 0.78 0.83 0.28 0.52–1.65 3 1.11 1.06 0.23 0.81–1.26 Chronic 12 0.75 0.73 0.12 0.44–0.91 10 0.87 0.92 0.17 0.68–1.19 Combined 28 0.77 0.78 0.21 0.44–1.65 13 0.87 0.95 0.19 0.68–1.26

B Acute 20 0.93 0.90 0.33 0.34–2.05 19 0.95 1.00 0.20 0.79–1.59 Chronic 26 0.82 0.84 0.13 0.64–1.10 29 1.18 1.13 0.16 0.84–1.42 Combined 46 0.86 0.86 0.24 0.34–2.05 48 1.09 1.08 0.19 0.79–1.59

C Acute 42 0.89 0.89 0.29 0.40–2.51 16 1.04 1.11 0.26 0.81–1.71 Chronic 33 0.79 0.81 0.10 0.64–1.07 53 1.26 1.26 0.30 0.68–2.01 Combined 75 0.84 0.85 0.23 0.40–2.51 69 1.22 1.23 0.29 0.68–2.01 * Data from Lu and Comanor (1998). Downloaded from http://direct.mit.edu/rest/article-pdf/85/2/298/1613500/003465303765299828.pdf by guest on 30 September 2021

B. Regression Results there is substantially weaker correlation between launch prices and the number of (brand) substitutes in the Swedish The regression results on the determinants of relative data than in the U.S. data. The variable for the number of introductory prices [equation (1)] are summarized in table 3. branded substitutes has the expected sign in both samples. In the sample from the regulated market, the variables A and B indicating therapeutic class have the expected sign However, it falls short of being significant at the 10% level and are statistically significant at the 5% level and at the 1% in the Swedish sample, whereas the corresponding variable level, respectively. This strongly suggests that A and B is highly significant in the U.S. sample. Also, the coefficient drugs are launched at higher relative prices than C drugs. for the number of substitutes is significantly larger in the The coefficient for A drugs is also, as expected, higher than U.S. data, suggesting not only a stronger correlation but also for B drugs, but this difference between A and B drugs is not a more profound influence on launch prices. Moreover, the significant. These results are qualitatively similar to the ones extent of generic competition does not show a significant found by LC, suggesting that therapeutic values is indeed effect in the price-regulated market either. We also observe reflected in launch premia in both market. Interestingly, that the variables indicating acute drugs have different signs

TABLE 3.—DETERMINANTS OF RELATIVE INTRODUCTORY PRICES Sweden United States DG, LG DG LG DG, LG DG LG Variable Excluded Included Included Excluded Included Included Constant 0.485 0.486 0.486 0.295 0.230 0.221 (2.90)a (2.89)a (2.89)a (2.32)b (1.82) (2.32)b

A 0.920 0.920 0.921 0.374 0.388 0.387 (2.45)b (2.48)b (2.44)b (1.76) (1.87) (1.88)

B 0.463 0.463 0.462 0.267 0.293 0.285 (2.99)a (2.98)a (2.98)a (2.13)b (2.39)b (2.34)b

Acute Ϫ0.197 Ϫ0.197 Ϫ0.200 0.334 0.336 0.335 (Ϫ1.37) (Ϫ1.35) (Ϫ1.38) (2.75)a (2.84)a (2.84)a

LNS Ϫ0.124 Ϫ0.124 Ϫ0.119 Ϫ0.322 Ϫ0.386 Ϫ0.376 (Ϫ1.24) (Ϫ1.14) (Ϫ1.12) (3.75)a (4.45)a (4.41)a

DG — 0.004 ——0.332 — (0.02) (2.78)a

LG ——Ϫ0.077 ——0.086 (Ϫ0.15) (2.96)a

N 218 218 218 130 130 130 R2 0.08 0.08 0.08 0.27 0.32 0.32 R2 (adj.) 0.07 0.06 0.06 0.25 0.29 0.29 F 4.79a 3.82a 3.82a na na na Numbers in parentheses are t-values. a Statistically signiﬁcant at the 1% level. b Statistically signiﬁcant at the 5% level. * The data under the U.S. heading is reproduced from Lu, Z. J., and W. S. Comanor, “Strategic Pricing of New Pharmaceuticals,” Review of Economics and Statistics 80:1, 108–118. PHARMACEUTICAL PRICING IN A REGULATED MARKET 305

TABLE 4.—RATES OF CHANGE OF REAL PRICES FOUR YEARS AFTER LAUNCH Sweden DALP, DBLP, DCLP Excluded DALP, DBLP, DCLP Included United Statesa BG, LGS BG LGS BG, LGS BG LGS BG, LGS BG BG Variable Excluded Included Included Excluded Included Included Excluded Included Excluded Constant Ϫ0.202 Ϫ0.203 Ϫ0.201 Ϫ0.175 Ϫ0.174 Ϫ0.168 0.302 0.295 0.300 (Ϫ4.81)b (Ϫ4.81)b (Ϫ4.78)b (Ϫ3.04)b (Ϫ3.00)b (Ϫ2.90)b (6.56)b (6.00)b (7.59)b

A Ϫ0.138 Ϫ0.138 Ϫ0.140 Ϫ0.126 Ϫ0.130 Ϫ0.137 Ϫ0.339 Ϫ0.337 Ϫ0.339 (Ϫ2.62)b (Ϫ2.62)b (Ϫ2.67)b (Ϫ0.85) (Ϫ0.88) (Ϫ0.93) (4.14)b (4.08)b (4.11)b

B 0.032 0.032 0.032 0.003 0.004 0.005 Ϫ0.092 Ϫ0.092 Ϫ0.093 (0.73) (0.75) (0.72) (0.04) (0.05) (0.06) (2.15)c (2.12)c (2.16)c

Ϫ Ϫ Ϫ

Acute 0.069 0.068 0.064 0.076 0.077 0.075 0.055 0.052 0.054 Downloaded from http://direct.mit.edu/rest/article-pdf/85/2/298/1613500/003465303765299828.pdf by guest on 30 September 2021 (1.86) (1.84) (1.74) (1.93) (1.94) (1.90) (1.33) (1.26) (1.29)

DALP ———Ϫ0.008 Ϫ0.008 Ϫ0.008 Ϫ0.007 Ϫ0.008 Ϫ0.007 (Ϫ0.33) (Ϫ0.32) (Ϫ0.31) (0.13) (0.15) (0.13)

DBLP ———Ϫ0.001 Ϫ0.001 Ϫ0.003 Ϫ0.108 Ϫ0.107 Ϫ0.106 (Ϫ0.04) (Ϫ0.08) (Ϫ0.20) (2.50)c (2.47)c (2.46)c

DCLP ———Ϫ0.009 Ϫ0.009 Ϫ0.010 Ϫ0.139 Ϫ0.136 Ϫ0.138 (Ϫ0.76) (Ϫ0.71) (Ϫ0.87) (2.66)b (2.57)b (2.62)b

LNS Ϫ0.020 Ϫ0.016 Ϫ0.011 Ϫ0.008 Ϫ0.014 Ϫ0.010 Ϫ0.077 Ϫ0.075 Ϫ0.076 (Ϫ0.87) (Ϫ0.62) (Ϫ0.48) (Ϫ0.09) (Ϫ0.54) (Ϫ0.40) (2.82)b (2.74)b (2.79)b

BG — Ϫ0.023 ——Ϫ0.032 ——0.002 — (Ϫ0.36) (Ϫ0.47) (0.05)

LGS ——Ϫ0.280 ——Ϫ0.313 ——0.004 (Ϫ1.17) (Ϫ1.28) (0.39)

N 149 149 149 128 128 128 130 130 130 R2 0.09 0.10 0.10 0.09 0.10 0.10 0.25 0.25 0.25 R2 (adj.) 0.06 0.06 0.07 0.05 0.04 0.05 0.21 0.20 0.20 F 3.57a 2.86b 3.14a 2.09b 1.85 2.04b na na na Numbers in parentheses are t-values. a Data reproduced from Lu and Comanor (1998). b Statistically significant at the 1% level. c Statistically significant at the 5% level. in the Swedish and U.S. samples, but the variable fails to be that there is no significant difference between B and C significantly different from 0 in the former. drugs. The variable for acute drugs has a positive coeffi- Finally, we note that the explanatory power of the in- cient, but fails to be significant at the 5% level. The cluded variables is substantially lower in the Swedish data variables for generic competition and brandname competi- than in the U.S. data. The model explains some 30% of tion have no significant effect on the change in prices over the variation in relative launch prices in the latter sample, time. Likewise, the interaction variables between drug clas- while in the former case the corresponding figure is only sification and relative introductory price are insignificant. about 8%. Except for the fact that the prices for class A drugs declined The results for the determinants of the real price changes over the four-year period in both samples, there were no that appear in equation (2) are shown in table 4. other significant similarities between the results found here We estimated the equations both with and without the and the findings in LC. interaction terms between relative introduction prices and Also in this model, the explanatory power of the regres- the drug classifications. Consequently, we may also include sion equations is low for the present sample, with an R2 in the latter the class A drugs that had no substitute drug at value of about 10%. the time of introduction. When we drop the interaction terms, we find that the VI. Discussion variable for class A drugs is negative and highly significant. The variable for class B drugs, however, is not significant. The main purpose of this paper was to compare NCE We interpret this as evidence that the prices for A drugs pricing outcomes in a regulated market (Sweden) with those decline faster over time than prices for B and C drugs, but reported from a market without price regulation (the United 306 THE REVIEW OF ECONOMICS AND STATISTICS

States). In accord with LC, we found that introductory centralized pricing introduces additional randomness in the prices are positively correlated with the degree of therapeu- data. However, we observe that the regulating agency’s tic innovation. The patterns were similar in the two studies, principal concern relates to total expenditures rather than but the average relative introductory prices were higher for relative price levels per se, and this fact may in part explain all classes of drugs in the regulated market. The inflation- the low explanatory power of the regression models. In sum, adjusted price change over time differed substantially be- we note that the differences in price behavior between the tween the studies. In the present analysis, real prices de- U.S. and Swedish markets suggest that the competitive clined over time for all classes of drugs. This result is in forces at work in the two regimes are different. In the United contrast to the substantial price increase that was reported States, price seems to be a key dimension along which firms for C drugs in the United States. We note that the NCE compete, whereas the Swedish data appears to show that price samples used in the Swedish and American studies are taken is not a crucial strategic variable. Clearly, price caps limit the in different time periods. The samples are roughly of the strategic possibilities open to pharmaceutical firms. The data same length, but the present sample starts some ten years on pharmaceutical pricing under the Swedish regulatory sys- Downloaded from http://direct.mit.edu/rest/article-pdf/85/2/298/1613500/003465303765299828.pdf by guest on 30 September 2021 after the one considered by LC. This discrepancy may have tem show this to be true. Furthermore, it generally appears that some effect on the validity of the comparison. some of the pricing constraints imposed by the regulator offset Compared to the U.S. market, the Swedish market is thus some disciplinary effects that market forces bring. This obser- characterized by higher relative launch prices and falling vation is consistent with the empirical evidence observed by real prices over time. This pattern of high relative launch Danzon and Chao (2000b), who maintain that the price regu- prices and declining real prices over time is in line with lation regimes undermine price competition. The low explan- what one might expect from the price-cap regulation in atory power of the demand factor variables (and the associated Sweden. Given that price increases are generally ruled out high variance in introductory prices), and the lack of price under the price-cap regime, it is not surprising that the effects from branded competition in the Swedish market, are regulator compensates the producers by allowing a rela- examples of this. tively high introductory price. If, in addition, existing substitutes have experienced decreasing prices under their REFERENCES product cycles, a natural effect is then that the relative Abbott III, T. A., “Price Regulation in the Pharmaceutical Industry: Prescription or Placebo?” Journal of Health Economics 14 (1995), launch prices are above unity for pharmaceuticals in all 551–565. therapeutic classes. This could partly explain why class C Besley, T., and M. Gouveia, “Alternative Systems of Health Care Provi- pharmaceuticals on average were launched at substantial sion,” Economic Policy 9:19 (1994), 199–258. Caves, R. E., M. D. Whinston, and M. A. Hurwitz, “Patent Expiration, premia. We recognize, however, that the therapeutically Entry, and Competition in the U.S. Pharmaceutical Industry,” more advanced A and B drugs were on average launched at Brookings Papers on Economic Activity: Microeconomics (1991), substantially larger premia than the C drugs. This result can 3–65. Danzon, P. M., Pharmaceutical Price Regulation: National Policies ver- then be interpreted as evidence in favor of therapeutic value sus Global Interests (Washington: AEI Press, 1997). being rewarded by the regulating agency. Danzon, P. M., and L.-W. Chao, “Cross-National Price Differences for Although penetration pricing was not observed for any of Pharmaceuticals: How Large, and Why?” Journal of Health Eco- nomics 19 (2000a), 159–195. the classes of drugs, there was a small tendency for the “Does Regulation Drive out Competition in Pharmaceutical Mar- prices of class A drugs to fall more than the prices of class kets?” Journal of Law and Economics 43:2 (2000b), 311–358. B and class C drugs. This was the only similarity with the Dean, J., “Pricing of Pioneering Products,” Journal of Industrial Econom- ics 17 (1969), 165–179. U.S. market in pricing patterns over time. FASS (Pharmaceutical Specialties in Sweden), (Elanders, Kungsbacka: LC also found that branded substitutes had a substantial La¨kemedelsinformation AB, 1988–1998). negative effect on both introductory prices and price Frank, R. E., and D. S. Salkever, “Generic Entry and the Pricing of Pharmaceuticals,” Journal of Economics and Management Strat- changes over time, and that the extent of generic competi- egy 6:1 (1997), 75–90. tion tended to increase the relative launch prices. Although Grabowski, H. G., and J. M. Vernon, “Brand Loyalty, Entry, and Price there was a (weak) negative correlation between launch Competition in Pharmaceuticals after the 1984 Drug Act,” Journal of Law and Economics 35 (1992), 331–350. premia and the number of branded substitutes, it was not Lu, Z. J., and W. S. Comanor, “Strategic Pricing of New Pharmaceuticals,” significant at any conventional level. As expected, there was Review of Economics and Statistics 80:1 (1998), 108–118. no correlation between the presence of generic substances Nilsson, J. L. G., et al., “Anva¨ndningen av Generiska La¨kemedel i Sverige,” NEPI report (1998). and the launch prices. Moreover, we found no effect of the Reekie, W. D., “Price and Quality Competition in the United States Drug number of substitutes, branded or generic, on the price Industry,” Journal of Industrial Economics 26 (1978), 223–237. changes over time. We interpret this as evidence that the Roseń, A., and B. Beermann, “Rating innovative therapeutic benefits of medicines licensed in Sweden 1987–1997,” International Journal regulation effectively relaxes price competition in the of Pharmaceutical Medicine 13 (1999), 123–126. Swedish market. Statistical Yearbook of Sweden (Stockholm: Statistics Sweden, 1999). The explanatory power in the models applied to the Swedish Drug Market (SDM), Sales and price statistics compiled by The Swedish Association of the Pharmaceutical Industry (1987–1997). Swedish data is (substantially) lower than the ones in LC. WHO Collaborating Centre for Drug Statistics Methodology, ATC Index The higher variability in the Swedish data suggests that with DDD’s 1997 (Oslo, 1997).