An Aggregate Approach to Antitrust: Using New Data and Rulemaking to Preserve Drug Competition

Antitrust policymaking in the United States has a tension at its core. Antitrust law "maintain [s] certain basic rules of competition" as a way to preserve low prices, efficient production, and robust innovation. 1 In regulating a particular type of behavior, a decisionmaker may choose a rule that minimizes costly errors-false condemnations and false exonerations-even at the expense of accuracy in a particular case. Courts, as the actors charged with setting substantive antitrust policy, routinely make such choices. Unfortunately, courts lack the information needed to select optimal rules. Consider, for example, predatory pricing. Antitrust law permits price-cutting to exclude a rival, provided that the price does not fall below cost, on the view that a more aggressive rule yields too many false condemnations. 2 That lenient rule increases false exonerations, but the Supreme Court has concluded that these are unlikely, as predation is "rarely tried, and even more rarely successful." '3 But how does a court come to know this? And is a court the right institution to uncover the answer?

This Article identifies and examines an "aggregation deficit" in antitrust analysis: the troubling lack of information about the frequency and costliness of anticompetitive activity. Aggregation matters for both the substance and institutional structure of antitrust policy. In setting substantive antitrust rules, courts make rough guesses, informed by economic theory and the facts of a specific case, about the distribution of real world economic conduct. What a decisionmaker actually needs is aggregate information on which to base a cost-minimizing substantive antitrust rule. In selecting an antitrust decisionmaker, moreover, we ought to favor the institution that has superior access to aggregate information, all else being equal.

As a vehicle for considering the substantive and institutional dimensions of an aggregate approach, this Article focuses on a single antitrust issue: patent settlements between a brand-name drug maker and its generic rival. Settlements result from a generic drug maker's effort to market a competing version of a brand-name product. The brand-name firm responds with a patent infringement suit that claims its product is protected by one or more patents, and the generic firm counters that the patent is invalid or not infringed by the proposed generic product. The brand-name firm, rather than take a chance that the generic firm might win that argument in court, thereby ending its monopoly on the product, settles the litigation by paying the generic firm to abandon the challenge and delay entry. Does this agreement violate antitrust law?

This question is the most important unresolved issue in U.S. antitrust policy, measured by economic importance and high-level judicial attention. Recent settlements involve some of the world's most important drugs. 4 The largest two settlements alone insulate from competition more than $10 billion in annual brand-name sales. 5 The importance and difficulty of the question has prompted the Supreme Court to seek the 2. Brooke Group Ltd. v. Brown & Williamson Tobacco Corp., 509 U.S. 209, 223 (1993) (declaring that such price cuts are "beyond the practical ability of ajudicial tribunal to control without courting intolerable risks of chilling legitimate price-cutting").

3. Id. at 226 (quoting Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 589 (1986)).

4. Settlements in 2008 included Lipitor (more than $7 billion in annual U.S. sales), Pfizer Inc., Annual Report (Form 10-K) exh. 13, at 18 (Feb. 29, 2008), and Nexium (more than $3 billion), AstraZeneca PLC, Annual Report (Form 20-F), at 55 (Mar. 12, 2008).

COLUMBIA LAW REVIEW Solicitor General's views three times since 2004.6 As of early 2009, the Federal Trade Commission (FTC) is pursuing new litigation challenging settlements over two drugs, 7 new bills aiming to prohibit such settlements have been introduced in Congress, 8 and the President has included a ban on anticompetitive settlements in his annual budget proposal. 9 Identifying the proper scope of liability, however, is not a simple task. Some settlements do not raise pay-for-delay concerns. For other settlements, it is difficult to tell whether a payment was made. Before an optimal antitrust rule can be developed, policymakers need accurate information regarding the scope and nature of the problem. As an initial step toward erasing this deficit, this Article assesses the problem of entrydelaying settlements by aggregating publicly available data about these settlements and considering the overall picture that emerges. This approach draws upon a new dataset of drug patent settlements, developed from a wide range of public sources. The resulting dataset provides, for the first time, a vivid picture of the frequency and distribution of settlement activity. Viewing the settlements collectively permits new insights about enforcement priorities, the optimal substantive rule, and the choice of decisionmaker.

The analysis reveals an evolution in the terms of settlement. Whereas early settlements simply traded cash for delay, modern settlements show sophistication in the means by which payment and delay are provided. One example is the use of side deals, consummated at the same time as settlement of the patent litigation, in which the generic firm contributes unrelated value, such as a separate patent license, ostensibly in exchange for payment. That tactic undermines reliable case-by-case characterization of settlements as collusive or not: In a particular instance, it is difficult to tell whether the brand-name firm's payment is consideration for delay, for the unrelated value, or both. 1 0 6. See Joblove v. Barr Labs., Inc., 127 S. Ct. 1868 (2007) 9. Office of Mgmt. & Budget, A New Era of Responsibility: Renewing America's Promise 28 (2009) ("The Administration will prevent drug companies from blocking generic drugs from consumers by prohibiting anticompetitive agreements and collusion between brand name and generic drug manufacturers intended to keep generic drugs off the market.").

10. For example, in an important test case brought by the FTC, the case-specific approach produced divergent results at each level of review. Compare Schering-Plough Corp. v. FTC, 402 F.3d 1056, 1070-72 (11th Cir. 2005) (concluding that Schering's payment to Upsher-Smith was for value of licenses), and Schering-Plough Corp., 136 F.T.C. 956, 1092, 1241 (2003) (opinion of administrative law judge) (same), with Schering-[Vol. 109:629

An aggregate approach permits us to address the question in a different way. It reveals that these sorts of deals are a frequent component of settlements, but rare outside of settlement. Thus, the overall pattern suggests they provide a disguised means to confer payment. This supports the adoption of a presumption that a brand-name firm's payment to a generic firm, when contemporaneous with a generic firm's agreement to delay entry, is consideration for delay, not for the goods or services acquired in the side deal.

As an institutional matter, the aggregate approach undermines the case for courts as primary antitrust policymakers. A court is largely limited to the facts of a particular case. It lacks the capacity to collect information about the distribution of activity in the economy. To be sure, parties can supply the court with aggregate analyses based upon public information, but public disclosures contain important gaps. Moreover, courts are likely to have trouble processing this information. Agencies have a decisive advantage in collecting and synthesizing aggregate information, given their expertise, access to confidential information about regulated firms, and freedom to examine issues over a long period of time, outside the litigation context. Thus, the analysis suggests that the FTC should do more to exploit its informational advantage as a plaintiff, amicus, and rulemaker.

Finally, the aggregate perspective provides a basis for predicting the success or failure of antitrust enforcement over time. As applied to settlements, the prediction is pessimistic. Settlement has continued to evolve-even beyond side deals-in response to the enforcement emphases of particular litigants and courts. Settling parties have been able to achieve the same entry-delaying effect of the earliest settlements, while devising new disguises for payment or even the very existence of agreement. As litigants respond dynamically to judicial scrutiny with new and complex settlement structures, existing antitrust institutions have trouble keeping up.

The Article proceeds in four parts. Part I introduces the pay-for-delay settlement problem and the aggregation deficit in antitrust. Part II draws upon the new dataset, outlining the scope and changing structure of entry-delaying settlements, and spells out how these features recommend making the settlement issue an enforcement priority. Part III examines side deals from an aggregate approach, explaining why they should be presumed to convey payment when accompanied by an agreement to delay entry. Finally, Part IV addresses the question of institutional choice. It first shows why courts make poor aggregators, and proceeds to consider how agencies can help fill the gap by aggregating data and promulgating rules. I. THE PAY-FOR-DELAY SETTLEMENT PROBLEM Part I.A describes the pay-for-delay settlement problem. Although settlements have received a great deal of attention, almost all of it has focused upon the theoretical issues raised in individual cases, at the expense of important factual questions that also arise. Part I.B describes this neglect and its connection to the larger problem of an aggregation deficit in antitrust.

Pay-for-delay settlements restrict a particular kind of competition between brand-name and generic firms. The process begins when a brandname firm launches a new drug pursuant to the Hatch-Waxman Act, the industry-specific scheme that regulates pharmaceutical competition." 1 Once the brand-name firm places a patented drug on the market, a generic firm may seek to launch a competing version of the same drug, asserting that any applicable patents are invalid or not infringed. 1 2 The assertion is contained in an Abbreviated New Drug Application, or ANDA, that is filed with the Food and Drug Administration (FDA). 13 If the filing is successful, the generic firm can launch a competing product without repeating the costly safety and efficacy studies that the FDA requires as a condition of brand-name approval.

The first generic firm to file an ANDA is entitled, upon FDA approval, to a 180-day exclusive right to market a generic version in competition with the brand-name firm, effectively creating a duopoly during that period. 14 For some drugs, multiple generic firms file ANDAs on the same day, and thus share the exclusivity entitlement.' 5 For others, one generic firm files at least a day before the others.' 6 In response to the 11. See 21 U.S.C. § 355(b) (2006) (providing for launch of new drug after demonstration of safety and efficacy). This account is a simplification. For more details, see C. Scott Hemphill, Paying for Delay: Pharmaceutical Patent Settdement as a Regulatory Design Problem, 81 N.Y.U. L. Rev. 1553, 1564-66 (2006).

12. See § 355(j) (2) (A) (vii) (IV) (requiring certification to FDA and notification of rightsholder that any applicable patents are invalid or not infringed).

13. Id. The ANDA contains a so-called "Paragraph IV" certification that the applicable patent protection is invalid or not infringed. Not all ANDAs contain such a certification; often, the generic firm is content to wait until patent expiration before entering. FTC, Generic Drug Entry Prior to Patent Expiration 10 (2002), available at http://www.ftc.gov/os/2002/07/genericdrugstudy.pdf (on file with the Columbia Law Review) [hereinafter FTC, Generic Drug Entry] (reporting ninety-four percent of the more than 8,000 ANDAs filed between 1984 and 2000 lacked a Paragraph IV certification). As a general matter, the ANDAs discussed in this Article contain Paragraph IV certifications.

14. § 355(j)(5)(B)(iv). The "duopoly" characterization ignores the effect of authorized generics, discussed infra Part III.A.2. 15. Ctr. for Drug Evaluation & Research, FDA, Guidance for Industry: 180-Day Exclusivity When Multiple ANDAs Are Submitted on the Same Day 3-4 (2003), available at http://www.fda.gov/cder/guidance/5 7 10fnl.pdf (on file with the Columbia Law Review).

16. Multiple first filers may result when the brand-name drug contains no "active moiety" already approved in another New Drug Application, or NDA. In that case, the [Vol. 109:629 ANDA, the brand-name firm may file a patent infringement suit to establish validity and infringement. This pattern-launch, challenge, sue-is typical for major drugs. 1 7 The two drug makers have a powerful incentive to settle. For a blockbuster drug with billions of dollars in annual sales, a brand-name firm has billions to lose from generic competition. Moreover, entry hurts the brand-name firm more than it helps the generic firm. Entry lowers total producer profits by introducing price competition, particularly once other generic firms are free to enter after the 180-day period ends. 18 There is therefore a large gain from trade for the two firms. A settlement in which the brand-name firm pays the generic firm, and the generic firm agrees to delay entry, is profitable for both firms. Because later filers generally have much less incentive to challenge a brand-name drug patent, including no eligibility for the 180-day period, buying off the first filer is an effective means to remove the most potent entry threat. 1 9 Such settlements, if they include payment, reduce expected static consumer welfare. Early competition benefits consumers by lowering drug prices sooner. The consumer benefit is probabilistic, since it is not certain that entry would occur; the brand-name firm might win the suit. Settlements without payment reflect the perceived strength of the patent. For example, a generic firm's fifty percent chance of success would yield, roughly speaking, an entry date halfway between immediate entry and patent expiration. 2 0 That result is equal to the average result of litigation, in which the consumer has a fifty percent chance of enjoying the full benefit of immediate competition and a fifty percent chance of re-FDA must not accept an ANDA for four years after NDA approval. § 3550) (5) (F) (ii); 21 C.F.R. § 314.108(b) (2006). Aside from giving the brand-name firm several years of protected sales before a generic challenge can commence, it also affords generic firms plenty of time to devise a workaround strategy. For other drugs, by contrast, the generic firms are in an immediate race to devise a plausible legal and pharmaceutical strategy, and the firms will usually differ significantly both in their assessment that the challenge is sufficiently promising tojustify an investment, and in their skill and speed in developing a workaround. 17. For example, of the fourteen best-selling drugs of 2005, see Matthew Herper, The Best-Selling Drugs in America, Forbes, Feb. 27, 2006, at http://www.forbes.com/2006/02/ 27/pfizer-merck-genentech-cx mh 0224topsellingdrugs.html (on file with the Columbia Law Review), twelve faced pre-expiration patent challenges: Lipitor, Nexium, Prevacid, Plavix, Zoloft, Norvasc, Seroquel, Effexor XR, Zyprexa, Singulair, Protonix, and Risperdal. The two exceptions are Zocor and Advair Diskus. This calculation does not include biologic drugs not subject to the Hatch-Waxman regime.

18. For details and caveats, see Hemphill, supra note 11, at 1580-82. 19. See id. at 1585-86 (noting small incentive to file and vigorously pursue challenge); id. at 1605-06 (discussing free-rider problem among later filers resulting from nonmutual issue preclusion, particularly in invalidity challenges). In some instances, the settlement also creates a bottleneck for later filers, as discussed infra Part II.C.2. 20. This is an oversimplification, because it ignores the effect of the exclusivity period, which is a source of compensation for the generic firm. See infra notes 91-102 and accompanying text (describing use of exclusivity period in settlements); see also Hemphill, supra note 11, at 1588-94 (describing exclusivity period as source of compensation).

ceiving no benefit. By contrast, bargains that reflect not only perceived patent strength but also payments from brand-name to generic manufacturers will induce the generic firm to accept a later entry date, which decreases consumer welfare. Thus, a pay-for-delay settlement transfers wealth from consumers to drug makers, in the form of continued high pharmaceutical prices, with brand-name firms sharing a portion of that transfer with the generic firm. The higher price also alters the purchase decisions of consumers and insurance providers, introducing an additional welfare loss. 2 1

As I have argued elsewhere, the consumer-disregarding effect of pay-for-delay settlements requires their condemnation as a violation of antitrust law. 2 2 Allocating markets in this fashion is a restraint on trade in violation of section 1 of the Sherman Act, 23 and may also be condemned as illegal monopolization. 24 It is therefore no surprise that the FTC-the federal agency charged with antitrust enforcement in the pharmaceutical industry-has brought numerous cases, often together with state attorneys general, arguing that certain pay-fordelay settlements violate antitrust law. 25 Private parties have done so as well. 26 21. Assessing this welfare loss is complex. In an ordinary market, setting a price above marginal cost produces an allocative distortion and accompanying welfare loss for consumers, because consumers who value the good above its marginal cost, but below the prevailing price, are deflected to less desired substitutes. To the extent that public and private insurance secures the purchase of a drug, this distortion is reduced, though it is not eliminated (as insurance is incomplete). Moreover, the higher price produces new distortions (and hence inefficiency) in the decisionmaking process of the insurance provider, through decisions to charge higher premiums and not to reimburse drugs whose value exceeds their marginal cost. In a similar manner, the existence of incomplete insurance affects the assessment of the size of the transfer.

22. Settling parties have offered a variety of defenses. 27 The most fundamental is that permitting settlement increases the brand-name firm's profit, and hence its expected reward for developing innovative drugs, the marketing of which provides great benefits to consumers. Put another way, the static harm of settlement from high prices today must be weighed against the dynamic benefit of more and better drugs in the future. The potential scope of this argument is extremely broad: Any practice currently prohibited by antitrust law, as practiced by innovators seeking to increase their profits, could be defended upon this ground. Even simple price fixing could be excused. In general, antitrust lacks any such exemption for collusive behavior. 28 The case for making an exemption is particularly weak where, as here, the increase in innovative incentive from delaying competition is partially offset by the necessary payments to the generic firm. 29 Settling parties have offered several further objections. They assert that the suppressed competition is not cognizable because it is merely probabilistic. 30 That objection ignores the fact that the suppressed entry subject to antitrust regulation is almost always probabilistic. 3 1 A second objection is that settlements in other industries are similarly consumerdisregarding, raising the specter of a widespread expansion of liability if these settlements are prohibited.

patent settlement is a real possibility in other industries, and to that extent, antitrust liability may be warranted there too. In addition, the Hatch-Waxman Act reflects a specific effort to promote consumer access through litigated challenges, a feature that makes the case for prohibition particularly strong in this industry.1 3 A third objection-that prohibiting certain settlements increases litigation costs-is overwhelmed by the much larger adverse effect on consumer welfare.

Courts have tended to reject antitrust liability for brand-generic settlements. These courts have accepted, as a doctrinal matter, a maximalist view of the patent right. Most appellate courts that have considered the issue have adopted the view that any settlement is permissible, provided it restricts no more entry than the nominal scope of the patent if valid and infringed. 3 4 As a result, brand-name firms are effectively permitted to buy private term extensions to their patents. The maximalist view thus produces the absurd result that an ironclad patent and a trivial patent have the same exclusionary force. Each can support a settlement that restricts generic entry until the nominal expiration date of the patent.

The maximalist perspective also ignores the fact that the nominal scope of the patents at issue, particularly the expiration date of the lastexpiring patent, is highly malleable. A sophisticated brand-name drug maker can produce a steady stream of patents, with successively later expiration dates, which in turn support a settlement date that is even later than the expiration of effective protection. A settlement involving the blockbuster drug Lipitor, Pfizer's most important product, provides an example.

as to the first patent but losing as to the second. 3 6 Analysts therefore expected entry in March 2010, or at the very latest in June 2011. 3 7 However, when the parties eventually settled, generic entry was set for November 2011, later than the expiration of either patent. 38 The parties defended this result on the ground that, shortly before settlement, Pfizer had also sued Ranbaxy on two minor patents that expire in 2016. 39 The main effect of the inclusion of these patents was to permit the parties to choose an entry date later than the expiration of the two main patents at issue.

Whether pay-for-delay settlements violate antitrust law has generated tremendous scholarly interest and a wide variety of responses. 40 The maximalist view of the patent right has been rejected by the FTC, senior officials of the Department of Justice Antitrust Division, 4 1 and the Solicitor General, 42 but they, like commentators, have a variety of views on the subject. Some take the view that all settlements that combine payment with delayed entry are per se violations of antitrust law. 43 Others would impose a presumption of illegality. 44 Still others say that the matter should be judged through a more detailed examination of the strength of the patent, compared to the details of the settlement. 4 5 The stronger the patent, the less troubling a long delay in entry would be.

Antitrust law is not the only way to address the pay-for-delay settlement problem. For example, Congress could modify or eliminate the 180-day exclusivity period, particularly for settling parties, or provide a means and incentive for drug purchasers, including the government, to challenge pharmaceutical patents. Such changes could address the incentives that give rise to the pay-for-delay settlement problem in the first place. As an alternative, settlements could be challenged at the moment they are reached, by requiring the court conducting the patent infringement case to approve the settlement using procedures akin to those employed in class actions to prevent collusive settlements. Private "objectors" or the FTC could be recruited to try to persuade the court that the settlement ought to be rejected.

Putting aside the question of political feasibility, however, such changes would not determine the legal status of the many settlements [Vol. 109:629 that have already been reached. Thus, antitrust law is a necessary component of any complete resolution of the pay-for-delay issue.

Beyond this theoretical question-do pay-for-delay settlements violate antitrust law?-there is a set of factual questions that must be answered. For example, how frequently do pay-for-delay settlements occur? Knowing the answer is necessary to decide whether to make the settlement issue an enforcement priority. A second factual question arises in many modern settlements. If settlement and delay occur as part of a larger set of transactions between the two firms, how do we know that the payment was made in exchange for delay, rather than for some other valuable consideration? Often, this is a difficult question. In the only case involving a side deal that has been fully litigated so far, attempts to determine whether the particular settlement was anticompetitive produced divergent results at each level of review. 46 These factual questions have been neglected by scholars so far.

This gap in our understanding of modern settlement practice exemplifies a general problem in antitrust enforcement. Given a theoretical model of anticompetitive behavior, true under specific factual circumstances, how do we establish with confidence that those circumstances are present in a particular case? If that determination is imperfect, how do we identify a cost-minimizing rule-for instance, that alleged predation is reviewed leniently because predation is "rarely tried, and even more rarely successful," 4 7 or that resale price maintenance ought to be accorded rule of reason treatment because its procompetitive uses are not merely "infrequent or hypothetical"? 48 Because a court lacks the capacity to independently collect the information necessary to develop an optimal rule, it relies upon others, including academics and other governmental institutions. In considering predation, for example, the Supreme Court has explicitly relied upon a "consensus among commentators" that the practice is rarely tried or successful. 49 If the external consensus changes, the Court suggests, so too may the substantive rule. 50

For example, Justice Breyer, dissenting from the Court's recent decision to end a longstanding per se ban on resale price maintenance, thought any change should await solid information about "how often are harms or benefits [from the practice] likely to occur." 5 1 He also questioned how readily the two can be distinguished; in other words, "[h]ow easy is it to separate the beneficial sheep from the antitrust goats?" 5 2 Such information must be supplied by others, if it is to be collected at all, since courts, unlike Congress and the FIC are not "well-equipped to gather empirical evidence outside the context of a single case." '5 3 Real world evidence about the frequency and distribution of anticompetitive activity helps to build the requisite consensus among commentators. Such work has furthered our understanding of predation, 54 vertical contracting, 55 and other competitive practices. Industry-specific analyses have been important too. 56 In addition to measuring the aggregate costs of a class of antitrust violation, this study adds a distinctive dimension: the effort to understand the evolution of a practice over time. Understanding this evolution provides evidence about how well existing antitrust instruments can be expected to cope. Frequent or rapid mutations in the practices of regulated firms raise doubts about whether common law processes can effectively regulate those practices. [Vol. 109:629

Whether by legislative reform orjudicial decisions, "economic policy must be contrived with a view to the typical rather than the exceptional," 5 7 to use George Stigler's apt phrase. Both legislators and judges would benefit from a clear idea of how often and in what form settlements occur, and how effective we can expect judicial management to be. This is a fitting moment to examine real world evidence of settlements, before the Supreme Court or Congress establishes a new rule. The Supreme Court has not weighed in on the settlement question, but if and when it does, its rule will be difficult to undo, thanks to the infrequency of antitrust review, the operation of stare decisis, and a fear of upsetting reliance interests. 58 The next Part begins the examination necessary to formulate an optimal rule for pay-for-delay settlements.

An agency such as the FFC is well positioned to fill these informational gaps. The agency has a statutory mandate to collect, study, and publish information about particular industries. It has general authority to require firms to divulge confidential information relevant to antitrust policyrnaking. 59 In the particular context of settlement, the FTC's position is even stronger: It has unique access to the details of every brandgeneric settlement since December 2003, due to drug makers' special statutory obligation to file all such settlements with the agency. 60 This aggregate information complements other sources of FTC expertise developed and used in litigation, congressional testimony, and public hearings. 6 1

The FFC sometimes uses this advantage to good effect. In 2002, the agency published an important survey of brand-generic drug competition, drawing upon information supplied by drug makers under FTC compulsion as well as information collected independently by the FDA. 62 That study indicated the importance of the pay-for-delay settlement problem and made a variety of policy recommendations. But there has been

no follow-up to the 2002 study; more generally, industry studies-once a staple product of the FTC-have become less frequent. 6 3 The FTC's conclusions, based on its aggregate information, can be deployed in a variety of policymaking settings. In the case of the 2002 study, the conclusions were used in amicus briefs, legislative advocacy, and litigation brought by the Agency. 64 But in each of these settings, the Agency is essentially supplying its information to an external decisionmaker. 6 5 The Agency has available to it a more aggressive option, however, which emphasizes the FTC's role as a decisionmaker in its own right: antitrust rulemaking.

The FTC possesses the power to promulgate rules with the force of law that are subject to deference under Chevron U.S.A., Inc. v. Natural Resources Defense Council, Inc. ,66 which imposes upon courts a "duty to defer to reasonable agency interpretations . . . [of an ambiguous] statute that an agency is charged with administering." 6 7 At first, this assertion may seem startling, because its power is seldom used. The Agency has promulgated just one such antitrust rule, and that was more than forty years ago. 68 Since then, the Commission has considered promulgating antitrust rules from time to time, but has never followed through. 69 In Part IV, I argue that the FTC's aggregation advantage is a reason to favor antitrust rulemaking, and that pay-for-delay settlement is an attractive candidate for a rule. But first, I will lay out what an aggregate approach can tell us about drug patent settlements. This Part introduces an aggregate perspective to the issue of settlements between brand-name and generic drug makers. Part II.A outlines the data collection effort. Part II.B shows the magnitude and continuing importance of settlements with delayed entry. It proceeds to describe three sources of evolution in the form of settlement, and the effects of each. Part IL.C elaborates an initial payoff from the aggregate approach: a clear sense that settlements ought to be considered a top priority for antitrust enforcement.

To examine the frequency and evolution of brand-generic settlements since 1984, I collected a novel dataset. The object was to identify and synthesize all public information about the frequency and terms of settlement. The effort drew upon press releases, trade publications, financial analyst reports, analyst calls with management, court filings of patent and antitrust litigation, SEC filings, FDA dockets, and FTC reports. 7 0 For nine settlements, the actual settlement agreement was available. 7 ' In addition to the terms of settlement, I recorded the annual sales 70. The broadest search was a review of all articles in the Factiva database mentioning "settlement" and a "new drug application." The database includes newspapers, magazines, trade journals, press releases, company presentations at analyst conferences, and transcripts of calls between company executives and equity analysts. The search included linguistic variants of "settlement" and the abbreviations "NDA" and "ANDA." The Factiva search found a number of settlements that were not evident in other sources, such as analyst reports. In many cases, articles in Factiva filled in important settlement details.

The FTC's 2002 report provided a detailed accounting of terms for the earliest settlements, with the drug name disguised. FTC, Generic Drug Entry, supra note 13. In figures at the time of settlement and noted whether the generic firm was eligible for the exclusivity period. 72 I also determined whether a major provision of the Medicare Prescription Drug, Improvement, and Modernization Act of 2003 (MMA) applied to the settlement. 73 To be included in the set, the agreement must pertain to patent litigation resulting from an ANDA filing by a generic firm. 74 The search period extended from 1984, when the Hatch-Waxman Act was passed, through August 2008, and therefore ignores subsequent settlement activity. 7 5

This work yielded information for 143 settlements involving 101 brand-name drugs. For 28 drugs, the brand-name drug maker settled with multiple generic firms. Multiple settlements can be the result of settlements with multiple first filers sharing the exclusivity entitlement, 76 or settlements with later filers who lack eligibility for the exclusivity period. Although the focus of the subsequent analysis is settlements with first filers, in some cases settlements with later filers can raise pay-for-delay issues as well. 7 7 72. To determine eligibility, I assessed whether the drug was subject to the exclusivity period, whether the settling generic firm was a first filer, whether any exclusivity eligibility had already been triggered at the time of settlement, and whether the settlement itself included a forfeiture of retained exclusivity. The second determination is the most difficult, because the FDA considers the identity of the first filer to be confidential information, and because there are often multiple first filers. I based the determination on FDA letters granting ANDA approval with exclusivity (which are not confidential), generic-firm press releases reporting presumed first filer status, and a comparison of complaints in patent suits with FDA reports of the date of a first ANDA filing, which is not confidential.

73. Pub. L. No. 108-173, 117 Stat. 2066 (2003). The relevance of this fact is discussed infra Part II.C.3. 74. This criteria rules out, for example, an agreement over Ovcon 35, which was not a patent dispute but did feature an agreement that was challenged as anticompetitive by the FTC. See supra note 25. It also omits drugs, such as Advicor, where a settlement as to another drug discouraged the filing of an ANDA in the first place. See, e.g., Niaspan Agreement, supra note 71 (providing eventual entry as to Advicor, a drug on which the generic firm had not yet filed an ANDA).

75 77. This applies, for example, to the settlements involving K-Dur, AndroGel, and Hytrin discussed infra note 105. In these cases, a later filer received an entry-delaying settlement in addition to the first filer.

Several checks confirm that the dataset contains nearly all significant settlements that delay entry. 78 The dataset oversamples settlements that restrict entry for important drugs. Important drugs receive more extensive coverage in public disclosures, and settlements that restrict entry tend to receive more attention. Thus, omitted settlements are likely to be for minor drugs, or settlements that had no effect on entry. For those settlements in the dataset, publicly available information contains significant gaps. In particular, price terms are normally omitted, and detailed settlement terms are sometimes missing. Even with these limitations, the new dataset is a useful tool for examining the extent and evolution of settlement; indeed, it may be the most comprehensive examination of brand-generic settlements until and unless the FTC uses its power of compulsion to produce a complete dataset.

Of the 143 settlements in the dataset, 60 include both delayed generic entry and possible contemporaneous provision of value by the brand-name firm. The 60 settlements involve 51 out of the 101 drugs in the dataset. For an additional two drugs, the Hatch-Waxman dispute was resolved through acquisition: The generic firm bought out the brandname firm's rights to the drug, thus ending the possibility of competition between the two. 79 (Neither deal was challenged by the FTC, however, suggesting that the firms lacked market power in the first place.) As to the remaining 48 drugs, my data collection effort identified no pay-fordelay issue, and so far as I know, the settlements raise none. 8 0 Some such 78. For example, the FIC catalogued 14 troubling settlements in 2002, but did not name names: 8 cash or side deal settlements, 2 "supply agreements," and 4 retained exclusivity settlements. FTC, Generic Drug Entry, supra note 13, at 34. Of these, I can match 7, 2, and zero settlements, respectively, to my dataset. Of 80. Of the 81 settlements in this category, 67 pertain to 48 drugs whose settlements appear to raise no pay-for-delay issue. The remaining 14 settlements pertain to drugs in which the brand-name firm reached at least one other settlement that does raise a pay-for-

settlements include an agreement on the generic entry date, without any payment. These negotiated outcomes likely reflect the perceived strength of the relevant patents. Their existence demonstrates that settlement without payment is feasible. 8 Table 1 summarizes the three categories of agreement. For the 51 drugs raising pay-for-delay issues, payment and delay take a variety of forms. For 21 of the 51, the compensation was wholly or partly monetary. 8 2 Sometimes the payment was an open conferral of cash. For other drugs, the possible payment was embedded within a more complicated transaction. The caveat "possible" is used because in some cases public information leaves it unclear whether the settlement included compensation. 83 These 21 drugs are listed in Table 2, together with details about the various forms of payment, which are explained later in this Article. On average, they had annual U.S. sales, measured in the year of settlement and adjusted for inflation, of $1.3 billion.

The 21 drugs include blockbusters such as Lipitor (more than $7 billion in annual sales) and Nexium (more than $3 billion). Five drugs with annual sales exceeding $2 billion account for more than two-thirds of the total, measured by annual sales. More than half are new versions of existing therapeutic agents, whose patents are generally thought to be weaker because they tend to be obvious (and hence invalid) and are easily worked around. 8 4 Some of the settlements in Table 2 have lapsed, and generic entry has occurred, while others continue to block entry as of delay issue. To avoid double-counting, these latter settlements are not included in the number of drugs in this category.

81. See, e.g., Jon Leibowitz, Op-Ed., This Pill Not to Be Taken with Competition: How Collusion Is Keeping Generic Drugs off the Shelves, Wash. Post, Feb. 25, 2008, at A15 (pointing to feasibility of no-delay settlements as supporting conclusion that pay-for-delay settlements should be prohibited). Even if no-delay settlements were infeasible, however, the main reasons to condemn pay-for-delay settlements would still hold.

82. This category includes "underpayment" settlements discussed infra Part IIl.A.2. 83. This issue is explored in more detail infra Part III. 84. This group consists of Sinemet CR, K-Dur, Naprelan, Niaspan, Effexor XR, Propecia, Adderall XR, AndroGel, Wellbutrin XL, Nexium, and Aggrenox. Even for those drugs that are not new versions, some of the relevant patents are noticeably weak. For example, Altace was protected by a patent not on the basic compound, but an enantiomer, and was subsequently invalidated. Aventis Pharma Deutschland, GmbH v. Lupin, Ltd., 499 F.3d 1293, 1295, 1303 (Fed. Cir. 2007). Provigil is protected not by a compound patent, which expired, but by a particle-size patent. Provigil Complaint, supra note 7, at 2.

[Vol. 109:629 March 2009. Ten drugs in the latter category account for about $17 billion in annual sales.

The effect of delayed entry can be enormous. For the questionable settlements in Table 2, a one-year delay in generic entry represents, under conservative assumptions, a transfer from consumers to producers of about $14 billion. 8 5 One of the 21 settlements, Plavix, never took full effect; 8 6 with Plavix removed, the transfer from a one-year delay is $12 billion. Whether the one-year benchmark is an overestimate or an underestimate is often difficult to assess in a particular case using public information. Part of the total delay caused by settlement is attributable to the strength of the patent itself, rather than payment. Since the pre-expiration period covered by settlement is several years-the average period, weighted by sales (and excluding Plavix), is 4.1 years-the benchmark is likely conservative.

A more nuanced figure might be developed by offering a specific prediction about what would have happened in each case absent the settlement. The particular circumstances of a settlement can provide important indications of the likely alternative outcome. A weak patent, and likely early entry, might be identified by an analysis of the patent's validity and scope, or inferentially by a large payment. Another basis for inference is preparations by a generic firm to launch "at risk"-that is, to enter even before a court has ruled on invalidity or noninfringement. Launches at risk suggest that the patent protection is weak, because the generic firm does not fear the prospect of damages (which would exceed the generic firm's profits if imposed) or a preliminary injunction (which would spoil the expensive preparations for a generic launch).

For some drugs, public statements by management or the expectations of financial analysts help to provide a specific measure of delay. In the case of Provigil, for example, the drug maker's CEO said that due to settlements, "We were able to get six more years of patent protection. That's $4 billion in sales that no one expected. '87 The CEO's statement reflects the firm's pre-settlement expectation of entry in 2006,88 and set-85. Suppose generic entry achieves 75% penetration and that the generic product is priced at a two-thirds discount, relative to the brand-name drug. These figures are a simplification, because in reality, penetration and the discount (particularly during the 180-day period) are smaller at first, but quickly increase. Under these assumptions, the avoided transfer is one-half of annual sales, or $661 million per drug. Across 21 drugs, the total is about $14 billion. This figure does not include welfare losses caused by pricing distortions. See supra note 21. [Vol. 109:629 tlements delaying entry until 2012.89 In the case of Lipitor, the settlement delayed anticipated entry by nearly two years. 90 Overall, the $12 billion benchmark estimate is likely to be conservative.

For settlements involving 25 drugs, the brand-name firm compensated the generic firm as part of an entry-delaying agreement, but the compensation was not monetary. Instead, compensation took the form of retained exclusivity. As explained in Part I, the 180-day period is valuable to the generic firm. One hundred eighty days of duopoly is worth hundreds of millions of dollars in the case of a blockbuster. 9 1 The entitlement can also be sold to another generic firm. 9 2 The value of this opportunity, however, is discounted by the uncertainty that the generic firm might lose the litigation, and thus never enjoy the exclusivity pe- 92. A generic firm can either selectively waive its entitlement in favor of a particular later filer, or relinquish it entirely. This is a profitable strategy where the firm with the entitlement has been unable to secure FDA approval-for example, due to difficulties in formulating or manufacturing the product-and a later filer is ready to go to market, but for the fact that it is "bottled up" behind the first filer. For a fuller explanation of this bottleneck, see infra notes 118-120 and accompanying text. Selective waiver has been permitted for numerous drugs, including Zantac, Zoloft, and Wellbutrin XL. 2008)) (explaining FDA position that "applicant may selectively waive its exclusivity only after the 180-day exclusivity period has begun to run with the occurrence" of favorable court ruling or commercial marketing).

riod. 93 A brand-name firm's agreement to drop the patent fight-an arrangement that does not forfeit eligibility 94 -is valuable to the generic firm because it raises the probability of enjoying the exclusivity. These 25 drugs are listed in Table 3.95 The ability to settle with retained exclusivity disrupts the alignment of interests between the generic firm and consumers. Ordinarily, late en-93. Other risks include the possibility that a later-filing generic firm wins a patent suit, thereby triggering the first filer's exclusivity period before the generic firm secures FDA approval, or that the patent expires before the generic firm wins the suit.

94. Settlement does not remove entitlement to the exclusivity period. See infra Part II.C.2 (discussing factors influencing settlements).

95. The list omits settlements where there was no delayed entry or where the available data was ambiguous about continued entitlement to the exclusivity period. Entry as to one drug on the list, Exelon, was not disclosed, but was assumed to be at least 180 days prior to patent expiration. This list is underinclusive. For example, my data collection identified none of the four early retained exclusivity settlements discussed in the FTC report. See supra note 78 (comparing this Article's dataset with that of the FTC report).

[Vol. 109:629 try dates are bad for consumers, but also bad for the alleged infringer, whose profits are a function of the amount of time on the market, and who therefore can be expected to fight for an earlier entry date. Here, by contrast, the generic firm cares more about protecting its 180-day duopoly entitlement, and less about when exactly that entry occurs. It is therefore willing to trade a later entry date for the better chance to enjoy the 180 days. 9 6 Meanwhile, consumers and taxpayers finance the continued sale of drugs at the higher, brand-name price. This argument has an important limit. If the generic firm's pre-expiration entry lasts for less than 180 days, then its profits are, roughly speaking, linearly increasing as it pushes for an earlier entry date. In that case, the alignment between the generic firm and consumers is more nearly maintained. Of the 25 drugs listed in Table 3, 7 have entry dates so late that they have less than 180 days of exclusive sales. 97 For the remaining 18 drugs, the misalignment critique applies.

The 25 drugs have average annual sales of $580 million. Of these, the 18 drugs with "full" exclusivity have average sales of $442 million. If guaranteed exclusivity induces a delay of one year for each of these drugs, the transfer, using the same calculus described above, would be about $4 billion.

The preservation of exclusivity can take a second form. In some cases, a generic firm wins a patent challenge, but is blocked from approval by a second patent that the generic firm either did not challenge at all, or challenged unsuccessfully. In such a case, the generic firm "wastes" the exclusivity resulting from that partial victory, which is triggered and expires while the generic firm is blocked from entering by the second patent. 98 Once the second patent expires, the generic firm enters, but without exclusivity. A generic firm can avoid wasting its exclusivity by abandoning its challenge, and agreeing to enter with exclusivity upon the expiration of the second patent. 99 This benefits the brandname firm, and harms consumers, for the same reason: Prices are higher 96. For a detailed analysis, see Hemphill, supra note 11, at 1588-94. 97. Frequently, this occurs when a brand-name firm secures a six-month pediatric extension that is tacked onto the end of the patent term. during the (preserved) duopoly exclusivity period than with full competition from other generic firms.

In addition to the drugs for which the only form of compensation is retained exclusivity, most of the drugs in Table 2, after the first five, have secured an assured 180 days of generic sales. 10 0 Other settlements explicagreed to enter with exclusivity upon the expiration of the basic patent. See Zoloft Agreement, supra note 71.

Barr's challenge to Prozac raised a similar possibility. See Barr Labs., Inc., Amendment to a Previously Filed 10-K405 (Form 10-K405/A), at 10-11 (May 15, 2001) (noting that 180-day period could be wasted if challenge to one patent succeeded, triggering the generic firm's exclusivity as to it, while a second patent blocked FDA approval of the generic drug). As it turned out, the patent had expired by the time exclusivity was triggered, and only six days remained of the associated pediatric exclusivity period. The premature triggering question was limited to the six-day overlap: Was the 180-day period truncated by the overlap with pediatric exclusivity? Congress passed a statute providing for the full benefit of exclusivity in such circumstances, and generic entry was protected for the six days. Best Pharmaceuticals for Children Act (BPCA) § 10, 21 U.S.C. § 355a(k) (2006); Press Release, Barr Labs., Inc., Barr Confirms Prozac Exclusivity Runs Until January 29 (Jan. 9, 2002) (announcing letter from FDA stating that BPCA .extends" exclusivity by amount of overlap, in this case to January 29, 2002).

The Lipitor settlement appears to contain another variant. When Ranbaxy won its challenge to one patent in the Federal Circuit, this triggered exclusivity, but prematurely, since the other valid and infringed patent prevented FDA approval. Pfizer, Inc. v. Ranbaxy Labs., 457 F.3d 1284, 1290-92 (Fed. Cir. 2006). The combined result would have been to permit entry without exclusivity in March 2010. (The patents expiring in 2016 were never listed in the Orange Book, and did not affect that result.) However, Pfizer had three more Orange Book-listed patents in reserve, on which Ranbaxy was likely the first ANDA filer but Pfizer did not sue. Under pre-MMA law, each patent provided a fresh opportunity for exclusivity. See (2000), to provide separate exclusivity for separate patents). By declining to sue Ranbaxy on these patents, Pfizer preserved Ranbaxy's exclusivity despite the initial trigger, a preferable result for both parties. The MMA replaced this "patent-by-patent" approach to exclusivity with a single opportunity for each product. 21 U.S.C. § 3550) (5) (B) (iv) (I) (2006) (making exclusivity available only to "first applicant"); id. § 355(j) (5) (B) (iv) (II) (bb) (defining "first applicant" by reference to drug, not patent); see also John. R. Thomas, Pharmaceutical Patent Law 367 (2005) (explaining post-MMA scheme).

100. The first five settlements included no pre-expiration entry, for reasons discussed infra. In the case of Lamictal and AndroGel, the preserved exclusive sales is a synthetic construct achieved by contract. For Lamictal, the 180-day period expires when the relevant patent expires, and the generic firm is granted a license during the pediatric exclusivity. Five pay-for-delay settlements fit neither of these categories. Three are "interim" agreements, which restrict entry while the patent infringement suit is pending but do not resolve the suit. After such agreements were targeted for antitrust enforcement in the late 1990s, 1 03 parties turned to the monetary and retained exclusivity settlements discussed above. The remaining two settlements are supply agreements in which the generic firm did not retain exclusivity eligibility.' 0 4 A summary of the four categories of pay-for-delay settlements appears in Table 4. Again, the number of settlements is larger than the number of drugs, becausefor a few drugs-the brand-name firm entered multiple settlements. For some settlements, such as Yasmin, contracting for retained exclusivity (and the accompanying bottleneck) is not necessary because delay can be secured by other means. In the case of Yasmin, the brand-name firm sued the later filer on different patents, thus depriving the later filer of the benefit of the earlier litigation. For an argument that this tactic is anticompetitive, see Answer, Affirmative Defenses, and Counterclaims at 29-30, Bayer Schering Pharma AG v. Sandoz, Inc., No. 08-3710 (S.D.N.Y. July 11, 2008).

102. This is the case when the generic firm is not a first filer, or when the brand-name drug does not give rise to exclusivity eligibility.

103. Interim settlements were reached for Cardizem CD and Hytrin (tablets and capsules), which led to the FTC consent decrees cited in note 25 supra.

104. The drugs are Procardia XL and Wellbutrin SR. In the case of Procardia XL, the generic firm received an immediate license not only on the 30-milligram strength for which it was the first filer, but on two other strengths as well. Defendant Pfizer, Inc.'s Motion to Dismiss the Complaint at 4-6, Great Lakes Health Plan, Inc. v. Pfizer, Inc., No. 01-106 (N.D. W. Va. July 30, 2001). In the case of Wellbutrin SR, the generic firm relinquished any eligibility for the 180 days, and received a license to sell not only the 100milligram strength for which it was first filer, but also a second strength. See

Three factors have shaped a continuing evolution in the structure and content of brand-generic settlements: (1) the waxing and waning of antitrust enforcement, (2) a change in judicial interpretation of the Hatch-Waxman Act, and (3) major statutory amendments to the Act in 2003. This evolution poses challenges when choosing an optimal substantive antitrust rule and antitrust decisionmaker, topics taken up in Parts III and IV, respectively.

1. Antitrust Challenges. -The form of settlement varies significantly with the level of perceived antitrust risk, particularly as to monetary settlements. Table 2 depicts this pattern. Monetary settlements occurred at a rate of about one per year from 1993 through 1999. In 2000, the FTC initiated antitrust actions against several settlements, 1 12 and monetary settlements subsided. In 2005, the government and private purchaser plaintiffs lost antitrust suits in the Eleventh and Second Circuits, respectively. 113 That year saw monetary settlements as to three drugs, and in 2006, six more. Moreover, some settlements may be timed to correspond to a depletion in FFC enforcement capacity. In 2008, shortly after the FTC challenged one monetary settlement, there was a renewed flurry of monetary settlements, including Lipitor and Nexium.

The intensity of antitrust enforcement affects not only the fact, but also the form, of monetary settlements. The first monetary settlementsincluding the first five listed in Table 2-blocked entry until patent expiration, and the brand-name firm paid cash.

increasing frequency after 2000, settling firms changed the standard form of settlements in two ways, both likely responses to increased pressure from antitrust enforcers. 1 15 First, settlements began to include some preexpiration entry. That shift provides drug makers with the rhetorical opportunity to argue that the settlement guarantees some competition. Some entry looks better than no entry. From this perspective, the law has shifted in the drug makers' favor even further than they may have anticipated, given the prevailing view of appellate courts that it is fine to pay for settlements with no pre-expiration entry.' 16 Second, starting in 1997, settlements frequently included not only payment and delay, but also additional contractual terms that tend to obscure whether payment has occurred. The forms of these disguises, and their importance for case-by-case litigation, are discussed in Part III.

-The shift toward settlements with pre-expiration entry has a second cause. Prior to 1998, the FDA had insisted that, in order to enjoy the 180-day exclusivity period, a generic firm must successfully defend its pre-expiration challenge. In 1998, that view was defeated in the courts, on the ground that it was contrary to the text of the Hatch-Waxman Act. 117 After that, a first-filing generic firm could expect to enjoy exclusivity provided it did not lose the patent suit, even if it settled. That made it possible to compensate using retained exclusivity, provided that entry occurred before patent expiration.

The end of the successful defense requirement also created a new form of delay with respect to nonsettling firms. This is due to a statutory quirk in the 180-day exclusivity provision: A later-filed ANDA may not be approved until 180 days after either the first filer's initiation of commercial marketing or a court determination of invalidity or noninfringement. A settlement with the first filer eliminates the possibility of commercial marketing or a court ruling. The 180 days is never triggered, and the FTC, Generic Drug Entry, supra note 13, at 32 tbl.3-3); id. at 1569 & n.63 (inferring same for Zantac); Faulding Inc., Annual Report (Form 10-K), at 9 (Sept. 27, 1996) (Sinemet CR). In addition, "interim" agreements involving two drugs, Cardizem CD and Hytrin, included naked cash payments. [Vol. 109:629 later ANDA filer is stuck, for the FDA lacks authority to approve the application, blocking subsequent entry. 1 18 This resulting "bottleneck," however, is defeasible. If a second generic firm files an ANDA, is sued by the brand-name firm, and wins the patent suit, that decision triggers the first filer's exclusivity period. The second ANDA filer can enter 180 days later. 1 19 To avoid that outcome, the brand-name firm may decline to sue the second generic firm, in which case the generic firm must bring a declaratory judgment suit challenging the patents, 120 win that suit, and then wait 180 days.

3. Statutory Change. -Statutory change represents a third possible source of evolution, but here, the actual change has been unexpectedly small. In 2003, as noted above, Congress amended the Hatch-Waxman regime as part of the MMA. 12 1 These provisions were designed, in part, to curb anticompetitive settlements. The most important change was a new forfeiture procedure, which causes a generic firm to lose its entitlement to the exclusivity period under certain circumstances described below. 1 2 2 The MMA's passage led some to conclude that the settlement problem had been resolved. 123 118. Of the 21 monetary settlements described in Table 2, at least 11 appear to create a bottleneck. As for the others, the first 5 settlements predated the demise of the successful defense requirement, and so their effect, at least as of the date of settlement, is debatable. Four recent settlements-Wellbutrin XL, Nexium, Caduet, and Aggrenox-are governed by the new rules, considered below. In the remaining settlement, AndroGel, the first filer abandoned any claim to the bottleneck. Of the 25 drugs described in Table 3, 9 appear to create a bottleneck under the old rules. The remaining 16 are subject to the new rules discussed infra.

119. In several early settlements, the generic firm disavowed exclusivity eligibility by changing its certification from Paragraph IV to Paragraph III. See Ciprofloxacin, 544 F.3d at 1328-29 (Cipro); Tamoxifen, 466 F.3d at 193-94 (Nolvadex); Bristol-Myers, 135 F.T.C. at 453-54 (FTC Analysis to Aid Public Comment) (BuSpar). In the case of Nolvadex, however, the generic firm reasserted its continued entitlement to exclusivity, after other potential generic entrants emerged and the successful defense requirement was held invalid. Tamoxifen, 466 F.3d at 195-96; see also Cipro, 544 F.3d at 1340 n.14 (considering and dismissing plaintiffs' contention that later filers were discouraged by belief that first filer retained exclusivity). 120. For some settlements, this route was blocked by the Federal Circuit's view that the generic firm lacked standing to bring suit, a roadblock that was later cleared by judicial interpretation.

Five years after the MMA's passage, however, there is little evidence that settlements featuring both payment and delayed entry have become less popular. As noted above in Figure 2, monetary settlements have been a common occurrence after 2003; if anything, they appear to have increased in frequency. And the incidence of monetary settlements for blockbuster drugs has increased. The most important settlements, preserving brand-name profits on blockbusters such as Lipitor, Nexium, and Plavix, occurred after the statutory change. The only blockbuster settlement that predates the MMA is Zantac. That 1995 settlement also preceded significant antitrust enforcement efforts and avoided antitrust scrutiny.

One reason for the limited effect is that the new forfeiture regime only applies prospectively. It is limited to drugs for which the first ANDA was filed after December 2003.124 Most drugs, therefore, are governed by the old regime. Patent litigation frequently takes four or five years to reach settlement. In the Lipitor litigation, for example, a generic firm first filed an ANDA in 2003, but the firms did not settle until 2008. All but 4 of the 21 monetary settlements depicted in Table 2, and 9 of the 25 retained exclusivity settlements in Table 3, were reached under the pre-MMA rules. In short, even if the pre-MMA regime is only transitional, it remains important.

Moreover, even when fully applicable, the new forfeiture rules do little to curb pay-for-delay settlement. Like the old rules, they permit a brand-name firm to neutralize the first filer's challenge through settlement. That first filer still has the largest incentive to challenge the patent because only it is eligible to receive the 180-day reward. And the new rules still contain a bottleneck. 12 5 Forfeiture applies only upon the satisothers, so that a single settlement could block all generic competition on a compound. The law has since changed on this point, and the bottleneck is no longer an issue."); see also Brief for the United States as Amicus Curiae at 18, Andrx Pharms., Inc. v. Kroger Co., 543 U.S. 939 (2004) (No. 03-779), 2004 WL 1562075 [hereinafter Brief for the United States, Andrx] (concluding that MMA's passage lessened need for Supreme Court review). 124. To be more precise, December 8, 2003. MMA § 1102(b) (1), 117 Stat. at 2460. An exception is that one basis for forfeiture, an unappealed or unappealable determination that the agreement violates antitrust law, 21 U.S.C. § 355(j) (5) (D) (i) (V), applies also to "old" ANDAs, MMA § 1102(b) (2), 117 Stat. at 2460.

125. The FDA recently reached the same conclusion: Inherent in the structure of the "failure to market" forfeiture provisions is the possibility that a first applicant would be able to enter into a settlement agreement... in which a court does not enter a final judgment of invalidity or non-infringement (i.e., without a forfeiture event under subpart (bb) occurring), and that subsequent applicants would be unable to initiate a forfeiture with a declaratory judgment action. This inability... could result in [approval delays of other ANDAs]. This potential scenario is not one for which the statute currently provides a remedy. Letter from Gary J. Buehler, Dir., Office of Generic Drugs, FDA, to Marc A. Goshko, Executive Dir., Teva N. Am. 5 n.6 (Jan. 17, 2008), available at http://www.fda.gov/ohrms/ DOCKETS/dockets/07n0389/07n-0389-etOO03.pdf (on file with the Columbia Law Review). This is not the only possible interpretation, since a court might conclude instead

[Vol. 109:629 faction of two statutory conditions. 126 The first condition is relatively easy to satisfy. 12 7 The second is triggered only if an appeals court rules that the relevant patents are invalid or not infringed, or if a settlement reaches a similar result. 1 28 The new bottleneck, like the old one, is defeasible; a later-filing generic firm can break the logjam by winning its challenge and waiting 180 days. The post-MMA rules make the relevant condition for defeasement an appeals court win, rather than a district court win-a condition now applicable to both post-MMA and pre-MMA drugs.' 2 9 This change delays further the moment of generic entry.

The foregoing survey has several implications for antitrust enforcement. First, it demonstrates that the settlement issue is a first-order enforcement question. The size of the buyer overcharge from pay-for-delay settlements likely exceeds $16 billion. 13 0 The large implications for consumer welfare justify vigorous FTC and private enforcement efforts, continued scholarly investigation of the evolution and effect of settlements, and a concerted effort by the FTC and Antitrust Division to reach a full convergence of their historically divergent views of settlements. 1 3 ' that the certification asserting patent invalidity or noninfringement was not "lawfully maintained." § 355(j) (5) (D) (i) (I) (bb).

126. See § 355(j) (5) (D) (i) (I) (triggeringing forfeiture when "later of' (aa) and (bb) occurs). Aside from forfeiture for failure to market, there is also a provision for forfeiture in the case of certain illegal agreements, but that condition requires a successful government antitrust suit against the settling parties. § 355(j) (5) (D) (i) (V).

127. See § 355(j) (5) (D) (i) (I) (aa) (requiring satisfaction by "the earlier of' 75 days after the first filer's effective date, and 30 months after application filing).

128. See § 355(j) (5) (D) (i) (bb). There is also a third possibility, that the brandname firm withdraws the relevant patent information from the Orange Book. § 355(0) (5) (D) (i) (I) (bb) (CC).

129. Prior to the MMA, a generic firm's district court win triggered the running of the exclusivity period. Mylan Pharms., Inc. v. Shalala, 81 F. Supp. For new ANDAS, the rule is analogous. Forfeiture (rather than triggering) of exclusivity occurs 75 days after a generic firm's appeals court win, § 355(j) (5) (D) (i) (I) (bb) (AA) (setting failure to market trigger), and provided that the "easy-to-satisfy" condition discussed supra note 127 is also satisfied.

130. The one-year benchmark measures discussed in Part II, $12 billion for monetary settlements and $4 billion for retained exclusivity settlements, imply a total $16 billion transfer from buyers to sellers. Again, that figure leaves out any effect from increased utilization due to competitive prices.

The survey also underscores the importance of prompt Supreme Court review. 132 In terms of their practical importance, the impact of drug patent settlements is at least comparable to other antitrust issues on which the Supreme Court has granted certiorari. By way of comparison, resale price maintenance, the subject of a recent major Supreme Court case, has long been avoidable for most well-counseled firms.' 3 3

Moreover, settlement has become a patent issue, not only an antitrust issue. Although framed as an antitrust case by plaintiffs, the Federal Circuit has embraced the view that settlement is essentially a patent issue, governed by patent law-indeed, arguably governed by Federal Circuit law1 34 -and that patent law trumps antitrust doctrine within the nominal scope of the patent. The settlement issue fits well with other patent cases on which the Court has taken certiorari in recent years, and is of a piece with the Court's effort to combat perceived hypertrophy in the claimed extent of patent protection. The MMA provisions targeting anticompetitive settlements provide no basis for postponing review. The "transitional" pre-MMA rules continue to have a significant impact. One of the first pay-for-delay settlements concerned an ANDA filed in 1985; the certiorari petition in the resulting antitrust suit was filed 21 years later. For evidence of convergence, see Meyer, supra note 41, at 18 (expressing agreement of DOJ Antitrust Division official with FTC position that courts are too lenient toward settlements).

132. The courts of appeals have varied in their treatment of settlements, see supra note 34 (collecting and comparing cases). The Solicitor General, assessing the cases prior to the most recent Cipro decision of the Federal Circuit, took the view that these cases do not create a true circuit split. Brief for the United States, Joblove, supra note 42, at 15-16.

133. See Leegin Creative Leather Prods., Inc. v. PSKS, Inc., 127 S. Ct. 2705, 2722 (2007) (describing practice of avoiding discussions of pricing policy on advice of "counsel knowledgeable of the intricacies of the law").

134. In the Federal Circuit's recent opinion rejecting antitrust liability, In re Ciprofloxacin Hydrochloride Antitrust Litig., 544 F.3d 1323 (Fed. Cir. 2008), the court does not consistently rely upon the law of the relevant regional court of appeals-the Second Circuit, as the appeal was from a district court in that circuit. After an initial, general statement that the overall rule of reason method should be employed as a matter of Second Circuit law, id. at 1332, the court's detailed analysis gave no privileged place to Second Circuit analysis, see id. at 1332-36, and cited only its own cases at several points, see id. at 1333-34. The court's conclusion was presented as its own independent judgment, rather than as a view about what the Second Circuit would have concluded. Id This aggregate survey reveals a final advantage of prompt review. Antitrust challenges to early settlements are still making their way to the Court. 1 3 7 These contain payment and delay, but not much else. Later settlements, however, add contractual complexity. They add difficult factual layers-Was there payment? Was there delay?-atop the legal question of whether payment in exchange for delay violates antitrust law. For a Court that dislikes wading into factual complexity, the early cases provide a more attractive vehicle for setting a clear rule.

This Part examines how an aggregate approach affects the choice of a substantive antitrust rule. Part III.A highlights one particularly troubling element of the evolution in settlements: the rise of side deals that disguise the fact of payment in a pay-for-delay settlement. Part III.B demonstrates that the exchanges seen in these side deals, though common in settlements, are uncommon otherwise. Part III.C argues that the absence of similar deals outside the settlement context provides a basis for presuming that side deals are disguised payments for delay, not for value.

As explained in Part II.B, the earliest settlements were straightforward affairs. The brand-name firm paid cash in exchange for the generic firm's delayed entry. The largest naked cash payment was nearly $400 million, which Bayer agreed to pay Barr in settling litigation over Cipro, a major antibiotic. 1 38 In the wake of increased antitrust scrutiny, naked payments have given way to more complex arrangements. Today, side deals take two complementary forms: overpayment by the brand-name firm for value contributed by the generic firm, and underpayment by the generic firm for value provided by the brand-name firm.

1. Overpayment by the Brand-Name Firm. -In the most common type of side deal, the generic firm contributes-in addition to delayed entrysome further value, such as an unrelated product license. The additional term provides an opportunity to overstate the value contributed by the generic firm and claim that the cash is consideration for the contributed value, rather than for delayed entry. In reviewing K-Dur, the earliest settlement with this type of side deal, the Eleventh Circuit accepted such a factual assertion, which provided a basis for rejecting antitrust liability.1 3 9

137. An example is Cipro, which could yield petitions from both the Federal Circuit and the Second Circuit.

COLUMBIA LAW REVIEW Side deals are now a regular feature of entry-delaying settlements. The contributed value can include a wide range of product development, manufacturing, and promotional services. In some deals, the generic firm offers a product or patent license, or agrees to develop a new prod-uCt.

1 4 0 In one variant, the generic firm develops a new formulation of the brand-name drug. 14 1 In other deals, it agrees to furnish manufacturing services to the brand-name producer, 1 42 or to provide inventory, 143 or even to provide "backup" manufacturing services. 144 In some cases, the generic firm provides promotional services as to the product at issue, related drugs, or unrelated products. 145 For some drugs, the brand-name firm reaches entry-delaying settlements with multiple generic firms, each with side deals. 146 Some of these arrangements are suspect on their face. It may seem clear that the brand-name firm does not need a patent license that does not clearly cover its product, new drug development that is unrelated to its current core business, a new source of raw material supply, backup 140. For example, K-Dur (two settlements), Naprelan, Provigil (four settlements), and Adderall XR (two settlements) all involved a license or product development agreement. 146. This is the case for four of the drugs discussed supra note 105: Provigil (as to multiple first filers), Adderall XR (as to both a first filer and a later filer), AndroGel (same), and K-Dur (same). See supra notes 140-145.

[Vol. 109:629 manufacturing, or additional promotion. 14 7 However, not all such settlements are facially absurd. In some cases, the generic firm has plausible expertise in the subject of the side deal.' 4 It is very difficult to be certain that a deal is collusive without a deep and complex inquiry into the business judgment of the two drug makers.

2. Underpayment by the Generic Firm. -The brand-name firm, rather than paying too much, can charge too little. One mechanism involves "authorized generic" sales. These are sales made by a generic firm under the brand-name firm's FDA approval. The brand-name firm supplies the product to the generic firm at a discount, which the generic firm then resells under its own label at a profitable price. The compensation is buried in the discounted price offered by the brand-name firm.

In several early settlements, the authorized generic product was launched at the time of settlement. 1 49 This practice fell out of favor after a court concluded that the authorized generic sales triggered the 180-day period. 15 0 Some modern settlements avoid the trigger problem by providing for authorized generic sales only after another generic firm enters, or of a drug other than the subject of the generic firm's ANDA filing, 15 1 147. For example, in the case of a settlement involving the wakefulness drug Provigil, the brand-name firm, Cephalon, apparently was aware of one generic firm's intellectual property for three years before showing any interest in seeking a license. Provigil Complaint, supra note 7, at 16. 148. See, e.g., Adderall XR Shire-Barr Agreement, supra note 71 (describing Barr investments in drug delivery technology, to be exploited in new product development by brand-name firm as part of settlement). The agreement was later terminated, with substantial payments to Barr. Shire PLC, Current Report (Form 8-K), at 1.01 (Mar. 2, 2009) (reporting reimbursement of up to $30 million in expenses, one-time payment of $10 million, and $25 million in foregone revenue from license for authorized generic supply).

149. For example, Nolvadex and Procardia XL involved authorized generic sales.

COLUMBIA LAW REVIEW or in another country. 1 52 In a related form of discounted sale, which avoids the trigger issue, the brand-name firm sells an entire product line to the generic firm. One settlement involving an extended-release version of a drug, for example, transferred (for a possibly discounted price) the immediate-release version to the generic firm. 1 53 In a more complicated set of deals, a brand-name firm may have sold a generic firm rights to one product, and the generic firm delayed entry in two other products. 5 4 (A further variant of this strategy, simultaneous settlement of multiple drugs with uneven entry terms, is considered in Part IV.C.) Once again, it is very difficult as a practical matter for a decisionmaker to know whether the transfer price provides compensation from the brand-name firm to the generic firm, and if so, how much.

Outside of settlement, brand-name firms seldom contract with generic firms for help with the activities that form the basis of side deals. Indeed, as a general matter, brand-name and generic firms seldom execute major deals outside the settlement context, with the exception of authorized generic arrangements, which necessarily are reached between a brand-name firm and a generic firm.

A review of the annual securities filings of settling drug makers supports this proposition. To examine the extent of business dealings outside of settlement, five major brand-name firms 55 and five major generic firms 156 were chosen based upon their frequency of settlement activity and economic importance. For each brand-name firm, annual fll- ings between 2000 and 2007 were searched for the names of the five generic firms. 1 5 7 Each resulting "hit" led to further examination, to see whether the discussion indicated a business relationship between the two firms, as opposed to, say, a description of litigation or competition. The business transactions were examined further using articles in the trade press and other materials. The same exercise was performed for each of the generic firms, as to each of the five brand-name firms. 158 The resulting inquiry into twenty-five total brand-generic business dealings-each of five brand-name firms, with each of five generic firms-produced just two responsive business arrangements, both of them involving Ranbaxy: an unusual drug development deal with one brand-name firm, 159 and a purchase of rights to a set of minor dermatology drugs from another brand-name firm. 1 60 Several other business arrangements do not match the terms of the side deals discussed above.' 6 ' This evidence is not decisive; such non-settlement deals could exist, yet be too insignificant to report in an annual filing. If so, however, they are apparently not of first-rank importance to the operations of the firm.

Further evidence about the firms' limited business dealings, outside of settlement, is revealed by one specific type of side deal known as co-157. Form 10-K, in the case of Abbott, Bristol, and Pfizer; Form 20-F, in the case of AstraZeneca and GlaxoSmithKline.

158. Form 10-K, in the case of Barr, Mylan, and Watson; Form 20-F, in the case of Teva; and detailed annual reports filed under Indian securities law, in the case of Ranbaxy.

159. Glaxo and Ranbaxy have an unusual drug development initiative, in which Ranbaxy takes "hit" molecules from Glaxo that show initial promise, and helps develop and winnow them into "candidates" for further development by Glaxo. Ranbaxy Labs. Ltd 161. For example, Bristol agreed to commercialize EmSam, a patch treatment for depression, after it was already developed by a Mylan-Watson joint venture, and ready for FDA approval. B-MS and Somerset in EmSam Distribution Deal, Pharma Marketletter,Jan. 3, 2005, available at Factiva. Bristol and Barr have had complex marketing arrangements on several products, but this is the accidental result of an antitrust settlement between DuPont and Barr, inherited by Bristol when it bought DuPont's drug business. Rick Mullin, Bristol-Myers Untangles Barr-DuPont Agreements, Chemical Wk., May 8, 2002, at 27, available at Factiva. Ranbaxy bought Glaxo's generic drug operations in Spain and Italy. Ranbaxy, 2006 Annual Report, supra note 159, at 5. In 1999, Watson paid Glaxo to acquire the rights to Androderm, a testosterone patch, but this was a reacquisition of rights to a product developed by a company later acquired by Watson. Taren Grom, Generics: Best Years to Come, Med Ad News, Oct. 1, 1999, available at Factiva (describing Watson's acquisition of TheraTech); Watson Rights, Chain Drug Rev., June 28, 1999, available at Factiva (announcing reacquisition of Androderm rights).

promotion. Brand-name firms frequently enter co-promotion arrangements to augment their promotion efforts-for example, to reach physicians that their own detailing team does not visit. In a second search, the same annual filings were reviewed for mentions of promotion, and those mentions which pertained to product promotion were examined further. That search produced many examples in which a brand-name firm recruited other brand-name firms to help promote a drug, but no significant examples, outside the settlement context, in which the brand-name firm recruited a generic firm to promote a brand-name drug. 162 On the other hand, generic firms do occasionally have significant branded drugs, and the search did reveal instances when they have hired brand-name firms to help market the drug. 163 This result is not surprising, considering the business of generic firms. Generally, they do not have substantial promotion teams, for they seldom have major branded drugs to promote. The absence of generic provision of other services, outside the settlement context, is equally unsurprising. Although some generic firms have made efforts to develop a brand-name drug business, 164 as a general matter, their research and development capacity is limited; this is not their core business. Nor do they have powerful manufacturing capabilities such that they would be the obvious and efficient alternative supplier for a brand-name firm. 165 The contrast is less severe in side deals featuring transferred assets. It is quite common for a brand-name firm to set up an authorized generic arrangement with some generic firm. Transfers of product lines to other drug makers are common as well.

Viewed in isolation, it is difficult to tell whether a side deal represents payment for value or disguised payment for delayed generic entry. A broader comparison of side deals in conjunction with settlements, versus brand-generic deals outside this context, tells a different story. At least with respect to overpayment side deals, the absence of brand-generic 162. A minor exception is promotion efforts outside the United States. In particular, Ranbaxy promotes a Sanofi vaccine in India. See Ranbaxy to Market Aventis Vaccines, Bus. Line, Oct. 6, 2002, available at Factiva (describing agreement to market six vaccines); see also Aventis Arm in Vaccine Tie-Up with India's Ranbaxy, Reuters News, Oct. 4, 2002, available at Factiva (describing marketing agreement).

163. For example, Teva recruited a predecessor of Sanofi-Aventis to help sell its multiple sclerosis drug Copaxone. Sanofi-Aventis, Annual Report (Form 20-F), at 61 (Mar. 7, 2008); Teva Pharmaceutical Industries Ltd., Annual Report (Form 20-F), at 20 (Mar. 31, 2001). Another example is the EmSam deal discussed supra note 161.

164. See, e.g., Teva Pharms. Inc., Innovative Research & Development, at http://www. tevapharm.com/research (on file with the Columbia Law Review) (describing efforts to develop innovative drugs that have yielded two products, Copaxone and Azilect).

165. For example, Cephalon agreed to buy Provigil's active ingredient from a third generic firm, even though the firm had not manufactured the product and Cephalon already had an adequate source of supply. Provigil Complaint, supra note 7.

[Vol. 109:629 deals outside of settlement is a strong reason to suspect that the deals are used to pay for delay.

In such cases, it is appropriate to impose a presumption that the side deal provides disguised payment to the generic firm. Under this pay-fordelay presumption, drug makers would be free to come forward with evidence that their unusual deal was for value and therefore raises no anticompetitive issues. That burden is most appropriately placed upon them, as the least-cost providers of the necessary information. An alternative approach, also supportable by the evidence from aggregation, would make this presumption conclusive.

That conclusion is not, by itself, enough to impose liability. It resolves the "factual" question of whether a settlement containing a side deal constitutes payment for delay, but not the "theoretical" question of whether pay-for-delay settlements violate antitrust law. 1 6 6 This proposal, like any aggressive antitrust rule, is potentially overinclusive. It raises the probability of false condemnation. But here, the rarity of such arrangements outside of settlement lowers the likelihood of false positives. The error cost analysis has a further component: How costly are false positives when they occur? Not very costly, as it turns out, because the generic firm is seldom a distinctive source of the particular value in question.

The rule comports with the comparative rigor with which we treat collusive activity generally. Antitrust's lenient approach to exclusionary conduct reflects an error cost calculation focused upon false positives.1 67 As noted in the introduction, decisionmakers think that true positives are rare and difficult to distinguish, and also that false positives are particularly costly, because they amount to condemnation of the "very conduct" (competitive price cuts) that antitrust is supposed to protect. 68 As other commentators have noted, false negatives are an important countervailing problem. 169 For collusion, by contrast, avoiding false negatives is the important goal, particularly where false positives are rare and low- cost, and where no significant equilibrating factors tend to restore competition. That relatively aggressive approach is shared even by "Chicago School" analysts, who support an enforcement emphasis upon collusion, 1 70 What about underpayment side deals? The likelihood of false positives is higher, compared to overpayment deals, because authorized generic arrangements and product transfers frequently occur outside the context of settlement. The cost of false positives remains low, however, due to the absence of distinctive value arising from dealing with this particular generic firm, which happens to be locked in a patent suit with the brand-name firm, as the counterparty in a transaction with this particular brand-name firm.

The high cost of false negatives and low cost of false positives support a presumption in the underpayment context, just as in the overpayment context. A more conservative alternative would be to make the presumption applicable only to future settlements. That way, parties have ample notice that they must not reach underpayment deals with parties with which they are settling. Given the absence of distinctive value offered by the settling firm, that route places at most a minimal burden upon parties that wish to reach authorized generic or asset transfer arrangements.

This policy suggestion could be implemented by several routes. For example, it could be adopted by a court considering a particular case, using the federal courts' common lawmaking authority under the Sherman Act. Alternatively, it could be instituted through new congressional legislation, or promulgated as an agency rule by the FTC. The next Part considers the strengths and weaknesses of these alternative routes.

This Part turns to the institutional question of who should employ this aggregate approach to antitrust questions. Part IV.A explains why an agency-here, the FTC-is better positioned to collect and synthesize aggregate information, relative to courts. Part IV.B argues that this advantage in wielding aggregate information favors a shift in substantive policymaking authority from courts to agencies. [Vol. 109:629

A court establishing antitrust policy faces the fundamental problem that it has little capacity to collect aggregate data. The disadvantage of a court as a fact-finder is a familiar idea from the literature on institutional choice. 1 7 1 The problem is particularly acute here. At best, a single court needs many years to develop a sense of the overall distribution of cases, as antitrust cases appear only rarely on its generalist docket. The Supreme Court is in a slightly better position, since it is exposed to appeals from all over the country. But many instances of anticompetitive behavior are never litigated, and courts have particularly limited ability to observe nonpublic data about settlements outside the case at bar.

Private parties cannot entirely fill the gap. These plaintiffs struggle to learn the content of settlements, with some early agreements escaping notice entirely.' 72 Later settlements have been shielded from scrutiny due to the difficulty of discerning, from public information, the extent of pay-for-delay deals. This information gap partially explains why so few of the most recent settlements have been challenged. This Article helps fill the gap, but it is not a complete solution. My data does not include nonpublic details that would help build confidence about whether a side deal conveys payment. For example, how much did the brand-name firm agree to pay for a co-promotion agreement? How much did a generic firm pay for a product transfer? Is payment conditioned on successful performance by the other party? Was a particular product development deal a long-felt need of the firm, which shopped for alternative sources? How was the service provided valued internally by the payor? Public data for most settlements lack these details.

Outside the context of side deals, two other issues are important. First, do the parties expect the generic firm to retain exclusivity when it enters the market? In some cases, one or both parties divulge their view publicly, but in other cases they do not. Second, how often does the brand-name firm contract with this counterparty and other generic firms 171. For an excellent discussion of this literature, see Margaret H. Lemos, The Commerce Power and Criminal Punishment: Presumption of Constitutionality or Presumption of Innocence?, 84 Tex. L. Rev. 1203, 1251-57 (2006) (reviewing argument that courts are weak fact-finders, limited by lack of expertise, investigative capacity, and access to facts beyond those of single case before them). Among the many sources cited there, see, e.g., Benjamin Cardozo, The Growth of the Law 116-17 (1924) ("Some of the errors of courts have their origin in imperfect knowledge of the economic and social consequences of a decision, or of the economic and social needs to which a decision will respond."); Cass Sunstein, The Partial Constitution 147 (1993) ("Courts are rarely experts in the area at hand. Moreover, the focus on the litigated case makes it hard for judges to understand the complex, often unpredictable effects of legal intervention. Knowledge of these effects is crucial but sometimes inaccessible."); William W. Buzbee & Robert A. Schapiro, Legislative Record Review, 54 Stan. L. Rev. 87, 143 (2001) (observing that courts "are not well-suited to gather the evidence necessary to assess the magnitude of complex social practices .... ).

172. See, e.g., Reguly, supra note 111, at 25 (reporting Zantac settlement).

outside the context of settlement? Reciprocally, what is each generic firm's experience with brand-name firms outside the context of settlement? Public information of the type collected in Part II paints only an incomplete picture of the frequency of particular arrangements outside the settlement context. With details such as these, an inference of payment for each case could be strengthened, and-more importantly-the inference of payment across cases could be strengthened as well.

The FTC already has in place all the tools it needs to perform this task. As noted in Part I.B, it receives information about each settlement and has statutory authority to require firms to produce additional information of the types discussed above.' 73 That authority ought to be used to collect two types of information. First, the Agency should seek full details about each settlement-at least enough information to answer the questions listed above. Some of these questions may be answerable by examination of the agreement itself. To the extent they are not, the gaps could be filled using voluntary questionnaires or, if necessary, compulsory process. Second, the Agency should collect from each brand-name firm a detailed catalogue of its dealings with generic firms, and vice versa for generic firms.

This information would be the key input in a comprehensive study of side deals. It would provide a firm basis for the Agency to endorse or reject the conclusion offered in Part II, based upon public information, that contemporaneous side deals should possess a presumption of payment. If the information is sufficiently lopsided, error cost minimization might suggest the more aggressive rule should be instituted, making the presumption of illegality conclusive and effectively banning contemporaneous side deals.

In this respect, the analysis in Part II provides a rough draft for a more comprehensive, future agency report. The public data presents a prima facie case that something is amiss regarding the increasing utilization of side deals. For skeptical readers of this Article, who may think that the survey results reported in Part II are too weak to justify a presumption of payment through side deals, the case for deploying the agency as an aggregator should be even stronger; agency action is necessary to fill these informational gaps and better explain whether and when compensation is conferred for delay.

The FTC has not fully exploited its information gathering advantage. Of the drugs with monetary settlements in Table 2, two-thirds occurred after the end of the FTC's last major study in 2002. Moreover, all of the retained exclusivity settlements in Table 3 post-date the study. To be sure, the FTC evaluates each individual agreement to determine whether further investigation is appropriate, and no doubt it asks some of the questions detailed above in considering its response. But it does not synthesize the resulting information, aside from very general annual summa-173. See supra notes 59-60 and accompanying text.

[Vol. 109:629 ries of settlement activity. As this Article reveals, only through such an aggregate approach can we expect to generate a useful picture of-and rule for-brand-generic settlement.

The previous section advocates a focused increase in the FTC's "competition policy research and development." 174 If the F-C accepted the suggestion, it would eventually reach a firm, empirically grounded conclusion about the optimal policy for side deals, and thus either confirm or reject the conclusion reached in Part II. That conclusion could be deployed in a variety of policymaking settings, including litigation brought by the Agency, amicus practice, and advocacy for congressional legislation. This section considers a further possibility, that a comprehensive aggregate study of settlement practice could form the basis for substantive policymaking by the Agency in the form of rulemaking.

There is of course an enormous literature on the choice of courts versus agencies, adjudication versus rulemaking, and rules versus standards, and this Article does not engage the full complexity of those debates. My goal here is simply to suggest how the virtues of an aggregate perspective on settlement practice shift the balance in a way that favors agency rulemaking. In other words, the settlement issue highlights certain advantages of moving away from a court-centered model of antitrust law.

Why bother with rulemaking? Even if the expert agency is better than a court at arriving at a correct policy conclusion, thanks to its superior capacity for aggregation, it does not necessarily follow that the agency ought to set policy. It could instead simply furnish the information to a court or Congress, which might then implement the same conclusion, with some of the same benefits-for example, efficiency compared to case-by-case adjudication, and certainty for businesses about the range of acceptable practices.

Put another way, why would we care whether the agency itself makes policy in the first instance, rather than acting as an input to a court? The question suggests a bureaucratic version of the Coase Theorem. If there is no friction in communicating an expert policy conclusion from the agency to the court, then it does not matter which of the two has policymaking authority. If, on the other hand, the agency's message arrives garbled or is ignored by the court, that provides reason to prefer that the agency reach a substantive policyjudgment of its own, rather than merely furnishing advice to the court.

One reason to expect the court to do a less effectivejob is that courts have trouble correctly identifying anticompetitive strategic behavior, 175 particularly in a setting as complex as the Hatch-Waxman Act. That view is borne out by a recent appeals court opinion about settlement. The court relied, as a reason to deny antitrust liability, upon the mistaken idea that a settlement with one generic firm would spur other generic firms to action, and that these firms would have the large incentive provided by the exclusivity period. 1 76 In fact, later filers are ineligible for the exclusivity period. This error was unforced; the point does not appear to have been argued below. The same court took comfort in the view that often there is more than one generic challenger, and the court concluded that multiple challengers are difficult to buy off. 1 77 In fact, however, multiple settlements do happen.

Courts have also had trouble evaluating the facts of particular cases. For example, in the case discussed above, the plaintiff had argued that the brand-name firm compensated the generic firm not only with cash, but also through authorized generic sales. 178 The court ignored this idea entirely. 1 79 In a second case focused on side deals, the appeals court essentially ignored the extensive evidence that the payment was for delay, rather than the separate value offered by the generic firm.' 80 This pattern is likely to continue, given the evidence of complexity discussed in Part III.A.

An expert agency, essentially by definition, is less likely to make mistakes identifying the strategic behavior of parties. To be sure, this information could be communicated to a court. But as a practical matter, courts have not welcomed the information about settlements supplied by the FTC. In a key case brought by the FTC, the appeals court largely ignored the analysis employed by the Agency, granted essentially no deference to its findings of fact, and indeed berated the Agency for failing to 175. See Herbert Hovenkamp, The Antitrust Enterprise 47 (2005) ("[T]here is relatively little disagreement about the basic proposition that often our general judicial system is not competent to apply the economic theory necessary for identifying strategic behavior as anticompetitive."); Christopher R. Leslie [Vol. 109:629

follow the appeals court's earlier rule. 18 ' For the most part, courts have also ignored the results of the FTC's extensive 2002 study and its subsequent annual summary updates, as well as its amicus recommendations based on this data. 1 8 2

A second reason to expect courts to be less effective than the FTC is that antitrust courts are obliged to impose treble damages when they condemn behavior as a violation of the Sherman Act. The large measure of damages may strike a court as excessive, particularly where the conduct seems ambiguous or complicated, such that the parties might not be expected to know that their behavior violated antitrust law. 1 83 That impression may be reinforced where the conduct is out in the open, rather than hidden, so that a usual justification for a damages multiple-the difficulty of detection-is missing. The combined effect is to make a court gun shy, and to cause it to select a deliberately underinclusive antitrust rule. 18 4 And indeed, courts rejecting antitrust liability for settlements have repeatedly adverted to treble damages in their analysis. 18 5 The FTC is less constrained. Its substantive conclusions would be made under the Federal Trade Commission Act's prohibition of "unfair 183. In other contexts, courts are thought to narrow substantive rights when the consequence of their violation is believed to be too severe. See, e.g., Akhil Reed Amar, Fourth Amendment First Principles, 107 Harv. L. Rev. 757, 799 (1994) ("The exclusionary rule renders the Fourth Amendment contemptible in the eyes of judges and citizens. Judges do not like excluding bloody knives, so they distort doctrine, claiming the Fourth Amendment was not really violated.").

methods of competition,"' 1 8 6 rather than under the Sherman Act. The Supreme Court has stated repeatedly that the FTC Act's prohibitions are broader than those of the Sherman Act. 18 7 Thus, behavior that constitutes unfair competition does not necessarily also violate the Sherman Act's prohibitions of unreasonable restraints of trade or monopolization.

This conclusion is resisted by some observers, who think it is "no longer tenable" to treat the FTC Act as broader than the Sherman Act. 18 8 However, the Supreme Court's rejection of strict equivalence can bejustified on an eminently pragmatic ground. The argument for equivalence rests upon the proposition that, as Richard Posner puts it, "the Sherman and Clayton Acts have been interpreted so broadly that they no longer contain gaps that a broad interpretation of Section 5 of the FTC Act might be needed to fill.' 89 But to the extent that the Sherman Act as actually interpreted by courts contains important gaps, as exemplified by the lack of liability for pay-for-delay settlements, the quoted statement does not hold. Where, as here, courts are reaching incorrect conclusions about liability, distinguishing the two statutes is useful, because it allows the FFC to enjoin settlements without being automatically reversed by a court equipped with the (erroneous) view that antitrust law does not extend so far.

Moreover, nonequivalence is particularly useful where, as here, treble damages lead courts to constrict the scope of liability. Even if it is appropriate for courts to constrict liability to compensate for the heightened false-positive risk created by treble damages, it does not follow that the FTC must adhere to the same path. The FTC seeks injunctive relief, not treble damages. That difference reduces concerns about false positives and overdeterrence. Put another way, the FTC's optimal scope of liability may well be broader than the courts'. Nonequivalence allows the FTC to take advantage of that difference, compared to the Sherman Act, which applies a harsher penalty to a narrower class of activity. 190 186. FTC Act § 5(a)(1), 15 U.S.C. § 45(a)(1) (2006). 187. See, e.g., FTC v. Ind. Fed'n of Dentists, 476 U.S. 447, 454 (1986) (holding that section 5 covers "not only practices that violate the Sherman Act and other antitrust laws, but also practices that the Commission determines are against public policy for other reasons" (citations omitted)); FTC v. Brown Shoe Co., 384 U.S. 316, 321 (1966) (holding that section 5 reaches "practices which conflict with the basic policies" underlying antitrust law, as well as incipient violations of antitrust law); FTC v. R.F. Keppel

A third advantage of the FTC is that it is less subject to the constraint of stare decisis. Lower courts are bound by their own or Supreme Court precedent. 1 9 1 The Supreme Court, for its part, is not quick to revisit antitrust doctrine,' 19 2 and frequently feels constrained to follow its own previous views. 1 93 The FTC is freer to change course, provided that the new interpretation is a reasonable understanding of the FTC Act. 194 One way for the FTC to exploit these advantages is to promulgate a legislative rule-that is, a rule having the force of law and entitled to Chevron deference by a court. 1 95 FTC rulemaking has been suggested periodically by commentators as a way to shift decisionmaking authority to the FTC and fill gaps in the coverage of other antitrust statutes. 1 9 6 The other statutes, 20 3 and overruling Petroleum Refiners today would jeopardize rulemaking in these other contexts, including the grant analyzed in Chevron itself. 20 4 These prudential considerations, and a later congressional enactment, 20 5 tend to confirm the viability of rulemaking authority. Although the FTC reportedly sought candidates for antitrust rulemaking after Petroleum Refiners, 20 6 it has not yet found any. A rulemaking focused on settlements is an attractive candidate if this procedural route is pursued again.

Rulemaking is not the only way to shift substantive policymaking authority from courts to the FTC. The FTC can bring individual cases through agency adjudication, 20 7 reviewed in a court of appeals of a respondent's choosing, 208 or directly in an action in district court. The FTC has taken both routes in attacking settlements. The agency adjudication route resulted in an appeals court loss; two cases in district court are pending.

Rulemaking has significant, familiar advantages over the adjudicatory route. Rulemaking permits affected parties to test aggregate data in an open way, with ample opportunity for rebuttal. 20 205. After Petroleum Refiners, Congress authorized legislative rulemaking in the consumer protection sphere, while preserving whatever antitrust rulemaking authority already existed. See 15 U.S.C. § 57a (authorizing rulemaking regarding "unfair or deceptive acts or practices in or affecting commerce"). This legislative action took place against the backdrop of both Petroleum Refiners and the previously promulgated antitrust rule. The decision not to disturb these indications of antitrust rulemaking authority, while altering the contours of consumer protection rulemaking authority, is arguably a ratification of the FTC and D.C. Circuit's views. On the other hand, an examination of the legislative history paints a more skeptical view. See Einer Elhauge & Damien Gerardin, Global Antitrust Law and Economics 5 n.l (2007) (concluding, based upon legislative history, that Congress came to no considered view about existence or absence of antitrust rulemaking authority when it passed Magnusson-Moss Act).

206. FTC Staff Narrows Rulemaking Possibilities to Three Areas, supra note 69, at A-13 (noting FTC staff's interest in rulemaking about "delivered pricing in the cement industry, physician influence over health insurance payments, and mergers").

207. FTC Act § 5(b), 15 U.S.C. §45(b). 208. Id. § 5(c). The chosen court of appeals must be one in which the condemned practice was used, or in which the respondent does business. In the settlement context, that means as a practical matter that the administrative ruling will be reviewed in a court of appeals already known to be hostile to liability. 209. See, e.g., Arthur Selwyn Miller & Jerome A. Barron, The Supreme Court, the Adversary System, and the Flow of Information to the Justices: A Preliminary Inquiry, 61 Va. L. Rev. 1187, 1211-18 (1975) (assessing problems that result when judges use data "not subject to test or challenge by the losing party").

for input and testing tends to produce superior policy. 2 10 The resulting rule thus has a superior claim to judicial deference, compared to judicial review of a single case: The rule has been thoroughly vetted under notice and comment, after a broad, deep review of the full terrain of behavior by regulated parties. It is this superior breadth and greater vetting, rather than the doctrinal force of Chevron itself, 211 that presents the strongest reason to think that a rule might succeed where adjudication has failed.

Rulemaking helps in another way. The FTC Act is broader than the Sherman Act, as noted above, but the degree of its additional breadth has been a subject of controversy. Some lower courts have regarded with skepticism the FTC's efforts to regulate behavior not already governed by the Sherman Act. 212 A powerful way for the FTC to overcome this skepticism would be to support its claim to authority with aggregation, buttressed by notice-and-comment rulemaking. In this manner, the FTC could combine, in a mutually reinforcing manner, the two ways in which its authority stands out relative to ordinary, judicial antitrust policymaking: in having a statute with broader reach than the Sherman Act, and in possessing the power to collect information beyond the reach of the judiciary.

Rulemaking has a further effect: It attracts congressional attention to an important policy issue where adjudication may not. The FTC's first controversial foray into rulemaking was the Cigarette Rule, 2 1 3 a consumer protection rule promulgated in 1964 that governed the advertising and labeling of cigarettes. One powerful effect of the rule was to focus Congress on the question, in part because the industry argued that the FTC had usurped congressional prerogatives. The rule was withdrawn the following year, replaced by a watered-down statute. 21 4 A modern antitrust rule might be expected to create a similar provocation. Whether that is an argument in favor of rulemaking is less cer-tain. In the case of cigarette regulation, congressional action preempted the FTC's rule in key respects. 21 5 However, the FTC stayed deeply engaged in congressional debates on the issue, and played an important role in promoting further statutory change. 2 1 6 Increased congressional attention might therefore be regarded as a modest positive overall, or at least not a negative.

At the same time, some of agencies' distinctive disadvantages seem less pronounced here. A shift from courts to agencies raises concerns about an agency's comparatively greater vulnerability to capture by regulated parties. 2 17 As applied to the FTC, this concern finds some support in the early history of the Agency, where a protectionist attitude toward small businesses in certain industries can be plausibly attributed to capture. 2 18 Moreover, the settlement issue is currently of concentrated interest only to the pharmaceutical industry, making the capture concern particularly salient, although one could imagine insurers and other drug purchasers providing a counterweight.

On the other hand, the modern FTC is a much more effective organization today than the agency that received so much criticism several decades ago, and has erased the taint of the earlier capture critique. 2 19 Its newfound success can be attributed in part to a bipartisan consensus about the role of economic analysis in modern antitrust law. That consensus has had a further effect, which is to help neutralize a second attribute of agencies, namely their sensitivity to political changes over time. 2 20 In any event, whatever the general merits of this characterization, it seems inapplicable to the settlement issue, where FTC commissioners across the political spectrum have been unanimous in their view that settlements raise serious competitive concerns.

Settlement practice continues to evolve to exploit regulatory complexity. The usual assumptions about settlement are that it entails an agreement, by which cash or its equivalent is exchanged for entry, for an entry date that is constrained to be no later than patent expiration. In fact, the forms of payment and even the fact of agreement are manipulable . The following examples from recent settlement practice bear this out.

1. Multiple Settlement with Uneven Entry. -In some instances, the brand-name and generic drug makers settle several disputes at the same time, affording the brand-name firm an opportunity to pay the generic firm for delayed entry on one drug by granting early generic entry on a second drug. Consider, for example, Lamictal, a blockbuster epilepsy treatment that is offered in both chewable and nonchewable forms. A generic firm launched a pre-expiration challenge to each form; both centered upon the same patent. 2 21 In the joint settlement of both disputes, the generic firm received a license to the chewable version that permitted entry three years before entry on the nonchewable version. 222 Uneven entry does not automatically raise pay-for-delay concerns. For example, a one-year delay as to one drug might exactly offset a oneyear acceleration of entry on a second drug of equal importance. More generally, if a generic firm's interests are aligned with consumer interests, there is little to worry about, because a generic firm will insist upon early enough entry (and increased consumer welfare) on one drug to compensate for the reduced generic entry (and consumer welfare) on the other drug. Of course, in such a situation, it is difficult to see why the parties would bother with uneven entry. The explanation is that the drug with early entry is one on which the parties expect comparatively little incremental entry from other generic firms. In the case of Lamictal, the nonchewable version is far more important than the chewable version; in fact, the chewable version had low enough sales as to be unlikely to attract additional generic challengers. -A special case of the strategy arises when entry as to one of the drugs has already occurred, and there are accrued damages-probabilistic, as the patent suit has not yet been resolvedthat the brand-name firm can forgive as part of the settlement. Lipitor is again exemplary. Pfizer and Ranbaxy had done battle on a second significant drug, Accupril. Ranbaxy had launched a generic version of Accupril without waiting for a district court to rule whether Pfizer's patent was valid and infringed. 224 Pfizer secured a preliminary injunction, which was affirmed by the Federal Circuit. 225 At this point, Pfizer's damages claim against Ranbaxy, although probabilistic, was large in expected value. 226 The Lipitor settlement also "resolved" the Accupril dispute, likely by forgiving the accumulated expected damages. The residual uncertainty about the terms of this settlement helps illustrate why the FTC's role is so important.

The forgiveness strategy can be applied not only across several drugs, but also across several strengths of a single drug. For example, in Wellbutrin XL, the generic firm had challenged the patent applicable to two different strengths of the drug. It launched at risk as to only one strength. The subsequent settlement forgave accumulated damages on the first strength, and delayed entry on the second. 2 27 Release, Barr Pharms., Inc., Barr Granted Rights to Generic of Cephalon's ACTIQ Cancer Pain Treatments (Aug. 10, 2004), available at http://www.medicalnewstoday.com/articles/ 12020.php (on file with the Columbia Law Review) (describing initial Barr license); Provigil Barr Press Release, supra note 89 (describing earlier Actiq license). A third possible example is Optivar and Astelin, which had first filer challenges that pertain to the same patent. Meda AB, Interim Report, at 6 (May 6, 2008), available at http://www.meda.fi/ english/news/year_2008/interim_report-january-march_2008.pdf (on file with the Columbia Law Review). The settlement as to both drugs permits entry as to Optivar, the less important drug, three months earlier than Astelin. Id. at 6-7 (noting settlement terms); see also Meda

The logical next step is a launch followed by waiver for a single drug. 228 3. "No Authorized Generic" Provisions, -As previously explained, retained exclusivity is a source of compensation to a generic drug maker. 229 That compensation is reduced, however, if a brand-name drug maker launches an authorized generic product to compete with the generic entrant, in addition to the brand-name firm's existing branded product. The brand-name firm can increase the generic entrant's profits from exclusivity by agreeing not to launch an authorized generic product. This decision is costly to the brand-name firm, which foregoes extra profits from its own authorized generic competitor. Numerous recent agreements include a "no authorized generic" term. 23 0 4. Avoiding Agreement. -Through careful design, settling parties can arrange for delayed entry without any formal agreement as to timing. The parties can condition periodic payment upon nonentry, and make payment a function of brand-name profits that depend upon nonentryfor example, a royalty paid on brand-name sales. 23 15, 2006), available at Factiva (noting "no authorized generic" provision); Plavix Agreement, supra note 71, exh. 99.1 (permitting generic manufacturer "to sell its Plavix brand product, but not to launch an authorized generic"); Wyeth, Current Report (Form 8-K), at 1.01 (Jan. 13, 2006) (noting that Teva's patent license for Effexor XR is exclusive at first); Nexium Press Release, supra note 110 (describing Nexium generic entry as "exclusive"); see also Lamictal Press Release, supra note 100 (describing grant of entry in Lamictal settlement, in which generic firm entered during pediatric exclusivity period, as "exclusive").

This source of payment to induce delay has attracted some antitrust enforcement attention. The clause was reportedly one reason why antitrust enforcers rejected the Plavix agreement. See John Carreyrou et al., FBI Raids Offices at Bristol-Myers over Plavix Deal, Wall St. J., July 28, 2006, at A3 (reporting that FIC opposition to "no authorized generic" clause caused rejection of initial agreement); see also Declaration of Bernard Sherman at 10-12, Sanofi-Synthelabo v. Apotex, Inc., 492 F. Supp. 2d 353 (S.D.N.Y. 2006) (No. 02-2255) (declaring that Sanofi orally offered to secretly include "no authorized generic" term in revised deal, after initial agreement containing that term was rejected by regulators). Logically, if a "no authorized generic" provision raises an antitrust problem, then so does retained exclusivity itself, for the effect of the provision is to raise the value of retained exclusivity.

231. Naprelan adopted this strategy. See Andrx Pharms., Inc. v. Elan Corp., No. 00-3481, slip op. at 6 (S.D. Fl. Apr. 24, 2003) (order granting motion for judgment on the pleadings) (describing royalty on brand-name sales, but finding allegation insufficient to survive dismissal on the pleadings). A promotion deal could be structured this way too.

volving the drug Altace, appears to have used a variant of this strategy. There, the brand-name firm acquired a new tablet formulation of the drug and agreed to pay a royalty on its sales. 2 32 This gave the generic firm an incentive not to enter precipitously, as early entry might jeopardize the orderly transition to a new and more profitable formulation. In addition, periodic cash payments, purportedly in exchange for developing the new formulation, were made contingent on unspecified events. This may have been directly for nonentry or indirectly for a successful transition; it is impossible to tell based on the limited data available.

These examples demonstrate that drug makers are adept at achieving a particular substantive outcome-brand-name compensation of generic firms, combined with delayed generic entry-while altering the form of settlement to evade the most obvious risks of antitrust liability. The continuing shift in strategy here resembles the economics of tax shelters. As a regulatory prohibition becomes more stringent, the cost of noncompliance rises. Some regulated parties will give up and simply comply, resulting in a welfare gain. Others will continue to avoid regulation, and instead shift to new strategies. These strategies are more costly to the firm, for otherwise they would have been chosen in the first place; this shift represents a social loss. 2 3 3 Thus, whether an increase in enforcement is warranted depends upon the amount of residual noncompliance and the increase in social costliness of the new behavior. 23 4 The review of settlement behavior in this Article paints a mixed picture. To be sure, this process of continuing evolution threatens the ability of existing antitrust institutions, particularly courts, to keep pace. Courts are increasingly unlikely to be an effective check on settlement. In part, this is because they are poor aggregators. In addition, courts must be fed cases by either a government agency or a private plaintiff. The FTC has limited case-by-case enforcement capacity; practically speak-232. King Pharms., Inc., Quarterly Report (Form 10-Q), at 10 (Aug. 7, 2007) (describing King's exercise of previously secured option to buy Cobalt's tablet NDA, and to pay Cobalt to manufacture and supply the tablet form). A royalty on sales for the acquired tablet product does not appear in the parties' disclosure of the agreements, but it is mentioned in the FTC's 2006 update describing a "complex set of transactions" that fit the Altace transactions. 234. This account is incomplete; also important are the cost of administering the system itself and costs resulting from overinclusion-for example, restricting some valueincreasing side deals. For a discussion of why the latter cost is likely small, see supra Part III.C. ing, it can bring at most a few pharmaceutical antitrust cases at a time, and they are likely to last for five years or more. That capacity is small, compared to the frequency of pay-for-delay settlements-although a successful enforcement action or two would likely reduce the frequency. 23 5 Private plaintiffs, meanwhile, are reluctant to bring cases. Having lost the simplest cash-for-delay agreements, why should they take a chance challenging more complex settlements?

Continuing evolution makes the crisis in case-by-case adjudication more acute for another reason. A single appellate or Supreme Court opinion imposing liability does not fully resolve liability for the newest settlements. A win on a simple case is a very helpful start, but only sets the stage in making sense of the more complicated cases. Thus, even if it were settled as a theoretical matter that paying for delayed entry is prohibited, and settled as a factual matter that side deals provide a disguised means to pay for delay, it does not necessarily follow that the newest settlements also violate antitrust law. Given the malleability of side deals, even an on-point judicial decision imposing liability would not preclude firms from arguing that their arrangements were conceptually or factually distinct. On the other hand, if a court is forced to start with one of the most complex cases, without the benefit of affirmative precedent on the simpler cases, correctly identifying liability seems less likely.

Here, too, FTC rulemaking can help. As to new forms of payment, for example, the FTC could set a rule stating that any conferral of value by a brand-name firm, if made contemporaneously with a generic firm's agreement to delay entry, will be considered to exchange payment for delay. Probabilistic damages would clearly fit within that definition. Agreements to preserve exclusivity, whether simply by agreeing not to contest ANDA approval or by an affirmative agreement not to launch an authorized generic product, would also be included, as just another form of payment. In these examples, the aggregation approach helps to identify and respond to emergent settlement practices before they become more prevalent. For settlements without formal agreement, moreover, FTC rulemaking would be even more effective, because the applicability of the FTC Act, unlike section 1 of the Sherman Act, is not conditioned on the existence of an agreement. 236 Agency rulemaking is not the only possible route for implementing a broadly applicable rule. If a court can be persuaded to think broadly about the implications of settlement, it may adopt a similarly broad ruling, though the considerations above tend to make that less likely. Legis- [Vol. 109:629 lative action is also a possibility. The MMA closed some loopholes, though it preserved others, including the bottleneck and retained exclusivity. Its requirement that an appeals court trigger exclusivity also worsened the delays in an important respect, through a provision buried seven steps deep in the statutory structure: 21 U.S.C.

§ 355(0) (5) (D) (i) (I) (bb) (AA). Thus, although Congress could directly implement any of the presumptions or rules discussed above, its ability to do so quickly and correctly is open to significant doubt. These difficulties also suggest a reframing of the legislative project. Rather than closing identified loopholes with another new layer of complexity, it would be better to remove existing complexity-in particular, by ending retained exclusivity. Simply put, if a generic firm ends its litigation against a brand-name firm, it should no longer be eligible for the exclusivity period.

In a sense, this provision would offer a partial return to the FDA's original view that a generic firm must earn exclusivity by winning a patent suit.

23 7 It would reduce both the amount of payment conferred in a settlement, and the extent to which a settlement delays entry. The provision should apply to both new and existing settlements, like other statutory provisions that have been given retrospective effect. Finally, the political economy of such a statutory change is attractive. If, as some observers might argue, retained exclusivity is not a valuable form of compensation for delay, then its omission from the settlement equation will not be missed, and there is little reason for drug makers to resist this statutory change.

A rule directed to all contemporaneous conferrals of value by a brand-name firm would appear to resolve the pay-for-delay issue, closing the avenue for escape to yet further forms of regulatory avoidance. This appears to be true even as to informal contemporaneous understandings reached between parties, as in the Altace example above. In tax planning, informal alternatives to contract greatly expand the opportunity for avoidance. 238 That problem is less severe for drug patent settlements, where repeat interactions are much less frequent, 239

COLUMBIA LAW REVIEW consequence of curtailing brand-generic interactions-of tolerating an overinclusive ban on the content of side deals-is small. Eliminating continued entitlement to the exclusivity period, despite settlement, would also simplify consideration of any arrangement reached by the brandname and generic firms. Thus, it seems unlikely that effective avoidance would survive the promulgation of a strong rule. If that judgment is incorrect, the agency's ability to respond flexibly, without being subject to stare decisis, may prove to be a significant advantage.

Examining in detail the terms and effects of drug patent settlements reveals several important points. Drug patent settlements that restrict generic competition are an increasingly important, unresolved problem in antitrust enforcement. The evolution in settlement structure makes it less likely that courts will correctly identify and condemn them. There is therefore much reason to fear a continuation and intensification of false negatives if the current policy persists. Case-by-case evaluation is a failure and is likely to remain so, at least absent intervention by the Supreme Court. One partial response is to impose a presumption of payment where side deals accompany delayed entry. This would force firms to explain their increasingly questionable side deals, and would potentially discourage such complex dealmaking in the future.

The analysis supports several further measures. This study reveals persistent gaps in public knowledge about settlements, both in their existence and their terms. These are gaps that the FrC is uniquely positioned to fill. The Agency should step in to collate the extensive information it already has, supplement it with additional factfinding, and disseminate authoritative information of the type offered here. In that respect, this study represents a prima facie case that additional information gathering is necessary, and can serve as a first draft for the FTC's future work. So long as settlements and their terms remain hidden, it will be difficult to do integrative work of the kind suggested here, and difficult to develop the "consensus among commentators" that is a key step in developing appropriate antitrust policy. The additional insight will help academics and policymakers in revising, if necessary, the initial conclusion presented here that pay-for-delay settlements are frequently tried and frequently successful.