Friday, January 31, 2014

Suzanne Scotchmer

I was very sorry to hear this evening about the death of University of California Professor Suzanne Scotchmer.  She was one of the giants of innovation economics.  Here is a link to a tribute by Professor Joshua Gans.

Thursday, January 30, 2014

Astrazeneca AB v. Apotex Corp. Part 2: Why Reasonable Royalties?

Earlier this week, I published Professor Siebrasse's comments on the Astrazeneca case, which centered on the issue of noninfringing alternatives.  Another aspect of the case that struck me as interesting is that the patent owner agreed to an award of a reasonable royalty rather than lost profits (see opinion p.4).  On one view, that seems odd, because  normally the brand-name drug patent owner's interest lies in exclusion, not licensing; the brand-name drug company stands to lose more in terms of lost profits than the would-be generic competitor stands to gain from competing.  See Roger D. Blair & Thomas F. Cotter, Are Settlements of Patent Disputes Illegal Per Se?, 47 Antitrust Bulletin 491, 524-25 (2002).  In this case in particular, Judge Cote concluded that Astra would not have licensed its patent to Apotex.  See opinion p.43 ("There is no evidence that Astra would have initiated or encouraged licensing discussions with Apotex."), p.115 (". . . Astra would have had no desire in November 2003, or indeed at any time, to issue a license to Apotex.").  So why not ask for lost profits instead of a reasonable royalty? There are at least three possible reasons.

One possible reason is that the defendant might have succeeded in showing that it could have turned to a noninfringing alternative.  If a noninfringing alternative were available and would have enabled Apotex to compete for Astra's clientele, Astra's lost profits caused by the infringement could be zero, in which case Astra would have been entitled only to a reasonable royalty.  (If the noninfringing alternative was not a perfect substitute in the minds of some of Astra's customers, however, Astra's lost profits caused by the infringement could be greater than zero, though whether it would be worthwhile to try to distinguish which sales Astra would have lost to Apotex even if Apotex had used the imperfect alternative, and which sales Astra would have made itself, depends on the facts.)  Perhaps Astra wasn't sure it would win on the noninfringing alternative point and decided to simplify matters by asking for only a reasonable royalty.

A second possibility is that Astra didn't want to have to disclose evidence of its own profit margins, which would be necessary to quantify its lost profits.  I don't know whether that is the case or not, but I understand that in many countries one of the reasons patent plaintiffs often opt for reasonable royalties or, where available, awards of defendant's profits is that they don't want to disclose this sort of information, particularly not to a competitor.

A third possibility is that Astra decided it would have been too difficult to try to quantify its lost profit, given the presence of multiple generic competitors, some of which were not infringing.  (Proving the amount of one's lost profits often is a problem, both in the U.S. and elsewhere.)  In that event, the reasonable royalty is really just a more-easily-proven but imperfect substitute for lost profits.  Roger D. Blair & Thomas F. Cotter, Intellectual Property:  Economic and Legal Dimensions of Rights and Remedies 231 (Cambridge Univ. Press 2005); Mark A. Lemley, Distinguishing Lost Profits from Reasonable Royalties,  51 Wm. & Mary L. Rev. 655, 662, 666-67, 671-72 (2009).  In the present case, Judge Cote set the royalty at 50% of Apotex's profits derived from the infringement over the applicable time period; and perhaps that comes close to approximating the profit that Astra lost as a result of the infringement.  (For the evidence from which she concluded that 50% was "within the range of negotiations that occurred in connection with the patents in suit," see opinion pp. 120-26.)  If, however, as suggested above, the brand-name drug company's lost profit normally exceeds the generic firm's profit attributable to the infringement, wouldn't an award of 100% of Apotex's profit come even closer to approximating the amount of Astra's lost profit?  Of course, it's not realistic to think that the infringer would have agreed to such terms ex ante, but the whole willing licensor-willing licensee framework is unrealistic on these facts, and ex ante a 100% royalty rate would come closer to approximating the amount of the plaintiff's lost profit than would a lower royalty rate multiplied by the infringer's actual sales or profits. 

Perhaps such an approach would be unduly speculative, though, and in any event an award of 100% of the infringer's profit as a reasonable royalty would be in some tension with the fact that (rightly or wrongly) the U.S. did away with awards of infringers' profits for the infringement of utility patents in 1946.  So even if it were clear that the patentee's expected lost profit exceeded the infringer's expected profit, we would need to impose an additional condition before awarding 100% of the infringer's profit as a reasonable royalty, or else we would be opening the door to allowing patent owners to recover infringers' profits even when their own lost profits are lower (that is, to awarding infringer's profits as such and not as a reasonable royalty in a case like the present one).  The additional condition would be evidence that the patentee's actual (though unquantifiable) lost profit exceeds the infringer's actual profit.  This second condition may often be present in cases involving brand-name versus generic drug companies, but of course expectations may be confounded if the infringer's product proves more popular than expected and other, perhaps unexpected, competition would have kept the patentee from achieving its expected profit even absent the infringement. 

Requiring such proof may be too complicated, however, in which case an award of something less than 100% may be the safer route, even if it does risk some degree of undercompensation.  I also tend to think that an award of 100% of the infringer’s profit as a reasonable royalty would be inappropriate in the first two scenarios above.  In the first scenario, where there is a noninfringing alternative, voluntary licensing would have been rational and a 100% royalty would not make much sense.  In the second, where the patentee chose not to seek lost profits because it wanted to keep its profit margin secret, I’m less concerned about making the patentee whole; it made its choice and has to live with it.  To the extent it’s difficult to tell whether the patentee opted for reasonable royalties over lost profits due to reasons 1, 2, or 3, though I suppose that weighs against awarding 100% of the defendant’s profit as a reasonable royalty under any of the scenarios. 

Either that, or we could amend the law to permit straightforward awards of defendant's profits (as many countries do), though I'm not sure I'm ready to go that route just yet.

*                 *                   *

In other news, last week a federal jury in California awarded the Alfred E. Mann Foundation for Scientific Research $131 million in patent damages against Cochlear.  Judgment has yet to be entered, however.  Here is a link to a report on Bloomberg News.  I may blog about this case at some point as it develops further.

Monday, January 27, 2014

Astrazeneca AB v. Apotex Corp.: U.S. District Court Awards Drug Company $76 Million in Patent Infringement Damages

I mentioned last month that I would be publishing a further post on this opinion authored by Judge Denise Cote of the U.S. District Court for the Southern District of New York, which issued in early December.  Here is a link to the opinion; it can also be found on Westlaw at 2013 WL 6244425.  As it turns out, there’s enough food for thought in the opinion to merit at least two posts.  Today’s post by Professor Norman Siebrasse of the University of New Brunswick focuses on the noninfringing alternatives.  Later this week, I will post my comments on the patentee’s decision to opt for reasonable royalties over lost profits.

Siebrasse: Apotex sold generic omeprazole (Prilosec) by from November 2003 until 2007, when it was held to have infringed AstraZeneca’s  patent. Last December Judge Cote released her damages decision, after a bench trial: AstraZeneca AB v Apotex Corp – F.Supp.2d –, 2013 WL 6244425 (SDNY, 2013). On a purely factual level, the decision is interesting for its detailed discussion of the dynamics of brand pricing in response to generic entry. The case also raises an interesting question of principle, which is a twist on the question of whether damages should be assessed with knowledge of validity and infringement of the patent.

A reasonable royalty is determined on the basis of a hypothetical negotiation between the parties at the time of the infringement. A key question in any such negotiation is what non-infringing alternatives (NIA) the infringer would have had available to it. A potential licensee will not license a patent if the royalty is so high that it would make more money using the best non-infringing alternative, so the availability of an NIA sets a cap on the reasonable royalty.

In this case, the patent at issue is a formulation patent for omeprazole. Mylan and Lek launched at risk in August of 2003, while Apotex launched shortly afterwards, in November 2003. Because the patent at issue was a formulation patent, infringement was not a foregone conclusion. Ultimately, Mylan and Lek were found not to have infringed, while Apotex did infringe (p 5, fn 7). Apotex argued that the appropriate NIA was the non-infringing Mylan and Lek product, notwithstanding that Apotex did not know at the time of the hypothetical negotiation that the products were non-infringing. Judge Cote rejected this argument as a matter of law (21, my emphasis):

Apotex suggests that because it is assumed to have known in November 2003 that it infringed the Patents under the hypothetical negotiation framework, it should also be assumed to have known that Mylan and Lek did not infringe. Apotex has not cited any law in support of this view. The hypothetical negotiation framework does not treat the parties as having knowledge of all events between the negotiation and the finding of infringement simply because it requires them to assume that the Patents are valid and infringed.

I’m not sure I agree with Judge Cote’s conclusion, as least with respect to the specific point of whether the infringer should be assumed to know whether the alternatives are in fact infringing. The principled reason to consider the NIA is that the true value of the invention is the marginal value of NIA as compared with the patented invention. If the patent can in fact be invented around relatively easily, then it is not a valuable contribution, even though it may be difficult to know in advance which of the various easily developed alternatives actually infringe. Patents are intended to reward valuable technical innovations, not to reward the uncertainty inherent in the patent system.

The particular facts of this case illustrates the point. Because of the dynamics of the pharma market, and the unusual dynamics of this product in particular, entry by another generic might be expected to cause a significant drop in the market price. In particular, Mylan and Lex kept their prices high precisely because they were uncertain as to whether they were infringing, so as to minimize their potential liability to AstraZeneca. So the policy question is whether the effective extension of protection due to uncertainty as to infringement is part of the legitimate return to a patentee. I am inclined to think not.

But Judge Cote was making the broader point that just because the parties are assumed to bargain with knowledge of the validity and infringement of the patent at issue, this does not imply that the parties should be treated as “having knowledge of all events between the negotiation and the finding of infringement.” The assumption of knowledge regarding validity of the patent at issue is usually justified to avoid a double discounting problem (see Blair & Cotter, “Intellectual Property: Economic and Legal Dimensions of Rights and Remedies” at 230) That policy does not seem to justify a broader principle of knowledge of all intervening facts.

While it is counter-intuitive to assume knowledge of all intervening events, I’m not sure it is wrong in principle. If, at the time of the hypothetical bargain, Apotex knew that its product infringed, but it did not know whether Mylan’s process infringed, Mylan’s process would nonetheless set Apotex’s maximum willingness to pay, albeit discounted for the fact that it might infringe. On the other hand, if Apotex is assumed to know whether Mylan’s process infringed, it would also set the maximum willingness to pay – which might be higher or lower than the discounted value, depending on whether it infringed. On average, the reasonable royalty – and thus the incentive to create and deterrence to infringement – will be the same whichever rule is used. (I don’t see any selection bias problem. The defendant will invoke the alternative when it turns out to be non-infringing, but not otherwise.) But using the actual facts, rather than trying to recreate Apotex’s expectations, is both administratively simpler and more accurate. Suppose that potential infringers are systematically irrationally pessimistic about whether the alternative infringes. If that were so, using an expectations approach would inflate the return to weak patents. Conversely, the incentive to create would be undermined if it turns out that infringers are systematically over-optimistic on this issue. The reward to invention should not turn on the quirks of behavioural psychology, any more than on the uncertainty inherent in the legal system.

Finally, I should note that on the facts the Judge Cote held that Apotex would not easily have been able to easily copy Mylan’s product or develop its own similar NIA, so the holding that the bargaining should be carried out in ignorance of whether those alternatives infringed was not determinative.

Friday, January 24, 2014

Bobst v. Heidelberg: A recent French case on lost profits

The case is Bobst SA v. Heidelberg Postpress Deutschland GmbH, TGI Paris, Sept. 27, 2013, PIBD No. 995, III, at 1585.  The patent in suit is the French portion of European Patent 1170228 A2, for (from the English translation, here) a device for controlling the means for feeding sheets into a machine.  The German Bundespatentgericht annulled the German portion of the patent in 2011, but the patent remains valid in France.

A little background first on calculating patent damages under French law.  As I explain in my book (pp. 264-65, citations omitted):
French courts often begin their damages analysis by calculating the infringing turnover (la masse contrefaisante), which is equal to the number of infringing products for which the defendant is responsible, multiplied by the price at which the defendant sold them.  To calculate lost profits, however, the court estimates the number of sales the plaintiff would have made but for the infringement, taking into account such factors as the plaintiff’s capacity, customer preferences, possible noninfringing substitutes, and so on. It then multiplies this number by (a) the price at which the plaintiff would have sold the products and (b) the plaintiff’s profit margin.
In the present case, there were three sales of infringing machines:  one in May 2006 to Cartolux; one in July 2007 to TPG; and one to Fellmann that was installed in August 2006, but then returned to the defendant in July 2009.  Among the contested issues were which of these sales should be included within the masse contrefaisante.

First, the court noted first that the parties accepted the expert’s conclusion that the sale to TPG had cost Bobst a sale.  The court also accepted the expert’s conclusion that the average sales price of a Bobst machine in 2007 was €1,231,000, and that Bobst’s margin over variable costs was 43%.  Bobst therefore suffered a lost profit of €529,300 on the sale lost to TPG.  

Second, the expert had noted that Bobst had not made an offer to Catolux, and concluded that it wasn’t clear that Bobst would have made this sale absent the infringement; he therefore wound up excluding this sale from the masse contrefaisante.  The court concluded, however, that Heidelberg’s aggressive pricing policy had caused Bobst to lose its chance of making an offer that might have interested Cartolux, and that absent Heidelberg’s infringing conduct Cartolux naturally would have turned toward Bobst.  As a result, the court estimated that Bobst had a 50% chance of making the sale to Cartolux absent the infringement, and awarded it one-half of the profits it would have earned on sale to Cartolux.  Using the average 2006 price as the base, the damages came to €240,000.  (Note that this is clearly different from U.S. law, where a 50% chance of making the sale would not, it seems to me, demonstrate by a preponderance of the evidence that the sale would have been made.  Therefore, Bobst would have received zero lost profits on this sale under U.S. law, though it would have been entitled to a royalty.  On the other hand, with a 51% chance of making the sale, Bobst would prevail by a preponderance of the evidence and get its entire lost profit on that sale.)

Third, as to Fellmann, the court concluded that the masse contrefaisante is determined from the date the infringement began up until the date of judgment, here May 27, 2009.  Nevertheless, the defendant argued that the machine sold to Fellmann—which the court characterized as having been provided on a trial basis (à titre d’essai), and which was, after all, returned to the defendant, albeit a couple of months after May 27, 2009—should not be included.  The court disagreed, noting that according to the court-appointed expert the plaintiff had made two offers to Fellmann in 2006, which were rejected, but that it had succeeded in selling Fellmann a machine in 2009. The court therefore concluded that Heidelberg’s infringing competition had cost Bobst the ability to make the sale at an earlier period.  (Elsewhere in the opinion, the court refers to the defendant’s aggressive pricing policy as a factor in its ability to take away business from the plaintiff.)  Bobst concluded a sale with Fellmann in 2009 for €860,897, a discount of €587,043 (40.54%) off the €1,447,940 sticker price.  Bobst claimed that it was entitled to recover the discount, on the theory that it wouldn’t have had to offer it if it had the sale three years earlier.  The expert concluded, however, that Bobst habitually gave clients discounts, and that only 10% off of the €1,447,940 sticker price, or €144,794, could be allocated to the infringement.

Finally, the court awarded damages for lost sales of complementary goods that would have been sold to TPG and Cartolux, amounting altogether to €205,000 and €93,000, respectively.  The court was not convinced that Bobst was entitled to additional damages it claimed to have suffered as a result of granting higher discounts to other customers during the period of Heidelberg’s activity.  

Wednesday, January 22, 2014

U.S. Supreme Court Reverses Federal Circuit on Declaratory Judgment Burden of Proof; Comparison with Germany and the U.K.

This morning the U.S. Supreme Court handed down a unanimous opinion, authored by Justice Breyer, in Medtronic, Inc. v. Mirowski Family Ventures, LLC, available here.  As I wrote in a previous post:

In September [2012], the Federal Circuit held that, where a licensee in good standing files an action seeking a declaratory judgment of noninfringement (permissible ever since the Supreme Court's 2007 decision in MedImmune v. Genentech. 540 U.S. 118), the licensee bears the burden of proving noninfringement--rather than the patentee bearing the burden of proving infringement.  Writing for the court, Judge Linn concluded that "A contrary result would allow licensees to use MedImmune's shield as a sword--haling licensors into court and forcing them to assert and prove what had already been resolved by license."  The question presented in the cert petition is:  "whether, in such a declaratory judgment action brought by a licensee under MedImmune, the licensee has the burden to prove that its products do not infringe the patent, or whether (as is the case in all other patent litigation, including other declaratory judgment actions), the patentee must prove infringement."  I don't have strong feelings about this one, but Dennis Crouch predicts a 9-0 reversal.

Dennis Crouch was right.  (And I eventually did come around to the view that reversal was proper and signed off on John Golden's amicus brief to that effect, see here.)  In my view, the result also is consistent with the tenor of the oral argument, which I blogged about here.

From the opinion (with citations omitted):

Simple legal logic, resting upon settled case law, strongly supports our conclusion. It is well established that the burden of proving infringement generally rests upon the patentee.  We have long considered “the operation of the Declaratory Judgment Act” to be only “procedural,” leaving “substantive rights unchanged.” 
And we have held that “the burden of proof” is a “‘substantive’ aspect of a claim.” 
Taken together these three legal propositions indicate that, in a licensee’s declaratory judgment action, the burden of proving infringement should remain with the patentee.
Several practical considerations lead to the same conclusion. To shift the burden depending upon the form of the action could create postlitigation uncertainty about the scope of the patent. Suppose the evidence is inconclusive, and an alleged infringer loses his declaratory judgment action because he failed to prove noninfringement. The alleged infringer, or others, might continue to engage in the same allegedly infringing behavior, leaving it to the patentee to bring an infringement action. If the burden shifts, the patentee might lose that action because, the evidence being inconclusive, he failed to prove infringement. So, both sides might lose as to infringement, leaving the infringement question undecided, creating uncertainty among the parties and others who seek to know just what products and processes they are free to use. . . .
Moreover, to shift the burden can, at least on occasion, create unnecessary complexity by making it difficult for the licensee to understand upon just what theory the patentee’s infringement claim rests. A complex patent can contain many pages of claims and limitations. A patentholder is in a better position than an alleged infringer to know, and to be able to point out, just where, how, and why a product (or process) infringes a claim of that patent. Until he does so, however, the alleged infringer may have to work in the dark, seeking, in his declaratory judgment complaint, to negate every conceivable infringement theory.
Finally burden shifting here is difficult to reconcile witha basic purpose of the Declaratory Judgment Act. . . .

As I mentioned in my May 21, 2013 post, according to Wolfgang von Meibom & Christian Harmsen's article The Rare German Negative Declaratory Action, in Patent Practice in Japan and Europe:  Liber Amicorum for Guntram Rahn 327 (Bernd Hansen & Dirk Schüssler-Langeheine eds. 2011), in  Germany "the patentee always has the burden of proof with regard to the infringement issue, even in negative declaratory actions.  It is thus not legitimate to arbitrarily impose this burden of proof on the owner of the IP right unless it has provided a reason to do so" (citing Higher Regional Court Cologne, NJW-RR 1997, 483, 484).  On the other hand, as I noted more recently in another post, a case from the U.K. (Generics (UK) Ltd. t/a Mylan v. Yeda Res. & Dev. Co. Ltd and Teva Pharm. Ind. Ltd [2013] EWCA Civ 925 (Ct. App. 2013)), states that "the burden of establishing non-infringement fell on [declaratory judgment plaintiff] Mylan" (para. 108).  (Hat tip to IPKat.)  I would be interested in learning if this issue has come up elsewhere.

By the way, I mentioned on January 13 that with my teaching responsibilities resuming, I would be blogging "only twice a week for the next few months--most likely Mondays and Thursdays--though I will chime in more frequently, as needed, if and when something of particular importance (e.g., an important Supreme Court case) comes down."  Well, this is an important Supreme Court case, so I'm chiming in more frequently.  I will go ahead and publish the post I was planning for tomorrow, on a recent French case on lost profits, later this week. 

Monday, January 20, 2014

Petit on Injunctions for FRAND-Pledged SEPs

I mentioned Nicolas Petit's article Injunctions for FRAND-Pledged SEPs:  The Quest for an Appropriate Test of Abuse Under Article 102 TFEU , 9 Euro. Comp. J. 677 in a recent post, while noting that I hadn't yet finished reading it.  I now have, and although I don't agree with everything the author says (what fun would that be?), I certainly recommend taking a look at Professor Petit's very interesting paper.  His thesis is that, under European competition law, it generally should not be an abuse of dominant position for the owner of a FRAND-encumbered SEP to seek injunctive relief.  In reaching this conclusion, Professor Petit surveys the circumstances under which EU law to date has found an abuse of dominant position premised on abusive refusals to supply, abusive litigation, or abusive anticompetitive foreclosure; and he concludes that in most cases the act of seeking an injunction for the infringement of a FRAND-encumbered SEP does not comfortably fit within any of these theories.  Of course, the possibility remains of fashioning a new theory of abuse--which could be some version of the German "Orange Book Standard" defense, or the standard the European Commission has articulated in its proceedings against Samsung and Google, or something else.  In this regard, Professor Petit argues that a useful criterion for selecting among competing theories or proposals is "consistency," which as he sees it has four dimensions:  internal consistency (consistency with existing tests of abuse); transversal consistency (consistency with "legal principles adopted in the other areas of EU competition law"); constitutional consistency; and economic consistency.  Professor Petit concludes that none of the new theories under which the act of seeking an injunction for the infringement of a FRAND-encumbered SEP would qualify as an abuse are sufficiently "consistent" along these different dimensions.

I agree with Professor Petit that competition law is not the ideal vehicle for addressing the sort of conduct that is at issue in FRAND/SEP cases, and I have reached conclusions similar to Professor Petit's under U.S. antitrust law (see here).  I am not as skeptical as he is, however, about the existence of patent holdup (see Professor Petit's article, p.714), and I am more skeptical than he about the ability of standard users to design around an SEP (see p.716).  As I have argued in my own work, though, I think the better option is to interpret the law of remedies as permitting courts to grant prospective royalties rather than permanent injunctions in cases in which there is a substantial risk of holdup or royalty stacking, i.e., in cases in which the patentee's bargaining power ex post is largely attributable to the incorporation of its technology into a standard.  To the extent that solution is not at present a realistic one under most European nations' patent laws, however, I'm moderately inclined to think that purely on a policy basis it may be acceptable to (in effect) achieve the same result by means of competition law (though I think the doctrinal issues Professor Petit has highlighted are indeed serious ones).

The article is also interesting in presenting (at p.715) patentee win rates in the field of telecommunications and electronic communications patents from 2000-13 in different European countries. 

Friday, January 17, 2014

Grunwald on the Recovery of Design-Around Costs Under German Law

Dr. Marc Grunwald has published an article titled Kosten für eine Ausweichtechnik als Vollstreckungsschaden? (“Design-around costs as enforcement damages?”) in 12/2013 Mitteilungen der deutschen Patentanwälte.  The topic relates to one I have touched on previously, namely the liability under some countries’ laws on the part of a patent owner when it enforces a patent judgment that is subsequently reversed on appeal or, even in a completely separate proceeding, invalidated.

By way of introduction, the general rule in federal courts in the United States is that a permanent injunction embodied in a district court’s final judgment is enforceable immediately, unless the district court or the Court of Appeals stays it.  The damages portion of the judgment is automatically stayed for 14 days, but any additional stay pending appeal is discretionary and is normally permitted only if the defendant posts a “supersedeas” bond in the amount of the damages judgment.  See Fed. R. Civ. P. 62; Fed. R. App. P. 8; 11 Charles Alan Wright et al., Federal Practice & Procedure §§ 2901-05 (3d ed. 2013).  Normally then if the judgment is reversed on appeal, the defendant hasn’t paid any damages yet; but it doesn’t get to recover any additional damages it suffered in the interim by having to comply with a non-stayed injunction.  If for whatever reason the damages judgment was not stayed pending appeal and the judgment is reversed, the defendant is entitled to get back the damages it paid, see 11 Wright et al., supra, § 2905, but, again, nothing in addition to that (absent fraud or something else that would give rise to further liability on the part of the plaintiff/appellee).

In Germany, unless a stay pending appeal is granted the judgment of the court of first instance becomes provisionally enforceable pending appeal if the prevailing plaintiff posts a bond.  See Thomas Kühnen, Patent Litigation Proceedings in Germany, paras. 1987-95, 2013-42 (Frank Peterreins tr., 6th ed. 2013).  According to Kühnen (paras. 1995-96):
The security bond serves to protect the interest of the party against whom enforcement is to be carried out.  It is intended to compensate it for any disadvantages it may suffer in the event of any compulsory enforcement of the judgment which subsequently turns out to be unjustified.

Alongside the fees of the plaintiff’s attorney and the court costs, account must also be taken of a possible claim for damages by the obligor pursuant to Section 717(2) of the German Code of Civil Procedure.
Section 717 reads as follows, first in the original German and then in an English language translation available here:

§ 717:  Wirkungen eines aufhebenden oder abändernden Urteils
(1) Die vorläufige Vollstreckbarkeit tritt mit der Verkündung eines Urteils, das die Entscheidung in der Hauptsache oder die Vollstreckbarkeitserklärung aufhebt oder abändert, insoweit außer Kraft, als die Aufhebung oder Abänderung ergeht.

(2) Wird ein für vorläufig vollstreckbar erklärtes Urteil aufgehoben oder abgeändert, so ist der Kläger zum Ersatz des Schadens verpflichtet, der dem Beklagten durch die Vollstreckung des Urteils oder durch eine zur Abwendung der Vollstreckung gemachte Leistung entstanden ist. Der Beklagte kann den Anspruch auf Schadensersatz in dem anhängigen Rechtsstreit geltend machen; wird der Anspruch geltend gemacht, so ist er als zur Zeit der Zahlung oder Leistung rechtshängig geworden anzusehen.

(3) Die Vorschriften des Absatzes 2 sind auf die im § 708 Nr. 10 bezeichneten Berufungsurteile, mit Ausnahme der Versäumnisurteile, nicht anzuwenden. Soweit ein solches Urteil aufgehoben oder abgeändert wird, ist der Kläger auf Antrag des Beklagten zur Erstattung des von diesem auf Grund des Urteils Gezahlten oder Geleisteten zu verurteilen. Die Erstattungspflicht des Klägers bestimmt sich nach den Vorschriften über die Herausgabe einer ungerechtfertigten Bereicherung. Wird der Antrag gestellt, so ist der Anspruch auf Erstattung als zur Zeit der Zahlung oder Leistung rechtshängig geworden anzusehen; die mit der Rechtshängigkeit nach den Vorschriften des bürgerlichen Rechts verbundenen Wirkungen treten mit der Zahlung oder Leistung auch dann ein, wenn der Antrag nicht gestellt wird.
Section 717:  Effects of a judgment reversing or modifying the original judgment
(1) Upon a judgment being pronounced that reverses or modifies the decision in the main action, or the declaration of enforceability, the judgment shall cease to be provisionally enforceable to the extent to which it is reversed or modified.

(2) If a judgment declared provisionally enforceable is reversed or modified, the plaintiff shall be obligated to compensate the defendant for the damages he has suffered by the judgment being enforced, or by the payments he had to make, or any other actions he had to take in order to avert enforcement. The defendant may assert the claim to compensation of damages in the pending legal dispute; once this claim is asserted and filed, it is to be deemed as having become pending at the time at which the payment was made or other action was taken.

(3) The stipulations of subsection (2) are not to be applied to the appellate judgments designated in section 708 number 10, to the exception of default judgments. Insofar as such a judgment is reversed or modified, the plaintiff is to be sentenced, upon a corresponding petition having been filed by the defendant, to reimburse the latter for the payments made or other actions taken on the basis of that prior judgment. The obligation of the plaintiff to so reimburse the defendant is determined by the rules as to the surrender of the result of any unjust enrichment. Once the petition has been filed, the claim to reimbursement is to be deemed as having become pending at the time at which the payment was made or other action was taken; even where the petition is not filed, the effects tied to the pendency of the matter pursuant to the stipulations under civil law shall occur with the payment being made or other action being taken.
(The above-mentioned section 708(10) states that "The following are to be declared provisionally enforceable without any provision of security . . . Appellate judgments in disputes under property law. Where leave to appeal is denied by a judgment or court order pursuant to section 522(2), this is to mandate that the judgment is provisionally enforceable without any provision of security.")

Dr. Grunwald’s article discusses the types of damages the former defendant may recover under § 717, and argues that the damages recoverable when a first instance judgment is reversed (717 para. 2) include not only any damages the former defendant has paid and any profits it has lost as a result of having to pull its product from the market, but also the costs the defendant has incurred pending appeal as a result of having to design around by adopting a noninfringing alternative technology.  (As a matter of policy, this would appear to be correct if the goal is to ensure that the defendant is no worse off than it would have been if the trial court had gotten it right.  Under U.S. law, by contrast, as noted above the defendant is out of luck if the injunction wasn’t stayed pending appeal; there is no recourse for recovering these costs.)  On the other hand, if I'm understanding this correctly, under 717 para. 3 sentence 2, when a second instance (first appellate) judgment is reversed on further appeal, the defendant generally is entitled to recover only those expenses it incurred as a result of complying with the judgment (not ts own lost profits), and only to the extent this has benefited the plaintiff.  The defendant doesn't get his lost profits or design-around expenses, in Dr. Grunwald's view.

Note also that, under German law even when a judgment is affirmed on appeal and the case would appear to be over, if the patent is invalidated in a subsequent proceeding the former defendant in the infringement suit can sue to recover the damages it paid the patent owner.  See my post of July 4, 2013, here (noting that this may be the law now in the U.K. as well, but not in France); Kühnen, supra, paras. 1917-57.  In addition, based on discussions with Dr. Christopher Heath, my understanding as expressed in this post of November 7, 2013, is that the former defendant also can recover damages for having had to comply with the judgment in the interim.   According to Dr. Heath, however, in most cases the parties settle the matter and he is not aware of any German cases precisely on point.